File: | llvm/include/llvm/CodeGen/SelectionDAGNodes.h |
Warning: | line 1122, column 10 Called C++ object pointer is null |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
1 | //===-- PPCISelLowering.cpp - PPC DAG Lowering Implementation -------------===// | |||
2 | // | |||
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. | |||
4 | // See https://llvm.org/LICENSE.txt for license information. | |||
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception | |||
6 | // | |||
7 | //===----------------------------------------------------------------------===// | |||
8 | // | |||
9 | // This file implements the PPCISelLowering class. | |||
10 | // | |||
11 | //===----------------------------------------------------------------------===// | |||
12 | ||||
13 | #include "PPCISelLowering.h" | |||
14 | #include "MCTargetDesc/PPCPredicates.h" | |||
15 | #include "PPC.h" | |||
16 | #include "PPCCCState.h" | |||
17 | #include "PPCCallingConv.h" | |||
18 | #include "PPCFrameLowering.h" | |||
19 | #include "PPCInstrInfo.h" | |||
20 | #include "PPCMachineFunctionInfo.h" | |||
21 | #include "PPCPerfectShuffle.h" | |||
22 | #include "PPCRegisterInfo.h" | |||
23 | #include "PPCSubtarget.h" | |||
24 | #include "PPCTargetMachine.h" | |||
25 | #include "llvm/ADT/APFloat.h" | |||
26 | #include "llvm/ADT/APInt.h" | |||
27 | #include "llvm/ADT/ArrayRef.h" | |||
28 | #include "llvm/ADT/DenseMap.h" | |||
29 | #include "llvm/ADT/None.h" | |||
30 | #include "llvm/ADT/STLExtras.h" | |||
31 | #include "llvm/ADT/SmallPtrSet.h" | |||
32 | #include "llvm/ADT/SmallSet.h" | |||
33 | #include "llvm/ADT/SmallVector.h" | |||
34 | #include "llvm/ADT/Statistic.h" | |||
35 | #include "llvm/ADT/StringRef.h" | |||
36 | #include "llvm/ADT/StringSwitch.h" | |||
37 | #include "llvm/CodeGen/CallingConvLower.h" | |||
38 | #include "llvm/CodeGen/ISDOpcodes.h" | |||
39 | #include "llvm/CodeGen/MachineBasicBlock.h" | |||
40 | #include "llvm/CodeGen/MachineFrameInfo.h" | |||
41 | #include "llvm/CodeGen/MachineFunction.h" | |||
42 | #include "llvm/CodeGen/MachineInstr.h" | |||
43 | #include "llvm/CodeGen/MachineInstrBuilder.h" | |||
44 | #include "llvm/CodeGen/MachineJumpTableInfo.h" | |||
45 | #include "llvm/CodeGen/MachineLoopInfo.h" | |||
46 | #include "llvm/CodeGen/MachineMemOperand.h" | |||
47 | #include "llvm/CodeGen/MachineModuleInfo.h" | |||
48 | #include "llvm/CodeGen/MachineOperand.h" | |||
49 | #include "llvm/CodeGen/MachineRegisterInfo.h" | |||
50 | #include "llvm/CodeGen/RuntimeLibcalls.h" | |||
51 | #include "llvm/CodeGen/SelectionDAG.h" | |||
52 | #include "llvm/CodeGen/SelectionDAGNodes.h" | |||
53 | #include "llvm/CodeGen/TargetInstrInfo.h" | |||
54 | #include "llvm/CodeGen/TargetLowering.h" | |||
55 | #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h" | |||
56 | #include "llvm/CodeGen/TargetRegisterInfo.h" | |||
57 | #include "llvm/CodeGen/ValueTypes.h" | |||
58 | #include "llvm/IR/CallingConv.h" | |||
59 | #include "llvm/IR/Constant.h" | |||
60 | #include "llvm/IR/Constants.h" | |||
61 | #include "llvm/IR/DataLayout.h" | |||
62 | #include "llvm/IR/DebugLoc.h" | |||
63 | #include "llvm/IR/DerivedTypes.h" | |||
64 | #include "llvm/IR/Function.h" | |||
65 | #include "llvm/IR/GlobalValue.h" | |||
66 | #include "llvm/IR/IRBuilder.h" | |||
67 | #include "llvm/IR/Instructions.h" | |||
68 | #include "llvm/IR/Intrinsics.h" | |||
69 | #include "llvm/IR/IntrinsicsPowerPC.h" | |||
70 | #include "llvm/IR/Module.h" | |||
71 | #include "llvm/IR/Type.h" | |||
72 | #include "llvm/IR/Use.h" | |||
73 | #include "llvm/IR/Value.h" | |||
74 | #include "llvm/MC/MCContext.h" | |||
75 | #include "llvm/MC/MCExpr.h" | |||
76 | #include "llvm/MC/MCRegisterInfo.h" | |||
77 | #include "llvm/MC/MCSectionXCOFF.h" | |||
78 | #include "llvm/MC/MCSymbolXCOFF.h" | |||
79 | #include "llvm/Support/AtomicOrdering.h" | |||
80 | #include "llvm/Support/BranchProbability.h" | |||
81 | #include "llvm/Support/Casting.h" | |||
82 | #include "llvm/Support/CodeGen.h" | |||
83 | #include "llvm/Support/CommandLine.h" | |||
84 | #include "llvm/Support/Compiler.h" | |||
85 | #include "llvm/Support/Debug.h" | |||
86 | #include "llvm/Support/ErrorHandling.h" | |||
87 | #include "llvm/Support/Format.h" | |||
88 | #include "llvm/Support/KnownBits.h" | |||
89 | #include "llvm/Support/MachineValueType.h" | |||
90 | #include "llvm/Support/MathExtras.h" | |||
91 | #include "llvm/Support/raw_ostream.h" | |||
92 | #include "llvm/Target/TargetMachine.h" | |||
93 | #include "llvm/Target/TargetOptions.h" | |||
94 | #include <algorithm> | |||
95 | #include <cassert> | |||
96 | #include <cstdint> | |||
97 | #include <iterator> | |||
98 | #include <list> | |||
99 | #include <utility> | |||
100 | #include <vector> | |||
101 | ||||
102 | using namespace llvm; | |||
103 | ||||
104 | #define DEBUG_TYPE"ppc-lowering" "ppc-lowering" | |||
105 | ||||
106 | static cl::opt<bool> DisablePPCPreinc("disable-ppc-preinc", | |||
107 | cl::desc("disable preincrement load/store generation on PPC"), cl::Hidden); | |||
108 | ||||
109 | static cl::opt<bool> DisableILPPref("disable-ppc-ilp-pref", | |||
110 | cl::desc("disable setting the node scheduling preference to ILP on PPC"), cl::Hidden); | |||
111 | ||||
112 | static cl::opt<bool> DisablePPCUnaligned("disable-ppc-unaligned", | |||
113 | cl::desc("disable unaligned load/store generation on PPC"), cl::Hidden); | |||
114 | ||||
115 | static cl::opt<bool> DisableSCO("disable-ppc-sco", | |||
116 | cl::desc("disable sibling call optimization on ppc"), cl::Hidden); | |||
117 | ||||
118 | static cl::opt<bool> DisableInnermostLoopAlign32("disable-ppc-innermost-loop-align32", | |||
119 | cl::desc("don't always align innermost loop to 32 bytes on ppc"), cl::Hidden); | |||
120 | ||||
121 | static cl::opt<bool> UseAbsoluteJumpTables("ppc-use-absolute-jumptables", | |||
122 | cl::desc("use absolute jump tables on ppc"), cl::Hidden); | |||
123 | ||||
124 | static cl::opt<bool> EnableQuadwordAtomics( | |||
125 | "ppc-quadword-atomics", | |||
126 | cl::desc("enable quadword lock-free atomic operations"), cl::init(false), | |||
127 | cl::Hidden); | |||
128 | ||||
129 | STATISTIC(NumTailCalls, "Number of tail calls")static llvm::Statistic NumTailCalls = {"ppc-lowering", "NumTailCalls" , "Number of tail calls"}; | |||
130 | STATISTIC(NumSiblingCalls, "Number of sibling calls")static llvm::Statistic NumSiblingCalls = {"ppc-lowering", "NumSiblingCalls" , "Number of sibling calls"}; | |||
131 | STATISTIC(ShufflesHandledWithVPERM, "Number of shuffles lowered to a VPERM")static llvm::Statistic ShufflesHandledWithVPERM = {"ppc-lowering" , "ShufflesHandledWithVPERM", "Number of shuffles lowered to a VPERM" }; | |||
132 | STATISTIC(NumDynamicAllocaProbed, "Number of dynamic stack allocation probed")static llvm::Statistic NumDynamicAllocaProbed = {"ppc-lowering" , "NumDynamicAllocaProbed", "Number of dynamic stack allocation probed" }; | |||
133 | ||||
134 | static bool isNByteElemShuffleMask(ShuffleVectorSDNode *, unsigned, int); | |||
135 | ||||
136 | static SDValue widenVec(SelectionDAG &DAG, SDValue Vec, const SDLoc &dl); | |||
137 | ||||
138 | static const char AIXSSPCanaryWordName[] = "__ssp_canary_word"; | |||
139 | ||||
140 | // FIXME: Remove this once the bug has been fixed! | |||
141 | extern cl::opt<bool> ANDIGlueBug; | |||
142 | ||||
143 | PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM, | |||
144 | const PPCSubtarget &STI) | |||
145 | : TargetLowering(TM), Subtarget(STI) { | |||
146 | // Initialize map that relates the PPC addressing modes to the computed flags | |||
147 | // of a load/store instruction. The map is used to determine the optimal | |||
148 | // addressing mode when selecting load and stores. | |||
149 | initializeAddrModeMap(); | |||
150 | // On PPC32/64, arguments smaller than 4/8 bytes are extended, so all | |||
151 | // arguments are at least 4/8 bytes aligned. | |||
152 | bool isPPC64 = Subtarget.isPPC64(); | |||
153 | setMinStackArgumentAlignment(isPPC64 ? Align(8) : Align(4)); | |||
154 | ||||
155 | // Set up the register classes. | |||
156 | addRegisterClass(MVT::i32, &PPC::GPRCRegClass); | |||
157 | if (!useSoftFloat()) { | |||
158 | if (hasSPE()) { | |||
159 | addRegisterClass(MVT::f32, &PPC::GPRCRegClass); | |||
160 | // EFPU2 APU only supports f32 | |||
161 | if (!Subtarget.hasEFPU2()) | |||
162 | addRegisterClass(MVT::f64, &PPC::SPERCRegClass); | |||
163 | } else { | |||
164 | addRegisterClass(MVT::f32, &PPC::F4RCRegClass); | |||
165 | addRegisterClass(MVT::f64, &PPC::F8RCRegClass); | |||
166 | } | |||
167 | } | |||
168 | ||||
169 | // Match BITREVERSE to customized fast code sequence in the td file. | |||
170 | setOperationAction(ISD::BITREVERSE, MVT::i32, Legal); | |||
171 | setOperationAction(ISD::BITREVERSE, MVT::i64, Legal); | |||
172 | ||||
173 | // Sub-word ATOMIC_CMP_SWAP need to ensure that the input is zero-extended. | |||
174 | setOperationAction(ISD::ATOMIC_CMP_SWAP, MVT::i32, Custom); | |||
175 | ||||
176 | // Custom lower inline assembly to check for special registers. | |||
177 | setOperationAction(ISD::INLINEASM, MVT::Other, Custom); | |||
178 | setOperationAction(ISD::INLINEASM_BR, MVT::Other, Custom); | |||
179 | ||||
180 | // PowerPC has an i16 but no i8 (or i1) SEXTLOAD. | |||
181 | for (MVT VT : MVT::integer_valuetypes()) { | |||
182 | setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote); | |||
183 | setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i8, Expand); | |||
184 | } | |||
185 | ||||
186 | if (Subtarget.isISA3_0()) { | |||
187 | setLoadExtAction(ISD::EXTLOAD, MVT::f64, MVT::f16, Legal); | |||
188 | setLoadExtAction(ISD::EXTLOAD, MVT::f32, MVT::f16, Legal); | |||
189 | setTruncStoreAction(MVT::f64, MVT::f16, Legal); | |||
190 | setTruncStoreAction(MVT::f32, MVT::f16, Legal); | |||
191 | } else { | |||
192 | // No extending loads from f16 or HW conversions back and forth. | |||
193 | setLoadExtAction(ISD::EXTLOAD, MVT::f64, MVT::f16, Expand); | |||
194 | setOperationAction(ISD::FP16_TO_FP, MVT::f64, Expand); | |||
195 | setOperationAction(ISD::FP_TO_FP16, MVT::f64, Expand); | |||
196 | setLoadExtAction(ISD::EXTLOAD, MVT::f32, MVT::f16, Expand); | |||
197 | setOperationAction(ISD::FP16_TO_FP, MVT::f32, Expand); | |||
198 | setOperationAction(ISD::FP_TO_FP16, MVT::f32, Expand); | |||
199 | setTruncStoreAction(MVT::f64, MVT::f16, Expand); | |||
200 | setTruncStoreAction(MVT::f32, MVT::f16, Expand); | |||
201 | } | |||
202 | ||||
203 | setTruncStoreAction(MVT::f64, MVT::f32, Expand); | |||
204 | ||||
205 | // PowerPC has pre-inc load and store's. | |||
206 | setIndexedLoadAction(ISD::PRE_INC, MVT::i1, Legal); | |||
207 | setIndexedLoadAction(ISD::PRE_INC, MVT::i8, Legal); | |||
208 | setIndexedLoadAction(ISD::PRE_INC, MVT::i16, Legal); | |||
209 | setIndexedLoadAction(ISD::PRE_INC, MVT::i32, Legal); | |||
210 | setIndexedLoadAction(ISD::PRE_INC, MVT::i64, Legal); | |||
211 | setIndexedStoreAction(ISD::PRE_INC, MVT::i1, Legal); | |||
212 | setIndexedStoreAction(ISD::PRE_INC, MVT::i8, Legal); | |||
213 | setIndexedStoreAction(ISD::PRE_INC, MVT::i16, Legal); | |||
214 | setIndexedStoreAction(ISD::PRE_INC, MVT::i32, Legal); | |||
215 | setIndexedStoreAction(ISD::PRE_INC, MVT::i64, Legal); | |||
216 | if (!Subtarget.hasSPE()) { | |||
217 | setIndexedLoadAction(ISD::PRE_INC, MVT::f32, Legal); | |||
218 | setIndexedLoadAction(ISD::PRE_INC, MVT::f64, Legal); | |||
219 | setIndexedStoreAction(ISD::PRE_INC, MVT::f32, Legal); | |||
220 | setIndexedStoreAction(ISD::PRE_INC, MVT::f64, Legal); | |||
221 | } | |||
222 | ||||
223 | // PowerPC uses ADDC/ADDE/SUBC/SUBE to propagate carry. | |||
224 | const MVT ScalarIntVTs[] = { MVT::i32, MVT::i64 }; | |||
225 | for (MVT VT : ScalarIntVTs) { | |||
226 | setOperationAction(ISD::ADDC, VT, Legal); | |||
227 | setOperationAction(ISD::ADDE, VT, Legal); | |||
228 | setOperationAction(ISD::SUBC, VT, Legal); | |||
229 | setOperationAction(ISD::SUBE, VT, Legal); | |||
230 | } | |||
231 | ||||
232 | if (Subtarget.useCRBits()) { | |||
233 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Expand); | |||
234 | ||||
235 | if (isPPC64 || Subtarget.hasFPCVT()) { | |||
236 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i1, Promote); | |||
237 | AddPromotedToType(ISD::STRICT_SINT_TO_FP, MVT::i1, | |||
238 | isPPC64 ? MVT::i64 : MVT::i32); | |||
239 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i1, Promote); | |||
240 | AddPromotedToType(ISD::STRICT_UINT_TO_FP, MVT::i1, | |||
241 | isPPC64 ? MVT::i64 : MVT::i32); | |||
242 | ||||
243 | setOperationAction(ISD::SINT_TO_FP, MVT::i1, Promote); | |||
244 | AddPromotedToType (ISD::SINT_TO_FP, MVT::i1, | |||
245 | isPPC64 ? MVT::i64 : MVT::i32); | |||
246 | setOperationAction(ISD::UINT_TO_FP, MVT::i1, Promote); | |||
247 | AddPromotedToType(ISD::UINT_TO_FP, MVT::i1, | |||
248 | isPPC64 ? MVT::i64 : MVT::i32); | |||
249 | ||||
250 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i1, Promote); | |||
251 | AddPromotedToType(ISD::STRICT_FP_TO_SINT, MVT::i1, | |||
252 | isPPC64 ? MVT::i64 : MVT::i32); | |||
253 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i1, Promote); | |||
254 | AddPromotedToType(ISD::STRICT_FP_TO_UINT, MVT::i1, | |||
255 | isPPC64 ? MVT::i64 : MVT::i32); | |||
256 | ||||
257 | setOperationAction(ISD::FP_TO_SINT, MVT::i1, Promote); | |||
258 | AddPromotedToType(ISD::FP_TO_SINT, MVT::i1, | |||
259 | isPPC64 ? MVT::i64 : MVT::i32); | |||
260 | setOperationAction(ISD::FP_TO_UINT, MVT::i1, Promote); | |||
261 | AddPromotedToType(ISD::FP_TO_UINT, MVT::i1, | |||
262 | isPPC64 ? MVT::i64 : MVT::i32); | |||
263 | } else { | |||
264 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i1, Custom); | |||
265 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i1, Custom); | |||
266 | setOperationAction(ISD::SINT_TO_FP, MVT::i1, Custom); | |||
267 | setOperationAction(ISD::UINT_TO_FP, MVT::i1, Custom); | |||
268 | } | |||
269 | ||||
270 | // PowerPC does not support direct load/store of condition registers. | |||
271 | setOperationAction(ISD::LOAD, MVT::i1, Custom); | |||
272 | setOperationAction(ISD::STORE, MVT::i1, Custom); | |||
273 | ||||
274 | // FIXME: Remove this once the ANDI glue bug is fixed: | |||
275 | if (ANDIGlueBug) | |||
276 | setOperationAction(ISD::TRUNCATE, MVT::i1, Custom); | |||
277 | ||||
278 | for (MVT VT : MVT::integer_valuetypes()) { | |||
279 | setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote); | |||
280 | setLoadExtAction(ISD::ZEXTLOAD, VT, MVT::i1, Promote); | |||
281 | setTruncStoreAction(VT, MVT::i1, Expand); | |||
282 | } | |||
283 | ||||
284 | addRegisterClass(MVT::i1, &PPC::CRBITRCRegClass); | |||
285 | } | |||
286 | ||||
287 | // Expand ppcf128 to i32 by hand for the benefit of llvm-gcc bootstrap on | |||
288 | // PPC (the libcall is not available). | |||
289 | setOperationAction(ISD::FP_TO_SINT, MVT::ppcf128, Custom); | |||
290 | setOperationAction(ISD::FP_TO_UINT, MVT::ppcf128, Custom); | |||
291 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::ppcf128, Custom); | |||
292 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::ppcf128, Custom); | |||
293 | ||||
294 | // We do not currently implement these libm ops for PowerPC. | |||
295 | setOperationAction(ISD::FFLOOR, MVT::ppcf128, Expand); | |||
296 | setOperationAction(ISD::FCEIL, MVT::ppcf128, Expand); | |||
297 | setOperationAction(ISD::FTRUNC, MVT::ppcf128, Expand); | |||
298 | setOperationAction(ISD::FRINT, MVT::ppcf128, Expand); | |||
299 | setOperationAction(ISD::FNEARBYINT, MVT::ppcf128, Expand); | |||
300 | setOperationAction(ISD::FREM, MVT::ppcf128, Expand); | |||
301 | ||||
302 | // PowerPC has no SREM/UREM instructions unless we are on P9 | |||
303 | // On P9 we may use a hardware instruction to compute the remainder. | |||
304 | // When the result of both the remainder and the division is required it is | |||
305 | // more efficient to compute the remainder from the result of the division | |||
306 | // rather than use the remainder instruction. The instructions are legalized | |||
307 | // directly because the DivRemPairsPass performs the transformation at the IR | |||
308 | // level. | |||
309 | if (Subtarget.isISA3_0()) { | |||
310 | setOperationAction(ISD::SREM, MVT::i32, Legal); | |||
311 | setOperationAction(ISD::UREM, MVT::i32, Legal); | |||
312 | setOperationAction(ISD::SREM, MVT::i64, Legal); | |||
313 | setOperationAction(ISD::UREM, MVT::i64, Legal); | |||
314 | } else { | |||
315 | setOperationAction(ISD::SREM, MVT::i32, Expand); | |||
316 | setOperationAction(ISD::UREM, MVT::i32, Expand); | |||
317 | setOperationAction(ISD::SREM, MVT::i64, Expand); | |||
318 | setOperationAction(ISD::UREM, MVT::i64, Expand); | |||
319 | } | |||
320 | ||||
321 | // Don't use SMUL_LOHI/UMUL_LOHI or SDIVREM/UDIVREM to lower SREM/UREM. | |||
322 | setOperationAction(ISD::UMUL_LOHI, MVT::i32, Expand); | |||
323 | setOperationAction(ISD::SMUL_LOHI, MVT::i32, Expand); | |||
324 | setOperationAction(ISD::UMUL_LOHI, MVT::i64, Expand); | |||
325 | setOperationAction(ISD::SMUL_LOHI, MVT::i64, Expand); | |||
326 | setOperationAction(ISD::UDIVREM, MVT::i32, Expand); | |||
327 | setOperationAction(ISD::SDIVREM, MVT::i32, Expand); | |||
328 | setOperationAction(ISD::UDIVREM, MVT::i64, Expand); | |||
329 | setOperationAction(ISD::SDIVREM, MVT::i64, Expand); | |||
330 | ||||
331 | // Handle constrained floating-point operations of scalar. | |||
332 | // TODO: Handle SPE specific operation. | |||
333 | setOperationAction(ISD::STRICT_FADD, MVT::f32, Legal); | |||
334 | setOperationAction(ISD::STRICT_FSUB, MVT::f32, Legal); | |||
335 | setOperationAction(ISD::STRICT_FMUL, MVT::f32, Legal); | |||
336 | setOperationAction(ISD::STRICT_FDIV, MVT::f32, Legal); | |||
337 | setOperationAction(ISD::STRICT_FP_ROUND, MVT::f32, Legal); | |||
338 | ||||
339 | setOperationAction(ISD::STRICT_FADD, MVT::f64, Legal); | |||
340 | setOperationAction(ISD::STRICT_FSUB, MVT::f64, Legal); | |||
341 | setOperationAction(ISD::STRICT_FMUL, MVT::f64, Legal); | |||
342 | setOperationAction(ISD::STRICT_FDIV, MVT::f64, Legal); | |||
343 | ||||
344 | if (!Subtarget.hasSPE()) { | |||
345 | setOperationAction(ISD::STRICT_FMA, MVT::f32, Legal); | |||
346 | setOperationAction(ISD::STRICT_FMA, MVT::f64, Legal); | |||
347 | } | |||
348 | ||||
349 | if (Subtarget.hasVSX()) { | |||
350 | setOperationAction(ISD::STRICT_FRINT, MVT::f32, Legal); | |||
351 | setOperationAction(ISD::STRICT_FRINT, MVT::f64, Legal); | |||
352 | } | |||
353 | ||||
354 | if (Subtarget.hasFSQRT()) { | |||
355 | setOperationAction(ISD::STRICT_FSQRT, MVT::f32, Legal); | |||
356 | setOperationAction(ISD::STRICT_FSQRT, MVT::f64, Legal); | |||
357 | } | |||
358 | ||||
359 | if (Subtarget.hasFPRND()) { | |||
360 | setOperationAction(ISD::STRICT_FFLOOR, MVT::f32, Legal); | |||
361 | setOperationAction(ISD::STRICT_FCEIL, MVT::f32, Legal); | |||
362 | setOperationAction(ISD::STRICT_FTRUNC, MVT::f32, Legal); | |||
363 | setOperationAction(ISD::STRICT_FROUND, MVT::f32, Legal); | |||
364 | ||||
365 | setOperationAction(ISD::STRICT_FFLOOR, MVT::f64, Legal); | |||
366 | setOperationAction(ISD::STRICT_FCEIL, MVT::f64, Legal); | |||
367 | setOperationAction(ISD::STRICT_FTRUNC, MVT::f64, Legal); | |||
368 | setOperationAction(ISD::STRICT_FROUND, MVT::f64, Legal); | |||
369 | } | |||
370 | ||||
371 | // We don't support sin/cos/sqrt/fmod/pow | |||
372 | setOperationAction(ISD::FSIN , MVT::f64, Expand); | |||
373 | setOperationAction(ISD::FCOS , MVT::f64, Expand); | |||
374 | setOperationAction(ISD::FSINCOS, MVT::f64, Expand); | |||
375 | setOperationAction(ISD::FREM , MVT::f64, Expand); | |||
376 | setOperationAction(ISD::FPOW , MVT::f64, Expand); | |||
377 | setOperationAction(ISD::FSIN , MVT::f32, Expand); | |||
378 | setOperationAction(ISD::FCOS , MVT::f32, Expand); | |||
379 | setOperationAction(ISD::FSINCOS, MVT::f32, Expand); | |||
380 | setOperationAction(ISD::FREM , MVT::f32, Expand); | |||
381 | setOperationAction(ISD::FPOW , MVT::f32, Expand); | |||
382 | if (Subtarget.hasSPE()) { | |||
383 | setOperationAction(ISD::FMA , MVT::f64, Expand); | |||
384 | setOperationAction(ISD::FMA , MVT::f32, Expand); | |||
385 | } else { | |||
386 | setOperationAction(ISD::FMA , MVT::f64, Legal); | |||
387 | setOperationAction(ISD::FMA , MVT::f32, Legal); | |||
388 | } | |||
389 | ||||
390 | if (Subtarget.hasSPE()) | |||
391 | setLoadExtAction(ISD::EXTLOAD, MVT::f64, MVT::f32, Expand); | |||
392 | ||||
393 | setOperationAction(ISD::FLT_ROUNDS_, MVT::i32, Custom); | |||
394 | ||||
395 | // If we're enabling GP optimizations, use hardware square root | |||
396 | if (!Subtarget.hasFSQRT() && | |||
397 | !(TM.Options.UnsafeFPMath && Subtarget.hasFRSQRTE() && | |||
398 | Subtarget.hasFRE())) | |||
399 | setOperationAction(ISD::FSQRT, MVT::f64, Expand); | |||
400 | ||||
401 | if (!Subtarget.hasFSQRT() && | |||
402 | !(TM.Options.UnsafeFPMath && Subtarget.hasFRSQRTES() && | |||
403 | Subtarget.hasFRES())) | |||
404 | setOperationAction(ISD::FSQRT, MVT::f32, Expand); | |||
405 | ||||
406 | if (Subtarget.hasFCPSGN()) { | |||
407 | setOperationAction(ISD::FCOPYSIGN, MVT::f64, Legal); | |||
408 | setOperationAction(ISD::FCOPYSIGN, MVT::f32, Legal); | |||
409 | } else { | |||
410 | setOperationAction(ISD::FCOPYSIGN, MVT::f64, Expand); | |||
411 | setOperationAction(ISD::FCOPYSIGN, MVT::f32, Expand); | |||
412 | } | |||
413 | ||||
414 | if (Subtarget.hasFPRND()) { | |||
415 | setOperationAction(ISD::FFLOOR, MVT::f64, Legal); | |||
416 | setOperationAction(ISD::FCEIL, MVT::f64, Legal); | |||
417 | setOperationAction(ISD::FTRUNC, MVT::f64, Legal); | |||
418 | setOperationAction(ISD::FROUND, MVT::f64, Legal); | |||
419 | ||||
420 | setOperationAction(ISD::FFLOOR, MVT::f32, Legal); | |||
421 | setOperationAction(ISD::FCEIL, MVT::f32, Legal); | |||
422 | setOperationAction(ISD::FTRUNC, MVT::f32, Legal); | |||
423 | setOperationAction(ISD::FROUND, MVT::f32, Legal); | |||
424 | } | |||
425 | ||||
426 | // PowerPC does not have BSWAP, but we can use vector BSWAP instruction xxbrd | |||
427 | // to speed up scalar BSWAP64. | |||
428 | // CTPOP or CTTZ were introduced in P8/P9 respectively | |||
429 | setOperationAction(ISD::BSWAP, MVT::i32 , Expand); | |||
430 | if (Subtarget.hasP9Vector() && Subtarget.isPPC64()) | |||
431 | setOperationAction(ISD::BSWAP, MVT::i64 , Custom); | |||
432 | else | |||
433 | setOperationAction(ISD::BSWAP, MVT::i64 , Expand); | |||
434 | if (Subtarget.isISA3_0()) { | |||
435 | setOperationAction(ISD::CTTZ , MVT::i32 , Legal); | |||
436 | setOperationAction(ISD::CTTZ , MVT::i64 , Legal); | |||
437 | } else { | |||
438 | setOperationAction(ISD::CTTZ , MVT::i32 , Expand); | |||
439 | setOperationAction(ISD::CTTZ , MVT::i64 , Expand); | |||
440 | } | |||
441 | ||||
442 | if (Subtarget.hasPOPCNTD() == PPCSubtarget::POPCNTD_Fast) { | |||
443 | setOperationAction(ISD::CTPOP, MVT::i32 , Legal); | |||
444 | setOperationAction(ISD::CTPOP, MVT::i64 , Legal); | |||
445 | } else { | |||
446 | setOperationAction(ISD::CTPOP, MVT::i32 , Expand); | |||
447 | setOperationAction(ISD::CTPOP, MVT::i64 , Expand); | |||
448 | } | |||
449 | ||||
450 | // PowerPC does not have ROTR | |||
451 | setOperationAction(ISD::ROTR, MVT::i32 , Expand); | |||
452 | setOperationAction(ISD::ROTR, MVT::i64 , Expand); | |||
453 | ||||
454 | if (!Subtarget.useCRBits()) { | |||
455 | // PowerPC does not have Select | |||
456 | setOperationAction(ISD::SELECT, MVT::i32, Expand); | |||
457 | setOperationAction(ISD::SELECT, MVT::i64, Expand); | |||
458 | setOperationAction(ISD::SELECT, MVT::f32, Expand); | |||
459 | setOperationAction(ISD::SELECT, MVT::f64, Expand); | |||
460 | } | |||
461 | ||||
462 | // PowerPC wants to turn select_cc of FP into fsel when possible. | |||
463 | setOperationAction(ISD::SELECT_CC, MVT::f32, Custom); | |||
464 | setOperationAction(ISD::SELECT_CC, MVT::f64, Custom); | |||
465 | ||||
466 | // PowerPC wants to optimize integer setcc a bit | |||
467 | if (!Subtarget.useCRBits()) | |||
468 | setOperationAction(ISD::SETCC, MVT::i32, Custom); | |||
469 | ||||
470 | if (Subtarget.hasFPU()) { | |||
471 | setOperationAction(ISD::STRICT_FSETCC, MVT::f32, Legal); | |||
472 | setOperationAction(ISD::STRICT_FSETCC, MVT::f64, Legal); | |||
473 | setOperationAction(ISD::STRICT_FSETCC, MVT::f128, Legal); | |||
474 | ||||
475 | setOperationAction(ISD::STRICT_FSETCCS, MVT::f32, Legal); | |||
476 | setOperationAction(ISD::STRICT_FSETCCS, MVT::f64, Legal); | |||
477 | setOperationAction(ISD::STRICT_FSETCCS, MVT::f128, Legal); | |||
478 | } | |||
479 | ||||
480 | // PowerPC does not have BRCOND which requires SetCC | |||
481 | if (!Subtarget.useCRBits()) | |||
482 | setOperationAction(ISD::BRCOND, MVT::Other, Expand); | |||
483 | ||||
484 | setOperationAction(ISD::BR_JT, MVT::Other, Expand); | |||
485 | ||||
486 | if (Subtarget.hasSPE()) { | |||
487 | // SPE has built-in conversions | |||
488 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i32, Legal); | |||
489 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i32, Legal); | |||
490 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i32, Legal); | |||
491 | setOperationAction(ISD::FP_TO_SINT, MVT::i32, Legal); | |||
492 | setOperationAction(ISD::SINT_TO_FP, MVT::i32, Legal); | |||
493 | setOperationAction(ISD::UINT_TO_FP, MVT::i32, Legal); | |||
494 | ||||
495 | // SPE supports signaling compare of f32/f64. | |||
496 | setOperationAction(ISD::STRICT_FSETCCS, MVT::f32, Legal); | |||
497 | setOperationAction(ISD::STRICT_FSETCCS, MVT::f64, Legal); | |||
498 | } else { | |||
499 | // PowerPC turns FP_TO_SINT into FCTIWZ and some load/stores. | |||
500 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i32, Custom); | |||
501 | setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom); | |||
502 | ||||
503 | // PowerPC does not have [U|S]INT_TO_FP | |||
504 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i32, Expand); | |||
505 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i32, Expand); | |||
506 | setOperationAction(ISD::SINT_TO_FP, MVT::i32, Expand); | |||
507 | setOperationAction(ISD::UINT_TO_FP, MVT::i32, Expand); | |||
508 | } | |||
509 | ||||
510 | if (Subtarget.hasDirectMove() && isPPC64) { | |||
511 | setOperationAction(ISD::BITCAST, MVT::f32, Legal); | |||
512 | setOperationAction(ISD::BITCAST, MVT::i32, Legal); | |||
513 | setOperationAction(ISD::BITCAST, MVT::i64, Legal); | |||
514 | setOperationAction(ISD::BITCAST, MVT::f64, Legal); | |||
515 | if (TM.Options.UnsafeFPMath) { | |||
516 | setOperationAction(ISD::LRINT, MVT::f64, Legal); | |||
517 | setOperationAction(ISD::LRINT, MVT::f32, Legal); | |||
518 | setOperationAction(ISD::LLRINT, MVT::f64, Legal); | |||
519 | setOperationAction(ISD::LLRINT, MVT::f32, Legal); | |||
520 | setOperationAction(ISD::LROUND, MVT::f64, Legal); | |||
521 | setOperationAction(ISD::LROUND, MVT::f32, Legal); | |||
522 | setOperationAction(ISD::LLROUND, MVT::f64, Legal); | |||
523 | setOperationAction(ISD::LLROUND, MVT::f32, Legal); | |||
524 | } | |||
525 | } else { | |||
526 | setOperationAction(ISD::BITCAST, MVT::f32, Expand); | |||
527 | setOperationAction(ISD::BITCAST, MVT::i32, Expand); | |||
528 | setOperationAction(ISD::BITCAST, MVT::i64, Expand); | |||
529 | setOperationAction(ISD::BITCAST, MVT::f64, Expand); | |||
530 | } | |||
531 | ||||
532 | // We cannot sextinreg(i1). Expand to shifts. | |||
533 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Expand); | |||
534 | ||||
535 | // NOTE: EH_SJLJ_SETJMP/_LONGJMP supported here is NOT intended to support | |||
536 | // SjLj exception handling but a light-weight setjmp/longjmp replacement to | |||
537 | // support continuation, user-level threading, and etc.. As a result, no | |||
538 | // other SjLj exception interfaces are implemented and please don't build | |||
539 | // your own exception handling based on them. | |||
540 | // LLVM/Clang supports zero-cost DWARF exception handling. | |||
541 | setOperationAction(ISD::EH_SJLJ_SETJMP, MVT::i32, Custom); | |||
542 | setOperationAction(ISD::EH_SJLJ_LONGJMP, MVT::Other, Custom); | |||
543 | ||||
544 | // We want to legalize GlobalAddress and ConstantPool nodes into the | |||
545 | // appropriate instructions to materialize the address. | |||
546 | setOperationAction(ISD::GlobalAddress, MVT::i32, Custom); | |||
547 | setOperationAction(ISD::GlobalTLSAddress, MVT::i32, Custom); | |||
548 | setOperationAction(ISD::BlockAddress, MVT::i32, Custom); | |||
549 | setOperationAction(ISD::ConstantPool, MVT::i32, Custom); | |||
550 | setOperationAction(ISD::JumpTable, MVT::i32, Custom); | |||
551 | setOperationAction(ISD::GlobalAddress, MVT::i64, Custom); | |||
552 | setOperationAction(ISD::GlobalTLSAddress, MVT::i64, Custom); | |||
553 | setOperationAction(ISD::BlockAddress, MVT::i64, Custom); | |||
554 | setOperationAction(ISD::ConstantPool, MVT::i64, Custom); | |||
555 | setOperationAction(ISD::JumpTable, MVT::i64, Custom); | |||
556 | ||||
557 | // TRAP is legal. | |||
558 | setOperationAction(ISD::TRAP, MVT::Other, Legal); | |||
559 | ||||
560 | // TRAMPOLINE is custom lowered. | |||
561 | setOperationAction(ISD::INIT_TRAMPOLINE, MVT::Other, Custom); | |||
562 | setOperationAction(ISD::ADJUST_TRAMPOLINE, MVT::Other, Custom); | |||
563 | ||||
564 | // VASTART needs to be custom lowered to use the VarArgsFrameIndex | |||
565 | setOperationAction(ISD::VASTART , MVT::Other, Custom); | |||
566 | ||||
567 | if (Subtarget.is64BitELFABI()) { | |||
568 | // VAARG always uses double-word chunks, so promote anything smaller. | |||
569 | setOperationAction(ISD::VAARG, MVT::i1, Promote); | |||
570 | AddPromotedToType(ISD::VAARG, MVT::i1, MVT::i64); | |||
571 | setOperationAction(ISD::VAARG, MVT::i8, Promote); | |||
572 | AddPromotedToType(ISD::VAARG, MVT::i8, MVT::i64); | |||
573 | setOperationAction(ISD::VAARG, MVT::i16, Promote); | |||
574 | AddPromotedToType(ISD::VAARG, MVT::i16, MVT::i64); | |||
575 | setOperationAction(ISD::VAARG, MVT::i32, Promote); | |||
576 | AddPromotedToType(ISD::VAARG, MVT::i32, MVT::i64); | |||
577 | setOperationAction(ISD::VAARG, MVT::Other, Expand); | |||
578 | } else if (Subtarget.is32BitELFABI()) { | |||
579 | // VAARG is custom lowered with the 32-bit SVR4 ABI. | |||
580 | setOperationAction(ISD::VAARG, MVT::Other, Custom); | |||
581 | setOperationAction(ISD::VAARG, MVT::i64, Custom); | |||
582 | } else | |||
583 | setOperationAction(ISD::VAARG, MVT::Other, Expand); | |||
584 | ||||
585 | // VACOPY is custom lowered with the 32-bit SVR4 ABI. | |||
586 | if (Subtarget.is32BitELFABI()) | |||
587 | setOperationAction(ISD::VACOPY , MVT::Other, Custom); | |||
588 | else | |||
589 | setOperationAction(ISD::VACOPY , MVT::Other, Expand); | |||
590 | ||||
591 | // Use the default implementation. | |||
592 | setOperationAction(ISD::VAEND , MVT::Other, Expand); | |||
593 | setOperationAction(ISD::STACKSAVE , MVT::Other, Expand); | |||
594 | setOperationAction(ISD::STACKRESTORE , MVT::Other, Custom); | |||
595 | setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i32 , Custom); | |||
596 | setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i64 , Custom); | |||
597 | setOperationAction(ISD::GET_DYNAMIC_AREA_OFFSET, MVT::i32, Custom); | |||
598 | setOperationAction(ISD::GET_DYNAMIC_AREA_OFFSET, MVT::i64, Custom); | |||
599 | setOperationAction(ISD::EH_DWARF_CFA, MVT::i32, Custom); | |||
600 | setOperationAction(ISD::EH_DWARF_CFA, MVT::i64, Custom); | |||
601 | ||||
602 | // We want to custom lower some of our intrinsics. | |||
603 | setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom); | |||
604 | ||||
605 | // To handle counter-based loop conditions. | |||
606 | setOperationAction(ISD::INTRINSIC_W_CHAIN, MVT::i1, Custom); | |||
607 | ||||
608 | setOperationAction(ISD::INTRINSIC_VOID, MVT::i8, Custom); | |||
609 | setOperationAction(ISD::INTRINSIC_VOID, MVT::i16, Custom); | |||
610 | setOperationAction(ISD::INTRINSIC_VOID, MVT::i32, Custom); | |||
611 | setOperationAction(ISD::INTRINSIC_VOID, MVT::Other, Custom); | |||
612 | ||||
613 | // Comparisons that require checking two conditions. | |||
614 | if (Subtarget.hasSPE()) { | |||
615 | setCondCodeAction(ISD::SETO, MVT::f32, Expand); | |||
616 | setCondCodeAction(ISD::SETO, MVT::f64, Expand); | |||
617 | setCondCodeAction(ISD::SETUO, MVT::f32, Expand); | |||
618 | setCondCodeAction(ISD::SETUO, MVT::f64, Expand); | |||
619 | } | |||
620 | setCondCodeAction(ISD::SETULT, MVT::f32, Expand); | |||
621 | setCondCodeAction(ISD::SETULT, MVT::f64, Expand); | |||
622 | setCondCodeAction(ISD::SETUGT, MVT::f32, Expand); | |||
623 | setCondCodeAction(ISD::SETUGT, MVT::f64, Expand); | |||
624 | setCondCodeAction(ISD::SETUEQ, MVT::f32, Expand); | |||
625 | setCondCodeAction(ISD::SETUEQ, MVT::f64, Expand); | |||
626 | setCondCodeAction(ISD::SETOGE, MVT::f32, Expand); | |||
627 | setCondCodeAction(ISD::SETOGE, MVT::f64, Expand); | |||
628 | setCondCodeAction(ISD::SETOLE, MVT::f32, Expand); | |||
629 | setCondCodeAction(ISD::SETOLE, MVT::f64, Expand); | |||
630 | setCondCodeAction(ISD::SETONE, MVT::f32, Expand); | |||
631 | setCondCodeAction(ISD::SETONE, MVT::f64, Expand); | |||
632 | ||||
633 | setOperationAction(ISD::STRICT_FP_EXTEND, MVT::f32, Legal); | |||
634 | setOperationAction(ISD::STRICT_FP_EXTEND, MVT::f64, Legal); | |||
635 | ||||
636 | if (Subtarget.has64BitSupport()) { | |||
637 | // They also have instructions for converting between i64 and fp. | |||
638 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i64, Custom); | |||
639 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i64, Expand); | |||
640 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i64, Custom); | |||
641 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i64, Expand); | |||
642 | setOperationAction(ISD::FP_TO_SINT, MVT::i64, Custom); | |||
643 | setOperationAction(ISD::FP_TO_UINT, MVT::i64, Expand); | |||
644 | setOperationAction(ISD::SINT_TO_FP, MVT::i64, Custom); | |||
645 | setOperationAction(ISD::UINT_TO_FP, MVT::i64, Expand); | |||
646 | // This is just the low 32 bits of a (signed) fp->i64 conversion. | |||
647 | // We cannot do this with Promote because i64 is not a legal type. | |||
648 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i32, Custom); | |||
649 | setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom); | |||
650 | ||||
651 | if (Subtarget.hasLFIWAX() || Subtarget.isPPC64()) { | |||
652 | setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom); | |||
653 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i32, Custom); | |||
654 | } | |||
655 | } else { | |||
656 | // PowerPC does not have FP_TO_UINT on 32-bit implementations. | |||
657 | if (Subtarget.hasSPE()) { | |||
658 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i32, Legal); | |||
659 | setOperationAction(ISD::FP_TO_UINT, MVT::i32, Legal); | |||
660 | } else { | |||
661 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i32, Expand); | |||
662 | setOperationAction(ISD::FP_TO_UINT, MVT::i32, Expand); | |||
663 | } | |||
664 | } | |||
665 | ||||
666 | // With the instructions enabled under FPCVT, we can do everything. | |||
667 | if (Subtarget.hasFPCVT()) { | |||
668 | if (Subtarget.has64BitSupport()) { | |||
669 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i64, Custom); | |||
670 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i64, Custom); | |||
671 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i64, Custom); | |||
672 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i64, Custom); | |||
673 | setOperationAction(ISD::FP_TO_SINT, MVT::i64, Custom); | |||
674 | setOperationAction(ISD::FP_TO_UINT, MVT::i64, Custom); | |||
675 | setOperationAction(ISD::SINT_TO_FP, MVT::i64, Custom); | |||
676 | setOperationAction(ISD::UINT_TO_FP, MVT::i64, Custom); | |||
677 | } | |||
678 | ||||
679 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i32, Custom); | |||
680 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i32, Custom); | |||
681 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i32, Custom); | |||
682 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i32, Custom); | |||
683 | setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom); | |||
684 | setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom); | |||
685 | setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom); | |||
686 | setOperationAction(ISD::UINT_TO_FP, MVT::i32, Custom); | |||
687 | } | |||
688 | ||||
689 | if (Subtarget.use64BitRegs()) { | |||
690 | // 64-bit PowerPC implementations can support i64 types directly | |||
691 | addRegisterClass(MVT::i64, &PPC::G8RCRegClass); | |||
692 | // BUILD_PAIR can't be handled natively, and should be expanded to shl/or | |||
693 | setOperationAction(ISD::BUILD_PAIR, MVT::i64, Expand); | |||
694 | // 64-bit PowerPC wants to expand i128 shifts itself. | |||
695 | setOperationAction(ISD::SHL_PARTS, MVT::i64, Custom); | |||
696 | setOperationAction(ISD::SRA_PARTS, MVT::i64, Custom); | |||
697 | setOperationAction(ISD::SRL_PARTS, MVT::i64, Custom); | |||
698 | } else { | |||
699 | // 32-bit PowerPC wants to expand i64 shifts itself. | |||
700 | setOperationAction(ISD::SHL_PARTS, MVT::i32, Custom); | |||
701 | setOperationAction(ISD::SRA_PARTS, MVT::i32, Custom); | |||
702 | setOperationAction(ISD::SRL_PARTS, MVT::i32, Custom); | |||
703 | } | |||
704 | ||||
705 | // PowerPC has better expansions for funnel shifts than the generic | |||
706 | // TargetLowering::expandFunnelShift. | |||
707 | if (Subtarget.has64BitSupport()) { | |||
708 | setOperationAction(ISD::FSHL, MVT::i64, Custom); | |||
709 | setOperationAction(ISD::FSHR, MVT::i64, Custom); | |||
710 | } | |||
711 | setOperationAction(ISD::FSHL, MVT::i32, Custom); | |||
712 | setOperationAction(ISD::FSHR, MVT::i32, Custom); | |||
713 | ||||
714 | if (Subtarget.hasVSX()) { | |||
715 | setOperationAction(ISD::FMAXNUM_IEEE, MVT::f64, Legal); | |||
716 | setOperationAction(ISD::FMAXNUM_IEEE, MVT::f32, Legal); | |||
717 | setOperationAction(ISD::FMINNUM_IEEE, MVT::f64, Legal); | |||
718 | setOperationAction(ISD::FMINNUM_IEEE, MVT::f32, Legal); | |||
719 | } | |||
720 | ||||
721 | if (Subtarget.hasAltivec()) { | |||
722 | for (MVT VT : { MVT::v16i8, MVT::v8i16, MVT::v4i32 }) { | |||
723 | setOperationAction(ISD::SADDSAT, VT, Legal); | |||
724 | setOperationAction(ISD::SSUBSAT, VT, Legal); | |||
725 | setOperationAction(ISD::UADDSAT, VT, Legal); | |||
726 | setOperationAction(ISD::USUBSAT, VT, Legal); | |||
727 | } | |||
728 | // First set operation action for all vector types to expand. Then we | |||
729 | // will selectively turn on ones that can be effectively codegen'd. | |||
730 | for (MVT VT : MVT::fixedlen_vector_valuetypes()) { | |||
731 | // add/sub are legal for all supported vector VT's. | |||
732 | setOperationAction(ISD::ADD, VT, Legal); | |||
733 | setOperationAction(ISD::SUB, VT, Legal); | |||
734 | ||||
735 | // For v2i64, these are only valid with P8Vector. This is corrected after | |||
736 | // the loop. | |||
737 | if (VT.getSizeInBits() <= 128 && VT.getScalarSizeInBits() <= 64) { | |||
738 | setOperationAction(ISD::SMAX, VT, Legal); | |||
739 | setOperationAction(ISD::SMIN, VT, Legal); | |||
740 | setOperationAction(ISD::UMAX, VT, Legal); | |||
741 | setOperationAction(ISD::UMIN, VT, Legal); | |||
742 | } | |||
743 | else { | |||
744 | setOperationAction(ISD::SMAX, VT, Expand); | |||
745 | setOperationAction(ISD::SMIN, VT, Expand); | |||
746 | setOperationAction(ISD::UMAX, VT, Expand); | |||
747 | setOperationAction(ISD::UMIN, VT, Expand); | |||
748 | } | |||
749 | ||||
750 | if (Subtarget.hasVSX()) { | |||
751 | setOperationAction(ISD::FMAXNUM, VT, Legal); | |||
752 | setOperationAction(ISD::FMINNUM, VT, Legal); | |||
753 | } | |||
754 | ||||
755 | // Vector instructions introduced in P8 | |||
756 | if (Subtarget.hasP8Altivec() && (VT.SimpleTy != MVT::v1i128)) { | |||
757 | setOperationAction(ISD::CTPOP, VT, Legal); | |||
758 | setOperationAction(ISD::CTLZ, VT, Legal); | |||
759 | } | |||
760 | else { | |||
761 | setOperationAction(ISD::CTPOP, VT, Expand); | |||
762 | setOperationAction(ISD::CTLZ, VT, Expand); | |||
763 | } | |||
764 | ||||
765 | // Vector instructions introduced in P9 | |||
766 | if (Subtarget.hasP9Altivec() && (VT.SimpleTy != MVT::v1i128)) | |||
767 | setOperationAction(ISD::CTTZ, VT, Legal); | |||
768 | else | |||
769 | setOperationAction(ISD::CTTZ, VT, Expand); | |||
770 | ||||
771 | // We promote all shuffles to v16i8. | |||
772 | setOperationAction(ISD::VECTOR_SHUFFLE, VT, Promote); | |||
773 | AddPromotedToType (ISD::VECTOR_SHUFFLE, VT, MVT::v16i8); | |||
774 | ||||
775 | // We promote all non-typed operations to v4i32. | |||
776 | setOperationAction(ISD::AND , VT, Promote); | |||
777 | AddPromotedToType (ISD::AND , VT, MVT::v4i32); | |||
778 | setOperationAction(ISD::OR , VT, Promote); | |||
779 | AddPromotedToType (ISD::OR , VT, MVT::v4i32); | |||
780 | setOperationAction(ISD::XOR , VT, Promote); | |||
781 | AddPromotedToType (ISD::XOR , VT, MVT::v4i32); | |||
782 | setOperationAction(ISD::LOAD , VT, Promote); | |||
783 | AddPromotedToType (ISD::LOAD , VT, MVT::v4i32); | |||
784 | setOperationAction(ISD::SELECT, VT, Promote); | |||
785 | AddPromotedToType (ISD::SELECT, VT, MVT::v4i32); | |||
786 | setOperationAction(ISD::VSELECT, VT, Legal); | |||
787 | setOperationAction(ISD::SELECT_CC, VT, Promote); | |||
788 | AddPromotedToType (ISD::SELECT_CC, VT, MVT::v4i32); | |||
789 | setOperationAction(ISD::STORE, VT, Promote); | |||
790 | AddPromotedToType (ISD::STORE, VT, MVT::v4i32); | |||
791 | ||||
792 | // No other operations are legal. | |||
793 | setOperationAction(ISD::MUL , VT, Expand); | |||
794 | setOperationAction(ISD::SDIV, VT, Expand); | |||
795 | setOperationAction(ISD::SREM, VT, Expand); | |||
796 | setOperationAction(ISD::UDIV, VT, Expand); | |||
797 | setOperationAction(ISD::UREM, VT, Expand); | |||
798 | setOperationAction(ISD::FDIV, VT, Expand); | |||
799 | setOperationAction(ISD::FREM, VT, Expand); | |||
800 | setOperationAction(ISD::FNEG, VT, Expand); | |||
801 | setOperationAction(ISD::FSQRT, VT, Expand); | |||
802 | setOperationAction(ISD::FLOG, VT, Expand); | |||
803 | setOperationAction(ISD::FLOG10, VT, Expand); | |||
804 | setOperationAction(ISD::FLOG2, VT, Expand); | |||
805 | setOperationAction(ISD::FEXP, VT, Expand); | |||
806 | setOperationAction(ISD::FEXP2, VT, Expand); | |||
807 | setOperationAction(ISD::FSIN, VT, Expand); | |||
808 | setOperationAction(ISD::FCOS, VT, Expand); | |||
809 | setOperationAction(ISD::FABS, VT, Expand); | |||
810 | setOperationAction(ISD::FFLOOR, VT, Expand); | |||
811 | setOperationAction(ISD::FCEIL, VT, Expand); | |||
812 | setOperationAction(ISD::FTRUNC, VT, Expand); | |||
813 | setOperationAction(ISD::FRINT, VT, Expand); | |||
814 | setOperationAction(ISD::FNEARBYINT, VT, Expand); | |||
815 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Expand); | |||
816 | setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Expand); | |||
817 | setOperationAction(ISD::BUILD_VECTOR, VT, Expand); | |||
818 | setOperationAction(ISD::MULHU, VT, Expand); | |||
819 | setOperationAction(ISD::MULHS, VT, Expand); | |||
820 | setOperationAction(ISD::UMUL_LOHI, VT, Expand); | |||
821 | setOperationAction(ISD::SMUL_LOHI, VT, Expand); | |||
822 | setOperationAction(ISD::UDIVREM, VT, Expand); | |||
823 | setOperationAction(ISD::SDIVREM, VT, Expand); | |||
824 | setOperationAction(ISD::SCALAR_TO_VECTOR, VT, Expand); | |||
825 | setOperationAction(ISD::FPOW, VT, Expand); | |||
826 | setOperationAction(ISD::BSWAP, VT, Expand); | |||
827 | setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Expand); | |||
828 | setOperationAction(ISD::ROTL, VT, Expand); | |||
829 | setOperationAction(ISD::ROTR, VT, Expand); | |||
830 | ||||
831 | for (MVT InnerVT : MVT::fixedlen_vector_valuetypes()) { | |||
832 | setTruncStoreAction(VT, InnerVT, Expand); | |||
833 | setLoadExtAction(ISD::SEXTLOAD, VT, InnerVT, Expand); | |||
834 | setLoadExtAction(ISD::ZEXTLOAD, VT, InnerVT, Expand); | |||
835 | setLoadExtAction(ISD::EXTLOAD, VT, InnerVT, Expand); | |||
836 | } | |||
837 | } | |||
838 | setOperationAction(ISD::SELECT_CC, MVT::v4i32, Expand); | |||
839 | if (!Subtarget.hasP8Vector()) { | |||
840 | setOperationAction(ISD::SMAX, MVT::v2i64, Expand); | |||
841 | setOperationAction(ISD::SMIN, MVT::v2i64, Expand); | |||
842 | setOperationAction(ISD::UMAX, MVT::v2i64, Expand); | |||
843 | setOperationAction(ISD::UMIN, MVT::v2i64, Expand); | |||
844 | } | |||
845 | ||||
846 | // We can custom expand all VECTOR_SHUFFLEs to VPERM, others we can handle | |||
847 | // with merges, splats, etc. | |||
848 | setOperationAction(ISD::VECTOR_SHUFFLE, MVT::v16i8, Custom); | |||
849 | ||||
850 | // Vector truncates to sub-word integer that fit in an Altivec/VSX register | |||
851 | // are cheap, so handle them before they get expanded to scalar. | |||
852 | setOperationAction(ISD::TRUNCATE, MVT::v8i8, Custom); | |||
853 | setOperationAction(ISD::TRUNCATE, MVT::v4i8, Custom); | |||
854 | setOperationAction(ISD::TRUNCATE, MVT::v2i8, Custom); | |||
855 | setOperationAction(ISD::TRUNCATE, MVT::v4i16, Custom); | |||
856 | setOperationAction(ISD::TRUNCATE, MVT::v2i16, Custom); | |||
857 | ||||
858 | setOperationAction(ISD::AND , MVT::v4i32, Legal); | |||
859 | setOperationAction(ISD::OR , MVT::v4i32, Legal); | |||
860 | setOperationAction(ISD::XOR , MVT::v4i32, Legal); | |||
861 | setOperationAction(ISD::LOAD , MVT::v4i32, Legal); | |||
862 | setOperationAction(ISD::SELECT, MVT::v4i32, | |||
863 | Subtarget.useCRBits() ? Legal : Expand); | |||
864 | setOperationAction(ISD::STORE , MVT::v4i32, Legal); | |||
865 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::v4i32, Legal); | |||
866 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::v4i32, Legal); | |||
867 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v4i32, Legal); | |||
868 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v4i32, Legal); | |||
869 | setOperationAction(ISD::FP_TO_SINT, MVT::v4i32, Legal); | |||
870 | setOperationAction(ISD::FP_TO_UINT, MVT::v4i32, Legal); | |||
871 | setOperationAction(ISD::SINT_TO_FP, MVT::v4i32, Legal); | |||
872 | setOperationAction(ISD::UINT_TO_FP, MVT::v4i32, Legal); | |||
873 | setOperationAction(ISD::FFLOOR, MVT::v4f32, Legal); | |||
874 | setOperationAction(ISD::FCEIL, MVT::v4f32, Legal); | |||
875 | setOperationAction(ISD::FTRUNC, MVT::v4f32, Legal); | |||
876 | setOperationAction(ISD::FNEARBYINT, MVT::v4f32, Legal); | |||
877 | ||||
878 | // Custom lowering ROTL v1i128 to VECTOR_SHUFFLE v16i8. | |||
879 | setOperationAction(ISD::ROTL, MVT::v1i128, Custom); | |||
880 | // With hasAltivec set, we can lower ISD::ROTL to vrl(b|h|w). | |||
881 | if (Subtarget.hasAltivec()) | |||
882 | for (auto VT : {MVT::v4i32, MVT::v8i16, MVT::v16i8}) | |||
883 | setOperationAction(ISD::ROTL, VT, Legal); | |||
884 | // With hasP8Altivec set, we can lower ISD::ROTL to vrld. | |||
885 | if (Subtarget.hasP8Altivec()) | |||
886 | setOperationAction(ISD::ROTL, MVT::v2i64, Legal); | |||
887 | ||||
888 | addRegisterClass(MVT::v4f32, &PPC::VRRCRegClass); | |||
889 | addRegisterClass(MVT::v4i32, &PPC::VRRCRegClass); | |||
890 | addRegisterClass(MVT::v8i16, &PPC::VRRCRegClass); | |||
891 | addRegisterClass(MVT::v16i8, &PPC::VRRCRegClass); | |||
892 | ||||
893 | setOperationAction(ISD::MUL, MVT::v4f32, Legal); | |||
894 | setOperationAction(ISD::FMA, MVT::v4f32, Legal); | |||
895 | ||||
896 | if (Subtarget.hasVSX()) { | |||
897 | setOperationAction(ISD::FDIV, MVT::v4f32, Legal); | |||
898 | setOperationAction(ISD::FSQRT, MVT::v4f32, Legal); | |||
899 | setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v2f64, Custom); | |||
900 | } | |||
901 | ||||
902 | if (Subtarget.hasP8Altivec()) | |||
903 | setOperationAction(ISD::MUL, MVT::v4i32, Legal); | |||
904 | else | |||
905 | setOperationAction(ISD::MUL, MVT::v4i32, Custom); | |||
906 | ||||
907 | if (Subtarget.isISA3_1()) { | |||
908 | setOperationAction(ISD::MUL, MVT::v2i64, Legal); | |||
909 | setOperationAction(ISD::MULHS, MVT::v2i64, Legal); | |||
910 | setOperationAction(ISD::MULHU, MVT::v2i64, Legal); | |||
911 | setOperationAction(ISD::MULHS, MVT::v4i32, Legal); | |||
912 | setOperationAction(ISD::MULHU, MVT::v4i32, Legal); | |||
913 | setOperationAction(ISD::UDIV, MVT::v2i64, Legal); | |||
914 | setOperationAction(ISD::SDIV, MVT::v2i64, Legal); | |||
915 | setOperationAction(ISD::UDIV, MVT::v4i32, Legal); | |||
916 | setOperationAction(ISD::SDIV, MVT::v4i32, Legal); | |||
917 | setOperationAction(ISD::UREM, MVT::v2i64, Legal); | |||
918 | setOperationAction(ISD::SREM, MVT::v2i64, Legal); | |||
919 | setOperationAction(ISD::UREM, MVT::v4i32, Legal); | |||
920 | setOperationAction(ISD::SREM, MVT::v4i32, Legal); | |||
921 | setOperationAction(ISD::UREM, MVT::v1i128, Legal); | |||
922 | setOperationAction(ISD::SREM, MVT::v1i128, Legal); | |||
923 | setOperationAction(ISD::UDIV, MVT::v1i128, Legal); | |||
924 | setOperationAction(ISD::SDIV, MVT::v1i128, Legal); | |||
925 | setOperationAction(ISD::ROTL, MVT::v1i128, Legal); | |||
926 | } | |||
927 | ||||
928 | setOperationAction(ISD::MUL, MVT::v8i16, Legal); | |||
929 | setOperationAction(ISD::MUL, MVT::v16i8, Custom); | |||
930 | ||||
931 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v4f32, Custom); | |||
932 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v4i32, Custom); | |||
933 | ||||
934 | setOperationAction(ISD::BUILD_VECTOR, MVT::v16i8, Custom); | |||
935 | setOperationAction(ISD::BUILD_VECTOR, MVT::v8i16, Custom); | |||
936 | setOperationAction(ISD::BUILD_VECTOR, MVT::v4i32, Custom); | |||
937 | setOperationAction(ISD::BUILD_VECTOR, MVT::v4f32, Custom); | |||
938 | ||||
939 | // Altivec does not contain unordered floating-point compare instructions | |||
940 | setCondCodeAction(ISD::SETUO, MVT::v4f32, Expand); | |||
941 | setCondCodeAction(ISD::SETUEQ, MVT::v4f32, Expand); | |||
942 | setCondCodeAction(ISD::SETO, MVT::v4f32, Expand); | |||
943 | setCondCodeAction(ISD::SETONE, MVT::v4f32, Expand); | |||
944 | ||||
945 | if (Subtarget.hasVSX()) { | |||
946 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v2f64, Legal); | |||
947 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v2f64, Legal); | |||
948 | if (Subtarget.hasP8Vector()) { | |||
949 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v4f32, Legal); | |||
950 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4f32, Legal); | |||
951 | } | |||
952 | if (Subtarget.hasDirectMove() && isPPC64) { | |||
953 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v16i8, Legal); | |||
954 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v8i16, Legal); | |||
955 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v4i32, Legal); | |||
956 | setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v2i64, Legal); | |||
957 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v16i8, Legal); | |||
958 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v8i16, Legal); | |||
959 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v4i32, Legal); | |||
960 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v2i64, Legal); | |||
961 | } | |||
962 | setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v2f64, Legal); | |||
963 | ||||
964 | // The nearbyint variants are not allowed to raise the inexact exception | |||
965 | // so we can only code-gen them with unsafe math. | |||
966 | if (TM.Options.UnsafeFPMath) { | |||
967 | setOperationAction(ISD::FNEARBYINT, MVT::f64, Legal); | |||
968 | setOperationAction(ISD::FNEARBYINT, MVT::f32, Legal); | |||
969 | } | |||
970 | ||||
971 | setOperationAction(ISD::FFLOOR, MVT::v2f64, Legal); | |||
972 | setOperationAction(ISD::FCEIL, MVT::v2f64, Legal); | |||
973 | setOperationAction(ISD::FTRUNC, MVT::v2f64, Legal); | |||
974 | setOperationAction(ISD::FNEARBYINT, MVT::v2f64, Legal); | |||
975 | setOperationAction(ISD::FRINT, MVT::v2f64, Legal); | |||
976 | setOperationAction(ISD::FROUND, MVT::v2f64, Legal); | |||
977 | setOperationAction(ISD::FROUND, MVT::f64, Legal); | |||
978 | setOperationAction(ISD::FRINT, MVT::f64, Legal); | |||
979 | ||||
980 | setOperationAction(ISD::FNEARBYINT, MVT::v4f32, Legal); | |||
981 | setOperationAction(ISD::FRINT, MVT::v4f32, Legal); | |||
982 | setOperationAction(ISD::FROUND, MVT::v4f32, Legal); | |||
983 | setOperationAction(ISD::FROUND, MVT::f32, Legal); | |||
984 | setOperationAction(ISD::FRINT, MVT::f32, Legal); | |||
985 | ||||
986 | setOperationAction(ISD::MUL, MVT::v2f64, Legal); | |||
987 | setOperationAction(ISD::FMA, MVT::v2f64, Legal); | |||
988 | ||||
989 | setOperationAction(ISD::FDIV, MVT::v2f64, Legal); | |||
990 | setOperationAction(ISD::FSQRT, MVT::v2f64, Legal); | |||
991 | ||||
992 | // Share the Altivec comparison restrictions. | |||
993 | setCondCodeAction(ISD::SETUO, MVT::v2f64, Expand); | |||
994 | setCondCodeAction(ISD::SETUEQ, MVT::v2f64, Expand); | |||
995 | setCondCodeAction(ISD::SETO, MVT::v2f64, Expand); | |||
996 | setCondCodeAction(ISD::SETONE, MVT::v2f64, Expand); | |||
997 | ||||
998 | setOperationAction(ISD::LOAD, MVT::v2f64, Legal); | |||
999 | setOperationAction(ISD::STORE, MVT::v2f64, Legal); | |||
1000 | ||||
1001 | setOperationAction(ISD::VECTOR_SHUFFLE, MVT::v2f64, Legal); | |||
1002 | ||||
1003 | if (Subtarget.hasP8Vector()) | |||
1004 | addRegisterClass(MVT::f32, &PPC::VSSRCRegClass); | |||
1005 | ||||
1006 | addRegisterClass(MVT::f64, &PPC::VSFRCRegClass); | |||
1007 | ||||
1008 | addRegisterClass(MVT::v4i32, &PPC::VSRCRegClass); | |||
1009 | addRegisterClass(MVT::v4f32, &PPC::VSRCRegClass); | |||
1010 | addRegisterClass(MVT::v2f64, &PPC::VSRCRegClass); | |||
1011 | ||||
1012 | if (Subtarget.hasP8Altivec()) { | |||
1013 | setOperationAction(ISD::SHL, MVT::v2i64, Legal); | |||
1014 | setOperationAction(ISD::SRA, MVT::v2i64, Legal); | |||
1015 | setOperationAction(ISD::SRL, MVT::v2i64, Legal); | |||
1016 | ||||
1017 | // 128 bit shifts can be accomplished via 3 instructions for SHL and | |||
1018 | // SRL, but not for SRA because of the instructions available: | |||
1019 | // VS{RL} and VS{RL}O. However due to direct move costs, it's not worth | |||
1020 | // doing | |||
1021 | setOperationAction(ISD::SHL, MVT::v1i128, Expand); | |||
1022 | setOperationAction(ISD::SRL, MVT::v1i128, Expand); | |||
1023 | setOperationAction(ISD::SRA, MVT::v1i128, Expand); | |||
1024 | ||||
1025 | setOperationAction(ISD::SETCC, MVT::v2i64, Legal); | |||
1026 | } | |||
1027 | else { | |||
1028 | setOperationAction(ISD::SHL, MVT::v2i64, Expand); | |||
1029 | setOperationAction(ISD::SRA, MVT::v2i64, Expand); | |||
1030 | setOperationAction(ISD::SRL, MVT::v2i64, Expand); | |||
1031 | ||||
1032 | setOperationAction(ISD::SETCC, MVT::v2i64, Custom); | |||
1033 | ||||
1034 | // VSX v2i64 only supports non-arithmetic operations. | |||
1035 | setOperationAction(ISD::ADD, MVT::v2i64, Expand); | |||
1036 | setOperationAction(ISD::SUB, MVT::v2i64, Expand); | |||
1037 | } | |||
1038 | ||||
1039 | if (Subtarget.isISA3_1()) | |||
1040 | setOperationAction(ISD::SETCC, MVT::v1i128, Legal); | |||
1041 | else | |||
1042 | setOperationAction(ISD::SETCC, MVT::v1i128, Expand); | |||
1043 | ||||
1044 | setOperationAction(ISD::LOAD, MVT::v2i64, Promote); | |||
1045 | AddPromotedToType (ISD::LOAD, MVT::v2i64, MVT::v2f64); | |||
1046 | setOperationAction(ISD::STORE, MVT::v2i64, Promote); | |||
1047 | AddPromotedToType (ISD::STORE, MVT::v2i64, MVT::v2f64); | |||
1048 | ||||
1049 | setOperationAction(ISD::VECTOR_SHUFFLE, MVT::v2i64, Legal); | |||
1050 | ||||
1051 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v2i64, Legal); | |||
1052 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v2i64, Legal); | |||
1053 | setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::v2i64, Legal); | |||
1054 | setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::v2i64, Legal); | |||
1055 | setOperationAction(ISD::SINT_TO_FP, MVT::v2i64, Legal); | |||
1056 | setOperationAction(ISD::UINT_TO_FP, MVT::v2i64, Legal); | |||
1057 | setOperationAction(ISD::FP_TO_SINT, MVT::v2i64, Legal); | |||
1058 | setOperationAction(ISD::FP_TO_UINT, MVT::v2i64, Legal); | |||
1059 | ||||
1060 | // Custom handling for partial vectors of integers converted to | |||
1061 | // floating point. We already have optimal handling for v2i32 through | |||
1062 | // the DAG combine, so those aren't necessary. | |||
1063 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v2i8, Custom); | |||
1064 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v4i8, Custom); | |||
1065 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v2i16, Custom); | |||
1066 | setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::v4i16, Custom); | |||
1067 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v2i8, Custom); | |||
1068 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v4i8, Custom); | |||
1069 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v2i16, Custom); | |||
1070 | setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::v4i16, Custom); | |||
1071 | setOperationAction(ISD::UINT_TO_FP, MVT::v2i8, Custom); | |||
1072 | setOperationAction(ISD::UINT_TO_FP, MVT::v4i8, Custom); | |||
1073 | setOperationAction(ISD::UINT_TO_FP, MVT::v2i16, Custom); | |||
1074 | setOperationAction(ISD::UINT_TO_FP, MVT::v4i16, Custom); | |||
1075 | setOperationAction(ISD::SINT_TO_FP, MVT::v2i8, Custom); | |||
1076 | setOperationAction(ISD::SINT_TO_FP, MVT::v4i8, Custom); | |||
1077 | setOperationAction(ISD::SINT_TO_FP, MVT::v2i16, Custom); | |||
1078 | setOperationAction(ISD::SINT_TO_FP, MVT::v4i16, Custom); | |||
1079 | ||||
1080 | setOperationAction(ISD::FNEG, MVT::v4f32, Legal); | |||
1081 | setOperationAction(ISD::FNEG, MVT::v2f64, Legal); | |||
1082 | setOperationAction(ISD::FABS, MVT::v4f32, Legal); | |||
1083 | setOperationAction(ISD::FABS, MVT::v2f64, Legal); | |||
1084 | setOperationAction(ISD::FCOPYSIGN, MVT::v4f32, Legal); | |||
1085 | setOperationAction(ISD::FCOPYSIGN, MVT::v2f64, Legal); | |||
1086 | ||||
1087 | setOperationAction(ISD::BUILD_VECTOR, MVT::v2i64, Custom); | |||
1088 | setOperationAction(ISD::BUILD_VECTOR, MVT::v2f64, Custom); | |||
1089 | ||||
1090 | // Handle constrained floating-point operations of vector. | |||
1091 | // The predictor is `hasVSX` because altivec instruction has | |||
1092 | // no exception but VSX vector instruction has. | |||
1093 | setOperationAction(ISD::STRICT_FADD, MVT::v4f32, Legal); | |||
1094 | setOperationAction(ISD::STRICT_FSUB, MVT::v4f32, Legal); | |||
1095 | setOperationAction(ISD::STRICT_FMUL, MVT::v4f32, Legal); | |||
1096 | setOperationAction(ISD::STRICT_FDIV, MVT::v4f32, Legal); | |||
1097 | setOperationAction(ISD::STRICT_FMA, MVT::v4f32, Legal); | |||
1098 | setOperationAction(ISD::STRICT_FSQRT, MVT::v4f32, Legal); | |||
1099 | setOperationAction(ISD::STRICT_FMAXNUM, MVT::v4f32, Legal); | |||
1100 | setOperationAction(ISD::STRICT_FMINNUM, MVT::v4f32, Legal); | |||
1101 | setOperationAction(ISD::STRICT_FRINT, MVT::v4f32, Legal); | |||
1102 | setOperationAction(ISD::STRICT_FFLOOR, MVT::v4f32, Legal); | |||
1103 | setOperationAction(ISD::STRICT_FCEIL, MVT::v4f32, Legal); | |||
1104 | setOperationAction(ISD::STRICT_FTRUNC, MVT::v4f32, Legal); | |||
1105 | setOperationAction(ISD::STRICT_FROUND, MVT::v4f32, Legal); | |||
1106 | ||||
1107 | setOperationAction(ISD::STRICT_FADD, MVT::v2f64, Legal); | |||
1108 | setOperationAction(ISD::STRICT_FSUB, MVT::v2f64, Legal); | |||
1109 | setOperationAction(ISD::STRICT_FMUL, MVT::v2f64, Legal); | |||
1110 | setOperationAction(ISD::STRICT_FDIV, MVT::v2f64, Legal); | |||
1111 | setOperationAction(ISD::STRICT_FMA, MVT::v2f64, Legal); | |||
1112 | setOperationAction(ISD::STRICT_FSQRT, MVT::v2f64, Legal); | |||
1113 | setOperationAction(ISD::STRICT_FMAXNUM, MVT::v2f64, Legal); | |||
1114 | setOperationAction(ISD::STRICT_FMINNUM, MVT::v2f64, Legal); | |||
1115 | setOperationAction(ISD::STRICT_FRINT, MVT::v2f64, Legal); | |||
1116 | setOperationAction(ISD::STRICT_FFLOOR, MVT::v2f64, Legal); | |||
1117 | setOperationAction(ISD::STRICT_FCEIL, MVT::v2f64, Legal); | |||
1118 | setOperationAction(ISD::STRICT_FTRUNC, MVT::v2f64, Legal); | |||
1119 | setOperationAction(ISD::STRICT_FROUND, MVT::v2f64, Legal); | |||
1120 | ||||
1121 | addRegisterClass(MVT::v2i64, &PPC::VSRCRegClass); | |||
1122 | addRegisterClass(MVT::f128, &PPC::VRRCRegClass); | |||
1123 | ||||
1124 | for (MVT FPT : MVT::fp_valuetypes()) | |||
1125 | setLoadExtAction(ISD::EXTLOAD, MVT::f128, FPT, Expand); | |||
1126 | ||||
1127 | // Expand the SELECT to SELECT_CC | |||
1128 | setOperationAction(ISD::SELECT, MVT::f128, Expand); | |||
1129 | ||||
1130 | setTruncStoreAction(MVT::f128, MVT::f64, Expand); | |||
1131 | setTruncStoreAction(MVT::f128, MVT::f32, Expand); | |||
1132 | ||||
1133 | // No implementation for these ops for PowerPC. | |||
1134 | setOperationAction(ISD::FSIN, MVT::f128, Expand); | |||
1135 | setOperationAction(ISD::FCOS, MVT::f128, Expand); | |||
1136 | setOperationAction(ISD::FPOW, MVT::f128, Expand); | |||
1137 | setOperationAction(ISD::FPOWI, MVT::f128, Expand); | |||
1138 | setOperationAction(ISD::FREM, MVT::f128, Expand); | |||
1139 | } | |||
1140 | ||||
1141 | if (Subtarget.hasP8Altivec()) { | |||
1142 | addRegisterClass(MVT::v2i64, &PPC::VRRCRegClass); | |||
1143 | addRegisterClass(MVT::v1i128, &PPC::VRRCRegClass); | |||
1144 | } | |||
1145 | ||||
1146 | if (Subtarget.hasP9Vector()) { | |||
1147 | setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4i32, Custom); | |||
1148 | setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v4f32, Custom); | |||
1149 | ||||
1150 | // 128 bit shifts can be accomplished via 3 instructions for SHL and | |||
1151 | // SRL, but not for SRA because of the instructions available: | |||
1152 | // VS{RL} and VS{RL}O. | |||
1153 | setOperationAction(ISD::SHL, MVT::v1i128, Legal); | |||
1154 | setOperationAction(ISD::SRL, MVT::v1i128, Legal); | |||
1155 | setOperationAction(ISD::SRA, MVT::v1i128, Expand); | |||
1156 | ||||
1157 | setOperationAction(ISD::FADD, MVT::f128, Legal); | |||
1158 | setOperationAction(ISD::FSUB, MVT::f128, Legal); | |||
1159 | setOperationAction(ISD::FDIV, MVT::f128, Legal); | |||
1160 | setOperationAction(ISD::FMUL, MVT::f128, Legal); | |||
1161 | setOperationAction(ISD::FP_EXTEND, MVT::f128, Legal); | |||
1162 | ||||
1163 | setOperationAction(ISD::FMA, MVT::f128, Legal); | |||
1164 | setCondCodeAction(ISD::SETULT, MVT::f128, Expand); | |||
1165 | setCondCodeAction(ISD::SETUGT, MVT::f128, Expand); | |||
1166 | setCondCodeAction(ISD::SETUEQ, MVT::f128, Expand); | |||
1167 | setCondCodeAction(ISD::SETOGE, MVT::f128, Expand); | |||
1168 | setCondCodeAction(ISD::SETOLE, MVT::f128, Expand); | |||
1169 | setCondCodeAction(ISD::SETONE, MVT::f128, Expand); | |||
1170 | ||||
1171 | setOperationAction(ISD::FTRUNC, MVT::f128, Legal); | |||
1172 | setOperationAction(ISD::FRINT, MVT::f128, Legal); | |||
1173 | setOperationAction(ISD::FFLOOR, MVT::f128, Legal); | |||
1174 | setOperationAction(ISD::FCEIL, MVT::f128, Legal); | |||
1175 | setOperationAction(ISD::FNEARBYINT, MVT::f128, Legal); | |||
1176 | setOperationAction(ISD::FROUND, MVT::f128, Legal); | |||
1177 | ||||
1178 | setOperationAction(ISD::FP_ROUND, MVT::f64, Legal); | |||
1179 | setOperationAction(ISD::FP_ROUND, MVT::f32, Legal); | |||
1180 | setOperationAction(ISD::BITCAST, MVT::i128, Custom); | |||
1181 | ||||
1182 | // Handle constrained floating-point operations of fp128 | |||
1183 | setOperationAction(ISD::STRICT_FADD, MVT::f128, Legal); | |||
1184 | setOperationAction(ISD::STRICT_FSUB, MVT::f128, Legal); | |||
1185 | setOperationAction(ISD::STRICT_FMUL, MVT::f128, Legal); | |||
1186 | setOperationAction(ISD::STRICT_FDIV, MVT::f128, Legal); | |||
1187 | setOperationAction(ISD::STRICT_FMA, MVT::f128, Legal); | |||
1188 | setOperationAction(ISD::STRICT_FSQRT, MVT::f128, Legal); | |||
1189 | setOperationAction(ISD::STRICT_FP_EXTEND, MVT::f128, Legal); | |||
1190 | setOperationAction(ISD::STRICT_FP_ROUND, MVT::f64, Legal); | |||
1191 | setOperationAction(ISD::STRICT_FP_ROUND, MVT::f32, Legal); | |||
1192 | setOperationAction(ISD::STRICT_FRINT, MVT::f128, Legal); | |||
1193 | setOperationAction(ISD::STRICT_FNEARBYINT, MVT::f128, Legal); | |||
1194 | setOperationAction(ISD::STRICT_FFLOOR, MVT::f128, Legal); | |||
1195 | setOperationAction(ISD::STRICT_FCEIL, MVT::f128, Legal); | |||
1196 | setOperationAction(ISD::STRICT_FTRUNC, MVT::f128, Legal); | |||
1197 | setOperationAction(ISD::STRICT_FROUND, MVT::f128, Legal); | |||
1198 | setOperationAction(ISD::FP_EXTEND, MVT::v2f32, Custom); | |||
1199 | setOperationAction(ISD::BSWAP, MVT::v8i16, Legal); | |||
1200 | setOperationAction(ISD::BSWAP, MVT::v4i32, Legal); | |||
1201 | setOperationAction(ISD::BSWAP, MVT::v2i64, Legal); | |||
1202 | setOperationAction(ISD::BSWAP, MVT::v1i128, Legal); | |||
1203 | } else if (Subtarget.hasVSX()) { | |||
1204 | setOperationAction(ISD::LOAD, MVT::f128, Promote); | |||
1205 | setOperationAction(ISD::STORE, MVT::f128, Promote); | |||
1206 | ||||
1207 | AddPromotedToType(ISD::LOAD, MVT::f128, MVT::v4i32); | |||
1208 | AddPromotedToType(ISD::STORE, MVT::f128, MVT::v4i32); | |||
1209 | ||||
1210 | // Set FADD/FSUB as libcall to avoid the legalizer to expand the | |||
1211 | // fp_to_uint and int_to_fp. | |||
1212 | setOperationAction(ISD::FADD, MVT::f128, LibCall); | |||
1213 | setOperationAction(ISD::FSUB, MVT::f128, LibCall); | |||
1214 | ||||
1215 | setOperationAction(ISD::FMUL, MVT::f128, Expand); | |||
1216 | setOperationAction(ISD::FDIV, MVT::f128, Expand); | |||
1217 | setOperationAction(ISD::FNEG, MVT::f128, Expand); | |||
1218 | setOperationAction(ISD::FABS, MVT::f128, Expand); | |||
1219 | setOperationAction(ISD::FSQRT, MVT::f128, Expand); | |||
1220 | setOperationAction(ISD::FMA, MVT::f128, Expand); | |||
1221 | setOperationAction(ISD::FCOPYSIGN, MVT::f128, Expand); | |||
1222 | ||||
1223 | // Expand the fp_extend if the target type is fp128. | |||
1224 | setOperationAction(ISD::FP_EXTEND, MVT::f128, Expand); | |||
1225 | setOperationAction(ISD::STRICT_FP_EXTEND, MVT::f128, Expand); | |||
1226 | ||||
1227 | // Expand the fp_round if the source type is fp128. | |||
1228 | for (MVT VT : {MVT::f32, MVT::f64}) { | |||
1229 | setOperationAction(ISD::FP_ROUND, VT, Custom); | |||
1230 | setOperationAction(ISD::STRICT_FP_ROUND, VT, Custom); | |||
1231 | } | |||
1232 | ||||
1233 | setOperationAction(ISD::SETCC, MVT::f128, Custom); | |||
1234 | setOperationAction(ISD::STRICT_FSETCC, MVT::f128, Custom); | |||
1235 | setOperationAction(ISD::STRICT_FSETCCS, MVT::f128, Custom); | |||
1236 | setOperationAction(ISD::BR_CC, MVT::f128, Expand); | |||
1237 | ||||
1238 | // Lower following f128 select_cc pattern: | |||
1239 | // select_cc x, y, tv, fv, cc -> select_cc (setcc x, y, cc), 0, tv, fv, NE | |||
1240 | setOperationAction(ISD::SELECT_CC, MVT::f128, Custom); | |||
1241 | ||||
1242 | // We need to handle f128 SELECT_CC with integer result type. | |||
1243 | setOperationAction(ISD::SELECT_CC, MVT::i32, Custom); | |||
1244 | setOperationAction(ISD::SELECT_CC, MVT::i64, isPPC64 ? Custom : Expand); | |||
1245 | } | |||
1246 | ||||
1247 | if (Subtarget.hasP9Altivec()) { | |||
1248 | setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v8i16, Custom); | |||
1249 | setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v16i8, Custom); | |||
1250 | ||||
1251 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i8, Legal); | |||
1252 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i16, Legal); | |||
1253 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i32, Legal); | |||
1254 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i8, Legal); | |||
1255 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i16, Legal); | |||
1256 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i32, Legal); | |||
1257 | setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i64, Legal); | |||
1258 | } | |||
1259 | ||||
1260 | if (Subtarget.isISA3_1()) | |||
1261 | setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::v2i64, Custom); | |||
1262 | } | |||
1263 | ||||
1264 | if (Subtarget.pairedVectorMemops()) { | |||
1265 | addRegisterClass(MVT::v256i1, &PPC::VSRpRCRegClass); | |||
1266 | setOperationAction(ISD::LOAD, MVT::v256i1, Custom); | |||
1267 | setOperationAction(ISD::STORE, MVT::v256i1, Custom); | |||
1268 | } | |||
1269 | if (Subtarget.hasMMA()) { | |||
1270 | addRegisterClass(MVT::v512i1, &PPC::UACCRCRegClass); | |||
1271 | setOperationAction(ISD::LOAD, MVT::v512i1, Custom); | |||
1272 | setOperationAction(ISD::STORE, MVT::v512i1, Custom); | |||
1273 | setOperationAction(ISD::BUILD_VECTOR, MVT::v512i1, Custom); | |||
1274 | } | |||
1275 | ||||
1276 | if (Subtarget.has64BitSupport()) | |||
1277 | setOperationAction(ISD::PREFETCH, MVT::Other, Legal); | |||
1278 | ||||
1279 | if (Subtarget.isISA3_1()) | |||
1280 | setOperationAction(ISD::SRA, MVT::v1i128, Legal); | |||
1281 | ||||
1282 | setOperationAction(ISD::READCYCLECOUNTER, MVT::i64, isPPC64 ? Legal : Custom); | |||
1283 | ||||
1284 | if (!isPPC64) { | |||
1285 | setOperationAction(ISD::ATOMIC_LOAD, MVT::i64, Expand); | |||
1286 | setOperationAction(ISD::ATOMIC_STORE, MVT::i64, Expand); | |||
1287 | } | |||
1288 | ||||
1289 | if (EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics()) | |||
1290 | setMaxAtomicSizeInBitsSupported(128); | |||
1291 | ||||
1292 | setBooleanContents(ZeroOrOneBooleanContent); | |||
1293 | ||||
1294 | if (Subtarget.hasAltivec()) { | |||
1295 | // Altivec instructions set fields to all zeros or all ones. | |||
1296 | setBooleanVectorContents(ZeroOrNegativeOneBooleanContent); | |||
1297 | } | |||
1298 | ||||
1299 | if (!isPPC64) { | |||
1300 | // These libcalls are not available in 32-bit. | |||
1301 | setLibcallName(RTLIB::SHL_I128, nullptr); | |||
1302 | setLibcallName(RTLIB::SRL_I128, nullptr); | |||
1303 | setLibcallName(RTLIB::SRA_I128, nullptr); | |||
1304 | } | |||
1305 | ||||
1306 | if (!isPPC64) | |||
1307 | setMaxAtomicSizeInBitsSupported(32); | |||
1308 | ||||
1309 | setStackPointerRegisterToSaveRestore(isPPC64 ? PPC::X1 : PPC::R1); | |||
1310 | ||||
1311 | // We have target-specific dag combine patterns for the following nodes: | |||
1312 | setTargetDAGCombine(ISD::ADD); | |||
1313 | setTargetDAGCombine(ISD::SHL); | |||
1314 | setTargetDAGCombine(ISD::SRA); | |||
1315 | setTargetDAGCombine(ISD::SRL); | |||
1316 | setTargetDAGCombine(ISD::MUL); | |||
1317 | setTargetDAGCombine(ISD::FMA); | |||
1318 | setTargetDAGCombine(ISD::SINT_TO_FP); | |||
1319 | setTargetDAGCombine(ISD::BUILD_VECTOR); | |||
1320 | if (Subtarget.hasFPCVT()) | |||
1321 | setTargetDAGCombine(ISD::UINT_TO_FP); | |||
1322 | setTargetDAGCombine(ISD::LOAD); | |||
1323 | setTargetDAGCombine(ISD::STORE); | |||
1324 | setTargetDAGCombine(ISD::BR_CC); | |||
1325 | if (Subtarget.useCRBits()) | |||
1326 | setTargetDAGCombine(ISD::BRCOND); | |||
1327 | setTargetDAGCombine(ISD::BSWAP); | |||
1328 | setTargetDAGCombine(ISD::INTRINSIC_WO_CHAIN); | |||
1329 | setTargetDAGCombine(ISD::INTRINSIC_W_CHAIN); | |||
1330 | setTargetDAGCombine(ISD::INTRINSIC_VOID); | |||
1331 | ||||
1332 | setTargetDAGCombine(ISD::SIGN_EXTEND); | |||
1333 | setTargetDAGCombine(ISD::ZERO_EXTEND); | |||
1334 | setTargetDAGCombine(ISD::ANY_EXTEND); | |||
1335 | ||||
1336 | setTargetDAGCombine(ISD::TRUNCATE); | |||
1337 | setTargetDAGCombine(ISD::VECTOR_SHUFFLE); | |||
1338 | ||||
1339 | ||||
1340 | if (Subtarget.useCRBits()) { | |||
1341 | setTargetDAGCombine(ISD::TRUNCATE); | |||
1342 | setTargetDAGCombine(ISD::SETCC); | |||
1343 | setTargetDAGCombine(ISD::SELECT_CC); | |||
1344 | } | |||
1345 | ||||
1346 | if (Subtarget.hasP9Altivec()) { | |||
1347 | setTargetDAGCombine(ISD::ABS); | |||
1348 | setTargetDAGCombine(ISD::VSELECT); | |||
1349 | } | |||
1350 | ||||
1351 | setLibcallName(RTLIB::LOG_F128, "logf128"); | |||
1352 | setLibcallName(RTLIB::LOG2_F128, "log2f128"); | |||
1353 | setLibcallName(RTLIB::LOG10_F128, "log10f128"); | |||
1354 | setLibcallName(RTLIB::EXP_F128, "expf128"); | |||
1355 | setLibcallName(RTLIB::EXP2_F128, "exp2f128"); | |||
1356 | setLibcallName(RTLIB::SIN_F128, "sinf128"); | |||
1357 | setLibcallName(RTLIB::COS_F128, "cosf128"); | |||
1358 | setLibcallName(RTLIB::POW_F128, "powf128"); | |||
1359 | setLibcallName(RTLIB::FMIN_F128, "fminf128"); | |||
1360 | setLibcallName(RTLIB::FMAX_F128, "fmaxf128"); | |||
1361 | setLibcallName(RTLIB::REM_F128, "fmodf128"); | |||
1362 | setLibcallName(RTLIB::SQRT_F128, "sqrtf128"); | |||
1363 | setLibcallName(RTLIB::CEIL_F128, "ceilf128"); | |||
1364 | setLibcallName(RTLIB::FLOOR_F128, "floorf128"); | |||
1365 | setLibcallName(RTLIB::TRUNC_F128, "truncf128"); | |||
1366 | setLibcallName(RTLIB::ROUND_F128, "roundf128"); | |||
1367 | setLibcallName(RTLIB::LROUND_F128, "lroundf128"); | |||
1368 | setLibcallName(RTLIB::LLROUND_F128, "llroundf128"); | |||
1369 | setLibcallName(RTLIB::RINT_F128, "rintf128"); | |||
1370 | setLibcallName(RTLIB::LRINT_F128, "lrintf128"); | |||
1371 | setLibcallName(RTLIB::LLRINT_F128, "llrintf128"); | |||
1372 | setLibcallName(RTLIB::NEARBYINT_F128, "nearbyintf128"); | |||
1373 | setLibcallName(RTLIB::FMA_F128, "fmaf128"); | |||
1374 | ||||
1375 | // With 32 condition bits, we don't need to sink (and duplicate) compares | |||
1376 | // aggressively in CodeGenPrep. | |||
1377 | if (Subtarget.useCRBits()) { | |||
1378 | setHasMultipleConditionRegisters(); | |||
1379 | setJumpIsExpensive(); | |||
1380 | } | |||
1381 | ||||
1382 | setMinFunctionAlignment(Align(4)); | |||
1383 | ||||
1384 | switch (Subtarget.getCPUDirective()) { | |||
1385 | default: break; | |||
1386 | case PPC::DIR_970: | |||
1387 | case PPC::DIR_A2: | |||
1388 | case PPC::DIR_E500: | |||
1389 | case PPC::DIR_E500mc: | |||
1390 | case PPC::DIR_E5500: | |||
1391 | case PPC::DIR_PWR4: | |||
1392 | case PPC::DIR_PWR5: | |||
1393 | case PPC::DIR_PWR5X: | |||
1394 | case PPC::DIR_PWR6: | |||
1395 | case PPC::DIR_PWR6X: | |||
1396 | case PPC::DIR_PWR7: | |||
1397 | case PPC::DIR_PWR8: | |||
1398 | case PPC::DIR_PWR9: | |||
1399 | case PPC::DIR_PWR10: | |||
1400 | case PPC::DIR_PWR_FUTURE: | |||
1401 | setPrefLoopAlignment(Align(16)); | |||
1402 | setPrefFunctionAlignment(Align(16)); | |||
1403 | break; | |||
1404 | } | |||
1405 | ||||
1406 | if (Subtarget.enableMachineScheduler()) | |||
1407 | setSchedulingPreference(Sched::Source); | |||
1408 | else | |||
1409 | setSchedulingPreference(Sched::Hybrid); | |||
1410 | ||||
1411 | computeRegisterProperties(STI.getRegisterInfo()); | |||
1412 | ||||
1413 | // The Freescale cores do better with aggressive inlining of memcpy and | |||
1414 | // friends. GCC uses same threshold of 128 bytes (= 32 word stores). | |||
1415 | if (Subtarget.getCPUDirective() == PPC::DIR_E500mc || | |||
1416 | Subtarget.getCPUDirective() == PPC::DIR_E5500) { | |||
1417 | MaxStoresPerMemset = 32; | |||
1418 | MaxStoresPerMemsetOptSize = 16; | |||
1419 | MaxStoresPerMemcpy = 32; | |||
1420 | MaxStoresPerMemcpyOptSize = 8; | |||
1421 | MaxStoresPerMemmove = 32; | |||
1422 | MaxStoresPerMemmoveOptSize = 8; | |||
1423 | } else if (Subtarget.getCPUDirective() == PPC::DIR_A2) { | |||
1424 | // The A2 also benefits from (very) aggressive inlining of memcpy and | |||
1425 | // friends. The overhead of a the function call, even when warm, can be | |||
1426 | // over one hundred cycles. | |||
1427 | MaxStoresPerMemset = 128; | |||
1428 | MaxStoresPerMemcpy = 128; | |||
1429 | MaxStoresPerMemmove = 128; | |||
1430 | MaxLoadsPerMemcmp = 128; | |||
1431 | } else { | |||
1432 | MaxLoadsPerMemcmp = 8; | |||
1433 | MaxLoadsPerMemcmpOptSize = 4; | |||
1434 | } | |||
1435 | ||||
1436 | IsStrictFPEnabled = true; | |||
1437 | ||||
1438 | // Let the subtarget (CPU) decide if a predictable select is more expensive | |||
1439 | // than the corresponding branch. This information is used in CGP to decide | |||
1440 | // when to convert selects into branches. | |||
1441 | PredictableSelectIsExpensive = Subtarget.isPredictableSelectIsExpensive(); | |||
1442 | } | |||
1443 | ||||
1444 | // *********************************** NOTE ************************************ | |||
1445 | // For selecting load and store instructions, the addressing modes are defined | |||
1446 | // as ComplexPatterns in PPCInstrInfo.td, which are then utilized in the TD | |||
1447 | // patterns to match the load the store instructions. | |||
1448 | // | |||
1449 | // The TD definitions for the addressing modes correspond to their respective | |||
1450 | // Select<AddrMode>Form() function in PPCISelDAGToDAG.cpp. These functions rely | |||
1451 | // on SelectOptimalAddrMode(), which calls computeMOFlags() to compute the | |||
1452 | // address mode flags of a particular node. Afterwards, the computed address | |||
1453 | // flags are passed into getAddrModeForFlags() in order to retrieve the optimal | |||
1454 | // addressing mode. SelectOptimalAddrMode() then sets the Base and Displacement | |||
1455 | // accordingly, based on the preferred addressing mode. | |||
1456 | // | |||
1457 | // Within PPCISelLowering.h, there are two enums: MemOpFlags and AddrMode. | |||
1458 | // MemOpFlags contains all the possible flags that can be used to compute the | |||
1459 | // optimal addressing mode for load and store instructions. | |||
1460 | // AddrMode contains all the possible load and store addressing modes available | |||
1461 | // on Power (such as DForm, DSForm, DQForm, XForm, etc.) | |||
1462 | // | |||
1463 | // When adding new load and store instructions, it is possible that new address | |||
1464 | // flags may need to be added into MemOpFlags, and a new addressing mode will | |||
1465 | // need to be added to AddrMode. An entry of the new addressing mode (consisting | |||
1466 | // of the minimal and main distinguishing address flags for the new load/store | |||
1467 | // instructions) will need to be added into initializeAddrModeMap() below. | |||
1468 | // Finally, when adding new addressing modes, the getAddrModeForFlags() will | |||
1469 | // need to be updated to account for selecting the optimal addressing mode. | |||
1470 | // ***************************************************************************** | |||
1471 | /// Initialize the map that relates the different addressing modes of the load | |||
1472 | /// and store instructions to a set of flags. This ensures the load/store | |||
1473 | /// instruction is correctly matched during instruction selection. | |||
1474 | void PPCTargetLowering::initializeAddrModeMap() { | |||
1475 | AddrModesMap[PPC::AM_DForm] = { | |||
1476 | // LWZ, STW | |||
1477 | PPC::MOF_ZExt | PPC::MOF_RPlusSImm16 | PPC::MOF_WordInt, | |||
1478 | PPC::MOF_ZExt | PPC::MOF_RPlusLo | PPC::MOF_WordInt, | |||
1479 | PPC::MOF_ZExt | PPC::MOF_NotAddNorCst | PPC::MOF_WordInt, | |||
1480 | PPC::MOF_ZExt | PPC::MOF_AddrIsSImm32 | PPC::MOF_WordInt, | |||
1481 | // LBZ, LHZ, STB, STH | |||
1482 | PPC::MOF_ZExt | PPC::MOF_RPlusSImm16 | PPC::MOF_SubWordInt, | |||
1483 | PPC::MOF_ZExt | PPC::MOF_RPlusLo | PPC::MOF_SubWordInt, | |||
1484 | PPC::MOF_ZExt | PPC::MOF_NotAddNorCst | PPC::MOF_SubWordInt, | |||
1485 | PPC::MOF_ZExt | PPC::MOF_AddrIsSImm32 | PPC::MOF_SubWordInt, | |||
1486 | // LHA | |||
1487 | PPC::MOF_SExt | PPC::MOF_RPlusSImm16 | PPC::MOF_SubWordInt, | |||
1488 | PPC::MOF_SExt | PPC::MOF_RPlusLo | PPC::MOF_SubWordInt, | |||
1489 | PPC::MOF_SExt | PPC::MOF_NotAddNorCst | PPC::MOF_SubWordInt, | |||
1490 | PPC::MOF_SExt | PPC::MOF_AddrIsSImm32 | PPC::MOF_SubWordInt, | |||
1491 | // LFS, LFD, STFS, STFD | |||
1492 | PPC::MOF_RPlusSImm16 | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetBeforeP9, | |||
1493 | PPC::MOF_RPlusLo | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetBeforeP9, | |||
1494 | PPC::MOF_NotAddNorCst | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetBeforeP9, | |||
1495 | PPC::MOF_AddrIsSImm32 | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetBeforeP9, | |||
1496 | }; | |||
1497 | AddrModesMap[PPC::AM_DSForm] = { | |||
1498 | // LWA | |||
1499 | PPC::MOF_SExt | PPC::MOF_RPlusSImm16Mult4 | PPC::MOF_WordInt, | |||
1500 | PPC::MOF_SExt | PPC::MOF_NotAddNorCst | PPC::MOF_WordInt, | |||
1501 | PPC::MOF_SExt | PPC::MOF_AddrIsSImm32 | PPC::MOF_WordInt, | |||
1502 | // LD, STD | |||
1503 | PPC::MOF_RPlusSImm16Mult4 | PPC::MOF_DoubleWordInt, | |||
1504 | PPC::MOF_NotAddNorCst | PPC::MOF_DoubleWordInt, | |||
1505 | PPC::MOF_AddrIsSImm32 | PPC::MOF_DoubleWordInt, | |||
1506 | // DFLOADf32, DFLOADf64, DSTOREf32, DSTOREf64 | |||
1507 | PPC::MOF_RPlusSImm16Mult4 | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetP9, | |||
1508 | PPC::MOF_NotAddNorCst | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetP9, | |||
1509 | PPC::MOF_AddrIsSImm32 | PPC::MOF_ScalarFloat | PPC::MOF_SubtargetP9, | |||
1510 | }; | |||
1511 | AddrModesMap[PPC::AM_DQForm] = { | |||
1512 | // LXV, STXV | |||
1513 | PPC::MOF_RPlusSImm16Mult16 | PPC::MOF_Vector | PPC::MOF_SubtargetP9, | |||
1514 | PPC::MOF_NotAddNorCst | PPC::MOF_Vector | PPC::MOF_SubtargetP9, | |||
1515 | PPC::MOF_AddrIsSImm32 | PPC::MOF_Vector | PPC::MOF_SubtargetP9, | |||
1516 | PPC::MOF_RPlusSImm16Mult16 | PPC::MOF_Vector256 | PPC::MOF_SubtargetP10, | |||
1517 | PPC::MOF_NotAddNorCst | PPC::MOF_Vector256 | PPC::MOF_SubtargetP10, | |||
1518 | PPC::MOF_AddrIsSImm32 | PPC::MOF_Vector256 | PPC::MOF_SubtargetP10, | |||
1519 | }; | |||
1520 | } | |||
1521 | ||||
1522 | /// getMaxByValAlign - Helper for getByValTypeAlignment to determine | |||
1523 | /// the desired ByVal argument alignment. | |||
1524 | static void getMaxByValAlign(Type *Ty, Align &MaxAlign, Align MaxMaxAlign) { | |||
1525 | if (MaxAlign == MaxMaxAlign) | |||
1526 | return; | |||
1527 | if (VectorType *VTy = dyn_cast<VectorType>(Ty)) { | |||
1528 | if (MaxMaxAlign >= 32 && | |||
1529 | VTy->getPrimitiveSizeInBits().getFixedSize() >= 256) | |||
1530 | MaxAlign = Align(32); | |||
1531 | else if (VTy->getPrimitiveSizeInBits().getFixedSize() >= 128 && | |||
1532 | MaxAlign < 16) | |||
1533 | MaxAlign = Align(16); | |||
1534 | } else if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { | |||
1535 | Align EltAlign; | |||
1536 | getMaxByValAlign(ATy->getElementType(), EltAlign, MaxMaxAlign); | |||
1537 | if (EltAlign > MaxAlign) | |||
1538 | MaxAlign = EltAlign; | |||
1539 | } else if (StructType *STy = dyn_cast<StructType>(Ty)) { | |||
1540 | for (auto *EltTy : STy->elements()) { | |||
1541 | Align EltAlign; | |||
1542 | getMaxByValAlign(EltTy, EltAlign, MaxMaxAlign); | |||
1543 | if (EltAlign > MaxAlign) | |||
1544 | MaxAlign = EltAlign; | |||
1545 | if (MaxAlign == MaxMaxAlign) | |||
1546 | break; | |||
1547 | } | |||
1548 | } | |||
1549 | } | |||
1550 | ||||
1551 | /// getByValTypeAlignment - Return the desired alignment for ByVal aggregate | |||
1552 | /// function arguments in the caller parameter area. | |||
1553 | unsigned PPCTargetLowering::getByValTypeAlignment(Type *Ty, | |||
1554 | const DataLayout &DL) const { | |||
1555 | // 16byte and wider vectors are passed on 16byte boundary. | |||
1556 | // The rest is 8 on PPC64 and 4 on PPC32 boundary. | |||
1557 | Align Alignment = Subtarget.isPPC64() ? Align(8) : Align(4); | |||
1558 | if (Subtarget.hasAltivec()) | |||
1559 | getMaxByValAlign(Ty, Alignment, Align(16)); | |||
1560 | return Alignment.value(); | |||
1561 | } | |||
1562 | ||||
1563 | bool PPCTargetLowering::useSoftFloat() const { | |||
1564 | return Subtarget.useSoftFloat(); | |||
1565 | } | |||
1566 | ||||
1567 | bool PPCTargetLowering::hasSPE() const { | |||
1568 | return Subtarget.hasSPE(); | |||
1569 | } | |||
1570 | ||||
1571 | bool PPCTargetLowering::preferIncOfAddToSubOfNot(EVT VT) const { | |||
1572 | return VT.isScalarInteger(); | |||
1573 | } | |||
1574 | ||||
1575 | const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const { | |||
1576 | switch ((PPCISD::NodeType)Opcode) { | |||
1577 | case PPCISD::FIRST_NUMBER: break; | |||
1578 | case PPCISD::FSEL: return "PPCISD::FSEL"; | |||
1579 | case PPCISD::XSMAXCDP: return "PPCISD::XSMAXCDP"; | |||
1580 | case PPCISD::XSMINCDP: return "PPCISD::XSMINCDP"; | |||
1581 | case PPCISD::FCFID: return "PPCISD::FCFID"; | |||
1582 | case PPCISD::FCFIDU: return "PPCISD::FCFIDU"; | |||
1583 | case PPCISD::FCFIDS: return "PPCISD::FCFIDS"; | |||
1584 | case PPCISD::FCFIDUS: return "PPCISD::FCFIDUS"; | |||
1585 | case PPCISD::FCTIDZ: return "PPCISD::FCTIDZ"; | |||
1586 | case PPCISD::FCTIWZ: return "PPCISD::FCTIWZ"; | |||
1587 | case PPCISD::FCTIDUZ: return "PPCISD::FCTIDUZ"; | |||
1588 | case PPCISD::FCTIWUZ: return "PPCISD::FCTIWUZ"; | |||
1589 | case PPCISD::FP_TO_UINT_IN_VSR: | |||
1590 | return "PPCISD::FP_TO_UINT_IN_VSR,"; | |||
1591 | case PPCISD::FP_TO_SINT_IN_VSR: | |||
1592 | return "PPCISD::FP_TO_SINT_IN_VSR"; | |||
1593 | case PPCISD::FRE: return "PPCISD::FRE"; | |||
1594 | case PPCISD::FRSQRTE: return "PPCISD::FRSQRTE"; | |||
1595 | case PPCISD::FTSQRT: | |||
1596 | return "PPCISD::FTSQRT"; | |||
1597 | case PPCISD::FSQRT: | |||
1598 | return "PPCISD::FSQRT"; | |||
1599 | case PPCISD::STFIWX: return "PPCISD::STFIWX"; | |||
1600 | case PPCISD::VPERM: return "PPCISD::VPERM"; | |||
1601 | case PPCISD::XXSPLT: return "PPCISD::XXSPLT"; | |||
1602 | case PPCISD::XXSPLTI_SP_TO_DP: | |||
1603 | return "PPCISD::XXSPLTI_SP_TO_DP"; | |||
1604 | case PPCISD::XXSPLTI32DX: | |||
1605 | return "PPCISD::XXSPLTI32DX"; | |||
1606 | case PPCISD::VECINSERT: return "PPCISD::VECINSERT"; | |||
1607 | case PPCISD::XXPERMDI: return "PPCISD::XXPERMDI"; | |||
1608 | case PPCISD::VECSHL: return "PPCISD::VECSHL"; | |||
1609 | case PPCISD::CMPB: return "PPCISD::CMPB"; | |||
1610 | case PPCISD::Hi: return "PPCISD::Hi"; | |||
1611 | case PPCISD::Lo: return "PPCISD::Lo"; | |||
1612 | case PPCISD::TOC_ENTRY: return "PPCISD::TOC_ENTRY"; | |||
1613 | case PPCISD::ATOMIC_CMP_SWAP_8: return "PPCISD::ATOMIC_CMP_SWAP_8"; | |||
1614 | case PPCISD::ATOMIC_CMP_SWAP_16: return "PPCISD::ATOMIC_CMP_SWAP_16"; | |||
1615 | case PPCISD::DYNALLOC: return "PPCISD::DYNALLOC"; | |||
1616 | case PPCISD::DYNAREAOFFSET: return "PPCISD::DYNAREAOFFSET"; | |||
1617 | case PPCISD::PROBED_ALLOCA: return "PPCISD::PROBED_ALLOCA"; | |||
1618 | case PPCISD::GlobalBaseReg: return "PPCISD::GlobalBaseReg"; | |||
1619 | case PPCISD::SRL: return "PPCISD::SRL"; | |||
1620 | case PPCISD::SRA: return "PPCISD::SRA"; | |||
1621 | case PPCISD::SHL: return "PPCISD::SHL"; | |||
1622 | case PPCISD::SRA_ADDZE: return "PPCISD::SRA_ADDZE"; | |||
1623 | case PPCISD::CALL: return "PPCISD::CALL"; | |||
1624 | case PPCISD::CALL_NOP: return "PPCISD::CALL_NOP"; | |||
1625 | case PPCISD::CALL_NOTOC: return "PPCISD::CALL_NOTOC"; | |||
1626 | case PPCISD::MTCTR: return "PPCISD::MTCTR"; | |||
1627 | case PPCISD::BCTRL: return "PPCISD::BCTRL"; | |||
1628 | case PPCISD::BCTRL_LOAD_TOC: return "PPCISD::BCTRL_LOAD_TOC"; | |||
1629 | case PPCISD::RET_FLAG: return "PPCISD::RET_FLAG"; | |||
1630 | case PPCISD::READ_TIME_BASE: return "PPCISD::READ_TIME_BASE"; | |||
1631 | case PPCISD::EH_SJLJ_SETJMP: return "PPCISD::EH_SJLJ_SETJMP"; | |||
1632 | case PPCISD::EH_SJLJ_LONGJMP: return "PPCISD::EH_SJLJ_LONGJMP"; | |||
1633 | case PPCISD::MFOCRF: return "PPCISD::MFOCRF"; | |||
1634 | case PPCISD::MFVSR: return "PPCISD::MFVSR"; | |||
1635 | case PPCISD::MTVSRA: return "PPCISD::MTVSRA"; | |||
1636 | case PPCISD::MTVSRZ: return "PPCISD::MTVSRZ"; | |||
1637 | case PPCISD::SINT_VEC_TO_FP: return "PPCISD::SINT_VEC_TO_FP"; | |||
1638 | case PPCISD::UINT_VEC_TO_FP: return "PPCISD::UINT_VEC_TO_FP"; | |||
1639 | case PPCISD::SCALAR_TO_VECTOR_PERMUTED: | |||
1640 | return "PPCISD::SCALAR_TO_VECTOR_PERMUTED"; | |||
1641 | case PPCISD::ANDI_rec_1_EQ_BIT: | |||
1642 | return "PPCISD::ANDI_rec_1_EQ_BIT"; | |||
1643 | case PPCISD::ANDI_rec_1_GT_BIT: | |||
1644 | return "PPCISD::ANDI_rec_1_GT_BIT"; | |||
1645 | case PPCISD::VCMP: return "PPCISD::VCMP"; | |||
1646 | case PPCISD::VCMP_rec: return "PPCISD::VCMP_rec"; | |||
1647 | case PPCISD::LBRX: return "PPCISD::LBRX"; | |||
1648 | case PPCISD::STBRX: return "PPCISD::STBRX"; | |||
1649 | case PPCISD::LFIWAX: return "PPCISD::LFIWAX"; | |||
1650 | case PPCISD::LFIWZX: return "PPCISD::LFIWZX"; | |||
1651 | case PPCISD::LXSIZX: return "PPCISD::LXSIZX"; | |||
1652 | case PPCISD::STXSIX: return "PPCISD::STXSIX"; | |||
1653 | case PPCISD::VEXTS: return "PPCISD::VEXTS"; | |||
1654 | case PPCISD::LXVD2X: return "PPCISD::LXVD2X"; | |||
1655 | case PPCISD::STXVD2X: return "PPCISD::STXVD2X"; | |||
1656 | case PPCISD::LOAD_VEC_BE: return "PPCISD::LOAD_VEC_BE"; | |||
1657 | case PPCISD::STORE_VEC_BE: return "PPCISD::STORE_VEC_BE"; | |||
1658 | case PPCISD::ST_VSR_SCAL_INT: | |||
1659 | return "PPCISD::ST_VSR_SCAL_INT"; | |||
1660 | case PPCISD::COND_BRANCH: return "PPCISD::COND_BRANCH"; | |||
1661 | case PPCISD::BDNZ: return "PPCISD::BDNZ"; | |||
1662 | case PPCISD::BDZ: return "PPCISD::BDZ"; | |||
1663 | case PPCISD::MFFS: return "PPCISD::MFFS"; | |||
1664 | case PPCISD::FADDRTZ: return "PPCISD::FADDRTZ"; | |||
1665 | case PPCISD::TC_RETURN: return "PPCISD::TC_RETURN"; | |||
1666 | case PPCISD::CR6SET: return "PPCISD::CR6SET"; | |||
1667 | case PPCISD::CR6UNSET: return "PPCISD::CR6UNSET"; | |||
1668 | case PPCISD::PPC32_GOT: return "PPCISD::PPC32_GOT"; | |||
1669 | case PPCISD::PPC32_PICGOT: return "PPCISD::PPC32_PICGOT"; | |||
1670 | case PPCISD::ADDIS_GOT_TPREL_HA: return "PPCISD::ADDIS_GOT_TPREL_HA"; | |||
1671 | case PPCISD::LD_GOT_TPREL_L: return "PPCISD::LD_GOT_TPREL_L"; | |||
1672 | case PPCISD::ADD_TLS: return "PPCISD::ADD_TLS"; | |||
1673 | case PPCISD::ADDIS_TLSGD_HA: return "PPCISD::ADDIS_TLSGD_HA"; | |||
1674 | case PPCISD::ADDI_TLSGD_L: return "PPCISD::ADDI_TLSGD_L"; | |||
1675 | case PPCISD::GET_TLS_ADDR: return "PPCISD::GET_TLS_ADDR"; | |||
1676 | case PPCISD::ADDI_TLSGD_L_ADDR: return "PPCISD::ADDI_TLSGD_L_ADDR"; | |||
1677 | case PPCISD::TLSGD_AIX: return "PPCISD::TLSGD_AIX"; | |||
1678 | case PPCISD::ADDIS_TLSLD_HA: return "PPCISD::ADDIS_TLSLD_HA"; | |||
1679 | case PPCISD::ADDI_TLSLD_L: return "PPCISD::ADDI_TLSLD_L"; | |||
1680 | case PPCISD::GET_TLSLD_ADDR: return "PPCISD::GET_TLSLD_ADDR"; | |||
1681 | case PPCISD::ADDI_TLSLD_L_ADDR: return "PPCISD::ADDI_TLSLD_L_ADDR"; | |||
1682 | case PPCISD::ADDIS_DTPREL_HA: return "PPCISD::ADDIS_DTPREL_HA"; | |||
1683 | case PPCISD::ADDI_DTPREL_L: return "PPCISD::ADDI_DTPREL_L"; | |||
1684 | case PPCISD::PADDI_DTPREL: | |||
1685 | return "PPCISD::PADDI_DTPREL"; | |||
1686 | case PPCISD::VADD_SPLAT: return "PPCISD::VADD_SPLAT"; | |||
1687 | case PPCISD::SC: return "PPCISD::SC"; | |||
1688 | case PPCISD::CLRBHRB: return "PPCISD::CLRBHRB"; | |||
1689 | case PPCISD::MFBHRBE: return "PPCISD::MFBHRBE"; | |||
1690 | case PPCISD::RFEBB: return "PPCISD::RFEBB"; | |||
1691 | case PPCISD::XXSWAPD: return "PPCISD::XXSWAPD"; | |||
1692 | case PPCISD::SWAP_NO_CHAIN: return "PPCISD::SWAP_NO_CHAIN"; | |||
1693 | case PPCISD::VABSD: return "PPCISD::VABSD"; | |||
1694 | case PPCISD::BUILD_FP128: return "PPCISD::BUILD_FP128"; | |||
1695 | case PPCISD::BUILD_SPE64: return "PPCISD::BUILD_SPE64"; | |||
1696 | case PPCISD::EXTRACT_SPE: return "PPCISD::EXTRACT_SPE"; | |||
1697 | case PPCISD::EXTSWSLI: return "PPCISD::EXTSWSLI"; | |||
1698 | case PPCISD::LD_VSX_LH: return "PPCISD::LD_VSX_LH"; | |||
1699 | case PPCISD::FP_EXTEND_HALF: return "PPCISD::FP_EXTEND_HALF"; | |||
1700 | case PPCISD::MAT_PCREL_ADDR: return "PPCISD::MAT_PCREL_ADDR"; | |||
1701 | case PPCISD::TLS_DYNAMIC_MAT_PCREL_ADDR: | |||
1702 | return "PPCISD::TLS_DYNAMIC_MAT_PCREL_ADDR"; | |||
1703 | case PPCISD::TLS_LOCAL_EXEC_MAT_ADDR: | |||
1704 | return "PPCISD::TLS_LOCAL_EXEC_MAT_ADDR"; | |||
1705 | case PPCISD::ACC_BUILD: return "PPCISD::ACC_BUILD"; | |||
1706 | case PPCISD::PAIR_BUILD: return "PPCISD::PAIR_BUILD"; | |||
1707 | case PPCISD::EXTRACT_VSX_REG: return "PPCISD::EXTRACT_VSX_REG"; | |||
1708 | case PPCISD::XXMFACC: return "PPCISD::XXMFACC"; | |||
1709 | case PPCISD::LD_SPLAT: return "PPCISD::LD_SPLAT"; | |||
1710 | case PPCISD::FNMSUB: return "PPCISD::FNMSUB"; | |||
1711 | case PPCISD::STRICT_FADDRTZ: | |||
1712 | return "PPCISD::STRICT_FADDRTZ"; | |||
1713 | case PPCISD::STRICT_FCTIDZ: | |||
1714 | return "PPCISD::STRICT_FCTIDZ"; | |||
1715 | case PPCISD::STRICT_FCTIWZ: | |||
1716 | return "PPCISD::STRICT_FCTIWZ"; | |||
1717 | case PPCISD::STRICT_FCTIDUZ: | |||
1718 | return "PPCISD::STRICT_FCTIDUZ"; | |||
1719 | case PPCISD::STRICT_FCTIWUZ: | |||
1720 | return "PPCISD::STRICT_FCTIWUZ"; | |||
1721 | case PPCISD::STRICT_FCFID: | |||
1722 | return "PPCISD::STRICT_FCFID"; | |||
1723 | case PPCISD::STRICT_FCFIDU: | |||
1724 | return "PPCISD::STRICT_FCFIDU"; | |||
1725 | case PPCISD::STRICT_FCFIDS: | |||
1726 | return "PPCISD::STRICT_FCFIDS"; | |||
1727 | case PPCISD::STRICT_FCFIDUS: | |||
1728 | return "PPCISD::STRICT_FCFIDUS"; | |||
1729 | case PPCISD::LXVRZX: return "PPCISD::LXVRZX"; | |||
1730 | } | |||
1731 | return nullptr; | |||
1732 | } | |||
1733 | ||||
1734 | EVT PPCTargetLowering::getSetCCResultType(const DataLayout &DL, LLVMContext &C, | |||
1735 | EVT VT) const { | |||
1736 | if (!VT.isVector()) | |||
1737 | return Subtarget.useCRBits() ? MVT::i1 : MVT::i32; | |||
1738 | ||||
1739 | return VT.changeVectorElementTypeToInteger(); | |||
1740 | } | |||
1741 | ||||
1742 | bool PPCTargetLowering::enableAggressiveFMAFusion(EVT VT) const { | |||
1743 | assert(VT.isFloatingPoint() && "Non-floating-point FMA?")(static_cast <bool> (VT.isFloatingPoint() && "Non-floating-point FMA?" ) ? void (0) : __assert_fail ("VT.isFloatingPoint() && \"Non-floating-point FMA?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 1743, __extension__ __PRETTY_FUNCTION__)); | |||
1744 | return true; | |||
1745 | } | |||
1746 | ||||
1747 | //===----------------------------------------------------------------------===// | |||
1748 | // Node matching predicates, for use by the tblgen matching code. | |||
1749 | //===----------------------------------------------------------------------===// | |||
1750 | ||||
1751 | /// isFloatingPointZero - Return true if this is 0.0 or -0.0. | |||
1752 | static bool isFloatingPointZero(SDValue Op) { | |||
1753 | if (ConstantFPSDNode *CFP = dyn_cast<ConstantFPSDNode>(Op)) | |||
1754 | return CFP->getValueAPF().isZero(); | |||
1755 | else if (ISD::isEXTLoad(Op.getNode()) || ISD::isNON_EXTLoad(Op.getNode())) { | |||
1756 | // Maybe this has already been legalized into the constant pool? | |||
1757 | if (ConstantPoolSDNode *CP = dyn_cast<ConstantPoolSDNode>(Op.getOperand(1))) | |||
1758 | if (const ConstantFP *CFP = dyn_cast<ConstantFP>(CP->getConstVal())) | |||
1759 | return CFP->getValueAPF().isZero(); | |||
1760 | } | |||
1761 | return false; | |||
1762 | } | |||
1763 | ||||
1764 | /// isConstantOrUndef - Op is either an undef node or a ConstantSDNode. Return | |||
1765 | /// true if Op is undef or if it matches the specified value. | |||
1766 | static bool isConstantOrUndef(int Op, int Val) { | |||
1767 | return Op < 0 || Op == Val; | |||
1768 | } | |||
1769 | ||||
1770 | /// isVPKUHUMShuffleMask - Return true if this is the shuffle mask for a | |||
1771 | /// VPKUHUM instruction. | |||
1772 | /// The ShuffleKind distinguishes between big-endian operations with | |||
1773 | /// two different inputs (0), either-endian operations with two identical | |||
1774 | /// inputs (1), and little-endian operations with two different inputs (2). | |||
1775 | /// For the latter, the input operands are swapped (see PPCInstrAltivec.td). | |||
1776 | bool PPC::isVPKUHUMShuffleMask(ShuffleVectorSDNode *N, unsigned ShuffleKind, | |||
1777 | SelectionDAG &DAG) { | |||
1778 | bool IsLE = DAG.getDataLayout().isLittleEndian(); | |||
1779 | if (ShuffleKind == 0) { | |||
1780 | if (IsLE) | |||
1781 | return false; | |||
1782 | for (unsigned i = 0; i != 16; ++i) | |||
1783 | if (!isConstantOrUndef(N->getMaskElt(i), i*2+1)) | |||
1784 | return false; | |||
1785 | } else if (ShuffleKind == 2) { | |||
1786 | if (!IsLE) | |||
1787 | return false; | |||
1788 | for (unsigned i = 0; i != 16; ++i) | |||
1789 | if (!isConstantOrUndef(N->getMaskElt(i), i*2)) | |||
1790 | return false; | |||
1791 | } else if (ShuffleKind == 1) { | |||
1792 | unsigned j = IsLE ? 0 : 1; | |||
1793 | for (unsigned i = 0; i != 8; ++i) | |||
1794 | if (!isConstantOrUndef(N->getMaskElt(i), i*2+j) || | |||
1795 | !isConstantOrUndef(N->getMaskElt(i+8), i*2+j)) | |||
1796 | return false; | |||
1797 | } | |||
1798 | return true; | |||
1799 | } | |||
1800 | ||||
1801 | /// isVPKUWUMShuffleMask - Return true if this is the shuffle mask for a | |||
1802 | /// VPKUWUM instruction. | |||
1803 | /// The ShuffleKind distinguishes between big-endian operations with | |||
1804 | /// two different inputs (0), either-endian operations with two identical | |||
1805 | /// inputs (1), and little-endian operations with two different inputs (2). | |||
1806 | /// For the latter, the input operands are swapped (see PPCInstrAltivec.td). | |||
1807 | bool PPC::isVPKUWUMShuffleMask(ShuffleVectorSDNode *N, unsigned ShuffleKind, | |||
1808 | SelectionDAG &DAG) { | |||
1809 | bool IsLE = DAG.getDataLayout().isLittleEndian(); | |||
1810 | if (ShuffleKind == 0) { | |||
1811 | if (IsLE) | |||
1812 | return false; | |||
1813 | for (unsigned i = 0; i != 16; i += 2) | |||
1814 | if (!isConstantOrUndef(N->getMaskElt(i ), i*2+2) || | |||
1815 | !isConstantOrUndef(N->getMaskElt(i+1), i*2+3)) | |||
1816 | return false; | |||
1817 | } else if (ShuffleKind == 2) { | |||
1818 | if (!IsLE) | |||
1819 | return false; | |||
1820 | for (unsigned i = 0; i != 16; i += 2) | |||
1821 | if (!isConstantOrUndef(N->getMaskElt(i ), i*2) || | |||
1822 | !isConstantOrUndef(N->getMaskElt(i+1), i*2+1)) | |||
1823 | return false; | |||
1824 | } else if (ShuffleKind == 1) { | |||
1825 | unsigned j = IsLE ? 0 : 2; | |||
1826 | for (unsigned i = 0; i != 8; i += 2) | |||
1827 | if (!isConstantOrUndef(N->getMaskElt(i ), i*2+j) || | |||
1828 | !isConstantOrUndef(N->getMaskElt(i+1), i*2+j+1) || | |||
1829 | !isConstantOrUndef(N->getMaskElt(i+8), i*2+j) || | |||
1830 | !isConstantOrUndef(N->getMaskElt(i+9), i*2+j+1)) | |||
1831 | return false; | |||
1832 | } | |||
1833 | return true; | |||
1834 | } | |||
1835 | ||||
1836 | /// isVPKUDUMShuffleMask - Return true if this is the shuffle mask for a | |||
1837 | /// VPKUDUM instruction, AND the VPKUDUM instruction exists for the | |||
1838 | /// current subtarget. | |||
1839 | /// | |||
1840 | /// The ShuffleKind distinguishes between big-endian operations with | |||
1841 | /// two different inputs (0), either-endian operations with two identical | |||
1842 | /// inputs (1), and little-endian operations with two different inputs (2). | |||
1843 | /// For the latter, the input operands are swapped (see PPCInstrAltivec.td). | |||
1844 | bool PPC::isVPKUDUMShuffleMask(ShuffleVectorSDNode *N, unsigned ShuffleKind, | |||
1845 | SelectionDAG &DAG) { | |||
1846 | const PPCSubtarget& Subtarget = | |||
1847 | static_cast<const PPCSubtarget&>(DAG.getSubtarget()); | |||
1848 | if (!Subtarget.hasP8Vector()) | |||
1849 | return false; | |||
1850 | ||||
1851 | bool IsLE = DAG.getDataLayout().isLittleEndian(); | |||
1852 | if (ShuffleKind == 0) { | |||
1853 | if (IsLE) | |||
1854 | return false; | |||
1855 | for (unsigned i = 0; i != 16; i += 4) | |||
1856 | if (!isConstantOrUndef(N->getMaskElt(i ), i*2+4) || | |||
1857 | !isConstantOrUndef(N->getMaskElt(i+1), i*2+5) || | |||
1858 | !isConstantOrUndef(N->getMaskElt(i+2), i*2+6) || | |||
1859 | !isConstantOrUndef(N->getMaskElt(i+3), i*2+7)) | |||
1860 | return false; | |||
1861 | } else if (ShuffleKind == 2) { | |||
1862 | if (!IsLE) | |||
1863 | return false; | |||
1864 | for (unsigned i = 0; i != 16; i += 4) | |||
1865 | if (!isConstantOrUndef(N->getMaskElt(i ), i*2) || | |||
1866 | !isConstantOrUndef(N->getMaskElt(i+1), i*2+1) || | |||
1867 | !isConstantOrUndef(N->getMaskElt(i+2), i*2+2) || | |||
1868 | !isConstantOrUndef(N->getMaskElt(i+3), i*2+3)) | |||
1869 | return false; | |||
1870 | } else if (ShuffleKind == 1) { | |||
1871 | unsigned j = IsLE ? 0 : 4; | |||
1872 | for (unsigned i = 0; i != 8; i += 4) | |||
1873 | if (!isConstantOrUndef(N->getMaskElt(i ), i*2+j) || | |||
1874 | !isConstantOrUndef(N->getMaskElt(i+1), i*2+j+1) || | |||
1875 | !isConstantOrUndef(N->getMaskElt(i+2), i*2+j+2) || | |||
1876 | !isConstantOrUndef(N->getMaskElt(i+3), i*2+j+3) || | |||
1877 | !isConstantOrUndef(N->getMaskElt(i+8), i*2+j) || | |||
1878 | !isConstantOrUndef(N->getMaskElt(i+9), i*2+j+1) || | |||
1879 | !isConstantOrUndef(N->getMaskElt(i+10), i*2+j+2) || | |||
1880 | !isConstantOrUndef(N->getMaskElt(i+11), i*2+j+3)) | |||
1881 | return false; | |||
1882 | } | |||
1883 | return true; | |||
1884 | } | |||
1885 | ||||
1886 | /// isVMerge - Common function, used to match vmrg* shuffles. | |||
1887 | /// | |||
1888 | static bool isVMerge(ShuffleVectorSDNode *N, unsigned UnitSize, | |||
1889 | unsigned LHSStart, unsigned RHSStart) { | |||
1890 | if (N->getValueType(0) != MVT::v16i8) | |||
1891 | return false; | |||
1892 | assert((UnitSize == 1 || UnitSize == 2 || UnitSize == 4) &&(static_cast <bool> ((UnitSize == 1 || UnitSize == 2 || UnitSize == 4) && "Unsupported merge size!") ? void ( 0) : __assert_fail ("(UnitSize == 1 || UnitSize == 2 || UnitSize == 4) && \"Unsupported merge size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 1893, __extension__ __PRETTY_FUNCTION__)) | |||
1893 | "Unsupported merge size!")(static_cast <bool> ((UnitSize == 1 || UnitSize == 2 || UnitSize == 4) && "Unsupported merge size!") ? void ( 0) : __assert_fail ("(UnitSize == 1 || UnitSize == 2 || UnitSize == 4) && \"Unsupported merge size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 1893, __extension__ __PRETTY_FUNCTION__)); | |||
1894 | ||||
1895 | for (unsigned i = 0; i != 8/UnitSize; ++i) // Step over units | |||
1896 | for (unsigned j = 0; j != UnitSize; ++j) { // Step over bytes within unit | |||
1897 | if (!isConstantOrUndef(N->getMaskElt(i*UnitSize*2+j), | |||
1898 | LHSStart+j+i*UnitSize) || | |||
1899 | !isConstantOrUndef(N->getMaskElt(i*UnitSize*2+UnitSize+j), | |||
1900 | RHSStart+j+i*UnitSize)) | |||
1901 | return false; | |||
1902 | } | |||
1903 | return true; | |||
1904 | } | |||
1905 | ||||
1906 | /// isVMRGLShuffleMask - Return true if this is a shuffle mask suitable for | |||
1907 | /// a VMRGL* instruction with the specified unit size (1,2 or 4 bytes). | |||
1908 | /// The ShuffleKind distinguishes between big-endian merges with two | |||
1909 | /// different inputs (0), either-endian merges with two identical inputs (1), | |||
1910 | /// and little-endian merges with two different inputs (2). For the latter, | |||
1911 | /// the input operands are swapped (see PPCInstrAltivec.td). | |||
1912 | bool PPC::isVMRGLShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize, | |||
1913 | unsigned ShuffleKind, SelectionDAG &DAG) { | |||
1914 | if (DAG.getDataLayout().isLittleEndian()) { | |||
1915 | if (ShuffleKind == 1) // unary | |||
1916 | return isVMerge(N, UnitSize, 0, 0); | |||
1917 | else if (ShuffleKind == 2) // swapped | |||
1918 | return isVMerge(N, UnitSize, 0, 16); | |||
1919 | else | |||
1920 | return false; | |||
1921 | } else { | |||
1922 | if (ShuffleKind == 1) // unary | |||
1923 | return isVMerge(N, UnitSize, 8, 8); | |||
1924 | else if (ShuffleKind == 0) // normal | |||
1925 | return isVMerge(N, UnitSize, 8, 24); | |||
1926 | else | |||
1927 | return false; | |||
1928 | } | |||
1929 | } | |||
1930 | ||||
1931 | /// isVMRGHShuffleMask - Return true if this is a shuffle mask suitable for | |||
1932 | /// a VMRGH* instruction with the specified unit size (1,2 or 4 bytes). | |||
1933 | /// The ShuffleKind distinguishes between big-endian merges with two | |||
1934 | /// different inputs (0), either-endian merges with two identical inputs (1), | |||
1935 | /// and little-endian merges with two different inputs (2). For the latter, | |||
1936 | /// the input operands are swapped (see PPCInstrAltivec.td). | |||
1937 | bool PPC::isVMRGHShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize, | |||
1938 | unsigned ShuffleKind, SelectionDAG &DAG) { | |||
1939 | if (DAG.getDataLayout().isLittleEndian()) { | |||
1940 | if (ShuffleKind == 1) // unary | |||
1941 | return isVMerge(N, UnitSize, 8, 8); | |||
1942 | else if (ShuffleKind == 2) // swapped | |||
1943 | return isVMerge(N, UnitSize, 8, 24); | |||
1944 | else | |||
1945 | return false; | |||
1946 | } else { | |||
1947 | if (ShuffleKind == 1) // unary | |||
1948 | return isVMerge(N, UnitSize, 0, 0); | |||
1949 | else if (ShuffleKind == 0) // normal | |||
1950 | return isVMerge(N, UnitSize, 0, 16); | |||
1951 | else | |||
1952 | return false; | |||
1953 | } | |||
1954 | } | |||
1955 | ||||
1956 | /** | |||
1957 | * Common function used to match vmrgew and vmrgow shuffles | |||
1958 | * | |||
1959 | * The indexOffset determines whether to look for even or odd words in | |||
1960 | * the shuffle mask. This is based on the of the endianness of the target | |||
1961 | * machine. | |||
1962 | * - Little Endian: | |||
1963 | * - Use offset of 0 to check for odd elements | |||
1964 | * - Use offset of 4 to check for even elements | |||
1965 | * - Big Endian: | |||
1966 | * - Use offset of 0 to check for even elements | |||
1967 | * - Use offset of 4 to check for odd elements | |||
1968 | * A detailed description of the vector element ordering for little endian and | |||
1969 | * big endian can be found at | |||
1970 | * http://www.ibm.com/developerworks/library/l-ibm-xl-c-cpp-compiler/index.html | |||
1971 | * Targeting your applications - what little endian and big endian IBM XL C/C++ | |||
1972 | * compiler differences mean to you | |||
1973 | * | |||
1974 | * The mask to the shuffle vector instruction specifies the indices of the | |||
1975 | * elements from the two input vectors to place in the result. The elements are | |||
1976 | * numbered in array-access order, starting with the first vector. These vectors | |||
1977 | * are always of type v16i8, thus each vector will contain 16 elements of size | |||
1978 | * 8. More info on the shuffle vector can be found in the | |||
1979 | * http://llvm.org/docs/LangRef.html#shufflevector-instruction | |||
1980 | * Language Reference. | |||
1981 | * | |||
1982 | * The RHSStartValue indicates whether the same input vectors are used (unary) | |||
1983 | * or two different input vectors are used, based on the following: | |||
1984 | * - If the instruction uses the same vector for both inputs, the range of the | |||
1985 | * indices will be 0 to 15. In this case, the RHSStart value passed should | |||
1986 | * be 0. | |||
1987 | * - If the instruction has two different vectors then the range of the | |||
1988 | * indices will be 0 to 31. In this case, the RHSStart value passed should | |||
1989 | * be 16 (indices 0-15 specify elements in the first vector while indices 16 | |||
1990 | * to 31 specify elements in the second vector). | |||
1991 | * | |||
1992 | * \param[in] N The shuffle vector SD Node to analyze | |||
1993 | * \param[in] IndexOffset Specifies whether to look for even or odd elements | |||
1994 | * \param[in] RHSStartValue Specifies the starting index for the righthand input | |||
1995 | * vector to the shuffle_vector instruction | |||
1996 | * \return true iff this shuffle vector represents an even or odd word merge | |||
1997 | */ | |||
1998 | static bool isVMerge(ShuffleVectorSDNode *N, unsigned IndexOffset, | |||
1999 | unsigned RHSStartValue) { | |||
2000 | if (N->getValueType(0) != MVT::v16i8) | |||
2001 | return false; | |||
2002 | ||||
2003 | for (unsigned i = 0; i < 2; ++i) | |||
2004 | for (unsigned j = 0; j < 4; ++j) | |||
2005 | if (!isConstantOrUndef(N->getMaskElt(i*4+j), | |||
2006 | i*RHSStartValue+j+IndexOffset) || | |||
2007 | !isConstantOrUndef(N->getMaskElt(i*4+j+8), | |||
2008 | i*RHSStartValue+j+IndexOffset+8)) | |||
2009 | return false; | |||
2010 | return true; | |||
2011 | } | |||
2012 | ||||
2013 | /** | |||
2014 | * Determine if the specified shuffle mask is suitable for the vmrgew or | |||
2015 | * vmrgow instructions. | |||
2016 | * | |||
2017 | * \param[in] N The shuffle vector SD Node to analyze | |||
2018 | * \param[in] CheckEven Check for an even merge (true) or an odd merge (false) | |||
2019 | * \param[in] ShuffleKind Identify the type of merge: | |||
2020 | * - 0 = big-endian merge with two different inputs; | |||
2021 | * - 1 = either-endian merge with two identical inputs; | |||
2022 | * - 2 = little-endian merge with two different inputs (inputs are swapped for | |||
2023 | * little-endian merges). | |||
2024 | * \param[in] DAG The current SelectionDAG | |||
2025 | * \return true iff this shuffle mask | |||
2026 | */ | |||
2027 | bool PPC::isVMRGEOShuffleMask(ShuffleVectorSDNode *N, bool CheckEven, | |||
2028 | unsigned ShuffleKind, SelectionDAG &DAG) { | |||
2029 | if (DAG.getDataLayout().isLittleEndian()) { | |||
2030 | unsigned indexOffset = CheckEven ? 4 : 0; | |||
2031 | if (ShuffleKind == 1) // Unary | |||
2032 | return isVMerge(N, indexOffset, 0); | |||
2033 | else if (ShuffleKind == 2) // swapped | |||
2034 | return isVMerge(N, indexOffset, 16); | |||
2035 | else | |||
2036 | return false; | |||
2037 | } | |||
2038 | else { | |||
2039 | unsigned indexOffset = CheckEven ? 0 : 4; | |||
2040 | if (ShuffleKind == 1) // Unary | |||
2041 | return isVMerge(N, indexOffset, 0); | |||
2042 | else if (ShuffleKind == 0) // Normal | |||
2043 | return isVMerge(N, indexOffset, 16); | |||
2044 | else | |||
2045 | return false; | |||
2046 | } | |||
2047 | return false; | |||
2048 | } | |||
2049 | ||||
2050 | /// isVSLDOIShuffleMask - If this is a vsldoi shuffle mask, return the shift | |||
2051 | /// amount, otherwise return -1. | |||
2052 | /// The ShuffleKind distinguishes between big-endian operations with two | |||
2053 | /// different inputs (0), either-endian operations with two identical inputs | |||
2054 | /// (1), and little-endian operations with two different inputs (2). For the | |||
2055 | /// latter, the input operands are swapped (see PPCInstrAltivec.td). | |||
2056 | int PPC::isVSLDOIShuffleMask(SDNode *N, unsigned ShuffleKind, | |||
2057 | SelectionDAG &DAG) { | |||
2058 | if (N->getValueType(0) != MVT::v16i8) | |||
2059 | return -1; | |||
2060 | ||||
2061 | ShuffleVectorSDNode *SVOp = cast<ShuffleVectorSDNode>(N); | |||
2062 | ||||
2063 | // Find the first non-undef value in the shuffle mask. | |||
2064 | unsigned i; | |||
2065 | for (i = 0; i != 16 && SVOp->getMaskElt(i) < 0; ++i) | |||
2066 | /*search*/; | |||
2067 | ||||
2068 | if (i == 16) return -1; // all undef. | |||
2069 | ||||
2070 | // Otherwise, check to see if the rest of the elements are consecutively | |||
2071 | // numbered from this value. | |||
2072 | unsigned ShiftAmt = SVOp->getMaskElt(i); | |||
2073 | if (ShiftAmt < i) return -1; | |||
2074 | ||||
2075 | ShiftAmt -= i; | |||
2076 | bool isLE = DAG.getDataLayout().isLittleEndian(); | |||
2077 | ||||
2078 | if ((ShuffleKind == 0 && !isLE) || (ShuffleKind == 2 && isLE)) { | |||
2079 | // Check the rest of the elements to see if they are consecutive. | |||
2080 | for (++i; i != 16; ++i) | |||
2081 | if (!isConstantOrUndef(SVOp->getMaskElt(i), ShiftAmt+i)) | |||
2082 | return -1; | |||
2083 | } else if (ShuffleKind == 1) { | |||
2084 | // Check the rest of the elements to see if they are consecutive. | |||
2085 | for (++i; i != 16; ++i) | |||
2086 | if (!isConstantOrUndef(SVOp->getMaskElt(i), (ShiftAmt+i) & 15)) | |||
2087 | return -1; | |||
2088 | } else | |||
2089 | return -1; | |||
2090 | ||||
2091 | if (isLE) | |||
2092 | ShiftAmt = 16 - ShiftAmt; | |||
2093 | ||||
2094 | return ShiftAmt; | |||
2095 | } | |||
2096 | ||||
2097 | /// isSplatShuffleMask - Return true if the specified VECTOR_SHUFFLE operand | |||
2098 | /// specifies a splat of a single element that is suitable for input to | |||
2099 | /// one of the splat operations (VSPLTB/VSPLTH/VSPLTW/XXSPLTW/LXVDSX/etc.). | |||
2100 | bool PPC::isSplatShuffleMask(ShuffleVectorSDNode *N, unsigned EltSize) { | |||
2101 | assert(N->getValueType(0) == MVT::v16i8 && isPowerOf2_32(EltSize) &&(static_cast <bool> (N->getValueType(0) == MVT::v16i8 && isPowerOf2_32(EltSize) && EltSize <= 8 && "Can only handle 1,2,4,8 byte element sizes") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::v16i8 && isPowerOf2_32(EltSize) && EltSize <= 8 && \"Can only handle 1,2,4,8 byte element sizes\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2102, __extension__ __PRETTY_FUNCTION__)) | |||
2102 | EltSize <= 8 && "Can only handle 1,2,4,8 byte element sizes")(static_cast <bool> (N->getValueType(0) == MVT::v16i8 && isPowerOf2_32(EltSize) && EltSize <= 8 && "Can only handle 1,2,4,8 byte element sizes") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::v16i8 && isPowerOf2_32(EltSize) && EltSize <= 8 && \"Can only handle 1,2,4,8 byte element sizes\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2102, __extension__ __PRETTY_FUNCTION__)); | |||
2103 | ||||
2104 | // The consecutive indices need to specify an element, not part of two | |||
2105 | // different elements. So abandon ship early if this isn't the case. | |||
2106 | if (N->getMaskElt(0) % EltSize != 0) | |||
2107 | return false; | |||
2108 | ||||
2109 | // This is a splat operation if each element of the permute is the same, and | |||
2110 | // if the value doesn't reference the second vector. | |||
2111 | unsigned ElementBase = N->getMaskElt(0); | |||
2112 | ||||
2113 | // FIXME: Handle UNDEF elements too! | |||
2114 | if (ElementBase >= 16) | |||
2115 | return false; | |||
2116 | ||||
2117 | // Check that the indices are consecutive, in the case of a multi-byte element | |||
2118 | // splatted with a v16i8 mask. | |||
2119 | for (unsigned i = 1; i != EltSize; ++i) | |||
2120 | if (N->getMaskElt(i) < 0 || N->getMaskElt(i) != (int)(i+ElementBase)) | |||
2121 | return false; | |||
2122 | ||||
2123 | for (unsigned i = EltSize, e = 16; i != e; i += EltSize) { | |||
2124 | if (N->getMaskElt(i) < 0) continue; | |||
2125 | for (unsigned j = 0; j != EltSize; ++j) | |||
2126 | if (N->getMaskElt(i+j) != N->getMaskElt(j)) | |||
2127 | return false; | |||
2128 | } | |||
2129 | return true; | |||
2130 | } | |||
2131 | ||||
2132 | /// Check that the mask is shuffling N byte elements. Within each N byte | |||
2133 | /// element of the mask, the indices could be either in increasing or | |||
2134 | /// decreasing order as long as they are consecutive. | |||
2135 | /// \param[in] N the shuffle vector SD Node to analyze | |||
2136 | /// \param[in] Width the element width in bytes, could be 2/4/8/16 (HalfWord/ | |||
2137 | /// Word/DoubleWord/QuadWord). | |||
2138 | /// \param[in] StepLen the delta indices number among the N byte element, if | |||
2139 | /// the mask is in increasing/decreasing order then it is 1/-1. | |||
2140 | /// \return true iff the mask is shuffling N byte elements. | |||
2141 | static bool isNByteElemShuffleMask(ShuffleVectorSDNode *N, unsigned Width, | |||
2142 | int StepLen) { | |||
2143 | assert((Width == 2 || Width == 4 || Width == 8 || Width == 16) &&(static_cast <bool> ((Width == 2 || Width == 4 || Width == 8 || Width == 16) && "Unexpected element width.") ? void (0) : __assert_fail ("(Width == 2 || Width == 4 || Width == 8 || Width == 16) && \"Unexpected element width.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2144, __extension__ __PRETTY_FUNCTION__)) | |||
2144 | "Unexpected element width.")(static_cast <bool> ((Width == 2 || Width == 4 || Width == 8 || Width == 16) && "Unexpected element width.") ? void (0) : __assert_fail ("(Width == 2 || Width == 4 || Width == 8 || Width == 16) && \"Unexpected element width.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2144, __extension__ __PRETTY_FUNCTION__)); | |||
2145 | assert((StepLen == 1 || StepLen == -1) && "Unexpected element width.")(static_cast <bool> ((StepLen == 1 || StepLen == -1) && "Unexpected element width.") ? void (0) : __assert_fail ("(StepLen == 1 || StepLen == -1) && \"Unexpected element width.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2145, __extension__ __PRETTY_FUNCTION__)); | |||
2146 | ||||
2147 | unsigned NumOfElem = 16 / Width; | |||
2148 | unsigned MaskVal[16]; // Width is never greater than 16 | |||
2149 | for (unsigned i = 0; i < NumOfElem; ++i) { | |||
2150 | MaskVal[0] = N->getMaskElt(i * Width); | |||
2151 | if ((StepLen == 1) && (MaskVal[0] % Width)) { | |||
2152 | return false; | |||
2153 | } else if ((StepLen == -1) && ((MaskVal[0] + 1) % Width)) { | |||
2154 | return false; | |||
2155 | } | |||
2156 | ||||
2157 | for (unsigned int j = 1; j < Width; ++j) { | |||
2158 | MaskVal[j] = N->getMaskElt(i * Width + j); | |||
2159 | if (MaskVal[j] != MaskVal[j-1] + StepLen) { | |||
2160 | return false; | |||
2161 | } | |||
2162 | } | |||
2163 | } | |||
2164 | ||||
2165 | return true; | |||
2166 | } | |||
2167 | ||||
2168 | bool PPC::isXXINSERTWMask(ShuffleVectorSDNode *N, unsigned &ShiftElts, | |||
2169 | unsigned &InsertAtByte, bool &Swap, bool IsLE) { | |||
2170 | if (!isNByteElemShuffleMask(N, 4, 1)) | |||
2171 | return false; | |||
2172 | ||||
2173 | // Now we look at mask elements 0,4,8,12 | |||
2174 | unsigned M0 = N->getMaskElt(0) / 4; | |||
2175 | unsigned M1 = N->getMaskElt(4) / 4; | |||
2176 | unsigned M2 = N->getMaskElt(8) / 4; | |||
2177 | unsigned M3 = N->getMaskElt(12) / 4; | |||
2178 | unsigned LittleEndianShifts[] = { 2, 1, 0, 3 }; | |||
2179 | unsigned BigEndianShifts[] = { 3, 0, 1, 2 }; | |||
2180 | ||||
2181 | // Below, let H and L be arbitrary elements of the shuffle mask | |||
2182 | // where H is in the range [4,7] and L is in the range [0,3]. | |||
2183 | // H, 1, 2, 3 or L, 5, 6, 7 | |||
2184 | if ((M0 > 3 && M1 == 1 && M2 == 2 && M3 == 3) || | |||
2185 | (M0 < 4 && M1 == 5 && M2 == 6 && M3 == 7)) { | |||
2186 | ShiftElts = IsLE ? LittleEndianShifts[M0 & 0x3] : BigEndianShifts[M0 & 0x3]; | |||
2187 | InsertAtByte = IsLE ? 12 : 0; | |||
2188 | Swap = M0 < 4; | |||
2189 | return true; | |||
2190 | } | |||
2191 | // 0, H, 2, 3 or 4, L, 6, 7 | |||
2192 | if ((M1 > 3 && M0 == 0 && M2 == 2 && M3 == 3) || | |||
2193 | (M1 < 4 && M0 == 4 && M2 == 6 && M3 == 7)) { | |||
2194 | ShiftElts = IsLE ? LittleEndianShifts[M1 & 0x3] : BigEndianShifts[M1 & 0x3]; | |||
2195 | InsertAtByte = IsLE ? 8 : 4; | |||
2196 | Swap = M1 < 4; | |||
2197 | return true; | |||
2198 | } | |||
2199 | // 0, 1, H, 3 or 4, 5, L, 7 | |||
2200 | if ((M2 > 3 && M0 == 0 && M1 == 1 && M3 == 3) || | |||
2201 | (M2 < 4 && M0 == 4 && M1 == 5 && M3 == 7)) { | |||
2202 | ShiftElts = IsLE ? LittleEndianShifts[M2 & 0x3] : BigEndianShifts[M2 & 0x3]; | |||
2203 | InsertAtByte = IsLE ? 4 : 8; | |||
2204 | Swap = M2 < 4; | |||
2205 | return true; | |||
2206 | } | |||
2207 | // 0, 1, 2, H or 4, 5, 6, L | |||
2208 | if ((M3 > 3 && M0 == 0 && M1 == 1 && M2 == 2) || | |||
2209 | (M3 < 4 && M0 == 4 && M1 == 5 && M2 == 6)) { | |||
2210 | ShiftElts = IsLE ? LittleEndianShifts[M3 & 0x3] : BigEndianShifts[M3 & 0x3]; | |||
2211 | InsertAtByte = IsLE ? 0 : 12; | |||
2212 | Swap = M3 < 4; | |||
2213 | return true; | |||
2214 | } | |||
2215 | ||||
2216 | // If both vector operands for the shuffle are the same vector, the mask will | |||
2217 | // contain only elements from the first one and the second one will be undef. | |||
2218 | if (N->getOperand(1).isUndef()) { | |||
2219 | ShiftElts = 0; | |||
2220 | Swap = true; | |||
2221 | unsigned XXINSERTWSrcElem = IsLE ? 2 : 1; | |||
2222 | if (M0 == XXINSERTWSrcElem && M1 == 1 && M2 == 2 && M3 == 3) { | |||
2223 | InsertAtByte = IsLE ? 12 : 0; | |||
2224 | return true; | |||
2225 | } | |||
2226 | if (M0 == 0 && M1 == XXINSERTWSrcElem && M2 == 2 && M3 == 3) { | |||
2227 | InsertAtByte = IsLE ? 8 : 4; | |||
2228 | return true; | |||
2229 | } | |||
2230 | if (M0 == 0 && M1 == 1 && M2 == XXINSERTWSrcElem && M3 == 3) { | |||
2231 | InsertAtByte = IsLE ? 4 : 8; | |||
2232 | return true; | |||
2233 | } | |||
2234 | if (M0 == 0 && M1 == 1 && M2 == 2 && M3 == XXINSERTWSrcElem) { | |||
2235 | InsertAtByte = IsLE ? 0 : 12; | |||
2236 | return true; | |||
2237 | } | |||
2238 | } | |||
2239 | ||||
2240 | return false; | |||
2241 | } | |||
2242 | ||||
2243 | bool PPC::isXXSLDWIShuffleMask(ShuffleVectorSDNode *N, unsigned &ShiftElts, | |||
2244 | bool &Swap, bool IsLE) { | |||
2245 | assert(N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8")(static_cast <bool> (N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::v16i8 && \"Shuffle vector expects v16i8\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2245, __extension__ __PRETTY_FUNCTION__)); | |||
2246 | // Ensure each byte index of the word is consecutive. | |||
2247 | if (!isNByteElemShuffleMask(N, 4, 1)) | |||
2248 | return false; | |||
2249 | ||||
2250 | // Now we look at mask elements 0,4,8,12, which are the beginning of words. | |||
2251 | unsigned M0 = N->getMaskElt(0) / 4; | |||
2252 | unsigned M1 = N->getMaskElt(4) / 4; | |||
2253 | unsigned M2 = N->getMaskElt(8) / 4; | |||
2254 | unsigned M3 = N->getMaskElt(12) / 4; | |||
2255 | ||||
2256 | // If both vector operands for the shuffle are the same vector, the mask will | |||
2257 | // contain only elements from the first one and the second one will be undef. | |||
2258 | if (N->getOperand(1).isUndef()) { | |||
2259 | assert(M0 < 4 && "Indexing into an undef vector?")(static_cast <bool> (M0 < 4 && "Indexing into an undef vector?" ) ? void (0) : __assert_fail ("M0 < 4 && \"Indexing into an undef vector?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2259, __extension__ __PRETTY_FUNCTION__)); | |||
2260 | if (M1 != (M0 + 1) % 4 || M2 != (M1 + 1) % 4 || M3 != (M2 + 1) % 4) | |||
2261 | return false; | |||
2262 | ||||
2263 | ShiftElts = IsLE ? (4 - M0) % 4 : M0; | |||
2264 | Swap = false; | |||
2265 | return true; | |||
2266 | } | |||
2267 | ||||
2268 | // Ensure each word index of the ShuffleVector Mask is consecutive. | |||
2269 | if (M1 != (M0 + 1) % 8 || M2 != (M1 + 1) % 8 || M3 != (M2 + 1) % 8) | |||
2270 | return false; | |||
2271 | ||||
2272 | if (IsLE) { | |||
2273 | if (M0 == 0 || M0 == 7 || M0 == 6 || M0 == 5) { | |||
2274 | // Input vectors don't need to be swapped if the leading element | |||
2275 | // of the result is one of the 3 left elements of the second vector | |||
2276 | // (or if there is no shift to be done at all). | |||
2277 | Swap = false; | |||
2278 | ShiftElts = (8 - M0) % 8; | |||
2279 | } else if (M0 == 4 || M0 == 3 || M0 == 2 || M0 == 1) { | |||
2280 | // Input vectors need to be swapped if the leading element | |||
2281 | // of the result is one of the 3 left elements of the first vector | |||
2282 | // (or if we're shifting by 4 - thereby simply swapping the vectors). | |||
2283 | Swap = true; | |||
2284 | ShiftElts = (4 - M0) % 4; | |||
2285 | } | |||
2286 | ||||
2287 | return true; | |||
2288 | } else { // BE | |||
2289 | if (M0 == 0 || M0 == 1 || M0 == 2 || M0 == 3) { | |||
2290 | // Input vectors don't need to be swapped if the leading element | |||
2291 | // of the result is one of the 4 elements of the first vector. | |||
2292 | Swap = false; | |||
2293 | ShiftElts = M0; | |||
2294 | } else if (M0 == 4 || M0 == 5 || M0 == 6 || M0 == 7) { | |||
2295 | // Input vectors need to be swapped if the leading element | |||
2296 | // of the result is one of the 4 elements of the right vector. | |||
2297 | Swap = true; | |||
2298 | ShiftElts = M0 - 4; | |||
2299 | } | |||
2300 | ||||
2301 | return true; | |||
2302 | } | |||
2303 | } | |||
2304 | ||||
2305 | bool static isXXBRShuffleMaskHelper(ShuffleVectorSDNode *N, int Width) { | |||
2306 | assert(N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8")(static_cast <bool> (N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::v16i8 && \"Shuffle vector expects v16i8\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2306, __extension__ __PRETTY_FUNCTION__)); | |||
2307 | ||||
2308 | if (!isNByteElemShuffleMask(N, Width, -1)) | |||
2309 | return false; | |||
2310 | ||||
2311 | for (int i = 0; i < 16; i += Width) | |||
2312 | if (N->getMaskElt(i) != i + Width - 1) | |||
2313 | return false; | |||
2314 | ||||
2315 | return true; | |||
2316 | } | |||
2317 | ||||
2318 | bool PPC::isXXBRHShuffleMask(ShuffleVectorSDNode *N) { | |||
2319 | return isXXBRShuffleMaskHelper(N, 2); | |||
2320 | } | |||
2321 | ||||
2322 | bool PPC::isXXBRWShuffleMask(ShuffleVectorSDNode *N) { | |||
2323 | return isXXBRShuffleMaskHelper(N, 4); | |||
2324 | } | |||
2325 | ||||
2326 | bool PPC::isXXBRDShuffleMask(ShuffleVectorSDNode *N) { | |||
2327 | return isXXBRShuffleMaskHelper(N, 8); | |||
2328 | } | |||
2329 | ||||
2330 | bool PPC::isXXBRQShuffleMask(ShuffleVectorSDNode *N) { | |||
2331 | return isXXBRShuffleMaskHelper(N, 16); | |||
2332 | } | |||
2333 | ||||
2334 | /// Can node \p N be lowered to an XXPERMDI instruction? If so, set \p Swap | |||
2335 | /// if the inputs to the instruction should be swapped and set \p DM to the | |||
2336 | /// value for the immediate. | |||
2337 | /// Specifically, set \p Swap to true only if \p N can be lowered to XXPERMDI | |||
2338 | /// AND element 0 of the result comes from the first input (LE) or second input | |||
2339 | /// (BE). Set \p DM to the calculated result (0-3) only if \p N can be lowered. | |||
2340 | /// \return true iff the given mask of shuffle node \p N is a XXPERMDI shuffle | |||
2341 | /// mask. | |||
2342 | bool PPC::isXXPERMDIShuffleMask(ShuffleVectorSDNode *N, unsigned &DM, | |||
2343 | bool &Swap, bool IsLE) { | |||
2344 | assert(N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8")(static_cast <bool> (N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::v16i8 && \"Shuffle vector expects v16i8\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2344, __extension__ __PRETTY_FUNCTION__)); | |||
2345 | ||||
2346 | // Ensure each byte index of the double word is consecutive. | |||
2347 | if (!isNByteElemShuffleMask(N, 8, 1)) | |||
2348 | return false; | |||
2349 | ||||
2350 | unsigned M0 = N->getMaskElt(0) / 8; | |||
2351 | unsigned M1 = N->getMaskElt(8) / 8; | |||
2352 | assert(((M0 | M1) < 4) && "A mask element out of bounds?")(static_cast <bool> (((M0 | M1) < 4) && "A mask element out of bounds?" ) ? void (0) : __assert_fail ("((M0 | M1) < 4) && \"A mask element out of bounds?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2352, __extension__ __PRETTY_FUNCTION__)); | |||
2353 | ||||
2354 | // If both vector operands for the shuffle are the same vector, the mask will | |||
2355 | // contain only elements from the first one and the second one will be undef. | |||
2356 | if (N->getOperand(1).isUndef()) { | |||
2357 | if ((M0 | M1) < 2) { | |||
2358 | DM = IsLE ? (((~M1) & 1) << 1) + ((~M0) & 1) : (M0 << 1) + (M1 & 1); | |||
2359 | Swap = false; | |||
2360 | return true; | |||
2361 | } else | |||
2362 | return false; | |||
2363 | } | |||
2364 | ||||
2365 | if (IsLE) { | |||
2366 | if (M0 > 1 && M1 < 2) { | |||
2367 | Swap = false; | |||
2368 | } else if (M0 < 2 && M1 > 1) { | |||
2369 | M0 = (M0 + 2) % 4; | |||
2370 | M1 = (M1 + 2) % 4; | |||
2371 | Swap = true; | |||
2372 | } else | |||
2373 | return false; | |||
2374 | ||||
2375 | // Note: if control flow comes here that means Swap is already set above | |||
2376 | DM = (((~M1) & 1) << 1) + ((~M0) & 1); | |||
2377 | return true; | |||
2378 | } else { // BE | |||
2379 | if (M0 < 2 && M1 > 1) { | |||
2380 | Swap = false; | |||
2381 | } else if (M0 > 1 && M1 < 2) { | |||
2382 | M0 = (M0 + 2) % 4; | |||
2383 | M1 = (M1 + 2) % 4; | |||
2384 | Swap = true; | |||
2385 | } else | |||
2386 | return false; | |||
2387 | ||||
2388 | // Note: if control flow comes here that means Swap is already set above | |||
2389 | DM = (M0 << 1) + (M1 & 1); | |||
2390 | return true; | |||
2391 | } | |||
2392 | } | |||
2393 | ||||
2394 | ||||
2395 | /// getSplatIdxForPPCMnemonics - Return the splat index as a value that is | |||
2396 | /// appropriate for PPC mnemonics (which have a big endian bias - namely | |||
2397 | /// elements are counted from the left of the vector register). | |||
2398 | unsigned PPC::getSplatIdxForPPCMnemonics(SDNode *N, unsigned EltSize, | |||
2399 | SelectionDAG &DAG) { | |||
2400 | ShuffleVectorSDNode *SVOp = cast<ShuffleVectorSDNode>(N); | |||
2401 | assert(isSplatShuffleMask(SVOp, EltSize))(static_cast <bool> (isSplatShuffleMask(SVOp, EltSize)) ? void (0) : __assert_fail ("isSplatShuffleMask(SVOp, EltSize)" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2401, __extension__ __PRETTY_FUNCTION__)); | |||
2402 | if (DAG.getDataLayout().isLittleEndian()) | |||
2403 | return (16 / EltSize) - 1 - (SVOp->getMaskElt(0) / EltSize); | |||
2404 | else | |||
2405 | return SVOp->getMaskElt(0) / EltSize; | |||
2406 | } | |||
2407 | ||||
2408 | /// get_VSPLTI_elt - If this is a build_vector of constants which can be formed | |||
2409 | /// by using a vspltis[bhw] instruction of the specified element size, return | |||
2410 | /// the constant being splatted. The ByteSize field indicates the number of | |||
2411 | /// bytes of each element [124] -> [bhw]. | |||
2412 | SDValue PPC::get_VSPLTI_elt(SDNode *N, unsigned ByteSize, SelectionDAG &DAG) { | |||
2413 | SDValue OpVal(nullptr, 0); | |||
2414 | ||||
2415 | // If ByteSize of the splat is bigger than the element size of the | |||
2416 | // build_vector, then we have a case where we are checking for a splat where | |||
2417 | // multiple elements of the buildvector are folded together into a single | |||
2418 | // logical element of the splat (e.g. "vsplish 1" to splat {0,1}*8). | |||
2419 | unsigned EltSize = 16/N->getNumOperands(); | |||
2420 | if (EltSize < ByteSize) { | |||
2421 | unsigned Multiple = ByteSize/EltSize; // Number of BV entries per spltval. | |||
2422 | SDValue UniquedVals[4]; | |||
2423 | assert(Multiple > 1 && Multiple <= 4 && "How can this happen?")(static_cast <bool> (Multiple > 1 && Multiple <= 4 && "How can this happen?") ? void (0) : __assert_fail ("Multiple > 1 && Multiple <= 4 && \"How can this happen?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2423, __extension__ __PRETTY_FUNCTION__)); | |||
2424 | ||||
2425 | // See if all of the elements in the buildvector agree across. | |||
2426 | for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) { | |||
2427 | if (N->getOperand(i).isUndef()) continue; | |||
2428 | // If the element isn't a constant, bail fully out. | |||
2429 | if (!isa<ConstantSDNode>(N->getOperand(i))) return SDValue(); | |||
2430 | ||||
2431 | if (!UniquedVals[i&(Multiple-1)].getNode()) | |||
2432 | UniquedVals[i&(Multiple-1)] = N->getOperand(i); | |||
2433 | else if (UniquedVals[i&(Multiple-1)] != N->getOperand(i)) | |||
2434 | return SDValue(); // no match. | |||
2435 | } | |||
2436 | ||||
2437 | // Okay, if we reached this point, UniquedVals[0..Multiple-1] contains | |||
2438 | // either constant or undef values that are identical for each chunk. See | |||
2439 | // if these chunks can form into a larger vspltis*. | |||
2440 | ||||
2441 | // Check to see if all of the leading entries are either 0 or -1. If | |||
2442 | // neither, then this won't fit into the immediate field. | |||
2443 | bool LeadingZero = true; | |||
2444 | bool LeadingOnes = true; | |||
2445 | for (unsigned i = 0; i != Multiple-1; ++i) { | |||
2446 | if (!UniquedVals[i].getNode()) continue; // Must have been undefs. | |||
2447 | ||||
2448 | LeadingZero &= isNullConstant(UniquedVals[i]); | |||
2449 | LeadingOnes &= isAllOnesConstant(UniquedVals[i]); | |||
2450 | } | |||
2451 | // Finally, check the least significant entry. | |||
2452 | if (LeadingZero) { | |||
2453 | if (!UniquedVals[Multiple-1].getNode()) | |||
2454 | return DAG.getTargetConstant(0, SDLoc(N), MVT::i32); // 0,0,0,undef | |||
2455 | int Val = cast<ConstantSDNode>(UniquedVals[Multiple-1])->getZExtValue(); | |||
2456 | if (Val < 16) // 0,0,0,4 -> vspltisw(4) | |||
2457 | return DAG.getTargetConstant(Val, SDLoc(N), MVT::i32); | |||
2458 | } | |||
2459 | if (LeadingOnes) { | |||
2460 | if (!UniquedVals[Multiple-1].getNode()) | |||
2461 | return DAG.getTargetConstant(~0U, SDLoc(N), MVT::i32); // -1,-1,-1,undef | |||
2462 | int Val =cast<ConstantSDNode>(UniquedVals[Multiple-1])->getSExtValue(); | |||
2463 | if (Val >= -16) // -1,-1,-1,-2 -> vspltisw(-2) | |||
2464 | return DAG.getTargetConstant(Val, SDLoc(N), MVT::i32); | |||
2465 | } | |||
2466 | ||||
2467 | return SDValue(); | |||
2468 | } | |||
2469 | ||||
2470 | // Check to see if this buildvec has a single non-undef value in its elements. | |||
2471 | for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) { | |||
2472 | if (N->getOperand(i).isUndef()) continue; | |||
2473 | if (!OpVal.getNode()) | |||
2474 | OpVal = N->getOperand(i); | |||
2475 | else if (OpVal != N->getOperand(i)) | |||
2476 | return SDValue(); | |||
2477 | } | |||
2478 | ||||
2479 | if (!OpVal.getNode()) return SDValue(); // All UNDEF: use implicit def. | |||
2480 | ||||
2481 | unsigned ValSizeInBytes = EltSize; | |||
2482 | uint64_t Value = 0; | |||
2483 | if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(OpVal)) { | |||
2484 | Value = CN->getZExtValue(); | |||
2485 | } else if (ConstantFPSDNode *CN = dyn_cast<ConstantFPSDNode>(OpVal)) { | |||
2486 | assert(CN->getValueType(0) == MVT::f32 && "Only one legal FP vector type!")(static_cast <bool> (CN->getValueType(0) == MVT::f32 && "Only one legal FP vector type!") ? void (0) : __assert_fail ("CN->getValueType(0) == MVT::f32 && \"Only one legal FP vector type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2486, __extension__ __PRETTY_FUNCTION__)); | |||
2487 | Value = FloatToBits(CN->getValueAPF().convertToFloat()); | |||
2488 | } | |||
2489 | ||||
2490 | // If the splat value is larger than the element value, then we can never do | |||
2491 | // this splat. The only case that we could fit the replicated bits into our | |||
2492 | // immediate field for would be zero, and we prefer to use vxor for it. | |||
2493 | if (ValSizeInBytes < ByteSize) return SDValue(); | |||
2494 | ||||
2495 | // If the element value is larger than the splat value, check if it consists | |||
2496 | // of a repeated bit pattern of size ByteSize. | |||
2497 | if (!APInt(ValSizeInBytes * 8, Value).isSplat(ByteSize * 8)) | |||
2498 | return SDValue(); | |||
2499 | ||||
2500 | // Properly sign extend the value. | |||
2501 | int MaskVal = SignExtend32(Value, ByteSize * 8); | |||
2502 | ||||
2503 | // If this is zero, don't match, zero matches ISD::isBuildVectorAllZeros. | |||
2504 | if (MaskVal == 0) return SDValue(); | |||
2505 | ||||
2506 | // Finally, if this value fits in a 5 bit sext field, return it | |||
2507 | if (SignExtend32<5>(MaskVal) == MaskVal) | |||
2508 | return DAG.getTargetConstant(MaskVal, SDLoc(N), MVT::i32); | |||
2509 | return SDValue(); | |||
2510 | } | |||
2511 | ||||
2512 | //===----------------------------------------------------------------------===// | |||
2513 | // Addressing Mode Selection | |||
2514 | //===----------------------------------------------------------------------===// | |||
2515 | ||||
2516 | /// isIntS16Immediate - This method tests to see if the node is either a 32-bit | |||
2517 | /// or 64-bit immediate, and if the value can be accurately represented as a | |||
2518 | /// sign extension from a 16-bit value. If so, this returns true and the | |||
2519 | /// immediate. | |||
2520 | bool llvm::isIntS16Immediate(SDNode *N, int16_t &Imm) { | |||
2521 | if (!isa<ConstantSDNode>(N)) | |||
2522 | return false; | |||
2523 | ||||
2524 | Imm = (int16_t)cast<ConstantSDNode>(N)->getZExtValue(); | |||
2525 | if (N->getValueType(0) == MVT::i32) | |||
2526 | return Imm == (int32_t)cast<ConstantSDNode>(N)->getZExtValue(); | |||
2527 | else | |||
2528 | return Imm == (int64_t)cast<ConstantSDNode>(N)->getZExtValue(); | |||
2529 | } | |||
2530 | bool llvm::isIntS16Immediate(SDValue Op, int16_t &Imm) { | |||
2531 | return isIntS16Immediate(Op.getNode(), Imm); | |||
2532 | } | |||
2533 | ||||
2534 | /// Used when computing address flags for selecting loads and stores. | |||
2535 | /// If we have an OR, check if the LHS and RHS are provably disjoint. | |||
2536 | /// An OR of two provably disjoint values is equivalent to an ADD. | |||
2537 | /// Most PPC load/store instructions compute the effective address as a sum, | |||
2538 | /// so doing this conversion is useful. | |||
2539 | static bool provablyDisjointOr(SelectionDAG &DAG, const SDValue &N) { | |||
2540 | if (N.getOpcode() != ISD::OR) | |||
2541 | return false; | |||
2542 | KnownBits LHSKnown = DAG.computeKnownBits(N.getOperand(0)); | |||
2543 | if (!LHSKnown.Zero.getBoolValue()) | |||
2544 | return false; | |||
2545 | KnownBits RHSKnown = DAG.computeKnownBits(N.getOperand(1)); | |||
2546 | return (~(LHSKnown.Zero | RHSKnown.Zero) == 0); | |||
2547 | } | |||
2548 | ||||
2549 | /// SelectAddressEVXRegReg - Given the specified address, check to see if it can | |||
2550 | /// be represented as an indexed [r+r] operation. | |||
2551 | bool PPCTargetLowering::SelectAddressEVXRegReg(SDValue N, SDValue &Base, | |||
2552 | SDValue &Index, | |||
2553 | SelectionDAG &DAG) const { | |||
2554 | for (SDNode::use_iterator UI = N->use_begin(), E = N->use_end(); | |||
2555 | UI != E; ++UI) { | |||
2556 | if (MemSDNode *Memop = dyn_cast<MemSDNode>(*UI)) { | |||
2557 | if (Memop->getMemoryVT() == MVT::f64) { | |||
2558 | Base = N.getOperand(0); | |||
2559 | Index = N.getOperand(1); | |||
2560 | return true; | |||
2561 | } | |||
2562 | } | |||
2563 | } | |||
2564 | return false; | |||
2565 | } | |||
2566 | ||||
2567 | /// isIntS34Immediate - This method tests if value of node given can be | |||
2568 | /// accurately represented as a sign extension from a 34-bit value. If so, | |||
2569 | /// this returns true and the immediate. | |||
2570 | bool llvm::isIntS34Immediate(SDNode *N, int64_t &Imm) { | |||
2571 | if (!isa<ConstantSDNode>(N)) | |||
2572 | return false; | |||
2573 | ||||
2574 | Imm = (int64_t)cast<ConstantSDNode>(N)->getZExtValue(); | |||
2575 | return isInt<34>(Imm); | |||
2576 | } | |||
2577 | bool llvm::isIntS34Immediate(SDValue Op, int64_t &Imm) { | |||
2578 | return isIntS34Immediate(Op.getNode(), Imm); | |||
2579 | } | |||
2580 | ||||
2581 | /// SelectAddressRegReg - Given the specified addressed, check to see if it | |||
2582 | /// can be represented as an indexed [r+r] operation. Returns false if it | |||
2583 | /// can be more efficiently represented as [r+imm]. If \p EncodingAlignment is | |||
2584 | /// non-zero and N can be represented by a base register plus a signed 16-bit | |||
2585 | /// displacement, make a more precise judgement by checking (displacement % \p | |||
2586 | /// EncodingAlignment). | |||
2587 | bool PPCTargetLowering::SelectAddressRegReg( | |||
2588 | SDValue N, SDValue &Base, SDValue &Index, SelectionDAG &DAG, | |||
2589 | MaybeAlign EncodingAlignment) const { | |||
2590 | // If we have a PC Relative target flag don't select as [reg+reg]. It will be | |||
2591 | // a [pc+imm]. | |||
2592 | if (SelectAddressPCRel(N, Base)) | |||
2593 | return false; | |||
2594 | ||||
2595 | int16_t Imm = 0; | |||
2596 | if (N.getOpcode() == ISD::ADD) { | |||
2597 | // Is there any SPE load/store (f64), which can't handle 16bit offset? | |||
2598 | // SPE load/store can only handle 8-bit offsets. | |||
2599 | if (hasSPE() && SelectAddressEVXRegReg(N, Base, Index, DAG)) | |||
2600 | return true; | |||
2601 | if (isIntS16Immediate(N.getOperand(1), Imm) && | |||
2602 | (!EncodingAlignment || isAligned(*EncodingAlignment, Imm))) | |||
2603 | return false; // r+i | |||
2604 | if (N.getOperand(1).getOpcode() == PPCISD::Lo) | |||
2605 | return false; // r+i | |||
2606 | ||||
2607 | Base = N.getOperand(0); | |||
2608 | Index = N.getOperand(1); | |||
2609 | return true; | |||
2610 | } else if (N.getOpcode() == ISD::OR) { | |||
2611 | if (isIntS16Immediate(N.getOperand(1), Imm) && | |||
2612 | (!EncodingAlignment || isAligned(*EncodingAlignment, Imm))) | |||
2613 | return false; // r+i can fold it if we can. | |||
2614 | ||||
2615 | // If this is an or of disjoint bitfields, we can codegen this as an add | |||
2616 | // (for better address arithmetic) if the LHS and RHS of the OR are provably | |||
2617 | // disjoint. | |||
2618 | KnownBits LHSKnown = DAG.computeKnownBits(N.getOperand(0)); | |||
2619 | ||||
2620 | if (LHSKnown.Zero.getBoolValue()) { | |||
2621 | KnownBits RHSKnown = DAG.computeKnownBits(N.getOperand(1)); | |||
2622 | // If all of the bits are known zero on the LHS or RHS, the add won't | |||
2623 | // carry. | |||
2624 | if (~(LHSKnown.Zero | RHSKnown.Zero) == 0) { | |||
2625 | Base = N.getOperand(0); | |||
2626 | Index = N.getOperand(1); | |||
2627 | return true; | |||
2628 | } | |||
2629 | } | |||
2630 | } | |||
2631 | ||||
2632 | return false; | |||
2633 | } | |||
2634 | ||||
2635 | // If we happen to be doing an i64 load or store into a stack slot that has | |||
2636 | // less than a 4-byte alignment, then the frame-index elimination may need to | |||
2637 | // use an indexed load or store instruction (because the offset may not be a | |||
2638 | // multiple of 4). The extra register needed to hold the offset comes from the | |||
2639 | // register scavenger, and it is possible that the scavenger will need to use | |||
2640 | // an emergency spill slot. As a result, we need to make sure that a spill slot | |||
2641 | // is allocated when doing an i64 load/store into a less-than-4-byte-aligned | |||
2642 | // stack slot. | |||
2643 | static void fixupFuncForFI(SelectionDAG &DAG, int FrameIdx, EVT VT) { | |||
2644 | // FIXME: This does not handle the LWA case. | |||
2645 | if (VT != MVT::i64) | |||
2646 | return; | |||
2647 | ||||
2648 | // NOTE: We'll exclude negative FIs here, which come from argument | |||
2649 | // lowering, because there are no known test cases triggering this problem | |||
2650 | // using packed structures (or similar). We can remove this exclusion if | |||
2651 | // we find such a test case. The reason why this is so test-case driven is | |||
2652 | // because this entire 'fixup' is only to prevent crashes (from the | |||
2653 | // register scavenger) on not-really-valid inputs. For example, if we have: | |||
2654 | // %a = alloca i1 | |||
2655 | // %b = bitcast i1* %a to i64* | |||
2656 | // store i64* a, i64 b | |||
2657 | // then the store should really be marked as 'align 1', but is not. If it | |||
2658 | // were marked as 'align 1' then the indexed form would have been | |||
2659 | // instruction-selected initially, and the problem this 'fixup' is preventing | |||
2660 | // won't happen regardless. | |||
2661 | if (FrameIdx < 0) | |||
2662 | return; | |||
2663 | ||||
2664 | MachineFunction &MF = DAG.getMachineFunction(); | |||
2665 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
2666 | ||||
2667 | if (MFI.getObjectAlign(FrameIdx) >= Align(4)) | |||
2668 | return; | |||
2669 | ||||
2670 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
2671 | FuncInfo->setHasNonRISpills(); | |||
2672 | } | |||
2673 | ||||
2674 | /// Returns true if the address N can be represented by a base register plus | |||
2675 | /// a signed 16-bit displacement [r+imm], and if it is not better | |||
2676 | /// represented as reg+reg. If \p EncodingAlignment is non-zero, only accept | |||
2677 | /// displacements that are multiples of that value. | |||
2678 | bool PPCTargetLowering::SelectAddressRegImm( | |||
2679 | SDValue N, SDValue &Disp, SDValue &Base, SelectionDAG &DAG, | |||
2680 | MaybeAlign EncodingAlignment) const { | |||
2681 | // FIXME dl should come from parent load or store, not from address | |||
2682 | SDLoc dl(N); | |||
2683 | ||||
2684 | // If we have a PC Relative target flag don't select as [reg+imm]. It will be | |||
2685 | // a [pc+imm]. | |||
2686 | if (SelectAddressPCRel(N, Base)) | |||
2687 | return false; | |||
2688 | ||||
2689 | // If this can be more profitably realized as r+r, fail. | |||
2690 | if (SelectAddressRegReg(N, Disp, Base, DAG, EncodingAlignment)) | |||
2691 | return false; | |||
2692 | ||||
2693 | if (N.getOpcode() == ISD::ADD) { | |||
2694 | int16_t imm = 0; | |||
2695 | if (isIntS16Immediate(N.getOperand(1), imm) && | |||
2696 | (!EncodingAlignment || isAligned(*EncodingAlignment, imm))) { | |||
2697 | Disp = DAG.getTargetConstant(imm, dl, N.getValueType()); | |||
2698 | if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N.getOperand(0))) { | |||
2699 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
2700 | fixupFuncForFI(DAG, FI->getIndex(), N.getValueType()); | |||
2701 | } else { | |||
2702 | Base = N.getOperand(0); | |||
2703 | } | |||
2704 | return true; // [r+i] | |||
2705 | } else if (N.getOperand(1).getOpcode() == PPCISD::Lo) { | |||
2706 | // Match LOAD (ADD (X, Lo(G))). | |||
2707 | assert(!cast<ConstantSDNode>(N.getOperand(1).getOperand(1))->getZExtValue()(static_cast <bool> (!cast<ConstantSDNode>(N.getOperand (1).getOperand(1))->getZExtValue() && "Cannot handle constant offsets yet!" ) ? void (0) : __assert_fail ("!cast<ConstantSDNode>(N.getOperand(1).getOperand(1))->getZExtValue() && \"Cannot handle constant offsets yet!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2708, __extension__ __PRETTY_FUNCTION__)) | |||
2708 | && "Cannot handle constant offsets yet!")(static_cast <bool> (!cast<ConstantSDNode>(N.getOperand (1).getOperand(1))->getZExtValue() && "Cannot handle constant offsets yet!" ) ? void (0) : __assert_fail ("!cast<ConstantSDNode>(N.getOperand(1).getOperand(1))->getZExtValue() && \"Cannot handle constant offsets yet!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2708, __extension__ __PRETTY_FUNCTION__)); | |||
2709 | Disp = N.getOperand(1).getOperand(0); // The global address. | |||
2710 | assert(Disp.getOpcode() == ISD::TargetGlobalAddress ||(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2713, __extension__ __PRETTY_FUNCTION__)) | |||
2711 | Disp.getOpcode() == ISD::TargetGlobalTLSAddress ||(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2713, __extension__ __PRETTY_FUNCTION__)) | |||
2712 | Disp.getOpcode() == ISD::TargetConstantPool ||(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2713, __extension__ __PRETTY_FUNCTION__)) | |||
2713 | Disp.getOpcode() == ISD::TargetJumpTable)(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 2713, __extension__ __PRETTY_FUNCTION__)); | |||
2714 | Base = N.getOperand(0); | |||
2715 | return true; // [&g+r] | |||
2716 | } | |||
2717 | } else if (N.getOpcode() == ISD::OR) { | |||
2718 | int16_t imm = 0; | |||
2719 | if (isIntS16Immediate(N.getOperand(1), imm) && | |||
2720 | (!EncodingAlignment || isAligned(*EncodingAlignment, imm))) { | |||
2721 | // If this is an or of disjoint bitfields, we can codegen this as an add | |||
2722 | // (for better address arithmetic) if the LHS and RHS of the OR are | |||
2723 | // provably disjoint. | |||
2724 | KnownBits LHSKnown = DAG.computeKnownBits(N.getOperand(0)); | |||
2725 | ||||
2726 | if ((LHSKnown.Zero.getZExtValue()|~(uint64_t)imm) == ~0ULL) { | |||
2727 | // If all of the bits are known zero on the LHS or RHS, the add won't | |||
2728 | // carry. | |||
2729 | if (FrameIndexSDNode *FI = | |||
2730 | dyn_cast<FrameIndexSDNode>(N.getOperand(0))) { | |||
2731 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
2732 | fixupFuncForFI(DAG, FI->getIndex(), N.getValueType()); | |||
2733 | } else { | |||
2734 | Base = N.getOperand(0); | |||
2735 | } | |||
2736 | Disp = DAG.getTargetConstant(imm, dl, N.getValueType()); | |||
2737 | return true; | |||
2738 | } | |||
2739 | } | |||
2740 | } else if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N)) { | |||
2741 | // Loading from a constant address. | |||
2742 | ||||
2743 | // If this address fits entirely in a 16-bit sext immediate field, codegen | |||
2744 | // this as "d, 0" | |||
2745 | int16_t Imm; | |||
2746 | if (isIntS16Immediate(CN, Imm) && | |||
2747 | (!EncodingAlignment || isAligned(*EncodingAlignment, Imm))) { | |||
2748 | Disp = DAG.getTargetConstant(Imm, dl, CN->getValueType(0)); | |||
2749 | Base = DAG.getRegister(Subtarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, | |||
2750 | CN->getValueType(0)); | |||
2751 | return true; | |||
2752 | } | |||
2753 | ||||
2754 | // Handle 32-bit sext immediates with LIS + addr mode. | |||
2755 | if ((CN->getValueType(0) == MVT::i32 || | |||
2756 | (int64_t)CN->getZExtValue() == (int)CN->getZExtValue()) && | |||
2757 | (!EncodingAlignment || | |||
2758 | isAligned(*EncodingAlignment, CN->getZExtValue()))) { | |||
2759 | int Addr = (int)CN->getZExtValue(); | |||
2760 | ||||
2761 | // Otherwise, break this down into an LIS + disp. | |||
2762 | Disp = DAG.getTargetConstant((short)Addr, dl, MVT::i32); | |||
2763 | ||||
2764 | Base = DAG.getTargetConstant((Addr - (signed short)Addr) >> 16, dl, | |||
2765 | MVT::i32); | |||
2766 | unsigned Opc = CN->getValueType(0) == MVT::i32 ? PPC::LIS : PPC::LIS8; | |||
2767 | Base = SDValue(DAG.getMachineNode(Opc, dl, CN->getValueType(0), Base), 0); | |||
2768 | return true; | |||
2769 | } | |||
2770 | } | |||
2771 | ||||
2772 | Disp = DAG.getTargetConstant(0, dl, getPointerTy(DAG.getDataLayout())); | |||
2773 | if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N)) { | |||
2774 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
2775 | fixupFuncForFI(DAG, FI->getIndex(), N.getValueType()); | |||
2776 | } else | |||
2777 | Base = N; | |||
2778 | return true; // [r+0] | |||
2779 | } | |||
2780 | ||||
2781 | /// Similar to the 16-bit case but for instructions that take a 34-bit | |||
2782 | /// displacement field (prefixed loads/stores). | |||
2783 | bool PPCTargetLowering::SelectAddressRegImm34(SDValue N, SDValue &Disp, | |||
2784 | SDValue &Base, | |||
2785 | SelectionDAG &DAG) const { | |||
2786 | // Only on 64-bit targets. | |||
2787 | if (N.getValueType() != MVT::i64) | |||
2788 | return false; | |||
2789 | ||||
2790 | SDLoc dl(N); | |||
2791 | int64_t Imm = 0; | |||
2792 | ||||
2793 | if (N.getOpcode() == ISD::ADD) { | |||
2794 | if (!isIntS34Immediate(N.getOperand(1), Imm)) | |||
2795 | return false; | |||
2796 | Disp = DAG.getTargetConstant(Imm, dl, N.getValueType()); | |||
2797 | if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N.getOperand(0))) | |||
2798 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
2799 | else | |||
2800 | Base = N.getOperand(0); | |||
2801 | return true; | |||
2802 | } | |||
2803 | ||||
2804 | if (N.getOpcode() == ISD::OR) { | |||
2805 | if (!isIntS34Immediate(N.getOperand(1), Imm)) | |||
2806 | return false; | |||
2807 | // If this is an or of disjoint bitfields, we can codegen this as an add | |||
2808 | // (for better address arithmetic) if the LHS and RHS of the OR are | |||
2809 | // provably disjoint. | |||
2810 | KnownBits LHSKnown = DAG.computeKnownBits(N.getOperand(0)); | |||
2811 | if ((LHSKnown.Zero.getZExtValue() | ~(uint64_t)Imm) != ~0ULL) | |||
2812 | return false; | |||
2813 | if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N.getOperand(0))) | |||
2814 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
2815 | else | |||
2816 | Base = N.getOperand(0); | |||
2817 | Disp = DAG.getTargetConstant(Imm, dl, N.getValueType()); | |||
2818 | return true; | |||
2819 | } | |||
2820 | ||||
2821 | if (isIntS34Immediate(N, Imm)) { // If the address is a 34-bit const. | |||
2822 | Disp = DAG.getTargetConstant(Imm, dl, N.getValueType()); | |||
2823 | Base = DAG.getRegister(PPC::ZERO8, N.getValueType()); | |||
2824 | return true; | |||
2825 | } | |||
2826 | ||||
2827 | return false; | |||
2828 | } | |||
2829 | ||||
2830 | /// SelectAddressRegRegOnly - Given the specified addressed, force it to be | |||
2831 | /// represented as an indexed [r+r] operation. | |||
2832 | bool PPCTargetLowering::SelectAddressRegRegOnly(SDValue N, SDValue &Base, | |||
2833 | SDValue &Index, | |||
2834 | SelectionDAG &DAG) const { | |||
2835 | // Check to see if we can easily represent this as an [r+r] address. This | |||
2836 | // will fail if it thinks that the address is more profitably represented as | |||
2837 | // reg+imm, e.g. where imm = 0. | |||
2838 | if (SelectAddressRegReg(N, Base, Index, DAG)) | |||
2839 | return true; | |||
2840 | ||||
2841 | // If the address is the result of an add, we will utilize the fact that the | |||
2842 | // address calculation includes an implicit add. However, we can reduce | |||
2843 | // register pressure if we do not materialize a constant just for use as the | |||
2844 | // index register. We only get rid of the add if it is not an add of a | |||
2845 | // value and a 16-bit signed constant and both have a single use. | |||
2846 | int16_t imm = 0; | |||
2847 | if (N.getOpcode() == ISD::ADD && | |||
2848 | (!isIntS16Immediate(N.getOperand(1), imm) || | |||
2849 | !N.getOperand(1).hasOneUse() || !N.getOperand(0).hasOneUse())) { | |||
2850 | Base = N.getOperand(0); | |||
2851 | Index = N.getOperand(1); | |||
2852 | return true; | |||
2853 | } | |||
2854 | ||||
2855 | // Otherwise, do it the hard way, using R0 as the base register. | |||
2856 | Base = DAG.getRegister(Subtarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, | |||
2857 | N.getValueType()); | |||
2858 | Index = N; | |||
2859 | return true; | |||
2860 | } | |||
2861 | ||||
2862 | template <typename Ty> static bool isValidPCRelNode(SDValue N) { | |||
2863 | Ty *PCRelCand = dyn_cast<Ty>(N); | |||
2864 | return PCRelCand && (PCRelCand->getTargetFlags() & PPCII::MO_PCREL_FLAG); | |||
2865 | } | |||
2866 | ||||
2867 | /// Returns true if this address is a PC Relative address. | |||
2868 | /// PC Relative addresses are marked with the flag PPCII::MO_PCREL_FLAG | |||
2869 | /// or if the node opcode is PPCISD::MAT_PCREL_ADDR. | |||
2870 | bool PPCTargetLowering::SelectAddressPCRel(SDValue N, SDValue &Base) const { | |||
2871 | // This is a materialize PC Relative node. Always select this as PC Relative. | |||
2872 | Base = N; | |||
2873 | if (N.getOpcode() == PPCISD::MAT_PCREL_ADDR) | |||
2874 | return true; | |||
2875 | if (isValidPCRelNode<ConstantPoolSDNode>(N) || | |||
2876 | isValidPCRelNode<GlobalAddressSDNode>(N) || | |||
2877 | isValidPCRelNode<JumpTableSDNode>(N) || | |||
2878 | isValidPCRelNode<BlockAddressSDNode>(N)) | |||
2879 | return true; | |||
2880 | return false; | |||
2881 | } | |||
2882 | ||||
2883 | /// Returns true if we should use a direct load into vector instruction | |||
2884 | /// (such as lxsd or lfd), instead of a load into gpr + direct move sequence. | |||
2885 | static bool usePartialVectorLoads(SDNode *N, const PPCSubtarget& ST) { | |||
2886 | ||||
2887 | // If there are any other uses other than scalar to vector, then we should | |||
2888 | // keep it as a scalar load -> direct move pattern to prevent multiple | |||
2889 | // loads. | |||
2890 | LoadSDNode *LD = dyn_cast<LoadSDNode>(N); | |||
2891 | if (!LD) | |||
2892 | return false; | |||
2893 | ||||
2894 | EVT MemVT = LD->getMemoryVT(); | |||
2895 | if (!MemVT.isSimple()) | |||
2896 | return false; | |||
2897 | switch(MemVT.getSimpleVT().SimpleTy) { | |||
2898 | case MVT::i64: | |||
2899 | break; | |||
2900 | case MVT::i32: | |||
2901 | if (!ST.hasP8Vector()) | |||
2902 | return false; | |||
2903 | break; | |||
2904 | case MVT::i16: | |||
2905 | case MVT::i8: | |||
2906 | if (!ST.hasP9Vector()) | |||
2907 | return false; | |||
2908 | break; | |||
2909 | default: | |||
2910 | return false; | |||
2911 | } | |||
2912 | ||||
2913 | SDValue LoadedVal(N, 0); | |||
2914 | if (!LoadedVal.hasOneUse()) | |||
2915 | return false; | |||
2916 | ||||
2917 | for (SDNode::use_iterator UI = LD->use_begin(), UE = LD->use_end(); | |||
2918 | UI != UE; ++UI) | |||
2919 | if (UI.getUse().get().getResNo() == 0 && | |||
2920 | UI->getOpcode() != ISD::SCALAR_TO_VECTOR && | |||
2921 | UI->getOpcode() != PPCISD::SCALAR_TO_VECTOR_PERMUTED) | |||
2922 | return false; | |||
2923 | ||||
2924 | return true; | |||
2925 | } | |||
2926 | ||||
2927 | /// getPreIndexedAddressParts - returns true by value, base pointer and | |||
2928 | /// offset pointer and addressing mode by reference if the node's address | |||
2929 | /// can be legally represented as pre-indexed load / store address. | |||
2930 | bool PPCTargetLowering::getPreIndexedAddressParts(SDNode *N, SDValue &Base, | |||
2931 | SDValue &Offset, | |||
2932 | ISD::MemIndexedMode &AM, | |||
2933 | SelectionDAG &DAG) const { | |||
2934 | if (DisablePPCPreinc) return false; | |||
2935 | ||||
2936 | bool isLoad = true; | |||
2937 | SDValue Ptr; | |||
2938 | EVT VT; | |||
2939 | unsigned Alignment; | |||
2940 | if (LoadSDNode *LD = dyn_cast<LoadSDNode>(N)) { | |||
2941 | Ptr = LD->getBasePtr(); | |||
2942 | VT = LD->getMemoryVT(); | |||
2943 | Alignment = LD->getAlignment(); | |||
2944 | } else if (StoreSDNode *ST = dyn_cast<StoreSDNode>(N)) { | |||
2945 | Ptr = ST->getBasePtr(); | |||
2946 | VT = ST->getMemoryVT(); | |||
2947 | Alignment = ST->getAlignment(); | |||
2948 | isLoad = false; | |||
2949 | } else | |||
2950 | return false; | |||
2951 | ||||
2952 | // Do not generate pre-inc forms for specific loads that feed scalar_to_vector | |||
2953 | // instructions because we can fold these into a more efficient instruction | |||
2954 | // instead, (such as LXSD). | |||
2955 | if (isLoad && usePartialVectorLoads(N, Subtarget)) { | |||
2956 | return false; | |||
2957 | } | |||
2958 | ||||
2959 | // PowerPC doesn't have preinc load/store instructions for vectors | |||
2960 | if (VT.isVector()) | |||
2961 | return false; | |||
2962 | ||||
2963 | if (SelectAddressRegReg(Ptr, Base, Offset, DAG)) { | |||
2964 | // Common code will reject creating a pre-inc form if the base pointer | |||
2965 | // is a frame index, or if N is a store and the base pointer is either | |||
2966 | // the same as or a predecessor of the value being stored. Check for | |||
2967 | // those situations here, and try with swapped Base/Offset instead. | |||
2968 | bool Swap = false; | |||
2969 | ||||
2970 | if (isa<FrameIndexSDNode>(Base) || isa<RegisterSDNode>(Base)) | |||
2971 | Swap = true; | |||
2972 | else if (!isLoad) { | |||
2973 | SDValue Val = cast<StoreSDNode>(N)->getValue(); | |||
2974 | if (Val == Base || Base.getNode()->isPredecessorOf(Val.getNode())) | |||
2975 | Swap = true; | |||
2976 | } | |||
2977 | ||||
2978 | if (Swap) | |||
2979 | std::swap(Base, Offset); | |||
2980 | ||||
2981 | AM = ISD::PRE_INC; | |||
2982 | return true; | |||
2983 | } | |||
2984 | ||||
2985 | // LDU/STU can only handle immediates that are a multiple of 4. | |||
2986 | if (VT != MVT::i64) { | |||
2987 | if (!SelectAddressRegImm(Ptr, Offset, Base, DAG, None)) | |||
2988 | return false; | |||
2989 | } else { | |||
2990 | // LDU/STU need an address with at least 4-byte alignment. | |||
2991 | if (Alignment < 4) | |||
2992 | return false; | |||
2993 | ||||
2994 | if (!SelectAddressRegImm(Ptr, Offset, Base, DAG, Align(4))) | |||
2995 | return false; | |||
2996 | } | |||
2997 | ||||
2998 | if (LoadSDNode *LD = dyn_cast<LoadSDNode>(N)) { | |||
2999 | // PPC64 doesn't have lwau, but it does have lwaux. Reject preinc load of | |||
3000 | // sext i32 to i64 when addr mode is r+i. | |||
3001 | if (LD->getValueType(0) == MVT::i64 && LD->getMemoryVT() == MVT::i32 && | |||
3002 | LD->getExtensionType() == ISD::SEXTLOAD && | |||
3003 | isa<ConstantSDNode>(Offset)) | |||
3004 | return false; | |||
3005 | } | |||
3006 | ||||
3007 | AM = ISD::PRE_INC; | |||
3008 | return true; | |||
3009 | } | |||
3010 | ||||
3011 | //===----------------------------------------------------------------------===// | |||
3012 | // LowerOperation implementation | |||
3013 | //===----------------------------------------------------------------------===// | |||
3014 | ||||
3015 | /// Return true if we should reference labels using a PICBase, set the HiOpFlags | |||
3016 | /// and LoOpFlags to the target MO flags. | |||
3017 | static void getLabelAccessInfo(bool IsPIC, const PPCSubtarget &Subtarget, | |||
3018 | unsigned &HiOpFlags, unsigned &LoOpFlags, | |||
3019 | const GlobalValue *GV = nullptr) { | |||
3020 | HiOpFlags = PPCII::MO_HA; | |||
3021 | LoOpFlags = PPCII::MO_LO; | |||
3022 | ||||
3023 | // Don't use the pic base if not in PIC relocation model. | |||
3024 | if (IsPIC) { | |||
3025 | HiOpFlags |= PPCII::MO_PIC_FLAG; | |||
3026 | LoOpFlags |= PPCII::MO_PIC_FLAG; | |||
3027 | } | |||
3028 | } | |||
3029 | ||||
3030 | static SDValue LowerLabelRef(SDValue HiPart, SDValue LoPart, bool isPIC, | |||
3031 | SelectionDAG &DAG) { | |||
3032 | SDLoc DL(HiPart); | |||
3033 | EVT PtrVT = HiPart.getValueType(); | |||
3034 | SDValue Zero = DAG.getConstant(0, DL, PtrVT); | |||
3035 | ||||
3036 | SDValue Hi = DAG.getNode(PPCISD::Hi, DL, PtrVT, HiPart, Zero); | |||
3037 | SDValue Lo = DAG.getNode(PPCISD::Lo, DL, PtrVT, LoPart, Zero); | |||
3038 | ||||
3039 | // With PIC, the first instruction is actually "GR+hi(&G)". | |||
3040 | if (isPIC) | |||
3041 | Hi = DAG.getNode(ISD::ADD, DL, PtrVT, | |||
3042 | DAG.getNode(PPCISD::GlobalBaseReg, DL, PtrVT), Hi); | |||
3043 | ||||
3044 | // Generate non-pic code that has direct accesses to the constant pool. | |||
3045 | // The address of the global is just (hi(&g)+lo(&g)). | |||
3046 | return DAG.getNode(ISD::ADD, DL, PtrVT, Hi, Lo); | |||
3047 | } | |||
3048 | ||||
3049 | static void setUsesTOCBasePtr(MachineFunction &MF) { | |||
3050 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
3051 | FuncInfo->setUsesTOCBasePtr(); | |||
3052 | } | |||
3053 | ||||
3054 | static void setUsesTOCBasePtr(SelectionDAG &DAG) { | |||
3055 | setUsesTOCBasePtr(DAG.getMachineFunction()); | |||
3056 | } | |||
3057 | ||||
3058 | SDValue PPCTargetLowering::getTOCEntry(SelectionDAG &DAG, const SDLoc &dl, | |||
3059 | SDValue GA) const { | |||
3060 | const bool Is64Bit = Subtarget.isPPC64(); | |||
3061 | EVT VT = Is64Bit ? MVT::i64 : MVT::i32; | |||
3062 | SDValue Reg = Is64Bit ? DAG.getRegister(PPC::X2, VT) | |||
3063 | : Subtarget.isAIXABI() | |||
3064 | ? DAG.getRegister(PPC::R2, VT) | |||
3065 | : DAG.getNode(PPCISD::GlobalBaseReg, dl, VT); | |||
3066 | SDValue Ops[] = { GA, Reg }; | |||
3067 | return DAG.getMemIntrinsicNode( | |||
3068 | PPCISD::TOC_ENTRY, dl, DAG.getVTList(VT, MVT::Other), Ops, VT, | |||
3069 | MachinePointerInfo::getGOT(DAG.getMachineFunction()), None, | |||
3070 | MachineMemOperand::MOLoad); | |||
3071 | } | |||
3072 | ||||
3073 | SDValue PPCTargetLowering::LowerConstantPool(SDValue Op, | |||
3074 | SelectionDAG &DAG) const { | |||
3075 | EVT PtrVT = Op.getValueType(); | |||
3076 | ConstantPoolSDNode *CP = cast<ConstantPoolSDNode>(Op); | |||
3077 | const Constant *C = CP->getConstVal(); | |||
3078 | ||||
3079 | // 64-bit SVR4 ABI and AIX ABI code are always position-independent. | |||
3080 | // The actual address of the GlobalValue is stored in the TOC. | |||
3081 | if (Subtarget.is64BitELFABI() || Subtarget.isAIXABI()) { | |||
3082 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3083 | SDLoc DL(CP); | |||
3084 | EVT Ty = getPointerTy(DAG.getDataLayout()); | |||
3085 | SDValue ConstPool = DAG.getTargetConstantPool( | |||
3086 | C, Ty, CP->getAlign(), CP->getOffset(), PPCII::MO_PCREL_FLAG); | |||
3087 | return DAG.getNode(PPCISD::MAT_PCREL_ADDR, DL, Ty, ConstPool); | |||
3088 | } | |||
3089 | setUsesTOCBasePtr(DAG); | |||
3090 | SDValue GA = DAG.getTargetConstantPool(C, PtrVT, CP->getAlign(), 0); | |||
3091 | return getTOCEntry(DAG, SDLoc(CP), GA); | |||
3092 | } | |||
3093 | ||||
3094 | unsigned MOHiFlag, MOLoFlag; | |||
3095 | bool IsPIC = isPositionIndependent(); | |||
3096 | getLabelAccessInfo(IsPIC, Subtarget, MOHiFlag, MOLoFlag); | |||
3097 | ||||
3098 | if (IsPIC && Subtarget.isSVR4ABI()) { | |||
3099 | SDValue GA = | |||
3100 | DAG.getTargetConstantPool(C, PtrVT, CP->getAlign(), PPCII::MO_PIC_FLAG); | |||
3101 | return getTOCEntry(DAG, SDLoc(CP), GA); | |||
3102 | } | |||
3103 | ||||
3104 | SDValue CPIHi = | |||
3105 | DAG.getTargetConstantPool(C, PtrVT, CP->getAlign(), 0, MOHiFlag); | |||
3106 | SDValue CPILo = | |||
3107 | DAG.getTargetConstantPool(C, PtrVT, CP->getAlign(), 0, MOLoFlag); | |||
3108 | return LowerLabelRef(CPIHi, CPILo, IsPIC, DAG); | |||
3109 | } | |||
3110 | ||||
3111 | // For 64-bit PowerPC, prefer the more compact relative encodings. | |||
3112 | // This trades 32 bits per jump table entry for one or two instructions | |||
3113 | // on the jump site. | |||
3114 | unsigned PPCTargetLowering::getJumpTableEncoding() const { | |||
3115 | if (isJumpTableRelative()) | |||
3116 | return MachineJumpTableInfo::EK_LabelDifference32; | |||
3117 | ||||
3118 | return TargetLowering::getJumpTableEncoding(); | |||
3119 | } | |||
3120 | ||||
3121 | bool PPCTargetLowering::isJumpTableRelative() const { | |||
3122 | if (UseAbsoluteJumpTables) | |||
3123 | return false; | |||
3124 | if (Subtarget.isPPC64() || Subtarget.isAIXABI()) | |||
3125 | return true; | |||
3126 | return TargetLowering::isJumpTableRelative(); | |||
3127 | } | |||
3128 | ||||
3129 | SDValue PPCTargetLowering::getPICJumpTableRelocBase(SDValue Table, | |||
3130 | SelectionDAG &DAG) const { | |||
3131 | if (!Subtarget.isPPC64() || Subtarget.isAIXABI()) | |||
3132 | return TargetLowering::getPICJumpTableRelocBase(Table, DAG); | |||
3133 | ||||
3134 | switch (getTargetMachine().getCodeModel()) { | |||
3135 | case CodeModel::Small: | |||
3136 | case CodeModel::Medium: | |||
3137 | return TargetLowering::getPICJumpTableRelocBase(Table, DAG); | |||
3138 | default: | |||
3139 | return DAG.getNode(PPCISD::GlobalBaseReg, SDLoc(), | |||
3140 | getPointerTy(DAG.getDataLayout())); | |||
3141 | } | |||
3142 | } | |||
3143 | ||||
3144 | const MCExpr * | |||
3145 | PPCTargetLowering::getPICJumpTableRelocBaseExpr(const MachineFunction *MF, | |||
3146 | unsigned JTI, | |||
3147 | MCContext &Ctx) const { | |||
3148 | if (!Subtarget.isPPC64() || Subtarget.isAIXABI()) | |||
3149 | return TargetLowering::getPICJumpTableRelocBaseExpr(MF, JTI, Ctx); | |||
3150 | ||||
3151 | switch (getTargetMachine().getCodeModel()) { | |||
3152 | case CodeModel::Small: | |||
3153 | case CodeModel::Medium: | |||
3154 | return TargetLowering::getPICJumpTableRelocBaseExpr(MF, JTI, Ctx); | |||
3155 | default: | |||
3156 | return MCSymbolRefExpr::create(MF->getPICBaseSymbol(), Ctx); | |||
3157 | } | |||
3158 | } | |||
3159 | ||||
3160 | SDValue PPCTargetLowering::LowerJumpTable(SDValue Op, SelectionDAG &DAG) const { | |||
3161 | EVT PtrVT = Op.getValueType(); | |||
3162 | JumpTableSDNode *JT = cast<JumpTableSDNode>(Op); | |||
3163 | ||||
3164 | // isUsingPCRelativeCalls() returns true when PCRelative is enabled | |||
3165 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3166 | SDLoc DL(JT); | |||
3167 | EVT Ty = getPointerTy(DAG.getDataLayout()); | |||
3168 | SDValue GA = | |||
3169 | DAG.getTargetJumpTable(JT->getIndex(), Ty, PPCII::MO_PCREL_FLAG); | |||
3170 | SDValue MatAddr = DAG.getNode(PPCISD::MAT_PCREL_ADDR, DL, Ty, GA); | |||
3171 | return MatAddr; | |||
3172 | } | |||
3173 | ||||
3174 | // 64-bit SVR4 ABI and AIX ABI code are always position-independent. | |||
3175 | // The actual address of the GlobalValue is stored in the TOC. | |||
3176 | if (Subtarget.is64BitELFABI() || Subtarget.isAIXABI()) { | |||
3177 | setUsesTOCBasePtr(DAG); | |||
3178 | SDValue GA = DAG.getTargetJumpTable(JT->getIndex(), PtrVT); | |||
3179 | return getTOCEntry(DAG, SDLoc(JT), GA); | |||
3180 | } | |||
3181 | ||||
3182 | unsigned MOHiFlag, MOLoFlag; | |||
3183 | bool IsPIC = isPositionIndependent(); | |||
3184 | getLabelAccessInfo(IsPIC, Subtarget, MOHiFlag, MOLoFlag); | |||
3185 | ||||
3186 | if (IsPIC && Subtarget.isSVR4ABI()) { | |||
3187 | SDValue GA = DAG.getTargetJumpTable(JT->getIndex(), PtrVT, | |||
3188 | PPCII::MO_PIC_FLAG); | |||
3189 | return getTOCEntry(DAG, SDLoc(GA), GA); | |||
3190 | } | |||
3191 | ||||
3192 | SDValue JTIHi = DAG.getTargetJumpTable(JT->getIndex(), PtrVT, MOHiFlag); | |||
3193 | SDValue JTILo = DAG.getTargetJumpTable(JT->getIndex(), PtrVT, MOLoFlag); | |||
3194 | return LowerLabelRef(JTIHi, JTILo, IsPIC, DAG); | |||
3195 | } | |||
3196 | ||||
3197 | SDValue PPCTargetLowering::LowerBlockAddress(SDValue Op, | |||
3198 | SelectionDAG &DAG) const { | |||
3199 | EVT PtrVT = Op.getValueType(); | |||
3200 | BlockAddressSDNode *BASDN = cast<BlockAddressSDNode>(Op); | |||
3201 | const BlockAddress *BA = BASDN->getBlockAddress(); | |||
3202 | ||||
3203 | // isUsingPCRelativeCalls() returns true when PCRelative is enabled | |||
3204 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3205 | SDLoc DL(BASDN); | |||
3206 | EVT Ty = getPointerTy(DAG.getDataLayout()); | |||
3207 | SDValue GA = DAG.getTargetBlockAddress(BA, Ty, BASDN->getOffset(), | |||
3208 | PPCII::MO_PCREL_FLAG); | |||
3209 | SDValue MatAddr = DAG.getNode(PPCISD::MAT_PCREL_ADDR, DL, Ty, GA); | |||
3210 | return MatAddr; | |||
3211 | } | |||
3212 | ||||
3213 | // 64-bit SVR4 ABI and AIX ABI code are always position-independent. | |||
3214 | // The actual BlockAddress is stored in the TOC. | |||
3215 | if (Subtarget.is64BitELFABI() || Subtarget.isAIXABI()) { | |||
3216 | setUsesTOCBasePtr(DAG); | |||
3217 | SDValue GA = DAG.getTargetBlockAddress(BA, PtrVT, BASDN->getOffset()); | |||
3218 | return getTOCEntry(DAG, SDLoc(BASDN), GA); | |||
3219 | } | |||
3220 | ||||
3221 | // 32-bit position-independent ELF stores the BlockAddress in the .got. | |||
3222 | if (Subtarget.is32BitELFABI() && isPositionIndependent()) | |||
3223 | return getTOCEntry( | |||
3224 | DAG, SDLoc(BASDN), | |||
3225 | DAG.getTargetBlockAddress(BA, PtrVT, BASDN->getOffset())); | |||
3226 | ||||
3227 | unsigned MOHiFlag, MOLoFlag; | |||
3228 | bool IsPIC = isPositionIndependent(); | |||
3229 | getLabelAccessInfo(IsPIC, Subtarget, MOHiFlag, MOLoFlag); | |||
3230 | SDValue TgtBAHi = DAG.getTargetBlockAddress(BA, PtrVT, 0, MOHiFlag); | |||
3231 | SDValue TgtBALo = DAG.getTargetBlockAddress(BA, PtrVT, 0, MOLoFlag); | |||
3232 | return LowerLabelRef(TgtBAHi, TgtBALo, IsPIC, DAG); | |||
3233 | } | |||
3234 | ||||
3235 | SDValue PPCTargetLowering::LowerGlobalTLSAddress(SDValue Op, | |||
3236 | SelectionDAG &DAG) const { | |||
3237 | if (Subtarget.isAIXABI()) | |||
3238 | return LowerGlobalTLSAddressAIX(Op, DAG); | |||
3239 | ||||
3240 | return LowerGlobalTLSAddressLinux(Op, DAG); | |||
3241 | } | |||
3242 | ||||
3243 | SDValue PPCTargetLowering::LowerGlobalTLSAddressAIX(SDValue Op, | |||
3244 | SelectionDAG &DAG) const { | |||
3245 | GlobalAddressSDNode *GA = cast<GlobalAddressSDNode>(Op); | |||
3246 | ||||
3247 | if (DAG.getTarget().useEmulatedTLS()) | |||
3248 | report_fatal_error("Emulated TLS is not yet supported on AIX"); | |||
3249 | ||||
3250 | SDLoc dl(GA); | |||
3251 | const GlobalValue *GV = GA->getGlobal(); | |||
3252 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
3253 | ||||
3254 | // The general-dynamic model is the only access model supported for now, so | |||
3255 | // all the GlobalTLSAddress nodes are lowered with this model. | |||
3256 | // We need to generate two TOC entries, one for the variable offset, one for | |||
3257 | // the region handle. The global address for the TOC entry of the region | |||
3258 | // handle is created with the MO_TLSGDM_FLAG flag and the global address | |||
3259 | // for the TOC entry of the variable offset is created with MO_TLSGD_FLAG. | |||
3260 | SDValue VariableOffsetTGA = | |||
3261 | DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, PPCII::MO_TLSGD_FLAG); | |||
3262 | SDValue RegionHandleTGA = | |||
3263 | DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, PPCII::MO_TLSGDM_FLAG); | |||
3264 | SDValue VariableOffset = getTOCEntry(DAG, dl, VariableOffsetTGA); | |||
3265 | SDValue RegionHandle = getTOCEntry(DAG, dl, RegionHandleTGA); | |||
3266 | return DAG.getNode(PPCISD::TLSGD_AIX, dl, PtrVT, VariableOffset, | |||
3267 | RegionHandle); | |||
3268 | } | |||
3269 | ||||
3270 | SDValue PPCTargetLowering::LowerGlobalTLSAddressLinux(SDValue Op, | |||
3271 | SelectionDAG &DAG) const { | |||
3272 | // FIXME: TLS addresses currently use medium model code sequences, | |||
3273 | // which is the most useful form. Eventually support for small and | |||
3274 | // large models could be added if users need it, at the cost of | |||
3275 | // additional complexity. | |||
3276 | GlobalAddressSDNode *GA = cast<GlobalAddressSDNode>(Op); | |||
3277 | if (DAG.getTarget().useEmulatedTLS()) | |||
3278 | return LowerToTLSEmulatedModel(GA, DAG); | |||
3279 | ||||
3280 | SDLoc dl(GA); | |||
3281 | const GlobalValue *GV = GA->getGlobal(); | |||
3282 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
3283 | bool is64bit = Subtarget.isPPC64(); | |||
3284 | const Module *M = DAG.getMachineFunction().getFunction().getParent(); | |||
3285 | PICLevel::Level picLevel = M->getPICLevel(); | |||
3286 | ||||
3287 | const TargetMachine &TM = getTargetMachine(); | |||
3288 | TLSModel::Model Model = TM.getTLSModel(GV); | |||
3289 | ||||
3290 | if (Model == TLSModel::LocalExec) { | |||
3291 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3292 | SDValue TLSReg = DAG.getRegister(PPC::X13, MVT::i64); | |||
3293 | SDValue TGA = DAG.getTargetGlobalAddress( | |||
3294 | GV, dl, PtrVT, 0, (PPCII::MO_PCREL_FLAG | PPCII::MO_TPREL_FLAG)); | |||
3295 | SDValue MatAddr = | |||
3296 | DAG.getNode(PPCISD::TLS_LOCAL_EXEC_MAT_ADDR, dl, PtrVT, TGA); | |||
3297 | return DAG.getNode(PPCISD::ADD_TLS, dl, PtrVT, TLSReg, MatAddr); | |||
3298 | } | |||
3299 | ||||
3300 | SDValue TGAHi = DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, | |||
3301 | PPCII::MO_TPREL_HA); | |||
3302 | SDValue TGALo = DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, | |||
3303 | PPCII::MO_TPREL_LO); | |||
3304 | SDValue TLSReg = is64bit ? DAG.getRegister(PPC::X13, MVT::i64) | |||
3305 | : DAG.getRegister(PPC::R2, MVT::i32); | |||
3306 | ||||
3307 | SDValue Hi = DAG.getNode(PPCISD::Hi, dl, PtrVT, TGAHi, TLSReg); | |||
3308 | return DAG.getNode(PPCISD::Lo, dl, PtrVT, TGALo, Hi); | |||
3309 | } | |||
3310 | ||||
3311 | if (Model == TLSModel::InitialExec) { | |||
3312 | bool IsPCRel = Subtarget.isUsingPCRelativeCalls(); | |||
3313 | SDValue TGA = DAG.getTargetGlobalAddress( | |||
3314 | GV, dl, PtrVT, 0, IsPCRel ? PPCII::MO_GOT_TPREL_PCREL_FLAG : 0); | |||
3315 | SDValue TGATLS = DAG.getTargetGlobalAddress( | |||
3316 | GV, dl, PtrVT, 0, | |||
3317 | IsPCRel ? (PPCII::MO_TLS | PPCII::MO_PCREL_FLAG) : PPCII::MO_TLS); | |||
3318 | SDValue TPOffset; | |||
3319 | if (IsPCRel) { | |||
3320 | SDValue MatPCRel = DAG.getNode(PPCISD::MAT_PCREL_ADDR, dl, PtrVT, TGA); | |||
3321 | TPOffset = DAG.getLoad(MVT::i64, dl, DAG.getEntryNode(), MatPCRel, | |||
3322 | MachinePointerInfo()); | |||
3323 | } else { | |||
3324 | SDValue GOTPtr; | |||
3325 | if (is64bit) { | |||
3326 | setUsesTOCBasePtr(DAG); | |||
3327 | SDValue GOTReg = DAG.getRegister(PPC::X2, MVT::i64); | |||
3328 | GOTPtr = | |||
3329 | DAG.getNode(PPCISD::ADDIS_GOT_TPREL_HA, dl, PtrVT, GOTReg, TGA); | |||
3330 | } else { | |||
3331 | if (!TM.isPositionIndependent()) | |||
3332 | GOTPtr = DAG.getNode(PPCISD::PPC32_GOT, dl, PtrVT); | |||
3333 | else if (picLevel == PICLevel::SmallPIC) | |||
3334 | GOTPtr = DAG.getNode(PPCISD::GlobalBaseReg, dl, PtrVT); | |||
3335 | else | |||
3336 | GOTPtr = DAG.getNode(PPCISD::PPC32_PICGOT, dl, PtrVT); | |||
3337 | } | |||
3338 | TPOffset = DAG.getNode(PPCISD::LD_GOT_TPREL_L, dl, PtrVT, TGA, GOTPtr); | |||
3339 | } | |||
3340 | return DAG.getNode(PPCISD::ADD_TLS, dl, PtrVT, TPOffset, TGATLS); | |||
3341 | } | |||
3342 | ||||
3343 | if (Model == TLSModel::GeneralDynamic) { | |||
3344 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3345 | SDValue TGA = DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, | |||
3346 | PPCII::MO_GOT_TLSGD_PCREL_FLAG); | |||
3347 | return DAG.getNode(PPCISD::TLS_DYNAMIC_MAT_PCREL_ADDR, dl, PtrVT, TGA); | |||
3348 | } | |||
3349 | ||||
3350 | SDValue TGA = DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, 0); | |||
3351 | SDValue GOTPtr; | |||
3352 | if (is64bit) { | |||
3353 | setUsesTOCBasePtr(DAG); | |||
3354 | SDValue GOTReg = DAG.getRegister(PPC::X2, MVT::i64); | |||
3355 | GOTPtr = DAG.getNode(PPCISD::ADDIS_TLSGD_HA, dl, PtrVT, | |||
3356 | GOTReg, TGA); | |||
3357 | } else { | |||
3358 | if (picLevel == PICLevel::SmallPIC) | |||
3359 | GOTPtr = DAG.getNode(PPCISD::GlobalBaseReg, dl, PtrVT); | |||
3360 | else | |||
3361 | GOTPtr = DAG.getNode(PPCISD::PPC32_PICGOT, dl, PtrVT); | |||
3362 | } | |||
3363 | return DAG.getNode(PPCISD::ADDI_TLSGD_L_ADDR, dl, PtrVT, | |||
3364 | GOTPtr, TGA, TGA); | |||
3365 | } | |||
3366 | ||||
3367 | if (Model == TLSModel::LocalDynamic) { | |||
3368 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3369 | SDValue TGA = DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, | |||
3370 | PPCII::MO_GOT_TLSLD_PCREL_FLAG); | |||
3371 | SDValue MatPCRel = | |||
3372 | DAG.getNode(PPCISD::TLS_DYNAMIC_MAT_PCREL_ADDR, dl, PtrVT, TGA); | |||
3373 | return DAG.getNode(PPCISD::PADDI_DTPREL, dl, PtrVT, MatPCRel, TGA); | |||
3374 | } | |||
3375 | ||||
3376 | SDValue TGA = DAG.getTargetGlobalAddress(GV, dl, PtrVT, 0, 0); | |||
3377 | SDValue GOTPtr; | |||
3378 | if (is64bit) { | |||
3379 | setUsesTOCBasePtr(DAG); | |||
3380 | SDValue GOTReg = DAG.getRegister(PPC::X2, MVT::i64); | |||
3381 | GOTPtr = DAG.getNode(PPCISD::ADDIS_TLSLD_HA, dl, PtrVT, | |||
3382 | GOTReg, TGA); | |||
3383 | } else { | |||
3384 | if (picLevel == PICLevel::SmallPIC) | |||
3385 | GOTPtr = DAG.getNode(PPCISD::GlobalBaseReg, dl, PtrVT); | |||
3386 | else | |||
3387 | GOTPtr = DAG.getNode(PPCISD::PPC32_PICGOT, dl, PtrVT); | |||
3388 | } | |||
3389 | SDValue TLSAddr = DAG.getNode(PPCISD::ADDI_TLSLD_L_ADDR, dl, | |||
3390 | PtrVT, GOTPtr, TGA, TGA); | |||
3391 | SDValue DtvOffsetHi = DAG.getNode(PPCISD::ADDIS_DTPREL_HA, dl, | |||
3392 | PtrVT, TLSAddr, TGA); | |||
3393 | return DAG.getNode(PPCISD::ADDI_DTPREL_L, dl, PtrVT, DtvOffsetHi, TGA); | |||
3394 | } | |||
3395 | ||||
3396 | llvm_unreachable("Unknown TLS model!")::llvm::llvm_unreachable_internal("Unknown TLS model!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3396); | |||
3397 | } | |||
3398 | ||||
3399 | SDValue PPCTargetLowering::LowerGlobalAddress(SDValue Op, | |||
3400 | SelectionDAG &DAG) const { | |||
3401 | EVT PtrVT = Op.getValueType(); | |||
3402 | GlobalAddressSDNode *GSDN = cast<GlobalAddressSDNode>(Op); | |||
3403 | SDLoc DL(GSDN); | |||
3404 | const GlobalValue *GV = GSDN->getGlobal(); | |||
3405 | ||||
3406 | // 64-bit SVR4 ABI & AIX ABI code is always position-independent. | |||
3407 | // The actual address of the GlobalValue is stored in the TOC. | |||
3408 | if (Subtarget.is64BitELFABI() || Subtarget.isAIXABI()) { | |||
3409 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
3410 | EVT Ty = getPointerTy(DAG.getDataLayout()); | |||
3411 | if (isAccessedAsGotIndirect(Op)) { | |||
3412 | SDValue GA = DAG.getTargetGlobalAddress(GV, DL, Ty, GSDN->getOffset(), | |||
3413 | PPCII::MO_PCREL_FLAG | | |||
3414 | PPCII::MO_GOT_FLAG); | |||
3415 | SDValue MatPCRel = DAG.getNode(PPCISD::MAT_PCREL_ADDR, DL, Ty, GA); | |||
3416 | SDValue Load = DAG.getLoad(MVT::i64, DL, DAG.getEntryNode(), MatPCRel, | |||
3417 | MachinePointerInfo()); | |||
3418 | return Load; | |||
3419 | } else { | |||
3420 | SDValue GA = DAG.getTargetGlobalAddress(GV, DL, Ty, GSDN->getOffset(), | |||
3421 | PPCII::MO_PCREL_FLAG); | |||
3422 | return DAG.getNode(PPCISD::MAT_PCREL_ADDR, DL, Ty, GA); | |||
3423 | } | |||
3424 | } | |||
3425 | setUsesTOCBasePtr(DAG); | |||
3426 | SDValue GA = DAG.getTargetGlobalAddress(GV, DL, PtrVT, GSDN->getOffset()); | |||
3427 | return getTOCEntry(DAG, DL, GA); | |||
3428 | } | |||
3429 | ||||
3430 | unsigned MOHiFlag, MOLoFlag; | |||
3431 | bool IsPIC = isPositionIndependent(); | |||
3432 | getLabelAccessInfo(IsPIC, Subtarget, MOHiFlag, MOLoFlag, GV); | |||
3433 | ||||
3434 | if (IsPIC && Subtarget.isSVR4ABI()) { | |||
3435 | SDValue GA = DAG.getTargetGlobalAddress(GV, DL, PtrVT, | |||
3436 | GSDN->getOffset(), | |||
3437 | PPCII::MO_PIC_FLAG); | |||
3438 | return getTOCEntry(DAG, DL, GA); | |||
3439 | } | |||
3440 | ||||
3441 | SDValue GAHi = | |||
3442 | DAG.getTargetGlobalAddress(GV, DL, PtrVT, GSDN->getOffset(), MOHiFlag); | |||
3443 | SDValue GALo = | |||
3444 | DAG.getTargetGlobalAddress(GV, DL, PtrVT, GSDN->getOffset(), MOLoFlag); | |||
3445 | ||||
3446 | return LowerLabelRef(GAHi, GALo, IsPIC, DAG); | |||
3447 | } | |||
3448 | ||||
3449 | SDValue PPCTargetLowering::LowerSETCC(SDValue Op, SelectionDAG &DAG) const { | |||
3450 | bool IsStrict = Op->isStrictFPOpcode(); | |||
3451 | ISD::CondCode CC = | |||
3452 | cast<CondCodeSDNode>(Op.getOperand(IsStrict ? 3 : 2))->get(); | |||
3453 | SDValue LHS = Op.getOperand(IsStrict ? 1 : 0); | |||
3454 | SDValue RHS = Op.getOperand(IsStrict ? 2 : 1); | |||
3455 | SDValue Chain = IsStrict ? Op.getOperand(0) : SDValue(); | |||
3456 | EVT LHSVT = LHS.getValueType(); | |||
3457 | SDLoc dl(Op); | |||
3458 | ||||
3459 | // Soften the setcc with libcall if it is fp128. | |||
3460 | if (LHSVT == MVT::f128) { | |||
3461 | assert(!Subtarget.hasP9Vector() &&(static_cast <bool> (!Subtarget.hasP9Vector() && "SETCC for f128 is already legal under Power9!") ? void (0) : __assert_fail ("!Subtarget.hasP9Vector() && \"SETCC for f128 is already legal under Power9!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3462, __extension__ __PRETTY_FUNCTION__)) | |||
3462 | "SETCC for f128 is already legal under Power9!")(static_cast <bool> (!Subtarget.hasP9Vector() && "SETCC for f128 is already legal under Power9!") ? void (0) : __assert_fail ("!Subtarget.hasP9Vector() && \"SETCC for f128 is already legal under Power9!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3462, __extension__ __PRETTY_FUNCTION__)); | |||
3463 | softenSetCCOperands(DAG, LHSVT, LHS, RHS, CC, dl, LHS, RHS, Chain, | |||
3464 | Op->getOpcode() == ISD::STRICT_FSETCCS); | |||
3465 | if (RHS.getNode()) | |||
3466 | LHS = DAG.getNode(ISD::SETCC, dl, Op.getValueType(), LHS, RHS, | |||
3467 | DAG.getCondCode(CC)); | |||
3468 | if (IsStrict) | |||
3469 | return DAG.getMergeValues({LHS, Chain}, dl); | |||
3470 | return LHS; | |||
3471 | } | |||
3472 | ||||
3473 | assert(!IsStrict && "Don't know how to handle STRICT_FSETCC!")(static_cast <bool> (!IsStrict && "Don't know how to handle STRICT_FSETCC!" ) ? void (0) : __assert_fail ("!IsStrict && \"Don't know how to handle STRICT_FSETCC!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3473, __extension__ __PRETTY_FUNCTION__)); | |||
3474 | ||||
3475 | if (Op.getValueType() == MVT::v2i64) { | |||
3476 | // When the operands themselves are v2i64 values, we need to do something | |||
3477 | // special because VSX has no underlying comparison operations for these. | |||
3478 | if (LHS.getValueType() == MVT::v2i64) { | |||
3479 | // Equality can be handled by casting to the legal type for Altivec | |||
3480 | // comparisons, everything else needs to be expanded. | |||
3481 | if (CC == ISD::SETEQ || CC == ISD::SETNE) { | |||
3482 | return DAG.getNode( | |||
3483 | ISD::BITCAST, dl, MVT::v2i64, | |||
3484 | DAG.getSetCC(dl, MVT::v4i32, | |||
3485 | DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, LHS), | |||
3486 | DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, RHS), CC)); | |||
3487 | } | |||
3488 | ||||
3489 | return SDValue(); | |||
3490 | } | |||
3491 | ||||
3492 | // We handle most of these in the usual way. | |||
3493 | return Op; | |||
3494 | } | |||
3495 | ||||
3496 | // If we're comparing for equality to zero, expose the fact that this is | |||
3497 | // implemented as a ctlz/srl pair on ppc, so that the dag combiner can | |||
3498 | // fold the new nodes. | |||
3499 | if (SDValue V = lowerCmpEqZeroToCtlzSrl(Op, DAG)) | |||
3500 | return V; | |||
3501 | ||||
3502 | if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(RHS)) { | |||
3503 | // Leave comparisons against 0 and -1 alone for now, since they're usually | |||
3504 | // optimized. FIXME: revisit this when we can custom lower all setcc | |||
3505 | // optimizations. | |||
3506 | if (C->isAllOnesValue() || C->isNullValue()) | |||
3507 | return SDValue(); | |||
3508 | } | |||
3509 | ||||
3510 | // If we have an integer seteq/setne, turn it into a compare against zero | |||
3511 | // by xor'ing the rhs with the lhs, which is faster than setting a | |||
3512 | // condition register, reading it back out, and masking the correct bit. The | |||
3513 | // normal approach here uses sub to do this instead of xor. Using xor exposes | |||
3514 | // the result to other bit-twiddling opportunities. | |||
3515 | if (LHSVT.isInteger() && (CC == ISD::SETEQ || CC == ISD::SETNE)) { | |||
3516 | EVT VT = Op.getValueType(); | |||
3517 | SDValue Sub = DAG.getNode(ISD::XOR, dl, LHSVT, LHS, RHS); | |||
3518 | return DAG.getSetCC(dl, VT, Sub, DAG.getConstant(0, dl, LHSVT), CC); | |||
3519 | } | |||
3520 | return SDValue(); | |||
3521 | } | |||
3522 | ||||
3523 | SDValue PPCTargetLowering::LowerVAARG(SDValue Op, SelectionDAG &DAG) const { | |||
3524 | SDNode *Node = Op.getNode(); | |||
3525 | EVT VT = Node->getValueType(0); | |||
3526 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
3527 | SDValue InChain = Node->getOperand(0); | |||
3528 | SDValue VAListPtr = Node->getOperand(1); | |||
3529 | const Value *SV = cast<SrcValueSDNode>(Node->getOperand(2))->getValue(); | |||
3530 | SDLoc dl(Node); | |||
3531 | ||||
3532 | assert(!Subtarget.isPPC64() && "LowerVAARG is PPC32 only")(static_cast <bool> (!Subtarget.isPPC64() && "LowerVAARG is PPC32 only" ) ? void (0) : __assert_fail ("!Subtarget.isPPC64() && \"LowerVAARG is PPC32 only\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3532, __extension__ __PRETTY_FUNCTION__)); | |||
3533 | ||||
3534 | // gpr_index | |||
3535 | SDValue GprIndex = DAG.getExtLoad(ISD::ZEXTLOAD, dl, MVT::i32, InChain, | |||
3536 | VAListPtr, MachinePointerInfo(SV), MVT::i8); | |||
3537 | InChain = GprIndex.getValue(1); | |||
3538 | ||||
3539 | if (VT == MVT::i64) { | |||
3540 | // Check if GprIndex is even | |||
3541 | SDValue GprAnd = DAG.getNode(ISD::AND, dl, MVT::i32, GprIndex, | |||
3542 | DAG.getConstant(1, dl, MVT::i32)); | |||
3543 | SDValue CC64 = DAG.getSetCC(dl, MVT::i32, GprAnd, | |||
3544 | DAG.getConstant(0, dl, MVT::i32), ISD::SETNE); | |||
3545 | SDValue GprIndexPlusOne = DAG.getNode(ISD::ADD, dl, MVT::i32, GprIndex, | |||
3546 | DAG.getConstant(1, dl, MVT::i32)); | |||
3547 | // Align GprIndex to be even if it isn't | |||
3548 | GprIndex = DAG.getNode(ISD::SELECT, dl, MVT::i32, CC64, GprIndexPlusOne, | |||
3549 | GprIndex); | |||
3550 | } | |||
3551 | ||||
3552 | // fpr index is 1 byte after gpr | |||
3553 | SDValue FprPtr = DAG.getNode(ISD::ADD, dl, PtrVT, VAListPtr, | |||
3554 | DAG.getConstant(1, dl, MVT::i32)); | |||
3555 | ||||
3556 | // fpr | |||
3557 | SDValue FprIndex = DAG.getExtLoad(ISD::ZEXTLOAD, dl, MVT::i32, InChain, | |||
3558 | FprPtr, MachinePointerInfo(SV), MVT::i8); | |||
3559 | InChain = FprIndex.getValue(1); | |||
3560 | ||||
3561 | SDValue RegSaveAreaPtr = DAG.getNode(ISD::ADD, dl, PtrVT, VAListPtr, | |||
3562 | DAG.getConstant(8, dl, MVT::i32)); | |||
3563 | ||||
3564 | SDValue OverflowAreaPtr = DAG.getNode(ISD::ADD, dl, PtrVT, VAListPtr, | |||
3565 | DAG.getConstant(4, dl, MVT::i32)); | |||
3566 | ||||
3567 | // areas | |||
3568 | SDValue OverflowArea = | |||
3569 | DAG.getLoad(MVT::i32, dl, InChain, OverflowAreaPtr, MachinePointerInfo()); | |||
3570 | InChain = OverflowArea.getValue(1); | |||
3571 | ||||
3572 | SDValue RegSaveArea = | |||
3573 | DAG.getLoad(MVT::i32, dl, InChain, RegSaveAreaPtr, MachinePointerInfo()); | |||
3574 | InChain = RegSaveArea.getValue(1); | |||
3575 | ||||
3576 | // select overflow_area if index > 8 | |||
3577 | SDValue CC = DAG.getSetCC(dl, MVT::i32, VT.isInteger() ? GprIndex : FprIndex, | |||
3578 | DAG.getConstant(8, dl, MVT::i32), ISD::SETLT); | |||
3579 | ||||
3580 | // adjustment constant gpr_index * 4/8 | |||
3581 | SDValue RegConstant = DAG.getNode(ISD::MUL, dl, MVT::i32, | |||
3582 | VT.isInteger() ? GprIndex : FprIndex, | |||
3583 | DAG.getConstant(VT.isInteger() ? 4 : 8, dl, | |||
3584 | MVT::i32)); | |||
3585 | ||||
3586 | // OurReg = RegSaveArea + RegConstant | |||
3587 | SDValue OurReg = DAG.getNode(ISD::ADD, dl, PtrVT, RegSaveArea, | |||
3588 | RegConstant); | |||
3589 | ||||
3590 | // Floating types are 32 bytes into RegSaveArea | |||
3591 | if (VT.isFloatingPoint()) | |||
3592 | OurReg = DAG.getNode(ISD::ADD, dl, PtrVT, OurReg, | |||
3593 | DAG.getConstant(32, dl, MVT::i32)); | |||
3594 | ||||
3595 | // increase {f,g}pr_index by 1 (or 2 if VT is i64) | |||
3596 | SDValue IndexPlus1 = DAG.getNode(ISD::ADD, dl, MVT::i32, | |||
3597 | VT.isInteger() ? GprIndex : FprIndex, | |||
3598 | DAG.getConstant(VT == MVT::i64 ? 2 : 1, dl, | |||
3599 | MVT::i32)); | |||
3600 | ||||
3601 | InChain = DAG.getTruncStore(InChain, dl, IndexPlus1, | |||
3602 | VT.isInteger() ? VAListPtr : FprPtr, | |||
3603 | MachinePointerInfo(SV), MVT::i8); | |||
3604 | ||||
3605 | // determine if we should load from reg_save_area or overflow_area | |||
3606 | SDValue Result = DAG.getNode(ISD::SELECT, dl, PtrVT, CC, OurReg, OverflowArea); | |||
3607 | ||||
3608 | // increase overflow_area by 4/8 if gpr/fpr > 8 | |||
3609 | SDValue OverflowAreaPlusN = DAG.getNode(ISD::ADD, dl, PtrVT, OverflowArea, | |||
3610 | DAG.getConstant(VT.isInteger() ? 4 : 8, | |||
3611 | dl, MVT::i32)); | |||
3612 | ||||
3613 | OverflowArea = DAG.getNode(ISD::SELECT, dl, MVT::i32, CC, OverflowArea, | |||
3614 | OverflowAreaPlusN); | |||
3615 | ||||
3616 | InChain = DAG.getTruncStore(InChain, dl, OverflowArea, OverflowAreaPtr, | |||
3617 | MachinePointerInfo(), MVT::i32); | |||
3618 | ||||
3619 | return DAG.getLoad(VT, dl, InChain, Result, MachinePointerInfo()); | |||
3620 | } | |||
3621 | ||||
3622 | SDValue PPCTargetLowering::LowerVACOPY(SDValue Op, SelectionDAG &DAG) const { | |||
3623 | assert(!Subtarget.isPPC64() && "LowerVACOPY is PPC32 only")(static_cast <bool> (!Subtarget.isPPC64() && "LowerVACOPY is PPC32 only" ) ? void (0) : __assert_fail ("!Subtarget.isPPC64() && \"LowerVACOPY is PPC32 only\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3623, __extension__ __PRETTY_FUNCTION__)); | |||
3624 | ||||
3625 | // We have to copy the entire va_list struct: | |||
3626 | // 2*sizeof(char) + 2 Byte alignment + 2*sizeof(char*) = 12 Byte | |||
3627 | return DAG.getMemcpy(Op.getOperand(0), Op, Op.getOperand(1), Op.getOperand(2), | |||
3628 | DAG.getConstant(12, SDLoc(Op), MVT::i32), Align(8), | |||
3629 | false, true, false, MachinePointerInfo(), | |||
3630 | MachinePointerInfo()); | |||
3631 | } | |||
3632 | ||||
3633 | SDValue PPCTargetLowering::LowerADJUST_TRAMPOLINE(SDValue Op, | |||
3634 | SelectionDAG &DAG) const { | |||
3635 | if (Subtarget.isAIXABI()) | |||
3636 | report_fatal_error("ADJUST_TRAMPOLINE operation is not supported on AIX."); | |||
3637 | ||||
3638 | return Op.getOperand(0); | |||
3639 | } | |||
3640 | ||||
3641 | SDValue PPCTargetLowering::LowerINLINEASM(SDValue Op, SelectionDAG &DAG) const { | |||
3642 | MachineFunction &MF = DAG.getMachineFunction(); | |||
3643 | PPCFunctionInfo &MFI = *MF.getInfo<PPCFunctionInfo>(); | |||
3644 | ||||
3645 | assert((Op.getOpcode() == ISD::INLINEASM ||(static_cast <bool> ((Op.getOpcode() == ISD::INLINEASM || Op.getOpcode() == ISD::INLINEASM_BR) && "Expecting Inline ASM node." ) ? void (0) : __assert_fail ("(Op.getOpcode() == ISD::INLINEASM || Op.getOpcode() == ISD::INLINEASM_BR) && \"Expecting Inline ASM node.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3647, __extension__ __PRETTY_FUNCTION__)) | |||
3646 | Op.getOpcode() == ISD::INLINEASM_BR) &&(static_cast <bool> ((Op.getOpcode() == ISD::INLINEASM || Op.getOpcode() == ISD::INLINEASM_BR) && "Expecting Inline ASM node." ) ? void (0) : __assert_fail ("(Op.getOpcode() == ISD::INLINEASM || Op.getOpcode() == ISD::INLINEASM_BR) && \"Expecting Inline ASM node.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3647, __extension__ __PRETTY_FUNCTION__)) | |||
3647 | "Expecting Inline ASM node.")(static_cast <bool> ((Op.getOpcode() == ISD::INLINEASM || Op.getOpcode() == ISD::INLINEASM_BR) && "Expecting Inline ASM node." ) ? void (0) : __assert_fail ("(Op.getOpcode() == ISD::INLINEASM || Op.getOpcode() == ISD::INLINEASM_BR) && \"Expecting Inline ASM node.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3647, __extension__ __PRETTY_FUNCTION__)); | |||
3648 | ||||
3649 | // If an LR store is already known to be required then there is not point in | |||
3650 | // checking this ASM as well. | |||
3651 | if (MFI.isLRStoreRequired()) | |||
3652 | return Op; | |||
3653 | ||||
3654 | // Inline ASM nodes have an optional last operand that is an incoming Flag of | |||
3655 | // type MVT::Glue. We want to ignore this last operand if that is the case. | |||
3656 | unsigned NumOps = Op.getNumOperands(); | |||
3657 | if (Op.getOperand(NumOps - 1).getValueType() == MVT::Glue) | |||
3658 | --NumOps; | |||
3659 | ||||
3660 | // Check all operands that may contain the LR. | |||
3661 | for (unsigned i = InlineAsm::Op_FirstOperand; i != NumOps;) { | |||
3662 | unsigned Flags = cast<ConstantSDNode>(Op.getOperand(i))->getZExtValue(); | |||
3663 | unsigned NumVals = InlineAsm::getNumOperandRegisters(Flags); | |||
3664 | ++i; // Skip the ID value. | |||
3665 | ||||
3666 | switch (InlineAsm::getKind(Flags)) { | |||
3667 | default: | |||
3668 | llvm_unreachable("Bad flags!")::llvm::llvm_unreachable_internal("Bad flags!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3668); | |||
3669 | case InlineAsm::Kind_RegUse: | |||
3670 | case InlineAsm::Kind_Imm: | |||
3671 | case InlineAsm::Kind_Mem: | |||
3672 | i += NumVals; | |||
3673 | break; | |||
3674 | case InlineAsm::Kind_Clobber: | |||
3675 | case InlineAsm::Kind_RegDef: | |||
3676 | case InlineAsm::Kind_RegDefEarlyClobber: { | |||
3677 | for (; NumVals; --NumVals, ++i) { | |||
3678 | Register Reg = cast<RegisterSDNode>(Op.getOperand(i))->getReg(); | |||
3679 | if (Reg != PPC::LR && Reg != PPC::LR8) | |||
3680 | continue; | |||
3681 | MFI.setLRStoreRequired(); | |||
3682 | return Op; | |||
3683 | } | |||
3684 | break; | |||
3685 | } | |||
3686 | } | |||
3687 | } | |||
3688 | ||||
3689 | return Op; | |||
3690 | } | |||
3691 | ||||
3692 | SDValue PPCTargetLowering::LowerINIT_TRAMPOLINE(SDValue Op, | |||
3693 | SelectionDAG &DAG) const { | |||
3694 | if (Subtarget.isAIXABI()) | |||
3695 | report_fatal_error("INIT_TRAMPOLINE operation is not supported on AIX."); | |||
3696 | ||||
3697 | SDValue Chain = Op.getOperand(0); | |||
3698 | SDValue Trmp = Op.getOperand(1); // trampoline | |||
3699 | SDValue FPtr = Op.getOperand(2); // nested function | |||
3700 | SDValue Nest = Op.getOperand(3); // 'nest' parameter value | |||
3701 | SDLoc dl(Op); | |||
3702 | ||||
3703 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
3704 | bool isPPC64 = (PtrVT == MVT::i64); | |||
3705 | Type *IntPtrTy = DAG.getDataLayout().getIntPtrType(*DAG.getContext()); | |||
3706 | ||||
3707 | TargetLowering::ArgListTy Args; | |||
3708 | TargetLowering::ArgListEntry Entry; | |||
3709 | ||||
3710 | Entry.Ty = IntPtrTy; | |||
3711 | Entry.Node = Trmp; Args.push_back(Entry); | |||
3712 | ||||
3713 | // TrampSize == (isPPC64 ? 48 : 40); | |||
3714 | Entry.Node = DAG.getConstant(isPPC64 ? 48 : 40, dl, | |||
3715 | isPPC64 ? MVT::i64 : MVT::i32); | |||
3716 | Args.push_back(Entry); | |||
3717 | ||||
3718 | Entry.Node = FPtr; Args.push_back(Entry); | |||
3719 | Entry.Node = Nest; Args.push_back(Entry); | |||
3720 | ||||
3721 | // Lower to a call to __trampoline_setup(Trmp, TrampSize, FPtr, ctx_reg) | |||
3722 | TargetLowering::CallLoweringInfo CLI(DAG); | |||
3723 | CLI.setDebugLoc(dl).setChain(Chain).setLibCallee( | |||
3724 | CallingConv::C, Type::getVoidTy(*DAG.getContext()), | |||
3725 | DAG.getExternalSymbol("__trampoline_setup", PtrVT), std::move(Args)); | |||
3726 | ||||
3727 | std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI); | |||
3728 | return CallResult.second; | |||
3729 | } | |||
3730 | ||||
3731 | SDValue PPCTargetLowering::LowerVASTART(SDValue Op, SelectionDAG &DAG) const { | |||
3732 | MachineFunction &MF = DAG.getMachineFunction(); | |||
3733 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
3734 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
3735 | ||||
3736 | SDLoc dl(Op); | |||
3737 | ||||
3738 | if (Subtarget.isPPC64() || Subtarget.isAIXABI()) { | |||
3739 | // vastart just stores the address of the VarArgsFrameIndex slot into the | |||
3740 | // memory location argument. | |||
3741 | SDValue FR = DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT); | |||
3742 | const Value *SV = cast<SrcValueSDNode>(Op.getOperand(2))->getValue(); | |||
3743 | return DAG.getStore(Op.getOperand(0), dl, FR, Op.getOperand(1), | |||
3744 | MachinePointerInfo(SV)); | |||
3745 | } | |||
3746 | ||||
3747 | // For the 32-bit SVR4 ABI we follow the layout of the va_list struct. | |||
3748 | // We suppose the given va_list is already allocated. | |||
3749 | // | |||
3750 | // typedef struct { | |||
3751 | // char gpr; /* index into the array of 8 GPRs | |||
3752 | // * stored in the register save area | |||
3753 | // * gpr=0 corresponds to r3, | |||
3754 | // * gpr=1 to r4, etc. | |||
3755 | // */ | |||
3756 | // char fpr; /* index into the array of 8 FPRs | |||
3757 | // * stored in the register save area | |||
3758 | // * fpr=0 corresponds to f1, | |||
3759 | // * fpr=1 to f2, etc. | |||
3760 | // */ | |||
3761 | // char *overflow_arg_area; | |||
3762 | // /* location on stack that holds | |||
3763 | // * the next overflow argument | |||
3764 | // */ | |||
3765 | // char *reg_save_area; | |||
3766 | // /* where r3:r10 and f1:f8 (if saved) | |||
3767 | // * are stored | |||
3768 | // */ | |||
3769 | // } va_list[1]; | |||
3770 | ||||
3771 | SDValue ArgGPR = DAG.getConstant(FuncInfo->getVarArgsNumGPR(), dl, MVT::i32); | |||
3772 | SDValue ArgFPR = DAG.getConstant(FuncInfo->getVarArgsNumFPR(), dl, MVT::i32); | |||
3773 | SDValue StackOffsetFI = DAG.getFrameIndex(FuncInfo->getVarArgsStackOffset(), | |||
3774 | PtrVT); | |||
3775 | SDValue FR = DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), | |||
3776 | PtrVT); | |||
3777 | ||||
3778 | uint64_t FrameOffset = PtrVT.getSizeInBits()/8; | |||
3779 | SDValue ConstFrameOffset = DAG.getConstant(FrameOffset, dl, PtrVT); | |||
3780 | ||||
3781 | uint64_t StackOffset = PtrVT.getSizeInBits()/8 - 1; | |||
3782 | SDValue ConstStackOffset = DAG.getConstant(StackOffset, dl, PtrVT); | |||
3783 | ||||
3784 | uint64_t FPROffset = 1; | |||
3785 | SDValue ConstFPROffset = DAG.getConstant(FPROffset, dl, PtrVT); | |||
3786 | ||||
3787 | const Value *SV = cast<SrcValueSDNode>(Op.getOperand(2))->getValue(); | |||
3788 | ||||
3789 | // Store first byte : number of int regs | |||
3790 | SDValue firstStore = | |||
3791 | DAG.getTruncStore(Op.getOperand(0), dl, ArgGPR, Op.getOperand(1), | |||
3792 | MachinePointerInfo(SV), MVT::i8); | |||
3793 | uint64_t nextOffset = FPROffset; | |||
3794 | SDValue nextPtr = DAG.getNode(ISD::ADD, dl, PtrVT, Op.getOperand(1), | |||
3795 | ConstFPROffset); | |||
3796 | ||||
3797 | // Store second byte : number of float regs | |||
3798 | SDValue secondStore = | |||
3799 | DAG.getTruncStore(firstStore, dl, ArgFPR, nextPtr, | |||
3800 | MachinePointerInfo(SV, nextOffset), MVT::i8); | |||
3801 | nextOffset += StackOffset; | |||
3802 | nextPtr = DAG.getNode(ISD::ADD, dl, PtrVT, nextPtr, ConstStackOffset); | |||
3803 | ||||
3804 | // Store second word : arguments given on stack | |||
3805 | SDValue thirdStore = DAG.getStore(secondStore, dl, StackOffsetFI, nextPtr, | |||
3806 | MachinePointerInfo(SV, nextOffset)); | |||
3807 | nextOffset += FrameOffset; | |||
3808 | nextPtr = DAG.getNode(ISD::ADD, dl, PtrVT, nextPtr, ConstFrameOffset); | |||
3809 | ||||
3810 | // Store third word : arguments given in registers | |||
3811 | return DAG.getStore(thirdStore, dl, FR, nextPtr, | |||
3812 | MachinePointerInfo(SV, nextOffset)); | |||
3813 | } | |||
3814 | ||||
3815 | /// FPR - The set of FP registers that should be allocated for arguments | |||
3816 | /// on Darwin and AIX. | |||
3817 | static const MCPhysReg FPR[] = {PPC::F1, PPC::F2, PPC::F3, PPC::F4, PPC::F5, | |||
3818 | PPC::F6, PPC::F7, PPC::F8, PPC::F9, PPC::F10, | |||
3819 | PPC::F11, PPC::F12, PPC::F13}; | |||
3820 | ||||
3821 | /// CalculateStackSlotSize - Calculates the size reserved for this argument on | |||
3822 | /// the stack. | |||
3823 | static unsigned CalculateStackSlotSize(EVT ArgVT, ISD::ArgFlagsTy Flags, | |||
3824 | unsigned PtrByteSize) { | |||
3825 | unsigned ArgSize = ArgVT.getStoreSize(); | |||
3826 | if (Flags.isByVal()) | |||
3827 | ArgSize = Flags.getByValSize(); | |||
3828 | ||||
3829 | // Round up to multiples of the pointer size, except for array members, | |||
3830 | // which are always packed. | |||
3831 | if (!Flags.isInConsecutiveRegs()) | |||
3832 | ArgSize = ((ArgSize + PtrByteSize - 1)/PtrByteSize) * PtrByteSize; | |||
3833 | ||||
3834 | return ArgSize; | |||
3835 | } | |||
3836 | ||||
3837 | /// CalculateStackSlotAlignment - Calculates the alignment of this argument | |||
3838 | /// on the stack. | |||
3839 | static Align CalculateStackSlotAlignment(EVT ArgVT, EVT OrigVT, | |||
3840 | ISD::ArgFlagsTy Flags, | |||
3841 | unsigned PtrByteSize) { | |||
3842 | Align Alignment(PtrByteSize); | |||
3843 | ||||
3844 | // Altivec parameters are padded to a 16 byte boundary. | |||
3845 | if (ArgVT == MVT::v4f32 || ArgVT == MVT::v4i32 || | |||
3846 | ArgVT == MVT::v8i16 || ArgVT == MVT::v16i8 || | |||
3847 | ArgVT == MVT::v2f64 || ArgVT == MVT::v2i64 || | |||
3848 | ArgVT == MVT::v1i128 || ArgVT == MVT::f128) | |||
3849 | Alignment = Align(16); | |||
3850 | ||||
3851 | // ByVal parameters are aligned as requested. | |||
3852 | if (Flags.isByVal()) { | |||
3853 | auto BVAlign = Flags.getNonZeroByValAlign(); | |||
3854 | if (BVAlign > PtrByteSize) { | |||
3855 | if (BVAlign.value() % PtrByteSize != 0) | |||
3856 | llvm_unreachable(::llvm::llvm_unreachable_internal("ByVal alignment is not a multiple of the pointer size" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3857) | |||
3857 | "ByVal alignment is not a multiple of the pointer size")::llvm::llvm_unreachable_internal("ByVal alignment is not a multiple of the pointer size" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3857); | |||
3858 | ||||
3859 | Alignment = BVAlign; | |||
3860 | } | |||
3861 | } | |||
3862 | ||||
3863 | // Array members are always packed to their original alignment. | |||
3864 | if (Flags.isInConsecutiveRegs()) { | |||
3865 | // If the array member was split into multiple registers, the first | |||
3866 | // needs to be aligned to the size of the full type. (Except for | |||
3867 | // ppcf128, which is only aligned as its f64 components.) | |||
3868 | if (Flags.isSplit() && OrigVT != MVT::ppcf128) | |||
3869 | Alignment = Align(OrigVT.getStoreSize()); | |||
3870 | else | |||
3871 | Alignment = Align(ArgVT.getStoreSize()); | |||
3872 | } | |||
3873 | ||||
3874 | return Alignment; | |||
3875 | } | |||
3876 | ||||
3877 | /// CalculateStackSlotUsed - Return whether this argument will use its | |||
3878 | /// stack slot (instead of being passed in registers). ArgOffset, | |||
3879 | /// AvailableFPRs, and AvailableVRs must hold the current argument | |||
3880 | /// position, and will be updated to account for this argument. | |||
3881 | static bool CalculateStackSlotUsed(const PPCSubtarget &Subtarget, EVT ArgVT, | |||
3882 | EVT OrigVT, ISD::ArgFlagsTy Flags, | |||
3883 | unsigned PtrByteSize, unsigned LinkageSize, | |||
3884 | unsigned ParamAreaSize, unsigned &ArgOffset, | |||
3885 | unsigned &AvailableFPRs, | |||
3886 | unsigned &AvailableVRs) { | |||
3887 | bool UseMemory = false; | |||
3888 | ||||
3889 | // Respect alignment of argument on the stack. | |||
3890 | Align Alignment = | |||
3891 | CalculateStackSlotAlignment(ArgVT, OrigVT, Flags, PtrByteSize); | |||
3892 | ArgOffset = alignTo(ArgOffset, Alignment); | |||
3893 | // If there's no space left in the argument save area, we must | |||
3894 | // use memory (this check also catches zero-sized arguments). | |||
3895 | if (ArgOffset >= LinkageSize + ParamAreaSize) | |||
3896 | UseMemory = true; | |||
3897 | ||||
3898 | // Allocate argument on the stack. | |||
3899 | ArgOffset += CalculateStackSlotSize(ArgVT, Flags, PtrByteSize); | |||
3900 | if (Flags.isInConsecutiveRegsLast()) | |||
3901 | ArgOffset = ((ArgOffset + PtrByteSize - 1)/PtrByteSize) * PtrByteSize; | |||
3902 | // If we overran the argument save area, we must use memory | |||
3903 | // (this check catches arguments passed partially in memory) | |||
3904 | if (ArgOffset > LinkageSize + ParamAreaSize) | |||
3905 | UseMemory = true; | |||
3906 | ||||
3907 | // However, if the argument is actually passed in an FPR or a VR, | |||
3908 | // we don't use memory after all. | |||
3909 | if (!Flags.isByVal()) { | |||
3910 | if (ArgVT == MVT::f32 || ArgVT == MVT::f64) | |||
3911 | if (AvailableFPRs > 0) { | |||
3912 | --AvailableFPRs; | |||
3913 | return false; | |||
3914 | } | |||
3915 | if (ArgVT == MVT::v4f32 || ArgVT == MVT::v4i32 || | |||
3916 | ArgVT == MVT::v8i16 || ArgVT == MVT::v16i8 || | |||
3917 | ArgVT == MVT::v2f64 || ArgVT == MVT::v2i64 || | |||
3918 | ArgVT == MVT::v1i128 || ArgVT == MVT::f128) | |||
3919 | if (AvailableVRs > 0) { | |||
3920 | --AvailableVRs; | |||
3921 | return false; | |||
3922 | } | |||
3923 | } else if (Subtarget.isPPC64() && Subtarget.isELFv2ABI() && | |||
3924 | Flags.getByValSize() >= 8) | |||
3925 | // For 64-bit ELF v2, passing by value object whose size is no less than 8 | |||
3926 | // bytes will be copied to parameter save area. This is for compatibility | |||
3927 | // for other compiler which requires byval parameters to be stored in | |||
3928 | // caller's parameter save area. | |||
3929 | return true; | |||
3930 | ||||
3931 | return UseMemory; | |||
3932 | } | |||
3933 | ||||
3934 | /// EnsureStackAlignment - Round stack frame size up from NumBytes to | |||
3935 | /// ensure minimum alignment required for target. | |||
3936 | static unsigned EnsureStackAlignment(const PPCFrameLowering *Lowering, | |||
3937 | unsigned NumBytes) { | |||
3938 | return alignTo(NumBytes, Lowering->getStackAlign()); | |||
3939 | } | |||
3940 | ||||
3941 | SDValue PPCTargetLowering::LowerFormalArguments( | |||
3942 | SDValue Chain, CallingConv::ID CallConv, bool isVarArg, | |||
3943 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
3944 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const { | |||
3945 | if (Subtarget.isAIXABI()) | |||
3946 | return LowerFormalArguments_AIX(Chain, CallConv, isVarArg, Ins, dl, DAG, | |||
3947 | InVals); | |||
3948 | if (Subtarget.is64BitELFABI()) | |||
3949 | return LowerFormalArguments_64SVR4(Chain, CallConv, isVarArg, Ins, dl, DAG, | |||
3950 | InVals); | |||
3951 | assert(Subtarget.is32BitELFABI())(static_cast <bool> (Subtarget.is32BitELFABI()) ? void ( 0) : __assert_fail ("Subtarget.is32BitELFABI()", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 3951, __extension__ __PRETTY_FUNCTION__)); | |||
3952 | return LowerFormalArguments_32SVR4(Chain, CallConv, isVarArg, Ins, dl, DAG, | |||
3953 | InVals); | |||
3954 | } | |||
3955 | ||||
3956 | SDValue PPCTargetLowering::LowerFormalArguments_32SVR4( | |||
3957 | SDValue Chain, CallingConv::ID CallConv, bool isVarArg, | |||
3958 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
3959 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const { | |||
3960 | ||||
3961 | // 32-bit SVR4 ABI Stack Frame Layout: | |||
3962 | // +-----------------------------------+ | |||
3963 | // +--> | Back chain | | |||
3964 | // | +-----------------------------------+ | |||
3965 | // | | Floating-point register save area | | |||
3966 | // | +-----------------------------------+ | |||
3967 | // | | General register save area | | |||
3968 | // | +-----------------------------------+ | |||
3969 | // | | CR save word | | |||
3970 | // | +-----------------------------------+ | |||
3971 | // | | VRSAVE save word | | |||
3972 | // | +-----------------------------------+ | |||
3973 | // | | Alignment padding | | |||
3974 | // | +-----------------------------------+ | |||
3975 | // | | Vector register save area | | |||
3976 | // | +-----------------------------------+ | |||
3977 | // | | Local variable space | | |||
3978 | // | +-----------------------------------+ | |||
3979 | // | | Parameter list area | | |||
3980 | // | +-----------------------------------+ | |||
3981 | // | | LR save word | | |||
3982 | // | +-----------------------------------+ | |||
3983 | // SP--> +--- | Back chain | | |||
3984 | // +-----------------------------------+ | |||
3985 | // | |||
3986 | // Specifications: | |||
3987 | // System V Application Binary Interface PowerPC Processor Supplement | |||
3988 | // AltiVec Technology Programming Interface Manual | |||
3989 | ||||
3990 | MachineFunction &MF = DAG.getMachineFunction(); | |||
3991 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
3992 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
3993 | ||||
3994 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
3995 | // Potential tail calls could cause overwriting of argument stack slots. | |||
3996 | bool isImmutable = !(getTargetMachine().Options.GuaranteedTailCallOpt && | |||
3997 | (CallConv == CallingConv::Fast)); | |||
3998 | const Align PtrAlign(4); | |||
3999 | ||||
4000 | // Assign locations to all of the incoming arguments. | |||
4001 | SmallVector<CCValAssign, 16> ArgLocs; | |||
4002 | PPCCCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs, | |||
4003 | *DAG.getContext()); | |||
4004 | ||||
4005 | // Reserve space for the linkage area on the stack. | |||
4006 | unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize(); | |||
4007 | CCInfo.AllocateStack(LinkageSize, PtrAlign); | |||
4008 | if (useSoftFloat()) | |||
4009 | CCInfo.PreAnalyzeFormalArguments(Ins); | |||
4010 | ||||
4011 | CCInfo.AnalyzeFormalArguments(Ins, CC_PPC32_SVR4); | |||
4012 | CCInfo.clearWasPPCF128(); | |||
4013 | ||||
4014 | for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) { | |||
4015 | CCValAssign &VA = ArgLocs[i]; | |||
4016 | ||||
4017 | // Arguments stored in registers. | |||
4018 | if (VA.isRegLoc()) { | |||
4019 | const TargetRegisterClass *RC; | |||
4020 | EVT ValVT = VA.getValVT(); | |||
4021 | ||||
4022 | switch (ValVT.getSimpleVT().SimpleTy) { | |||
4023 | default: | |||
4024 | llvm_unreachable("ValVT not supported by formal arguments Lowering")::llvm::llvm_unreachable_internal("ValVT not supported by formal arguments Lowering" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4024); | |||
4025 | case MVT::i1: | |||
4026 | case MVT::i32: | |||
4027 | RC = &PPC::GPRCRegClass; | |||
4028 | break; | |||
4029 | case MVT::f32: | |||
4030 | if (Subtarget.hasP8Vector()) | |||
4031 | RC = &PPC::VSSRCRegClass; | |||
4032 | else if (Subtarget.hasSPE()) | |||
4033 | RC = &PPC::GPRCRegClass; | |||
4034 | else | |||
4035 | RC = &PPC::F4RCRegClass; | |||
4036 | break; | |||
4037 | case MVT::f64: | |||
4038 | if (Subtarget.hasVSX()) | |||
4039 | RC = &PPC::VSFRCRegClass; | |||
4040 | else if (Subtarget.hasSPE()) | |||
4041 | // SPE passes doubles in GPR pairs. | |||
4042 | RC = &PPC::GPRCRegClass; | |||
4043 | else | |||
4044 | RC = &PPC::F8RCRegClass; | |||
4045 | break; | |||
4046 | case MVT::v16i8: | |||
4047 | case MVT::v8i16: | |||
4048 | case MVT::v4i32: | |||
4049 | RC = &PPC::VRRCRegClass; | |||
4050 | break; | |||
4051 | case MVT::v4f32: | |||
4052 | RC = &PPC::VRRCRegClass; | |||
4053 | break; | |||
4054 | case MVT::v2f64: | |||
4055 | case MVT::v2i64: | |||
4056 | RC = &PPC::VRRCRegClass; | |||
4057 | break; | |||
4058 | } | |||
4059 | ||||
4060 | SDValue ArgValue; | |||
4061 | // Transform the arguments stored in physical registers into | |||
4062 | // virtual ones. | |||
4063 | if (VA.getLocVT() == MVT::f64 && Subtarget.hasSPE()) { | |||
4064 | assert(i + 1 < e && "No second half of double precision argument")(static_cast <bool> (i + 1 < e && "No second half of double precision argument" ) ? void (0) : __assert_fail ("i + 1 < e && \"No second half of double precision argument\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4064, __extension__ __PRETTY_FUNCTION__)); | |||
4065 | unsigned RegLo = MF.addLiveIn(VA.getLocReg(), RC); | |||
4066 | unsigned RegHi = MF.addLiveIn(ArgLocs[++i].getLocReg(), RC); | |||
4067 | SDValue ArgValueLo = DAG.getCopyFromReg(Chain, dl, RegLo, MVT::i32); | |||
4068 | SDValue ArgValueHi = DAG.getCopyFromReg(Chain, dl, RegHi, MVT::i32); | |||
4069 | if (!Subtarget.isLittleEndian()) | |||
4070 | std::swap (ArgValueLo, ArgValueHi); | |||
4071 | ArgValue = DAG.getNode(PPCISD::BUILD_SPE64, dl, MVT::f64, ArgValueLo, | |||
4072 | ArgValueHi); | |||
4073 | } else { | |||
4074 | unsigned Reg = MF.addLiveIn(VA.getLocReg(), RC); | |||
4075 | ArgValue = DAG.getCopyFromReg(Chain, dl, Reg, | |||
4076 | ValVT == MVT::i1 ? MVT::i32 : ValVT); | |||
4077 | if (ValVT == MVT::i1) | |||
4078 | ArgValue = DAG.getNode(ISD::TRUNCATE, dl, MVT::i1, ArgValue); | |||
4079 | } | |||
4080 | ||||
4081 | InVals.push_back(ArgValue); | |||
4082 | } else { | |||
4083 | // Argument stored in memory. | |||
4084 | assert(VA.isMemLoc())(static_cast <bool> (VA.isMemLoc()) ? void (0) : __assert_fail ("VA.isMemLoc()", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4084, __extension__ __PRETTY_FUNCTION__)); | |||
4085 | ||||
4086 | // Get the extended size of the argument type in stack | |||
4087 | unsigned ArgSize = VA.getLocVT().getStoreSize(); | |||
4088 | // Get the actual size of the argument type | |||
4089 | unsigned ObjSize = VA.getValVT().getStoreSize(); | |||
4090 | unsigned ArgOffset = VA.getLocMemOffset(); | |||
4091 | // Stack objects in PPC32 are right justified. | |||
4092 | ArgOffset += ArgSize - ObjSize; | |||
4093 | int FI = MFI.CreateFixedObject(ArgSize, ArgOffset, isImmutable); | |||
4094 | ||||
4095 | // Create load nodes to retrieve arguments from the stack. | |||
4096 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
4097 | InVals.push_back( | |||
4098 | DAG.getLoad(VA.getValVT(), dl, Chain, FIN, MachinePointerInfo())); | |||
4099 | } | |||
4100 | } | |||
4101 | ||||
4102 | // Assign locations to all of the incoming aggregate by value arguments. | |||
4103 | // Aggregates passed by value are stored in the local variable space of the | |||
4104 | // caller's stack frame, right above the parameter list area. | |||
4105 | SmallVector<CCValAssign, 16> ByValArgLocs; | |||
4106 | CCState CCByValInfo(CallConv, isVarArg, DAG.getMachineFunction(), | |||
4107 | ByValArgLocs, *DAG.getContext()); | |||
4108 | ||||
4109 | // Reserve stack space for the allocations in CCInfo. | |||
4110 | CCByValInfo.AllocateStack(CCInfo.getNextStackOffset(), PtrAlign); | |||
4111 | ||||
4112 | CCByValInfo.AnalyzeFormalArguments(Ins, CC_PPC32_SVR4_ByVal); | |||
4113 | ||||
4114 | // Area that is at least reserved in the caller of this function. | |||
4115 | unsigned MinReservedArea = CCByValInfo.getNextStackOffset(); | |||
4116 | MinReservedArea = std::max(MinReservedArea, LinkageSize); | |||
4117 | ||||
4118 | // Set the size that is at least reserved in caller of this function. Tail | |||
4119 | // call optimized function's reserved stack space needs to be aligned so that | |||
4120 | // taking the difference between two stack areas will result in an aligned | |||
4121 | // stack. | |||
4122 | MinReservedArea = | |||
4123 | EnsureStackAlignment(Subtarget.getFrameLowering(), MinReservedArea); | |||
4124 | FuncInfo->setMinReservedArea(MinReservedArea); | |||
4125 | ||||
4126 | SmallVector<SDValue, 8> MemOps; | |||
4127 | ||||
4128 | // If the function takes variable number of arguments, make a frame index for | |||
4129 | // the start of the first vararg value... for expansion of llvm.va_start. | |||
4130 | if (isVarArg) { | |||
4131 | static const MCPhysReg GPArgRegs[] = { | |||
4132 | PPC::R3, PPC::R4, PPC::R5, PPC::R6, | |||
4133 | PPC::R7, PPC::R8, PPC::R9, PPC::R10, | |||
4134 | }; | |||
4135 | const unsigned NumGPArgRegs = array_lengthof(GPArgRegs); | |||
4136 | ||||
4137 | static const MCPhysReg FPArgRegs[] = { | |||
4138 | PPC::F1, PPC::F2, PPC::F3, PPC::F4, PPC::F5, PPC::F6, PPC::F7, | |||
4139 | PPC::F8 | |||
4140 | }; | |||
4141 | unsigned NumFPArgRegs = array_lengthof(FPArgRegs); | |||
4142 | ||||
4143 | if (useSoftFloat() || hasSPE()) | |||
4144 | NumFPArgRegs = 0; | |||
4145 | ||||
4146 | FuncInfo->setVarArgsNumGPR(CCInfo.getFirstUnallocated(GPArgRegs)); | |||
4147 | FuncInfo->setVarArgsNumFPR(CCInfo.getFirstUnallocated(FPArgRegs)); | |||
4148 | ||||
4149 | // Make room for NumGPArgRegs and NumFPArgRegs. | |||
4150 | int Depth = NumGPArgRegs * PtrVT.getSizeInBits()/8 + | |||
4151 | NumFPArgRegs * MVT(MVT::f64).getSizeInBits()/8; | |||
4152 | ||||
4153 | FuncInfo->setVarArgsStackOffset( | |||
4154 | MFI.CreateFixedObject(PtrVT.getSizeInBits()/8, | |||
4155 | CCInfo.getNextStackOffset(), true)); | |||
4156 | ||||
4157 | FuncInfo->setVarArgsFrameIndex( | |||
4158 | MFI.CreateStackObject(Depth, Align(8), false)); | |||
4159 | SDValue FIN = DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT); | |||
4160 | ||||
4161 | // The fixed integer arguments of a variadic function are stored to the | |||
4162 | // VarArgsFrameIndex on the stack so that they may be loaded by | |||
4163 | // dereferencing the result of va_next. | |||
4164 | for (unsigned GPRIndex = 0; GPRIndex != NumGPArgRegs; ++GPRIndex) { | |||
4165 | // Get an existing live-in vreg, or add a new one. | |||
4166 | unsigned VReg = MF.getRegInfo().getLiveInVirtReg(GPArgRegs[GPRIndex]); | |||
4167 | if (!VReg) | |||
4168 | VReg = MF.addLiveIn(GPArgRegs[GPRIndex], &PPC::GPRCRegClass); | |||
4169 | ||||
4170 | SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT); | |||
4171 | SDValue Store = | |||
4172 | DAG.getStore(Val.getValue(1), dl, Val, FIN, MachinePointerInfo()); | |||
4173 | MemOps.push_back(Store); | |||
4174 | // Increment the address by four for the next argument to store | |||
4175 | SDValue PtrOff = DAG.getConstant(PtrVT.getSizeInBits()/8, dl, PtrVT); | |||
4176 | FIN = DAG.getNode(ISD::ADD, dl, PtrOff.getValueType(), FIN, PtrOff); | |||
4177 | } | |||
4178 | ||||
4179 | // FIXME 32-bit SVR4: We only need to save FP argument registers if CR bit 6 | |||
4180 | // is set. | |||
4181 | // The double arguments are stored to the VarArgsFrameIndex | |||
4182 | // on the stack. | |||
4183 | for (unsigned FPRIndex = 0; FPRIndex != NumFPArgRegs; ++FPRIndex) { | |||
4184 | // Get an existing live-in vreg, or add a new one. | |||
4185 | unsigned VReg = MF.getRegInfo().getLiveInVirtReg(FPArgRegs[FPRIndex]); | |||
4186 | if (!VReg) | |||
4187 | VReg = MF.addLiveIn(FPArgRegs[FPRIndex], &PPC::F8RCRegClass); | |||
4188 | ||||
4189 | SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, MVT::f64); | |||
4190 | SDValue Store = | |||
4191 | DAG.getStore(Val.getValue(1), dl, Val, FIN, MachinePointerInfo()); | |||
4192 | MemOps.push_back(Store); | |||
4193 | // Increment the address by eight for the next argument to store | |||
4194 | SDValue PtrOff = DAG.getConstant(MVT(MVT::f64).getSizeInBits()/8, dl, | |||
4195 | PtrVT); | |||
4196 | FIN = DAG.getNode(ISD::ADD, dl, PtrOff.getValueType(), FIN, PtrOff); | |||
4197 | } | |||
4198 | } | |||
4199 | ||||
4200 | if (!MemOps.empty()) | |||
4201 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOps); | |||
4202 | ||||
4203 | return Chain; | |||
4204 | } | |||
4205 | ||||
4206 | // PPC64 passes i8, i16, and i32 values in i64 registers. Promote | |||
4207 | // value to MVT::i64 and then truncate to the correct register size. | |||
4208 | SDValue PPCTargetLowering::extendArgForPPC64(ISD::ArgFlagsTy Flags, | |||
4209 | EVT ObjectVT, SelectionDAG &DAG, | |||
4210 | SDValue ArgVal, | |||
4211 | const SDLoc &dl) const { | |||
4212 | if (Flags.isSExt()) | |||
4213 | ArgVal = DAG.getNode(ISD::AssertSext, dl, MVT::i64, ArgVal, | |||
4214 | DAG.getValueType(ObjectVT)); | |||
4215 | else if (Flags.isZExt()) | |||
4216 | ArgVal = DAG.getNode(ISD::AssertZext, dl, MVT::i64, ArgVal, | |||
4217 | DAG.getValueType(ObjectVT)); | |||
4218 | ||||
4219 | return DAG.getNode(ISD::TRUNCATE, dl, ObjectVT, ArgVal); | |||
4220 | } | |||
4221 | ||||
4222 | SDValue PPCTargetLowering::LowerFormalArguments_64SVR4( | |||
4223 | SDValue Chain, CallingConv::ID CallConv, bool isVarArg, | |||
4224 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
4225 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const { | |||
4226 | // TODO: add description of PPC stack frame format, or at least some docs. | |||
4227 | // | |||
4228 | bool isELFv2ABI = Subtarget.isELFv2ABI(); | |||
4229 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
4230 | MachineFunction &MF = DAG.getMachineFunction(); | |||
4231 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
4232 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
4233 | ||||
4234 | assert(!(CallConv == CallingConv::Fast && isVarArg) &&(static_cast <bool> (!(CallConv == CallingConv::Fast && isVarArg) && "fastcc not supported on varargs functions" ) ? void (0) : __assert_fail ("!(CallConv == CallingConv::Fast && isVarArg) && \"fastcc not supported on varargs functions\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4235, __extension__ __PRETTY_FUNCTION__)) | |||
4235 | "fastcc not supported on varargs functions")(static_cast <bool> (!(CallConv == CallingConv::Fast && isVarArg) && "fastcc not supported on varargs functions" ) ? void (0) : __assert_fail ("!(CallConv == CallingConv::Fast && isVarArg) && \"fastcc not supported on varargs functions\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4235, __extension__ __PRETTY_FUNCTION__)); | |||
4236 | ||||
4237 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
4238 | // Potential tail calls could cause overwriting of argument stack slots. | |||
4239 | bool isImmutable = !(getTargetMachine().Options.GuaranteedTailCallOpt && | |||
4240 | (CallConv == CallingConv::Fast)); | |||
4241 | unsigned PtrByteSize = 8; | |||
4242 | unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize(); | |||
4243 | ||||
4244 | static const MCPhysReg GPR[] = { | |||
4245 | PPC::X3, PPC::X4, PPC::X5, PPC::X6, | |||
4246 | PPC::X7, PPC::X8, PPC::X9, PPC::X10, | |||
4247 | }; | |||
4248 | static const MCPhysReg VR[] = { | |||
4249 | PPC::V2, PPC::V3, PPC::V4, PPC::V5, PPC::V6, PPC::V7, PPC::V8, | |||
4250 | PPC::V9, PPC::V10, PPC::V11, PPC::V12, PPC::V13 | |||
4251 | }; | |||
4252 | ||||
4253 | const unsigned Num_GPR_Regs = array_lengthof(GPR); | |||
4254 | const unsigned Num_FPR_Regs = useSoftFloat() ? 0 : 13; | |||
4255 | const unsigned Num_VR_Regs = array_lengthof(VR); | |||
4256 | ||||
4257 | // Do a first pass over the arguments to determine whether the ABI | |||
4258 | // guarantees that our caller has allocated the parameter save area | |||
4259 | // on its stack frame. In the ELFv1 ABI, this is always the case; | |||
4260 | // in the ELFv2 ABI, it is true if this is a vararg function or if | |||
4261 | // any parameter is located in a stack slot. | |||
4262 | ||||
4263 | bool HasParameterArea = !isELFv2ABI || isVarArg; | |||
4264 | unsigned ParamAreaSize = Num_GPR_Regs * PtrByteSize; | |||
4265 | unsigned NumBytes = LinkageSize; | |||
4266 | unsigned AvailableFPRs = Num_FPR_Regs; | |||
4267 | unsigned AvailableVRs = Num_VR_Regs; | |||
4268 | for (unsigned i = 0, e = Ins.size(); i != e; ++i) { | |||
4269 | if (Ins[i].Flags.isNest()) | |||
4270 | continue; | |||
4271 | ||||
4272 | if (CalculateStackSlotUsed(Subtarget, Ins[i].VT, Ins[i].ArgVT, Ins[i].Flags, | |||
4273 | PtrByteSize, LinkageSize, ParamAreaSize, | |||
4274 | NumBytes, AvailableFPRs, AvailableVRs)) | |||
4275 | HasParameterArea = true; | |||
4276 | } | |||
4277 | ||||
4278 | // Add DAG nodes to load the arguments or copy them out of registers. On | |||
4279 | // entry to a function on PPC, the arguments start after the linkage area, | |||
4280 | // although the first ones are often in registers. | |||
4281 | ||||
4282 | unsigned ArgOffset = LinkageSize; | |||
4283 | unsigned GPR_idx = 0, FPR_idx = 0, VR_idx = 0; | |||
4284 | SmallVector<SDValue, 8> MemOps; | |||
4285 | Function::const_arg_iterator FuncArg = MF.getFunction().arg_begin(); | |||
4286 | unsigned CurArgIdx = 0; | |||
4287 | for (unsigned ArgNo = 0, e = Ins.size(); ArgNo != e; ++ArgNo) { | |||
4288 | SDValue ArgVal; | |||
4289 | bool needsLoad = false; | |||
4290 | EVT ObjectVT = Ins[ArgNo].VT; | |||
4291 | EVT OrigVT = Ins[ArgNo].ArgVT; | |||
4292 | unsigned ObjSize = ObjectVT.getStoreSize(); | |||
4293 | unsigned ArgSize = ObjSize; | |||
4294 | ISD::ArgFlagsTy Flags = Ins[ArgNo].Flags; | |||
4295 | if (Ins[ArgNo].isOrigArg()) { | |||
4296 | std::advance(FuncArg, Ins[ArgNo].getOrigArgIndex() - CurArgIdx); | |||
4297 | CurArgIdx = Ins[ArgNo].getOrigArgIndex(); | |||
4298 | } | |||
4299 | // We re-align the argument offset for each argument, except when using the | |||
4300 | // fast calling convention, when we need to make sure we do that only when | |||
4301 | // we'll actually use a stack slot. | |||
4302 | unsigned CurArgOffset; | |||
4303 | Align Alignment; | |||
4304 | auto ComputeArgOffset = [&]() { | |||
4305 | /* Respect alignment of argument on the stack. */ | |||
4306 | Alignment = | |||
4307 | CalculateStackSlotAlignment(ObjectVT, OrigVT, Flags, PtrByteSize); | |||
4308 | ArgOffset = alignTo(ArgOffset, Alignment); | |||
4309 | CurArgOffset = ArgOffset; | |||
4310 | }; | |||
4311 | ||||
4312 | if (CallConv != CallingConv::Fast) { | |||
4313 | ComputeArgOffset(); | |||
4314 | ||||
4315 | /* Compute GPR index associated with argument offset. */ | |||
4316 | GPR_idx = (ArgOffset - LinkageSize) / PtrByteSize; | |||
4317 | GPR_idx = std::min(GPR_idx, Num_GPR_Regs); | |||
4318 | } | |||
4319 | ||||
4320 | // FIXME the codegen can be much improved in some cases. | |||
4321 | // We do not have to keep everything in memory. | |||
4322 | if (Flags.isByVal()) { | |||
4323 | assert(Ins[ArgNo].isOrigArg() && "Byval arguments cannot be implicit")(static_cast <bool> (Ins[ArgNo].isOrigArg() && "Byval arguments cannot be implicit" ) ? void (0) : __assert_fail ("Ins[ArgNo].isOrigArg() && \"Byval arguments cannot be implicit\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4323, __extension__ __PRETTY_FUNCTION__)); | |||
4324 | ||||
4325 | if (CallConv == CallingConv::Fast) | |||
4326 | ComputeArgOffset(); | |||
4327 | ||||
4328 | // ObjSize is the true size, ArgSize rounded up to multiple of registers. | |||
4329 | ObjSize = Flags.getByValSize(); | |||
4330 | ArgSize = ((ObjSize + PtrByteSize - 1)/PtrByteSize) * PtrByteSize; | |||
4331 | // Empty aggregate parameters do not take up registers. Examples: | |||
4332 | // struct { } a; | |||
4333 | // union { } b; | |||
4334 | // int c[0]; | |||
4335 | // etc. However, we have to provide a place-holder in InVals, so | |||
4336 | // pretend we have an 8-byte item at the current address for that | |||
4337 | // purpose. | |||
4338 | if (!ObjSize) { | |||
4339 | int FI = MFI.CreateFixedObject(PtrByteSize, ArgOffset, true); | |||
4340 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
4341 | InVals.push_back(FIN); | |||
4342 | continue; | |||
4343 | } | |||
4344 | ||||
4345 | // Create a stack object covering all stack doublewords occupied | |||
4346 | // by the argument. If the argument is (fully or partially) on | |||
4347 | // the stack, or if the argument is fully in registers but the | |||
4348 | // caller has allocated the parameter save anyway, we can refer | |||
4349 | // directly to the caller's stack frame. Otherwise, create a | |||
4350 | // local copy in our own frame. | |||
4351 | int FI; | |||
4352 | if (HasParameterArea || | |||
4353 | ArgSize + ArgOffset > LinkageSize + Num_GPR_Regs * PtrByteSize) | |||
4354 | FI = MFI.CreateFixedObject(ArgSize, ArgOffset, false, true); | |||
4355 | else | |||
4356 | FI = MFI.CreateStackObject(ArgSize, Alignment, false); | |||
4357 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
4358 | ||||
4359 | // Handle aggregates smaller than 8 bytes. | |||
4360 | if (ObjSize < PtrByteSize) { | |||
4361 | // The value of the object is its address, which differs from the | |||
4362 | // address of the enclosing doubleword on big-endian systems. | |||
4363 | SDValue Arg = FIN; | |||
4364 | if (!isLittleEndian) { | |||
4365 | SDValue ArgOff = DAG.getConstant(PtrByteSize - ObjSize, dl, PtrVT); | |||
4366 | Arg = DAG.getNode(ISD::ADD, dl, ArgOff.getValueType(), Arg, ArgOff); | |||
4367 | } | |||
4368 | InVals.push_back(Arg); | |||
4369 | ||||
4370 | if (GPR_idx != Num_GPR_Regs) { | |||
4371 | unsigned VReg = MF.addLiveIn(GPR[GPR_idx++], &PPC::G8RCRegClass); | |||
4372 | FuncInfo->addLiveInAttr(VReg, Flags); | |||
4373 | SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT); | |||
4374 | SDValue Store; | |||
4375 | ||||
4376 | if (ObjSize==1 || ObjSize==2 || ObjSize==4) { | |||
4377 | EVT ObjType = (ObjSize == 1 ? MVT::i8 : | |||
4378 | (ObjSize == 2 ? MVT::i16 : MVT::i32)); | |||
4379 | Store = DAG.getTruncStore(Val.getValue(1), dl, Val, Arg, | |||
4380 | MachinePointerInfo(&*FuncArg), ObjType); | |||
4381 | } else { | |||
4382 | // For sizes that don't fit a truncating store (3, 5, 6, 7), | |||
4383 | // store the whole register as-is to the parameter save area | |||
4384 | // slot. | |||
4385 | Store = DAG.getStore(Val.getValue(1), dl, Val, FIN, | |||
4386 | MachinePointerInfo(&*FuncArg)); | |||
4387 | } | |||
4388 | ||||
4389 | MemOps.push_back(Store); | |||
4390 | } | |||
4391 | // Whether we copied from a register or not, advance the offset | |||
4392 | // into the parameter save area by a full doubleword. | |||
4393 | ArgOffset += PtrByteSize; | |||
4394 | continue; | |||
4395 | } | |||
4396 | ||||
4397 | // The value of the object is its address, which is the address of | |||
4398 | // its first stack doubleword. | |||
4399 | InVals.push_back(FIN); | |||
4400 | ||||
4401 | // Store whatever pieces of the object are in registers to memory. | |||
4402 | for (unsigned j = 0; j < ArgSize; j += PtrByteSize) { | |||
4403 | if (GPR_idx == Num_GPR_Regs) | |||
4404 | break; | |||
4405 | ||||
4406 | unsigned VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::G8RCRegClass); | |||
4407 | FuncInfo->addLiveInAttr(VReg, Flags); | |||
4408 | SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT); | |||
4409 | SDValue Addr = FIN; | |||
4410 | if (j) { | |||
4411 | SDValue Off = DAG.getConstant(j, dl, PtrVT); | |||
4412 | Addr = DAG.getNode(ISD::ADD, dl, Off.getValueType(), Addr, Off); | |||
4413 | } | |||
4414 | SDValue Store = DAG.getStore(Val.getValue(1), dl, Val, Addr, | |||
4415 | MachinePointerInfo(&*FuncArg, j)); | |||
4416 | MemOps.push_back(Store); | |||
4417 | ++GPR_idx; | |||
4418 | } | |||
4419 | ArgOffset += ArgSize; | |||
4420 | continue; | |||
4421 | } | |||
4422 | ||||
4423 | switch (ObjectVT.getSimpleVT().SimpleTy) { | |||
4424 | default: llvm_unreachable("Unhandled argument type!")::llvm::llvm_unreachable_internal("Unhandled argument type!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4424); | |||
4425 | case MVT::i1: | |||
4426 | case MVT::i32: | |||
4427 | case MVT::i64: | |||
4428 | if (Flags.isNest()) { | |||
4429 | // The 'nest' parameter, if any, is passed in R11. | |||
4430 | unsigned VReg = MF.addLiveIn(PPC::X11, &PPC::G8RCRegClass); | |||
4431 | ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i64); | |||
4432 | ||||
4433 | if (ObjectVT == MVT::i32 || ObjectVT == MVT::i1) | |||
4434 | ArgVal = extendArgForPPC64(Flags, ObjectVT, DAG, ArgVal, dl); | |||
4435 | ||||
4436 | break; | |||
4437 | } | |||
4438 | ||||
4439 | // These can be scalar arguments or elements of an integer array type | |||
4440 | // passed directly. Clang may use those instead of "byval" aggregate | |||
4441 | // types to avoid forcing arguments to memory unnecessarily. | |||
4442 | if (GPR_idx != Num_GPR_Regs) { | |||
4443 | unsigned VReg = MF.addLiveIn(GPR[GPR_idx++], &PPC::G8RCRegClass); | |||
4444 | FuncInfo->addLiveInAttr(VReg, Flags); | |||
4445 | ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i64); | |||
4446 | ||||
4447 | if (ObjectVT == MVT::i32 || ObjectVT == MVT::i1) | |||
4448 | // PPC64 passes i8, i16, and i32 values in i64 registers. Promote | |||
4449 | // value to MVT::i64 and then truncate to the correct register size. | |||
4450 | ArgVal = extendArgForPPC64(Flags, ObjectVT, DAG, ArgVal, dl); | |||
4451 | } else { | |||
4452 | if (CallConv == CallingConv::Fast) | |||
4453 | ComputeArgOffset(); | |||
4454 | ||||
4455 | needsLoad = true; | |||
4456 | ArgSize = PtrByteSize; | |||
4457 | } | |||
4458 | if (CallConv != CallingConv::Fast || needsLoad) | |||
4459 | ArgOffset += 8; | |||
4460 | break; | |||
4461 | ||||
4462 | case MVT::f32: | |||
4463 | case MVT::f64: | |||
4464 | // These can be scalar arguments or elements of a float array type | |||
4465 | // passed directly. The latter are used to implement ELFv2 homogenous | |||
4466 | // float aggregates. | |||
4467 | if (FPR_idx != Num_FPR_Regs) { | |||
4468 | unsigned VReg; | |||
4469 | ||||
4470 | if (ObjectVT == MVT::f32) | |||
4471 | VReg = MF.addLiveIn(FPR[FPR_idx], | |||
4472 | Subtarget.hasP8Vector() | |||
4473 | ? &PPC::VSSRCRegClass | |||
4474 | : &PPC::F4RCRegClass); | |||
4475 | else | |||
4476 | VReg = MF.addLiveIn(FPR[FPR_idx], Subtarget.hasVSX() | |||
4477 | ? &PPC::VSFRCRegClass | |||
4478 | : &PPC::F8RCRegClass); | |||
4479 | ||||
4480 | ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, ObjectVT); | |||
4481 | ++FPR_idx; | |||
4482 | } else if (GPR_idx != Num_GPR_Regs && CallConv != CallingConv::Fast) { | |||
4483 | // FIXME: We may want to re-enable this for CallingConv::Fast on the P8 | |||
4484 | // once we support fp <-> gpr moves. | |||
4485 | ||||
4486 | // This can only ever happen in the presence of f32 array types, | |||
4487 | // since otherwise we never run out of FPRs before running out | |||
4488 | // of GPRs. | |||
4489 | unsigned VReg = MF.addLiveIn(GPR[GPR_idx++], &PPC::G8RCRegClass); | |||
4490 | FuncInfo->addLiveInAttr(VReg, Flags); | |||
4491 | ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i64); | |||
4492 | ||||
4493 | if (ObjectVT == MVT::f32) { | |||
4494 | if ((ArgOffset % PtrByteSize) == (isLittleEndian ? 4 : 0)) | |||
4495 | ArgVal = DAG.getNode(ISD::SRL, dl, MVT::i64, ArgVal, | |||
4496 | DAG.getConstant(32, dl, MVT::i32)); | |||
4497 | ArgVal = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, ArgVal); | |||
4498 | } | |||
4499 | ||||
4500 | ArgVal = DAG.getNode(ISD::BITCAST, dl, ObjectVT, ArgVal); | |||
4501 | } else { | |||
4502 | if (CallConv == CallingConv::Fast) | |||
4503 | ComputeArgOffset(); | |||
4504 | ||||
4505 | needsLoad = true; | |||
4506 | } | |||
4507 | ||||
4508 | // When passing an array of floats, the array occupies consecutive | |||
4509 | // space in the argument area; only round up to the next doubleword | |||
4510 | // at the end of the array. Otherwise, each float takes 8 bytes. | |||
4511 | if (CallConv != CallingConv::Fast || needsLoad) { | |||
4512 | ArgSize = Flags.isInConsecutiveRegs() ? ObjSize : PtrByteSize; | |||
4513 | ArgOffset += ArgSize; | |||
4514 | if (Flags.isInConsecutiveRegsLast()) | |||
4515 | ArgOffset = ((ArgOffset + PtrByteSize - 1)/PtrByteSize) * PtrByteSize; | |||
4516 | } | |||
4517 | break; | |||
4518 | case MVT::v4f32: | |||
4519 | case MVT::v4i32: | |||
4520 | case MVT::v8i16: | |||
4521 | case MVT::v16i8: | |||
4522 | case MVT::v2f64: | |||
4523 | case MVT::v2i64: | |||
4524 | case MVT::v1i128: | |||
4525 | case MVT::f128: | |||
4526 | // These can be scalar arguments or elements of a vector array type | |||
4527 | // passed directly. The latter are used to implement ELFv2 homogenous | |||
4528 | // vector aggregates. | |||
4529 | if (VR_idx != Num_VR_Regs) { | |||
4530 | unsigned VReg = MF.addLiveIn(VR[VR_idx], &PPC::VRRCRegClass); | |||
4531 | ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, ObjectVT); | |||
4532 | ++VR_idx; | |||
4533 | } else { | |||
4534 | if (CallConv == CallingConv::Fast) | |||
4535 | ComputeArgOffset(); | |||
4536 | needsLoad = true; | |||
4537 | } | |||
4538 | if (CallConv != CallingConv::Fast || needsLoad) | |||
4539 | ArgOffset += 16; | |||
4540 | break; | |||
4541 | } | |||
4542 | ||||
4543 | // We need to load the argument to a virtual register if we determined | |||
4544 | // above that we ran out of physical registers of the appropriate type. | |||
4545 | if (needsLoad) { | |||
4546 | if (ObjSize < ArgSize && !isLittleEndian) | |||
4547 | CurArgOffset += ArgSize - ObjSize; | |||
4548 | int FI = MFI.CreateFixedObject(ObjSize, CurArgOffset, isImmutable); | |||
4549 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
4550 | ArgVal = DAG.getLoad(ObjectVT, dl, Chain, FIN, MachinePointerInfo()); | |||
4551 | } | |||
4552 | ||||
4553 | InVals.push_back(ArgVal); | |||
4554 | } | |||
4555 | ||||
4556 | // Area that is at least reserved in the caller of this function. | |||
4557 | unsigned MinReservedArea; | |||
4558 | if (HasParameterArea) | |||
4559 | MinReservedArea = std::max(ArgOffset, LinkageSize + 8 * PtrByteSize); | |||
4560 | else | |||
4561 | MinReservedArea = LinkageSize; | |||
4562 | ||||
4563 | // Set the size that is at least reserved in caller of this function. Tail | |||
4564 | // call optimized functions' reserved stack space needs to be aligned so that | |||
4565 | // taking the difference between two stack areas will result in an aligned | |||
4566 | // stack. | |||
4567 | MinReservedArea = | |||
4568 | EnsureStackAlignment(Subtarget.getFrameLowering(), MinReservedArea); | |||
4569 | FuncInfo->setMinReservedArea(MinReservedArea); | |||
4570 | ||||
4571 | // If the function takes variable number of arguments, make a frame index for | |||
4572 | // the start of the first vararg value... for expansion of llvm.va_start. | |||
4573 | // On ELFv2ABI spec, it writes: | |||
4574 | // C programs that are intended to be *portable* across different compilers | |||
4575 | // and architectures must use the header file <stdarg.h> to deal with variable | |||
4576 | // argument lists. | |||
4577 | if (isVarArg && MFI.hasVAStart()) { | |||
4578 | int Depth = ArgOffset; | |||
4579 | ||||
4580 | FuncInfo->setVarArgsFrameIndex( | |||
4581 | MFI.CreateFixedObject(PtrByteSize, Depth, true)); | |||
4582 | SDValue FIN = DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT); | |||
4583 | ||||
4584 | // If this function is vararg, store any remaining integer argument regs | |||
4585 | // to their spots on the stack so that they may be loaded by dereferencing | |||
4586 | // the result of va_next. | |||
4587 | for (GPR_idx = (ArgOffset - LinkageSize) / PtrByteSize; | |||
4588 | GPR_idx < Num_GPR_Regs; ++GPR_idx) { | |||
4589 | unsigned VReg = MF.addLiveIn(GPR[GPR_idx], &PPC::G8RCRegClass); | |||
4590 | SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT); | |||
4591 | SDValue Store = | |||
4592 | DAG.getStore(Val.getValue(1), dl, Val, FIN, MachinePointerInfo()); | |||
4593 | MemOps.push_back(Store); | |||
4594 | // Increment the address by four for the next argument to store | |||
4595 | SDValue PtrOff = DAG.getConstant(PtrByteSize, dl, PtrVT); | |||
4596 | FIN = DAG.getNode(ISD::ADD, dl, PtrOff.getValueType(), FIN, PtrOff); | |||
4597 | } | |||
4598 | } | |||
4599 | ||||
4600 | if (!MemOps.empty()) | |||
4601 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOps); | |||
4602 | ||||
4603 | return Chain; | |||
4604 | } | |||
4605 | ||||
4606 | /// CalculateTailCallSPDiff - Get the amount the stack pointer has to be | |||
4607 | /// adjusted to accommodate the arguments for the tailcall. | |||
4608 | static int CalculateTailCallSPDiff(SelectionDAG& DAG, bool isTailCall, | |||
4609 | unsigned ParamSize) { | |||
4610 | ||||
4611 | if (!isTailCall) return 0; | |||
4612 | ||||
4613 | PPCFunctionInfo *FI = DAG.getMachineFunction().getInfo<PPCFunctionInfo>(); | |||
4614 | unsigned CallerMinReservedArea = FI->getMinReservedArea(); | |||
4615 | int SPDiff = (int)CallerMinReservedArea - (int)ParamSize; | |||
4616 | // Remember only if the new adjustment is bigger. | |||
4617 | if (SPDiff < FI->getTailCallSPDelta()) | |||
4618 | FI->setTailCallSPDelta(SPDiff); | |||
4619 | ||||
4620 | return SPDiff; | |||
4621 | } | |||
4622 | ||||
4623 | static bool isFunctionGlobalAddress(SDValue Callee); | |||
4624 | ||||
4625 | static bool callsShareTOCBase(const Function *Caller, SDValue Callee, | |||
4626 | const TargetMachine &TM) { | |||
4627 | // It does not make sense to call callsShareTOCBase() with a caller that | |||
4628 | // is PC Relative since PC Relative callers do not have a TOC. | |||
4629 | #ifndef NDEBUG | |||
4630 | const PPCSubtarget *STICaller = &TM.getSubtarget<PPCSubtarget>(*Caller); | |||
4631 | assert(!STICaller->isUsingPCRelativeCalls() &&(static_cast <bool> (!STICaller->isUsingPCRelativeCalls () && "PC Relative callers do not have a TOC and cannot share a TOC Base" ) ? void (0) : __assert_fail ("!STICaller->isUsingPCRelativeCalls() && \"PC Relative callers do not have a TOC and cannot share a TOC Base\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4632, __extension__ __PRETTY_FUNCTION__)) | |||
4632 | "PC Relative callers do not have a TOC and cannot share a TOC Base")(static_cast <bool> (!STICaller->isUsingPCRelativeCalls () && "PC Relative callers do not have a TOC and cannot share a TOC Base" ) ? void (0) : __assert_fail ("!STICaller->isUsingPCRelativeCalls() && \"PC Relative callers do not have a TOC and cannot share a TOC Base\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4632, __extension__ __PRETTY_FUNCTION__)); | |||
4633 | #endif | |||
4634 | ||||
4635 | // Callee is either a GlobalAddress or an ExternalSymbol. ExternalSymbols | |||
4636 | // don't have enough information to determine if the caller and callee share | |||
4637 | // the same TOC base, so we have to pessimistically assume they don't for | |||
4638 | // correctness. | |||
4639 | GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee); | |||
4640 | if (!G) | |||
4641 | return false; | |||
4642 | ||||
4643 | const GlobalValue *GV = G->getGlobal(); | |||
4644 | ||||
4645 | // If the callee is preemptable, then the static linker will use a plt-stub | |||
4646 | // which saves the toc to the stack, and needs a nop after the call | |||
4647 | // instruction to convert to a toc-restore. | |||
4648 | if (!TM.shouldAssumeDSOLocal(*Caller->getParent(), GV)) | |||
4649 | return false; | |||
4650 | ||||
4651 | // Functions with PC Relative enabled may clobber the TOC in the same DSO. | |||
4652 | // We may need a TOC restore in the situation where the caller requires a | |||
4653 | // valid TOC but the callee is PC Relative and does not. | |||
4654 | const Function *F = dyn_cast<Function>(GV); | |||
4655 | const GlobalAlias *Alias = dyn_cast<GlobalAlias>(GV); | |||
4656 | ||||
4657 | // If we have an Alias we can try to get the function from there. | |||
4658 | if (Alias) { | |||
4659 | const GlobalObject *GlobalObj = Alias->getBaseObject(); | |||
4660 | F = dyn_cast<Function>(GlobalObj); | |||
4661 | } | |||
4662 | ||||
4663 | // If we still have no valid function pointer we do not have enough | |||
4664 | // information to determine if the callee uses PC Relative calls so we must | |||
4665 | // assume that it does. | |||
4666 | if (!F) | |||
4667 | return false; | |||
4668 | ||||
4669 | // If the callee uses PC Relative we cannot guarantee that the callee won't | |||
4670 | // clobber the TOC of the caller and so we must assume that the two | |||
4671 | // functions do not share a TOC base. | |||
4672 | const PPCSubtarget *STICallee = &TM.getSubtarget<PPCSubtarget>(*F); | |||
4673 | if (STICallee->isUsingPCRelativeCalls()) | |||
4674 | return false; | |||
4675 | ||||
4676 | // If the GV is not a strong definition then we need to assume it can be | |||
4677 | // replaced by another function at link time. The function that replaces | |||
4678 | // it may not share the same TOC as the caller since the callee may be | |||
4679 | // replaced by a PC Relative version of the same function. | |||
4680 | if (!GV->isStrongDefinitionForLinker()) | |||
4681 | return false; | |||
4682 | ||||
4683 | // The medium and large code models are expected to provide a sufficiently | |||
4684 | // large TOC to provide all data addressing needs of a module with a | |||
4685 | // single TOC. | |||
4686 | if (CodeModel::Medium == TM.getCodeModel() || | |||
4687 | CodeModel::Large == TM.getCodeModel()) | |||
4688 | return true; | |||
4689 | ||||
4690 | // Any explicitly-specified sections and section prefixes must also match. | |||
4691 | // Also, if we're using -ffunction-sections, then each function is always in | |||
4692 | // a different section (the same is true for COMDAT functions). | |||
4693 | if (TM.getFunctionSections() || GV->hasComdat() || Caller->hasComdat() || | |||
4694 | GV->getSection() != Caller->getSection()) | |||
4695 | return false; | |||
4696 | if (const auto *F = dyn_cast<Function>(GV)) { | |||
4697 | if (F->getSectionPrefix() != Caller->getSectionPrefix()) | |||
4698 | return false; | |||
4699 | } | |||
4700 | ||||
4701 | return true; | |||
4702 | } | |||
4703 | ||||
4704 | static bool | |||
4705 | needStackSlotPassParameters(const PPCSubtarget &Subtarget, | |||
4706 | const SmallVectorImpl<ISD::OutputArg> &Outs) { | |||
4707 | assert(Subtarget.is64BitELFABI())(static_cast <bool> (Subtarget.is64BitELFABI()) ? void ( 0) : __assert_fail ("Subtarget.is64BitELFABI()", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 4707, __extension__ __PRETTY_FUNCTION__)); | |||
4708 | ||||
4709 | const unsigned PtrByteSize = 8; | |||
4710 | const unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize(); | |||
4711 | ||||
4712 | static const MCPhysReg GPR[] = { | |||
4713 | PPC::X3, PPC::X4, PPC::X5, PPC::X6, | |||
4714 | PPC::X7, PPC::X8, PPC::X9, PPC::X10, | |||
4715 | }; | |||
4716 | static const MCPhysReg VR[] = { | |||
4717 | PPC::V2, PPC::V3, PPC::V4, PPC::V5, PPC::V6, PPC::V7, PPC::V8, | |||
4718 | PPC::V9, PPC::V10, PPC::V11, PPC::V12, PPC::V13 | |||
4719 | }; | |||
4720 | ||||
4721 | const unsigned NumGPRs = array_lengthof(GPR); | |||
4722 | const unsigned NumFPRs = 13; | |||
4723 | const unsigned NumVRs = array_lengthof(VR); | |||
4724 | const unsigned ParamAreaSize = NumGPRs * PtrByteSize; | |||
4725 | ||||
4726 | unsigned NumBytes = LinkageSize; | |||
4727 | unsigned AvailableFPRs = NumFPRs; | |||
4728 | unsigned AvailableVRs = NumVRs; | |||
4729 | ||||
4730 | for (const ISD::OutputArg& Param : Outs) { | |||
4731 | if (Param.Flags.isNest()) continue; | |||
4732 | ||||
4733 | if (CalculateStackSlotUsed(Subtarget, Param.VT, Param.ArgVT, Param.Flags, | |||
4734 | PtrByteSize, LinkageSize, ParamAreaSize, | |||
4735 | NumBytes, AvailableFPRs, AvailableVRs)) | |||
4736 | return true; | |||
4737 | } | |||
4738 | return false; | |||
4739 | } | |||
4740 | ||||
4741 | static bool hasSameArgumentList(const Function *CallerFn, const CallBase &CB) { | |||
4742 | if (CB.arg_size() != CallerFn->arg_size()) | |||
4743 | return false; | |||
4744 | ||||
4745 | auto CalleeArgIter = CB.arg_begin(); | |||
4746 | auto CalleeArgEnd = CB.arg_end(); | |||
4747 | Function::const_arg_iterator CallerArgIter = CallerFn->arg_begin(); | |||
4748 | ||||
4749 | for (; CalleeArgIter != CalleeArgEnd; ++CalleeArgIter, ++CallerArgIter) { | |||
4750 | const Value* CalleeArg = *CalleeArgIter; | |||
4751 | const Value* CallerArg = &(*CallerArgIter); | |||
4752 | if (CalleeArg == CallerArg) | |||
4753 | continue; | |||
4754 | ||||
4755 | // e.g. @caller([4 x i64] %a, [4 x i64] %b) { | |||
4756 | // tail call @callee([4 x i64] undef, [4 x i64] %b) | |||
4757 | // } | |||
4758 | // 1st argument of callee is undef and has the same type as caller. | |||
4759 | if (CalleeArg->getType() == CallerArg->getType() && | |||
4760 | isa<UndefValue>(CalleeArg)) | |||
4761 | continue; | |||
4762 | ||||
4763 | return false; | |||
4764 | } | |||
4765 | ||||
4766 | return true; | |||
4767 | } | |||
4768 | ||||
4769 | // Returns true if TCO is possible between the callers and callees | |||
4770 | // calling conventions. | |||
4771 | static bool | |||
4772 | areCallingConvEligibleForTCO_64SVR4(CallingConv::ID CallerCC, | |||
4773 | CallingConv::ID CalleeCC) { | |||
4774 | // Tail calls are possible with fastcc and ccc. | |||
4775 | auto isTailCallableCC = [] (CallingConv::ID CC){ | |||
4776 | return CC == CallingConv::C || CC == CallingConv::Fast; | |||
4777 | }; | |||
4778 | if (!isTailCallableCC(CallerCC) || !isTailCallableCC(CalleeCC)) | |||
4779 | return false; | |||
4780 | ||||
4781 | // We can safely tail call both fastcc and ccc callees from a c calling | |||
4782 | // convention caller. If the caller is fastcc, we may have less stack space | |||
4783 | // than a non-fastcc caller with the same signature so disable tail-calls in | |||
4784 | // that case. | |||
4785 | return CallerCC == CallingConv::C || CallerCC == CalleeCC; | |||
4786 | } | |||
4787 | ||||
4788 | bool PPCTargetLowering::IsEligibleForTailCallOptimization_64SVR4( | |||
4789 | SDValue Callee, CallingConv::ID CalleeCC, const CallBase *CB, bool isVarArg, | |||
4790 | const SmallVectorImpl<ISD::OutputArg> &Outs, | |||
4791 | const SmallVectorImpl<ISD::InputArg> &Ins, SelectionDAG &DAG) const { | |||
4792 | bool TailCallOpt = getTargetMachine().Options.GuaranteedTailCallOpt; | |||
4793 | ||||
4794 | if (DisableSCO && !TailCallOpt) return false; | |||
4795 | ||||
4796 | // Variadic argument functions are not supported. | |||
4797 | if (isVarArg) return false; | |||
4798 | ||||
4799 | auto &Caller = DAG.getMachineFunction().getFunction(); | |||
4800 | // Check that the calling conventions are compatible for tco. | |||
4801 | if (!areCallingConvEligibleForTCO_64SVR4(Caller.getCallingConv(), CalleeCC)) | |||
4802 | return false; | |||
4803 | ||||
4804 | // Caller contains any byval parameter is not supported. | |||
4805 | if (any_of(Ins, [](const ISD::InputArg &IA) { return IA.Flags.isByVal(); })) | |||
4806 | return false; | |||
4807 | ||||
4808 | // Callee contains any byval parameter is not supported, too. | |||
4809 | // Note: This is a quick work around, because in some cases, e.g. | |||
4810 | // caller's stack size > callee's stack size, we are still able to apply | |||
4811 | // sibling call optimization. For example, gcc is able to do SCO for caller1 | |||
4812 | // in the following example, but not for caller2. | |||
4813 | // struct test { | |||
4814 | // long int a; | |||
4815 | // char ary[56]; | |||
4816 | // } gTest; | |||
4817 | // __attribute__((noinline)) int callee(struct test v, struct test *b) { | |||
4818 | // b->a = v.a; | |||
4819 | // return 0; | |||
4820 | // } | |||
4821 | // void caller1(struct test a, struct test c, struct test *b) { | |||
4822 | // callee(gTest, b); } | |||
4823 | // void caller2(struct test *b) { callee(gTest, b); } | |||
4824 | if (any_of(Outs, [](const ISD::OutputArg& OA) { return OA.Flags.isByVal(); })) | |||
4825 | return false; | |||
4826 | ||||
4827 | // If callee and caller use different calling conventions, we cannot pass | |||
4828 | // parameters on stack since offsets for the parameter area may be different. | |||
4829 | if (Caller.getCallingConv() != CalleeCC && | |||
4830 | needStackSlotPassParameters(Subtarget, Outs)) | |||
4831 | return false; | |||
4832 | ||||
4833 | // All variants of 64-bit ELF ABIs without PC-Relative addressing require that | |||
4834 | // the caller and callee share the same TOC for TCO/SCO. If the caller and | |||
4835 | // callee potentially have different TOC bases then we cannot tail call since | |||
4836 | // we need to restore the TOC pointer after the call. | |||
4837 | // ref: https://bugzilla.mozilla.org/show_bug.cgi?id=973977 | |||
4838 | // We cannot guarantee this for indirect calls or calls to external functions. | |||
4839 | // When PC-Relative addressing is used, the concept of the TOC is no longer | |||
4840 | // applicable so this check is not required. | |||
4841 | // Check first for indirect calls. | |||
4842 | if (!Subtarget.isUsingPCRelativeCalls() && | |||
4843 | !isFunctionGlobalAddress(Callee) && !isa<ExternalSymbolSDNode>(Callee)) | |||
4844 | return false; | |||
4845 | ||||
4846 | // Check if we share the TOC base. | |||
4847 | if (!Subtarget.isUsingPCRelativeCalls() && | |||
4848 | !callsShareTOCBase(&Caller, Callee, getTargetMachine())) | |||
4849 | return false; | |||
4850 | ||||
4851 | // TCO allows altering callee ABI, so we don't have to check further. | |||
4852 | if (CalleeCC == CallingConv::Fast && TailCallOpt) | |||
4853 | return true; | |||
4854 | ||||
4855 | if (DisableSCO) return false; | |||
4856 | ||||
4857 | // If callee use the same argument list that caller is using, then we can | |||
4858 | // apply SCO on this case. If it is not, then we need to check if callee needs | |||
4859 | // stack for passing arguments. | |||
4860 | // PC Relative tail calls may not have a CallBase. | |||
4861 | // If there is no CallBase we cannot verify if we have the same argument | |||
4862 | // list so assume that we don't have the same argument list. | |||
4863 | if (CB && !hasSameArgumentList(&Caller, *CB) && | |||
4864 | needStackSlotPassParameters(Subtarget, Outs)) | |||
4865 | return false; | |||
4866 | else if (!CB && needStackSlotPassParameters(Subtarget, Outs)) | |||
4867 | return false; | |||
4868 | ||||
4869 | return true; | |||
4870 | } | |||
4871 | ||||
4872 | /// IsEligibleForTailCallOptimization - Check whether the call is eligible | |||
4873 | /// for tail call optimization. Targets which want to do tail call | |||
4874 | /// optimization should implement this function. | |||
4875 | bool | |||
4876 | PPCTargetLowering::IsEligibleForTailCallOptimization(SDValue Callee, | |||
4877 | CallingConv::ID CalleeCC, | |||
4878 | bool isVarArg, | |||
4879 | const SmallVectorImpl<ISD::InputArg> &Ins, | |||
4880 | SelectionDAG& DAG) const { | |||
4881 | if (!getTargetMachine().Options.GuaranteedTailCallOpt) | |||
4882 | return false; | |||
4883 | ||||
4884 | // Variable argument functions are not supported. | |||
4885 | if (isVarArg) | |||
4886 | return false; | |||
4887 | ||||
4888 | MachineFunction &MF = DAG.getMachineFunction(); | |||
4889 | CallingConv::ID CallerCC = MF.getFunction().getCallingConv(); | |||
4890 | if (CalleeCC == CallingConv::Fast && CallerCC == CalleeCC) { | |||
4891 | // Functions containing by val parameters are not supported. | |||
4892 | for (unsigned i = 0; i != Ins.size(); i++) { | |||
4893 | ISD::ArgFlagsTy Flags = Ins[i].Flags; | |||
4894 | if (Flags.isByVal()) return false; | |||
4895 | } | |||
4896 | ||||
4897 | // Non-PIC/GOT tail calls are supported. | |||
4898 | if (getTargetMachine().getRelocationModel() != Reloc::PIC_) | |||
4899 | return true; | |||
4900 | ||||
4901 | // At the moment we can only do local tail calls (in same module, hidden | |||
4902 | // or protected) if we are generating PIC. | |||
4903 | if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) | |||
4904 | return G->getGlobal()->hasHiddenVisibility() | |||
4905 | || G->getGlobal()->hasProtectedVisibility(); | |||
4906 | } | |||
4907 | ||||
4908 | return false; | |||
4909 | } | |||
4910 | ||||
4911 | /// isCallCompatibleAddress - Return the immediate to use if the specified | |||
4912 | /// 32-bit value is representable in the immediate field of a BxA instruction. | |||
4913 | static SDNode *isBLACompatibleAddress(SDValue Op, SelectionDAG &DAG) { | |||
4914 | ConstantSDNode *C = dyn_cast<ConstantSDNode>(Op); | |||
4915 | if (!C) return nullptr; | |||
4916 | ||||
4917 | int Addr = C->getZExtValue(); | |||
4918 | if ((Addr & 3) != 0 || // Low 2 bits are implicitly zero. | |||
4919 | SignExtend32<26>(Addr) != Addr) | |||
4920 | return nullptr; // Top 6 bits have to be sext of immediate. | |||
4921 | ||||
4922 | return DAG | |||
4923 | .getConstant( | |||
4924 | (int)C->getZExtValue() >> 2, SDLoc(Op), | |||
4925 | DAG.getTargetLoweringInfo().getPointerTy(DAG.getDataLayout())) | |||
4926 | .getNode(); | |||
4927 | } | |||
4928 | ||||
4929 | namespace { | |||
4930 | ||||
4931 | struct TailCallArgumentInfo { | |||
4932 | SDValue Arg; | |||
4933 | SDValue FrameIdxOp; | |||
4934 | int FrameIdx = 0; | |||
4935 | ||||
4936 | TailCallArgumentInfo() = default; | |||
4937 | }; | |||
4938 | ||||
4939 | } // end anonymous namespace | |||
4940 | ||||
4941 | /// StoreTailCallArgumentsToStackSlot - Stores arguments to their stack slot. | |||
4942 | static void StoreTailCallArgumentsToStackSlot( | |||
4943 | SelectionDAG &DAG, SDValue Chain, | |||
4944 | const SmallVectorImpl<TailCallArgumentInfo> &TailCallArgs, | |||
4945 | SmallVectorImpl<SDValue> &MemOpChains, const SDLoc &dl) { | |||
4946 | for (unsigned i = 0, e = TailCallArgs.size(); i != e; ++i) { | |||
4947 | SDValue Arg = TailCallArgs[i].Arg; | |||
4948 | SDValue FIN = TailCallArgs[i].FrameIdxOp; | |||
4949 | int FI = TailCallArgs[i].FrameIdx; | |||
4950 | // Store relative to framepointer. | |||
4951 | MemOpChains.push_back(DAG.getStore( | |||
4952 | Chain, dl, Arg, FIN, | |||
4953 | MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI))); | |||
4954 | } | |||
4955 | } | |||
4956 | ||||
4957 | /// EmitTailCallStoreFPAndRetAddr - Move the frame pointer and return address to | |||
4958 | /// the appropriate stack slot for the tail call optimized function call. | |||
4959 | static SDValue EmitTailCallStoreFPAndRetAddr(SelectionDAG &DAG, SDValue Chain, | |||
4960 | SDValue OldRetAddr, SDValue OldFP, | |||
4961 | int SPDiff, const SDLoc &dl) { | |||
4962 | if (SPDiff) { | |||
4963 | // Calculate the new stack slot for the return address. | |||
4964 | MachineFunction &MF = DAG.getMachineFunction(); | |||
4965 | const PPCSubtarget &Subtarget = MF.getSubtarget<PPCSubtarget>(); | |||
4966 | const PPCFrameLowering *FL = Subtarget.getFrameLowering(); | |||
4967 | bool isPPC64 = Subtarget.isPPC64(); | |||
4968 | int SlotSize = isPPC64 ? 8 : 4; | |||
4969 | int NewRetAddrLoc = SPDiff + FL->getReturnSaveOffset(); | |||
4970 | int NewRetAddr = MF.getFrameInfo().CreateFixedObject(SlotSize, | |||
4971 | NewRetAddrLoc, true); | |||
4972 | EVT VT = isPPC64 ? MVT::i64 : MVT::i32; | |||
4973 | SDValue NewRetAddrFrIdx = DAG.getFrameIndex(NewRetAddr, VT); | |||
4974 | Chain = DAG.getStore(Chain, dl, OldRetAddr, NewRetAddrFrIdx, | |||
4975 | MachinePointerInfo::getFixedStack(MF, NewRetAddr)); | |||
4976 | } | |||
4977 | return Chain; | |||
4978 | } | |||
4979 | ||||
4980 | /// CalculateTailCallArgDest - Remember Argument for later processing. Calculate | |||
4981 | /// the position of the argument. | |||
4982 | static void | |||
4983 | CalculateTailCallArgDest(SelectionDAG &DAG, MachineFunction &MF, bool isPPC64, | |||
4984 | SDValue Arg, int SPDiff, unsigned ArgOffset, | |||
4985 | SmallVectorImpl<TailCallArgumentInfo>& TailCallArguments) { | |||
4986 | int Offset = ArgOffset + SPDiff; | |||
4987 | uint32_t OpSize = (Arg.getValueSizeInBits() + 7) / 8; | |||
4988 | int FI = MF.getFrameInfo().CreateFixedObject(OpSize, Offset, true); | |||
4989 | EVT VT = isPPC64 ? MVT::i64 : MVT::i32; | |||
4990 | SDValue FIN = DAG.getFrameIndex(FI, VT); | |||
4991 | TailCallArgumentInfo Info; | |||
4992 | Info.Arg = Arg; | |||
4993 | Info.FrameIdxOp = FIN; | |||
4994 | Info.FrameIdx = FI; | |||
4995 | TailCallArguments.push_back(Info); | |||
4996 | } | |||
4997 | ||||
4998 | /// EmitTCFPAndRetAddrLoad - Emit load from frame pointer and return address | |||
4999 | /// stack slot. Returns the chain as result and the loaded frame pointers in | |||
5000 | /// LROpOut/FPOpout. Used when tail calling. | |||
5001 | SDValue PPCTargetLowering::EmitTailCallLoadFPAndRetAddr( | |||
5002 | SelectionDAG &DAG, int SPDiff, SDValue Chain, SDValue &LROpOut, | |||
5003 | SDValue &FPOpOut, const SDLoc &dl) const { | |||
5004 | if (SPDiff) { | |||
5005 | // Load the LR and FP stack slot for later adjusting. | |||
5006 | EVT VT = Subtarget.isPPC64() ? MVT::i64 : MVT::i32; | |||
5007 | LROpOut = getReturnAddrFrameIndex(DAG); | |||
5008 | LROpOut = DAG.getLoad(VT, dl, Chain, LROpOut, MachinePointerInfo()); | |||
5009 | Chain = SDValue(LROpOut.getNode(), 1); | |||
5010 | } | |||
5011 | return Chain; | |||
5012 | } | |||
5013 | ||||
5014 | /// CreateCopyOfByValArgument - Make a copy of an aggregate at address specified | |||
5015 | /// by "Src" to address "Dst" of size "Size". Alignment information is | |||
5016 | /// specified by the specific parameter attribute. The copy will be passed as | |||
5017 | /// a byval function parameter. | |||
5018 | /// Sometimes what we are copying is the end of a larger object, the part that | |||
5019 | /// does not fit in registers. | |||
5020 | static SDValue CreateCopyOfByValArgument(SDValue Src, SDValue Dst, | |||
5021 | SDValue Chain, ISD::ArgFlagsTy Flags, | |||
5022 | SelectionDAG &DAG, const SDLoc &dl) { | |||
5023 | SDValue SizeNode = DAG.getConstant(Flags.getByValSize(), dl, MVT::i32); | |||
5024 | return DAG.getMemcpy(Chain, dl, Dst, Src, SizeNode, | |||
5025 | Flags.getNonZeroByValAlign(), false, false, false, | |||
5026 | MachinePointerInfo(), MachinePointerInfo()); | |||
5027 | } | |||
5028 | ||||
5029 | /// LowerMemOpCallTo - Store the argument to the stack or remember it in case of | |||
5030 | /// tail calls. | |||
5031 | static void LowerMemOpCallTo( | |||
5032 | SelectionDAG &DAG, MachineFunction &MF, SDValue Chain, SDValue Arg, | |||
5033 | SDValue PtrOff, int SPDiff, unsigned ArgOffset, bool isPPC64, | |||
5034 | bool isTailCall, bool isVector, SmallVectorImpl<SDValue> &MemOpChains, | |||
5035 | SmallVectorImpl<TailCallArgumentInfo> &TailCallArguments, const SDLoc &dl) { | |||
5036 | EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy(DAG.getDataLayout()); | |||
5037 | if (!isTailCall) { | |||
5038 | if (isVector) { | |||
5039 | SDValue StackPtr; | |||
5040 | if (isPPC64) | |||
5041 | StackPtr = DAG.getRegister(PPC::X1, MVT::i64); | |||
5042 | else | |||
5043 | StackPtr = DAG.getRegister(PPC::R1, MVT::i32); | |||
5044 | PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, | |||
5045 | DAG.getConstant(ArgOffset, dl, PtrVT)); | |||
5046 | } | |||
5047 | MemOpChains.push_back( | |||
5048 | DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo())); | |||
5049 | // Calculate and remember argument location. | |||
5050 | } else CalculateTailCallArgDest(DAG, MF, isPPC64, Arg, SPDiff, ArgOffset, | |||
5051 | TailCallArguments); | |||
5052 | } | |||
5053 | ||||
5054 | static void | |||
5055 | PrepareTailCall(SelectionDAG &DAG, SDValue &InFlag, SDValue &Chain, | |||
5056 | const SDLoc &dl, int SPDiff, unsigned NumBytes, SDValue LROp, | |||
5057 | SDValue FPOp, | |||
5058 | SmallVectorImpl<TailCallArgumentInfo> &TailCallArguments) { | |||
5059 | // Emit a sequence of copyto/copyfrom virtual registers for arguments that | |||
5060 | // might overwrite each other in case of tail call optimization. | |||
5061 | SmallVector<SDValue, 8> MemOpChains2; | |||
5062 | // Do not flag preceding copytoreg stuff together with the following stuff. | |||
5063 | InFlag = SDValue(); | |||
5064 | StoreTailCallArgumentsToStackSlot(DAG, Chain, TailCallArguments, | |||
5065 | MemOpChains2, dl); | |||
5066 | if (!MemOpChains2.empty()) | |||
5067 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains2); | |||
5068 | ||||
5069 | // Store the return address to the appropriate stack slot. | |||
5070 | Chain = EmitTailCallStoreFPAndRetAddr(DAG, Chain, LROp, FPOp, SPDiff, dl); | |||
5071 | ||||
5072 | // Emit callseq_end just before tailcall node. | |||
5073 | Chain = DAG.getCALLSEQ_END(Chain, DAG.getIntPtrConstant(NumBytes, dl, true), | |||
5074 | DAG.getIntPtrConstant(0, dl, true), InFlag, dl); | |||
5075 | InFlag = Chain.getValue(1); | |||
5076 | } | |||
5077 | ||||
5078 | // Is this global address that of a function that can be called by name? (as | |||
5079 | // opposed to something that must hold a descriptor for an indirect call). | |||
5080 | static bool isFunctionGlobalAddress(SDValue Callee) { | |||
5081 | if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) { | |||
5082 | if (Callee.getOpcode() == ISD::GlobalTLSAddress || | |||
5083 | Callee.getOpcode() == ISD::TargetGlobalTLSAddress) | |||
5084 | return false; | |||
5085 | ||||
5086 | return G->getGlobal()->getValueType()->isFunctionTy(); | |||
5087 | } | |||
5088 | ||||
5089 | return false; | |||
5090 | } | |||
5091 | ||||
5092 | SDValue PPCTargetLowering::LowerCallResult( | |||
5093 | SDValue Chain, SDValue InFlag, CallingConv::ID CallConv, bool isVarArg, | |||
5094 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
5095 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const { | |||
5096 | SmallVector<CCValAssign, 16> RVLocs; | |||
5097 | CCState CCRetInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs, | |||
5098 | *DAG.getContext()); | |||
5099 | ||||
5100 | CCRetInfo.AnalyzeCallResult( | |||
5101 | Ins, (Subtarget.isSVR4ABI() && CallConv == CallingConv::Cold) | |||
5102 | ? RetCC_PPC_Cold | |||
5103 | : RetCC_PPC); | |||
5104 | ||||
5105 | // Copy all of the result registers out of their specified physreg. | |||
5106 | for (unsigned i = 0, e = RVLocs.size(); i != e; ++i) { | |||
5107 | CCValAssign &VA = RVLocs[i]; | |||
5108 | assert(VA.isRegLoc() && "Can only return in registers!")(static_cast <bool> (VA.isRegLoc() && "Can only return in registers!" ) ? void (0) : __assert_fail ("VA.isRegLoc() && \"Can only return in registers!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5108, __extension__ __PRETTY_FUNCTION__)); | |||
5109 | ||||
5110 | SDValue Val; | |||
5111 | ||||
5112 | if (Subtarget.hasSPE() && VA.getLocVT() == MVT::f64) { | |||
5113 | SDValue Lo = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32, | |||
5114 | InFlag); | |||
5115 | Chain = Lo.getValue(1); | |||
5116 | InFlag = Lo.getValue(2); | |||
5117 | VA = RVLocs[++i]; // skip ahead to next loc | |||
5118 | SDValue Hi = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32, | |||
5119 | InFlag); | |||
5120 | Chain = Hi.getValue(1); | |||
5121 | InFlag = Hi.getValue(2); | |||
5122 | if (!Subtarget.isLittleEndian()) | |||
5123 | std::swap (Lo, Hi); | |||
5124 | Val = DAG.getNode(PPCISD::BUILD_SPE64, dl, MVT::f64, Lo, Hi); | |||
5125 | } else { | |||
5126 | Val = DAG.getCopyFromReg(Chain, dl, | |||
5127 | VA.getLocReg(), VA.getLocVT(), InFlag); | |||
5128 | Chain = Val.getValue(1); | |||
5129 | InFlag = Val.getValue(2); | |||
5130 | } | |||
5131 | ||||
5132 | switch (VA.getLocInfo()) { | |||
5133 | default: llvm_unreachable("Unknown loc info!")::llvm::llvm_unreachable_internal("Unknown loc info!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5133); | |||
5134 | case CCValAssign::Full: break; | |||
5135 | case CCValAssign::AExt: | |||
5136 | Val = DAG.getNode(ISD::TRUNCATE, dl, VA.getValVT(), Val); | |||
5137 | break; | |||
5138 | case CCValAssign::ZExt: | |||
5139 | Val = DAG.getNode(ISD::AssertZext, dl, VA.getLocVT(), Val, | |||
5140 | DAG.getValueType(VA.getValVT())); | |||
5141 | Val = DAG.getNode(ISD::TRUNCATE, dl, VA.getValVT(), Val); | |||
5142 | break; | |||
5143 | case CCValAssign::SExt: | |||
5144 | Val = DAG.getNode(ISD::AssertSext, dl, VA.getLocVT(), Val, | |||
5145 | DAG.getValueType(VA.getValVT())); | |||
5146 | Val = DAG.getNode(ISD::TRUNCATE, dl, VA.getValVT(), Val); | |||
5147 | break; | |||
5148 | } | |||
5149 | ||||
5150 | InVals.push_back(Val); | |||
5151 | } | |||
5152 | ||||
5153 | return Chain; | |||
5154 | } | |||
5155 | ||||
5156 | static bool isIndirectCall(const SDValue &Callee, SelectionDAG &DAG, | |||
5157 | const PPCSubtarget &Subtarget, bool isPatchPoint) { | |||
5158 | // PatchPoint calls are not indirect. | |||
5159 | if (isPatchPoint) | |||
5160 | return false; | |||
5161 | ||||
5162 | if (isFunctionGlobalAddress(Callee) || isa<ExternalSymbolSDNode>(Callee)) | |||
5163 | return false; | |||
5164 | ||||
5165 | // Darwin, and 32-bit ELF can use a BLA. The descriptor based ABIs can not | |||
5166 | // becuase the immediate function pointer points to a descriptor instead of | |||
5167 | // a function entry point. The ELFv2 ABI cannot use a BLA because the function | |||
5168 | // pointer immediate points to the global entry point, while the BLA would | |||
5169 | // need to jump to the local entry point (see rL211174). | |||
5170 | if (!Subtarget.usesFunctionDescriptors() && !Subtarget.isELFv2ABI() && | |||
5171 | isBLACompatibleAddress(Callee, DAG)) | |||
5172 | return false; | |||
5173 | ||||
5174 | return true; | |||
5175 | } | |||
5176 | ||||
5177 | // AIX and 64-bit ELF ABIs w/o PCRel require a TOC save/restore around calls. | |||
5178 | static inline bool isTOCSaveRestoreRequired(const PPCSubtarget &Subtarget) { | |||
5179 | return Subtarget.isAIXABI() || | |||
5180 | (Subtarget.is64BitELFABI() && !Subtarget.isUsingPCRelativeCalls()); | |||
5181 | } | |||
5182 | ||||
5183 | static unsigned getCallOpcode(PPCTargetLowering::CallFlags CFlags, | |||
5184 | const Function &Caller, | |||
5185 | const SDValue &Callee, | |||
5186 | const PPCSubtarget &Subtarget, | |||
5187 | const TargetMachine &TM) { | |||
5188 | if (CFlags.IsTailCall) | |||
5189 | return PPCISD::TC_RETURN; | |||
5190 | ||||
5191 | // This is a call through a function pointer. | |||
5192 | if (CFlags.IsIndirect) { | |||
5193 | // AIX and the 64-bit ELF ABIs need to maintain the TOC pointer accross | |||
5194 | // indirect calls. The save of the caller's TOC pointer to the stack will be | |||
5195 | // inserted into the DAG as part of call lowering. The restore of the TOC | |||
5196 | // pointer is modeled by using a pseudo instruction for the call opcode that | |||
5197 | // represents the 2 instruction sequence of an indirect branch and link, | |||
5198 | // immediately followed by a load of the TOC pointer from the the stack save | |||
5199 | // slot into gpr2. For 64-bit ELFv2 ABI with PCRel, do not restore the TOC | |||
5200 | // as it is not saved or used. | |||
5201 | return isTOCSaveRestoreRequired(Subtarget) ? PPCISD::BCTRL_LOAD_TOC | |||
5202 | : PPCISD::BCTRL; | |||
5203 | } | |||
5204 | ||||
5205 | if (Subtarget.isUsingPCRelativeCalls()) { | |||
5206 | assert(Subtarget.is64BitELFABI() && "PC Relative is only on ELF ABI.")(static_cast <bool> (Subtarget.is64BitELFABI() && "PC Relative is only on ELF ABI.") ? void (0) : __assert_fail ("Subtarget.is64BitELFABI() && \"PC Relative is only on ELF ABI.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5206, __extension__ __PRETTY_FUNCTION__)); | |||
5207 | return PPCISD::CALL_NOTOC; | |||
5208 | } | |||
5209 | ||||
5210 | // The ABIs that maintain a TOC pointer accross calls need to have a nop | |||
5211 | // immediately following the call instruction if the caller and callee may | |||
5212 | // have different TOC bases. At link time if the linker determines the calls | |||
5213 | // may not share a TOC base, the call is redirected to a trampoline inserted | |||
5214 | // by the linker. The trampoline will (among other things) save the callers | |||
5215 | // TOC pointer at an ABI designated offset in the linkage area and the linker | |||
5216 | // will rewrite the nop to be a load of the TOC pointer from the linkage area | |||
5217 | // into gpr2. | |||
5218 | if (Subtarget.isAIXABI() || Subtarget.is64BitELFABI()) | |||
5219 | return callsShareTOCBase(&Caller, Callee, TM) ? PPCISD::CALL | |||
5220 | : PPCISD::CALL_NOP; | |||
5221 | ||||
5222 | return PPCISD::CALL; | |||
5223 | } | |||
5224 | ||||
5225 | static SDValue transformCallee(const SDValue &Callee, SelectionDAG &DAG, | |||
5226 | const SDLoc &dl, const PPCSubtarget &Subtarget) { | |||
5227 | if (!Subtarget.usesFunctionDescriptors() && !Subtarget.isELFv2ABI()) | |||
5228 | if (SDNode *Dest = isBLACompatibleAddress(Callee, DAG)) | |||
5229 | return SDValue(Dest, 0); | |||
5230 | ||||
5231 | // Returns true if the callee is local, and false otherwise. | |||
5232 | auto isLocalCallee = [&]() { | |||
5233 | const GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee); | |||
5234 | const Module *Mod = DAG.getMachineFunction().getFunction().getParent(); | |||
5235 | const GlobalValue *GV = G ? G->getGlobal() : nullptr; | |||
5236 | ||||
5237 | return DAG.getTarget().shouldAssumeDSOLocal(*Mod, GV) && | |||
5238 | !dyn_cast_or_null<GlobalIFunc>(GV); | |||
5239 | }; | |||
5240 | ||||
5241 | // The PLT is only used in 32-bit ELF PIC mode. Attempting to use the PLT in | |||
5242 | // a static relocation model causes some versions of GNU LD (2.17.50, at | |||
5243 | // least) to force BSS-PLT, instead of secure-PLT, even if all objects are | |||
5244 | // built with secure-PLT. | |||
5245 | bool UsePlt = | |||
5246 | Subtarget.is32BitELFABI() && !isLocalCallee() && | |||
5247 | Subtarget.getTargetMachine().getRelocationModel() == Reloc::PIC_; | |||
5248 | ||||
5249 | const auto getAIXFuncEntryPointSymbolSDNode = [&](const GlobalValue *GV) { | |||
5250 | const TargetMachine &TM = Subtarget.getTargetMachine(); | |||
5251 | const TargetLoweringObjectFile *TLOF = TM.getObjFileLowering(); | |||
5252 | MCSymbolXCOFF *S = | |||
5253 | cast<MCSymbolXCOFF>(TLOF->getFunctionEntryPointSymbol(GV, TM)); | |||
5254 | ||||
5255 | MVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy(DAG.getDataLayout()); | |||
5256 | return DAG.getMCSymbol(S, PtrVT); | |||
5257 | }; | |||
5258 | ||||
5259 | if (isFunctionGlobalAddress(Callee)) { | |||
5260 | const GlobalValue *GV = cast<GlobalAddressSDNode>(Callee)->getGlobal(); | |||
5261 | ||||
5262 | if (Subtarget.isAIXABI()) { | |||
5263 | assert(!isa<GlobalIFunc>(GV) && "IFunc is not supported on AIX.")(static_cast <bool> (!isa<GlobalIFunc>(GV) && "IFunc is not supported on AIX.") ? void (0) : __assert_fail ("!isa<GlobalIFunc>(GV) && \"IFunc is not supported on AIX.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5263, __extension__ __PRETTY_FUNCTION__)); | |||
5264 | return getAIXFuncEntryPointSymbolSDNode(GV); | |||
5265 | } | |||
5266 | return DAG.getTargetGlobalAddress(GV, dl, Callee.getValueType(), 0, | |||
5267 | UsePlt ? PPCII::MO_PLT : 0); | |||
5268 | } | |||
5269 | ||||
5270 | if (ExternalSymbolSDNode *S = dyn_cast<ExternalSymbolSDNode>(Callee)) { | |||
5271 | const char *SymName = S->getSymbol(); | |||
5272 | if (Subtarget.isAIXABI()) { | |||
5273 | // If there exists a user-declared function whose name is the same as the | |||
5274 | // ExternalSymbol's, then we pick up the user-declared version. | |||
5275 | const Module *Mod = DAG.getMachineFunction().getFunction().getParent(); | |||
5276 | if (const Function *F = | |||
5277 | dyn_cast_or_null<Function>(Mod->getNamedValue(SymName))) | |||
5278 | return getAIXFuncEntryPointSymbolSDNode(F); | |||
5279 | ||||
5280 | // On AIX, direct function calls reference the symbol for the function's | |||
5281 | // entry point, which is named by prepending a "." before the function's | |||
5282 | // C-linkage name. A Qualname is returned here because an external | |||
5283 | // function entry point is a csect with XTY_ER property. | |||
5284 | const auto getExternalFunctionEntryPointSymbol = [&](StringRef SymName) { | |||
5285 | auto &Context = DAG.getMachineFunction().getMMI().getContext(); | |||
5286 | MCSectionXCOFF *Sec = Context.getXCOFFSection( | |||
5287 | (Twine(".") + Twine(SymName)).str(), SectionKind::getMetadata(), | |||
5288 | XCOFF::CsectProperties(XCOFF::XMC_PR, XCOFF::XTY_ER)); | |||
5289 | return Sec->getQualNameSymbol(); | |||
5290 | }; | |||
5291 | ||||
5292 | SymName = getExternalFunctionEntryPointSymbol(SymName)->getName().data(); | |||
5293 | } | |||
5294 | return DAG.getTargetExternalSymbol(SymName, Callee.getValueType(), | |||
5295 | UsePlt ? PPCII::MO_PLT : 0); | |||
5296 | } | |||
5297 | ||||
5298 | // No transformation needed. | |||
5299 | assert(Callee.getNode() && "What no callee?")(static_cast <bool> (Callee.getNode() && "What no callee?" ) ? void (0) : __assert_fail ("Callee.getNode() && \"What no callee?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5299, __extension__ __PRETTY_FUNCTION__)); | |||
5300 | return Callee; | |||
5301 | } | |||
5302 | ||||
5303 | static SDValue getOutputChainFromCallSeq(SDValue CallSeqStart) { | |||
5304 | assert(CallSeqStart.getOpcode() == ISD::CALLSEQ_START &&(static_cast <bool> (CallSeqStart.getOpcode() == ISD::CALLSEQ_START && "Expected a CALLSEQ_STARTSDNode.") ? void (0) : __assert_fail ("CallSeqStart.getOpcode() == ISD::CALLSEQ_START && \"Expected a CALLSEQ_STARTSDNode.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5305, __extension__ __PRETTY_FUNCTION__)) | |||
5305 | "Expected a CALLSEQ_STARTSDNode.")(static_cast <bool> (CallSeqStart.getOpcode() == ISD::CALLSEQ_START && "Expected a CALLSEQ_STARTSDNode.") ? void (0) : __assert_fail ("CallSeqStart.getOpcode() == ISD::CALLSEQ_START && \"Expected a CALLSEQ_STARTSDNode.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5305, __extension__ __PRETTY_FUNCTION__)); | |||
5306 | ||||
5307 | // The last operand is the chain, except when the node has glue. If the node | |||
5308 | // has glue, then the last operand is the glue, and the chain is the second | |||
5309 | // last operand. | |||
5310 | SDValue LastValue = CallSeqStart.getValue(CallSeqStart->getNumValues() - 1); | |||
5311 | if (LastValue.getValueType() != MVT::Glue) | |||
5312 | return LastValue; | |||
5313 | ||||
5314 | return CallSeqStart.getValue(CallSeqStart->getNumValues() - 2); | |||
5315 | } | |||
5316 | ||||
5317 | // Creates the node that moves a functions address into the count register | |||
5318 | // to prepare for an indirect call instruction. | |||
5319 | static void prepareIndirectCall(SelectionDAG &DAG, SDValue &Callee, | |||
5320 | SDValue &Glue, SDValue &Chain, | |||
5321 | const SDLoc &dl) { | |||
5322 | SDValue MTCTROps[] = {Chain, Callee, Glue}; | |||
5323 | EVT ReturnTypes[] = {MVT::Other, MVT::Glue}; | |||
5324 | Chain = DAG.getNode(PPCISD::MTCTR, dl, makeArrayRef(ReturnTypes, 2), | |||
5325 | makeArrayRef(MTCTROps, Glue.getNode() ? 3 : 2)); | |||
5326 | // The glue is the second value produced. | |||
5327 | Glue = Chain.getValue(1); | |||
5328 | } | |||
5329 | ||||
5330 | static void prepareDescriptorIndirectCall(SelectionDAG &DAG, SDValue &Callee, | |||
5331 | SDValue &Glue, SDValue &Chain, | |||
5332 | SDValue CallSeqStart, | |||
5333 | const CallBase *CB, const SDLoc &dl, | |||
5334 | bool hasNest, | |||
5335 | const PPCSubtarget &Subtarget) { | |||
5336 | // Function pointers in the 64-bit SVR4 ABI do not point to the function | |||
5337 | // entry point, but to the function descriptor (the function entry point | |||
5338 | // address is part of the function descriptor though). | |||
5339 | // The function descriptor is a three doubleword structure with the | |||
5340 | // following fields: function entry point, TOC base address and | |||
5341 | // environment pointer. | |||
5342 | // Thus for a call through a function pointer, the following actions need | |||
5343 | // to be performed: | |||
5344 | // 1. Save the TOC of the caller in the TOC save area of its stack | |||
5345 | // frame (this is done in LowerCall_Darwin() or LowerCall_64SVR4()). | |||
5346 | // 2. Load the address of the function entry point from the function | |||
5347 | // descriptor. | |||
5348 | // 3. Load the TOC of the callee from the function descriptor into r2. | |||
5349 | // 4. Load the environment pointer from the function descriptor into | |||
5350 | // r11. | |||
5351 | // 5. Branch to the function entry point address. | |||
5352 | // 6. On return of the callee, the TOC of the caller needs to be | |||
5353 | // restored (this is done in FinishCall()). | |||
5354 | // | |||
5355 | // The loads are scheduled at the beginning of the call sequence, and the | |||
5356 | // register copies are flagged together to ensure that no other | |||
5357 | // operations can be scheduled in between. E.g. without flagging the | |||
5358 | // copies together, a TOC access in the caller could be scheduled between | |||
5359 | // the assignment of the callee TOC and the branch to the callee, which leads | |||
5360 | // to incorrect code. | |||
5361 | ||||
5362 | // Start by loading the function address from the descriptor. | |||
5363 | SDValue LDChain = getOutputChainFromCallSeq(CallSeqStart); | |||
5364 | auto MMOFlags = Subtarget.hasInvariantFunctionDescriptors() | |||
5365 | ? (MachineMemOperand::MODereferenceable | | |||
5366 | MachineMemOperand::MOInvariant) | |||
5367 | : MachineMemOperand::MONone; | |||
5368 | ||||
5369 | MachinePointerInfo MPI(CB ? CB->getCalledOperand() : nullptr); | |||
5370 | ||||
5371 | // Registers used in building the DAG. | |||
5372 | const MCRegister EnvPtrReg = Subtarget.getEnvironmentPointerRegister(); | |||
5373 | const MCRegister TOCReg = Subtarget.getTOCPointerRegister(); | |||
5374 | ||||
5375 | // Offsets of descriptor members. | |||
5376 | const unsigned TOCAnchorOffset = Subtarget.descriptorTOCAnchorOffset(); | |||
5377 | const unsigned EnvPtrOffset = Subtarget.descriptorEnvironmentPointerOffset(); | |||
5378 | ||||
5379 | const MVT RegVT = Subtarget.isPPC64() ? MVT::i64 : MVT::i32; | |||
5380 | const unsigned Alignment = Subtarget.isPPC64() ? 8 : 4; | |||
5381 | ||||
5382 | // One load for the functions entry point address. | |||
5383 | SDValue LoadFuncPtr = DAG.getLoad(RegVT, dl, LDChain, Callee, MPI, | |||
5384 | Alignment, MMOFlags); | |||
5385 | ||||
5386 | // One for loading the TOC anchor for the module that contains the called | |||
5387 | // function. | |||
5388 | SDValue TOCOff = DAG.getIntPtrConstant(TOCAnchorOffset, dl); | |||
5389 | SDValue AddTOC = DAG.getNode(ISD::ADD, dl, RegVT, Callee, TOCOff); | |||
5390 | SDValue TOCPtr = | |||
5391 | DAG.getLoad(RegVT, dl, LDChain, AddTOC, | |||
5392 | MPI.getWithOffset(TOCAnchorOffset), Alignment, MMOFlags); | |||
5393 | ||||
5394 | // One for loading the environment pointer. | |||
5395 | SDValue PtrOff = DAG.getIntPtrConstant(EnvPtrOffset, dl); | |||
5396 | SDValue AddPtr = DAG.getNode(ISD::ADD, dl, RegVT, Callee, PtrOff); | |||
5397 | SDValue LoadEnvPtr = | |||
5398 | DAG.getLoad(RegVT, dl, LDChain, AddPtr, | |||
5399 | MPI.getWithOffset(EnvPtrOffset), Alignment, MMOFlags); | |||
5400 | ||||
5401 | ||||
5402 | // Then copy the newly loaded TOC anchor to the TOC pointer. | |||
5403 | SDValue TOCVal = DAG.getCopyToReg(Chain, dl, TOCReg, TOCPtr, Glue); | |||
5404 | Chain = TOCVal.getValue(0); | |||
5405 | Glue = TOCVal.getValue(1); | |||
5406 | ||||
5407 | // If the function call has an explicit 'nest' parameter, it takes the | |||
5408 | // place of the environment pointer. | |||
5409 | assert((!hasNest || !Subtarget.isAIXABI()) &&(static_cast <bool> ((!hasNest || !Subtarget.isAIXABI() ) && "Nest parameter is not supported on AIX.") ? void (0) : __assert_fail ("(!hasNest || !Subtarget.isAIXABI()) && \"Nest parameter is not supported on AIX.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5410, __extension__ __PRETTY_FUNCTION__)) | |||
5410 | "Nest parameter is not supported on AIX.")(static_cast <bool> ((!hasNest || !Subtarget.isAIXABI() ) && "Nest parameter is not supported on AIX.") ? void (0) : __assert_fail ("(!hasNest || !Subtarget.isAIXABI()) && \"Nest parameter is not supported on AIX.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5410, __extension__ __PRETTY_FUNCTION__)); | |||
5411 | if (!hasNest) { | |||
5412 | SDValue EnvVal = DAG.getCopyToReg(Chain, dl, EnvPtrReg, LoadEnvPtr, Glue); | |||
5413 | Chain = EnvVal.getValue(0); | |||
5414 | Glue = EnvVal.getValue(1); | |||
5415 | } | |||
5416 | ||||
5417 | // The rest of the indirect call sequence is the same as the non-descriptor | |||
5418 | // DAG. | |||
5419 | prepareIndirectCall(DAG, LoadFuncPtr, Glue, Chain, dl); | |||
5420 | } | |||
5421 | ||||
5422 | static void | |||
5423 | buildCallOperands(SmallVectorImpl<SDValue> &Ops, | |||
5424 | PPCTargetLowering::CallFlags CFlags, const SDLoc &dl, | |||
5425 | SelectionDAG &DAG, | |||
5426 | SmallVector<std::pair<unsigned, SDValue>, 8> &RegsToPass, | |||
5427 | SDValue Glue, SDValue Chain, SDValue &Callee, int SPDiff, | |||
5428 | const PPCSubtarget &Subtarget) { | |||
5429 | const bool IsPPC64 = Subtarget.isPPC64(); | |||
5430 | // MVT for a general purpose register. | |||
5431 | const MVT RegVT = IsPPC64 ? MVT::i64 : MVT::i32; | |||
5432 | ||||
5433 | // First operand is always the chain. | |||
5434 | Ops.push_back(Chain); | |||
5435 | ||||
5436 | // If it's a direct call pass the callee as the second operand. | |||
5437 | if (!CFlags.IsIndirect) | |||
5438 | Ops.push_back(Callee); | |||
5439 | else { | |||
5440 | assert(!CFlags.IsPatchPoint && "Patch point calls are not indirect.")(static_cast <bool> (!CFlags.IsPatchPoint && "Patch point calls are not indirect." ) ? void (0) : __assert_fail ("!CFlags.IsPatchPoint && \"Patch point calls are not indirect.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5440, __extension__ __PRETTY_FUNCTION__)); | |||
5441 | ||||
5442 | // For the TOC based ABIs, we have saved the TOC pointer to the linkage area | |||
5443 | // on the stack (this would have been done in `LowerCall_64SVR4` or | |||
5444 | // `LowerCall_AIX`). The call instruction is a pseudo instruction that | |||
5445 | // represents both the indirect branch and a load that restores the TOC | |||
5446 | // pointer from the linkage area. The operand for the TOC restore is an add | |||
5447 | // of the TOC save offset to the stack pointer. This must be the second | |||
5448 | // operand: after the chain input but before any other variadic arguments. | |||
5449 | // For 64-bit ELFv2 ABI with PCRel, do not restore the TOC as it is not | |||
5450 | // saved or used. | |||
5451 | if (isTOCSaveRestoreRequired(Subtarget)) { | |||
5452 | const MCRegister StackPtrReg = Subtarget.getStackPointerRegister(); | |||
5453 | ||||
5454 | SDValue StackPtr = DAG.getRegister(StackPtrReg, RegVT); | |||
5455 | unsigned TOCSaveOffset = Subtarget.getFrameLowering()->getTOCSaveOffset(); | |||
5456 | SDValue TOCOff = DAG.getIntPtrConstant(TOCSaveOffset, dl); | |||
5457 | SDValue AddTOC = DAG.getNode(ISD::ADD, dl, RegVT, StackPtr, TOCOff); | |||
5458 | Ops.push_back(AddTOC); | |||
5459 | } | |||
5460 | ||||
5461 | // Add the register used for the environment pointer. | |||
5462 | if (Subtarget.usesFunctionDescriptors() && !CFlags.HasNest) | |||
5463 | Ops.push_back(DAG.getRegister(Subtarget.getEnvironmentPointerRegister(), | |||
5464 | RegVT)); | |||
5465 | ||||
5466 | ||||
5467 | // Add CTR register as callee so a bctr can be emitted later. | |||
5468 | if (CFlags.IsTailCall) | |||
5469 | Ops.push_back(DAG.getRegister(IsPPC64 ? PPC::CTR8 : PPC::CTR, RegVT)); | |||
5470 | } | |||
5471 | ||||
5472 | // If this is a tail call add stack pointer delta. | |||
5473 | if (CFlags.IsTailCall) | |||
5474 | Ops.push_back(DAG.getConstant(SPDiff, dl, MVT::i32)); | |||
5475 | ||||
5476 | // Add argument registers to the end of the list so that they are known live | |||
5477 | // into the call. | |||
5478 | for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) | |||
5479 | Ops.push_back(DAG.getRegister(RegsToPass[i].first, | |||
5480 | RegsToPass[i].second.getValueType())); | |||
5481 | ||||
5482 | // We cannot add R2/X2 as an operand here for PATCHPOINT, because there is | |||
5483 | // no way to mark dependencies as implicit here. | |||
5484 | // We will add the R2/X2 dependency in EmitInstrWithCustomInserter. | |||
5485 | if ((Subtarget.is64BitELFABI() || Subtarget.isAIXABI()) && | |||
5486 | !CFlags.IsPatchPoint && !Subtarget.isUsingPCRelativeCalls()) | |||
5487 | Ops.push_back(DAG.getRegister(Subtarget.getTOCPointerRegister(), RegVT)); | |||
5488 | ||||
5489 | // Add implicit use of CR bit 6 for 32-bit SVR4 vararg calls | |||
5490 | if (CFlags.IsVarArg && Subtarget.is32BitELFABI()) | |||
5491 | Ops.push_back(DAG.getRegister(PPC::CR1EQ, MVT::i32)); | |||
5492 | ||||
5493 | // Add a register mask operand representing the call-preserved registers. | |||
5494 | const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo(); | |||
5495 | const uint32_t *Mask = | |||
5496 | TRI->getCallPreservedMask(DAG.getMachineFunction(), CFlags.CallConv); | |||
5497 | assert(Mask && "Missing call preserved mask for calling convention")(static_cast <bool> (Mask && "Missing call preserved mask for calling convention" ) ? void (0) : __assert_fail ("Mask && \"Missing call preserved mask for calling convention\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5497, __extension__ __PRETTY_FUNCTION__)); | |||
5498 | Ops.push_back(DAG.getRegisterMask(Mask)); | |||
5499 | ||||
5500 | // If the glue is valid, it is the last operand. | |||
5501 | if (Glue.getNode()) | |||
5502 | Ops.push_back(Glue); | |||
5503 | } | |||
5504 | ||||
5505 | SDValue PPCTargetLowering::FinishCall( | |||
5506 | CallFlags CFlags, const SDLoc &dl, SelectionDAG &DAG, | |||
5507 | SmallVector<std::pair<unsigned, SDValue>, 8> &RegsToPass, SDValue Glue, | |||
5508 | SDValue Chain, SDValue CallSeqStart, SDValue &Callee, int SPDiff, | |||
5509 | unsigned NumBytes, const SmallVectorImpl<ISD::InputArg> &Ins, | |||
5510 | SmallVectorImpl<SDValue> &InVals, const CallBase *CB) const { | |||
5511 | ||||
5512 | if ((Subtarget.is64BitELFABI() && !Subtarget.isUsingPCRelativeCalls()) || | |||
5513 | Subtarget.isAIXABI()) | |||
5514 | setUsesTOCBasePtr(DAG); | |||
5515 | ||||
5516 | unsigned CallOpc = | |||
5517 | getCallOpcode(CFlags, DAG.getMachineFunction().getFunction(), Callee, | |||
5518 | Subtarget, DAG.getTarget()); | |||
5519 | ||||
5520 | if (!CFlags.IsIndirect) | |||
5521 | Callee = transformCallee(Callee, DAG, dl, Subtarget); | |||
5522 | else if (Subtarget.usesFunctionDescriptors()) | |||
5523 | prepareDescriptorIndirectCall(DAG, Callee, Glue, Chain, CallSeqStart, CB, | |||
5524 | dl, CFlags.HasNest, Subtarget); | |||
5525 | else | |||
5526 | prepareIndirectCall(DAG, Callee, Glue, Chain, dl); | |||
5527 | ||||
5528 | // Build the operand list for the call instruction. | |||
5529 | SmallVector<SDValue, 8> Ops; | |||
5530 | buildCallOperands(Ops, CFlags, dl, DAG, RegsToPass, Glue, Chain, Callee, | |||
5531 | SPDiff, Subtarget); | |||
5532 | ||||
5533 | // Emit tail call. | |||
5534 | if (CFlags.IsTailCall) { | |||
5535 | // Indirect tail call when using PC Relative calls do not have the same | |||
5536 | // constraints. | |||
5537 | assert(((Callee.getOpcode() == ISD::Register &&(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5538 | cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) ||(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5539 | Callee.getOpcode() == ISD::TargetExternalSymbol ||(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5540 | Callee.getOpcode() == ISD::TargetGlobalAddress ||(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5541 | isa<ConstantSDNode>(Callee) ||(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5542 | (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) &&(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5543 | "Expecting a global address, external symbol, absolute value, "(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5544 | "register or an indirect tail call when PC Relative calls are "(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)) | |||
5545 | "used.")(static_cast <bool> (((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa< ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget .isUsingPCRelativeCalls())) && "Expecting a global address, external symbol, absolute value, " "register or an indirect tail call when PC Relative calls are " "used.") ? void (0) : __assert_fail ("((Callee.getOpcode() == ISD::Register && cast<RegisterSDNode>(Callee)->getReg() == PPC::CTR) || Callee.getOpcode() == ISD::TargetExternalSymbol || Callee.getOpcode() == ISD::TargetGlobalAddress || isa<ConstantSDNode>(Callee) || (CFlags.IsIndirect && Subtarget.isUsingPCRelativeCalls())) && \"Expecting a global address, external symbol, absolute value, \" \"register or an indirect tail call when PC Relative calls are \" \"used.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5545, __extension__ __PRETTY_FUNCTION__)); | |||
5546 | // PC Relative calls also use TC_RETURN as the way to mark tail calls. | |||
5547 | assert(CallOpc == PPCISD::TC_RETURN &&(static_cast <bool> (CallOpc == PPCISD::TC_RETURN && "Unexpected call opcode for a tail call.") ? void (0) : __assert_fail ("CallOpc == PPCISD::TC_RETURN && \"Unexpected call opcode for a tail call.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5548, __extension__ __PRETTY_FUNCTION__)) | |||
5548 | "Unexpected call opcode for a tail call.")(static_cast <bool> (CallOpc == PPCISD::TC_RETURN && "Unexpected call opcode for a tail call.") ? void (0) : __assert_fail ("CallOpc == PPCISD::TC_RETURN && \"Unexpected call opcode for a tail call.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5548, __extension__ __PRETTY_FUNCTION__)); | |||
5549 | DAG.getMachineFunction().getFrameInfo().setHasTailCall(); | |||
5550 | return DAG.getNode(CallOpc, dl, MVT::Other, Ops); | |||
5551 | } | |||
5552 | ||||
5553 | std::array<EVT, 2> ReturnTypes = {{MVT::Other, MVT::Glue}}; | |||
5554 | Chain = DAG.getNode(CallOpc, dl, ReturnTypes, Ops); | |||
5555 | DAG.addNoMergeSiteInfo(Chain.getNode(), CFlags.NoMerge); | |||
5556 | Glue = Chain.getValue(1); | |||
5557 | ||||
5558 | // When performing tail call optimization the callee pops its arguments off | |||
5559 | // the stack. Account for this here so these bytes can be pushed back on in | |||
5560 | // PPCFrameLowering::eliminateCallFramePseudoInstr. | |||
5561 | int BytesCalleePops = (CFlags.CallConv == CallingConv::Fast && | |||
5562 | getTargetMachine().Options.GuaranteedTailCallOpt) | |||
5563 | ? NumBytes | |||
5564 | : 0; | |||
5565 | ||||
5566 | Chain = DAG.getCALLSEQ_END(Chain, DAG.getIntPtrConstant(NumBytes, dl, true), | |||
5567 | DAG.getIntPtrConstant(BytesCalleePops, dl, true), | |||
5568 | Glue, dl); | |||
5569 | Glue = Chain.getValue(1); | |||
5570 | ||||
5571 | return LowerCallResult(Chain, Glue, CFlags.CallConv, CFlags.IsVarArg, Ins, dl, | |||
5572 | DAG, InVals); | |||
5573 | } | |||
5574 | ||||
5575 | SDValue | |||
5576 | PPCTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI, | |||
5577 | SmallVectorImpl<SDValue> &InVals) const { | |||
5578 | SelectionDAG &DAG = CLI.DAG; | |||
5579 | SDLoc &dl = CLI.DL; | |||
5580 | SmallVectorImpl<ISD::OutputArg> &Outs = CLI.Outs; | |||
5581 | SmallVectorImpl<SDValue> &OutVals = CLI.OutVals; | |||
5582 | SmallVectorImpl<ISD::InputArg> &Ins = CLI.Ins; | |||
5583 | SDValue Chain = CLI.Chain; | |||
5584 | SDValue Callee = CLI.Callee; | |||
5585 | bool &isTailCall = CLI.IsTailCall; | |||
5586 | CallingConv::ID CallConv = CLI.CallConv; | |||
5587 | bool isVarArg = CLI.IsVarArg; | |||
5588 | bool isPatchPoint = CLI.IsPatchPoint; | |||
5589 | const CallBase *CB = CLI.CB; | |||
5590 | ||||
5591 | if (isTailCall) { | |||
5592 | if (Subtarget.useLongCalls() && !(CB && CB->isMustTailCall())) | |||
5593 | isTailCall = false; | |||
5594 | else if (Subtarget.isSVR4ABI() && Subtarget.isPPC64()) | |||
5595 | isTailCall = IsEligibleForTailCallOptimization_64SVR4( | |||
5596 | Callee, CallConv, CB, isVarArg, Outs, Ins, DAG); | |||
5597 | else | |||
5598 | isTailCall = IsEligibleForTailCallOptimization(Callee, CallConv, isVarArg, | |||
5599 | Ins, DAG); | |||
5600 | if (isTailCall) { | |||
5601 | ++NumTailCalls; | |||
5602 | if (!getTargetMachine().Options.GuaranteedTailCallOpt) | |||
5603 | ++NumSiblingCalls; | |||
5604 | ||||
5605 | // PC Relative calls no longer guarantee that the callee is a Global | |||
5606 | // Address Node. The callee could be an indirect tail call in which | |||
5607 | // case the SDValue for the callee could be a load (to load the address | |||
5608 | // of a function pointer) or it may be a register copy (to move the | |||
5609 | // address of the callee from a function parameter into a virtual | |||
5610 | // register). It may also be an ExternalSymbolSDNode (ex memcopy). | |||
5611 | assert((Subtarget.isUsingPCRelativeCalls() ||(static_cast <bool> ((Subtarget.isUsingPCRelativeCalls( ) || isa<GlobalAddressSDNode>(Callee)) && "Callee should be an llvm::Function object." ) ? void (0) : __assert_fail ("(Subtarget.isUsingPCRelativeCalls() || isa<GlobalAddressSDNode>(Callee)) && \"Callee should be an llvm::Function object.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5613, __extension__ __PRETTY_FUNCTION__)) | |||
5612 | isa<GlobalAddressSDNode>(Callee)) &&(static_cast <bool> ((Subtarget.isUsingPCRelativeCalls( ) || isa<GlobalAddressSDNode>(Callee)) && "Callee should be an llvm::Function object." ) ? void (0) : __assert_fail ("(Subtarget.isUsingPCRelativeCalls() || isa<GlobalAddressSDNode>(Callee)) && \"Callee should be an llvm::Function object.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5613, __extension__ __PRETTY_FUNCTION__)) | |||
5613 | "Callee should be an llvm::Function object.")(static_cast <bool> ((Subtarget.isUsingPCRelativeCalls( ) || isa<GlobalAddressSDNode>(Callee)) && "Callee should be an llvm::Function object." ) ? void (0) : __assert_fail ("(Subtarget.isUsingPCRelativeCalls() || isa<GlobalAddressSDNode>(Callee)) && \"Callee should be an llvm::Function object.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5613, __extension__ __PRETTY_FUNCTION__)); | |||
5614 | ||||
5615 | LLVM_DEBUG(dbgs() << "TCO caller: " << DAG.getMachineFunction().getName()do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { dbgs() << "TCO caller: " << DAG .getMachineFunction().getName() << "\nTCO callee: "; } } while (false) | |||
5616 | << "\nTCO callee: ")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { dbgs() << "TCO caller: " << DAG .getMachineFunction().getName() << "\nTCO callee: "; } } while (false); | |||
5617 | LLVM_DEBUG(Callee.dump())do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { Callee.dump(); } } while (false); | |||
5618 | } | |||
5619 | } | |||
5620 | ||||
5621 | if (!isTailCall && CB && CB->isMustTailCall()) | |||
5622 | report_fatal_error("failed to perform tail call elimination on a call " | |||
5623 | "site marked musttail"); | |||
5624 | ||||
5625 | // When long calls (i.e. indirect calls) are always used, calls are always | |||
5626 | // made via function pointer. If we have a function name, first translate it | |||
5627 | // into a pointer. | |||
5628 | if (Subtarget.useLongCalls() && isa<GlobalAddressSDNode>(Callee) && | |||
5629 | !isTailCall) | |||
5630 | Callee = LowerGlobalAddress(Callee, DAG); | |||
5631 | ||||
5632 | CallFlags CFlags( | |||
5633 | CallConv, isTailCall, isVarArg, isPatchPoint, | |||
5634 | isIndirectCall(Callee, DAG, Subtarget, isPatchPoint), | |||
5635 | // hasNest | |||
5636 | Subtarget.is64BitELFABI() && | |||
5637 | any_of(Outs, [](ISD::OutputArg Arg) { return Arg.Flags.isNest(); }), | |||
5638 | CLI.NoMerge); | |||
5639 | ||||
5640 | if (Subtarget.isAIXABI()) | |||
5641 | return LowerCall_AIX(Chain, Callee, CFlags, Outs, OutVals, Ins, dl, DAG, | |||
5642 | InVals, CB); | |||
5643 | ||||
5644 | assert(Subtarget.isSVR4ABI())(static_cast <bool> (Subtarget.isSVR4ABI()) ? void (0) : __assert_fail ("Subtarget.isSVR4ABI()", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5644, __extension__ __PRETTY_FUNCTION__)); | |||
5645 | if (Subtarget.isPPC64()) | |||
5646 | return LowerCall_64SVR4(Chain, Callee, CFlags, Outs, OutVals, Ins, dl, DAG, | |||
5647 | InVals, CB); | |||
5648 | return LowerCall_32SVR4(Chain, Callee, CFlags, Outs, OutVals, Ins, dl, DAG, | |||
5649 | InVals, CB); | |||
5650 | } | |||
5651 | ||||
5652 | SDValue PPCTargetLowering::LowerCall_32SVR4( | |||
5653 | SDValue Chain, SDValue Callee, CallFlags CFlags, | |||
5654 | const SmallVectorImpl<ISD::OutputArg> &Outs, | |||
5655 | const SmallVectorImpl<SDValue> &OutVals, | |||
5656 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
5657 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals, | |||
5658 | const CallBase *CB) const { | |||
5659 | // See PPCTargetLowering::LowerFormalArguments_32SVR4() for a description | |||
5660 | // of the 32-bit SVR4 ABI stack frame layout. | |||
5661 | ||||
5662 | const CallingConv::ID CallConv = CFlags.CallConv; | |||
5663 | const bool IsVarArg = CFlags.IsVarArg; | |||
5664 | const bool IsTailCall = CFlags.IsTailCall; | |||
5665 | ||||
5666 | assert((CallConv == CallingConv::C ||(static_cast <bool> ((CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && "Unknown calling convention!") ? void (0) : __assert_fail ("(CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && \"Unknown calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5668, __extension__ __PRETTY_FUNCTION__)) | |||
5667 | CallConv == CallingConv::Cold ||(static_cast <bool> ((CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && "Unknown calling convention!") ? void (0) : __assert_fail ("(CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && \"Unknown calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5668, __extension__ __PRETTY_FUNCTION__)) | |||
5668 | CallConv == CallingConv::Fast) && "Unknown calling convention!")(static_cast <bool> ((CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && "Unknown calling convention!") ? void (0) : __assert_fail ("(CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && \"Unknown calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5668, __extension__ __PRETTY_FUNCTION__)); | |||
5669 | ||||
5670 | const Align PtrAlign(4); | |||
5671 | ||||
5672 | MachineFunction &MF = DAG.getMachineFunction(); | |||
5673 | ||||
5674 | // Mark this function as potentially containing a function that contains a | |||
5675 | // tail call. As a consequence the frame pointer will be used for dynamicalloc | |||
5676 | // and restoring the callers stack pointer in this functions epilog. This is | |||
5677 | // done because by tail calling the called function might overwrite the value | |||
5678 | // in this function's (MF) stack pointer stack slot 0(SP). | |||
5679 | if (getTargetMachine().Options.GuaranteedTailCallOpt && | |||
5680 | CallConv == CallingConv::Fast) | |||
5681 | MF.getInfo<PPCFunctionInfo>()->setHasFastCall(); | |||
5682 | ||||
5683 | // Count how many bytes are to be pushed on the stack, including the linkage | |||
5684 | // area, parameter list area and the part of the local variable space which | |||
5685 | // contains copies of aggregates which are passed by value. | |||
5686 | ||||
5687 | // Assign locations to all of the outgoing arguments. | |||
5688 | SmallVector<CCValAssign, 16> ArgLocs; | |||
5689 | PPCCCState CCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext()); | |||
5690 | ||||
5691 | // Reserve space for the linkage area on the stack. | |||
5692 | CCInfo.AllocateStack(Subtarget.getFrameLowering()->getLinkageSize(), | |||
5693 | PtrAlign); | |||
5694 | if (useSoftFloat()) | |||
5695 | CCInfo.PreAnalyzeCallOperands(Outs); | |||
5696 | ||||
5697 | if (IsVarArg) { | |||
5698 | // Handle fixed and variable vector arguments differently. | |||
5699 | // Fixed vector arguments go into registers as long as registers are | |||
5700 | // available. Variable vector arguments always go into memory. | |||
5701 | unsigned NumArgs = Outs.size(); | |||
5702 | ||||
5703 | for (unsigned i = 0; i != NumArgs; ++i) { | |||
5704 | MVT ArgVT = Outs[i].VT; | |||
5705 | ISD::ArgFlagsTy ArgFlags = Outs[i].Flags; | |||
5706 | bool Result; | |||
5707 | ||||
5708 | if (Outs[i].IsFixed) { | |||
5709 | Result = CC_PPC32_SVR4(i, ArgVT, ArgVT, CCValAssign::Full, ArgFlags, | |||
5710 | CCInfo); | |||
5711 | } else { | |||
5712 | Result = CC_PPC32_SVR4_VarArg(i, ArgVT, ArgVT, CCValAssign::Full, | |||
5713 | ArgFlags, CCInfo); | |||
5714 | } | |||
5715 | ||||
5716 | if (Result) { | |||
5717 | #ifndef NDEBUG | |||
5718 | errs() << "Call operand #" << i << " has unhandled type " | |||
5719 | << EVT(ArgVT).getEVTString() << "\n"; | |||
5720 | #endif | |||
5721 | llvm_unreachable(nullptr)::llvm::llvm_unreachable_internal(nullptr, "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5721); | |||
5722 | } | |||
5723 | } | |||
5724 | } else { | |||
5725 | // All arguments are treated the same. | |||
5726 | CCInfo.AnalyzeCallOperands(Outs, CC_PPC32_SVR4); | |||
5727 | } | |||
5728 | CCInfo.clearWasPPCF128(); | |||
5729 | ||||
5730 | // Assign locations to all of the outgoing aggregate by value arguments. | |||
5731 | SmallVector<CCValAssign, 16> ByValArgLocs; | |||
5732 | CCState CCByValInfo(CallConv, IsVarArg, MF, ByValArgLocs, *DAG.getContext()); | |||
5733 | ||||
5734 | // Reserve stack space for the allocations in CCInfo. | |||
5735 | CCByValInfo.AllocateStack(CCInfo.getNextStackOffset(), PtrAlign); | |||
5736 | ||||
5737 | CCByValInfo.AnalyzeCallOperands(Outs, CC_PPC32_SVR4_ByVal); | |||
5738 | ||||
5739 | // Size of the linkage area, parameter list area and the part of the local | |||
5740 | // space variable where copies of aggregates which are passed by value are | |||
5741 | // stored. | |||
5742 | unsigned NumBytes = CCByValInfo.getNextStackOffset(); | |||
5743 | ||||
5744 | // Calculate by how many bytes the stack has to be adjusted in case of tail | |||
5745 | // call optimization. | |||
5746 | int SPDiff = CalculateTailCallSPDiff(DAG, IsTailCall, NumBytes); | |||
5747 | ||||
5748 | // Adjust the stack pointer for the new arguments... | |||
5749 | // These operations are automatically eliminated by the prolog/epilog pass | |||
5750 | Chain = DAG.getCALLSEQ_START(Chain, NumBytes, 0, dl); | |||
5751 | SDValue CallSeqStart = Chain; | |||
5752 | ||||
5753 | // Load the return address and frame pointer so it can be moved somewhere else | |||
5754 | // later. | |||
5755 | SDValue LROp, FPOp; | |||
5756 | Chain = EmitTailCallLoadFPAndRetAddr(DAG, SPDiff, Chain, LROp, FPOp, dl); | |||
5757 | ||||
5758 | // Set up a copy of the stack pointer for use loading and storing any | |||
5759 | // arguments that may not fit in the registers available for argument | |||
5760 | // passing. | |||
5761 | SDValue StackPtr = DAG.getRegister(PPC::R1, MVT::i32); | |||
5762 | ||||
5763 | SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass; | |||
5764 | SmallVector<TailCallArgumentInfo, 8> TailCallArguments; | |||
5765 | SmallVector<SDValue, 8> MemOpChains; | |||
5766 | ||||
5767 | bool seenFloatArg = false; | |||
5768 | // Walk the register/memloc assignments, inserting copies/loads. | |||
5769 | // i - Tracks the index into the list of registers allocated for the call | |||
5770 | // RealArgIdx - Tracks the index into the list of actual function arguments | |||
5771 | // j - Tracks the index into the list of byval arguments | |||
5772 | for (unsigned i = 0, RealArgIdx = 0, j = 0, e = ArgLocs.size(); | |||
5773 | i != e; | |||
5774 | ++i, ++RealArgIdx) { | |||
5775 | CCValAssign &VA = ArgLocs[i]; | |||
5776 | SDValue Arg = OutVals[RealArgIdx]; | |||
5777 | ISD::ArgFlagsTy Flags = Outs[RealArgIdx].Flags; | |||
5778 | ||||
5779 | if (Flags.isByVal()) { | |||
5780 | // Argument is an aggregate which is passed by value, thus we need to | |||
5781 | // create a copy of it in the local variable space of the current stack | |||
5782 | // frame (which is the stack frame of the caller) and pass the address of | |||
5783 | // this copy to the callee. | |||
5784 | assert((j < ByValArgLocs.size()) && "Index out of bounds!")(static_cast <bool> ((j < ByValArgLocs.size()) && "Index out of bounds!") ? void (0) : __assert_fail ("(j < ByValArgLocs.size()) && \"Index out of bounds!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5784, __extension__ __PRETTY_FUNCTION__)); | |||
5785 | CCValAssign &ByValVA = ByValArgLocs[j++]; | |||
5786 | assert((VA.getValNo() == ByValVA.getValNo()) && "ValNo mismatch!")(static_cast <bool> ((VA.getValNo() == ByValVA.getValNo ()) && "ValNo mismatch!") ? void (0) : __assert_fail ( "(VA.getValNo() == ByValVA.getValNo()) && \"ValNo mismatch!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5786, __extension__ __PRETTY_FUNCTION__)); | |||
5787 | ||||
5788 | // Memory reserved in the local variable space of the callers stack frame. | |||
5789 | unsigned LocMemOffset = ByValVA.getLocMemOffset(); | |||
5790 | ||||
5791 | SDValue PtrOff = DAG.getIntPtrConstant(LocMemOffset, dl); | |||
5792 | PtrOff = DAG.getNode(ISD::ADD, dl, getPointerTy(MF.getDataLayout()), | |||
5793 | StackPtr, PtrOff); | |||
5794 | ||||
5795 | // Create a copy of the argument in the local area of the current | |||
5796 | // stack frame. | |||
5797 | SDValue MemcpyCall = | |||
5798 | CreateCopyOfByValArgument(Arg, PtrOff, | |||
5799 | CallSeqStart.getNode()->getOperand(0), | |||
5800 | Flags, DAG, dl); | |||
5801 | ||||
5802 | // This must go outside the CALLSEQ_START..END. | |||
5803 | SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall, NumBytes, 0, | |||
5804 | SDLoc(MemcpyCall)); | |||
5805 | DAG.ReplaceAllUsesWith(CallSeqStart.getNode(), | |||
5806 | NewCallSeqStart.getNode()); | |||
5807 | Chain = CallSeqStart = NewCallSeqStart; | |||
5808 | ||||
5809 | // Pass the address of the aggregate copy on the stack either in a | |||
5810 | // physical register or in the parameter list area of the current stack | |||
5811 | // frame to the callee. | |||
5812 | Arg = PtrOff; | |||
5813 | } | |||
5814 | ||||
5815 | // When useCRBits() is true, there can be i1 arguments. | |||
5816 | // It is because getRegisterType(MVT::i1) => MVT::i1, | |||
5817 | // and for other integer types getRegisterType() => MVT::i32. | |||
5818 | // Extend i1 and ensure callee will get i32. | |||
5819 | if (Arg.getValueType() == MVT::i1) | |||
5820 | Arg = DAG.getNode(Flags.isSExt() ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND, | |||
5821 | dl, MVT::i32, Arg); | |||
5822 | ||||
5823 | if (VA.isRegLoc()) { | |||
5824 | seenFloatArg |= VA.getLocVT().isFloatingPoint(); | |||
5825 | // Put argument in a physical register. | |||
5826 | if (Subtarget.hasSPE() && Arg.getValueType() == MVT::f64) { | |||
5827 | bool IsLE = Subtarget.isLittleEndian(); | |||
5828 | SDValue SVal = DAG.getNode(PPCISD::EXTRACT_SPE, dl, MVT::i32, Arg, | |||
5829 | DAG.getIntPtrConstant(IsLE ? 0 : 1, dl)); | |||
5830 | RegsToPass.push_back(std::make_pair(VA.getLocReg(), SVal.getValue(0))); | |||
5831 | SVal = DAG.getNode(PPCISD::EXTRACT_SPE, dl, MVT::i32, Arg, | |||
5832 | DAG.getIntPtrConstant(IsLE ? 1 : 0, dl)); | |||
5833 | RegsToPass.push_back(std::make_pair(ArgLocs[++i].getLocReg(), | |||
5834 | SVal.getValue(0))); | |||
5835 | } else | |||
5836 | RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg)); | |||
5837 | } else { | |||
5838 | // Put argument in the parameter list area of the current stack frame. | |||
5839 | assert(VA.isMemLoc())(static_cast <bool> (VA.isMemLoc()) ? void (0) : __assert_fail ("VA.isMemLoc()", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5839, __extension__ __PRETTY_FUNCTION__)); | |||
5840 | unsigned LocMemOffset = VA.getLocMemOffset(); | |||
5841 | ||||
5842 | if (!IsTailCall) { | |||
5843 | SDValue PtrOff = DAG.getIntPtrConstant(LocMemOffset, dl); | |||
5844 | PtrOff = DAG.getNode(ISD::ADD, dl, getPointerTy(MF.getDataLayout()), | |||
5845 | StackPtr, PtrOff); | |||
5846 | ||||
5847 | MemOpChains.push_back( | |||
5848 | DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo())); | |||
5849 | } else { | |||
5850 | // Calculate and remember argument location. | |||
5851 | CalculateTailCallArgDest(DAG, MF, false, Arg, SPDiff, LocMemOffset, | |||
5852 | TailCallArguments); | |||
5853 | } | |||
5854 | } | |||
5855 | } | |||
5856 | ||||
5857 | if (!MemOpChains.empty()) | |||
5858 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains); | |||
5859 | ||||
5860 | // Build a sequence of copy-to-reg nodes chained together with token chain | |||
5861 | // and flag operands which copy the outgoing args into the appropriate regs. | |||
5862 | SDValue InFlag; | |||
5863 | for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) { | |||
5864 | Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first, | |||
5865 | RegsToPass[i].second, InFlag); | |||
5866 | InFlag = Chain.getValue(1); | |||
5867 | } | |||
5868 | ||||
5869 | // Set CR bit 6 to true if this is a vararg call with floating args passed in | |||
5870 | // registers. | |||
5871 | if (IsVarArg) { | |||
5872 | SDVTList VTs = DAG.getVTList(MVT::Other, MVT::Glue); | |||
5873 | SDValue Ops[] = { Chain, InFlag }; | |||
5874 | ||||
5875 | Chain = DAG.getNode(seenFloatArg ? PPCISD::CR6SET : PPCISD::CR6UNSET, | |||
5876 | dl, VTs, makeArrayRef(Ops, InFlag.getNode() ? 2 : 1)); | |||
5877 | ||||
5878 | InFlag = Chain.getValue(1); | |||
5879 | } | |||
5880 | ||||
5881 | if (IsTailCall) | |||
5882 | PrepareTailCall(DAG, InFlag, Chain, dl, SPDiff, NumBytes, LROp, FPOp, | |||
5883 | TailCallArguments); | |||
5884 | ||||
5885 | return FinishCall(CFlags, dl, DAG, RegsToPass, InFlag, Chain, CallSeqStart, | |||
5886 | Callee, SPDiff, NumBytes, Ins, InVals, CB); | |||
5887 | } | |||
5888 | ||||
5889 | // Copy an argument into memory, being careful to do this outside the | |||
5890 | // call sequence for the call to which the argument belongs. | |||
5891 | SDValue PPCTargetLowering::createMemcpyOutsideCallSeq( | |||
5892 | SDValue Arg, SDValue PtrOff, SDValue CallSeqStart, ISD::ArgFlagsTy Flags, | |||
5893 | SelectionDAG &DAG, const SDLoc &dl) const { | |||
5894 | SDValue MemcpyCall = CreateCopyOfByValArgument(Arg, PtrOff, | |||
5895 | CallSeqStart.getNode()->getOperand(0), | |||
5896 | Flags, DAG, dl); | |||
5897 | // The MEMCPY must go outside the CALLSEQ_START..END. | |||
5898 | int64_t FrameSize = CallSeqStart.getConstantOperandVal(1); | |||
5899 | SDValue NewCallSeqStart = DAG.getCALLSEQ_START(MemcpyCall, FrameSize, 0, | |||
5900 | SDLoc(MemcpyCall)); | |||
5901 | DAG.ReplaceAllUsesWith(CallSeqStart.getNode(), | |||
5902 | NewCallSeqStart.getNode()); | |||
5903 | return NewCallSeqStart; | |||
5904 | } | |||
5905 | ||||
5906 | SDValue PPCTargetLowering::LowerCall_64SVR4( | |||
5907 | SDValue Chain, SDValue Callee, CallFlags CFlags, | |||
5908 | const SmallVectorImpl<ISD::OutputArg> &Outs, | |||
5909 | const SmallVectorImpl<SDValue> &OutVals, | |||
5910 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
5911 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals, | |||
5912 | const CallBase *CB) const { | |||
5913 | bool isELFv2ABI = Subtarget.isELFv2ABI(); | |||
5914 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
5915 | unsigned NumOps = Outs.size(); | |||
5916 | bool IsSibCall = false; | |||
5917 | bool IsFastCall = CFlags.CallConv == CallingConv::Fast; | |||
5918 | ||||
5919 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
5920 | unsigned PtrByteSize = 8; | |||
5921 | ||||
5922 | MachineFunction &MF = DAG.getMachineFunction(); | |||
5923 | ||||
5924 | if (CFlags.IsTailCall && !getTargetMachine().Options.GuaranteedTailCallOpt) | |||
5925 | IsSibCall = true; | |||
5926 | ||||
5927 | // Mark this function as potentially containing a function that contains a | |||
5928 | // tail call. As a consequence the frame pointer will be used for dynamicalloc | |||
5929 | // and restoring the callers stack pointer in this functions epilog. This is | |||
5930 | // done because by tail calling the called function might overwrite the value | |||
5931 | // in this function's (MF) stack pointer stack slot 0(SP). | |||
5932 | if (getTargetMachine().Options.GuaranteedTailCallOpt && IsFastCall) | |||
5933 | MF.getInfo<PPCFunctionInfo>()->setHasFastCall(); | |||
5934 | ||||
5935 | assert(!(IsFastCall && CFlags.IsVarArg) &&(static_cast <bool> (!(IsFastCall && CFlags.IsVarArg ) && "fastcc not supported on varargs functions") ? void (0) : __assert_fail ("!(IsFastCall && CFlags.IsVarArg) && \"fastcc not supported on varargs functions\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5936, __extension__ __PRETTY_FUNCTION__)) | |||
5936 | "fastcc not supported on varargs functions")(static_cast <bool> (!(IsFastCall && CFlags.IsVarArg ) && "fastcc not supported on varargs functions") ? void (0) : __assert_fail ("!(IsFastCall && CFlags.IsVarArg) && \"fastcc not supported on varargs functions\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 5936, __extension__ __PRETTY_FUNCTION__)); | |||
5937 | ||||
5938 | // Count how many bytes are to be pushed on the stack, including the linkage | |||
5939 | // area, and parameter passing area. On ELFv1, the linkage area is 48 bytes | |||
5940 | // reserved space for [SP][CR][LR][2 x unused][TOC]; on ELFv2, the linkage | |||
5941 | // area is 32 bytes reserved space for [SP][CR][LR][TOC]. | |||
5942 | unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize(); | |||
5943 | unsigned NumBytes = LinkageSize; | |||
5944 | unsigned GPR_idx = 0, FPR_idx = 0, VR_idx = 0; | |||
5945 | ||||
5946 | static const MCPhysReg GPR[] = { | |||
5947 | PPC::X3, PPC::X4, PPC::X5, PPC::X6, | |||
5948 | PPC::X7, PPC::X8, PPC::X9, PPC::X10, | |||
5949 | }; | |||
5950 | static const MCPhysReg VR[] = { | |||
5951 | PPC::V2, PPC::V3, PPC::V4, PPC::V5, PPC::V6, PPC::V7, PPC::V8, | |||
5952 | PPC::V9, PPC::V10, PPC::V11, PPC::V12, PPC::V13 | |||
5953 | }; | |||
5954 | ||||
5955 | const unsigned NumGPRs = array_lengthof(GPR); | |||
5956 | const unsigned NumFPRs = useSoftFloat() ? 0 : 13; | |||
5957 | const unsigned NumVRs = array_lengthof(VR); | |||
5958 | ||||
5959 | // On ELFv2, we can avoid allocating the parameter area if all the arguments | |||
5960 | // can be passed to the callee in registers. | |||
5961 | // For the fast calling convention, there is another check below. | |||
5962 | // Note: We should keep consistent with LowerFormalArguments_64SVR4() | |||
5963 | bool HasParameterArea = !isELFv2ABI || CFlags.IsVarArg || IsFastCall; | |||
5964 | if (!HasParameterArea) { | |||
5965 | unsigned ParamAreaSize = NumGPRs * PtrByteSize; | |||
5966 | unsigned AvailableFPRs = NumFPRs; | |||
5967 | unsigned AvailableVRs = NumVRs; | |||
5968 | unsigned NumBytesTmp = NumBytes; | |||
5969 | for (unsigned i = 0; i != NumOps; ++i) { | |||
5970 | if (Outs[i].Flags.isNest()) continue; | |||
5971 | if (CalculateStackSlotUsed(Subtarget, Outs[i].VT, Outs[i].ArgVT, | |||
5972 | Outs[i].Flags, PtrByteSize, LinkageSize, | |||
5973 | ParamAreaSize, NumBytesTmp, AvailableFPRs, | |||
5974 | AvailableVRs)) | |||
5975 | HasParameterArea = true; | |||
5976 | } | |||
5977 | } | |||
5978 | ||||
5979 | // When using the fast calling convention, we don't provide backing for | |||
5980 | // arguments that will be in registers. | |||
5981 | unsigned NumGPRsUsed = 0, NumFPRsUsed = 0, NumVRsUsed = 0; | |||
5982 | ||||
5983 | // Avoid allocating parameter area for fastcc functions if all the arguments | |||
5984 | // can be passed in the registers. | |||
5985 | if (IsFastCall) | |||
5986 | HasParameterArea = false; | |||
5987 | ||||
5988 | // Add up all the space actually used. | |||
5989 | for (unsigned i = 0; i != NumOps; ++i) { | |||
5990 | ISD::ArgFlagsTy Flags = Outs[i].Flags; | |||
5991 | EVT ArgVT = Outs[i].VT; | |||
5992 | EVT OrigVT = Outs[i].ArgVT; | |||
5993 | ||||
5994 | if (Flags.isNest()) | |||
5995 | continue; | |||
5996 | ||||
5997 | if (IsFastCall) { | |||
5998 | if (Flags.isByVal()) { | |||
5999 | NumGPRsUsed += (Flags.getByValSize()+7)/8; | |||
6000 | if (NumGPRsUsed > NumGPRs) | |||
6001 | HasParameterArea = true; | |||
6002 | } else { | |||
6003 | switch (ArgVT.getSimpleVT().SimpleTy) { | |||
6004 | default: llvm_unreachable("Unexpected ValueType for argument!")::llvm::llvm_unreachable_internal("Unexpected ValueType for argument!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6004); | |||
6005 | case MVT::i1: | |||
6006 | case MVT::i32: | |||
6007 | case MVT::i64: | |||
6008 | if (++NumGPRsUsed <= NumGPRs) | |||
6009 | continue; | |||
6010 | break; | |||
6011 | case MVT::v4i32: | |||
6012 | case MVT::v8i16: | |||
6013 | case MVT::v16i8: | |||
6014 | case MVT::v2f64: | |||
6015 | case MVT::v2i64: | |||
6016 | case MVT::v1i128: | |||
6017 | case MVT::f128: | |||
6018 | if (++NumVRsUsed <= NumVRs) | |||
6019 | continue; | |||
6020 | break; | |||
6021 | case MVT::v4f32: | |||
6022 | if (++NumVRsUsed <= NumVRs) | |||
6023 | continue; | |||
6024 | break; | |||
6025 | case MVT::f32: | |||
6026 | case MVT::f64: | |||
6027 | if (++NumFPRsUsed <= NumFPRs) | |||
6028 | continue; | |||
6029 | break; | |||
6030 | } | |||
6031 | HasParameterArea = true; | |||
6032 | } | |||
6033 | } | |||
6034 | ||||
6035 | /* Respect alignment of argument on the stack. */ | |||
6036 | auto Alignement = | |||
6037 | CalculateStackSlotAlignment(ArgVT, OrigVT, Flags, PtrByteSize); | |||
6038 | NumBytes = alignTo(NumBytes, Alignement); | |||
6039 | ||||
6040 | NumBytes += CalculateStackSlotSize(ArgVT, Flags, PtrByteSize); | |||
6041 | if (Flags.isInConsecutiveRegsLast()) | |||
6042 | NumBytes = ((NumBytes + PtrByteSize - 1)/PtrByteSize) * PtrByteSize; | |||
6043 | } | |||
6044 | ||||
6045 | unsigned NumBytesActuallyUsed = NumBytes; | |||
6046 | ||||
6047 | // In the old ELFv1 ABI, | |||
6048 | // the prolog code of the callee may store up to 8 GPR argument registers to | |||
6049 | // the stack, allowing va_start to index over them in memory if its varargs. | |||
6050 | // Because we cannot tell if this is needed on the caller side, we have to | |||
6051 | // conservatively assume that it is needed. As such, make sure we have at | |||
6052 | // least enough stack space for the caller to store the 8 GPRs. | |||
6053 | // In the ELFv2 ABI, we allocate the parameter area iff a callee | |||
6054 | // really requires memory operands, e.g. a vararg function. | |||
6055 | if (HasParameterArea) | |||
6056 | NumBytes = std::max(NumBytes, LinkageSize + 8 * PtrByteSize); | |||
6057 | else | |||
6058 | NumBytes = LinkageSize; | |||
6059 | ||||
6060 | // Tail call needs the stack to be aligned. | |||
6061 | if (getTargetMachine().Options.GuaranteedTailCallOpt && IsFastCall) | |||
6062 | NumBytes = EnsureStackAlignment(Subtarget.getFrameLowering(), NumBytes); | |||
6063 | ||||
6064 | int SPDiff = 0; | |||
6065 | ||||
6066 | // Calculate by how many bytes the stack has to be adjusted in case of tail | |||
6067 | // call optimization. | |||
6068 | if (!IsSibCall) | |||
6069 | SPDiff = CalculateTailCallSPDiff(DAG, CFlags.IsTailCall, NumBytes); | |||
6070 | ||||
6071 | // To protect arguments on the stack from being clobbered in a tail call, | |||
6072 | // force all the loads to happen before doing any other lowering. | |||
6073 | if (CFlags.IsTailCall) | |||
6074 | Chain = DAG.getStackArgumentTokenFactor(Chain); | |||
6075 | ||||
6076 | // Adjust the stack pointer for the new arguments... | |||
6077 | // These operations are automatically eliminated by the prolog/epilog pass | |||
6078 | if (!IsSibCall) | |||
6079 | Chain = DAG.getCALLSEQ_START(Chain, NumBytes, 0, dl); | |||
6080 | SDValue CallSeqStart = Chain; | |||
6081 | ||||
6082 | // Load the return address and frame pointer so it can be move somewhere else | |||
6083 | // later. | |||
6084 | SDValue LROp, FPOp; | |||
6085 | Chain = EmitTailCallLoadFPAndRetAddr(DAG, SPDiff, Chain, LROp, FPOp, dl); | |||
6086 | ||||
6087 | // Set up a copy of the stack pointer for use loading and storing any | |||
6088 | // arguments that may not fit in the registers available for argument | |||
6089 | // passing. | |||
6090 | SDValue StackPtr = DAG.getRegister(PPC::X1, MVT::i64); | |||
6091 | ||||
6092 | // Figure out which arguments are going to go in registers, and which in | |||
6093 | // memory. Also, if this is a vararg function, floating point operations | |||
6094 | // must be stored to our stack, and loaded into integer regs as well, if | |||
6095 | // any integer regs are available for argument passing. | |||
6096 | unsigned ArgOffset = LinkageSize; | |||
6097 | ||||
6098 | SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass; | |||
6099 | SmallVector<TailCallArgumentInfo, 8> TailCallArguments; | |||
6100 | ||||
6101 | SmallVector<SDValue, 8> MemOpChains; | |||
6102 | for (unsigned i = 0; i != NumOps; ++i) { | |||
6103 | SDValue Arg = OutVals[i]; | |||
6104 | ISD::ArgFlagsTy Flags = Outs[i].Flags; | |||
6105 | EVT ArgVT = Outs[i].VT; | |||
6106 | EVT OrigVT = Outs[i].ArgVT; | |||
6107 | ||||
6108 | // PtrOff will be used to store the current argument to the stack if a | |||
6109 | // register cannot be found for it. | |||
6110 | SDValue PtrOff; | |||
6111 | ||||
6112 | // We re-align the argument offset for each argument, except when using the | |||
6113 | // fast calling convention, when we need to make sure we do that only when | |||
6114 | // we'll actually use a stack slot. | |||
6115 | auto ComputePtrOff = [&]() { | |||
6116 | /* Respect alignment of argument on the stack. */ | |||
6117 | auto Alignment = | |||
6118 | CalculateStackSlotAlignment(ArgVT, OrigVT, Flags, PtrByteSize); | |||
6119 | ArgOffset = alignTo(ArgOffset, Alignment); | |||
6120 | ||||
6121 | PtrOff = DAG.getConstant(ArgOffset, dl, StackPtr.getValueType()); | |||
6122 | ||||
6123 | PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff); | |||
6124 | }; | |||
6125 | ||||
6126 | if (!IsFastCall) { | |||
6127 | ComputePtrOff(); | |||
6128 | ||||
6129 | /* Compute GPR index associated with argument offset. */ | |||
6130 | GPR_idx = (ArgOffset - LinkageSize) / PtrByteSize; | |||
6131 | GPR_idx = std::min(GPR_idx, NumGPRs); | |||
6132 | } | |||
6133 | ||||
6134 | // Promote integers to 64-bit values. | |||
6135 | if (Arg.getValueType() == MVT::i32 || Arg.getValueType() == MVT::i1) { | |||
6136 | // FIXME: Should this use ANY_EXTEND if neither sext nor zext? | |||
6137 | unsigned ExtOp = Flags.isSExt() ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND; | |||
6138 | Arg = DAG.getNode(ExtOp, dl, MVT::i64, Arg); | |||
6139 | } | |||
6140 | ||||
6141 | // FIXME memcpy is used way more than necessary. Correctness first. | |||
6142 | // Note: "by value" is code for passing a structure by value, not | |||
6143 | // basic types. | |||
6144 | if (Flags.isByVal()) { | |||
6145 | // Note: Size includes alignment padding, so | |||
6146 | // struct x { short a; char b; } | |||
6147 | // will have Size = 4. With #pragma pack(1), it will have Size = 3. | |||
6148 | // These are the proper values we need for right-justifying the | |||
6149 | // aggregate in a parameter register. | |||
6150 | unsigned Size = Flags.getByValSize(); | |||
6151 | ||||
6152 | // An empty aggregate parameter takes up no storage and no | |||
6153 | // registers. | |||
6154 | if (Size == 0) | |||
6155 | continue; | |||
6156 | ||||
6157 | if (IsFastCall) | |||
6158 | ComputePtrOff(); | |||
6159 | ||||
6160 | // All aggregates smaller than 8 bytes must be passed right-justified. | |||
6161 | if (Size==1 || Size==2 || Size==4) { | |||
6162 | EVT VT = (Size==1) ? MVT::i8 : ((Size==2) ? MVT::i16 : MVT::i32); | |||
6163 | if (GPR_idx != NumGPRs) { | |||
6164 | SDValue Load = DAG.getExtLoad(ISD::EXTLOAD, dl, PtrVT, Chain, Arg, | |||
6165 | MachinePointerInfo(), VT); | |||
6166 | MemOpChains.push_back(Load.getValue(1)); | |||
6167 | RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load)); | |||
6168 | ||||
6169 | ArgOffset += PtrByteSize; | |||
6170 | continue; | |||
6171 | } | |||
6172 | } | |||
6173 | ||||
6174 | if (GPR_idx == NumGPRs && Size < 8) { | |||
6175 | SDValue AddPtr = PtrOff; | |||
6176 | if (!isLittleEndian) { | |||
6177 | SDValue Const = DAG.getConstant(PtrByteSize - Size, dl, | |||
6178 | PtrOff.getValueType()); | |||
6179 | AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, Const); | |||
6180 | } | |||
6181 | Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, AddPtr, | |||
6182 | CallSeqStart, | |||
6183 | Flags, DAG, dl); | |||
6184 | ArgOffset += PtrByteSize; | |||
6185 | continue; | |||
6186 | } | |||
6187 | // Copy entire object into memory. There are cases where gcc-generated | |||
6188 | // code assumes it is there, even if it could be put entirely into | |||
6189 | // registers. (This is not what the doc says.) | |||
6190 | ||||
6191 | // FIXME: The above statement is likely due to a misunderstanding of the | |||
6192 | // documents. All arguments must be copied into the parameter area BY | |||
6193 | // THE CALLEE in the event that the callee takes the address of any | |||
6194 | // formal argument. That has not yet been implemented. However, it is | |||
6195 | // reasonable to use the stack area as a staging area for the register | |||
6196 | // load. | |||
6197 | ||||
6198 | // Skip this for small aggregates, as we will use the same slot for a | |||
6199 | // right-justified copy, below. | |||
6200 | if (Size >= 8) | |||
6201 | Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, PtrOff, | |||
6202 | CallSeqStart, | |||
6203 | Flags, DAG, dl); | |||
6204 | ||||
6205 | // When a register is available, pass a small aggregate right-justified. | |||
6206 | if (Size < 8 && GPR_idx != NumGPRs) { | |||
6207 | // The easiest way to get this right-justified in a register | |||
6208 | // is to copy the structure into the rightmost portion of a | |||
6209 | // local variable slot, then load the whole slot into the | |||
6210 | // register. | |||
6211 | // FIXME: The memcpy seems to produce pretty awful code for | |||
6212 | // small aggregates, particularly for packed ones. | |||
6213 | // FIXME: It would be preferable to use the slot in the | |||
6214 | // parameter save area instead of a new local variable. | |||
6215 | SDValue AddPtr = PtrOff; | |||
6216 | if (!isLittleEndian) { | |||
6217 | SDValue Const = DAG.getConstant(8 - Size, dl, PtrOff.getValueType()); | |||
6218 | AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, Const); | |||
6219 | } | |||
6220 | Chain = CallSeqStart = createMemcpyOutsideCallSeq(Arg, AddPtr, | |||
6221 | CallSeqStart, | |||
6222 | Flags, DAG, dl); | |||
6223 | ||||
6224 | // Load the slot into the register. | |||
6225 | SDValue Load = | |||
6226 | DAG.getLoad(PtrVT, dl, Chain, PtrOff, MachinePointerInfo()); | |||
6227 | MemOpChains.push_back(Load.getValue(1)); | |||
6228 | RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load)); | |||
6229 | ||||
6230 | // Done with this argument. | |||
6231 | ArgOffset += PtrByteSize; | |||
6232 | continue; | |||
6233 | } | |||
6234 | ||||
6235 | // For aggregates larger than PtrByteSize, copy the pieces of the | |||
6236 | // object that fit into registers from the parameter save area. | |||
6237 | for (unsigned j=0; j<Size; j+=PtrByteSize) { | |||
6238 | SDValue Const = DAG.getConstant(j, dl, PtrOff.getValueType()); | |||
6239 | SDValue AddArg = DAG.getNode(ISD::ADD, dl, PtrVT, Arg, Const); | |||
6240 | if (GPR_idx != NumGPRs) { | |||
6241 | SDValue Load = | |||
6242 | DAG.getLoad(PtrVT, dl, Chain, AddArg, MachinePointerInfo()); | |||
6243 | MemOpChains.push_back(Load.getValue(1)); | |||
6244 | RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load)); | |||
6245 | ArgOffset += PtrByteSize; | |||
6246 | } else { | |||
6247 | ArgOffset += ((Size - j + PtrByteSize-1)/PtrByteSize)*PtrByteSize; | |||
6248 | break; | |||
6249 | } | |||
6250 | } | |||
6251 | continue; | |||
6252 | } | |||
6253 | ||||
6254 | switch (Arg.getSimpleValueType().SimpleTy) { | |||
6255 | default: llvm_unreachable("Unexpected ValueType for argument!")::llvm::llvm_unreachable_internal("Unexpected ValueType for argument!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6255); | |||
6256 | case MVT::i1: | |||
6257 | case MVT::i32: | |||
6258 | case MVT::i64: | |||
6259 | if (Flags.isNest()) { | |||
6260 | // The 'nest' parameter, if any, is passed in R11. | |||
6261 | RegsToPass.push_back(std::make_pair(PPC::X11, Arg)); | |||
6262 | break; | |||
6263 | } | |||
6264 | ||||
6265 | // These can be scalar arguments or elements of an integer array type | |||
6266 | // passed directly. Clang may use those instead of "byval" aggregate | |||
6267 | // types to avoid forcing arguments to memory unnecessarily. | |||
6268 | if (GPR_idx != NumGPRs) { | |||
6269 | RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Arg)); | |||
6270 | } else { | |||
6271 | if (IsFastCall) | |||
6272 | ComputePtrOff(); | |||
6273 | ||||
6274 | assert(HasParameterArea &&(static_cast <bool> (HasParameterArea && "Parameter area must exist to pass an argument in memory." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist to pass an argument in memory.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6275, __extension__ __PRETTY_FUNCTION__)) | |||
6275 | "Parameter area must exist to pass an argument in memory.")(static_cast <bool> (HasParameterArea && "Parameter area must exist to pass an argument in memory." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist to pass an argument in memory.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6275, __extension__ __PRETTY_FUNCTION__)); | |||
6276 | LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset, | |||
6277 | true, CFlags.IsTailCall, false, MemOpChains, | |||
6278 | TailCallArguments, dl); | |||
6279 | if (IsFastCall) | |||
6280 | ArgOffset += PtrByteSize; | |||
6281 | } | |||
6282 | if (!IsFastCall) | |||
6283 | ArgOffset += PtrByteSize; | |||
6284 | break; | |||
6285 | case MVT::f32: | |||
6286 | case MVT::f64: { | |||
6287 | // These can be scalar arguments or elements of a float array type | |||
6288 | // passed directly. The latter are used to implement ELFv2 homogenous | |||
6289 | // float aggregates. | |||
6290 | ||||
6291 | // Named arguments go into FPRs first, and once they overflow, the | |||
6292 | // remaining arguments go into GPRs and then the parameter save area. | |||
6293 | // Unnamed arguments for vararg functions always go to GPRs and | |||
6294 | // then the parameter save area. For now, put all arguments to vararg | |||
6295 | // routines always in both locations (FPR *and* GPR or stack slot). | |||
6296 | bool NeedGPROrStack = CFlags.IsVarArg || FPR_idx == NumFPRs; | |||
6297 | bool NeededLoad = false; | |||
6298 | ||||
6299 | // First load the argument into the next available FPR. | |||
6300 | if (FPR_idx != NumFPRs) | |||
6301 | RegsToPass.push_back(std::make_pair(FPR[FPR_idx++], Arg)); | |||
6302 | ||||
6303 | // Next, load the argument into GPR or stack slot if needed. | |||
6304 | if (!NeedGPROrStack) | |||
6305 | ; | |||
6306 | else if (GPR_idx != NumGPRs && !IsFastCall) { | |||
6307 | // FIXME: We may want to re-enable this for CallingConv::Fast on the P8 | |||
6308 | // once we support fp <-> gpr moves. | |||
6309 | ||||
6310 | // In the non-vararg case, this can only ever happen in the | |||
6311 | // presence of f32 array types, since otherwise we never run | |||
6312 | // out of FPRs before running out of GPRs. | |||
6313 | SDValue ArgVal; | |||
6314 | ||||
6315 | // Double values are always passed in a single GPR. | |||
6316 | if (Arg.getValueType() != MVT::f32) { | |||
6317 | ArgVal = DAG.getNode(ISD::BITCAST, dl, MVT::i64, Arg); | |||
6318 | ||||
6319 | // Non-array float values are extended and passed in a GPR. | |||
6320 | } else if (!Flags.isInConsecutiveRegs()) { | |||
6321 | ArgVal = DAG.getNode(ISD::BITCAST, dl, MVT::i32, Arg); | |||
6322 | ArgVal = DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i64, ArgVal); | |||
6323 | ||||
6324 | // If we have an array of floats, we collect every odd element | |||
6325 | // together with its predecessor into one GPR. | |||
6326 | } else if (ArgOffset % PtrByteSize != 0) { | |||
6327 | SDValue Lo, Hi; | |||
6328 | Lo = DAG.getNode(ISD::BITCAST, dl, MVT::i32, OutVals[i - 1]); | |||
6329 | Hi = DAG.getNode(ISD::BITCAST, dl, MVT::i32, Arg); | |||
6330 | if (!isLittleEndian) | |||
6331 | std::swap(Lo, Hi); | |||
6332 | ArgVal = DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi); | |||
6333 | ||||
6334 | // The final element, if even, goes into the first half of a GPR. | |||
6335 | } else if (Flags.isInConsecutiveRegsLast()) { | |||
6336 | ArgVal = DAG.getNode(ISD::BITCAST, dl, MVT::i32, Arg); | |||
6337 | ArgVal = DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i64, ArgVal); | |||
6338 | if (!isLittleEndian) | |||
6339 | ArgVal = DAG.getNode(ISD::SHL, dl, MVT::i64, ArgVal, | |||
6340 | DAG.getConstant(32, dl, MVT::i32)); | |||
6341 | ||||
6342 | // Non-final even elements are skipped; they will be handled | |||
6343 | // together the with subsequent argument on the next go-around. | |||
6344 | } else | |||
6345 | ArgVal = SDValue(); | |||
6346 | ||||
6347 | if (ArgVal.getNode()) | |||
6348 | RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], ArgVal)); | |||
6349 | } else { | |||
6350 | if (IsFastCall) | |||
6351 | ComputePtrOff(); | |||
6352 | ||||
6353 | // Single-precision floating-point values are mapped to the | |||
6354 | // second (rightmost) word of the stack doubleword. | |||
6355 | if (Arg.getValueType() == MVT::f32 && | |||
6356 | !isLittleEndian && !Flags.isInConsecutiveRegs()) { | |||
6357 | SDValue ConstFour = DAG.getConstant(4, dl, PtrOff.getValueType()); | |||
6358 | PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, ConstFour); | |||
6359 | } | |||
6360 | ||||
6361 | assert(HasParameterArea &&(static_cast <bool> (HasParameterArea && "Parameter area must exist to pass an argument in memory." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist to pass an argument in memory.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6362, __extension__ __PRETTY_FUNCTION__)) | |||
6362 | "Parameter area must exist to pass an argument in memory.")(static_cast <bool> (HasParameterArea && "Parameter area must exist to pass an argument in memory." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist to pass an argument in memory.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6362, __extension__ __PRETTY_FUNCTION__)); | |||
6363 | LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset, | |||
6364 | true, CFlags.IsTailCall, false, MemOpChains, | |||
6365 | TailCallArguments, dl); | |||
6366 | ||||
6367 | NeededLoad = true; | |||
6368 | } | |||
6369 | // When passing an array of floats, the array occupies consecutive | |||
6370 | // space in the argument area; only round up to the next doubleword | |||
6371 | // at the end of the array. Otherwise, each float takes 8 bytes. | |||
6372 | if (!IsFastCall || NeededLoad) { | |||
6373 | ArgOffset += (Arg.getValueType() == MVT::f32 && | |||
6374 | Flags.isInConsecutiveRegs()) ? 4 : 8; | |||
6375 | if (Flags.isInConsecutiveRegsLast()) | |||
6376 | ArgOffset = ((ArgOffset + PtrByteSize - 1)/PtrByteSize) * PtrByteSize; | |||
6377 | } | |||
6378 | break; | |||
6379 | } | |||
6380 | case MVT::v4f32: | |||
6381 | case MVT::v4i32: | |||
6382 | case MVT::v8i16: | |||
6383 | case MVT::v16i8: | |||
6384 | case MVT::v2f64: | |||
6385 | case MVT::v2i64: | |||
6386 | case MVT::v1i128: | |||
6387 | case MVT::f128: | |||
6388 | // These can be scalar arguments or elements of a vector array type | |||
6389 | // passed directly. The latter are used to implement ELFv2 homogenous | |||
6390 | // vector aggregates. | |||
6391 | ||||
6392 | // For a varargs call, named arguments go into VRs or on the stack as | |||
6393 | // usual; unnamed arguments always go to the stack or the corresponding | |||
6394 | // GPRs when within range. For now, we always put the value in both | |||
6395 | // locations (or even all three). | |||
6396 | if (CFlags.IsVarArg) { | |||
6397 | assert(HasParameterArea &&(static_cast <bool> (HasParameterArea && "Parameter area must exist if we have a varargs call." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist if we have a varargs call.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6398, __extension__ __PRETTY_FUNCTION__)) | |||
6398 | "Parameter area must exist if we have a varargs call.")(static_cast <bool> (HasParameterArea && "Parameter area must exist if we have a varargs call." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist if we have a varargs call.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6398, __extension__ __PRETTY_FUNCTION__)); | |||
6399 | // We could elide this store in the case where the object fits | |||
6400 | // entirely in R registers. Maybe later. | |||
6401 | SDValue Store = | |||
6402 | DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo()); | |||
6403 | MemOpChains.push_back(Store); | |||
6404 | if (VR_idx != NumVRs) { | |||
6405 | SDValue Load = | |||
6406 | DAG.getLoad(MVT::v4f32, dl, Store, PtrOff, MachinePointerInfo()); | |||
6407 | MemOpChains.push_back(Load.getValue(1)); | |||
6408 | RegsToPass.push_back(std::make_pair(VR[VR_idx++], Load)); | |||
6409 | } | |||
6410 | ArgOffset += 16; | |||
6411 | for (unsigned i=0; i<16; i+=PtrByteSize) { | |||
6412 | if (GPR_idx == NumGPRs) | |||
6413 | break; | |||
6414 | SDValue Ix = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, | |||
6415 | DAG.getConstant(i, dl, PtrVT)); | |||
6416 | SDValue Load = | |||
6417 | DAG.getLoad(PtrVT, dl, Store, Ix, MachinePointerInfo()); | |||
6418 | MemOpChains.push_back(Load.getValue(1)); | |||
6419 | RegsToPass.push_back(std::make_pair(GPR[GPR_idx++], Load)); | |||
6420 | } | |||
6421 | break; | |||
6422 | } | |||
6423 | ||||
6424 | // Non-varargs Altivec params go into VRs or on the stack. | |||
6425 | if (VR_idx != NumVRs) { | |||
6426 | RegsToPass.push_back(std::make_pair(VR[VR_idx++], Arg)); | |||
6427 | } else { | |||
6428 | if (IsFastCall) | |||
6429 | ComputePtrOff(); | |||
6430 | ||||
6431 | assert(HasParameterArea &&(static_cast <bool> (HasParameterArea && "Parameter area must exist to pass an argument in memory." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist to pass an argument in memory.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6432, __extension__ __PRETTY_FUNCTION__)) | |||
6432 | "Parameter area must exist to pass an argument in memory.")(static_cast <bool> (HasParameterArea && "Parameter area must exist to pass an argument in memory." ) ? void (0) : __assert_fail ("HasParameterArea && \"Parameter area must exist to pass an argument in memory.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6432, __extension__ __PRETTY_FUNCTION__)); | |||
6433 | LowerMemOpCallTo(DAG, MF, Chain, Arg, PtrOff, SPDiff, ArgOffset, | |||
6434 | true, CFlags.IsTailCall, true, MemOpChains, | |||
6435 | TailCallArguments, dl); | |||
6436 | if (IsFastCall) | |||
6437 | ArgOffset += 16; | |||
6438 | } | |||
6439 | ||||
6440 | if (!IsFastCall) | |||
6441 | ArgOffset += 16; | |||
6442 | break; | |||
6443 | } | |||
6444 | } | |||
6445 | ||||
6446 | assert((!HasParameterArea || NumBytesActuallyUsed == ArgOffset) &&(static_cast <bool> ((!HasParameterArea || NumBytesActuallyUsed == ArgOffset) && "mismatch in size of parameter area" ) ? void (0) : __assert_fail ("(!HasParameterArea || NumBytesActuallyUsed == ArgOffset) && \"mismatch in size of parameter area\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6447, __extension__ __PRETTY_FUNCTION__)) | |||
6447 | "mismatch in size of parameter area")(static_cast <bool> ((!HasParameterArea || NumBytesActuallyUsed == ArgOffset) && "mismatch in size of parameter area" ) ? void (0) : __assert_fail ("(!HasParameterArea || NumBytesActuallyUsed == ArgOffset) && \"mismatch in size of parameter area\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6447, __extension__ __PRETTY_FUNCTION__)); | |||
6448 | (void)NumBytesActuallyUsed; | |||
6449 | ||||
6450 | if (!MemOpChains.empty()) | |||
6451 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains); | |||
6452 | ||||
6453 | // Check if this is an indirect call (MTCTR/BCTRL). | |||
6454 | // See prepareDescriptorIndirectCall and buildCallOperands for more | |||
6455 | // information about calls through function pointers in the 64-bit SVR4 ABI. | |||
6456 | if (CFlags.IsIndirect) { | |||
6457 | // For 64-bit ELFv2 ABI with PCRel, do not save the TOC of the | |||
6458 | // caller in the TOC save area. | |||
6459 | if (isTOCSaveRestoreRequired(Subtarget)) { | |||
6460 | assert(!CFlags.IsTailCall && "Indirect tails calls not supported")(static_cast <bool> (!CFlags.IsTailCall && "Indirect tails calls not supported" ) ? void (0) : __assert_fail ("!CFlags.IsTailCall && \"Indirect tails calls not supported\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6460, __extension__ __PRETTY_FUNCTION__)); | |||
6461 | // Load r2 into a virtual register and store it to the TOC save area. | |||
6462 | setUsesTOCBasePtr(DAG); | |||
6463 | SDValue Val = DAG.getCopyFromReg(Chain, dl, PPC::X2, MVT::i64); | |||
6464 | // TOC save area offset. | |||
6465 | unsigned TOCSaveOffset = Subtarget.getFrameLowering()->getTOCSaveOffset(); | |||
6466 | SDValue PtrOff = DAG.getIntPtrConstant(TOCSaveOffset, dl); | |||
6467 | SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff); | |||
6468 | Chain = DAG.getStore(Val.getValue(1), dl, Val, AddPtr, | |||
6469 | MachinePointerInfo::getStack( | |||
6470 | DAG.getMachineFunction(), TOCSaveOffset)); | |||
6471 | } | |||
6472 | // In the ELFv2 ABI, R12 must contain the address of an indirect callee. | |||
6473 | // This does not mean the MTCTR instruction must use R12; it's easier | |||
6474 | // to model this as an extra parameter, so do that. | |||
6475 | if (isELFv2ABI && !CFlags.IsPatchPoint) | |||
6476 | RegsToPass.push_back(std::make_pair((unsigned)PPC::X12, Callee)); | |||
6477 | } | |||
6478 | ||||
6479 | // Build a sequence of copy-to-reg nodes chained together with token chain | |||
6480 | // and flag operands which copy the outgoing args into the appropriate regs. | |||
6481 | SDValue InFlag; | |||
6482 | for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) { | |||
6483 | Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first, | |||
6484 | RegsToPass[i].second, InFlag); | |||
6485 | InFlag = Chain.getValue(1); | |||
6486 | } | |||
6487 | ||||
6488 | if (CFlags.IsTailCall && !IsSibCall) | |||
6489 | PrepareTailCall(DAG, InFlag, Chain, dl, SPDiff, NumBytes, LROp, FPOp, | |||
6490 | TailCallArguments); | |||
6491 | ||||
6492 | return FinishCall(CFlags, dl, DAG, RegsToPass, InFlag, Chain, CallSeqStart, | |||
6493 | Callee, SPDiff, NumBytes, Ins, InVals, CB); | |||
6494 | } | |||
6495 | ||||
6496 | // Returns true when the shadow of a general purpose argument register | |||
6497 | // in the parameter save area is aligned to at least 'RequiredAlign'. | |||
6498 | static bool isGPRShadowAligned(MCPhysReg Reg, Align RequiredAlign) { | |||
6499 | assert(RequiredAlign.value() <= 16 &&(static_cast <bool> (RequiredAlign.value() <= 16 && "Required alignment greater than stack alignment.") ? void ( 0) : __assert_fail ("RequiredAlign.value() <= 16 && \"Required alignment greater than stack alignment.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6500, __extension__ __PRETTY_FUNCTION__)) | |||
6500 | "Required alignment greater than stack alignment.")(static_cast <bool> (RequiredAlign.value() <= 16 && "Required alignment greater than stack alignment.") ? void ( 0) : __assert_fail ("RequiredAlign.value() <= 16 && \"Required alignment greater than stack alignment.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6500, __extension__ __PRETTY_FUNCTION__)); | |||
6501 | switch (Reg) { | |||
6502 | default: | |||
6503 | report_fatal_error("called on invalid register."); | |||
6504 | case PPC::R5: | |||
6505 | case PPC::R9: | |||
6506 | case PPC::X3: | |||
6507 | case PPC::X5: | |||
6508 | case PPC::X7: | |||
6509 | case PPC::X9: | |||
6510 | // These registers are 16 byte aligned which is the most strict aligment | |||
6511 | // we can support. | |||
6512 | return true; | |||
6513 | case PPC::R3: | |||
6514 | case PPC::R7: | |||
6515 | case PPC::X4: | |||
6516 | case PPC::X6: | |||
6517 | case PPC::X8: | |||
6518 | case PPC::X10: | |||
6519 | // The shadow of these registers in the PSA is 8 byte aligned. | |||
6520 | return RequiredAlign <= 8; | |||
6521 | case PPC::R4: | |||
6522 | case PPC::R6: | |||
6523 | case PPC::R8: | |||
6524 | case PPC::R10: | |||
6525 | return RequiredAlign <= 4; | |||
6526 | } | |||
6527 | } | |||
6528 | ||||
6529 | static bool CC_AIX(unsigned ValNo, MVT ValVT, MVT LocVT, | |||
6530 | CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags, | |||
6531 | CCState &S) { | |||
6532 | AIXCCState &State = static_cast<AIXCCState &>(S); | |||
6533 | const PPCSubtarget &Subtarget = static_cast<const PPCSubtarget &>( | |||
6534 | State.getMachineFunction().getSubtarget()); | |||
6535 | const bool IsPPC64 = Subtarget.isPPC64(); | |||
6536 | const Align PtrAlign = IsPPC64 ? Align(8) : Align(4); | |||
6537 | const MVT RegVT = IsPPC64 ? MVT::i64 : MVT::i32; | |||
6538 | ||||
6539 | if (ValVT == MVT::f128) | |||
6540 | report_fatal_error("f128 is unimplemented on AIX."); | |||
6541 | ||||
6542 | if (ArgFlags.isNest()) | |||
6543 | report_fatal_error("Nest arguments are unimplemented."); | |||
6544 | ||||
6545 | static const MCPhysReg GPR_32[] = {// 32-bit registers. | |||
6546 | PPC::R3, PPC::R4, PPC::R5, PPC::R6, | |||
6547 | PPC::R7, PPC::R8, PPC::R9, PPC::R10}; | |||
6548 | static const MCPhysReg GPR_64[] = {// 64-bit registers. | |||
6549 | PPC::X3, PPC::X4, PPC::X5, PPC::X6, | |||
6550 | PPC::X7, PPC::X8, PPC::X9, PPC::X10}; | |||
6551 | ||||
6552 | static const MCPhysReg VR[] = {// Vector registers. | |||
6553 | PPC::V2, PPC::V3, PPC::V4, PPC::V5, | |||
6554 | PPC::V6, PPC::V7, PPC::V8, PPC::V9, | |||
6555 | PPC::V10, PPC::V11, PPC::V12, PPC::V13}; | |||
6556 | ||||
6557 | if (ArgFlags.isByVal()) { | |||
6558 | if (ArgFlags.getNonZeroByValAlign() > PtrAlign) | |||
6559 | report_fatal_error("Pass-by-value arguments with alignment greater than " | |||
6560 | "register width are not supported."); | |||
6561 | ||||
6562 | const unsigned ByValSize = ArgFlags.getByValSize(); | |||
6563 | ||||
6564 | // An empty aggregate parameter takes up no storage and no registers, | |||
6565 | // but needs a MemLoc for a stack slot for the formal arguments side. | |||
6566 | if (ByValSize == 0) { | |||
6567 | State.addLoc(CCValAssign::getMem(ValNo, MVT::INVALID_SIMPLE_VALUE_TYPE, | |||
6568 | State.getNextStackOffset(), RegVT, | |||
6569 | LocInfo)); | |||
6570 | return false; | |||
6571 | } | |||
6572 | ||||
6573 | const unsigned StackSize = alignTo(ByValSize, PtrAlign); | |||
6574 | unsigned Offset = State.AllocateStack(StackSize, PtrAlign); | |||
6575 | for (const unsigned E = Offset + StackSize; Offset < E; | |||
6576 | Offset += PtrAlign.value()) { | |||
6577 | if (unsigned Reg = State.AllocateReg(IsPPC64 ? GPR_64 : GPR_32)) | |||
6578 | State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, RegVT, LocInfo)); | |||
6579 | else { | |||
6580 | State.addLoc(CCValAssign::getMem(ValNo, MVT::INVALID_SIMPLE_VALUE_TYPE, | |||
6581 | Offset, MVT::INVALID_SIMPLE_VALUE_TYPE, | |||
6582 | LocInfo)); | |||
6583 | break; | |||
6584 | } | |||
6585 | } | |||
6586 | return false; | |||
6587 | } | |||
6588 | ||||
6589 | // Arguments always reserve parameter save area. | |||
6590 | switch (ValVT.SimpleTy) { | |||
6591 | default: | |||
6592 | report_fatal_error("Unhandled value type for argument."); | |||
6593 | case MVT::i64: | |||
6594 | // i64 arguments should have been split to i32 for PPC32. | |||
6595 | assert(IsPPC64 && "PPC32 should have split i64 values.")(static_cast <bool> (IsPPC64 && "PPC32 should have split i64 values." ) ? void (0) : __assert_fail ("IsPPC64 && \"PPC32 should have split i64 values.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6595, __extension__ __PRETTY_FUNCTION__)); | |||
6596 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
6597 | case MVT::i1: | |||
6598 | case MVT::i32: { | |||
6599 | const unsigned Offset = State.AllocateStack(PtrAlign.value(), PtrAlign); | |||
6600 | // AIX integer arguments are always passed in register width. | |||
6601 | if (ValVT.getFixedSizeInBits() < RegVT.getFixedSizeInBits()) | |||
6602 | LocInfo = ArgFlags.isSExt() ? CCValAssign::LocInfo::SExt | |||
6603 | : CCValAssign::LocInfo::ZExt; | |||
6604 | if (unsigned Reg = State.AllocateReg(IsPPC64 ? GPR_64 : GPR_32)) | |||
6605 | State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, RegVT, LocInfo)); | |||
6606 | else | |||
6607 | State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, RegVT, LocInfo)); | |||
6608 | ||||
6609 | return false; | |||
6610 | } | |||
6611 | case MVT::f32: | |||
6612 | case MVT::f64: { | |||
6613 | // Parameter save area (PSA) is reserved even if the float passes in fpr. | |||
6614 | const unsigned StoreSize = LocVT.getStoreSize(); | |||
6615 | // Floats are always 4-byte aligned in the PSA on AIX. | |||
6616 | // This includes f64 in 64-bit mode for ABI compatibility. | |||
6617 | const unsigned Offset = | |||
6618 | State.AllocateStack(IsPPC64 ? 8 : StoreSize, Align(4)); | |||
6619 | unsigned FReg = State.AllocateReg(FPR); | |||
6620 | if (FReg) | |||
6621 | State.addLoc(CCValAssign::getReg(ValNo, ValVT, FReg, LocVT, LocInfo)); | |||
6622 | ||||
6623 | // Reserve and initialize GPRs or initialize the PSA as required. | |||
6624 | for (unsigned I = 0; I < StoreSize; I += PtrAlign.value()) { | |||
6625 | if (unsigned Reg = State.AllocateReg(IsPPC64 ? GPR_64 : GPR_32)) { | |||
6626 | assert(FReg && "An FPR should be available when a GPR is reserved.")(static_cast <bool> (FReg && "An FPR should be available when a GPR is reserved." ) ? void (0) : __assert_fail ("FReg && \"An FPR should be available when a GPR is reserved.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6626, __extension__ __PRETTY_FUNCTION__)); | |||
6627 | if (State.isVarArg()) { | |||
6628 | // Successfully reserved GPRs are only initialized for vararg calls. | |||
6629 | // Custom handling is required for: | |||
6630 | // f64 in PPC32 needs to be split into 2 GPRs. | |||
6631 | // f32 in PPC64 needs to occupy only lower 32 bits of 64-bit GPR. | |||
6632 | State.addLoc( | |||
6633 | CCValAssign::getCustomReg(ValNo, ValVT, Reg, RegVT, LocInfo)); | |||
6634 | } | |||
6635 | } else { | |||
6636 | // If there are insufficient GPRs, the PSA needs to be initialized. | |||
6637 | // Initialization occurs even if an FPR was initialized for | |||
6638 | // compatibility with the AIX XL compiler. The full memory for the | |||
6639 | // argument will be initialized even if a prior word is saved in GPR. | |||
6640 | // A custom memLoc is used when the argument also passes in FPR so | |||
6641 | // that the callee handling can skip over it easily. | |||
6642 | State.addLoc( | |||
6643 | FReg ? CCValAssign::getCustomMem(ValNo, ValVT, Offset, LocVT, | |||
6644 | LocInfo) | |||
6645 | : CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo)); | |||
6646 | break; | |||
6647 | } | |||
6648 | } | |||
6649 | ||||
6650 | return false; | |||
6651 | } | |||
6652 | case MVT::v4f32: | |||
6653 | case MVT::v4i32: | |||
6654 | case MVT::v8i16: | |||
6655 | case MVT::v16i8: | |||
6656 | case MVT::v2i64: | |||
6657 | case MVT::v2f64: | |||
6658 | case MVT::v1i128: { | |||
6659 | const unsigned VecSize = 16; | |||
6660 | const Align VecAlign(VecSize); | |||
6661 | ||||
6662 | if (!State.isVarArg()) { | |||
6663 | // If there are vector registers remaining we don't consume any stack | |||
6664 | // space. | |||
6665 | if (unsigned VReg = State.AllocateReg(VR)) { | |||
6666 | State.addLoc(CCValAssign::getReg(ValNo, ValVT, VReg, LocVT, LocInfo)); | |||
6667 | return false; | |||
6668 | } | |||
6669 | // Vectors passed on the stack do not shadow GPRs or FPRs even though they | |||
6670 | // might be allocated in the portion of the PSA that is shadowed by the | |||
6671 | // GPRs. | |||
6672 | const unsigned Offset = State.AllocateStack(VecSize, VecAlign); | |||
6673 | State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo)); | |||
6674 | return false; | |||
6675 | } | |||
6676 | ||||
6677 | const unsigned PtrSize = IsPPC64 ? 8 : 4; | |||
6678 | ArrayRef<MCPhysReg> GPRs = IsPPC64 ? GPR_64 : GPR_32; | |||
6679 | ||||
6680 | unsigned NextRegIndex = State.getFirstUnallocated(GPRs); | |||
6681 | // Burn any underaligned registers and their shadowed stack space until | |||
6682 | // we reach the required alignment. | |||
6683 | while (NextRegIndex != GPRs.size() && | |||
6684 | !isGPRShadowAligned(GPRs[NextRegIndex], VecAlign)) { | |||
6685 | // Shadow allocate register and its stack shadow. | |||
6686 | unsigned Reg = State.AllocateReg(GPRs); | |||
6687 | State.AllocateStack(PtrSize, PtrAlign); | |||
6688 | assert(Reg && "Allocating register unexpectedly failed.")(static_cast <bool> (Reg && "Allocating register unexpectedly failed." ) ? void (0) : __assert_fail ("Reg && \"Allocating register unexpectedly failed.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6688, __extension__ __PRETTY_FUNCTION__)); | |||
6689 | (void)Reg; | |||
6690 | NextRegIndex = State.getFirstUnallocated(GPRs); | |||
6691 | } | |||
6692 | ||||
6693 | // Vectors that are passed as fixed arguments are handled differently. | |||
6694 | // They are passed in VRs if any are available (unlike arguments passed | |||
6695 | // through ellipses) and shadow GPRs (unlike arguments to non-vaarg | |||
6696 | // functions) | |||
6697 | if (State.isFixed(ValNo)) { | |||
6698 | if (unsigned VReg = State.AllocateReg(VR)) { | |||
6699 | State.addLoc(CCValAssign::getReg(ValNo, ValVT, VReg, LocVT, LocInfo)); | |||
6700 | // Shadow allocate GPRs and stack space even though we pass in a VR. | |||
6701 | for (unsigned I = 0; I != VecSize; I += PtrSize) | |||
6702 | State.AllocateReg(GPRs); | |||
6703 | State.AllocateStack(VecSize, VecAlign); | |||
6704 | return false; | |||
6705 | } | |||
6706 | // No vector registers remain so pass on the stack. | |||
6707 | const unsigned Offset = State.AllocateStack(VecSize, VecAlign); | |||
6708 | State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo)); | |||
6709 | return false; | |||
6710 | } | |||
6711 | ||||
6712 | // If all GPRS are consumed then we pass the argument fully on the stack. | |||
6713 | if (NextRegIndex == GPRs.size()) { | |||
6714 | const unsigned Offset = State.AllocateStack(VecSize, VecAlign); | |||
6715 | State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo)); | |||
6716 | return false; | |||
6717 | } | |||
6718 | ||||
6719 | // Corner case for 32-bit codegen. We have 2 registers to pass the first | |||
6720 | // half of the argument, and then need to pass the remaining half on the | |||
6721 | // stack. | |||
6722 | if (GPRs[NextRegIndex] == PPC::R9) { | |||
6723 | const unsigned Offset = State.AllocateStack(VecSize, VecAlign); | |||
6724 | State.addLoc( | |||
6725 | CCValAssign::getCustomMem(ValNo, ValVT, Offset, LocVT, LocInfo)); | |||
6726 | ||||
6727 | const unsigned FirstReg = State.AllocateReg(PPC::R9); | |||
6728 | const unsigned SecondReg = State.AllocateReg(PPC::R10); | |||
6729 | assert(FirstReg && SecondReg &&(static_cast <bool> (FirstReg && SecondReg && "Allocating R9 or R10 unexpectedly failed.") ? void (0) : __assert_fail ("FirstReg && SecondReg && \"Allocating R9 or R10 unexpectedly failed.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6730, __extension__ __PRETTY_FUNCTION__)) | |||
6730 | "Allocating R9 or R10 unexpectedly failed.")(static_cast <bool> (FirstReg && SecondReg && "Allocating R9 or R10 unexpectedly failed.") ? void (0) : __assert_fail ("FirstReg && SecondReg && \"Allocating R9 or R10 unexpectedly failed.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6730, __extension__ __PRETTY_FUNCTION__)); | |||
6731 | State.addLoc( | |||
6732 | CCValAssign::getCustomReg(ValNo, ValVT, FirstReg, RegVT, LocInfo)); | |||
6733 | State.addLoc( | |||
6734 | CCValAssign::getCustomReg(ValNo, ValVT, SecondReg, RegVT, LocInfo)); | |||
6735 | return false; | |||
6736 | } | |||
6737 | ||||
6738 | // We have enough GPRs to fully pass the vector argument, and we have | |||
6739 | // already consumed any underaligned registers. Start with the custom | |||
6740 | // MemLoc and then the custom RegLocs. | |||
6741 | const unsigned Offset = State.AllocateStack(VecSize, VecAlign); | |||
6742 | State.addLoc( | |||
6743 | CCValAssign::getCustomMem(ValNo, ValVT, Offset, LocVT, LocInfo)); | |||
6744 | for (unsigned I = 0; I != VecSize; I += PtrSize) { | |||
6745 | const unsigned Reg = State.AllocateReg(GPRs); | |||
6746 | assert(Reg && "Failed to allocated register for vararg vector argument")(static_cast <bool> (Reg && "Failed to allocated register for vararg vector argument" ) ? void (0) : __assert_fail ("Reg && \"Failed to allocated register for vararg vector argument\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6746, __extension__ __PRETTY_FUNCTION__)); | |||
6747 | State.addLoc( | |||
6748 | CCValAssign::getCustomReg(ValNo, ValVT, Reg, RegVT, LocInfo)); | |||
6749 | } | |||
6750 | return false; | |||
6751 | } | |||
6752 | } | |||
6753 | return true; | |||
6754 | } | |||
6755 | ||||
6756 | // So far, this function is only used by LowerFormalArguments_AIX() | |||
6757 | static const TargetRegisterClass *getRegClassForSVT(MVT::SimpleValueType SVT, | |||
6758 | bool IsPPC64, | |||
6759 | bool HasP8Vector, | |||
6760 | bool HasVSX) { | |||
6761 | assert((IsPPC64 || SVT != MVT::i64) &&(static_cast <bool> ((IsPPC64 || SVT != MVT::i64) && "i64 should have been split for 32-bit codegen.") ? void (0) : __assert_fail ("(IsPPC64 || SVT != MVT::i64) && \"i64 should have been split for 32-bit codegen.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6762, __extension__ __PRETTY_FUNCTION__)) | |||
6762 | "i64 should have been split for 32-bit codegen.")(static_cast <bool> ((IsPPC64 || SVT != MVT::i64) && "i64 should have been split for 32-bit codegen.") ? void (0) : __assert_fail ("(IsPPC64 || SVT != MVT::i64) && \"i64 should have been split for 32-bit codegen.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6762, __extension__ __PRETTY_FUNCTION__)); | |||
6763 | ||||
6764 | switch (SVT) { | |||
6765 | default: | |||
6766 | report_fatal_error("Unexpected value type for formal argument"); | |||
6767 | case MVT::i1: | |||
6768 | case MVT::i32: | |||
6769 | case MVT::i64: | |||
6770 | return IsPPC64 ? &PPC::G8RCRegClass : &PPC::GPRCRegClass; | |||
6771 | case MVT::f32: | |||
6772 | return HasP8Vector ? &PPC::VSSRCRegClass : &PPC::F4RCRegClass; | |||
6773 | case MVT::f64: | |||
6774 | return HasVSX ? &PPC::VSFRCRegClass : &PPC::F8RCRegClass; | |||
6775 | case MVT::v4f32: | |||
6776 | case MVT::v4i32: | |||
6777 | case MVT::v8i16: | |||
6778 | case MVT::v16i8: | |||
6779 | case MVT::v2i64: | |||
6780 | case MVT::v2f64: | |||
6781 | case MVT::v1i128: | |||
6782 | return &PPC::VRRCRegClass; | |||
6783 | } | |||
6784 | } | |||
6785 | ||||
6786 | static SDValue truncateScalarIntegerArg(ISD::ArgFlagsTy Flags, EVT ValVT, | |||
6787 | SelectionDAG &DAG, SDValue ArgValue, | |||
6788 | MVT LocVT, const SDLoc &dl) { | |||
6789 | assert(ValVT.isScalarInteger() && LocVT.isScalarInteger())(static_cast <bool> (ValVT.isScalarInteger() && LocVT.isScalarInteger()) ? void (0) : __assert_fail ("ValVT.isScalarInteger() && LocVT.isScalarInteger()" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6789, __extension__ __PRETTY_FUNCTION__)); | |||
6790 | assert(ValVT.getFixedSizeInBits() < LocVT.getFixedSizeInBits())(static_cast <bool> (ValVT.getFixedSizeInBits() < LocVT .getFixedSizeInBits()) ? void (0) : __assert_fail ("ValVT.getFixedSizeInBits() < LocVT.getFixedSizeInBits()" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6790, __extension__ __PRETTY_FUNCTION__)); | |||
6791 | ||||
6792 | if (Flags.isSExt()) | |||
6793 | ArgValue = DAG.getNode(ISD::AssertSext, dl, LocVT, ArgValue, | |||
6794 | DAG.getValueType(ValVT)); | |||
6795 | else if (Flags.isZExt()) | |||
6796 | ArgValue = DAG.getNode(ISD::AssertZext, dl, LocVT, ArgValue, | |||
6797 | DAG.getValueType(ValVT)); | |||
6798 | ||||
6799 | return DAG.getNode(ISD::TRUNCATE, dl, ValVT, ArgValue); | |||
6800 | } | |||
6801 | ||||
6802 | static unsigned mapArgRegToOffsetAIX(unsigned Reg, const PPCFrameLowering *FL) { | |||
6803 | const unsigned LASize = FL->getLinkageSize(); | |||
6804 | ||||
6805 | if (PPC::GPRCRegClass.contains(Reg)) { | |||
6806 | assert(Reg >= PPC::R3 && Reg <= PPC::R10 &&(static_cast <bool> (Reg >= PPC::R3 && Reg <= PPC::R10 && "Reg must be a valid argument register!" ) ? void (0) : __assert_fail ("Reg >= PPC::R3 && Reg <= PPC::R10 && \"Reg must be a valid argument register!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6807, __extension__ __PRETTY_FUNCTION__)) | |||
6807 | "Reg must be a valid argument register!")(static_cast <bool> (Reg >= PPC::R3 && Reg <= PPC::R10 && "Reg must be a valid argument register!" ) ? void (0) : __assert_fail ("Reg >= PPC::R3 && Reg <= PPC::R10 && \"Reg must be a valid argument register!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6807, __extension__ __PRETTY_FUNCTION__)); | |||
6808 | return LASize + 4 * (Reg - PPC::R3); | |||
6809 | } | |||
6810 | ||||
6811 | if (PPC::G8RCRegClass.contains(Reg)) { | |||
6812 | assert(Reg >= PPC::X3 && Reg <= PPC::X10 &&(static_cast <bool> (Reg >= PPC::X3 && Reg <= PPC::X10 && "Reg must be a valid argument register!" ) ? void (0) : __assert_fail ("Reg >= PPC::X3 && Reg <= PPC::X10 && \"Reg must be a valid argument register!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6813, __extension__ __PRETTY_FUNCTION__)) | |||
6813 | "Reg must be a valid argument register!")(static_cast <bool> (Reg >= PPC::X3 && Reg <= PPC::X10 && "Reg must be a valid argument register!" ) ? void (0) : __assert_fail ("Reg >= PPC::X3 && Reg <= PPC::X10 && \"Reg must be a valid argument register!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6813, __extension__ __PRETTY_FUNCTION__)); | |||
6814 | return LASize + 8 * (Reg - PPC::X3); | |||
6815 | } | |||
6816 | ||||
6817 | llvm_unreachable("Only general purpose registers expected.")::llvm::llvm_unreachable_internal("Only general purpose registers expected." , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6817); | |||
6818 | } | |||
6819 | ||||
6820 | // AIX ABI Stack Frame Layout: | |||
6821 | // | |||
6822 | // Low Memory +--------------------------------------------+ | |||
6823 | // SP +---> | Back chain | ---+ | |||
6824 | // | +--------------------------------------------+ | | |||
6825 | // | | Saved Condition Register | | | |||
6826 | // | +--------------------------------------------+ | | |||
6827 | // | | Saved Linkage Register | | | |||
6828 | // | +--------------------------------------------+ | Linkage Area | |||
6829 | // | | Reserved for compilers | | | |||
6830 | // | +--------------------------------------------+ | | |||
6831 | // | | Reserved for binders | | | |||
6832 | // | +--------------------------------------------+ | | |||
6833 | // | | Saved TOC pointer | ---+ | |||
6834 | // | +--------------------------------------------+ | |||
6835 | // | | Parameter save area | | |||
6836 | // | +--------------------------------------------+ | |||
6837 | // | | Alloca space | | |||
6838 | // | +--------------------------------------------+ | |||
6839 | // | | Local variable space | | |||
6840 | // | +--------------------------------------------+ | |||
6841 | // | | Float/int conversion temporary | | |||
6842 | // | +--------------------------------------------+ | |||
6843 | // | | Save area for AltiVec registers | | |||
6844 | // | +--------------------------------------------+ | |||
6845 | // | | AltiVec alignment padding | | |||
6846 | // | +--------------------------------------------+ | |||
6847 | // | | Save area for VRSAVE register | | |||
6848 | // | +--------------------------------------------+ | |||
6849 | // | | Save area for General Purpose registers | | |||
6850 | // | +--------------------------------------------+ | |||
6851 | // | | Save area for Floating Point registers | | |||
6852 | // | +--------------------------------------------+ | |||
6853 | // +---- | Back chain | | |||
6854 | // High Memory +--------------------------------------------+ | |||
6855 | // | |||
6856 | // Specifications: | |||
6857 | // AIX 7.2 Assembler Language Reference | |||
6858 | // Subroutine linkage convention | |||
6859 | ||||
6860 | SDValue PPCTargetLowering::LowerFormalArguments_AIX( | |||
6861 | SDValue Chain, CallingConv::ID CallConv, bool isVarArg, | |||
6862 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
6863 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const { | |||
6864 | ||||
6865 | assert((CallConv == CallingConv::C || CallConv == CallingConv::Cold ||(static_cast <bool> ((CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && "Unexpected calling convention!") ? void (0) : __assert_fail ("(CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6867, __extension__ __PRETTY_FUNCTION__)) | |||
6866 | CallConv == CallingConv::Fast) &&(static_cast <bool> ((CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && "Unexpected calling convention!") ? void (0) : __assert_fail ("(CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6867, __extension__ __PRETTY_FUNCTION__)) | |||
6867 | "Unexpected calling convention!")(static_cast <bool> ((CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && "Unexpected calling convention!") ? void (0) : __assert_fail ("(CallConv == CallingConv::C || CallConv == CallingConv::Cold || CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6867, __extension__ __PRETTY_FUNCTION__)); | |||
6868 | ||||
6869 | if (getTargetMachine().Options.GuaranteedTailCallOpt) | |||
6870 | report_fatal_error("Tail call support is unimplemented on AIX."); | |||
6871 | ||||
6872 | if (useSoftFloat()) | |||
6873 | report_fatal_error("Soft float support is unimplemented on AIX."); | |||
6874 | ||||
6875 | const PPCSubtarget &Subtarget = | |||
6876 | static_cast<const PPCSubtarget &>(DAG.getSubtarget()); | |||
6877 | ||||
6878 | const bool IsPPC64 = Subtarget.isPPC64(); | |||
6879 | const unsigned PtrByteSize = IsPPC64 ? 8 : 4; | |||
6880 | ||||
6881 | // Assign locations to all of the incoming arguments. | |||
6882 | SmallVector<CCValAssign, 16> ArgLocs; | |||
6883 | MachineFunction &MF = DAG.getMachineFunction(); | |||
6884 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
6885 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
6886 | AIXCCState CCInfo(CallConv, isVarArg, MF, ArgLocs, *DAG.getContext()); | |||
6887 | ||||
6888 | const EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
6889 | // Reserve space for the linkage area on the stack. | |||
6890 | const unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize(); | |||
6891 | CCInfo.AllocateStack(LinkageSize, Align(PtrByteSize)); | |||
6892 | CCInfo.AnalyzeFormalArguments(Ins, CC_AIX); | |||
6893 | ||||
6894 | SmallVector<SDValue, 8> MemOps; | |||
6895 | ||||
6896 | for (size_t I = 0, End = ArgLocs.size(); I != End; /* No increment here */) { | |||
6897 | CCValAssign &VA = ArgLocs[I++]; | |||
6898 | MVT LocVT = VA.getLocVT(); | |||
6899 | MVT ValVT = VA.getValVT(); | |||
6900 | ISD::ArgFlagsTy Flags = Ins[VA.getValNo()].Flags; | |||
6901 | // For compatibility with the AIX XL compiler, the float args in the | |||
6902 | // parameter save area are initialized even if the argument is available | |||
6903 | // in register. The caller is required to initialize both the register | |||
6904 | // and memory, however, the callee can choose to expect it in either. | |||
6905 | // The memloc is dismissed here because the argument is retrieved from | |||
6906 | // the register. | |||
6907 | if (VA.isMemLoc() && VA.needsCustom() && ValVT.isFloatingPoint()) | |||
6908 | continue; | |||
6909 | ||||
6910 | auto HandleMemLoc = [&]() { | |||
6911 | const unsigned LocSize = LocVT.getStoreSize(); | |||
6912 | const unsigned ValSize = ValVT.getStoreSize(); | |||
6913 | assert((ValSize <= LocSize) &&(static_cast <bool> ((ValSize <= LocSize) && "Object size is larger than size of MemLoc") ? void (0) : __assert_fail ("(ValSize <= LocSize) && \"Object size is larger than size of MemLoc\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6914, __extension__ __PRETTY_FUNCTION__)) | |||
6914 | "Object size is larger than size of MemLoc")(static_cast <bool> ((ValSize <= LocSize) && "Object size is larger than size of MemLoc") ? void (0) : __assert_fail ("(ValSize <= LocSize) && \"Object size is larger than size of MemLoc\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6914, __extension__ __PRETTY_FUNCTION__)); | |||
6915 | int CurArgOffset = VA.getLocMemOffset(); | |||
6916 | // Objects are right-justified because AIX is big-endian. | |||
6917 | if (LocSize > ValSize) | |||
6918 | CurArgOffset += LocSize - ValSize; | |||
6919 | // Potential tail calls could cause overwriting of argument stack slots. | |||
6920 | const bool IsImmutable = | |||
6921 | !(getTargetMachine().Options.GuaranteedTailCallOpt && | |||
6922 | (CallConv == CallingConv::Fast)); | |||
6923 | int FI = MFI.CreateFixedObject(ValSize, CurArgOffset, IsImmutable); | |||
6924 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
6925 | SDValue ArgValue = | |||
6926 | DAG.getLoad(ValVT, dl, Chain, FIN, MachinePointerInfo()); | |||
6927 | InVals.push_back(ArgValue); | |||
6928 | }; | |||
6929 | ||||
6930 | // Vector arguments to VaArg functions are passed both on the stack, and | |||
6931 | // in any available GPRs. Load the value from the stack and add the GPRs | |||
6932 | // as live ins. | |||
6933 | if (VA.isMemLoc() && VA.needsCustom()) { | |||
6934 | assert(ValVT.isVector() && "Unexpected Custom MemLoc type.")(static_cast <bool> (ValVT.isVector() && "Unexpected Custom MemLoc type." ) ? void (0) : __assert_fail ("ValVT.isVector() && \"Unexpected Custom MemLoc type.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6934, __extension__ __PRETTY_FUNCTION__)); | |||
6935 | assert(isVarArg && "Only use custom memloc for vararg.")(static_cast <bool> (isVarArg && "Only use custom memloc for vararg." ) ? void (0) : __assert_fail ("isVarArg && \"Only use custom memloc for vararg.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6935, __extension__ __PRETTY_FUNCTION__)); | |||
6936 | // ValNo of the custom MemLoc, so we can compare it to the ValNo of the | |||
6937 | // matching custom RegLocs. | |||
6938 | const unsigned OriginalValNo = VA.getValNo(); | |||
6939 | (void)OriginalValNo; | |||
6940 | ||||
6941 | auto HandleCustomVecRegLoc = [&]() { | |||
6942 | assert(I != End && ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() &&(static_cast <bool> (I != End && ArgLocs[I].isRegLoc () && ArgLocs[I].needsCustom() && "Missing custom RegLoc." ) ? void (0) : __assert_fail ("I != End && ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() && \"Missing custom RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6943, __extension__ __PRETTY_FUNCTION__)) | |||
6943 | "Missing custom RegLoc.")(static_cast <bool> (I != End && ArgLocs[I].isRegLoc () && ArgLocs[I].needsCustom() && "Missing custom RegLoc." ) ? void (0) : __assert_fail ("I != End && ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() && \"Missing custom RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6943, __extension__ __PRETTY_FUNCTION__)); | |||
6944 | VA = ArgLocs[I++]; | |||
6945 | assert(VA.getValVT().isVector() &&(static_cast <bool> (VA.getValVT().isVector() && "Unexpected Val type for custom RegLoc.") ? void (0) : __assert_fail ("VA.getValVT().isVector() && \"Unexpected Val type for custom RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6946, __extension__ __PRETTY_FUNCTION__)) | |||
6946 | "Unexpected Val type for custom RegLoc.")(static_cast <bool> (VA.getValVT().isVector() && "Unexpected Val type for custom RegLoc.") ? void (0) : __assert_fail ("VA.getValVT().isVector() && \"Unexpected Val type for custom RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6946, __extension__ __PRETTY_FUNCTION__)); | |||
6947 | assert(VA.getValNo() == OriginalValNo &&(static_cast <bool> (VA.getValNo() == OriginalValNo && "ValNo mismatch between custom MemLoc and RegLoc.") ? void ( 0) : __assert_fail ("VA.getValNo() == OriginalValNo && \"ValNo mismatch between custom MemLoc and RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6948, __extension__ __PRETTY_FUNCTION__)) | |||
6948 | "ValNo mismatch between custom MemLoc and RegLoc.")(static_cast <bool> (VA.getValNo() == OriginalValNo && "ValNo mismatch between custom MemLoc and RegLoc.") ? void ( 0) : __assert_fail ("VA.getValNo() == OriginalValNo && \"ValNo mismatch between custom MemLoc and RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6948, __extension__ __PRETTY_FUNCTION__)); | |||
6949 | MVT::SimpleValueType SVT = VA.getLocVT().SimpleTy; | |||
6950 | MF.addLiveIn(VA.getLocReg(), | |||
6951 | getRegClassForSVT(SVT, IsPPC64, Subtarget.hasP8Vector(), | |||
6952 | Subtarget.hasVSX())); | |||
6953 | }; | |||
6954 | ||||
6955 | HandleMemLoc(); | |||
6956 | // In 64-bit there will be exactly 2 custom RegLocs that follow, and in | |||
6957 | // in 32-bit there will be 2 custom RegLocs if we are passing in R9 and | |||
6958 | // R10. | |||
6959 | HandleCustomVecRegLoc(); | |||
6960 | HandleCustomVecRegLoc(); | |||
6961 | ||||
6962 | // If we are targeting 32-bit, there might be 2 extra custom RegLocs if | |||
6963 | // we passed the vector in R5, R6, R7 and R8. | |||
6964 | if (I != End && ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom()) { | |||
6965 | assert(!IsPPC64 &&(static_cast <bool> (!IsPPC64 && "Only 2 custom RegLocs expected for 64-bit codegen." ) ? void (0) : __assert_fail ("!IsPPC64 && \"Only 2 custom RegLocs expected for 64-bit codegen.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6966, __extension__ __PRETTY_FUNCTION__)) | |||
6966 | "Only 2 custom RegLocs expected for 64-bit codegen.")(static_cast <bool> (!IsPPC64 && "Only 2 custom RegLocs expected for 64-bit codegen." ) ? void (0) : __assert_fail ("!IsPPC64 && \"Only 2 custom RegLocs expected for 64-bit codegen.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 6966, __extension__ __PRETTY_FUNCTION__)); | |||
6967 | HandleCustomVecRegLoc(); | |||
6968 | HandleCustomVecRegLoc(); | |||
6969 | } | |||
6970 | ||||
6971 | continue; | |||
6972 | } | |||
6973 | ||||
6974 | if (VA.isRegLoc()) { | |||
6975 | if (VA.getValVT().isScalarInteger()) | |||
6976 | FuncInfo->appendParameterType(PPCFunctionInfo::FixedType); | |||
6977 | else if (VA.getValVT().isFloatingPoint() && !VA.getValVT().isVector()) { | |||
6978 | switch (VA.getValVT().SimpleTy) { | |||
6979 | default: | |||
6980 | report_fatal_error("Unhandled value type for argument."); | |||
6981 | case MVT::f32: | |||
6982 | FuncInfo->appendParameterType(PPCFunctionInfo::ShortFloatingPoint); | |||
6983 | break; | |||
6984 | case MVT::f64: | |||
6985 | FuncInfo->appendParameterType(PPCFunctionInfo::LongFloatingPoint); | |||
6986 | break; | |||
6987 | } | |||
6988 | } else if (VA.getValVT().isVector()) { | |||
6989 | switch (VA.getValVT().SimpleTy) { | |||
6990 | default: | |||
6991 | report_fatal_error("Unhandled value type for argument."); | |||
6992 | case MVT::v16i8: | |||
6993 | FuncInfo->appendParameterType(PPCFunctionInfo::VectorChar); | |||
6994 | break; | |||
6995 | case MVT::v8i16: | |||
6996 | FuncInfo->appendParameterType(PPCFunctionInfo::VectorShort); | |||
6997 | break; | |||
6998 | case MVT::v4i32: | |||
6999 | case MVT::v2i64: | |||
7000 | case MVT::v1i128: | |||
7001 | FuncInfo->appendParameterType(PPCFunctionInfo::VectorInt); | |||
7002 | break; | |||
7003 | case MVT::v4f32: | |||
7004 | case MVT::v2f64: | |||
7005 | FuncInfo->appendParameterType(PPCFunctionInfo::VectorFloat); | |||
7006 | break; | |||
7007 | } | |||
7008 | } | |||
7009 | } | |||
7010 | ||||
7011 | if (Flags.isByVal() && VA.isMemLoc()) { | |||
7012 | const unsigned Size = | |||
7013 | alignTo(Flags.getByValSize() ? Flags.getByValSize() : PtrByteSize, | |||
7014 | PtrByteSize); | |||
7015 | const int FI = MF.getFrameInfo().CreateFixedObject( | |||
7016 | Size, VA.getLocMemOffset(), /* IsImmutable */ false, | |||
7017 | /* IsAliased */ true); | |||
7018 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
7019 | InVals.push_back(FIN); | |||
7020 | ||||
7021 | continue; | |||
7022 | } | |||
7023 | ||||
7024 | if (Flags.isByVal()) { | |||
7025 | assert(VA.isRegLoc() && "MemLocs should already be handled.")(static_cast <bool> (VA.isRegLoc() && "MemLocs should already be handled." ) ? void (0) : __assert_fail ("VA.isRegLoc() && \"MemLocs should already be handled.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7025, __extension__ __PRETTY_FUNCTION__)); | |||
7026 | ||||
7027 | const MCPhysReg ArgReg = VA.getLocReg(); | |||
7028 | const PPCFrameLowering *FL = Subtarget.getFrameLowering(); | |||
7029 | ||||
7030 | if (Flags.getNonZeroByValAlign() > PtrByteSize) | |||
7031 | report_fatal_error("Over aligned byvals not supported yet."); | |||
7032 | ||||
7033 | const unsigned StackSize = alignTo(Flags.getByValSize(), PtrByteSize); | |||
7034 | const int FI = MF.getFrameInfo().CreateFixedObject( | |||
7035 | StackSize, mapArgRegToOffsetAIX(ArgReg, FL), /* IsImmutable */ false, | |||
7036 | /* IsAliased */ true); | |||
7037 | SDValue FIN = DAG.getFrameIndex(FI, PtrVT); | |||
7038 | InVals.push_back(FIN); | |||
7039 | ||||
7040 | // Add live ins for all the RegLocs for the same ByVal. | |||
7041 | const TargetRegisterClass *RegClass = | |||
7042 | IsPPC64 ? &PPC::G8RCRegClass : &PPC::GPRCRegClass; | |||
7043 | ||||
7044 | auto HandleRegLoc = [&, RegClass, LocVT](const MCPhysReg PhysReg, | |||
7045 | unsigned Offset) { | |||
7046 | const unsigned VReg = MF.addLiveIn(PhysReg, RegClass); | |||
7047 | // Since the callers side has left justified the aggregate in the | |||
7048 | // register, we can simply store the entire register into the stack | |||
7049 | // slot. | |||
7050 | SDValue CopyFrom = DAG.getCopyFromReg(Chain, dl, VReg, LocVT); | |||
7051 | // The store to the fixedstack object is needed becuase accessing a | |||
7052 | // field of the ByVal will use a gep and load. Ideally we will optimize | |||
7053 | // to extracting the value from the register directly, and elide the | |||
7054 | // stores when the arguments address is not taken, but that will need to | |||
7055 | // be future work. | |||
7056 | SDValue Store = DAG.getStore( | |||
7057 | CopyFrom.getValue(1), dl, CopyFrom, | |||
7058 | DAG.getObjectPtrOffset(dl, FIN, TypeSize::Fixed(Offset)), | |||
7059 | MachinePointerInfo::getFixedStack(MF, FI, Offset)); | |||
7060 | ||||
7061 | MemOps.push_back(Store); | |||
7062 | }; | |||
7063 | ||||
7064 | unsigned Offset = 0; | |||
7065 | HandleRegLoc(VA.getLocReg(), Offset); | |||
7066 | Offset += PtrByteSize; | |||
7067 | for (; Offset != StackSize && ArgLocs[I].isRegLoc(); | |||
7068 | Offset += PtrByteSize) { | |||
7069 | assert(ArgLocs[I].getValNo() == VA.getValNo() &&(static_cast <bool> (ArgLocs[I].getValNo() == VA.getValNo () && "RegLocs should be for ByVal argument.") ? void (0) : __assert_fail ("ArgLocs[I].getValNo() == VA.getValNo() && \"RegLocs should be for ByVal argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7070, __extension__ __PRETTY_FUNCTION__)) | |||
7070 | "RegLocs should be for ByVal argument.")(static_cast <bool> (ArgLocs[I].getValNo() == VA.getValNo () && "RegLocs should be for ByVal argument.") ? void (0) : __assert_fail ("ArgLocs[I].getValNo() == VA.getValNo() && \"RegLocs should be for ByVal argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7070, __extension__ __PRETTY_FUNCTION__)); | |||
7071 | ||||
7072 | const CCValAssign RL = ArgLocs[I++]; | |||
7073 | HandleRegLoc(RL.getLocReg(), Offset); | |||
7074 | FuncInfo->appendParameterType(PPCFunctionInfo::FixedType); | |||
7075 | } | |||
7076 | ||||
7077 | if (Offset != StackSize) { | |||
7078 | assert(ArgLocs[I].getValNo() == VA.getValNo() &&(static_cast <bool> (ArgLocs[I].getValNo() == VA.getValNo () && "Expected MemLoc for remaining bytes.") ? void ( 0) : __assert_fail ("ArgLocs[I].getValNo() == VA.getValNo() && \"Expected MemLoc for remaining bytes.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7079, __extension__ __PRETTY_FUNCTION__)) | |||
7079 | "Expected MemLoc for remaining bytes.")(static_cast <bool> (ArgLocs[I].getValNo() == VA.getValNo () && "Expected MemLoc for remaining bytes.") ? void ( 0) : __assert_fail ("ArgLocs[I].getValNo() == VA.getValNo() && \"Expected MemLoc for remaining bytes.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7079, __extension__ __PRETTY_FUNCTION__)); | |||
7080 | assert(ArgLocs[I].isMemLoc() && "Expected MemLoc for remaining bytes.")(static_cast <bool> (ArgLocs[I].isMemLoc() && "Expected MemLoc for remaining bytes." ) ? void (0) : __assert_fail ("ArgLocs[I].isMemLoc() && \"Expected MemLoc for remaining bytes.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7080, __extension__ __PRETTY_FUNCTION__)); | |||
7081 | // Consume the MemLoc.The InVal has already been emitted, so nothing | |||
7082 | // more needs to be done. | |||
7083 | ++I; | |||
7084 | } | |||
7085 | ||||
7086 | continue; | |||
7087 | } | |||
7088 | ||||
7089 | if (VA.isRegLoc() && !VA.needsCustom()) { | |||
7090 | MVT::SimpleValueType SVT = ValVT.SimpleTy; | |||
7091 | Register VReg = | |||
7092 | MF.addLiveIn(VA.getLocReg(), | |||
7093 | getRegClassForSVT(SVT, IsPPC64, Subtarget.hasP8Vector(), | |||
7094 | Subtarget.hasVSX())); | |||
7095 | SDValue ArgValue = DAG.getCopyFromReg(Chain, dl, VReg, LocVT); | |||
7096 | if (ValVT.isScalarInteger() && | |||
7097 | (ValVT.getFixedSizeInBits() < LocVT.getFixedSizeInBits())) { | |||
7098 | ArgValue = | |||
7099 | truncateScalarIntegerArg(Flags, ValVT, DAG, ArgValue, LocVT, dl); | |||
7100 | } | |||
7101 | InVals.push_back(ArgValue); | |||
7102 | continue; | |||
7103 | } | |||
7104 | if (VA.isMemLoc()) { | |||
7105 | HandleMemLoc(); | |||
7106 | continue; | |||
7107 | } | |||
7108 | } | |||
7109 | ||||
7110 | // On AIX a minimum of 8 words is saved to the parameter save area. | |||
7111 | const unsigned MinParameterSaveArea = 8 * PtrByteSize; | |||
7112 | // Area that is at least reserved in the caller of this function. | |||
7113 | unsigned CallerReservedArea = | |||
7114 | std::max(CCInfo.getNextStackOffset(), LinkageSize + MinParameterSaveArea); | |||
7115 | ||||
7116 | // Set the size that is at least reserved in caller of this function. Tail | |||
7117 | // call optimized function's reserved stack space needs to be aligned so | |||
7118 | // that taking the difference between two stack areas will result in an | |||
7119 | // aligned stack. | |||
7120 | CallerReservedArea = | |||
7121 | EnsureStackAlignment(Subtarget.getFrameLowering(), CallerReservedArea); | |||
7122 | FuncInfo->setMinReservedArea(CallerReservedArea); | |||
7123 | ||||
7124 | if (isVarArg) { | |||
7125 | FuncInfo->setVarArgsFrameIndex( | |||
7126 | MFI.CreateFixedObject(PtrByteSize, CCInfo.getNextStackOffset(), true)); | |||
7127 | SDValue FIN = DAG.getFrameIndex(FuncInfo->getVarArgsFrameIndex(), PtrVT); | |||
7128 | ||||
7129 | static const MCPhysReg GPR_32[] = {PPC::R3, PPC::R4, PPC::R5, PPC::R6, | |||
7130 | PPC::R7, PPC::R8, PPC::R9, PPC::R10}; | |||
7131 | ||||
7132 | static const MCPhysReg GPR_64[] = {PPC::X3, PPC::X4, PPC::X5, PPC::X6, | |||
7133 | PPC::X7, PPC::X8, PPC::X9, PPC::X10}; | |||
7134 | const unsigned NumGPArgRegs = array_lengthof(IsPPC64 ? GPR_64 : GPR_32); | |||
7135 | ||||
7136 | // The fixed integer arguments of a variadic function are stored to the | |||
7137 | // VarArgsFrameIndex on the stack so that they may be loaded by | |||
7138 | // dereferencing the result of va_next. | |||
7139 | for (unsigned GPRIndex = | |||
7140 | (CCInfo.getNextStackOffset() - LinkageSize) / PtrByteSize; | |||
7141 | GPRIndex < NumGPArgRegs; ++GPRIndex) { | |||
7142 | ||||
7143 | const unsigned VReg = | |||
7144 | IsPPC64 ? MF.addLiveIn(GPR_64[GPRIndex], &PPC::G8RCRegClass) | |||
7145 | : MF.addLiveIn(GPR_32[GPRIndex], &PPC::GPRCRegClass); | |||
7146 | ||||
7147 | SDValue Val = DAG.getCopyFromReg(Chain, dl, VReg, PtrVT); | |||
7148 | SDValue Store = | |||
7149 | DAG.getStore(Val.getValue(1), dl, Val, FIN, MachinePointerInfo()); | |||
7150 | MemOps.push_back(Store); | |||
7151 | // Increment the address for the next argument to store. | |||
7152 | SDValue PtrOff = DAG.getConstant(PtrByteSize, dl, PtrVT); | |||
7153 | FIN = DAG.getNode(ISD::ADD, dl, PtrOff.getValueType(), FIN, PtrOff); | |||
7154 | } | |||
7155 | } | |||
7156 | ||||
7157 | if (!MemOps.empty()) | |||
7158 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOps); | |||
7159 | ||||
7160 | return Chain; | |||
7161 | } | |||
7162 | ||||
7163 | SDValue PPCTargetLowering::LowerCall_AIX( | |||
7164 | SDValue Chain, SDValue Callee, CallFlags CFlags, | |||
7165 | const SmallVectorImpl<ISD::OutputArg> &Outs, | |||
7166 | const SmallVectorImpl<SDValue> &OutVals, | |||
7167 | const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl, | |||
7168 | SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals, | |||
7169 | const CallBase *CB) const { | |||
7170 | // See PPCTargetLowering::LowerFormalArguments_AIX() for a description of the | |||
7171 | // AIX ABI stack frame layout. | |||
7172 | ||||
7173 | assert((CFlags.CallConv == CallingConv::C ||(static_cast <bool> ((CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && "Unexpected calling convention!" ) ? void (0) : __assert_fail ("(CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7176, __extension__ __PRETTY_FUNCTION__)) | |||
7174 | CFlags.CallConv == CallingConv::Cold ||(static_cast <bool> ((CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && "Unexpected calling convention!" ) ? void (0) : __assert_fail ("(CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7176, __extension__ __PRETTY_FUNCTION__)) | |||
7175 | CFlags.CallConv == CallingConv::Fast) &&(static_cast <bool> ((CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && "Unexpected calling convention!" ) ? void (0) : __assert_fail ("(CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7176, __extension__ __PRETTY_FUNCTION__)) | |||
7176 | "Unexpected calling convention!")(static_cast <bool> ((CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && "Unexpected calling convention!" ) ? void (0) : __assert_fail ("(CFlags.CallConv == CallingConv::C || CFlags.CallConv == CallingConv::Cold || CFlags.CallConv == CallingConv::Fast) && \"Unexpected calling convention!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7176, __extension__ __PRETTY_FUNCTION__)); | |||
7177 | ||||
7178 | if (CFlags.IsPatchPoint) | |||
7179 | report_fatal_error("This call type is unimplemented on AIX."); | |||
7180 | ||||
7181 | const PPCSubtarget& Subtarget = | |||
7182 | static_cast<const PPCSubtarget&>(DAG.getSubtarget()); | |||
7183 | ||||
7184 | MachineFunction &MF = DAG.getMachineFunction(); | |||
7185 | SmallVector<CCValAssign, 16> ArgLocs; | |||
7186 | AIXCCState CCInfo(CFlags.CallConv, CFlags.IsVarArg, MF, ArgLocs, | |||
7187 | *DAG.getContext()); | |||
7188 | ||||
7189 | // Reserve space for the linkage save area (LSA) on the stack. | |||
7190 | // In both PPC32 and PPC64 there are 6 reserved slots in the LSA: | |||
7191 | // [SP][CR][LR][2 x reserved][TOC]. | |||
7192 | // The LSA is 24 bytes (6x4) in PPC32 and 48 bytes (6x8) in PPC64. | |||
7193 | const unsigned LinkageSize = Subtarget.getFrameLowering()->getLinkageSize(); | |||
7194 | const bool IsPPC64 = Subtarget.isPPC64(); | |||
7195 | const EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
7196 | const unsigned PtrByteSize = IsPPC64 ? 8 : 4; | |||
7197 | CCInfo.AllocateStack(LinkageSize, Align(PtrByteSize)); | |||
7198 | CCInfo.AnalyzeCallOperands(Outs, CC_AIX); | |||
7199 | ||||
7200 | // The prolog code of the callee may store up to 8 GPR argument registers to | |||
7201 | // the stack, allowing va_start to index over them in memory if the callee | |||
7202 | // is variadic. | |||
7203 | // Because we cannot tell if this is needed on the caller side, we have to | |||
7204 | // conservatively assume that it is needed. As such, make sure we have at | |||
7205 | // least enough stack space for the caller to store the 8 GPRs. | |||
7206 | const unsigned MinParameterSaveAreaSize = 8 * PtrByteSize; | |||
7207 | const unsigned NumBytes = std::max(LinkageSize + MinParameterSaveAreaSize, | |||
7208 | CCInfo.getNextStackOffset()); | |||
7209 | ||||
7210 | // Adjust the stack pointer for the new arguments... | |||
7211 | // These operations are automatically eliminated by the prolog/epilog pass. | |||
7212 | Chain = DAG.getCALLSEQ_START(Chain, NumBytes, 0, dl); | |||
7213 | SDValue CallSeqStart = Chain; | |||
7214 | ||||
7215 | SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass; | |||
7216 | SmallVector<SDValue, 8> MemOpChains; | |||
7217 | ||||
7218 | // Set up a copy of the stack pointer for loading and storing any | |||
7219 | // arguments that may not fit in the registers available for argument | |||
7220 | // passing. | |||
7221 | const SDValue StackPtr = IsPPC64 ? DAG.getRegister(PPC::X1, MVT::i64) | |||
7222 | : DAG.getRegister(PPC::R1, MVT::i32); | |||
7223 | ||||
7224 | for (unsigned I = 0, E = ArgLocs.size(); I != E;) { | |||
7225 | const unsigned ValNo = ArgLocs[I].getValNo(); | |||
7226 | SDValue Arg = OutVals[ValNo]; | |||
7227 | ISD::ArgFlagsTy Flags = Outs[ValNo].Flags; | |||
7228 | ||||
7229 | if (Flags.isByVal()) { | |||
7230 | const unsigned ByValSize = Flags.getByValSize(); | |||
7231 | ||||
7232 | // Nothing to do for zero-sized ByVals on the caller side. | |||
7233 | if (!ByValSize) { | |||
7234 | ++I; | |||
7235 | continue; | |||
7236 | } | |||
7237 | ||||
7238 | auto GetLoad = [&](EVT VT, unsigned LoadOffset) { | |||
7239 | return DAG.getExtLoad( | |||
7240 | ISD::ZEXTLOAD, dl, PtrVT, Chain, | |||
7241 | (LoadOffset != 0) | |||
7242 | ? DAG.getObjectPtrOffset(dl, Arg, TypeSize::Fixed(LoadOffset)) | |||
7243 | : Arg, | |||
7244 | MachinePointerInfo(), VT); | |||
7245 | }; | |||
7246 | ||||
7247 | unsigned LoadOffset = 0; | |||
7248 | ||||
7249 | // Initialize registers, which are fully occupied by the by-val argument. | |||
7250 | while (LoadOffset + PtrByteSize <= ByValSize && ArgLocs[I].isRegLoc()) { | |||
7251 | SDValue Load = GetLoad(PtrVT, LoadOffset); | |||
7252 | MemOpChains.push_back(Load.getValue(1)); | |||
7253 | LoadOffset += PtrByteSize; | |||
7254 | const CCValAssign &ByValVA = ArgLocs[I++]; | |||
7255 | assert(ByValVA.getValNo() == ValNo &&(static_cast <bool> (ByValVA.getValNo() == ValNo && "Unexpected location for pass-by-value argument.") ? void (0 ) : __assert_fail ("ByValVA.getValNo() == ValNo && \"Unexpected location for pass-by-value argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7256, __extension__ __PRETTY_FUNCTION__)) | |||
7256 | "Unexpected location for pass-by-value argument.")(static_cast <bool> (ByValVA.getValNo() == ValNo && "Unexpected location for pass-by-value argument.") ? void (0 ) : __assert_fail ("ByValVA.getValNo() == ValNo && \"Unexpected location for pass-by-value argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7256, __extension__ __PRETTY_FUNCTION__)); | |||
7257 | RegsToPass.push_back(std::make_pair(ByValVA.getLocReg(), Load)); | |||
7258 | } | |||
7259 | ||||
7260 | if (LoadOffset == ByValSize) | |||
7261 | continue; | |||
7262 | ||||
7263 | // There must be one more loc to handle the remainder. | |||
7264 | assert(ArgLocs[I].getValNo() == ValNo &&(static_cast <bool> (ArgLocs[I].getValNo() == ValNo && "Expected additional location for by-value argument.") ? void (0) : __assert_fail ("ArgLocs[I].getValNo() == ValNo && \"Expected additional location for by-value argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7265, __extension__ __PRETTY_FUNCTION__)) | |||
7265 | "Expected additional location for by-value argument.")(static_cast <bool> (ArgLocs[I].getValNo() == ValNo && "Expected additional location for by-value argument.") ? void (0) : __assert_fail ("ArgLocs[I].getValNo() == ValNo && \"Expected additional location for by-value argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7265, __extension__ __PRETTY_FUNCTION__)); | |||
7266 | ||||
7267 | if (ArgLocs[I].isMemLoc()) { | |||
7268 | assert(LoadOffset < ByValSize && "Unexpected memloc for by-val arg.")(static_cast <bool> (LoadOffset < ByValSize && "Unexpected memloc for by-val arg.") ? void (0) : __assert_fail ("LoadOffset < ByValSize && \"Unexpected memloc for by-val arg.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7268, __extension__ __PRETTY_FUNCTION__)); | |||
7269 | const CCValAssign &ByValVA = ArgLocs[I++]; | |||
7270 | ISD::ArgFlagsTy MemcpyFlags = Flags; | |||
7271 | // Only memcpy the bytes that don't pass in register. | |||
7272 | MemcpyFlags.setByValSize(ByValSize - LoadOffset); | |||
7273 | Chain = CallSeqStart = createMemcpyOutsideCallSeq( | |||
7274 | (LoadOffset != 0) | |||
7275 | ? DAG.getObjectPtrOffset(dl, Arg, TypeSize::Fixed(LoadOffset)) | |||
7276 | : Arg, | |||
7277 | DAG.getObjectPtrOffset(dl, StackPtr, | |||
7278 | TypeSize::Fixed(ByValVA.getLocMemOffset())), | |||
7279 | CallSeqStart, MemcpyFlags, DAG, dl); | |||
7280 | continue; | |||
7281 | } | |||
7282 | ||||
7283 | // Initialize the final register residue. | |||
7284 | // Any residue that occupies the final by-val arg register must be | |||
7285 | // left-justified on AIX. Loads must be a power-of-2 size and cannot be | |||
7286 | // larger than the ByValSize. For example: a 7 byte by-val arg requires 4, | |||
7287 | // 2 and 1 byte loads. | |||
7288 | const unsigned ResidueBytes = ByValSize % PtrByteSize; | |||
7289 | assert(ResidueBytes != 0 && LoadOffset + PtrByteSize > ByValSize &&(static_cast <bool> (ResidueBytes != 0 && LoadOffset + PtrByteSize > ByValSize && "Unexpected register residue for by-value argument." ) ? void (0) : __assert_fail ("ResidueBytes != 0 && LoadOffset + PtrByteSize > ByValSize && \"Unexpected register residue for by-value argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7290, __extension__ __PRETTY_FUNCTION__)) | |||
7290 | "Unexpected register residue for by-value argument.")(static_cast <bool> (ResidueBytes != 0 && LoadOffset + PtrByteSize > ByValSize && "Unexpected register residue for by-value argument." ) ? void (0) : __assert_fail ("ResidueBytes != 0 && LoadOffset + PtrByteSize > ByValSize && \"Unexpected register residue for by-value argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7290, __extension__ __PRETTY_FUNCTION__)); | |||
7291 | SDValue ResidueVal; | |||
7292 | for (unsigned Bytes = 0; Bytes != ResidueBytes;) { | |||
7293 | const unsigned N = PowerOf2Floor(ResidueBytes - Bytes); | |||
7294 | const MVT VT = | |||
7295 | N == 1 ? MVT::i8 | |||
7296 | : ((N == 2) ? MVT::i16 : (N == 4 ? MVT::i32 : MVT::i64)); | |||
7297 | SDValue Load = GetLoad(VT, LoadOffset); | |||
7298 | MemOpChains.push_back(Load.getValue(1)); | |||
7299 | LoadOffset += N; | |||
7300 | Bytes += N; | |||
7301 | ||||
7302 | // By-val arguments are passed left-justfied in register. | |||
7303 | // Every load here needs to be shifted, otherwise a full register load | |||
7304 | // should have been used. | |||
7305 | assert(PtrVT.getSimpleVT().getSizeInBits() > (Bytes * 8) &&(static_cast <bool> (PtrVT.getSimpleVT().getSizeInBits( ) > (Bytes * 8) && "Unexpected load emitted during handling of pass-by-value " "argument.") ? void (0) : __assert_fail ("PtrVT.getSimpleVT().getSizeInBits() > (Bytes * 8) && \"Unexpected load emitted during handling of pass-by-value \" \"argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7307, __extension__ __PRETTY_FUNCTION__)) | |||
7306 | "Unexpected load emitted during handling of pass-by-value "(static_cast <bool> (PtrVT.getSimpleVT().getSizeInBits( ) > (Bytes * 8) && "Unexpected load emitted during handling of pass-by-value " "argument.") ? void (0) : __assert_fail ("PtrVT.getSimpleVT().getSizeInBits() > (Bytes * 8) && \"Unexpected load emitted during handling of pass-by-value \" \"argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7307, __extension__ __PRETTY_FUNCTION__)) | |||
7307 | "argument.")(static_cast <bool> (PtrVT.getSimpleVT().getSizeInBits( ) > (Bytes * 8) && "Unexpected load emitted during handling of pass-by-value " "argument.") ? void (0) : __assert_fail ("PtrVT.getSimpleVT().getSizeInBits() > (Bytes * 8) && \"Unexpected load emitted during handling of pass-by-value \" \"argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7307, __extension__ __PRETTY_FUNCTION__)); | |||
7308 | unsigned NumSHLBits = PtrVT.getSimpleVT().getSizeInBits() - (Bytes * 8); | |||
7309 | EVT ShiftAmountTy = | |||
7310 | getShiftAmountTy(Load->getValueType(0), DAG.getDataLayout()); | |||
7311 | SDValue SHLAmt = DAG.getConstant(NumSHLBits, dl, ShiftAmountTy); | |||
7312 | SDValue ShiftedLoad = | |||
7313 | DAG.getNode(ISD::SHL, dl, Load.getValueType(), Load, SHLAmt); | |||
7314 | ResidueVal = ResidueVal ? DAG.getNode(ISD::OR, dl, PtrVT, ResidueVal, | |||
7315 | ShiftedLoad) | |||
7316 | : ShiftedLoad; | |||
7317 | } | |||
7318 | ||||
7319 | const CCValAssign &ByValVA = ArgLocs[I++]; | |||
7320 | RegsToPass.push_back(std::make_pair(ByValVA.getLocReg(), ResidueVal)); | |||
7321 | continue; | |||
7322 | } | |||
7323 | ||||
7324 | CCValAssign &VA = ArgLocs[I++]; | |||
7325 | const MVT LocVT = VA.getLocVT(); | |||
7326 | const MVT ValVT = VA.getValVT(); | |||
7327 | ||||
7328 | switch (VA.getLocInfo()) { | |||
7329 | default: | |||
7330 | report_fatal_error("Unexpected argument extension type."); | |||
7331 | case CCValAssign::Full: | |||
7332 | break; | |||
7333 | case CCValAssign::ZExt: | |||
7334 | Arg = DAG.getNode(ISD::ZERO_EXTEND, dl, VA.getLocVT(), Arg); | |||
7335 | break; | |||
7336 | case CCValAssign::SExt: | |||
7337 | Arg = DAG.getNode(ISD::SIGN_EXTEND, dl, VA.getLocVT(), Arg); | |||
7338 | break; | |||
7339 | } | |||
7340 | ||||
7341 | if (VA.isRegLoc() && !VA.needsCustom()) { | |||
7342 | RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg)); | |||
7343 | continue; | |||
7344 | } | |||
7345 | ||||
7346 | // Vector arguments passed to VarArg functions need custom handling when | |||
7347 | // they are passed (at least partially) in GPRs. | |||
7348 | if (VA.isMemLoc() && VA.needsCustom() && ValVT.isVector()) { | |||
7349 | assert(CFlags.IsVarArg && "Custom MemLocs only used for Vector args.")(static_cast <bool> (CFlags.IsVarArg && "Custom MemLocs only used for Vector args." ) ? void (0) : __assert_fail ("CFlags.IsVarArg && \"Custom MemLocs only used for Vector args.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7349, __extension__ __PRETTY_FUNCTION__)); | |||
7350 | // Store value to its stack slot. | |||
7351 | SDValue PtrOff = | |||
7352 | DAG.getConstant(VA.getLocMemOffset(), dl, StackPtr.getValueType()); | |||
7353 | PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff); | |||
7354 | SDValue Store = | |||
7355 | DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo()); | |||
7356 | MemOpChains.push_back(Store); | |||
7357 | const unsigned OriginalValNo = VA.getValNo(); | |||
7358 | // Then load the GPRs from the stack | |||
7359 | unsigned LoadOffset = 0; | |||
7360 | auto HandleCustomVecRegLoc = [&]() { | |||
7361 | assert(I != E && "Unexpected end of CCvalAssigns.")(static_cast <bool> (I != E && "Unexpected end of CCvalAssigns." ) ? void (0) : __assert_fail ("I != E && \"Unexpected end of CCvalAssigns.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7361, __extension__ __PRETTY_FUNCTION__)); | |||
7362 | assert(ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() &&(static_cast <bool> (ArgLocs[I].isRegLoc() && ArgLocs [I].needsCustom() && "Expected custom RegLoc.") ? void (0) : __assert_fail ("ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() && \"Expected custom RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7363, __extension__ __PRETTY_FUNCTION__)) | |||
7363 | "Expected custom RegLoc.")(static_cast <bool> (ArgLocs[I].isRegLoc() && ArgLocs [I].needsCustom() && "Expected custom RegLoc.") ? void (0) : __assert_fail ("ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() && \"Expected custom RegLoc.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7363, __extension__ __PRETTY_FUNCTION__)); | |||
7364 | CCValAssign RegVA = ArgLocs[I++]; | |||
7365 | assert(RegVA.getValNo() == OriginalValNo &&(static_cast <bool> (RegVA.getValNo() == OriginalValNo && "Custom MemLoc ValNo and custom RegLoc ValNo must match.") ? void (0) : __assert_fail ("RegVA.getValNo() == OriginalValNo && \"Custom MemLoc ValNo and custom RegLoc ValNo must match.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7366, __extension__ __PRETTY_FUNCTION__)) | |||
7366 | "Custom MemLoc ValNo and custom RegLoc ValNo must match.")(static_cast <bool> (RegVA.getValNo() == OriginalValNo && "Custom MemLoc ValNo and custom RegLoc ValNo must match.") ? void (0) : __assert_fail ("RegVA.getValNo() == OriginalValNo && \"Custom MemLoc ValNo and custom RegLoc ValNo must match.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7366, __extension__ __PRETTY_FUNCTION__)); | |||
7367 | SDValue Add = DAG.getNode(ISD::ADD, dl, PtrVT, PtrOff, | |||
7368 | DAG.getConstant(LoadOffset, dl, PtrVT)); | |||
7369 | SDValue Load = DAG.getLoad(PtrVT, dl, Store, Add, MachinePointerInfo()); | |||
7370 | MemOpChains.push_back(Load.getValue(1)); | |||
7371 | RegsToPass.push_back(std::make_pair(RegVA.getLocReg(), Load)); | |||
7372 | LoadOffset += PtrByteSize; | |||
7373 | }; | |||
7374 | ||||
7375 | // In 64-bit there will be exactly 2 custom RegLocs that follow, and in | |||
7376 | // in 32-bit there will be 2 custom RegLocs if we are passing in R9 and | |||
7377 | // R10. | |||
7378 | HandleCustomVecRegLoc(); | |||
7379 | HandleCustomVecRegLoc(); | |||
7380 | ||||
7381 | if (I != E && ArgLocs[I].isRegLoc() && ArgLocs[I].needsCustom() && | |||
7382 | ArgLocs[I].getValNo() == OriginalValNo) { | |||
7383 | assert(!IsPPC64 &&(static_cast <bool> (!IsPPC64 && "Only 2 custom RegLocs expected for 64-bit codegen." ) ? void (0) : __assert_fail ("!IsPPC64 && \"Only 2 custom RegLocs expected for 64-bit codegen.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7384, __extension__ __PRETTY_FUNCTION__)) | |||
7384 | "Only 2 custom RegLocs expected for 64-bit codegen.")(static_cast <bool> (!IsPPC64 && "Only 2 custom RegLocs expected for 64-bit codegen." ) ? void (0) : __assert_fail ("!IsPPC64 && \"Only 2 custom RegLocs expected for 64-bit codegen.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7384, __extension__ __PRETTY_FUNCTION__)); | |||
7385 | HandleCustomVecRegLoc(); | |||
7386 | HandleCustomVecRegLoc(); | |||
7387 | } | |||
7388 | ||||
7389 | continue; | |||
7390 | } | |||
7391 | ||||
7392 | if (VA.isMemLoc()) { | |||
7393 | SDValue PtrOff = | |||
7394 | DAG.getConstant(VA.getLocMemOffset(), dl, StackPtr.getValueType()); | |||
7395 | PtrOff = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff); | |||
7396 | MemOpChains.push_back( | |||
7397 | DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo())); | |||
7398 | ||||
7399 | continue; | |||
7400 | } | |||
7401 | ||||
7402 | if (!ValVT.isFloatingPoint()) | |||
7403 | report_fatal_error( | |||
7404 | "Unexpected register handling for calling convention."); | |||
7405 | ||||
7406 | // Custom handling is used for GPR initializations for vararg float | |||
7407 | // arguments. | |||
7408 | assert(VA.isRegLoc() && VA.needsCustom() && CFlags.IsVarArg &&(static_cast <bool> (VA.isRegLoc() && VA.needsCustom () && CFlags.IsVarArg && LocVT.isInteger() && "Custom register handling only expected for VarArg.") ? void (0) : __assert_fail ("VA.isRegLoc() && VA.needsCustom() && CFlags.IsVarArg && LocVT.isInteger() && \"Custom register handling only expected for VarArg.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7410, __extension__ __PRETTY_FUNCTION__)) | |||
7409 | LocVT.isInteger() &&(static_cast <bool> (VA.isRegLoc() && VA.needsCustom () && CFlags.IsVarArg && LocVT.isInteger() && "Custom register handling only expected for VarArg.") ? void (0) : __assert_fail ("VA.isRegLoc() && VA.needsCustom() && CFlags.IsVarArg && LocVT.isInteger() && \"Custom register handling only expected for VarArg.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7410, __extension__ __PRETTY_FUNCTION__)) | |||
7410 | "Custom register handling only expected for VarArg.")(static_cast <bool> (VA.isRegLoc() && VA.needsCustom () && CFlags.IsVarArg && LocVT.isInteger() && "Custom register handling only expected for VarArg.") ? void (0) : __assert_fail ("VA.isRegLoc() && VA.needsCustom() && CFlags.IsVarArg && LocVT.isInteger() && \"Custom register handling only expected for VarArg.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7410, __extension__ __PRETTY_FUNCTION__)); | |||
7411 | ||||
7412 | SDValue ArgAsInt = | |||
7413 | DAG.getBitcast(MVT::getIntegerVT(ValVT.getSizeInBits()), Arg); | |||
7414 | ||||
7415 | if (Arg.getValueType().getStoreSize() == LocVT.getStoreSize()) | |||
7416 | // f32 in 32-bit GPR | |||
7417 | // f64 in 64-bit GPR | |||
7418 | RegsToPass.push_back(std::make_pair(VA.getLocReg(), ArgAsInt)); | |||
7419 | else if (Arg.getValueType().getFixedSizeInBits() < | |||
7420 | LocVT.getFixedSizeInBits()) | |||
7421 | // f32 in 64-bit GPR. | |||
7422 | RegsToPass.push_back(std::make_pair( | |||
7423 | VA.getLocReg(), DAG.getZExtOrTrunc(ArgAsInt, dl, LocVT))); | |||
7424 | else { | |||
7425 | // f64 in two 32-bit GPRs | |||
7426 | // The 2 GPRs are marked custom and expected to be adjacent in ArgLocs. | |||
7427 | assert(Arg.getValueType() == MVT::f64 && CFlags.IsVarArg && !IsPPC64 &&(static_cast <bool> (Arg.getValueType() == MVT::f64 && CFlags.IsVarArg && !IsPPC64 && "Unexpected custom register for argument!" ) ? void (0) : __assert_fail ("Arg.getValueType() == MVT::f64 && CFlags.IsVarArg && !IsPPC64 && \"Unexpected custom register for argument!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7428, __extension__ __PRETTY_FUNCTION__)) | |||
7428 | "Unexpected custom register for argument!")(static_cast <bool> (Arg.getValueType() == MVT::f64 && CFlags.IsVarArg && !IsPPC64 && "Unexpected custom register for argument!" ) ? void (0) : __assert_fail ("Arg.getValueType() == MVT::f64 && CFlags.IsVarArg && !IsPPC64 && \"Unexpected custom register for argument!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7428, __extension__ __PRETTY_FUNCTION__)); | |||
7429 | CCValAssign &GPR1 = VA; | |||
7430 | SDValue MSWAsI64 = DAG.getNode(ISD::SRL, dl, MVT::i64, ArgAsInt, | |||
7431 | DAG.getConstant(32, dl, MVT::i8)); | |||
7432 | RegsToPass.push_back(std::make_pair( | |||
7433 | GPR1.getLocReg(), DAG.getZExtOrTrunc(MSWAsI64, dl, MVT::i32))); | |||
7434 | ||||
7435 | if (I != E) { | |||
7436 | // If only 1 GPR was available, there will only be one custom GPR and | |||
7437 | // the argument will also pass in memory. | |||
7438 | CCValAssign &PeekArg = ArgLocs[I]; | |||
7439 | if (PeekArg.isRegLoc() && PeekArg.getValNo() == PeekArg.getValNo()) { | |||
7440 | assert(PeekArg.needsCustom() && "A second custom GPR is expected.")(static_cast <bool> (PeekArg.needsCustom() && "A second custom GPR is expected." ) ? void (0) : __assert_fail ("PeekArg.needsCustom() && \"A second custom GPR is expected.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7440, __extension__ __PRETTY_FUNCTION__)); | |||
7441 | CCValAssign &GPR2 = ArgLocs[I++]; | |||
7442 | RegsToPass.push_back(std::make_pair( | |||
7443 | GPR2.getLocReg(), DAG.getZExtOrTrunc(ArgAsInt, dl, MVT::i32))); | |||
7444 | } | |||
7445 | } | |||
7446 | } | |||
7447 | } | |||
7448 | ||||
7449 | if (!MemOpChains.empty()) | |||
7450 | Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains); | |||
7451 | ||||
7452 | // For indirect calls, we need to save the TOC base to the stack for | |||
7453 | // restoration after the call. | |||
7454 | if (CFlags.IsIndirect) { | |||
7455 | assert(!CFlags.IsTailCall && "Indirect tail-calls not supported.")(static_cast <bool> (!CFlags.IsTailCall && "Indirect tail-calls not supported." ) ? void (0) : __assert_fail ("!CFlags.IsTailCall && \"Indirect tail-calls not supported.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7455, __extension__ __PRETTY_FUNCTION__)); | |||
7456 | const MCRegister TOCBaseReg = Subtarget.getTOCPointerRegister(); | |||
7457 | const MCRegister StackPtrReg = Subtarget.getStackPointerRegister(); | |||
7458 | const MVT PtrVT = Subtarget.isPPC64() ? MVT::i64 : MVT::i32; | |||
7459 | const unsigned TOCSaveOffset = | |||
7460 | Subtarget.getFrameLowering()->getTOCSaveOffset(); | |||
7461 | ||||
7462 | setUsesTOCBasePtr(DAG); | |||
7463 | SDValue Val = DAG.getCopyFromReg(Chain, dl, TOCBaseReg, PtrVT); | |||
7464 | SDValue PtrOff = DAG.getIntPtrConstant(TOCSaveOffset, dl); | |||
7465 | SDValue StackPtr = DAG.getRegister(StackPtrReg, PtrVT); | |||
7466 | SDValue AddPtr = DAG.getNode(ISD::ADD, dl, PtrVT, StackPtr, PtrOff); | |||
7467 | Chain = DAG.getStore( | |||
7468 | Val.getValue(1), dl, Val, AddPtr, | |||
7469 | MachinePointerInfo::getStack(DAG.getMachineFunction(), TOCSaveOffset)); | |||
7470 | } | |||
7471 | ||||
7472 | // Build a sequence of copy-to-reg nodes chained together with token chain | |||
7473 | // and flag operands which copy the outgoing args into the appropriate regs. | |||
7474 | SDValue InFlag; | |||
7475 | for (auto Reg : RegsToPass) { | |||
7476 | Chain = DAG.getCopyToReg(Chain, dl, Reg.first, Reg.second, InFlag); | |||
7477 | InFlag = Chain.getValue(1); | |||
7478 | } | |||
7479 | ||||
7480 | const int SPDiff = 0; | |||
7481 | return FinishCall(CFlags, dl, DAG, RegsToPass, InFlag, Chain, CallSeqStart, | |||
7482 | Callee, SPDiff, NumBytes, Ins, InVals, CB); | |||
7483 | } | |||
7484 | ||||
7485 | bool | |||
7486 | PPCTargetLowering::CanLowerReturn(CallingConv::ID CallConv, | |||
7487 | MachineFunction &MF, bool isVarArg, | |||
7488 | const SmallVectorImpl<ISD::OutputArg> &Outs, | |||
7489 | LLVMContext &Context) const { | |||
7490 | SmallVector<CCValAssign, 16> RVLocs; | |||
7491 | CCState CCInfo(CallConv, isVarArg, MF, RVLocs, Context); | |||
7492 | return CCInfo.CheckReturn( | |||
7493 | Outs, (Subtarget.isSVR4ABI() && CallConv == CallingConv::Cold) | |||
7494 | ? RetCC_PPC_Cold | |||
7495 | : RetCC_PPC); | |||
7496 | } | |||
7497 | ||||
7498 | SDValue | |||
7499 | PPCTargetLowering::LowerReturn(SDValue Chain, CallingConv::ID CallConv, | |||
7500 | bool isVarArg, | |||
7501 | const SmallVectorImpl<ISD::OutputArg> &Outs, | |||
7502 | const SmallVectorImpl<SDValue> &OutVals, | |||
7503 | const SDLoc &dl, SelectionDAG &DAG) const { | |||
7504 | SmallVector<CCValAssign, 16> RVLocs; | |||
7505 | CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs, | |||
7506 | *DAG.getContext()); | |||
7507 | CCInfo.AnalyzeReturn(Outs, | |||
7508 | (Subtarget.isSVR4ABI() && CallConv == CallingConv::Cold) | |||
7509 | ? RetCC_PPC_Cold | |||
7510 | : RetCC_PPC); | |||
7511 | ||||
7512 | SDValue Flag; | |||
7513 | SmallVector<SDValue, 4> RetOps(1, Chain); | |||
7514 | ||||
7515 | // Copy the result values into the output registers. | |||
7516 | for (unsigned i = 0, RealResIdx = 0; i != RVLocs.size(); ++i, ++RealResIdx) { | |||
7517 | CCValAssign &VA = RVLocs[i]; | |||
7518 | assert(VA.isRegLoc() && "Can only return in registers!")(static_cast <bool> (VA.isRegLoc() && "Can only return in registers!" ) ? void (0) : __assert_fail ("VA.isRegLoc() && \"Can only return in registers!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7518, __extension__ __PRETTY_FUNCTION__)); | |||
7519 | ||||
7520 | SDValue Arg = OutVals[RealResIdx]; | |||
7521 | ||||
7522 | switch (VA.getLocInfo()) { | |||
7523 | default: llvm_unreachable("Unknown loc info!")::llvm::llvm_unreachable_internal("Unknown loc info!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7523); | |||
7524 | case CCValAssign::Full: break; | |||
7525 | case CCValAssign::AExt: | |||
7526 | Arg = DAG.getNode(ISD::ANY_EXTEND, dl, VA.getLocVT(), Arg); | |||
7527 | break; | |||
7528 | case CCValAssign::ZExt: | |||
7529 | Arg = DAG.getNode(ISD::ZERO_EXTEND, dl, VA.getLocVT(), Arg); | |||
7530 | break; | |||
7531 | case CCValAssign::SExt: | |||
7532 | Arg = DAG.getNode(ISD::SIGN_EXTEND, dl, VA.getLocVT(), Arg); | |||
7533 | break; | |||
7534 | } | |||
7535 | if (Subtarget.hasSPE() && VA.getLocVT() == MVT::f64) { | |||
7536 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
7537 | // Legalize ret f64 -> ret 2 x i32. | |||
7538 | SDValue SVal = | |||
7539 | DAG.getNode(PPCISD::EXTRACT_SPE, dl, MVT::i32, Arg, | |||
7540 | DAG.getIntPtrConstant(isLittleEndian ? 0 : 1, dl)); | |||
7541 | Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), SVal, Flag); | |||
7542 | RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT())); | |||
7543 | SVal = DAG.getNode(PPCISD::EXTRACT_SPE, dl, MVT::i32, Arg, | |||
7544 | DAG.getIntPtrConstant(isLittleEndian ? 1 : 0, dl)); | |||
7545 | Flag = Chain.getValue(1); | |||
7546 | VA = RVLocs[++i]; // skip ahead to next loc | |||
7547 | Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), SVal, Flag); | |||
7548 | } else | |||
7549 | Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), Arg, Flag); | |||
7550 | Flag = Chain.getValue(1); | |||
7551 | RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT())); | |||
7552 | } | |||
7553 | ||||
7554 | RetOps[0] = Chain; // Update chain. | |||
7555 | ||||
7556 | // Add the flag if we have it. | |||
7557 | if (Flag.getNode()) | |||
7558 | RetOps.push_back(Flag); | |||
7559 | ||||
7560 | return DAG.getNode(PPCISD::RET_FLAG, dl, MVT::Other, RetOps); | |||
7561 | } | |||
7562 | ||||
7563 | SDValue | |||
7564 | PPCTargetLowering::LowerGET_DYNAMIC_AREA_OFFSET(SDValue Op, | |||
7565 | SelectionDAG &DAG) const { | |||
7566 | SDLoc dl(Op); | |||
7567 | ||||
7568 | // Get the correct type for integers. | |||
7569 | EVT IntVT = Op.getValueType(); | |||
7570 | ||||
7571 | // Get the inputs. | |||
7572 | SDValue Chain = Op.getOperand(0); | |||
7573 | SDValue FPSIdx = getFramePointerFrameIndex(DAG); | |||
7574 | // Build a DYNAREAOFFSET node. | |||
7575 | SDValue Ops[2] = {Chain, FPSIdx}; | |||
7576 | SDVTList VTs = DAG.getVTList(IntVT); | |||
7577 | return DAG.getNode(PPCISD::DYNAREAOFFSET, dl, VTs, Ops); | |||
7578 | } | |||
7579 | ||||
7580 | SDValue PPCTargetLowering::LowerSTACKRESTORE(SDValue Op, | |||
7581 | SelectionDAG &DAG) const { | |||
7582 | // When we pop the dynamic allocation we need to restore the SP link. | |||
7583 | SDLoc dl(Op); | |||
7584 | ||||
7585 | // Get the correct type for pointers. | |||
7586 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
7587 | ||||
7588 | // Construct the stack pointer operand. | |||
7589 | bool isPPC64 = Subtarget.isPPC64(); | |||
7590 | unsigned SP = isPPC64 ? PPC::X1 : PPC::R1; | |||
7591 | SDValue StackPtr = DAG.getRegister(SP, PtrVT); | |||
7592 | ||||
7593 | // Get the operands for the STACKRESTORE. | |||
7594 | SDValue Chain = Op.getOperand(0); | |||
7595 | SDValue SaveSP = Op.getOperand(1); | |||
7596 | ||||
7597 | // Load the old link SP. | |||
7598 | SDValue LoadLinkSP = | |||
7599 | DAG.getLoad(PtrVT, dl, Chain, StackPtr, MachinePointerInfo()); | |||
7600 | ||||
7601 | // Restore the stack pointer. | |||
7602 | Chain = DAG.getCopyToReg(LoadLinkSP.getValue(1), dl, SP, SaveSP); | |||
7603 | ||||
7604 | // Store the old link SP. | |||
7605 | return DAG.getStore(Chain, dl, LoadLinkSP, StackPtr, MachinePointerInfo()); | |||
7606 | } | |||
7607 | ||||
7608 | SDValue PPCTargetLowering::getReturnAddrFrameIndex(SelectionDAG &DAG) const { | |||
7609 | MachineFunction &MF = DAG.getMachineFunction(); | |||
7610 | bool isPPC64 = Subtarget.isPPC64(); | |||
7611 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
7612 | ||||
7613 | // Get current frame pointer save index. The users of this index will be | |||
7614 | // primarily DYNALLOC instructions. | |||
7615 | PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); | |||
7616 | int RASI = FI->getReturnAddrSaveIndex(); | |||
7617 | ||||
7618 | // If the frame pointer save index hasn't been defined yet. | |||
7619 | if (!RASI) { | |||
7620 | // Find out what the fix offset of the frame pointer save area. | |||
7621 | int LROffset = Subtarget.getFrameLowering()->getReturnSaveOffset(); | |||
7622 | // Allocate the frame index for frame pointer save area. | |||
7623 | RASI = MF.getFrameInfo().CreateFixedObject(isPPC64? 8 : 4, LROffset, false); | |||
7624 | // Save the result. | |||
7625 | FI->setReturnAddrSaveIndex(RASI); | |||
7626 | } | |||
7627 | return DAG.getFrameIndex(RASI, PtrVT); | |||
7628 | } | |||
7629 | ||||
7630 | SDValue | |||
7631 | PPCTargetLowering::getFramePointerFrameIndex(SelectionDAG & DAG) const { | |||
7632 | MachineFunction &MF = DAG.getMachineFunction(); | |||
7633 | bool isPPC64 = Subtarget.isPPC64(); | |||
7634 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
7635 | ||||
7636 | // Get current frame pointer save index. The users of this index will be | |||
7637 | // primarily DYNALLOC instructions. | |||
7638 | PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); | |||
7639 | int FPSI = FI->getFramePointerSaveIndex(); | |||
7640 | ||||
7641 | // If the frame pointer save index hasn't been defined yet. | |||
7642 | if (!FPSI) { | |||
7643 | // Find out what the fix offset of the frame pointer save area. | |||
7644 | int FPOffset = Subtarget.getFrameLowering()->getFramePointerSaveOffset(); | |||
7645 | // Allocate the frame index for frame pointer save area. | |||
7646 | FPSI = MF.getFrameInfo().CreateFixedObject(isPPC64? 8 : 4, FPOffset, true); | |||
7647 | // Save the result. | |||
7648 | FI->setFramePointerSaveIndex(FPSI); | |||
7649 | } | |||
7650 | return DAG.getFrameIndex(FPSI, PtrVT); | |||
7651 | } | |||
7652 | ||||
7653 | SDValue PPCTargetLowering::LowerDYNAMIC_STACKALLOC(SDValue Op, | |||
7654 | SelectionDAG &DAG) const { | |||
7655 | MachineFunction &MF = DAG.getMachineFunction(); | |||
7656 | // Get the inputs. | |||
7657 | SDValue Chain = Op.getOperand(0); | |||
7658 | SDValue Size = Op.getOperand(1); | |||
7659 | SDLoc dl(Op); | |||
7660 | ||||
7661 | // Get the correct type for pointers. | |||
7662 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
7663 | // Negate the size. | |||
7664 | SDValue NegSize = DAG.getNode(ISD::SUB, dl, PtrVT, | |||
7665 | DAG.getConstant(0, dl, PtrVT), Size); | |||
7666 | // Construct a node for the frame pointer save index. | |||
7667 | SDValue FPSIdx = getFramePointerFrameIndex(DAG); | |||
7668 | SDValue Ops[3] = { Chain, NegSize, FPSIdx }; | |||
7669 | SDVTList VTs = DAG.getVTList(PtrVT, MVT::Other); | |||
7670 | if (hasInlineStackProbe(MF)) | |||
7671 | return DAG.getNode(PPCISD::PROBED_ALLOCA, dl, VTs, Ops); | |||
7672 | return DAG.getNode(PPCISD::DYNALLOC, dl, VTs, Ops); | |||
7673 | } | |||
7674 | ||||
7675 | SDValue PPCTargetLowering::LowerEH_DWARF_CFA(SDValue Op, | |||
7676 | SelectionDAG &DAG) const { | |||
7677 | MachineFunction &MF = DAG.getMachineFunction(); | |||
7678 | ||||
7679 | bool isPPC64 = Subtarget.isPPC64(); | |||
7680 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
7681 | ||||
7682 | int FI = MF.getFrameInfo().CreateFixedObject(isPPC64 ? 8 : 4, 0, false); | |||
7683 | return DAG.getFrameIndex(FI, PtrVT); | |||
7684 | } | |||
7685 | ||||
7686 | SDValue PPCTargetLowering::lowerEH_SJLJ_SETJMP(SDValue Op, | |||
7687 | SelectionDAG &DAG) const { | |||
7688 | SDLoc DL(Op); | |||
7689 | return DAG.getNode(PPCISD::EH_SJLJ_SETJMP, DL, | |||
7690 | DAG.getVTList(MVT::i32, MVT::Other), | |||
7691 | Op.getOperand(0), Op.getOperand(1)); | |||
7692 | } | |||
7693 | ||||
7694 | SDValue PPCTargetLowering::lowerEH_SJLJ_LONGJMP(SDValue Op, | |||
7695 | SelectionDAG &DAG) const { | |||
7696 | SDLoc DL(Op); | |||
7697 | return DAG.getNode(PPCISD::EH_SJLJ_LONGJMP, DL, MVT::Other, | |||
7698 | Op.getOperand(0), Op.getOperand(1)); | |||
7699 | } | |||
7700 | ||||
7701 | SDValue PPCTargetLowering::LowerLOAD(SDValue Op, SelectionDAG &DAG) const { | |||
7702 | if (Op.getValueType().isVector()) | |||
7703 | return LowerVectorLoad(Op, DAG); | |||
7704 | ||||
7705 | assert(Op.getValueType() == MVT::i1 &&(static_cast <bool> (Op.getValueType() == MVT::i1 && "Custom lowering only for i1 loads") ? void (0) : __assert_fail ("Op.getValueType() == MVT::i1 && \"Custom lowering only for i1 loads\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7706, __extension__ __PRETTY_FUNCTION__)) | |||
7706 | "Custom lowering only for i1 loads")(static_cast <bool> (Op.getValueType() == MVT::i1 && "Custom lowering only for i1 loads") ? void (0) : __assert_fail ("Op.getValueType() == MVT::i1 && \"Custom lowering only for i1 loads\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7706, __extension__ __PRETTY_FUNCTION__)); | |||
7707 | ||||
7708 | // First, load 8 bits into 32 bits, then truncate to 1 bit. | |||
7709 | ||||
7710 | SDLoc dl(Op); | |||
7711 | LoadSDNode *LD = cast<LoadSDNode>(Op); | |||
7712 | ||||
7713 | SDValue Chain = LD->getChain(); | |||
7714 | SDValue BasePtr = LD->getBasePtr(); | |||
7715 | MachineMemOperand *MMO = LD->getMemOperand(); | |||
7716 | ||||
7717 | SDValue NewLD = | |||
7718 | DAG.getExtLoad(ISD::EXTLOAD, dl, getPointerTy(DAG.getDataLayout()), Chain, | |||
7719 | BasePtr, MVT::i8, MMO); | |||
7720 | SDValue Result = DAG.getNode(ISD::TRUNCATE, dl, MVT::i1, NewLD); | |||
7721 | ||||
7722 | SDValue Ops[] = { Result, SDValue(NewLD.getNode(), 1) }; | |||
7723 | return DAG.getMergeValues(Ops, dl); | |||
7724 | } | |||
7725 | ||||
7726 | SDValue PPCTargetLowering::LowerSTORE(SDValue Op, SelectionDAG &DAG) const { | |||
7727 | if (Op.getOperand(1).getValueType().isVector()) | |||
7728 | return LowerVectorStore(Op, DAG); | |||
7729 | ||||
7730 | assert(Op.getOperand(1).getValueType() == MVT::i1 &&(static_cast <bool> (Op.getOperand(1).getValueType() == MVT::i1 && "Custom lowering only for i1 stores") ? void (0) : __assert_fail ("Op.getOperand(1).getValueType() == MVT::i1 && \"Custom lowering only for i1 stores\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7731, __extension__ __PRETTY_FUNCTION__)) | |||
7731 | "Custom lowering only for i1 stores")(static_cast <bool> (Op.getOperand(1).getValueType() == MVT::i1 && "Custom lowering only for i1 stores") ? void (0) : __assert_fail ("Op.getOperand(1).getValueType() == MVT::i1 && \"Custom lowering only for i1 stores\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7731, __extension__ __PRETTY_FUNCTION__)); | |||
7732 | ||||
7733 | // First, zero extend to 32 bits, then use a truncating store to 8 bits. | |||
7734 | ||||
7735 | SDLoc dl(Op); | |||
7736 | StoreSDNode *ST = cast<StoreSDNode>(Op); | |||
7737 | ||||
7738 | SDValue Chain = ST->getChain(); | |||
7739 | SDValue BasePtr = ST->getBasePtr(); | |||
7740 | SDValue Value = ST->getValue(); | |||
7741 | MachineMemOperand *MMO = ST->getMemOperand(); | |||
7742 | ||||
7743 | Value = DAG.getNode(ISD::ZERO_EXTEND, dl, getPointerTy(DAG.getDataLayout()), | |||
7744 | Value); | |||
7745 | return DAG.getTruncStore(Chain, dl, Value, BasePtr, MVT::i8, MMO); | |||
7746 | } | |||
7747 | ||||
7748 | // FIXME: Remove this once the ANDI glue bug is fixed: | |||
7749 | SDValue PPCTargetLowering::LowerTRUNCATE(SDValue Op, SelectionDAG &DAG) const { | |||
7750 | assert(Op.getValueType() == MVT::i1 &&(static_cast <bool> (Op.getValueType() == MVT::i1 && "Custom lowering only for i1 results") ? void (0) : __assert_fail ("Op.getValueType() == MVT::i1 && \"Custom lowering only for i1 results\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7751, __extension__ __PRETTY_FUNCTION__)) | |||
7751 | "Custom lowering only for i1 results")(static_cast <bool> (Op.getValueType() == MVT::i1 && "Custom lowering only for i1 results") ? void (0) : __assert_fail ("Op.getValueType() == MVT::i1 && \"Custom lowering only for i1 results\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7751, __extension__ __PRETTY_FUNCTION__)); | |||
7752 | ||||
7753 | SDLoc DL(Op); | |||
7754 | return DAG.getNode(PPCISD::ANDI_rec_1_GT_BIT, DL, MVT::i1, Op.getOperand(0)); | |||
7755 | } | |||
7756 | ||||
7757 | SDValue PPCTargetLowering::LowerTRUNCATEVector(SDValue Op, | |||
7758 | SelectionDAG &DAG) const { | |||
7759 | ||||
7760 | // Implements a vector truncate that fits in a vector register as a shuffle. | |||
7761 | // We want to legalize vector truncates down to where the source fits in | |||
7762 | // a vector register (and target is therefore smaller than vector register | |||
7763 | // size). At that point legalization will try to custom lower the sub-legal | |||
7764 | // result and get here - where we can contain the truncate as a single target | |||
7765 | // operation. | |||
7766 | ||||
7767 | // For example a trunc <2 x i16> to <2 x i8> could be visualized as follows: | |||
7768 | // <MSB1|LSB1, MSB2|LSB2> to <LSB1, LSB2> | |||
7769 | // | |||
7770 | // We will implement it for big-endian ordering as this (where x denotes | |||
7771 | // undefined): | |||
7772 | // < MSB1|LSB1, MSB2|LSB2, uu, uu, uu, uu, uu, uu> to | |||
7773 | // < LSB1, LSB2, u, u, u, u, u, u, u, u, u, u, u, u, u, u> | |||
7774 | // | |||
7775 | // The same operation in little-endian ordering will be: | |||
7776 | // <uu, uu, uu, uu, uu, uu, LSB2|MSB2, LSB1|MSB1> to | |||
7777 | // <u, u, u, u, u, u, u, u, u, u, u, u, u, u, LSB2, LSB1> | |||
7778 | ||||
7779 | EVT TrgVT = Op.getValueType(); | |||
7780 | assert(TrgVT.isVector() && "Vector type expected.")(static_cast <bool> (TrgVT.isVector() && "Vector type expected." ) ? void (0) : __assert_fail ("TrgVT.isVector() && \"Vector type expected.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7780, __extension__ __PRETTY_FUNCTION__)); | |||
7781 | unsigned TrgNumElts = TrgVT.getVectorNumElements(); | |||
7782 | EVT EltVT = TrgVT.getVectorElementType(); | |||
7783 | if (!isOperationCustom(Op.getOpcode(), TrgVT) || | |||
7784 | TrgVT.getSizeInBits() > 128 || !isPowerOf2_32(TrgNumElts) || | |||
7785 | !isPowerOf2_32(EltVT.getSizeInBits())) | |||
7786 | return SDValue(); | |||
7787 | ||||
7788 | SDValue N1 = Op.getOperand(0); | |||
7789 | EVT SrcVT = N1.getValueType(); | |||
7790 | unsigned SrcSize = SrcVT.getSizeInBits(); | |||
7791 | if (SrcSize > 256 || | |||
7792 | !isPowerOf2_32(SrcVT.getVectorNumElements()) || | |||
7793 | !isPowerOf2_32(SrcVT.getVectorElementType().getSizeInBits())) | |||
7794 | return SDValue(); | |||
7795 | if (SrcSize == 256 && SrcVT.getVectorNumElements() < 2) | |||
7796 | return SDValue(); | |||
7797 | ||||
7798 | unsigned WideNumElts = 128 / EltVT.getSizeInBits(); | |||
7799 | EVT WideVT = EVT::getVectorVT(*DAG.getContext(), EltVT, WideNumElts); | |||
7800 | ||||
7801 | SDLoc DL(Op); | |||
7802 | SDValue Op1, Op2; | |||
7803 | if (SrcSize == 256) { | |||
7804 | EVT VecIdxTy = getVectorIdxTy(DAG.getDataLayout()); | |||
7805 | EVT SplitVT = | |||
7806 | N1.getValueType().getHalfNumVectorElementsVT(*DAG.getContext()); | |||
7807 | unsigned SplitNumElts = SplitVT.getVectorNumElements(); | |||
7808 | Op1 = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, SplitVT, N1, | |||
7809 | DAG.getConstant(0, DL, VecIdxTy)); | |||
7810 | Op2 = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, SplitVT, N1, | |||
7811 | DAG.getConstant(SplitNumElts, DL, VecIdxTy)); | |||
7812 | } | |||
7813 | else { | |||
7814 | Op1 = SrcSize == 128 ? N1 : widenVec(DAG, N1, DL); | |||
7815 | Op2 = DAG.getUNDEF(WideVT); | |||
7816 | } | |||
7817 | ||||
7818 | // First list the elements we want to keep. | |||
7819 | unsigned SizeMult = SrcSize / TrgVT.getSizeInBits(); | |||
7820 | SmallVector<int, 16> ShuffV; | |||
7821 | if (Subtarget.isLittleEndian()) | |||
7822 | for (unsigned i = 0; i < TrgNumElts; ++i) | |||
7823 | ShuffV.push_back(i * SizeMult); | |||
7824 | else | |||
7825 | for (unsigned i = 1; i <= TrgNumElts; ++i) | |||
7826 | ShuffV.push_back(i * SizeMult - 1); | |||
7827 | ||||
7828 | // Populate the remaining elements with undefs. | |||
7829 | for (unsigned i = TrgNumElts; i < WideNumElts; ++i) | |||
7830 | // ShuffV.push_back(i + WideNumElts); | |||
7831 | ShuffV.push_back(WideNumElts + 1); | |||
7832 | ||||
7833 | Op1 = DAG.getNode(ISD::BITCAST, DL, WideVT, Op1); | |||
7834 | Op2 = DAG.getNode(ISD::BITCAST, DL, WideVT, Op2); | |||
7835 | return DAG.getVectorShuffle(WideVT, DL, Op1, Op2, ShuffV); | |||
7836 | } | |||
7837 | ||||
7838 | /// LowerSELECT_CC - Lower floating point select_cc's into fsel instruction when | |||
7839 | /// possible. | |||
7840 | SDValue PPCTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const { | |||
7841 | ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(4))->get(); | |||
7842 | EVT ResVT = Op.getValueType(); | |||
7843 | EVT CmpVT = Op.getOperand(0).getValueType(); | |||
7844 | SDValue LHS = Op.getOperand(0), RHS = Op.getOperand(1); | |||
7845 | SDValue TV = Op.getOperand(2), FV = Op.getOperand(3); | |||
7846 | SDLoc dl(Op); | |||
7847 | ||||
7848 | // Without power9-vector, we don't have native instruction for f128 comparison. | |||
7849 | // Following transformation to libcall is needed for setcc: | |||
7850 | // select_cc lhs, rhs, tv, fv, cc -> select_cc (setcc cc, x, y), 0, tv, fv, NE | |||
7851 | if (!Subtarget.hasP9Vector() && CmpVT == MVT::f128) { | |||
7852 | SDValue Z = DAG.getSetCC( | |||
7853 | dl, getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), CmpVT), | |||
7854 | LHS, RHS, CC); | |||
7855 | SDValue Zero = DAG.getConstant(0, dl, Z.getValueType()); | |||
7856 | return DAG.getSelectCC(dl, Z, Zero, TV, FV, ISD::SETNE); | |||
7857 | } | |||
7858 | ||||
7859 | // Not FP, or using SPE? Not a fsel. | |||
7860 | if (!CmpVT.isFloatingPoint() || !TV.getValueType().isFloatingPoint() || | |||
7861 | Subtarget.hasSPE()) | |||
7862 | return Op; | |||
7863 | ||||
7864 | SDNodeFlags Flags = Op.getNode()->getFlags(); | |||
7865 | ||||
7866 | // We have xsmaxcdp/xsmincdp which are OK to emit even in the | |||
7867 | // presence of infinities. | |||
7868 | if (Subtarget.hasP9Vector() && LHS == TV && RHS == FV) { | |||
7869 | switch (CC) { | |||
7870 | default: | |||
7871 | break; | |||
7872 | case ISD::SETOGT: | |||
7873 | case ISD::SETGT: | |||
7874 | return DAG.getNode(PPCISD::XSMAXCDP, dl, Op.getValueType(), LHS, RHS); | |||
7875 | case ISD::SETOLT: | |||
7876 | case ISD::SETLT: | |||
7877 | return DAG.getNode(PPCISD::XSMINCDP, dl, Op.getValueType(), LHS, RHS); | |||
7878 | } | |||
7879 | } | |||
7880 | ||||
7881 | // We might be able to do better than this under some circumstances, but in | |||
7882 | // general, fsel-based lowering of select is a finite-math-only optimization. | |||
7883 | // For more information, see section F.3 of the 2.06 ISA specification. | |||
7884 | // With ISA 3.0 | |||
7885 | if ((!DAG.getTarget().Options.NoInfsFPMath && !Flags.hasNoInfs()) || | |||
7886 | (!DAG.getTarget().Options.NoNaNsFPMath && !Flags.hasNoNaNs())) | |||
7887 | return Op; | |||
7888 | ||||
7889 | // If the RHS of the comparison is a 0.0, we don't need to do the | |||
7890 | // subtraction at all. | |||
7891 | SDValue Sel1; | |||
7892 | if (isFloatingPointZero(RHS)) | |||
7893 | switch (CC) { | |||
7894 | default: break; // SETUO etc aren't handled by fsel. | |||
7895 | case ISD::SETNE: | |||
7896 | std::swap(TV, FV); | |||
7897 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
7898 | case ISD::SETEQ: | |||
7899 | if (LHS.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7900 | LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, LHS); | |||
7901 | Sel1 = DAG.getNode(PPCISD::FSEL, dl, ResVT, LHS, TV, FV); | |||
7902 | if (Sel1.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7903 | Sel1 = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Sel1); | |||
7904 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, | |||
7905 | DAG.getNode(ISD::FNEG, dl, MVT::f64, LHS), Sel1, FV); | |||
7906 | case ISD::SETULT: | |||
7907 | case ISD::SETLT: | |||
7908 | std::swap(TV, FV); // fsel is natively setge, swap operands for setlt | |||
7909 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
7910 | case ISD::SETOGE: | |||
7911 | case ISD::SETGE: | |||
7912 | if (LHS.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7913 | LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, LHS); | |||
7914 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, LHS, TV, FV); | |||
7915 | case ISD::SETUGT: | |||
7916 | case ISD::SETGT: | |||
7917 | std::swap(TV, FV); // fsel is natively setge, swap operands for setlt | |||
7918 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
7919 | case ISD::SETOLE: | |||
7920 | case ISD::SETLE: | |||
7921 | if (LHS.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7922 | LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, LHS); | |||
7923 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, | |||
7924 | DAG.getNode(ISD::FNEG, dl, MVT::f64, LHS), TV, FV); | |||
7925 | } | |||
7926 | ||||
7927 | SDValue Cmp; | |||
7928 | switch (CC) { | |||
7929 | default: break; // SETUO etc aren't handled by fsel. | |||
7930 | case ISD::SETNE: | |||
7931 | std::swap(TV, FV); | |||
7932 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
7933 | case ISD::SETEQ: | |||
7934 | Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, LHS, RHS, Flags); | |||
7935 | if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7936 | Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); | |||
7937 | Sel1 = DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); | |||
7938 | if (Sel1.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7939 | Sel1 = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Sel1); | |||
7940 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, | |||
7941 | DAG.getNode(ISD::FNEG, dl, MVT::f64, Cmp), Sel1, FV); | |||
7942 | case ISD::SETULT: | |||
7943 | case ISD::SETLT: | |||
7944 | Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, LHS, RHS, Flags); | |||
7945 | if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7946 | Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); | |||
7947 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, FV, TV); | |||
7948 | case ISD::SETOGE: | |||
7949 | case ISD::SETGE: | |||
7950 | Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, LHS, RHS, Flags); | |||
7951 | if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7952 | Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); | |||
7953 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); | |||
7954 | case ISD::SETUGT: | |||
7955 | case ISD::SETGT: | |||
7956 | Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, RHS, LHS, Flags); | |||
7957 | if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7958 | Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); | |||
7959 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, FV, TV); | |||
7960 | case ISD::SETOLE: | |||
7961 | case ISD::SETLE: | |||
7962 | Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, RHS, LHS, Flags); | |||
7963 | if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits | |||
7964 | Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); | |||
7965 | return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); | |||
7966 | } | |||
7967 | return Op; | |||
7968 | } | |||
7969 | ||||
7970 | static unsigned getPPCStrictOpcode(unsigned Opc) { | |||
7971 | switch (Opc) { | |||
7972 | default: | |||
7973 | llvm_unreachable("No strict version of this opcode!")::llvm::llvm_unreachable_internal("No strict version of this opcode!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 7973); | |||
7974 | case PPCISD::FCTIDZ: | |||
7975 | return PPCISD::STRICT_FCTIDZ; | |||
7976 | case PPCISD::FCTIWZ: | |||
7977 | return PPCISD::STRICT_FCTIWZ; | |||
7978 | case PPCISD::FCTIDUZ: | |||
7979 | return PPCISD::STRICT_FCTIDUZ; | |||
7980 | case PPCISD::FCTIWUZ: | |||
7981 | return PPCISD::STRICT_FCTIWUZ; | |||
7982 | case PPCISD::FCFID: | |||
7983 | return PPCISD::STRICT_FCFID; | |||
7984 | case PPCISD::FCFIDU: | |||
7985 | return PPCISD::STRICT_FCFIDU; | |||
7986 | case PPCISD::FCFIDS: | |||
7987 | return PPCISD::STRICT_FCFIDS; | |||
7988 | case PPCISD::FCFIDUS: | |||
7989 | return PPCISD::STRICT_FCFIDUS; | |||
7990 | } | |||
7991 | } | |||
7992 | ||||
7993 | static SDValue convertFPToInt(SDValue Op, SelectionDAG &DAG, | |||
7994 | const PPCSubtarget &Subtarget) { | |||
7995 | SDLoc dl(Op); | |||
7996 | bool IsStrict = Op->isStrictFPOpcode(); | |||
7997 | bool IsSigned = Op.getOpcode() == ISD::FP_TO_SINT || | |||
7998 | Op.getOpcode() == ISD::STRICT_FP_TO_SINT; | |||
7999 | ||||
8000 | // TODO: Any other flags to propagate? | |||
8001 | SDNodeFlags Flags; | |||
8002 | Flags.setNoFPExcept(Op->getFlags().hasNoFPExcept()); | |||
8003 | ||||
8004 | // For strict nodes, source is the second operand. | |||
8005 | SDValue Src = Op.getOperand(IsStrict ? 1 : 0); | |||
8006 | SDValue Chain = IsStrict ? Op.getOperand(0) : SDValue(); | |||
8007 | assert(Src.getValueType().isFloatingPoint())(static_cast <bool> (Src.getValueType().isFloatingPoint ()) ? void (0) : __assert_fail ("Src.getValueType().isFloatingPoint()" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8007, __extension__ __PRETTY_FUNCTION__)); | |||
8008 | if (Src.getValueType() == MVT::f32) { | |||
8009 | if (IsStrict) { | |||
8010 | Src = | |||
8011 | DAG.getNode(ISD::STRICT_FP_EXTEND, dl, | |||
8012 | DAG.getVTList(MVT::f64, MVT::Other), {Chain, Src}, Flags); | |||
8013 | Chain = Src.getValue(1); | |||
8014 | } else | |||
8015 | Src = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Src); | |||
8016 | } | |||
8017 | SDValue Conv; | |||
8018 | unsigned Opc = ISD::DELETED_NODE; | |||
8019 | switch (Op.getSimpleValueType().SimpleTy) { | |||
8020 | default: llvm_unreachable("Unhandled FP_TO_INT type in custom expander!")::llvm::llvm_unreachable_internal("Unhandled FP_TO_INT type in custom expander!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8020); | |||
8021 | case MVT::i32: | |||
8022 | Opc = IsSigned ? PPCISD::FCTIWZ | |||
8023 | : (Subtarget.hasFPCVT() ? PPCISD::FCTIWUZ : PPCISD::FCTIDZ); | |||
8024 | break; | |||
8025 | case MVT::i64: | |||
8026 | assert((IsSigned || Subtarget.hasFPCVT()) &&(static_cast <bool> ((IsSigned || Subtarget.hasFPCVT()) && "i64 FP_TO_UINT is supported only with FPCVT") ? void (0) : __assert_fail ("(IsSigned || Subtarget.hasFPCVT()) && \"i64 FP_TO_UINT is supported only with FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8027, __extension__ __PRETTY_FUNCTION__)) | |||
8027 | "i64 FP_TO_UINT is supported only with FPCVT")(static_cast <bool> ((IsSigned || Subtarget.hasFPCVT()) && "i64 FP_TO_UINT is supported only with FPCVT") ? void (0) : __assert_fail ("(IsSigned || Subtarget.hasFPCVT()) && \"i64 FP_TO_UINT is supported only with FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8027, __extension__ __PRETTY_FUNCTION__)); | |||
8028 | Opc = IsSigned ? PPCISD::FCTIDZ : PPCISD::FCTIDUZ; | |||
8029 | } | |||
8030 | if (IsStrict) { | |||
8031 | Opc = getPPCStrictOpcode(Opc); | |||
8032 | Conv = DAG.getNode(Opc, dl, DAG.getVTList(MVT::f64, MVT::Other), | |||
8033 | {Chain, Src}, Flags); | |||
8034 | } else { | |||
8035 | Conv = DAG.getNode(Opc, dl, MVT::f64, Src); | |||
8036 | } | |||
8037 | return Conv; | |||
8038 | } | |||
8039 | ||||
8040 | void PPCTargetLowering::LowerFP_TO_INTForReuse(SDValue Op, ReuseLoadInfo &RLI, | |||
8041 | SelectionDAG &DAG, | |||
8042 | const SDLoc &dl) const { | |||
8043 | SDValue Tmp = convertFPToInt(Op, DAG, Subtarget); | |||
8044 | bool IsSigned = Op.getOpcode() == ISD::FP_TO_SINT || | |||
8045 | Op.getOpcode() == ISD::STRICT_FP_TO_SINT; | |||
8046 | bool IsStrict = Op->isStrictFPOpcode(); | |||
8047 | ||||
8048 | // Convert the FP value to an int value through memory. | |||
8049 | bool i32Stack = Op.getValueType() == MVT::i32 && Subtarget.hasSTFIWX() && | |||
8050 | (IsSigned || Subtarget.hasFPCVT()); | |||
8051 | SDValue FIPtr = DAG.CreateStackTemporary(i32Stack ? MVT::i32 : MVT::f64); | |||
8052 | int FI = cast<FrameIndexSDNode>(FIPtr)->getIndex(); | |||
8053 | MachinePointerInfo MPI = | |||
8054 | MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI); | |||
8055 | ||||
8056 | // Emit a store to the stack slot. | |||
8057 | SDValue Chain = IsStrict ? Tmp.getValue(1) : DAG.getEntryNode(); | |||
8058 | Align Alignment(DAG.getEVTAlign(Tmp.getValueType())); | |||
8059 | if (i32Stack) { | |||
8060 | MachineFunction &MF = DAG.getMachineFunction(); | |||
8061 | Alignment = Align(4); | |||
8062 | MachineMemOperand *MMO = | |||
8063 | MF.getMachineMemOperand(MPI, MachineMemOperand::MOStore, 4, Alignment); | |||
8064 | SDValue Ops[] = { Chain, Tmp, FIPtr }; | |||
8065 | Chain = DAG.getMemIntrinsicNode(PPCISD::STFIWX, dl, | |||
8066 | DAG.getVTList(MVT::Other), Ops, MVT::i32, MMO); | |||
8067 | } else | |||
8068 | Chain = DAG.getStore(Chain, dl, Tmp, FIPtr, MPI, Alignment); | |||
8069 | ||||
8070 | // Result is a load from the stack slot. If loading 4 bytes, make sure to | |||
8071 | // add in a bias on big endian. | |||
8072 | if (Op.getValueType() == MVT::i32 && !i32Stack) { | |||
8073 | FIPtr = DAG.getNode(ISD::ADD, dl, FIPtr.getValueType(), FIPtr, | |||
8074 | DAG.getConstant(4, dl, FIPtr.getValueType())); | |||
8075 | MPI = MPI.getWithOffset(Subtarget.isLittleEndian() ? 0 : 4); | |||
8076 | } | |||
8077 | ||||
8078 | RLI.Chain = Chain; | |||
8079 | RLI.Ptr = FIPtr; | |||
8080 | RLI.MPI = MPI; | |||
8081 | RLI.Alignment = Alignment; | |||
8082 | } | |||
8083 | ||||
8084 | /// Custom lowers floating point to integer conversions to use | |||
8085 | /// the direct move instructions available in ISA 2.07 to avoid the | |||
8086 | /// need for load/store combinations. | |||
8087 | SDValue PPCTargetLowering::LowerFP_TO_INTDirectMove(SDValue Op, | |||
8088 | SelectionDAG &DAG, | |||
8089 | const SDLoc &dl) const { | |||
8090 | SDValue Conv = convertFPToInt(Op, DAG, Subtarget); | |||
8091 | SDValue Mov = DAG.getNode(PPCISD::MFVSR, dl, Op.getValueType(), Conv); | |||
8092 | if (Op->isStrictFPOpcode()) | |||
8093 | return DAG.getMergeValues({Mov, Conv.getValue(1)}, dl); | |||
8094 | else | |||
8095 | return Mov; | |||
8096 | } | |||
8097 | ||||
8098 | SDValue PPCTargetLowering::LowerFP_TO_INT(SDValue Op, SelectionDAG &DAG, | |||
8099 | const SDLoc &dl) const { | |||
8100 | bool IsStrict = Op->isStrictFPOpcode(); | |||
8101 | bool IsSigned = Op.getOpcode() == ISD::FP_TO_SINT || | |||
8102 | Op.getOpcode() == ISD::STRICT_FP_TO_SINT; | |||
8103 | SDValue Src = Op.getOperand(IsStrict ? 1 : 0); | |||
8104 | EVT SrcVT = Src.getValueType(); | |||
8105 | EVT DstVT = Op.getValueType(); | |||
8106 | ||||
8107 | // FP to INT conversions are legal for f128. | |||
8108 | if (SrcVT == MVT::f128) | |||
8109 | return Subtarget.hasP9Vector() ? Op : SDValue(); | |||
8110 | ||||
8111 | // Expand ppcf128 to i32 by hand for the benefit of llvm-gcc bootstrap on | |||
8112 | // PPC (the libcall is not available). | |||
8113 | if (SrcVT == MVT::ppcf128) { | |||
8114 | if (DstVT == MVT::i32) { | |||
8115 | // TODO: Conservatively pass only nofpexcept flag here. Need to check and | |||
8116 | // set other fast-math flags to FP operations in both strict and | |||
8117 | // non-strict cases. (FP_TO_SINT, FSUB) | |||
8118 | SDNodeFlags Flags; | |||
8119 | Flags.setNoFPExcept(Op->getFlags().hasNoFPExcept()); | |||
8120 | ||||
8121 | if (IsSigned) { | |||
8122 | SDValue Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::f64, Src, | |||
8123 | DAG.getIntPtrConstant(0, dl)); | |||
8124 | SDValue Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::f64, Src, | |||
8125 | DAG.getIntPtrConstant(1, dl)); | |||
8126 | ||||
8127 | // Add the two halves of the long double in round-to-zero mode, and use | |||
8128 | // a smaller FP_TO_SINT. | |||
8129 | if (IsStrict) { | |||
8130 | SDValue Res = DAG.getNode(PPCISD::STRICT_FADDRTZ, dl, | |||
8131 | DAG.getVTList(MVT::f64, MVT::Other), | |||
8132 | {Op.getOperand(0), Lo, Hi}, Flags); | |||
8133 | return DAG.getNode(ISD::STRICT_FP_TO_SINT, dl, | |||
8134 | DAG.getVTList(MVT::i32, MVT::Other), | |||
8135 | {Res.getValue(1), Res}, Flags); | |||
8136 | } else { | |||
8137 | SDValue Res = DAG.getNode(PPCISD::FADDRTZ, dl, MVT::f64, Lo, Hi); | |||
8138 | return DAG.getNode(ISD::FP_TO_SINT, dl, MVT::i32, Res); | |||
8139 | } | |||
8140 | } else { | |||
8141 | const uint64_t TwoE31[] = {0x41e0000000000000LL, 0}; | |||
8142 | APFloat APF = APFloat(APFloat::PPCDoubleDouble(), APInt(128, TwoE31)); | |||
8143 | SDValue Cst = DAG.getConstantFP(APF, dl, SrcVT); | |||
8144 | SDValue SignMask = DAG.getConstant(0x80000000, dl, DstVT); | |||
8145 | if (IsStrict) { | |||
8146 | // Sel = Src < 0x80000000 | |||
8147 | // FltOfs = select Sel, 0.0, 0x80000000 | |||
8148 | // IntOfs = select Sel, 0, 0x80000000 | |||
8149 | // Result = fp_to_sint(Src - FltOfs) ^ IntOfs | |||
8150 | SDValue Chain = Op.getOperand(0); | |||
8151 | EVT SetCCVT = | |||
8152 | getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), SrcVT); | |||
8153 | EVT DstSetCCVT = | |||
8154 | getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), DstVT); | |||
8155 | SDValue Sel = DAG.getSetCC(dl, SetCCVT, Src, Cst, ISD::SETLT, | |||
8156 | Chain, true); | |||
8157 | Chain = Sel.getValue(1); | |||
8158 | ||||
8159 | SDValue FltOfs = DAG.getSelect( | |||
8160 | dl, SrcVT, Sel, DAG.getConstantFP(0.0, dl, SrcVT), Cst); | |||
8161 | Sel = DAG.getBoolExtOrTrunc(Sel, dl, DstSetCCVT, DstVT); | |||
8162 | ||||
8163 | SDValue Val = DAG.getNode(ISD::STRICT_FSUB, dl, | |||
8164 | DAG.getVTList(SrcVT, MVT::Other), | |||
8165 | {Chain, Src, FltOfs}, Flags); | |||
8166 | Chain = Val.getValue(1); | |||
8167 | SDValue SInt = DAG.getNode(ISD::STRICT_FP_TO_SINT, dl, | |||
8168 | DAG.getVTList(DstVT, MVT::Other), | |||
8169 | {Chain, Val}, Flags); | |||
8170 | Chain = SInt.getValue(1); | |||
8171 | SDValue IntOfs = DAG.getSelect( | |||
8172 | dl, DstVT, Sel, DAG.getConstant(0, dl, DstVT), SignMask); | |||
8173 | SDValue Result = DAG.getNode(ISD::XOR, dl, DstVT, SInt, IntOfs); | |||
8174 | return DAG.getMergeValues({Result, Chain}, dl); | |||
8175 | } else { | |||
8176 | // X>=2^31 ? (int)(X-2^31)+0x80000000 : (int)X | |||
8177 | // FIXME: generated code sucks. | |||
8178 | SDValue True = DAG.getNode(ISD::FSUB, dl, MVT::ppcf128, Src, Cst); | |||
8179 | True = DAG.getNode(ISD::FP_TO_SINT, dl, MVT::i32, True); | |||
8180 | True = DAG.getNode(ISD::ADD, dl, MVT::i32, True, SignMask); | |||
8181 | SDValue False = DAG.getNode(ISD::FP_TO_SINT, dl, MVT::i32, Src); | |||
8182 | return DAG.getSelectCC(dl, Src, Cst, True, False, ISD::SETGE); | |||
8183 | } | |||
8184 | } | |||
8185 | } | |||
8186 | ||||
8187 | return SDValue(); | |||
8188 | } | |||
8189 | ||||
8190 | if (Subtarget.hasDirectMove() && Subtarget.isPPC64()) | |||
8191 | return LowerFP_TO_INTDirectMove(Op, DAG, dl); | |||
8192 | ||||
8193 | ReuseLoadInfo RLI; | |||
8194 | LowerFP_TO_INTForReuse(Op, RLI, DAG, dl); | |||
8195 | ||||
8196 | return DAG.getLoad(Op.getValueType(), dl, RLI.Chain, RLI.Ptr, RLI.MPI, | |||
8197 | RLI.Alignment, RLI.MMOFlags(), RLI.AAInfo, RLI.Ranges); | |||
8198 | } | |||
8199 | ||||
8200 | // We're trying to insert a regular store, S, and then a load, L. If the | |||
8201 | // incoming value, O, is a load, we might just be able to have our load use the | |||
8202 | // address used by O. However, we don't know if anything else will store to | |||
8203 | // that address before we can load from it. To prevent this situation, we need | |||
8204 | // to insert our load, L, into the chain as a peer of O. To do this, we give L | |||
8205 | // the same chain operand as O, we create a token factor from the chain results | |||
8206 | // of O and L, and we replace all uses of O's chain result with that token | |||
8207 | // factor (see spliceIntoChain below for this last part). | |||
8208 | bool PPCTargetLowering::canReuseLoadAddress(SDValue Op, EVT MemVT, | |||
8209 | ReuseLoadInfo &RLI, | |||
8210 | SelectionDAG &DAG, | |||
8211 | ISD::LoadExtType ET) const { | |||
8212 | // Conservatively skip reusing for constrained FP nodes. | |||
8213 | if (Op->isStrictFPOpcode()) | |||
8214 | return false; | |||
8215 | ||||
8216 | SDLoc dl(Op); | |||
8217 | bool ValidFPToUint = Op.getOpcode() == ISD::FP_TO_UINT && | |||
8218 | (Subtarget.hasFPCVT() || Op.getValueType() == MVT::i32); | |||
8219 | if (ET == ISD::NON_EXTLOAD && | |||
8220 | (ValidFPToUint || Op.getOpcode() == ISD::FP_TO_SINT) && | |||
8221 | isOperationLegalOrCustom(Op.getOpcode(), | |||
8222 | Op.getOperand(0).getValueType())) { | |||
8223 | ||||
8224 | LowerFP_TO_INTForReuse(Op, RLI, DAG, dl); | |||
8225 | return true; | |||
8226 | } | |||
8227 | ||||
8228 | LoadSDNode *LD = dyn_cast<LoadSDNode>(Op); | |||
8229 | if (!LD || LD->getExtensionType() != ET || LD->isVolatile() || | |||
8230 | LD->isNonTemporal()) | |||
8231 | return false; | |||
8232 | if (LD->getMemoryVT() != MemVT) | |||
8233 | return false; | |||
8234 | ||||
8235 | // If the result of the load is an illegal type, then we can't build a | |||
8236 | // valid chain for reuse since the legalised loads and token factor node that | |||
8237 | // ties the legalised loads together uses a different output chain then the | |||
8238 | // illegal load. | |||
8239 | if (!isTypeLegal(LD->getValueType(0))) | |||
8240 | return false; | |||
8241 | ||||
8242 | RLI.Ptr = LD->getBasePtr(); | |||
8243 | if (LD->isIndexed() && !LD->getOffset().isUndef()) { | |||
8244 | assert(LD->getAddressingMode() == ISD::PRE_INC &&(static_cast <bool> (LD->getAddressingMode() == ISD:: PRE_INC && "Non-pre-inc AM on PPC?") ? void (0) : __assert_fail ("LD->getAddressingMode() == ISD::PRE_INC && \"Non-pre-inc AM on PPC?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8245, __extension__ __PRETTY_FUNCTION__)) | |||
8245 | "Non-pre-inc AM on PPC?")(static_cast <bool> (LD->getAddressingMode() == ISD:: PRE_INC && "Non-pre-inc AM on PPC?") ? void (0) : __assert_fail ("LD->getAddressingMode() == ISD::PRE_INC && \"Non-pre-inc AM on PPC?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8245, __extension__ __PRETTY_FUNCTION__)); | |||
8246 | RLI.Ptr = DAG.getNode(ISD::ADD, dl, RLI.Ptr.getValueType(), RLI.Ptr, | |||
8247 | LD->getOffset()); | |||
8248 | } | |||
8249 | ||||
8250 | RLI.Chain = LD->getChain(); | |||
8251 | RLI.MPI = LD->getPointerInfo(); | |||
8252 | RLI.IsDereferenceable = LD->isDereferenceable(); | |||
8253 | RLI.IsInvariant = LD->isInvariant(); | |||
8254 | RLI.Alignment = LD->getAlign(); | |||
8255 | RLI.AAInfo = LD->getAAInfo(); | |||
8256 | RLI.Ranges = LD->getRanges(); | |||
8257 | ||||
8258 | RLI.ResChain = SDValue(LD, LD->isIndexed() ? 2 : 1); | |||
8259 | return true; | |||
8260 | } | |||
8261 | ||||
8262 | // Given the head of the old chain, ResChain, insert a token factor containing | |||
8263 | // it and NewResChain, and make users of ResChain now be users of that token | |||
8264 | // factor. | |||
8265 | // TODO: Remove and use DAG::makeEquivalentMemoryOrdering() instead. | |||
8266 | void PPCTargetLowering::spliceIntoChain(SDValue ResChain, | |||
8267 | SDValue NewResChain, | |||
8268 | SelectionDAG &DAG) const { | |||
8269 | if (!ResChain) | |||
8270 | return; | |||
8271 | ||||
8272 | SDLoc dl(NewResChain); | |||
8273 | ||||
8274 | SDValue TF = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, | |||
8275 | NewResChain, DAG.getUNDEF(MVT::Other)); | |||
8276 | assert(TF.getNode() != NewResChain.getNode() &&(static_cast <bool> (TF.getNode() != NewResChain.getNode () && "A new TF really is required here") ? void (0) : __assert_fail ("TF.getNode() != NewResChain.getNode() && \"A new TF really is required here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8277, __extension__ __PRETTY_FUNCTION__)) | |||
8277 | "A new TF really is required here")(static_cast <bool> (TF.getNode() != NewResChain.getNode () && "A new TF really is required here") ? void (0) : __assert_fail ("TF.getNode() != NewResChain.getNode() && \"A new TF really is required here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8277, __extension__ __PRETTY_FUNCTION__)); | |||
8278 | ||||
8279 | DAG.ReplaceAllUsesOfValueWith(ResChain, TF); | |||
8280 | DAG.UpdateNodeOperands(TF.getNode(), ResChain, NewResChain); | |||
8281 | } | |||
8282 | ||||
8283 | /// Analyze profitability of direct move | |||
8284 | /// prefer float load to int load plus direct move | |||
8285 | /// when there is no integer use of int load | |||
8286 | bool PPCTargetLowering::directMoveIsProfitable(const SDValue &Op) const { | |||
8287 | SDNode *Origin = Op.getOperand(0).getNode(); | |||
8288 | if (Origin->getOpcode() != ISD::LOAD) | |||
8289 | return true; | |||
8290 | ||||
8291 | // If there is no LXSIBZX/LXSIHZX, like Power8, | |||
8292 | // prefer direct move if the memory size is 1 or 2 bytes. | |||
8293 | MachineMemOperand *MMO = cast<LoadSDNode>(Origin)->getMemOperand(); | |||
8294 | if (!Subtarget.hasP9Vector() && MMO->getSize() <= 2) | |||
8295 | return true; | |||
8296 | ||||
8297 | for (SDNode::use_iterator UI = Origin->use_begin(), | |||
8298 | UE = Origin->use_end(); | |||
8299 | UI != UE; ++UI) { | |||
8300 | ||||
8301 | // Only look at the users of the loaded value. | |||
8302 | if (UI.getUse().get().getResNo() != 0) | |||
8303 | continue; | |||
8304 | ||||
8305 | if (UI->getOpcode() != ISD::SINT_TO_FP && | |||
8306 | UI->getOpcode() != ISD::UINT_TO_FP && | |||
8307 | UI->getOpcode() != ISD::STRICT_SINT_TO_FP && | |||
8308 | UI->getOpcode() != ISD::STRICT_UINT_TO_FP) | |||
8309 | return true; | |||
8310 | } | |||
8311 | ||||
8312 | return false; | |||
8313 | } | |||
8314 | ||||
8315 | static SDValue convertIntToFP(SDValue Op, SDValue Src, SelectionDAG &DAG, | |||
8316 | const PPCSubtarget &Subtarget, | |||
8317 | SDValue Chain = SDValue()) { | |||
8318 | bool IsSigned = Op.getOpcode() == ISD::SINT_TO_FP || | |||
8319 | Op.getOpcode() == ISD::STRICT_SINT_TO_FP; | |||
8320 | SDLoc dl(Op); | |||
8321 | ||||
8322 | // TODO: Any other flags to propagate? | |||
8323 | SDNodeFlags Flags; | |||
8324 | Flags.setNoFPExcept(Op->getFlags().hasNoFPExcept()); | |||
8325 | ||||
8326 | // If we have FCFIDS, then use it when converting to single-precision. | |||
8327 | // Otherwise, convert to double-precision and then round. | |||
8328 | bool IsSingle = Op.getValueType() == MVT::f32 && Subtarget.hasFPCVT(); | |||
8329 | unsigned ConvOpc = IsSingle ? (IsSigned ? PPCISD::FCFIDS : PPCISD::FCFIDUS) | |||
8330 | : (IsSigned ? PPCISD::FCFID : PPCISD::FCFIDU); | |||
8331 | EVT ConvTy = IsSingle ? MVT::f32 : MVT::f64; | |||
8332 | if (Op->isStrictFPOpcode()) { | |||
8333 | if (!Chain) | |||
8334 | Chain = Op.getOperand(0); | |||
8335 | return DAG.getNode(getPPCStrictOpcode(ConvOpc), dl, | |||
8336 | DAG.getVTList(ConvTy, MVT::Other), {Chain, Src}, Flags); | |||
8337 | } else | |||
8338 | return DAG.getNode(ConvOpc, dl, ConvTy, Src); | |||
8339 | } | |||
8340 | ||||
8341 | /// Custom lowers integer to floating point conversions to use | |||
8342 | /// the direct move instructions available in ISA 2.07 to avoid the | |||
8343 | /// need for load/store combinations. | |||
8344 | SDValue PPCTargetLowering::LowerINT_TO_FPDirectMove(SDValue Op, | |||
8345 | SelectionDAG &DAG, | |||
8346 | const SDLoc &dl) const { | |||
8347 | assert((Op.getValueType() == MVT::f32 ||(static_cast <bool> ((Op.getValueType() == MVT::f32 || Op .getValueType() == MVT::f64) && "Invalid floating point type as target of conversion" ) ? void (0) : __assert_fail ("(Op.getValueType() == MVT::f32 || Op.getValueType() == MVT::f64) && \"Invalid floating point type as target of conversion\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8349, __extension__ __PRETTY_FUNCTION__)) | |||
8348 | Op.getValueType() == MVT::f64) &&(static_cast <bool> ((Op.getValueType() == MVT::f32 || Op .getValueType() == MVT::f64) && "Invalid floating point type as target of conversion" ) ? void (0) : __assert_fail ("(Op.getValueType() == MVT::f32 || Op.getValueType() == MVT::f64) && \"Invalid floating point type as target of conversion\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8349, __extension__ __PRETTY_FUNCTION__)) | |||
8349 | "Invalid floating point type as target of conversion")(static_cast <bool> ((Op.getValueType() == MVT::f32 || Op .getValueType() == MVT::f64) && "Invalid floating point type as target of conversion" ) ? void (0) : __assert_fail ("(Op.getValueType() == MVT::f32 || Op.getValueType() == MVT::f64) && \"Invalid floating point type as target of conversion\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8349, __extension__ __PRETTY_FUNCTION__)); | |||
8350 | assert(Subtarget.hasFPCVT() &&(static_cast <bool> (Subtarget.hasFPCVT() && "Int to FP conversions with direct moves require FPCVT" ) ? void (0) : __assert_fail ("Subtarget.hasFPCVT() && \"Int to FP conversions with direct moves require FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8351, __extension__ __PRETTY_FUNCTION__)) | |||
8351 | "Int to FP conversions with direct moves require FPCVT")(static_cast <bool> (Subtarget.hasFPCVT() && "Int to FP conversions with direct moves require FPCVT" ) ? void (0) : __assert_fail ("Subtarget.hasFPCVT() && \"Int to FP conversions with direct moves require FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8351, __extension__ __PRETTY_FUNCTION__)); | |||
8352 | SDValue Src = Op.getOperand(Op->isStrictFPOpcode() ? 1 : 0); | |||
8353 | bool WordInt = Src.getSimpleValueType().SimpleTy == MVT::i32; | |||
8354 | bool Signed = Op.getOpcode() == ISD::SINT_TO_FP || | |||
8355 | Op.getOpcode() == ISD::STRICT_SINT_TO_FP; | |||
8356 | unsigned MovOpc = (WordInt && !Signed) ? PPCISD::MTVSRZ : PPCISD::MTVSRA; | |||
8357 | SDValue Mov = DAG.getNode(MovOpc, dl, MVT::f64, Src); | |||
8358 | return convertIntToFP(Op, Mov, DAG, Subtarget); | |||
8359 | } | |||
8360 | ||||
8361 | static SDValue widenVec(SelectionDAG &DAG, SDValue Vec, const SDLoc &dl) { | |||
8362 | ||||
8363 | EVT VecVT = Vec.getValueType(); | |||
8364 | assert(VecVT.isVector() && "Expected a vector type.")(static_cast <bool> (VecVT.isVector() && "Expected a vector type." ) ? void (0) : __assert_fail ("VecVT.isVector() && \"Expected a vector type.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8364, __extension__ __PRETTY_FUNCTION__)); | |||
8365 | assert(VecVT.getSizeInBits() < 128 && "Vector is already full width.")(static_cast <bool> (VecVT.getSizeInBits() < 128 && "Vector is already full width.") ? void (0) : __assert_fail ( "VecVT.getSizeInBits() < 128 && \"Vector is already full width.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8365, __extension__ __PRETTY_FUNCTION__)); | |||
8366 | ||||
8367 | EVT EltVT = VecVT.getVectorElementType(); | |||
8368 | unsigned WideNumElts = 128 / EltVT.getSizeInBits(); | |||
8369 | EVT WideVT = EVT::getVectorVT(*DAG.getContext(), EltVT, WideNumElts); | |||
8370 | ||||
8371 | unsigned NumConcat = WideNumElts / VecVT.getVectorNumElements(); | |||
8372 | SmallVector<SDValue, 16> Ops(NumConcat); | |||
8373 | Ops[0] = Vec; | |||
8374 | SDValue UndefVec = DAG.getUNDEF(VecVT); | |||
8375 | for (unsigned i = 1; i < NumConcat; ++i) | |||
8376 | Ops[i] = UndefVec; | |||
8377 | ||||
8378 | return DAG.getNode(ISD::CONCAT_VECTORS, dl, WideVT, Ops); | |||
8379 | } | |||
8380 | ||||
8381 | SDValue PPCTargetLowering::LowerINT_TO_FPVector(SDValue Op, SelectionDAG &DAG, | |||
8382 | const SDLoc &dl) const { | |||
8383 | bool IsStrict = Op->isStrictFPOpcode(); | |||
8384 | unsigned Opc = Op.getOpcode(); | |||
8385 | SDValue Src = Op.getOperand(IsStrict ? 1 : 0); | |||
8386 | assert((Opc == ISD::UINT_TO_FP || Opc == ISD::SINT_TO_FP ||(static_cast <bool> ((Opc == ISD::UINT_TO_FP || Opc == ISD ::SINT_TO_FP || Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD:: STRICT_SINT_TO_FP) && "Unexpected conversion type") ? void (0) : __assert_fail ("(Opc == ISD::UINT_TO_FP || Opc == ISD::SINT_TO_FP || Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD::STRICT_SINT_TO_FP) && \"Unexpected conversion type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8388, __extension__ __PRETTY_FUNCTION__)) | |||
8387 | Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD::STRICT_SINT_TO_FP) &&(static_cast <bool> ((Opc == ISD::UINT_TO_FP || Opc == ISD ::SINT_TO_FP || Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD:: STRICT_SINT_TO_FP) && "Unexpected conversion type") ? void (0) : __assert_fail ("(Opc == ISD::UINT_TO_FP || Opc == ISD::SINT_TO_FP || Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD::STRICT_SINT_TO_FP) && \"Unexpected conversion type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8388, __extension__ __PRETTY_FUNCTION__)) | |||
8388 | "Unexpected conversion type")(static_cast <bool> ((Opc == ISD::UINT_TO_FP || Opc == ISD ::SINT_TO_FP || Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD:: STRICT_SINT_TO_FP) && "Unexpected conversion type") ? void (0) : __assert_fail ("(Opc == ISD::UINT_TO_FP || Opc == ISD::SINT_TO_FP || Opc == ISD::STRICT_UINT_TO_FP || Opc == ISD::STRICT_SINT_TO_FP) && \"Unexpected conversion type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8388, __extension__ __PRETTY_FUNCTION__)); | |||
8389 | assert((Op.getValueType() == MVT::v2f64 || Op.getValueType() == MVT::v4f32) &&(static_cast <bool> ((Op.getValueType() == MVT::v2f64 || Op.getValueType() == MVT::v4f32) && "Supports conversions to v2f64/v4f32 only." ) ? void (0) : __assert_fail ("(Op.getValueType() == MVT::v2f64 || Op.getValueType() == MVT::v4f32) && \"Supports conversions to v2f64/v4f32 only.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8390, __extension__ __PRETTY_FUNCTION__)) | |||
8390 | "Supports conversions to v2f64/v4f32 only.")(static_cast <bool> ((Op.getValueType() == MVT::v2f64 || Op.getValueType() == MVT::v4f32) && "Supports conversions to v2f64/v4f32 only." ) ? void (0) : __assert_fail ("(Op.getValueType() == MVT::v2f64 || Op.getValueType() == MVT::v4f32) && \"Supports conversions to v2f64/v4f32 only.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8390, __extension__ __PRETTY_FUNCTION__)); | |||
8391 | ||||
8392 | // TODO: Any other flags to propagate? | |||
8393 | SDNodeFlags Flags; | |||
8394 | Flags.setNoFPExcept(Op->getFlags().hasNoFPExcept()); | |||
8395 | ||||
8396 | bool SignedConv = Opc == ISD::SINT_TO_FP || Opc == ISD::STRICT_SINT_TO_FP; | |||
8397 | bool FourEltRes = Op.getValueType() == MVT::v4f32; | |||
8398 | ||||
8399 | SDValue Wide = widenVec(DAG, Src, dl); | |||
8400 | EVT WideVT = Wide.getValueType(); | |||
8401 | unsigned WideNumElts = WideVT.getVectorNumElements(); | |||
8402 | MVT IntermediateVT = FourEltRes ? MVT::v4i32 : MVT::v2i64; | |||
8403 | ||||
8404 | SmallVector<int, 16> ShuffV; | |||
8405 | for (unsigned i = 0; i < WideNumElts; ++i) | |||
8406 | ShuffV.push_back(i + WideNumElts); | |||
8407 | ||||
8408 | int Stride = FourEltRes ? WideNumElts / 4 : WideNumElts / 2; | |||
8409 | int SaveElts = FourEltRes ? 4 : 2; | |||
8410 | if (Subtarget.isLittleEndian()) | |||
8411 | for (int i = 0; i < SaveElts; i++) | |||
8412 | ShuffV[i * Stride] = i; | |||
8413 | else | |||
8414 | for (int i = 1; i <= SaveElts; i++) | |||
8415 | ShuffV[i * Stride - 1] = i - 1; | |||
8416 | ||||
8417 | SDValue ShuffleSrc2 = | |||
8418 | SignedConv ? DAG.getUNDEF(WideVT) : DAG.getConstant(0, dl, WideVT); | |||
8419 | SDValue Arrange = DAG.getVectorShuffle(WideVT, dl, Wide, ShuffleSrc2, ShuffV); | |||
8420 | ||||
8421 | SDValue Extend; | |||
8422 | if (SignedConv) { | |||
8423 | Arrange = DAG.getBitcast(IntermediateVT, Arrange); | |||
8424 | EVT ExtVT = Src.getValueType(); | |||
8425 | if (Subtarget.hasP9Altivec()) | |||
8426 | ExtVT = EVT::getVectorVT(*DAG.getContext(), WideVT.getVectorElementType(), | |||
8427 | IntermediateVT.getVectorNumElements()); | |||
8428 | ||||
8429 | Extend = DAG.getNode(ISD::SIGN_EXTEND_INREG, dl, IntermediateVT, Arrange, | |||
8430 | DAG.getValueType(ExtVT)); | |||
8431 | } else | |||
8432 | Extend = DAG.getNode(ISD::BITCAST, dl, IntermediateVT, Arrange); | |||
8433 | ||||
8434 | if (IsStrict) | |||
8435 | return DAG.getNode(Opc, dl, DAG.getVTList(Op.getValueType(), MVT::Other), | |||
8436 | {Op.getOperand(0), Extend}, Flags); | |||
8437 | ||||
8438 | return DAG.getNode(Opc, dl, Op.getValueType(), Extend); | |||
8439 | } | |||
8440 | ||||
8441 | SDValue PPCTargetLowering::LowerINT_TO_FP(SDValue Op, | |||
8442 | SelectionDAG &DAG) const { | |||
8443 | SDLoc dl(Op); | |||
8444 | bool IsSigned = Op.getOpcode() == ISD::SINT_TO_FP || | |||
8445 | Op.getOpcode() == ISD::STRICT_SINT_TO_FP; | |||
8446 | bool IsStrict = Op->isStrictFPOpcode(); | |||
8447 | SDValue Src = Op.getOperand(IsStrict ? 1 : 0); | |||
8448 | SDValue Chain = IsStrict ? Op.getOperand(0) : DAG.getEntryNode(); | |||
8449 | ||||
8450 | // TODO: Any other flags to propagate? | |||
8451 | SDNodeFlags Flags; | |||
8452 | Flags.setNoFPExcept(Op->getFlags().hasNoFPExcept()); | |||
8453 | ||||
8454 | EVT InVT = Src.getValueType(); | |||
8455 | EVT OutVT = Op.getValueType(); | |||
8456 | if (OutVT.isVector() && OutVT.isFloatingPoint() && | |||
8457 | isOperationCustom(Op.getOpcode(), InVT)) | |||
8458 | return LowerINT_TO_FPVector(Op, DAG, dl); | |||
8459 | ||||
8460 | // Conversions to f128 are legal. | |||
8461 | if (Op.getValueType() == MVT::f128) | |||
8462 | return Subtarget.hasP9Vector() ? Op : SDValue(); | |||
8463 | ||||
8464 | // Don't handle ppc_fp128 here; let it be lowered to a libcall. | |||
8465 | if (Op.getValueType() != MVT::f32 && Op.getValueType() != MVT::f64) | |||
8466 | return SDValue(); | |||
8467 | ||||
8468 | if (Src.getValueType() == MVT::i1) { | |||
8469 | SDValue Sel = DAG.getNode(ISD::SELECT, dl, Op.getValueType(), Src, | |||
8470 | DAG.getConstantFP(1.0, dl, Op.getValueType()), | |||
8471 | DAG.getConstantFP(0.0, dl, Op.getValueType())); | |||
8472 | if (IsStrict) | |||
8473 | return DAG.getMergeValues({Sel, Chain}, dl); | |||
8474 | else | |||
8475 | return Sel; | |||
8476 | } | |||
8477 | ||||
8478 | // If we have direct moves, we can do all the conversion, skip the store/load | |||
8479 | // however, without FPCVT we can't do most conversions. | |||
8480 | if (Subtarget.hasDirectMove() && directMoveIsProfitable(Op) && | |||
8481 | Subtarget.isPPC64() && Subtarget.hasFPCVT()) | |||
8482 | return LowerINT_TO_FPDirectMove(Op, DAG, dl); | |||
8483 | ||||
8484 | assert((IsSigned || Subtarget.hasFPCVT()) &&(static_cast <bool> ((IsSigned || Subtarget.hasFPCVT()) && "UINT_TO_FP is supported only with FPCVT") ? void (0) : __assert_fail ("(IsSigned || Subtarget.hasFPCVT()) && \"UINT_TO_FP is supported only with FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8485, __extension__ __PRETTY_FUNCTION__)) | |||
8485 | "UINT_TO_FP is supported only with FPCVT")(static_cast <bool> ((IsSigned || Subtarget.hasFPCVT()) && "UINT_TO_FP is supported only with FPCVT") ? void (0) : __assert_fail ("(IsSigned || Subtarget.hasFPCVT()) && \"UINT_TO_FP is supported only with FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8485, __extension__ __PRETTY_FUNCTION__)); | |||
8486 | ||||
8487 | if (Src.getValueType() == MVT::i64) { | |||
8488 | SDValue SINT = Src; | |||
8489 | // When converting to single-precision, we actually need to convert | |||
8490 | // to double-precision first and then round to single-precision. | |||
8491 | // To avoid double-rounding effects during that operation, we have | |||
8492 | // to prepare the input operand. Bits that might be truncated when | |||
8493 | // converting to double-precision are replaced by a bit that won't | |||
8494 | // be lost at this stage, but is below the single-precision rounding | |||
8495 | // position. | |||
8496 | // | |||
8497 | // However, if -enable-unsafe-fp-math is in effect, accept double | |||
8498 | // rounding to avoid the extra overhead. | |||
8499 | if (Op.getValueType() == MVT::f32 && | |||
8500 | !Subtarget.hasFPCVT() && | |||
8501 | !DAG.getTarget().Options.UnsafeFPMath) { | |||
8502 | ||||
8503 | // Twiddle input to make sure the low 11 bits are zero. (If this | |||
8504 | // is the case, we are guaranteed the value will fit into the 53 bit | |||
8505 | // mantissa of an IEEE double-precision value without rounding.) | |||
8506 | // If any of those low 11 bits were not zero originally, make sure | |||
8507 | // bit 12 (value 2048) is set instead, so that the final rounding | |||
8508 | // to single-precision gets the correct result. | |||
8509 | SDValue Round = DAG.getNode(ISD::AND, dl, MVT::i64, | |||
8510 | SINT, DAG.getConstant(2047, dl, MVT::i64)); | |||
8511 | Round = DAG.getNode(ISD::ADD, dl, MVT::i64, | |||
8512 | Round, DAG.getConstant(2047, dl, MVT::i64)); | |||
8513 | Round = DAG.getNode(ISD::OR, dl, MVT::i64, Round, SINT); | |||
8514 | Round = DAG.getNode(ISD::AND, dl, MVT::i64, | |||
8515 | Round, DAG.getConstant(-2048, dl, MVT::i64)); | |||
8516 | ||||
8517 | // However, we cannot use that value unconditionally: if the magnitude | |||
8518 | // of the input value is small, the bit-twiddling we did above might | |||
8519 | // end up visibly changing the output. Fortunately, in that case, we | |||
8520 | // don't need to twiddle bits since the original input will convert | |||
8521 | // exactly to double-precision floating-point already. Therefore, | |||
8522 | // construct a conditional to use the original value if the top 11 | |||
8523 | // bits are all sign-bit copies, and use the rounded value computed | |||
8524 | // above otherwise. | |||
8525 | SDValue Cond = DAG.getNode(ISD::SRA, dl, MVT::i64, | |||
8526 | SINT, DAG.getConstant(53, dl, MVT::i32)); | |||
8527 | Cond = DAG.getNode(ISD::ADD, dl, MVT::i64, | |||
8528 | Cond, DAG.getConstant(1, dl, MVT::i64)); | |||
8529 | Cond = DAG.getSetCC( | |||
8530 | dl, | |||
8531 | getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), MVT::i64), | |||
8532 | Cond, DAG.getConstant(1, dl, MVT::i64), ISD::SETUGT); | |||
8533 | ||||
8534 | SINT = DAG.getNode(ISD::SELECT, dl, MVT::i64, Cond, Round, SINT); | |||
8535 | } | |||
8536 | ||||
8537 | ReuseLoadInfo RLI; | |||
8538 | SDValue Bits; | |||
8539 | ||||
8540 | MachineFunction &MF = DAG.getMachineFunction(); | |||
8541 | if (canReuseLoadAddress(SINT, MVT::i64, RLI, DAG)) { | |||
8542 | Bits = DAG.getLoad(MVT::f64, dl, RLI.Chain, RLI.Ptr, RLI.MPI, | |||
8543 | RLI.Alignment, RLI.MMOFlags(), RLI.AAInfo, RLI.Ranges); | |||
8544 | spliceIntoChain(RLI.ResChain, Bits.getValue(1), DAG); | |||
8545 | } else if (Subtarget.hasLFIWAX() && | |||
8546 | canReuseLoadAddress(SINT, MVT::i32, RLI, DAG, ISD::SEXTLOAD)) { | |||
8547 | MachineMemOperand *MMO = | |||
8548 | MF.getMachineMemOperand(RLI.MPI, MachineMemOperand::MOLoad, 4, | |||
8549 | RLI.Alignment, RLI.AAInfo, RLI.Ranges); | |||
8550 | SDValue Ops[] = { RLI.Chain, RLI.Ptr }; | |||
8551 | Bits = DAG.getMemIntrinsicNode(PPCISD::LFIWAX, dl, | |||
8552 | DAG.getVTList(MVT::f64, MVT::Other), | |||
8553 | Ops, MVT::i32, MMO); | |||
8554 | spliceIntoChain(RLI.ResChain, Bits.getValue(1), DAG); | |||
8555 | } else if (Subtarget.hasFPCVT() && | |||
8556 | canReuseLoadAddress(SINT, MVT::i32, RLI, DAG, ISD::ZEXTLOAD)) { | |||
8557 | MachineMemOperand *MMO = | |||
8558 | MF.getMachineMemOperand(RLI.MPI, MachineMemOperand::MOLoad, 4, | |||
8559 | RLI.Alignment, RLI.AAInfo, RLI.Ranges); | |||
8560 | SDValue Ops[] = { RLI.Chain, RLI.Ptr }; | |||
8561 | Bits = DAG.getMemIntrinsicNode(PPCISD::LFIWZX, dl, | |||
8562 | DAG.getVTList(MVT::f64, MVT::Other), | |||
8563 | Ops, MVT::i32, MMO); | |||
8564 | spliceIntoChain(RLI.ResChain, Bits.getValue(1), DAG); | |||
8565 | } else if (((Subtarget.hasLFIWAX() && | |||
8566 | SINT.getOpcode() == ISD::SIGN_EXTEND) || | |||
8567 | (Subtarget.hasFPCVT() && | |||
8568 | SINT.getOpcode() == ISD::ZERO_EXTEND)) && | |||
8569 | SINT.getOperand(0).getValueType() == MVT::i32) { | |||
8570 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
8571 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
8572 | ||||
8573 | int FrameIdx = MFI.CreateStackObject(4, Align(4), false); | |||
8574 | SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); | |||
8575 | ||||
8576 | SDValue Store = DAG.getStore(Chain, dl, SINT.getOperand(0), FIdx, | |||
8577 | MachinePointerInfo::getFixedStack( | |||
8578 | DAG.getMachineFunction(), FrameIdx)); | |||
8579 | Chain = Store; | |||
8580 | ||||
8581 | assert(cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 &&(static_cast <bool> (cast<StoreSDNode>(Store)-> getMemoryVT() == MVT::i32 && "Expected an i32 store") ? void (0) : __assert_fail ("cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 && \"Expected an i32 store\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8582, __extension__ __PRETTY_FUNCTION__)) | |||
8582 | "Expected an i32 store")(static_cast <bool> (cast<StoreSDNode>(Store)-> getMemoryVT() == MVT::i32 && "Expected an i32 store") ? void (0) : __assert_fail ("cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 && \"Expected an i32 store\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8582, __extension__ __PRETTY_FUNCTION__)); | |||
8583 | ||||
8584 | RLI.Ptr = FIdx; | |||
8585 | RLI.Chain = Chain; | |||
8586 | RLI.MPI = | |||
8587 | MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FrameIdx); | |||
8588 | RLI.Alignment = Align(4); | |||
8589 | ||||
8590 | MachineMemOperand *MMO = | |||
8591 | MF.getMachineMemOperand(RLI.MPI, MachineMemOperand::MOLoad, 4, | |||
8592 | RLI.Alignment, RLI.AAInfo, RLI.Ranges); | |||
8593 | SDValue Ops[] = { RLI.Chain, RLI.Ptr }; | |||
8594 | Bits = DAG.getMemIntrinsicNode(SINT.getOpcode() == ISD::ZERO_EXTEND ? | |||
8595 | PPCISD::LFIWZX : PPCISD::LFIWAX, | |||
8596 | dl, DAG.getVTList(MVT::f64, MVT::Other), | |||
8597 | Ops, MVT::i32, MMO); | |||
8598 | Chain = Bits.getValue(1); | |||
8599 | } else | |||
8600 | Bits = DAG.getNode(ISD::BITCAST, dl, MVT::f64, SINT); | |||
8601 | ||||
8602 | SDValue FP = convertIntToFP(Op, Bits, DAG, Subtarget, Chain); | |||
8603 | if (IsStrict) | |||
8604 | Chain = FP.getValue(1); | |||
8605 | ||||
8606 | if (Op.getValueType() == MVT::f32 && !Subtarget.hasFPCVT()) { | |||
8607 | if (IsStrict) | |||
8608 | FP = DAG.getNode(ISD::STRICT_FP_ROUND, dl, | |||
8609 | DAG.getVTList(MVT::f32, MVT::Other), | |||
8610 | {Chain, FP, DAG.getIntPtrConstant(0, dl)}, Flags); | |||
8611 | else | |||
8612 | FP = DAG.getNode(ISD::FP_ROUND, dl, MVT::f32, FP, | |||
8613 | DAG.getIntPtrConstant(0, dl)); | |||
8614 | } | |||
8615 | return FP; | |||
8616 | } | |||
8617 | ||||
8618 | assert(Src.getValueType() == MVT::i32 &&(static_cast <bool> (Src.getValueType() == MVT::i32 && "Unhandled INT_TO_FP type in custom expander!") ? void (0) : __assert_fail ("Src.getValueType() == MVT::i32 && \"Unhandled INT_TO_FP type in custom expander!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8619, __extension__ __PRETTY_FUNCTION__)) | |||
8619 | "Unhandled INT_TO_FP type in custom expander!")(static_cast <bool> (Src.getValueType() == MVT::i32 && "Unhandled INT_TO_FP type in custom expander!") ? void (0) : __assert_fail ("Src.getValueType() == MVT::i32 && \"Unhandled INT_TO_FP type in custom expander!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8619, __extension__ __PRETTY_FUNCTION__)); | |||
8620 | // Since we only generate this in 64-bit mode, we can take advantage of | |||
8621 | // 64-bit registers. In particular, sign extend the input value into the | |||
8622 | // 64-bit register with extsw, store the WHOLE 64-bit value into the stack | |||
8623 | // then lfd it and fcfid it. | |||
8624 | MachineFunction &MF = DAG.getMachineFunction(); | |||
8625 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
8626 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
8627 | ||||
8628 | SDValue Ld; | |||
8629 | if (Subtarget.hasLFIWAX() || Subtarget.hasFPCVT()) { | |||
8630 | ReuseLoadInfo RLI; | |||
8631 | bool ReusingLoad; | |||
8632 | if (!(ReusingLoad = canReuseLoadAddress(Src, MVT::i32, RLI, DAG))) { | |||
8633 | int FrameIdx = MFI.CreateStackObject(4, Align(4), false); | |||
8634 | SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); | |||
8635 | ||||
8636 | SDValue Store = DAG.getStore(Chain, dl, Src, FIdx, | |||
8637 | MachinePointerInfo::getFixedStack( | |||
8638 | DAG.getMachineFunction(), FrameIdx)); | |||
8639 | Chain = Store; | |||
8640 | ||||
8641 | assert(cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 &&(static_cast <bool> (cast<StoreSDNode>(Store)-> getMemoryVT() == MVT::i32 && "Expected an i32 store") ? void (0) : __assert_fail ("cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 && \"Expected an i32 store\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8642, __extension__ __PRETTY_FUNCTION__)) | |||
8642 | "Expected an i32 store")(static_cast <bool> (cast<StoreSDNode>(Store)-> getMemoryVT() == MVT::i32 && "Expected an i32 store") ? void (0) : __assert_fail ("cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 && \"Expected an i32 store\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8642, __extension__ __PRETTY_FUNCTION__)); | |||
8643 | ||||
8644 | RLI.Ptr = FIdx; | |||
8645 | RLI.Chain = Chain; | |||
8646 | RLI.MPI = | |||
8647 | MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FrameIdx); | |||
8648 | RLI.Alignment = Align(4); | |||
8649 | } | |||
8650 | ||||
8651 | MachineMemOperand *MMO = | |||
8652 | MF.getMachineMemOperand(RLI.MPI, MachineMemOperand::MOLoad, 4, | |||
8653 | RLI.Alignment, RLI.AAInfo, RLI.Ranges); | |||
8654 | SDValue Ops[] = { RLI.Chain, RLI.Ptr }; | |||
8655 | Ld = DAG.getMemIntrinsicNode(IsSigned ? PPCISD::LFIWAX : PPCISD::LFIWZX, dl, | |||
8656 | DAG.getVTList(MVT::f64, MVT::Other), Ops, | |||
8657 | MVT::i32, MMO); | |||
8658 | Chain = Ld.getValue(1); | |||
8659 | if (ReusingLoad) | |||
8660 | spliceIntoChain(RLI.ResChain, Ld.getValue(1), DAG); | |||
8661 | } else { | |||
8662 | assert(Subtarget.isPPC64() &&(static_cast <bool> (Subtarget.isPPC64() && "i32->FP without LFIWAX supported only on PPC64" ) ? void (0) : __assert_fail ("Subtarget.isPPC64() && \"i32->FP without LFIWAX supported only on PPC64\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8663, __extension__ __PRETTY_FUNCTION__)) | |||
8663 | "i32->FP without LFIWAX supported only on PPC64")(static_cast <bool> (Subtarget.isPPC64() && "i32->FP without LFIWAX supported only on PPC64" ) ? void (0) : __assert_fail ("Subtarget.isPPC64() && \"i32->FP without LFIWAX supported only on PPC64\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8663, __extension__ __PRETTY_FUNCTION__)); | |||
8664 | ||||
8665 | int FrameIdx = MFI.CreateStackObject(8, Align(8), false); | |||
8666 | SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); | |||
8667 | ||||
8668 | SDValue Ext64 = DAG.getNode(ISD::SIGN_EXTEND, dl, MVT::i64, Src); | |||
8669 | ||||
8670 | // STD the extended value into the stack slot. | |||
8671 | SDValue Store = DAG.getStore( | |||
8672 | Chain, dl, Ext64, FIdx, | |||
8673 | MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FrameIdx)); | |||
8674 | Chain = Store; | |||
8675 | ||||
8676 | // Load the value as a double. | |||
8677 | Ld = DAG.getLoad( | |||
8678 | MVT::f64, dl, Chain, FIdx, | |||
8679 | MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FrameIdx)); | |||
8680 | Chain = Ld.getValue(1); | |||
8681 | } | |||
8682 | ||||
8683 | // FCFID it and return it. | |||
8684 | SDValue FP = convertIntToFP(Op, Ld, DAG, Subtarget, Chain); | |||
8685 | if (IsStrict) | |||
8686 | Chain = FP.getValue(1); | |||
8687 | if (Op.getValueType() == MVT::f32 && !Subtarget.hasFPCVT()) { | |||
8688 | if (IsStrict) | |||
8689 | FP = DAG.getNode(ISD::STRICT_FP_ROUND, dl, | |||
8690 | DAG.getVTList(MVT::f32, MVT::Other), | |||
8691 | {Chain, FP, DAG.getIntPtrConstant(0, dl)}, Flags); | |||
8692 | else | |||
8693 | FP = DAG.getNode(ISD::FP_ROUND, dl, MVT::f32, FP, | |||
8694 | DAG.getIntPtrConstant(0, dl)); | |||
8695 | } | |||
8696 | return FP; | |||
8697 | } | |||
8698 | ||||
8699 | SDValue PPCTargetLowering::LowerFLT_ROUNDS_(SDValue Op, | |||
8700 | SelectionDAG &DAG) const { | |||
8701 | SDLoc dl(Op); | |||
8702 | /* | |||
8703 | The rounding mode is in bits 30:31 of FPSR, and has the following | |||
8704 | settings: | |||
8705 | 00 Round to nearest | |||
8706 | 01 Round to 0 | |||
8707 | 10 Round to +inf | |||
8708 | 11 Round to -inf | |||
8709 | ||||
8710 | FLT_ROUNDS, on the other hand, expects the following: | |||
8711 | -1 Undefined | |||
8712 | 0 Round to 0 | |||
8713 | 1 Round to nearest | |||
8714 | 2 Round to +inf | |||
8715 | 3 Round to -inf | |||
8716 | ||||
8717 | To perform the conversion, we do: | |||
8718 | ((FPSCR & 0x3) ^ ((~FPSCR & 0x3) >> 1)) | |||
8719 | */ | |||
8720 | ||||
8721 | MachineFunction &MF = DAG.getMachineFunction(); | |||
8722 | EVT VT = Op.getValueType(); | |||
8723 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
8724 | ||||
8725 | // Save FP Control Word to register | |||
8726 | SDValue Chain = Op.getOperand(0); | |||
8727 | SDValue MFFS = DAG.getNode(PPCISD::MFFS, dl, {MVT::f64, MVT::Other}, Chain); | |||
8728 | Chain = MFFS.getValue(1); | |||
8729 | ||||
8730 | SDValue CWD; | |||
8731 | if (isTypeLegal(MVT::i64)) { | |||
8732 | CWD = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, | |||
8733 | DAG.getNode(ISD::BITCAST, dl, MVT::i64, MFFS)); | |||
8734 | } else { | |||
8735 | // Save FP register to stack slot | |||
8736 | int SSFI = MF.getFrameInfo().CreateStackObject(8, Align(8), false); | |||
8737 | SDValue StackSlot = DAG.getFrameIndex(SSFI, PtrVT); | |||
8738 | Chain = DAG.getStore(Chain, dl, MFFS, StackSlot, MachinePointerInfo()); | |||
8739 | ||||
8740 | // Load FP Control Word from low 32 bits of stack slot. | |||
8741 | assert(hasBigEndianPartOrdering(MVT::i64, MF.getDataLayout()) &&(static_cast <bool> (hasBigEndianPartOrdering(MVT::i64, MF.getDataLayout()) && "Stack slot adjustment is valid only on big endian subtargets!" ) ? void (0) : __assert_fail ("hasBigEndianPartOrdering(MVT::i64, MF.getDataLayout()) && \"Stack slot adjustment is valid only on big endian subtargets!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8742, __extension__ __PRETTY_FUNCTION__)) | |||
8742 | "Stack slot adjustment is valid only on big endian subtargets!")(static_cast <bool> (hasBigEndianPartOrdering(MVT::i64, MF.getDataLayout()) && "Stack slot adjustment is valid only on big endian subtargets!" ) ? void (0) : __assert_fail ("hasBigEndianPartOrdering(MVT::i64, MF.getDataLayout()) && \"Stack slot adjustment is valid only on big endian subtargets!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8742, __extension__ __PRETTY_FUNCTION__)); | |||
8743 | SDValue Four = DAG.getConstant(4, dl, PtrVT); | |||
8744 | SDValue Addr = DAG.getNode(ISD::ADD, dl, PtrVT, StackSlot, Four); | |||
8745 | CWD = DAG.getLoad(MVT::i32, dl, Chain, Addr, MachinePointerInfo()); | |||
8746 | Chain = CWD.getValue(1); | |||
8747 | } | |||
8748 | ||||
8749 | // Transform as necessary | |||
8750 | SDValue CWD1 = | |||
8751 | DAG.getNode(ISD::AND, dl, MVT::i32, | |||
8752 | CWD, DAG.getConstant(3, dl, MVT::i32)); | |||
8753 | SDValue CWD2 = | |||
8754 | DAG.getNode(ISD::SRL, dl, MVT::i32, | |||
8755 | DAG.getNode(ISD::AND, dl, MVT::i32, | |||
8756 | DAG.getNode(ISD::XOR, dl, MVT::i32, | |||
8757 | CWD, DAG.getConstant(3, dl, MVT::i32)), | |||
8758 | DAG.getConstant(3, dl, MVT::i32)), | |||
8759 | DAG.getConstant(1, dl, MVT::i32)); | |||
8760 | ||||
8761 | SDValue RetVal = | |||
8762 | DAG.getNode(ISD::XOR, dl, MVT::i32, CWD1, CWD2); | |||
8763 | ||||
8764 | RetVal = | |||
8765 | DAG.getNode((VT.getSizeInBits() < 16 ? ISD::TRUNCATE : ISD::ZERO_EXTEND), | |||
8766 | dl, VT, RetVal); | |||
8767 | ||||
8768 | return DAG.getMergeValues({RetVal, Chain}, dl); | |||
8769 | } | |||
8770 | ||||
8771 | SDValue PPCTargetLowering::LowerSHL_PARTS(SDValue Op, SelectionDAG &DAG) const { | |||
8772 | EVT VT = Op.getValueType(); | |||
8773 | unsigned BitWidth = VT.getSizeInBits(); | |||
8774 | SDLoc dl(Op); | |||
8775 | assert(Op.getNumOperands() == 3 &&(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SHL!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SHL!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8777, __extension__ __PRETTY_FUNCTION__)) | |||
8776 | VT == Op.getOperand(1).getValueType() &&(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SHL!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SHL!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8777, __extension__ __PRETTY_FUNCTION__)) | |||
8777 | "Unexpected SHL!")(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SHL!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SHL!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8777, __extension__ __PRETTY_FUNCTION__)); | |||
8778 | ||||
8779 | // Expand into a bunch of logical ops. Note that these ops | |||
8780 | // depend on the PPC behavior for oversized shift amounts. | |||
8781 | SDValue Lo = Op.getOperand(0); | |||
8782 | SDValue Hi = Op.getOperand(1); | |||
8783 | SDValue Amt = Op.getOperand(2); | |||
8784 | EVT AmtVT = Amt.getValueType(); | |||
8785 | ||||
8786 | SDValue Tmp1 = DAG.getNode(ISD::SUB, dl, AmtVT, | |||
8787 | DAG.getConstant(BitWidth, dl, AmtVT), Amt); | |||
8788 | SDValue Tmp2 = DAG.getNode(PPCISD::SHL, dl, VT, Hi, Amt); | |||
8789 | SDValue Tmp3 = DAG.getNode(PPCISD::SRL, dl, VT, Lo, Tmp1); | |||
8790 | SDValue Tmp4 = DAG.getNode(ISD::OR , dl, VT, Tmp2, Tmp3); | |||
8791 | SDValue Tmp5 = DAG.getNode(ISD::ADD, dl, AmtVT, Amt, | |||
8792 | DAG.getConstant(-BitWidth, dl, AmtVT)); | |||
8793 | SDValue Tmp6 = DAG.getNode(PPCISD::SHL, dl, VT, Lo, Tmp5); | |||
8794 | SDValue OutHi = DAG.getNode(ISD::OR, dl, VT, Tmp4, Tmp6); | |||
8795 | SDValue OutLo = DAG.getNode(PPCISD::SHL, dl, VT, Lo, Amt); | |||
8796 | SDValue OutOps[] = { OutLo, OutHi }; | |||
8797 | return DAG.getMergeValues(OutOps, dl); | |||
8798 | } | |||
8799 | ||||
8800 | SDValue PPCTargetLowering::LowerSRL_PARTS(SDValue Op, SelectionDAG &DAG) const { | |||
8801 | EVT VT = Op.getValueType(); | |||
8802 | SDLoc dl(Op); | |||
8803 | unsigned BitWidth = VT.getSizeInBits(); | |||
8804 | assert(Op.getNumOperands() == 3 &&(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SRL!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SRL!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8806, __extension__ __PRETTY_FUNCTION__)) | |||
8805 | VT == Op.getOperand(1).getValueType() &&(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SRL!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SRL!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8806, __extension__ __PRETTY_FUNCTION__)) | |||
8806 | "Unexpected SRL!")(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SRL!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SRL!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8806, __extension__ __PRETTY_FUNCTION__)); | |||
8807 | ||||
8808 | // Expand into a bunch of logical ops. Note that these ops | |||
8809 | // depend on the PPC behavior for oversized shift amounts. | |||
8810 | SDValue Lo = Op.getOperand(0); | |||
8811 | SDValue Hi = Op.getOperand(1); | |||
8812 | SDValue Amt = Op.getOperand(2); | |||
8813 | EVT AmtVT = Amt.getValueType(); | |||
8814 | ||||
8815 | SDValue Tmp1 = DAG.getNode(ISD::SUB, dl, AmtVT, | |||
8816 | DAG.getConstant(BitWidth, dl, AmtVT), Amt); | |||
8817 | SDValue Tmp2 = DAG.getNode(PPCISD::SRL, dl, VT, Lo, Amt); | |||
8818 | SDValue Tmp3 = DAG.getNode(PPCISD::SHL, dl, VT, Hi, Tmp1); | |||
8819 | SDValue Tmp4 = DAG.getNode(ISD::OR, dl, VT, Tmp2, Tmp3); | |||
8820 | SDValue Tmp5 = DAG.getNode(ISD::ADD, dl, AmtVT, Amt, | |||
8821 | DAG.getConstant(-BitWidth, dl, AmtVT)); | |||
8822 | SDValue Tmp6 = DAG.getNode(PPCISD::SRL, dl, VT, Hi, Tmp5); | |||
8823 | SDValue OutLo = DAG.getNode(ISD::OR, dl, VT, Tmp4, Tmp6); | |||
8824 | SDValue OutHi = DAG.getNode(PPCISD::SRL, dl, VT, Hi, Amt); | |||
8825 | SDValue OutOps[] = { OutLo, OutHi }; | |||
8826 | return DAG.getMergeValues(OutOps, dl); | |||
8827 | } | |||
8828 | ||||
8829 | SDValue PPCTargetLowering::LowerSRA_PARTS(SDValue Op, SelectionDAG &DAG) const { | |||
8830 | SDLoc dl(Op); | |||
8831 | EVT VT = Op.getValueType(); | |||
8832 | unsigned BitWidth = VT.getSizeInBits(); | |||
8833 | assert(Op.getNumOperands() == 3 &&(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SRA!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SRA!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8835, __extension__ __PRETTY_FUNCTION__)) | |||
8834 | VT == Op.getOperand(1).getValueType() &&(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SRA!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SRA!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8835, __extension__ __PRETTY_FUNCTION__)) | |||
8835 | "Unexpected SRA!")(static_cast <bool> (Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && "Unexpected SRA!" ) ? void (0) : __assert_fail ("Op.getNumOperands() == 3 && VT == Op.getOperand(1).getValueType() && \"Unexpected SRA!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 8835, __extension__ __PRETTY_FUNCTION__)); | |||
8836 | ||||
8837 | // Expand into a bunch of logical ops, followed by a select_cc. | |||
8838 | SDValue Lo = Op.getOperand(0); | |||
8839 | SDValue Hi = Op.getOperand(1); | |||
8840 | SDValue Amt = Op.getOperand(2); | |||
8841 | EVT AmtVT = Amt.getValueType(); | |||
8842 | ||||
8843 | SDValue Tmp1 = DAG.getNode(ISD::SUB, dl, AmtVT, | |||
8844 | DAG.getConstant(BitWidth, dl, AmtVT), Amt); | |||
8845 | SDValue Tmp2 = DAG.getNode(PPCISD::SRL, dl, VT, Lo, Amt); | |||
8846 | SDValue Tmp3 = DAG.getNode(PPCISD::SHL, dl, VT, Hi, Tmp1); | |||
8847 | SDValue Tmp4 = DAG.getNode(ISD::OR, dl, VT, Tmp2, Tmp3); | |||
8848 | SDValue Tmp5 = DAG.getNode(ISD::ADD, dl, AmtVT, Amt, | |||
8849 | DAG.getConstant(-BitWidth, dl, AmtVT)); | |||
8850 | SDValue Tmp6 = DAG.getNode(PPCISD::SRA, dl, VT, Hi, Tmp5); | |||
8851 | SDValue OutHi = DAG.getNode(PPCISD::SRA, dl, VT, Hi, Amt); | |||
8852 | SDValue OutLo = DAG.getSelectCC(dl, Tmp5, DAG.getConstant(0, dl, AmtVT), | |||
8853 | Tmp4, Tmp6, ISD::SETLE); | |||
8854 | SDValue OutOps[] = { OutLo, OutHi }; | |||
8855 | return DAG.getMergeValues(OutOps, dl); | |||
8856 | } | |||
8857 | ||||
8858 | SDValue PPCTargetLowering::LowerFunnelShift(SDValue Op, | |||
8859 | SelectionDAG &DAG) const { | |||
8860 | SDLoc dl(Op); | |||
8861 | EVT VT = Op.getValueType(); | |||
8862 | unsigned BitWidth = VT.getSizeInBits(); | |||
8863 | ||||
8864 | bool IsFSHL = Op.getOpcode() == ISD::FSHL; | |||
8865 | SDValue X = Op.getOperand(0); | |||
8866 | SDValue Y = Op.getOperand(1); | |||
8867 | SDValue Z = Op.getOperand(2); | |||
8868 | EVT AmtVT = Z.getValueType(); | |||
8869 | ||||
8870 | // fshl: (X << (Z % BW)) | (Y >> (BW - (Z % BW))) | |||
8871 | // fshr: (X << (BW - (Z % BW))) | (Y >> (Z % BW)) | |||
8872 | // This is simpler than TargetLowering::expandFunnelShift because we can rely | |||
8873 | // on PowerPC shift by BW being well defined. | |||
8874 | Z = DAG.getNode(ISD::AND, dl, AmtVT, Z, | |||
8875 | DAG.getConstant(BitWidth - 1, dl, AmtVT)); | |||
8876 | SDValue SubZ = | |||
8877 | DAG.getNode(ISD::SUB, dl, AmtVT, DAG.getConstant(BitWidth, dl, AmtVT), Z); | |||
8878 | X = DAG.getNode(PPCISD::SHL, dl, VT, X, IsFSHL ? Z : SubZ); | |||
8879 | Y = DAG.getNode(PPCISD::SRL, dl, VT, Y, IsFSHL ? SubZ : Z); | |||
8880 | return DAG.getNode(ISD::OR, dl, VT, X, Y); | |||
8881 | } | |||
8882 | ||||
8883 | //===----------------------------------------------------------------------===// | |||
8884 | // Vector related lowering. | |||
8885 | // | |||
8886 | ||||
8887 | /// getCanonicalConstSplat - Build a canonical splat immediate of Val with an | |||
8888 | /// element size of SplatSize. Cast the result to VT. | |||
8889 | static SDValue getCanonicalConstSplat(uint64_t Val, unsigned SplatSize, EVT VT, | |||
8890 | SelectionDAG &DAG, const SDLoc &dl) { | |||
8891 | static const MVT VTys[] = { // canonical VT to use for each size. | |||
8892 | MVT::v16i8, MVT::v8i16, MVT::Other, MVT::v4i32 | |||
8893 | }; | |||
8894 | ||||
8895 | EVT ReqVT = VT != MVT::Other ? VT : VTys[SplatSize-1]; | |||
8896 | ||||
8897 | // For a splat with all ones, turn it to vspltisb 0xFF to canonicalize. | |||
8898 | if (Val == ((1LLU << (SplatSize * 8)) - 1)) { | |||
8899 | SplatSize = 1; | |||
8900 | Val = 0xFF; | |||
8901 | } | |||
8902 | ||||
8903 | EVT CanonicalVT = VTys[SplatSize-1]; | |||
8904 | ||||
8905 | // Build a canonical splat for this value. | |||
8906 | return DAG.getBitcast(ReqVT, DAG.getConstant(Val, dl, CanonicalVT)); | |||
8907 | } | |||
8908 | ||||
8909 | /// BuildIntrinsicOp - Return a unary operator intrinsic node with the | |||
8910 | /// specified intrinsic ID. | |||
8911 | static SDValue BuildIntrinsicOp(unsigned IID, SDValue Op, SelectionDAG &DAG, | |||
8912 | const SDLoc &dl, EVT DestVT = MVT::Other) { | |||
8913 | if (DestVT == MVT::Other) DestVT = Op.getValueType(); | |||
8914 | return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, DestVT, | |||
8915 | DAG.getConstant(IID, dl, MVT::i32), Op); | |||
8916 | } | |||
8917 | ||||
8918 | /// BuildIntrinsicOp - Return a binary operator intrinsic node with the | |||
8919 | /// specified intrinsic ID. | |||
8920 | static SDValue BuildIntrinsicOp(unsigned IID, SDValue LHS, SDValue RHS, | |||
8921 | SelectionDAG &DAG, const SDLoc &dl, | |||
8922 | EVT DestVT = MVT::Other) { | |||
8923 | if (DestVT == MVT::Other) DestVT = LHS.getValueType(); | |||
8924 | return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, DestVT, | |||
8925 | DAG.getConstant(IID, dl, MVT::i32), LHS, RHS); | |||
8926 | } | |||
8927 | ||||
8928 | /// BuildIntrinsicOp - Return a ternary operator intrinsic node with the | |||
8929 | /// specified intrinsic ID. | |||
8930 | static SDValue BuildIntrinsicOp(unsigned IID, SDValue Op0, SDValue Op1, | |||
8931 | SDValue Op2, SelectionDAG &DAG, const SDLoc &dl, | |||
8932 | EVT DestVT = MVT::Other) { | |||
8933 | if (DestVT == MVT::Other) DestVT = Op0.getValueType(); | |||
8934 | return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, DestVT, | |||
8935 | DAG.getConstant(IID, dl, MVT::i32), Op0, Op1, Op2); | |||
8936 | } | |||
8937 | ||||
8938 | /// BuildVSLDOI - Return a VECTOR_SHUFFLE that is a vsldoi of the specified | |||
8939 | /// amount. The result has the specified value type. | |||
8940 | static SDValue BuildVSLDOI(SDValue LHS, SDValue RHS, unsigned Amt, EVT VT, | |||
8941 | SelectionDAG &DAG, const SDLoc &dl) { | |||
8942 | // Force LHS/RHS to be the right type. | |||
8943 | LHS = DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, LHS); | |||
8944 | RHS = DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, RHS); | |||
8945 | ||||
8946 | int Ops[16]; | |||
8947 | for (unsigned i = 0; i != 16; ++i) | |||
8948 | Ops[i] = i + Amt; | |||
8949 | SDValue T = DAG.getVectorShuffle(MVT::v16i8, dl, LHS, RHS, Ops); | |||
8950 | return DAG.getNode(ISD::BITCAST, dl, VT, T); | |||
8951 | } | |||
8952 | ||||
8953 | /// Do we have an efficient pattern in a .td file for this node? | |||
8954 | /// | |||
8955 | /// \param V - pointer to the BuildVectorSDNode being matched | |||
8956 | /// \param HasDirectMove - does this subtarget have VSR <-> GPR direct moves? | |||
8957 | /// | |||
8958 | /// There are some patterns where it is beneficial to keep a BUILD_VECTOR | |||
8959 | /// node as a BUILD_VECTOR node rather than expanding it. The patterns where | |||
8960 | /// the opposite is true (expansion is beneficial) are: | |||
8961 | /// - The node builds a vector out of integers that are not 32 or 64-bits | |||
8962 | /// - The node builds a vector out of constants | |||
8963 | /// - The node is a "load-and-splat" | |||
8964 | /// In all other cases, we will choose to keep the BUILD_VECTOR. | |||
8965 | static bool haveEfficientBuildVectorPattern(BuildVectorSDNode *V, | |||
8966 | bool HasDirectMove, | |||
8967 | bool HasP8Vector) { | |||
8968 | EVT VecVT = V->getValueType(0); | |||
8969 | bool RightType = VecVT == MVT::v2f64 || | |||
8970 | (HasP8Vector && VecVT == MVT::v4f32) || | |||
8971 | (HasDirectMove && (VecVT == MVT::v2i64 || VecVT == MVT::v4i32)); | |||
8972 | if (!RightType) | |||
8973 | return false; | |||
8974 | ||||
8975 | bool IsSplat = true; | |||
8976 | bool IsLoad = false; | |||
8977 | SDValue Op0 = V->getOperand(0); | |||
8978 | ||||
8979 | // This function is called in a block that confirms the node is not a constant | |||
8980 | // splat. So a constant BUILD_VECTOR here means the vector is built out of | |||
8981 | // different constants. | |||
8982 | if (V->isConstant()) | |||
8983 | return false; | |||
8984 | for (int i = 0, e = V->getNumOperands(); i < e; ++i) { | |||
8985 | if (V->getOperand(i).isUndef()) | |||
8986 | return false; | |||
8987 | // We want to expand nodes that represent load-and-splat even if the | |||
8988 | // loaded value is a floating point truncation or conversion to int. | |||
8989 | if (V->getOperand(i).getOpcode() == ISD::LOAD || | |||
8990 | (V->getOperand(i).getOpcode() == ISD::FP_ROUND && | |||
8991 | V->getOperand(i).getOperand(0).getOpcode() == ISD::LOAD) || | |||
8992 | (V->getOperand(i).getOpcode() == ISD::FP_TO_SINT && | |||
8993 | V->getOperand(i).getOperand(0).getOpcode() == ISD::LOAD) || | |||
8994 | (V->getOperand(i).getOpcode() == ISD::FP_TO_UINT && | |||
8995 | V->getOperand(i).getOperand(0).getOpcode() == ISD::LOAD)) | |||
8996 | IsLoad = true; | |||
8997 | // If the operands are different or the input is not a load and has more | |||
8998 | // uses than just this BV node, then it isn't a splat. | |||
8999 | if (V->getOperand(i) != Op0 || | |||
9000 | (!IsLoad && !V->isOnlyUserOf(V->getOperand(i).getNode()))) | |||
9001 | IsSplat = false; | |||
9002 | } | |||
9003 | return !(IsSplat && IsLoad); | |||
9004 | } | |||
9005 | ||||
9006 | // Lower BITCAST(f128, (build_pair i64, i64)) to BUILD_FP128. | |||
9007 | SDValue PPCTargetLowering::LowerBITCAST(SDValue Op, SelectionDAG &DAG) const { | |||
9008 | ||||
9009 | SDLoc dl(Op); | |||
9010 | SDValue Op0 = Op->getOperand(0); | |||
9011 | ||||
9012 | if ((Op.getValueType() != MVT::f128) || | |||
9013 | (Op0.getOpcode() != ISD::BUILD_PAIR) || | |||
9014 | (Op0.getOperand(0).getValueType() != MVT::i64) || | |||
9015 | (Op0.getOperand(1).getValueType() != MVT::i64)) | |||
9016 | return SDValue(); | |||
9017 | ||||
9018 | return DAG.getNode(PPCISD::BUILD_FP128, dl, MVT::f128, Op0.getOperand(0), | |||
9019 | Op0.getOperand(1)); | |||
9020 | } | |||
9021 | ||||
9022 | static const SDValue *getNormalLoadInput(const SDValue &Op, bool &IsPermuted) { | |||
9023 | const SDValue *InputLoad = &Op; | |||
9024 | if (InputLoad->getOpcode() == ISD::BITCAST) | |||
9025 | InputLoad = &InputLoad->getOperand(0); | |||
9026 | if (InputLoad->getOpcode() == ISD::SCALAR_TO_VECTOR || | |||
9027 | InputLoad->getOpcode() == PPCISD::SCALAR_TO_VECTOR_PERMUTED) { | |||
9028 | IsPermuted = InputLoad->getOpcode() == PPCISD::SCALAR_TO_VECTOR_PERMUTED; | |||
9029 | InputLoad = &InputLoad->getOperand(0); | |||
9030 | } | |||
9031 | if (InputLoad->getOpcode() != ISD::LOAD) | |||
9032 | return nullptr; | |||
9033 | LoadSDNode *LD = cast<LoadSDNode>(*InputLoad); | |||
9034 | return ISD::isNormalLoad(LD) ? InputLoad : nullptr; | |||
9035 | } | |||
9036 | ||||
9037 | // Convert the argument APFloat to a single precision APFloat if there is no | |||
9038 | // loss in information during the conversion to single precision APFloat and the | |||
9039 | // resulting number is not a denormal number. Return true if successful. | |||
9040 | bool llvm::convertToNonDenormSingle(APFloat &ArgAPFloat) { | |||
9041 | APFloat APFloatToConvert = ArgAPFloat; | |||
9042 | bool LosesInfo = true; | |||
9043 | APFloatToConvert.convert(APFloat::IEEEsingle(), APFloat::rmNearestTiesToEven, | |||
9044 | &LosesInfo); | |||
9045 | bool Success = (!LosesInfo && !APFloatToConvert.isDenormal()); | |||
9046 | if (Success) | |||
9047 | ArgAPFloat = APFloatToConvert; | |||
9048 | return Success; | |||
9049 | } | |||
9050 | ||||
9051 | // Bitcast the argument APInt to a double and convert it to a single precision | |||
9052 | // APFloat, bitcast the APFloat to an APInt and assign it to the original | |||
9053 | // argument if there is no loss in information during the conversion from | |||
9054 | // double to single precision APFloat and the resulting number is not a denormal | |||
9055 | // number. Return true if successful. | |||
9056 | bool llvm::convertToNonDenormSingle(APInt &ArgAPInt) { | |||
9057 | double DpValue = ArgAPInt.bitsToDouble(); | |||
9058 | APFloat APFloatDp(DpValue); | |||
9059 | bool Success = convertToNonDenormSingle(APFloatDp); | |||
9060 | if (Success) | |||
9061 | ArgAPInt = APFloatDp.bitcastToAPInt(); | |||
9062 | return Success; | |||
9063 | } | |||
9064 | ||||
9065 | // Nondestructive check for convertTonNonDenormSingle. | |||
9066 | bool llvm::checkConvertToNonDenormSingle(APFloat &ArgAPFloat) { | |||
9067 | // Only convert if it loses info, since XXSPLTIDP should | |||
9068 | // handle the other case. | |||
9069 | APFloat APFloatToConvert = ArgAPFloat; | |||
9070 | bool LosesInfo = true; | |||
9071 | APFloatToConvert.convert(APFloat::IEEEsingle(), APFloat::rmNearestTiesToEven, | |||
9072 | &LosesInfo); | |||
9073 | ||||
9074 | return (!LosesInfo && !APFloatToConvert.isDenormal()); | |||
9075 | } | |||
9076 | ||||
9077 | // If this is a case we can't handle, return null and let the default | |||
9078 | // expansion code take care of it. If we CAN select this case, and if it | |||
9079 | // selects to a single instruction, return Op. Otherwise, if we can codegen | |||
9080 | // this case more efficiently than a constant pool load, lower it to the | |||
9081 | // sequence of ops that should be used. | |||
9082 | SDValue PPCTargetLowering::LowerBUILD_VECTOR(SDValue Op, | |||
9083 | SelectionDAG &DAG) const { | |||
9084 | SDLoc dl(Op); | |||
9085 | BuildVectorSDNode *BVN = dyn_cast<BuildVectorSDNode>(Op.getNode()); | |||
9086 | assert(BVN && "Expected a BuildVectorSDNode in LowerBUILD_VECTOR")(static_cast <bool> (BVN && "Expected a BuildVectorSDNode in LowerBUILD_VECTOR" ) ? void (0) : __assert_fail ("BVN && \"Expected a BuildVectorSDNode in LowerBUILD_VECTOR\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9086, __extension__ __PRETTY_FUNCTION__)); | |||
9087 | ||||
9088 | // Check if this is a splat of a constant value. | |||
9089 | APInt APSplatBits, APSplatUndef; | |||
9090 | unsigned SplatBitSize; | |||
9091 | bool HasAnyUndefs; | |||
9092 | bool BVNIsConstantSplat = | |||
9093 | BVN->isConstantSplat(APSplatBits, APSplatUndef, SplatBitSize, | |||
9094 | HasAnyUndefs, 0, !Subtarget.isLittleEndian()); | |||
9095 | ||||
9096 | // If it is a splat of a double, check if we can shrink it to a 32 bit | |||
9097 | // non-denormal float which when converted back to double gives us the same | |||
9098 | // double. This is to exploit the XXSPLTIDP instruction. | |||
9099 | // If we lose precision, we use XXSPLTI32DX. | |||
9100 | if (BVNIsConstantSplat && (SplatBitSize == 64) && | |||
9101 | Subtarget.hasPrefixInstrs()) { | |||
9102 | // Check the type first to short-circuit so we don't modify APSplatBits if | |||
9103 | // this block isn't executed. | |||
9104 | if ((Op->getValueType(0) == MVT::v2f64) && | |||
9105 | convertToNonDenormSingle(APSplatBits)) { | |||
9106 | SDValue SplatNode = DAG.getNode( | |||
9107 | PPCISD::XXSPLTI_SP_TO_DP, dl, MVT::v2f64, | |||
9108 | DAG.getTargetConstant(APSplatBits.getZExtValue(), dl, MVT::i32)); | |||
9109 | return DAG.getBitcast(Op.getValueType(), SplatNode); | |||
9110 | } else { | |||
9111 | // We may lose precision, so we have to use XXSPLTI32DX. | |||
9112 | ||||
9113 | uint32_t Hi = | |||
9114 | (uint32_t)((APSplatBits.getZExtValue() & 0xFFFFFFFF00000000LL) >> 32); | |||
9115 | uint32_t Lo = | |||
9116 | (uint32_t)(APSplatBits.getZExtValue() & 0xFFFFFFFF); | |||
9117 | SDValue SplatNode = DAG.getUNDEF(MVT::v2i64); | |||
9118 | ||||
9119 | if (!Hi || !Lo) | |||
9120 | // If either load is 0, then we should generate XXLXOR to set to 0. | |||
9121 | SplatNode = DAG.getTargetConstant(0, dl, MVT::v2i64); | |||
9122 | ||||
9123 | if (Hi) | |||
9124 | SplatNode = DAG.getNode( | |||
9125 | PPCISD::XXSPLTI32DX, dl, MVT::v2i64, SplatNode, | |||
9126 | DAG.getTargetConstant(0, dl, MVT::i32), | |||
9127 | DAG.getTargetConstant(Hi, dl, MVT::i32)); | |||
9128 | ||||
9129 | if (Lo) | |||
9130 | SplatNode = | |||
9131 | DAG.getNode(PPCISD::XXSPLTI32DX, dl, MVT::v2i64, SplatNode, | |||
9132 | DAG.getTargetConstant(1, dl, MVT::i32), | |||
9133 | DAG.getTargetConstant(Lo, dl, MVT::i32)); | |||
9134 | ||||
9135 | return DAG.getBitcast(Op.getValueType(), SplatNode); | |||
9136 | } | |||
9137 | } | |||
9138 | ||||
9139 | if (!BVNIsConstantSplat || SplatBitSize > 32) { | |||
9140 | ||||
9141 | bool IsPermutedLoad = false; | |||
9142 | const SDValue *InputLoad = | |||
9143 | getNormalLoadInput(Op.getOperand(0), IsPermutedLoad); | |||
9144 | // Handle load-and-splat patterns as we have instructions that will do this | |||
9145 | // in one go. | |||
9146 | if (InputLoad && DAG.isSplatValue(Op, true)) { | |||
9147 | LoadSDNode *LD = cast<LoadSDNode>(*InputLoad); | |||
9148 | ||||
9149 | // We have handling for 4 and 8 byte elements. | |||
9150 | unsigned ElementSize = LD->getMemoryVT().getScalarSizeInBits(); | |||
9151 | ||||
9152 | // Checking for a single use of this load, we have to check for vector | |||
9153 | // width (128 bits) / ElementSize uses (since each operand of the | |||
9154 | // BUILD_VECTOR is a separate use of the value. | |||
9155 | unsigned NumUsesOfInputLD = 128 / ElementSize; | |||
9156 | for (SDValue BVInOp : Op->ops()) | |||
9157 | if (BVInOp.isUndef()) | |||
9158 | NumUsesOfInputLD--; | |||
9159 | assert(NumUsesOfInputLD > 0 && "No uses of input LD of a build_vector?")(static_cast <bool> (NumUsesOfInputLD > 0 && "No uses of input LD of a build_vector?") ? void (0) : __assert_fail ("NumUsesOfInputLD > 0 && \"No uses of input LD of a build_vector?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9159, __extension__ __PRETTY_FUNCTION__)); | |||
9160 | if (InputLoad->getNode()->hasNUsesOfValue(NumUsesOfInputLD, 0) && | |||
9161 | ((Subtarget.hasVSX() && ElementSize == 64) || | |||
9162 | (Subtarget.hasP9Vector() && ElementSize == 32))) { | |||
9163 | SDValue Ops[] = { | |||
9164 | LD->getChain(), // Chain | |||
9165 | LD->getBasePtr(), // Ptr | |||
9166 | DAG.getValueType(Op.getValueType()) // VT | |||
9167 | }; | |||
9168 | SDValue LdSplt = DAG.getMemIntrinsicNode( | |||
9169 | PPCISD::LD_SPLAT, dl, DAG.getVTList(Op.getValueType(), MVT::Other), | |||
9170 | Ops, LD->getMemoryVT(), LD->getMemOperand()); | |||
9171 | // Replace all uses of the output chain of the original load with the | |||
9172 | // output chain of the new load. | |||
9173 | DAG.ReplaceAllUsesOfValueWith(InputLoad->getValue(1), | |||
9174 | LdSplt.getValue(1)); | |||
9175 | return LdSplt; | |||
9176 | } | |||
9177 | } | |||
9178 | ||||
9179 | // In 64BIT mode BUILD_VECTOR nodes that are not constant splats of up to | |||
9180 | // 32-bits can be lowered to VSX instructions under certain conditions. | |||
9181 | // Without VSX, there is no pattern more efficient than expanding the node. | |||
9182 | if (Subtarget.hasVSX() && Subtarget.isPPC64() && | |||
9183 | haveEfficientBuildVectorPattern(BVN, Subtarget.hasDirectMove(), | |||
9184 | Subtarget.hasP8Vector())) | |||
9185 | return Op; | |||
9186 | return SDValue(); | |||
9187 | } | |||
9188 | ||||
9189 | uint64_t SplatBits = APSplatBits.getZExtValue(); | |||
9190 | uint64_t SplatUndef = APSplatUndef.getZExtValue(); | |||
9191 | unsigned SplatSize = SplatBitSize / 8; | |||
9192 | ||||
9193 | // First, handle single instruction cases. | |||
9194 | ||||
9195 | // All zeros? | |||
9196 | if (SplatBits == 0) { | |||
9197 | // Canonicalize all zero vectors to be v4i32. | |||
9198 | if (Op.getValueType() != MVT::v4i32 || HasAnyUndefs) { | |||
9199 | SDValue Z = DAG.getConstant(0, dl, MVT::v4i32); | |||
9200 | Op = DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), Z); | |||
9201 | } | |||
9202 | return Op; | |||
9203 | } | |||
9204 | ||||
9205 | // We have XXSPLTIW for constant splats four bytes wide. | |||
9206 | // Given vector length is a multiple of 4, 2-byte splats can be replaced | |||
9207 | // with 4-byte splats. We replicate the SplatBits in case of 2-byte splat to | |||
9208 | // make a 4-byte splat element. For example: 2-byte splat of 0xABAB can be | |||
9209 | // turned into a 4-byte splat of 0xABABABAB. | |||
9210 | if (Subtarget.hasPrefixInstrs() && SplatSize == 2) | |||
9211 | return getCanonicalConstSplat(SplatBits | (SplatBits << 16), SplatSize * 2, | |||
9212 | Op.getValueType(), DAG, dl); | |||
9213 | ||||
9214 | if (Subtarget.hasPrefixInstrs() && SplatSize == 4) | |||
9215 | return getCanonicalConstSplat(SplatBits, SplatSize, Op.getValueType(), DAG, | |||
9216 | dl); | |||
9217 | ||||
9218 | // We have XXSPLTIB for constant splats one byte wide. | |||
9219 | if (Subtarget.hasP9Vector() && SplatSize == 1) | |||
9220 | return getCanonicalConstSplat(SplatBits, SplatSize, Op.getValueType(), DAG, | |||
9221 | dl); | |||
9222 | ||||
9223 | // If the sign extended value is in the range [-16,15], use VSPLTI[bhw]. | |||
9224 | int32_t SextVal= (int32_t(SplatBits << (32-SplatBitSize)) >> | |||
9225 | (32-SplatBitSize)); | |||
9226 | if (SextVal >= -16 && SextVal <= 15) | |||
9227 | return getCanonicalConstSplat(SextVal, SplatSize, Op.getValueType(), DAG, | |||
9228 | dl); | |||
9229 | ||||
9230 | // Two instruction sequences. | |||
9231 | ||||
9232 | // If this value is in the range [-32,30] and is even, use: | |||
9233 | // VSPLTI[bhw](val/2) + VSPLTI[bhw](val/2) | |||
9234 | // If this value is in the range [17,31] and is odd, use: | |||
9235 | // VSPLTI[bhw](val-16) - VSPLTI[bhw](-16) | |||
9236 | // If this value is in the range [-31,-17] and is odd, use: | |||
9237 | // VSPLTI[bhw](val+16) + VSPLTI[bhw](-16) | |||
9238 | // Note the last two are three-instruction sequences. | |||
9239 | if (SextVal >= -32 && SextVal <= 31) { | |||
9240 | // To avoid having these optimizations undone by constant folding, | |||
9241 | // we convert to a pseudo that will be expanded later into one of | |||
9242 | // the above forms. | |||
9243 | SDValue Elt = DAG.getConstant(SextVal, dl, MVT::i32); | |||
9244 | EVT VT = (SplatSize == 1 ? MVT::v16i8 : | |||
9245 | (SplatSize == 2 ? MVT::v8i16 : MVT::v4i32)); | |||
9246 | SDValue EltSize = DAG.getConstant(SplatSize, dl, MVT::i32); | |||
9247 | SDValue RetVal = DAG.getNode(PPCISD::VADD_SPLAT, dl, VT, Elt, EltSize); | |||
9248 | if (VT == Op.getValueType()) | |||
9249 | return RetVal; | |||
9250 | else | |||
9251 | return DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), RetVal); | |||
9252 | } | |||
9253 | ||||
9254 | // If this is 0x8000_0000 x 4, turn into vspltisw + vslw. If it is | |||
9255 | // 0x7FFF_FFFF x 4, turn it into not(0x8000_0000). This is important | |||
9256 | // for fneg/fabs. | |||
9257 | if (SplatSize == 4 && SplatBits == (0x7FFFFFFF&~SplatUndef)) { | |||
9258 | // Make -1 and vspltisw -1: | |||
9259 | SDValue OnesV = getCanonicalConstSplat(-1, 4, MVT::v4i32, DAG, dl); | |||
9260 | ||||
9261 | // Make the VSLW intrinsic, computing 0x8000_0000. | |||
9262 | SDValue Res = BuildIntrinsicOp(Intrinsic::ppc_altivec_vslw, OnesV, | |||
9263 | OnesV, DAG, dl); | |||
9264 | ||||
9265 | // xor by OnesV to invert it. | |||
9266 | Res = DAG.getNode(ISD::XOR, dl, MVT::v4i32, Res, OnesV); | |||
9267 | return DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), Res); | |||
9268 | } | |||
9269 | ||||
9270 | // Check to see if this is a wide variety of vsplti*, binop self cases. | |||
9271 | static const signed char SplatCsts[] = { | |||
9272 | -1, 1, -2, 2, -3, 3, -4, 4, -5, 5, -6, 6, -7, 7, | |||
9273 | -8, 8, -9, 9, -10, 10, -11, 11, -12, 12, -13, 13, 14, -14, 15, -15, -16 | |||
9274 | }; | |||
9275 | ||||
9276 | for (unsigned idx = 0; idx < array_lengthof(SplatCsts); ++idx) { | |||
9277 | // Indirect through the SplatCsts array so that we favor 'vsplti -1' for | |||
9278 | // cases which are ambiguous (e.g. formation of 0x8000_0000). 'vsplti -1' | |||
9279 | int i = SplatCsts[idx]; | |||
9280 | ||||
9281 | // Figure out what shift amount will be used by altivec if shifted by i in | |||
9282 | // this splat size. | |||
9283 | unsigned TypeShiftAmt = i & (SplatBitSize-1); | |||
9284 | ||||
9285 | // vsplti + shl self. | |||
9286 | if (SextVal == (int)((unsigned)i << TypeShiftAmt)) { | |||
9287 | SDValue Res = getCanonicalConstSplat(i, SplatSize, MVT::Other, DAG, dl); | |||
9288 | static const unsigned IIDs[] = { // Intrinsic to use for each size. | |||
9289 | Intrinsic::ppc_altivec_vslb, Intrinsic::ppc_altivec_vslh, 0, | |||
9290 | Intrinsic::ppc_altivec_vslw | |||
9291 | }; | |||
9292 | Res = BuildIntrinsicOp(IIDs[SplatSize-1], Res, Res, DAG, dl); | |||
9293 | return DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), Res); | |||
9294 | } | |||
9295 | ||||
9296 | // vsplti + srl self. | |||
9297 | if (SextVal == (int)((unsigned)i >> TypeShiftAmt)) { | |||
9298 | SDValue Res = getCanonicalConstSplat(i, SplatSize, MVT::Other, DAG, dl); | |||
9299 | static const unsigned IIDs[] = { // Intrinsic to use for each size. | |||
9300 | Intrinsic::ppc_altivec_vsrb, Intrinsic::ppc_altivec_vsrh, 0, | |||
9301 | Intrinsic::ppc_altivec_vsrw | |||
9302 | }; | |||
9303 | Res = BuildIntrinsicOp(IIDs[SplatSize-1], Res, Res, DAG, dl); | |||
9304 | return DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), Res); | |||
9305 | } | |||
9306 | ||||
9307 | // vsplti + rol self. | |||
9308 | if (SextVal == (int)(((unsigned)i << TypeShiftAmt) | | |||
9309 | ((unsigned)i >> (SplatBitSize-TypeShiftAmt)))) { | |||
9310 | SDValue Res = getCanonicalConstSplat(i, SplatSize, MVT::Other, DAG, dl); | |||
9311 | static const unsigned IIDs[] = { // Intrinsic to use for each size. | |||
9312 | Intrinsic::ppc_altivec_vrlb, Intrinsic::ppc_altivec_vrlh, 0, | |||
9313 | Intrinsic::ppc_altivec_vrlw | |||
9314 | }; | |||
9315 | Res = BuildIntrinsicOp(IIDs[SplatSize-1], Res, Res, DAG, dl); | |||
9316 | return DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), Res); | |||
9317 | } | |||
9318 | ||||
9319 | // t = vsplti c, result = vsldoi t, t, 1 | |||
9320 | if (SextVal == (int)(((unsigned)i << 8) | (i < 0 ? 0xFF : 0))) { | |||
9321 | SDValue T = getCanonicalConstSplat(i, SplatSize, MVT::v16i8, DAG, dl); | |||
9322 | unsigned Amt = Subtarget.isLittleEndian() ? 15 : 1; | |||
9323 | return BuildVSLDOI(T, T, Amt, Op.getValueType(), DAG, dl); | |||
9324 | } | |||
9325 | // t = vsplti c, result = vsldoi t, t, 2 | |||
9326 | if (SextVal == (int)(((unsigned)i << 16) | (i < 0 ? 0xFFFF : 0))) { | |||
9327 | SDValue T = getCanonicalConstSplat(i, SplatSize, MVT::v16i8, DAG, dl); | |||
9328 | unsigned Amt = Subtarget.isLittleEndian() ? 14 : 2; | |||
9329 | return BuildVSLDOI(T, T, Amt, Op.getValueType(), DAG, dl); | |||
9330 | } | |||
9331 | // t = vsplti c, result = vsldoi t, t, 3 | |||
9332 | if (SextVal == (int)(((unsigned)i << 24) | (i < 0 ? 0xFFFFFF : 0))) { | |||
9333 | SDValue T = getCanonicalConstSplat(i, SplatSize, MVT::v16i8, DAG, dl); | |||
9334 | unsigned Amt = Subtarget.isLittleEndian() ? 13 : 3; | |||
9335 | return BuildVSLDOI(T, T, Amt, Op.getValueType(), DAG, dl); | |||
9336 | } | |||
9337 | } | |||
9338 | ||||
9339 | return SDValue(); | |||
9340 | } | |||
9341 | ||||
9342 | /// GeneratePerfectShuffle - Given an entry in the perfect-shuffle table, emit | |||
9343 | /// the specified operations to build the shuffle. | |||
9344 | static SDValue GeneratePerfectShuffle(unsigned PFEntry, SDValue LHS, | |||
9345 | SDValue RHS, SelectionDAG &DAG, | |||
9346 | const SDLoc &dl) { | |||
9347 | unsigned OpNum = (PFEntry >> 26) & 0x0F; | |||
9348 | unsigned LHSID = (PFEntry >> 13) & ((1 << 13)-1); | |||
9349 | unsigned RHSID = (PFEntry >> 0) & ((1 << 13)-1); | |||
9350 | ||||
9351 | enum { | |||
9352 | OP_COPY = 0, // Copy, used for things like <u,u,u,3> to say it is <0,1,2,3> | |||
9353 | OP_VMRGHW, | |||
9354 | OP_VMRGLW, | |||
9355 | OP_VSPLTISW0, | |||
9356 | OP_VSPLTISW1, | |||
9357 | OP_VSPLTISW2, | |||
9358 | OP_VSPLTISW3, | |||
9359 | OP_VSLDOI4, | |||
9360 | OP_VSLDOI8, | |||
9361 | OP_VSLDOI12 | |||
9362 | }; | |||
9363 | ||||
9364 | if (OpNum == OP_COPY) { | |||
9365 | if (LHSID == (1*9+2)*9+3) return LHS; | |||
9366 | assert(LHSID == ((4*9+5)*9+6)*9+7 && "Illegal OP_COPY!")(static_cast <bool> (LHSID == ((4*9+5)*9+6)*9+7 && "Illegal OP_COPY!") ? void (0) : __assert_fail ("LHSID == ((4*9+5)*9+6)*9+7 && \"Illegal OP_COPY!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9366, __extension__ __PRETTY_FUNCTION__)); | |||
9367 | return RHS; | |||
9368 | } | |||
9369 | ||||
9370 | SDValue OpLHS, OpRHS; | |||
9371 | OpLHS = GeneratePerfectShuffle(PerfectShuffleTable[LHSID], LHS, RHS, DAG, dl); | |||
9372 | OpRHS = GeneratePerfectShuffle(PerfectShuffleTable[RHSID], LHS, RHS, DAG, dl); | |||
9373 | ||||
9374 | int ShufIdxs[16]; | |||
9375 | switch (OpNum) { | |||
9376 | default: llvm_unreachable("Unknown i32 permute!")::llvm::llvm_unreachable_internal("Unknown i32 permute!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9376); | |||
9377 | case OP_VMRGHW: | |||
9378 | ShufIdxs[ 0] = 0; ShufIdxs[ 1] = 1; ShufIdxs[ 2] = 2; ShufIdxs[ 3] = 3; | |||
9379 | ShufIdxs[ 4] = 16; ShufIdxs[ 5] = 17; ShufIdxs[ 6] = 18; ShufIdxs[ 7] = 19; | |||
9380 | ShufIdxs[ 8] = 4; ShufIdxs[ 9] = 5; ShufIdxs[10] = 6; ShufIdxs[11] = 7; | |||
9381 | ShufIdxs[12] = 20; ShufIdxs[13] = 21; ShufIdxs[14] = 22; ShufIdxs[15] = 23; | |||
9382 | break; | |||
9383 | case OP_VMRGLW: | |||
9384 | ShufIdxs[ 0] = 8; ShufIdxs[ 1] = 9; ShufIdxs[ 2] = 10; ShufIdxs[ 3] = 11; | |||
9385 | ShufIdxs[ 4] = 24; ShufIdxs[ 5] = 25; ShufIdxs[ 6] = 26; ShufIdxs[ 7] = 27; | |||
9386 | ShufIdxs[ 8] = 12; ShufIdxs[ 9] = 13; ShufIdxs[10] = 14; ShufIdxs[11] = 15; | |||
9387 | ShufIdxs[12] = 28; ShufIdxs[13] = 29; ShufIdxs[14] = 30; ShufIdxs[15] = 31; | |||
9388 | break; | |||
9389 | case OP_VSPLTISW0: | |||
9390 | for (unsigned i = 0; i != 16; ++i) | |||
9391 | ShufIdxs[i] = (i&3)+0; | |||
9392 | break; | |||
9393 | case OP_VSPLTISW1: | |||
9394 | for (unsigned i = 0; i != 16; ++i) | |||
9395 | ShufIdxs[i] = (i&3)+4; | |||
9396 | break; | |||
9397 | case OP_VSPLTISW2: | |||
9398 | for (unsigned i = 0; i != 16; ++i) | |||
9399 | ShufIdxs[i] = (i&3)+8; | |||
9400 | break; | |||
9401 | case OP_VSPLTISW3: | |||
9402 | for (unsigned i = 0; i != 16; ++i) | |||
9403 | ShufIdxs[i] = (i&3)+12; | |||
9404 | break; | |||
9405 | case OP_VSLDOI4: | |||
9406 | return BuildVSLDOI(OpLHS, OpRHS, 4, OpLHS.getValueType(), DAG, dl); | |||
9407 | case OP_VSLDOI8: | |||
9408 | return BuildVSLDOI(OpLHS, OpRHS, 8, OpLHS.getValueType(), DAG, dl); | |||
9409 | case OP_VSLDOI12: | |||
9410 | return BuildVSLDOI(OpLHS, OpRHS, 12, OpLHS.getValueType(), DAG, dl); | |||
9411 | } | |||
9412 | EVT VT = OpLHS.getValueType(); | |||
9413 | OpLHS = DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, OpLHS); | |||
9414 | OpRHS = DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, OpRHS); | |||
9415 | SDValue T = DAG.getVectorShuffle(MVT::v16i8, dl, OpLHS, OpRHS, ShufIdxs); | |||
9416 | return DAG.getNode(ISD::BITCAST, dl, VT, T); | |||
9417 | } | |||
9418 | ||||
9419 | /// lowerToVINSERTB - Return the SDValue if this VECTOR_SHUFFLE can be handled | |||
9420 | /// by the VINSERTB instruction introduced in ISA 3.0, else just return default | |||
9421 | /// SDValue. | |||
9422 | SDValue PPCTargetLowering::lowerToVINSERTB(ShuffleVectorSDNode *N, | |||
9423 | SelectionDAG &DAG) const { | |||
9424 | const unsigned BytesInVector = 16; | |||
9425 | bool IsLE = Subtarget.isLittleEndian(); | |||
9426 | SDLoc dl(N); | |||
9427 | SDValue V1 = N->getOperand(0); | |||
9428 | SDValue V2 = N->getOperand(1); | |||
9429 | unsigned ShiftElts = 0, InsertAtByte = 0; | |||
9430 | bool Swap = false; | |||
9431 | ||||
9432 | // Shifts required to get the byte we want at element 7. | |||
9433 | unsigned LittleEndianShifts[] = {8, 7, 6, 5, 4, 3, 2, 1, | |||
9434 | 0, 15, 14, 13, 12, 11, 10, 9}; | |||
9435 | unsigned BigEndianShifts[] = {9, 10, 11, 12, 13, 14, 15, 0, | |||
9436 | 1, 2, 3, 4, 5, 6, 7, 8}; | |||
9437 | ||||
9438 | ArrayRef<int> Mask = N->getMask(); | |||
9439 | int OriginalOrder[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15}; | |||
9440 | ||||
9441 | // For each mask element, find out if we're just inserting something | |||
9442 | // from V2 into V1 or vice versa. | |||
9443 | // Possible permutations inserting an element from V2 into V1: | |||
9444 | // X, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 | |||
9445 | // 0, X, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 | |||
9446 | // ... | |||
9447 | // 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, X | |||
9448 | // Inserting from V1 into V2 will be similar, except mask range will be | |||
9449 | // [16,31]. | |||
9450 | ||||
9451 | bool FoundCandidate = false; | |||
9452 | // If both vector operands for the shuffle are the same vector, the mask | |||
9453 | // will contain only elements from the first one and the second one will be | |||
9454 | // undef. | |||
9455 | unsigned VINSERTBSrcElem = IsLE ? 8 : 7; | |||
9456 | // Go through the mask of half-words to find an element that's being moved | |||
9457 | // from one vector to the other. | |||
9458 | for (unsigned i = 0; i < BytesInVector; ++i) { | |||
9459 | unsigned CurrentElement = Mask[i]; | |||
9460 | // If 2nd operand is undefined, we should only look for element 7 in the | |||
9461 | // Mask. | |||
9462 | if (V2.isUndef() && CurrentElement != VINSERTBSrcElem) | |||
9463 | continue; | |||
9464 | ||||
9465 | bool OtherElementsInOrder = true; | |||
9466 | // Examine the other elements in the Mask to see if they're in original | |||
9467 | // order. | |||
9468 | for (unsigned j = 0; j < BytesInVector; ++j) { | |||
9469 | if (j == i) | |||
9470 | continue; | |||
9471 | // If CurrentElement is from V1 [0,15], then we the rest of the Mask to be | |||
9472 | // from V2 [16,31] and vice versa. Unless the 2nd operand is undefined, | |||
9473 | // in which we always assume we're always picking from the 1st operand. | |||
9474 | int MaskOffset = | |||
9475 | (!V2.isUndef() && CurrentElement < BytesInVector) ? BytesInVector : 0; | |||
9476 | if (Mask[j] != OriginalOrder[j] + MaskOffset) { | |||
9477 | OtherElementsInOrder = false; | |||
9478 | break; | |||
9479 | } | |||
9480 | } | |||
9481 | // If other elements are in original order, we record the number of shifts | |||
9482 | // we need to get the element we want into element 7. Also record which byte | |||
9483 | // in the vector we should insert into. | |||
9484 | if (OtherElementsInOrder) { | |||
9485 | // If 2nd operand is undefined, we assume no shifts and no swapping. | |||
9486 | if (V2.isUndef()) { | |||
9487 | ShiftElts = 0; | |||
9488 | Swap = false; | |||
9489 | } else { | |||
9490 | // Only need the last 4-bits for shifts because operands will be swapped if CurrentElement is >= 2^4. | |||
9491 | ShiftElts = IsLE ? LittleEndianShifts[CurrentElement & 0xF] | |||
9492 | : BigEndianShifts[CurrentElement & 0xF]; | |||
9493 | Swap = CurrentElement < BytesInVector; | |||
9494 | } | |||
9495 | InsertAtByte = IsLE ? BytesInVector - (i + 1) : i; | |||
9496 | FoundCandidate = true; | |||
9497 | break; | |||
9498 | } | |||
9499 | } | |||
9500 | ||||
9501 | if (!FoundCandidate) | |||
9502 | return SDValue(); | |||
9503 | ||||
9504 | // Candidate found, construct the proper SDAG sequence with VINSERTB, | |||
9505 | // optionally with VECSHL if shift is required. | |||
9506 | if (Swap) | |||
9507 | std::swap(V1, V2); | |||
9508 | if (V2.isUndef()) | |||
9509 | V2 = V1; | |||
9510 | if (ShiftElts) { | |||
9511 | SDValue Shl = DAG.getNode(PPCISD::VECSHL, dl, MVT::v16i8, V2, V2, | |||
9512 | DAG.getConstant(ShiftElts, dl, MVT::i32)); | |||
9513 | return DAG.getNode(PPCISD::VECINSERT, dl, MVT::v16i8, V1, Shl, | |||
9514 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
9515 | } | |||
9516 | return DAG.getNode(PPCISD::VECINSERT, dl, MVT::v16i8, V1, V2, | |||
9517 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
9518 | } | |||
9519 | ||||
9520 | /// lowerToVINSERTH - Return the SDValue if this VECTOR_SHUFFLE can be handled | |||
9521 | /// by the VINSERTH instruction introduced in ISA 3.0, else just return default | |||
9522 | /// SDValue. | |||
9523 | SDValue PPCTargetLowering::lowerToVINSERTH(ShuffleVectorSDNode *N, | |||
9524 | SelectionDAG &DAG) const { | |||
9525 | const unsigned NumHalfWords = 8; | |||
9526 | const unsigned BytesInVector = NumHalfWords * 2; | |||
9527 | // Check that the shuffle is on half-words. | |||
9528 | if (!isNByteElemShuffleMask(N, 2, 1)) | |||
9529 | return SDValue(); | |||
9530 | ||||
9531 | bool IsLE = Subtarget.isLittleEndian(); | |||
9532 | SDLoc dl(N); | |||
9533 | SDValue V1 = N->getOperand(0); | |||
9534 | SDValue V2 = N->getOperand(1); | |||
9535 | unsigned ShiftElts = 0, InsertAtByte = 0; | |||
9536 | bool Swap = false; | |||
9537 | ||||
9538 | // Shifts required to get the half-word we want at element 3. | |||
9539 | unsigned LittleEndianShifts[] = {4, 3, 2, 1, 0, 7, 6, 5}; | |||
9540 | unsigned BigEndianShifts[] = {5, 6, 7, 0, 1, 2, 3, 4}; | |||
9541 | ||||
9542 | uint32_t Mask = 0; | |||
9543 | uint32_t OriginalOrderLow = 0x1234567; | |||
9544 | uint32_t OriginalOrderHigh = 0x89ABCDEF; | |||
9545 | // Now we look at mask elements 0,2,4,6,8,10,12,14. Pack the mask into a | |||
9546 | // 32-bit space, only need 4-bit nibbles per element. | |||
9547 | for (unsigned i = 0; i < NumHalfWords; ++i) { | |||
9548 | unsigned MaskShift = (NumHalfWords - 1 - i) * 4; | |||
9549 | Mask |= ((uint32_t)(N->getMaskElt(i * 2) / 2) << MaskShift); | |||
9550 | } | |||
9551 | ||||
9552 | // For each mask element, find out if we're just inserting something | |||
9553 | // from V2 into V1 or vice versa. Possible permutations inserting an element | |||
9554 | // from V2 into V1: | |||
9555 | // X, 1, 2, 3, 4, 5, 6, 7 | |||
9556 | // 0, X, 2, 3, 4, 5, 6, 7 | |||
9557 | // 0, 1, X, 3, 4, 5, 6, 7 | |||
9558 | // 0, 1, 2, X, 4, 5, 6, 7 | |||
9559 | // 0, 1, 2, 3, X, 5, 6, 7 | |||
9560 | // 0, 1, 2, 3, 4, X, 6, 7 | |||
9561 | // 0, 1, 2, 3, 4, 5, X, 7 | |||
9562 | // 0, 1, 2, 3, 4, 5, 6, X | |||
9563 | // Inserting from V1 into V2 will be similar, except mask range will be [8,15]. | |||
9564 | ||||
9565 | bool FoundCandidate = false; | |||
9566 | // Go through the mask of half-words to find an element that's being moved | |||
9567 | // from one vector to the other. | |||
9568 | for (unsigned i = 0; i < NumHalfWords; ++i) { | |||
9569 | unsigned MaskShift = (NumHalfWords - 1 - i) * 4; | |||
9570 | uint32_t MaskOneElt = (Mask >> MaskShift) & 0xF; | |||
9571 | uint32_t MaskOtherElts = ~(0xF << MaskShift); | |||
9572 | uint32_t TargetOrder = 0x0; | |||
9573 | ||||
9574 | // If both vector operands for the shuffle are the same vector, the mask | |||
9575 | // will contain only elements from the first one and the second one will be | |||
9576 | // undef. | |||
9577 | if (V2.isUndef()) { | |||
9578 | ShiftElts = 0; | |||
9579 | unsigned VINSERTHSrcElem = IsLE ? 4 : 3; | |||
9580 | TargetOrder = OriginalOrderLow; | |||
9581 | Swap = false; | |||
9582 | // Skip if not the correct element or mask of other elements don't equal | |||
9583 | // to our expected order. | |||
9584 | if (MaskOneElt == VINSERTHSrcElem && | |||
9585 | (Mask & MaskOtherElts) == (TargetOrder & MaskOtherElts)) { | |||
9586 | InsertAtByte = IsLE ? BytesInVector - (i + 1) * 2 : i * 2; | |||
9587 | FoundCandidate = true; | |||
9588 | break; | |||
9589 | } | |||
9590 | } else { // If both operands are defined. | |||
9591 | // Target order is [8,15] if the current mask is between [0,7]. | |||
9592 | TargetOrder = | |||
9593 | (MaskOneElt < NumHalfWords) ? OriginalOrderHigh : OriginalOrderLow; | |||
9594 | // Skip if mask of other elements don't equal our expected order. | |||
9595 | if ((Mask & MaskOtherElts) == (TargetOrder & MaskOtherElts)) { | |||
9596 | // We only need the last 3 bits for the number of shifts. | |||
9597 | ShiftElts = IsLE ? LittleEndianShifts[MaskOneElt & 0x7] | |||
9598 | : BigEndianShifts[MaskOneElt & 0x7]; | |||
9599 | InsertAtByte = IsLE ? BytesInVector - (i + 1) * 2 : i * 2; | |||
9600 | Swap = MaskOneElt < NumHalfWords; | |||
9601 | FoundCandidate = true; | |||
9602 | break; | |||
9603 | } | |||
9604 | } | |||
9605 | } | |||
9606 | ||||
9607 | if (!FoundCandidate) | |||
9608 | return SDValue(); | |||
9609 | ||||
9610 | // Candidate found, construct the proper SDAG sequence with VINSERTH, | |||
9611 | // optionally with VECSHL if shift is required. | |||
9612 | if (Swap) | |||
9613 | std::swap(V1, V2); | |||
9614 | if (V2.isUndef()) | |||
9615 | V2 = V1; | |||
9616 | SDValue Conv1 = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, V1); | |||
9617 | if (ShiftElts) { | |||
9618 | // Double ShiftElts because we're left shifting on v16i8 type. | |||
9619 | SDValue Shl = DAG.getNode(PPCISD::VECSHL, dl, MVT::v16i8, V2, V2, | |||
9620 | DAG.getConstant(2 * ShiftElts, dl, MVT::i32)); | |||
9621 | SDValue Conv2 = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, Shl); | |||
9622 | SDValue Ins = DAG.getNode(PPCISD::VECINSERT, dl, MVT::v8i16, Conv1, Conv2, | |||
9623 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
9624 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Ins); | |||
9625 | } | |||
9626 | SDValue Conv2 = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, V2); | |||
9627 | SDValue Ins = DAG.getNode(PPCISD::VECINSERT, dl, MVT::v8i16, Conv1, Conv2, | |||
9628 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
9629 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Ins); | |||
9630 | } | |||
9631 | ||||
9632 | /// lowerToXXSPLTI32DX - Return the SDValue if this VECTOR_SHUFFLE can be | |||
9633 | /// handled by the XXSPLTI32DX instruction introduced in ISA 3.1, otherwise | |||
9634 | /// return the default SDValue. | |||
9635 | SDValue PPCTargetLowering::lowerToXXSPLTI32DX(ShuffleVectorSDNode *SVN, | |||
9636 | SelectionDAG &DAG) const { | |||
9637 | // The LHS and RHS may be bitcasts to v16i8 as we canonicalize shuffles | |||
9638 | // to v16i8. Peek through the bitcasts to get the actual operands. | |||
9639 | SDValue LHS = peekThroughBitcasts(SVN->getOperand(0)); | |||
9640 | SDValue RHS = peekThroughBitcasts(SVN->getOperand(1)); | |||
9641 | ||||
9642 | auto ShuffleMask = SVN->getMask(); | |||
9643 | SDValue VecShuffle(SVN, 0); | |||
9644 | SDLoc DL(SVN); | |||
9645 | ||||
9646 | // Check that we have a four byte shuffle. | |||
9647 | if (!isNByteElemShuffleMask(SVN, 4, 1)) | |||
9648 | return SDValue(); | |||
9649 | ||||
9650 | // Canonicalize the RHS being a BUILD_VECTOR when lowering to xxsplti32dx. | |||
9651 | if (RHS->getOpcode() != ISD::BUILD_VECTOR) { | |||
9652 | std::swap(LHS, RHS); | |||
9653 | VecShuffle = DAG.getCommutedVectorShuffle(*SVN); | |||
9654 | ShuffleMask = cast<ShuffleVectorSDNode>(VecShuffle)->getMask(); | |||
9655 | } | |||
9656 | ||||
9657 | // Ensure that the RHS is a vector of constants. | |||
9658 | BuildVectorSDNode *BVN = dyn_cast<BuildVectorSDNode>(RHS.getNode()); | |||
9659 | if (!BVN) | |||
9660 | return SDValue(); | |||
9661 | ||||
9662 | // Check if RHS is a splat of 4-bytes (or smaller). | |||
9663 | APInt APSplatValue, APSplatUndef; | |||
9664 | unsigned SplatBitSize; | |||
9665 | bool HasAnyUndefs; | |||
9666 | if (!BVN->isConstantSplat(APSplatValue, APSplatUndef, SplatBitSize, | |||
9667 | HasAnyUndefs, 0, !Subtarget.isLittleEndian()) || | |||
9668 | SplatBitSize > 32) | |||
9669 | return SDValue(); | |||
9670 | ||||
9671 | // Check that the shuffle mask matches the semantics of XXSPLTI32DX. | |||
9672 | // The instruction splats a constant C into two words of the source vector | |||
9673 | // producing { C, Unchanged, C, Unchanged } or { Unchanged, C, Unchanged, C }. | |||
9674 | // Thus we check that the shuffle mask is the equivalent of | |||
9675 | // <0, [4-7], 2, [4-7]> or <[4-7], 1, [4-7], 3> respectively. | |||
9676 | // Note: the check above of isNByteElemShuffleMask() ensures that the bytes | |||
9677 | // within each word are consecutive, so we only need to check the first byte. | |||
9678 | SDValue Index; | |||
9679 | bool IsLE = Subtarget.isLittleEndian(); | |||
9680 | if ((ShuffleMask[0] == 0 && ShuffleMask[8] == 8) && | |||
9681 | (ShuffleMask[4] % 4 == 0 && ShuffleMask[12] % 4 == 0 && | |||
9682 | ShuffleMask[4] > 15 && ShuffleMask[12] > 15)) | |||
9683 | Index = DAG.getTargetConstant(IsLE ? 0 : 1, DL, MVT::i32); | |||
9684 | else if ((ShuffleMask[4] == 4 && ShuffleMask[12] == 12) && | |||
9685 | (ShuffleMask[0] % 4 == 0 && ShuffleMask[8] % 4 == 0 && | |||
9686 | ShuffleMask[0] > 15 && ShuffleMask[8] > 15)) | |||
9687 | Index = DAG.getTargetConstant(IsLE ? 1 : 0, DL, MVT::i32); | |||
9688 | else | |||
9689 | return SDValue(); | |||
9690 | ||||
9691 | // If the splat is narrower than 32-bits, we need to get the 32-bit value | |||
9692 | // for XXSPLTI32DX. | |||
9693 | unsigned SplatVal = APSplatValue.getZExtValue(); | |||
9694 | for (; SplatBitSize < 32; SplatBitSize <<= 1) | |||
9695 | SplatVal |= (SplatVal << SplatBitSize); | |||
9696 | ||||
9697 | SDValue SplatNode = DAG.getNode( | |||
9698 | PPCISD::XXSPLTI32DX, DL, MVT::v2i64, DAG.getBitcast(MVT::v2i64, LHS), | |||
9699 | Index, DAG.getTargetConstant(SplatVal, DL, MVT::i32)); | |||
9700 | return DAG.getNode(ISD::BITCAST, DL, MVT::v16i8, SplatNode); | |||
9701 | } | |||
9702 | ||||
9703 | /// LowerROTL - Custom lowering for ROTL(v1i128) to vector_shuffle(v16i8). | |||
9704 | /// We lower ROTL(v1i128) to vector_shuffle(v16i8) only if shift amount is | |||
9705 | /// a multiple of 8. Otherwise convert it to a scalar rotation(i128) | |||
9706 | /// i.e (or (shl x, C1), (srl x, 128-C1)). | |||
9707 | SDValue PPCTargetLowering::LowerROTL(SDValue Op, SelectionDAG &DAG) const { | |||
9708 | assert(Op.getOpcode() == ISD::ROTL && "Should only be called for ISD::ROTL")(static_cast <bool> (Op.getOpcode() == ISD::ROTL && "Should only be called for ISD::ROTL") ? void (0) : __assert_fail ("Op.getOpcode() == ISD::ROTL && \"Should only be called for ISD::ROTL\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9708, __extension__ __PRETTY_FUNCTION__)); | |||
9709 | assert(Op.getValueType() == MVT::v1i128 &&(static_cast <bool> (Op.getValueType() == MVT::v1i128 && "Only set v1i128 as custom, other type shouldn't reach here!" ) ? void (0) : __assert_fail ("Op.getValueType() == MVT::v1i128 && \"Only set v1i128 as custom, other type shouldn't reach here!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9710, __extension__ __PRETTY_FUNCTION__)) | |||
9710 | "Only set v1i128 as custom, other type shouldn't reach here!")(static_cast <bool> (Op.getValueType() == MVT::v1i128 && "Only set v1i128 as custom, other type shouldn't reach here!" ) ? void (0) : __assert_fail ("Op.getValueType() == MVT::v1i128 && \"Only set v1i128 as custom, other type shouldn't reach here!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9710, __extension__ __PRETTY_FUNCTION__)); | |||
9711 | SDLoc dl(Op); | |||
9712 | SDValue N0 = peekThroughBitcasts(Op.getOperand(0)); | |||
9713 | SDValue N1 = peekThroughBitcasts(Op.getOperand(1)); | |||
9714 | unsigned SHLAmt = N1.getConstantOperandVal(0); | |||
9715 | if (SHLAmt % 8 == 0) { | |||
9716 | SmallVector<int, 16> Mask(16, 0); | |||
9717 | std::iota(Mask.begin(), Mask.end(), 0); | |||
9718 | std::rotate(Mask.begin(), Mask.begin() + SHLAmt / 8, Mask.end()); | |||
9719 | if (SDValue Shuffle = | |||
9720 | DAG.getVectorShuffle(MVT::v16i8, dl, DAG.getBitcast(MVT::v16i8, N0), | |||
9721 | DAG.getUNDEF(MVT::v16i8), Mask)) | |||
9722 | return DAG.getNode(ISD::BITCAST, dl, MVT::v1i128, Shuffle); | |||
9723 | } | |||
9724 | SDValue ArgVal = DAG.getBitcast(MVT::i128, N0); | |||
9725 | SDValue SHLOp = DAG.getNode(ISD::SHL, dl, MVT::i128, ArgVal, | |||
9726 | DAG.getConstant(SHLAmt, dl, MVT::i32)); | |||
9727 | SDValue SRLOp = DAG.getNode(ISD::SRL, dl, MVT::i128, ArgVal, | |||
9728 | DAG.getConstant(128 - SHLAmt, dl, MVT::i32)); | |||
9729 | SDValue OROp = DAG.getNode(ISD::OR, dl, MVT::i128, SHLOp, SRLOp); | |||
9730 | return DAG.getNode(ISD::BITCAST, dl, MVT::v1i128, OROp); | |||
9731 | } | |||
9732 | ||||
9733 | /// LowerVECTOR_SHUFFLE - Return the code we lower for VECTOR_SHUFFLE. If this | |||
9734 | /// is a shuffle we can handle in a single instruction, return it. Otherwise, | |||
9735 | /// return the code it can be lowered into. Worst case, it can always be | |||
9736 | /// lowered into a vperm. | |||
9737 | SDValue PPCTargetLowering::LowerVECTOR_SHUFFLE(SDValue Op, | |||
9738 | SelectionDAG &DAG) const { | |||
9739 | SDLoc dl(Op); | |||
9740 | SDValue V1 = Op.getOperand(0); | |||
9741 | SDValue V2 = Op.getOperand(1); | |||
9742 | ShuffleVectorSDNode *SVOp = cast<ShuffleVectorSDNode>(Op); | |||
9743 | ||||
9744 | // Any nodes that were combined in the target-independent combiner prior | |||
9745 | // to vector legalization will not be sent to the target combine. Try to | |||
9746 | // combine it here. | |||
9747 | if (SDValue NewShuffle = combineVectorShuffle(SVOp, DAG)) { | |||
9748 | if (!isa<ShuffleVectorSDNode>(NewShuffle)) | |||
9749 | return NewShuffle; | |||
9750 | Op = NewShuffle; | |||
9751 | SVOp = cast<ShuffleVectorSDNode>(Op); | |||
9752 | V1 = Op.getOperand(0); | |||
9753 | V2 = Op.getOperand(1); | |||
9754 | } | |||
9755 | EVT VT = Op.getValueType(); | |||
9756 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
9757 | ||||
9758 | unsigned ShiftElts, InsertAtByte; | |||
9759 | bool Swap = false; | |||
9760 | ||||
9761 | // If this is a load-and-splat, we can do that with a single instruction | |||
9762 | // in some cases. However if the load has multiple uses, we don't want to | |||
9763 | // combine it because that will just produce multiple loads. | |||
9764 | bool IsPermutedLoad = false; | |||
9765 | const SDValue *InputLoad = getNormalLoadInput(V1, IsPermutedLoad); | |||
9766 | if (InputLoad && Subtarget.hasVSX() && V2.isUndef() && | |||
9767 | (PPC::isSplatShuffleMask(SVOp, 4) || PPC::isSplatShuffleMask(SVOp, 8)) && | |||
9768 | InputLoad->hasOneUse()) { | |||
9769 | bool IsFourByte = PPC::isSplatShuffleMask(SVOp, 4); | |||
9770 | int SplatIdx = | |||
9771 | PPC::getSplatIdxForPPCMnemonics(SVOp, IsFourByte ? 4 : 8, DAG); | |||
9772 | ||||
9773 | // The splat index for permuted loads will be in the left half of the vector | |||
9774 | // which is strictly wider than the loaded value by 8 bytes. So we need to | |||
9775 | // adjust the splat index to point to the correct address in memory. | |||
9776 | if (IsPermutedLoad) { | |||
9777 | assert((isLittleEndian || IsFourByte) &&(static_cast <bool> ((isLittleEndian || IsFourByte) && "Unexpected size for permuted load on big endian target") ? void (0) : __assert_fail ("(isLittleEndian || IsFourByte) && \"Unexpected size for permuted load on big endian target\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9778, __extension__ __PRETTY_FUNCTION__)) | |||
9778 | "Unexpected size for permuted load on big endian target")(static_cast <bool> ((isLittleEndian || IsFourByte) && "Unexpected size for permuted load on big endian target") ? void (0) : __assert_fail ("(isLittleEndian || IsFourByte) && \"Unexpected size for permuted load on big endian target\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9778, __extension__ __PRETTY_FUNCTION__)); | |||
9779 | SplatIdx += IsFourByte ? 2 : 1; | |||
9780 | assert((SplatIdx < (IsFourByte ? 4 : 2)) &&(static_cast <bool> ((SplatIdx < (IsFourByte ? 4 : 2 )) && "Splat of a value outside of the loaded memory" ) ? void (0) : __assert_fail ("(SplatIdx < (IsFourByte ? 4 : 2)) && \"Splat of a value outside of the loaded memory\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9781, __extension__ __PRETTY_FUNCTION__)) | |||
9781 | "Splat of a value outside of the loaded memory")(static_cast <bool> ((SplatIdx < (IsFourByte ? 4 : 2 )) && "Splat of a value outside of the loaded memory" ) ? void (0) : __assert_fail ("(SplatIdx < (IsFourByte ? 4 : 2)) && \"Splat of a value outside of the loaded memory\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 9781, __extension__ __PRETTY_FUNCTION__)); | |||
9782 | } | |||
9783 | ||||
9784 | LoadSDNode *LD = cast<LoadSDNode>(*InputLoad); | |||
9785 | // For 4-byte load-and-splat, we need Power9. | |||
9786 | if ((IsFourByte && Subtarget.hasP9Vector()) || !IsFourByte) { | |||
9787 | uint64_t Offset = 0; | |||
9788 | if (IsFourByte) | |||
9789 | Offset = isLittleEndian ? (3 - SplatIdx) * 4 : SplatIdx * 4; | |||
9790 | else | |||
9791 | Offset = isLittleEndian ? (1 - SplatIdx) * 8 : SplatIdx * 8; | |||
9792 | ||||
9793 | // If the width of the load is the same as the width of the splat, | |||
9794 | // loading with an offset would load the wrong memory. | |||
9795 | if (LD->getValueType(0).getSizeInBits() == (IsFourByte ? 32 : 64)) | |||
9796 | Offset = 0; | |||
9797 | ||||
9798 | SDValue BasePtr = LD->getBasePtr(); | |||
9799 | if (Offset != 0) | |||
9800 | BasePtr = DAG.getNode(ISD::ADD, dl, getPointerTy(DAG.getDataLayout()), | |||
9801 | BasePtr, DAG.getIntPtrConstant(Offset, dl)); | |||
9802 | SDValue Ops[] = { | |||
9803 | LD->getChain(), // Chain | |||
9804 | BasePtr, // BasePtr | |||
9805 | DAG.getValueType(Op.getValueType()) // VT | |||
9806 | }; | |||
9807 | SDVTList VTL = | |||
9808 | DAG.getVTList(IsFourByte ? MVT::v4i32 : MVT::v2i64, MVT::Other); | |||
9809 | SDValue LdSplt = | |||
9810 | DAG.getMemIntrinsicNode(PPCISD::LD_SPLAT, dl, VTL, | |||
9811 | Ops, LD->getMemoryVT(), LD->getMemOperand()); | |||
9812 | DAG.ReplaceAllUsesOfValueWith(InputLoad->getValue(1), LdSplt.getValue(1)); | |||
9813 | if (LdSplt.getValueType() != SVOp->getValueType(0)) | |||
9814 | LdSplt = DAG.getBitcast(SVOp->getValueType(0), LdSplt); | |||
9815 | return LdSplt; | |||
9816 | } | |||
9817 | } | |||
9818 | if (Subtarget.hasP9Vector() && | |||
9819 | PPC::isXXINSERTWMask(SVOp, ShiftElts, InsertAtByte, Swap, | |||
9820 | isLittleEndian)) { | |||
9821 | if (Swap) | |||
9822 | std::swap(V1, V2); | |||
9823 | SDValue Conv1 = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V1); | |||
9824 | SDValue Conv2 = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V2); | |||
9825 | if (ShiftElts) { | |||
9826 | SDValue Shl = DAG.getNode(PPCISD::VECSHL, dl, MVT::v4i32, Conv2, Conv2, | |||
9827 | DAG.getConstant(ShiftElts, dl, MVT::i32)); | |||
9828 | SDValue Ins = DAG.getNode(PPCISD::VECINSERT, dl, MVT::v4i32, Conv1, Shl, | |||
9829 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
9830 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Ins); | |||
9831 | } | |||
9832 | SDValue Ins = DAG.getNode(PPCISD::VECINSERT, dl, MVT::v4i32, Conv1, Conv2, | |||
9833 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
9834 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Ins); | |||
9835 | } | |||
9836 | ||||
9837 | if (Subtarget.hasPrefixInstrs()) { | |||
9838 | SDValue SplatInsertNode; | |||
9839 | if ((SplatInsertNode = lowerToXXSPLTI32DX(SVOp, DAG))) | |||
9840 | return SplatInsertNode; | |||
9841 | } | |||
9842 | ||||
9843 | if (Subtarget.hasP9Altivec()) { | |||
9844 | SDValue NewISDNode; | |||
9845 | if ((NewISDNode = lowerToVINSERTH(SVOp, DAG))) | |||
9846 | return NewISDNode; | |||
9847 | ||||
9848 | if ((NewISDNode = lowerToVINSERTB(SVOp, DAG))) | |||
9849 | return NewISDNode; | |||
9850 | } | |||
9851 | ||||
9852 | if (Subtarget.hasVSX() && | |||
9853 | PPC::isXXSLDWIShuffleMask(SVOp, ShiftElts, Swap, isLittleEndian)) { | |||
9854 | if (Swap) | |||
9855 | std::swap(V1, V2); | |||
9856 | SDValue Conv1 = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V1); | |||
9857 | SDValue Conv2 = | |||
9858 | DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V2.isUndef() ? V1 : V2); | |||
9859 | ||||
9860 | SDValue Shl = DAG.getNode(PPCISD::VECSHL, dl, MVT::v4i32, Conv1, Conv2, | |||
9861 | DAG.getConstant(ShiftElts, dl, MVT::i32)); | |||
9862 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Shl); | |||
9863 | } | |||
9864 | ||||
9865 | if (Subtarget.hasVSX() && | |||
9866 | PPC::isXXPERMDIShuffleMask(SVOp, ShiftElts, Swap, isLittleEndian)) { | |||
9867 | if (Swap) | |||
9868 | std::swap(V1, V2); | |||
9869 | SDValue Conv1 = DAG.getNode(ISD::BITCAST, dl, MVT::v2i64, V1); | |||
9870 | SDValue Conv2 = | |||
9871 | DAG.getNode(ISD::BITCAST, dl, MVT::v2i64, V2.isUndef() ? V1 : V2); | |||
9872 | ||||
9873 | SDValue PermDI = DAG.getNode(PPCISD::XXPERMDI, dl, MVT::v2i64, Conv1, Conv2, | |||
9874 | DAG.getConstant(ShiftElts, dl, MVT::i32)); | |||
9875 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, PermDI); | |||
9876 | } | |||
9877 | ||||
9878 | if (Subtarget.hasP9Vector()) { | |||
9879 | if (PPC::isXXBRHShuffleMask(SVOp)) { | |||
9880 | SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, V1); | |||
9881 | SDValue ReveHWord = DAG.getNode(ISD::BSWAP, dl, MVT::v8i16, Conv); | |||
9882 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveHWord); | |||
9883 | } else if (PPC::isXXBRWShuffleMask(SVOp)) { | |||
9884 | SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V1); | |||
9885 | SDValue ReveWord = DAG.getNode(ISD::BSWAP, dl, MVT::v4i32, Conv); | |||
9886 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveWord); | |||
9887 | } else if (PPC::isXXBRDShuffleMask(SVOp)) { | |||
9888 | SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v2i64, V1); | |||
9889 | SDValue ReveDWord = DAG.getNode(ISD::BSWAP, dl, MVT::v2i64, Conv); | |||
9890 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveDWord); | |||
9891 | } else if (PPC::isXXBRQShuffleMask(SVOp)) { | |||
9892 | SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v1i128, V1); | |||
9893 | SDValue ReveQWord = DAG.getNode(ISD::BSWAP, dl, MVT::v1i128, Conv); | |||
9894 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveQWord); | |||
9895 | } | |||
9896 | } | |||
9897 | ||||
9898 | if (Subtarget.hasVSX()) { | |||
9899 | if (V2.isUndef() && PPC::isSplatShuffleMask(SVOp, 4)) { | |||
9900 | int SplatIdx = PPC::getSplatIdxForPPCMnemonics(SVOp, 4, DAG); | |||
9901 | ||||
9902 | SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V1); | |||
9903 | SDValue Splat = DAG.getNode(PPCISD::XXSPLT, dl, MVT::v4i32, Conv, | |||
9904 | DAG.getConstant(SplatIdx, dl, MVT::i32)); | |||
9905 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Splat); | |||
9906 | } | |||
9907 | ||||
9908 | // Left shifts of 8 bytes are actually swaps. Convert accordingly. | |||
9909 | if (V2.isUndef() && PPC::isVSLDOIShuffleMask(SVOp, 1, DAG) == 8) { | |||
9910 | SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v2f64, V1); | |||
9911 | SDValue Swap = DAG.getNode(PPCISD::SWAP_NO_CHAIN, dl, MVT::v2f64, Conv); | |||
9912 | return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, Swap); | |||
9913 | } | |||
9914 | } | |||
9915 | ||||
9916 | // Cases that are handled by instructions that take permute immediates | |||
9917 | // (such as vsplt*) should be left as VECTOR_SHUFFLE nodes so they can be | |||
9918 | // selected by the instruction selector. | |||
9919 | if (V2.isUndef()) { | |||
9920 | if (PPC::isSplatShuffleMask(SVOp, 1) || | |||
9921 | PPC::isSplatShuffleMask(SVOp, 2) || | |||
9922 | PPC::isSplatShuffleMask(SVOp, 4) || | |||
9923 | PPC::isVPKUWUMShuffleMask(SVOp, 1, DAG) || | |||
9924 | PPC::isVPKUHUMShuffleMask(SVOp, 1, DAG) || | |||
9925 | PPC::isVSLDOIShuffleMask(SVOp, 1, DAG) != -1 || | |||
9926 | PPC::isVMRGLShuffleMask(SVOp, 1, 1, DAG) || | |||
9927 | PPC::isVMRGLShuffleMask(SVOp, 2, 1, DAG) || | |||
9928 | PPC::isVMRGLShuffleMask(SVOp, 4, 1, DAG) || | |||
9929 | PPC::isVMRGHShuffleMask(SVOp, 1, 1, DAG) || | |||
9930 | PPC::isVMRGHShuffleMask(SVOp, 2, 1, DAG) || | |||
9931 | PPC::isVMRGHShuffleMask(SVOp, 4, 1, DAG) || | |||
9932 | (Subtarget.hasP8Altivec() && ( | |||
9933 | PPC::isVPKUDUMShuffleMask(SVOp, 1, DAG) || | |||
9934 | PPC::isVMRGEOShuffleMask(SVOp, true, 1, DAG) || | |||
9935 | PPC::isVMRGEOShuffleMask(SVOp, false, 1, DAG)))) { | |||
9936 | return Op; | |||
9937 | } | |||
9938 | } | |||
9939 | ||||
9940 | // Altivec has a variety of "shuffle immediates" that take two vector inputs | |||
9941 | // and produce a fixed permutation. If any of these match, do not lower to | |||
9942 | // VPERM. | |||
9943 | unsigned int ShuffleKind = isLittleEndian ? 2 : 0; | |||
9944 | if (PPC::isVPKUWUMShuffleMask(SVOp, ShuffleKind, DAG) || | |||
9945 | PPC::isVPKUHUMShuffleMask(SVOp, ShuffleKind, DAG) || | |||
9946 | PPC::isVSLDOIShuffleMask(SVOp, ShuffleKind, DAG) != -1 || | |||
9947 | PPC::isVMRGLShuffleMask(SVOp, 1, ShuffleKind, DAG) || | |||
9948 | PPC::isVMRGLShuffleMask(SVOp, 2, ShuffleKind, DAG) || | |||
9949 | PPC::isVMRGLShuffleMask(SVOp, 4, ShuffleKind, DAG) || | |||
9950 | PPC::isVMRGHShuffleMask(SVOp, 1, ShuffleKind, DAG) || | |||
9951 | PPC::isVMRGHShuffleMask(SVOp, 2, ShuffleKind, DAG) || | |||
9952 | PPC::isVMRGHShuffleMask(SVOp, 4, ShuffleKind, DAG) || | |||
9953 | (Subtarget.hasP8Altivec() && ( | |||
9954 | PPC::isVPKUDUMShuffleMask(SVOp, ShuffleKind, DAG) || | |||
9955 | PPC::isVMRGEOShuffleMask(SVOp, true, ShuffleKind, DAG) || | |||
9956 | PPC::isVMRGEOShuffleMask(SVOp, false, ShuffleKind, DAG)))) | |||
9957 | return Op; | |||
9958 | ||||
9959 | // Check to see if this is a shuffle of 4-byte values. If so, we can use our | |||
9960 | // perfect shuffle table to emit an optimal matching sequence. | |||
9961 | ArrayRef<int> PermMask = SVOp->getMask(); | |||
9962 | ||||
9963 | unsigned PFIndexes[4]; | |||
9964 | bool isFourElementShuffle = true; | |||
9965 | for (unsigned i = 0; i != 4 && isFourElementShuffle; ++i) { // Element number | |||
9966 | unsigned EltNo = 8; // Start out undef. | |||
9967 | for (unsigned j = 0; j != 4; ++j) { // Intra-element byte. | |||
9968 | if (PermMask[i*4+j] < 0) | |||
9969 | continue; // Undef, ignore it. | |||
9970 | ||||
9971 | unsigned ByteSource = PermMask[i*4+j]; | |||
9972 | if ((ByteSource & 3) != j) { | |||
9973 | isFourElementShuffle = false; | |||
9974 | break; | |||
9975 | } | |||
9976 | ||||
9977 | if (EltNo == 8) { | |||
9978 | EltNo = ByteSource/4; | |||
9979 | } else if (EltNo != ByteSource/4) { | |||
9980 | isFourElementShuffle = false; | |||
9981 | break; | |||
9982 | } | |||
9983 | } | |||
9984 | PFIndexes[i] = EltNo; | |||
9985 | } | |||
9986 | ||||
9987 | // If this shuffle can be expressed as a shuffle of 4-byte elements, use the | |||
9988 | // perfect shuffle vector to determine if it is cost effective to do this as | |||
9989 | // discrete instructions, or whether we should use a vperm. | |||
9990 | // For now, we skip this for little endian until such time as we have a | |||
9991 | // little-endian perfect shuffle table. | |||
9992 | if (isFourElementShuffle && !isLittleEndian) { | |||
9993 | // Compute the index in the perfect shuffle table. | |||
9994 | unsigned PFTableIndex = | |||
9995 | PFIndexes[0]*9*9*9+PFIndexes[1]*9*9+PFIndexes[2]*9+PFIndexes[3]; | |||
9996 | ||||
9997 | unsigned PFEntry = PerfectShuffleTable[PFTableIndex]; | |||
9998 | unsigned Cost = (PFEntry >> 30); | |||
9999 | ||||
10000 | // Determining when to avoid vperm is tricky. Many things affect the cost | |||
10001 | // of vperm, particularly how many times the perm mask needs to be computed. | |||
10002 | // For example, if the perm mask can be hoisted out of a loop or is already | |||
10003 | // used (perhaps because there are multiple permutes with the same shuffle | |||
10004 | // mask?) the vperm has a cost of 1. OTOH, hoisting the permute mask out of | |||
10005 | // the loop requires an extra register. | |||
10006 | // | |||
10007 | // As a compromise, we only emit discrete instructions if the shuffle can be | |||
10008 | // generated in 3 or fewer operations. When we have loop information | |||
10009 | // available, if this block is within a loop, we should avoid using vperm | |||
10010 | // for 3-operation perms and use a constant pool load instead. | |||
10011 | if (Cost < 3) | |||
10012 | return GeneratePerfectShuffle(PFEntry, V1, V2, DAG, dl); | |||
10013 | } | |||
10014 | ||||
10015 | // Lower this to a VPERM(V1, V2, V3) expression, where V3 is a constant | |||
10016 | // vector that will get spilled to the constant pool. | |||
10017 | if (V2.isUndef()) V2 = V1; | |||
10018 | ||||
10019 | // The SHUFFLE_VECTOR mask is almost exactly what we want for vperm, except | |||
10020 | // that it is in input element units, not in bytes. Convert now. | |||
10021 | ||||
10022 | // For little endian, the order of the input vectors is reversed, and | |||
10023 | // the permutation mask is complemented with respect to 31. This is | |||
10024 | // necessary to produce proper semantics with the big-endian-biased vperm | |||
10025 | // instruction. | |||
10026 | EVT EltVT = V1.getValueType().getVectorElementType(); | |||
10027 | unsigned BytesPerElement = EltVT.getSizeInBits()/8; | |||
10028 | ||||
10029 | SmallVector<SDValue, 16> ResultMask; | |||
10030 | for (unsigned i = 0, e = VT.getVectorNumElements(); i != e; ++i) { | |||
10031 | unsigned SrcElt = PermMask[i] < 0 ? 0 : PermMask[i]; | |||
10032 | ||||
10033 | for (unsigned j = 0; j != BytesPerElement; ++j) | |||
10034 | if (isLittleEndian) | |||
10035 | ResultMask.push_back(DAG.getConstant(31 - (SrcElt*BytesPerElement + j), | |||
10036 | dl, MVT::i32)); | |||
10037 | else | |||
10038 | ResultMask.push_back(DAG.getConstant(SrcElt*BytesPerElement + j, dl, | |||
10039 | MVT::i32)); | |||
10040 | } | |||
10041 | ||||
10042 | ShufflesHandledWithVPERM++; | |||
10043 | SDValue VPermMask = DAG.getBuildVector(MVT::v16i8, dl, ResultMask); | |||
10044 | LLVM_DEBUG(dbgs() << "Emitting a VPERM for the following shuffle:\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { dbgs() << "Emitting a VPERM for the following shuffle:\n" ; } } while (false); | |||
10045 | LLVM_DEBUG(SVOp->dump())do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { SVOp->dump(); } } while (false); | |||
10046 | LLVM_DEBUG(dbgs() << "With the following permute control vector:\n")do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { dbgs() << "With the following permute control vector:\n" ; } } while (false); | |||
10047 | LLVM_DEBUG(VPermMask.dump())do { if (::llvm::DebugFlag && ::llvm::isCurrentDebugType ("ppc-lowering")) { VPermMask.dump(); } } while (false); | |||
10048 | ||||
10049 | if (isLittleEndian) | |||
10050 | return DAG.getNode(PPCISD::VPERM, dl, V1.getValueType(), | |||
10051 | V2, V1, VPermMask); | |||
10052 | else | |||
10053 | return DAG.getNode(PPCISD::VPERM, dl, V1.getValueType(), | |||
10054 | V1, V2, VPermMask); | |||
10055 | } | |||
10056 | ||||
10057 | /// getVectorCompareInfo - Given an intrinsic, return false if it is not a | |||
10058 | /// vector comparison. If it is, return true and fill in Opc/isDot with | |||
10059 | /// information about the intrinsic. | |||
10060 | static bool getVectorCompareInfo(SDValue Intrin, int &CompareOpc, | |||
10061 | bool &isDot, const PPCSubtarget &Subtarget) { | |||
10062 | unsigned IntrinsicID = | |||
10063 | cast<ConstantSDNode>(Intrin.getOperand(0))->getZExtValue(); | |||
10064 | CompareOpc = -1; | |||
10065 | isDot = false; | |||
10066 | switch (IntrinsicID) { | |||
10067 | default: | |||
10068 | return false; | |||
10069 | // Comparison predicates. | |||
10070 | case Intrinsic::ppc_altivec_vcmpbfp_p: | |||
10071 | CompareOpc = 966; | |||
10072 | isDot = true; | |||
10073 | break; | |||
10074 | case Intrinsic::ppc_altivec_vcmpeqfp_p: | |||
10075 | CompareOpc = 198; | |||
10076 | isDot = true; | |||
10077 | break; | |||
10078 | case Intrinsic::ppc_altivec_vcmpequb_p: | |||
10079 | CompareOpc = 6; | |||
10080 | isDot = true; | |||
10081 | break; | |||
10082 | case Intrinsic::ppc_altivec_vcmpequh_p: | |||
10083 | CompareOpc = 70; | |||
10084 | isDot = true; | |||
10085 | break; | |||
10086 | case Intrinsic::ppc_altivec_vcmpequw_p: | |||
10087 | CompareOpc = 134; | |||
10088 | isDot = true; | |||
10089 | break; | |||
10090 | case Intrinsic::ppc_altivec_vcmpequd_p: | |||
10091 | if (Subtarget.hasVSX() || Subtarget.hasP8Altivec()) { | |||
10092 | CompareOpc = 199; | |||
10093 | isDot = true; | |||
10094 | } else | |||
10095 | return false; | |||
10096 | break; | |||
10097 | case Intrinsic::ppc_altivec_vcmpneb_p: | |||
10098 | case Intrinsic::ppc_altivec_vcmpneh_p: | |||
10099 | case Intrinsic::ppc_altivec_vcmpnew_p: | |||
10100 | case Intrinsic::ppc_altivec_vcmpnezb_p: | |||
10101 | case Intrinsic::ppc_altivec_vcmpnezh_p: | |||
10102 | case Intrinsic::ppc_altivec_vcmpnezw_p: | |||
10103 | if (Subtarget.hasP9Altivec()) { | |||
10104 | switch (IntrinsicID) { | |||
10105 | default: | |||
10106 | llvm_unreachable("Unknown comparison intrinsic.")::llvm::llvm_unreachable_internal("Unknown comparison intrinsic." , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10106); | |||
10107 | case Intrinsic::ppc_altivec_vcmpneb_p: | |||
10108 | CompareOpc = 7; | |||
10109 | break; | |||
10110 | case Intrinsic::ppc_altivec_vcmpneh_p: | |||
10111 | CompareOpc = 71; | |||
10112 | break; | |||
10113 | case Intrinsic::ppc_altivec_vcmpnew_p: | |||
10114 | CompareOpc = 135; | |||
10115 | break; | |||
10116 | case Intrinsic::ppc_altivec_vcmpnezb_p: | |||
10117 | CompareOpc = 263; | |||
10118 | break; | |||
10119 | case Intrinsic::ppc_altivec_vcmpnezh_p: | |||
10120 | CompareOpc = 327; | |||
10121 | break; | |||
10122 | case Intrinsic::ppc_altivec_vcmpnezw_p: | |||
10123 | CompareOpc = 391; | |||
10124 | break; | |||
10125 | } | |||
10126 | isDot = true; | |||
10127 | } else | |||
10128 | return false; | |||
10129 | break; | |||
10130 | case Intrinsic::ppc_altivec_vcmpgefp_p: | |||
10131 | CompareOpc = 454; | |||
10132 | isDot = true; | |||
10133 | break; | |||
10134 | case Intrinsic::ppc_altivec_vcmpgtfp_p: | |||
10135 | CompareOpc = 710; | |||
10136 | isDot = true; | |||
10137 | break; | |||
10138 | case Intrinsic::ppc_altivec_vcmpgtsb_p: | |||
10139 | CompareOpc = 774; | |||
10140 | isDot = true; | |||
10141 | break; | |||
10142 | case Intrinsic::ppc_altivec_vcmpgtsh_p: | |||
10143 | CompareOpc = 838; | |||
10144 | isDot = true; | |||
10145 | break; | |||
10146 | case Intrinsic::ppc_altivec_vcmpgtsw_p: | |||
10147 | CompareOpc = 902; | |||
10148 | isDot = true; | |||
10149 | break; | |||
10150 | case Intrinsic::ppc_altivec_vcmpgtsd_p: | |||
10151 | if (Subtarget.hasVSX() || Subtarget.hasP8Altivec()) { | |||
10152 | CompareOpc = 967; | |||
10153 | isDot = true; | |||
10154 | } else | |||
10155 | return false; | |||
10156 | break; | |||
10157 | case Intrinsic::ppc_altivec_vcmpgtub_p: | |||
10158 | CompareOpc = 518; | |||
10159 | isDot = true; | |||
10160 | break; | |||
10161 | case Intrinsic::ppc_altivec_vcmpgtuh_p: | |||
10162 | CompareOpc = 582; | |||
10163 | isDot = true; | |||
10164 | break; | |||
10165 | case Intrinsic::ppc_altivec_vcmpgtuw_p: | |||
10166 | CompareOpc = 646; | |||
10167 | isDot = true; | |||
10168 | break; | |||
10169 | case Intrinsic::ppc_altivec_vcmpgtud_p: | |||
10170 | if (Subtarget.hasVSX() || Subtarget.hasP8Altivec()) { | |||
10171 | CompareOpc = 711; | |||
10172 | isDot = true; | |||
10173 | } else | |||
10174 | return false; | |||
10175 | break; | |||
10176 | ||||
10177 | case Intrinsic::ppc_altivec_vcmpequq: | |||
10178 | case Intrinsic::ppc_altivec_vcmpgtsq: | |||
10179 | case Intrinsic::ppc_altivec_vcmpgtuq: | |||
10180 | if (!Subtarget.isISA3_1()) | |||
10181 | return false; | |||
10182 | switch (IntrinsicID) { | |||
10183 | default: | |||
10184 | llvm_unreachable("Unknown comparison intrinsic.")::llvm::llvm_unreachable_internal("Unknown comparison intrinsic." , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10184); | |||
10185 | case Intrinsic::ppc_altivec_vcmpequq: | |||
10186 | CompareOpc = 455; | |||
10187 | break; | |||
10188 | case Intrinsic::ppc_altivec_vcmpgtsq: | |||
10189 | CompareOpc = 903; | |||
10190 | break; | |||
10191 | case Intrinsic::ppc_altivec_vcmpgtuq: | |||
10192 | CompareOpc = 647; | |||
10193 | break; | |||
10194 | } | |||
10195 | break; | |||
10196 | ||||
10197 | // VSX predicate comparisons use the same infrastructure | |||
10198 | case Intrinsic::ppc_vsx_xvcmpeqdp_p: | |||
10199 | case Intrinsic::ppc_vsx_xvcmpgedp_p: | |||
10200 | case Intrinsic::ppc_vsx_xvcmpgtdp_p: | |||
10201 | case Intrinsic::ppc_vsx_xvcmpeqsp_p: | |||
10202 | case Intrinsic::ppc_vsx_xvcmpgesp_p: | |||
10203 | case Intrinsic::ppc_vsx_xvcmpgtsp_p: | |||
10204 | if (Subtarget.hasVSX()) { | |||
10205 | switch (IntrinsicID) { | |||
10206 | case Intrinsic::ppc_vsx_xvcmpeqdp_p: | |||
10207 | CompareOpc = 99; | |||
10208 | break; | |||
10209 | case Intrinsic::ppc_vsx_xvcmpgedp_p: | |||
10210 | CompareOpc = 115; | |||
10211 | break; | |||
10212 | case Intrinsic::ppc_vsx_xvcmpgtdp_p: | |||
10213 | CompareOpc = 107; | |||
10214 | break; | |||
10215 | case Intrinsic::ppc_vsx_xvcmpeqsp_p: | |||
10216 | CompareOpc = 67; | |||
10217 | break; | |||
10218 | case Intrinsic::ppc_vsx_xvcmpgesp_p: | |||
10219 | CompareOpc = 83; | |||
10220 | break; | |||
10221 | case Intrinsic::ppc_vsx_xvcmpgtsp_p: | |||
10222 | CompareOpc = 75; | |||
10223 | break; | |||
10224 | } | |||
10225 | isDot = true; | |||
10226 | } else | |||
10227 | return false; | |||
10228 | break; | |||
10229 | ||||
10230 | // Normal Comparisons. | |||
10231 | case Intrinsic::ppc_altivec_vcmpbfp: | |||
10232 | CompareOpc = 966; | |||
10233 | break; | |||
10234 | case Intrinsic::ppc_altivec_vcmpeqfp: | |||
10235 | CompareOpc = 198; | |||
10236 | break; | |||
10237 | case Intrinsic::ppc_altivec_vcmpequb: | |||
10238 | CompareOpc = 6; | |||
10239 | break; | |||
10240 | case Intrinsic::ppc_altivec_vcmpequh: | |||
10241 | CompareOpc = 70; | |||
10242 | break; | |||
10243 | case Intrinsic::ppc_altivec_vcmpequw: | |||
10244 | CompareOpc = 134; | |||
10245 | break; | |||
10246 | case Intrinsic::ppc_altivec_vcmpequd: | |||
10247 | if (Subtarget.hasP8Altivec()) | |||
10248 | CompareOpc = 199; | |||
10249 | else | |||
10250 | return false; | |||
10251 | break; | |||
10252 | case Intrinsic::ppc_altivec_vcmpneb: | |||
10253 | case Intrinsic::ppc_altivec_vcmpneh: | |||
10254 | case Intrinsic::ppc_altivec_vcmpnew: | |||
10255 | case Intrinsic::ppc_altivec_vcmpnezb: | |||
10256 | case Intrinsic::ppc_altivec_vcmpnezh: | |||
10257 | case Intrinsic::ppc_altivec_vcmpnezw: | |||
10258 | if (Subtarget.hasP9Altivec()) | |||
10259 | switch (IntrinsicID) { | |||
10260 | default: | |||
10261 | llvm_unreachable("Unknown comparison intrinsic.")::llvm::llvm_unreachable_internal("Unknown comparison intrinsic." , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10261); | |||
10262 | case Intrinsic::ppc_altivec_vcmpneb: | |||
10263 | CompareOpc = 7; | |||
10264 | break; | |||
10265 | case Intrinsic::ppc_altivec_vcmpneh: | |||
10266 | CompareOpc = 71; | |||
10267 | break; | |||
10268 | case Intrinsic::ppc_altivec_vcmpnew: | |||
10269 | CompareOpc = 135; | |||
10270 | break; | |||
10271 | case Intrinsic::ppc_altivec_vcmpnezb: | |||
10272 | CompareOpc = 263; | |||
10273 | break; | |||
10274 | case Intrinsic::ppc_altivec_vcmpnezh: | |||
10275 | CompareOpc = 327; | |||
10276 | break; | |||
10277 | case Intrinsic::ppc_altivec_vcmpnezw: | |||
10278 | CompareOpc = 391; | |||
10279 | break; | |||
10280 | } | |||
10281 | else | |||
10282 | return false; | |||
10283 | break; | |||
10284 | case Intrinsic::ppc_altivec_vcmpgefp: | |||
10285 | CompareOpc = 454; | |||
10286 | break; | |||
10287 | case Intrinsic::ppc_altivec_vcmpgtfp: | |||
10288 | CompareOpc = 710; | |||
10289 | break; | |||
10290 | case Intrinsic::ppc_altivec_vcmpgtsb: | |||
10291 | CompareOpc = 774; | |||
10292 | break; | |||
10293 | case Intrinsic::ppc_altivec_vcmpgtsh: | |||
10294 | CompareOpc = 838; | |||
10295 | break; | |||
10296 | case Intrinsic::ppc_altivec_vcmpgtsw: | |||
10297 | CompareOpc = 902; | |||
10298 | break; | |||
10299 | case Intrinsic::ppc_altivec_vcmpgtsd: | |||
10300 | if (Subtarget.hasP8Altivec()) | |||
10301 | CompareOpc = 967; | |||
10302 | else | |||
10303 | return false; | |||
10304 | break; | |||
10305 | case Intrinsic::ppc_altivec_vcmpgtub: | |||
10306 | CompareOpc = 518; | |||
10307 | break; | |||
10308 | case Intrinsic::ppc_altivec_vcmpgtuh: | |||
10309 | CompareOpc = 582; | |||
10310 | break; | |||
10311 | case Intrinsic::ppc_altivec_vcmpgtuw: | |||
10312 | CompareOpc = 646; | |||
10313 | break; | |||
10314 | case Intrinsic::ppc_altivec_vcmpgtud: | |||
10315 | if (Subtarget.hasP8Altivec()) | |||
10316 | CompareOpc = 711; | |||
10317 | else | |||
10318 | return false; | |||
10319 | break; | |||
10320 | case Intrinsic::ppc_altivec_vcmpequq_p: | |||
10321 | case Intrinsic::ppc_altivec_vcmpgtsq_p: | |||
10322 | case Intrinsic::ppc_altivec_vcmpgtuq_p: | |||
10323 | if (!Subtarget.isISA3_1()) | |||
10324 | return false; | |||
10325 | switch (IntrinsicID) { | |||
10326 | default: | |||
10327 | llvm_unreachable("Unknown comparison intrinsic.")::llvm::llvm_unreachable_internal("Unknown comparison intrinsic." , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10327); | |||
10328 | case Intrinsic::ppc_altivec_vcmpequq_p: | |||
10329 | CompareOpc = 455; | |||
10330 | break; | |||
10331 | case Intrinsic::ppc_altivec_vcmpgtsq_p: | |||
10332 | CompareOpc = 903; | |||
10333 | break; | |||
10334 | case Intrinsic::ppc_altivec_vcmpgtuq_p: | |||
10335 | CompareOpc = 647; | |||
10336 | break; | |||
10337 | } | |||
10338 | isDot = true; | |||
10339 | break; | |||
10340 | } | |||
10341 | return true; | |||
10342 | } | |||
10343 | ||||
10344 | /// LowerINTRINSIC_WO_CHAIN - If this is an intrinsic that we want to custom | |||
10345 | /// lower, do it, otherwise return null. | |||
10346 | SDValue PPCTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op, | |||
10347 | SelectionDAG &DAG) const { | |||
10348 | unsigned IntrinsicID = | |||
10349 | cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue(); | |||
10350 | ||||
10351 | SDLoc dl(Op); | |||
10352 | ||||
10353 | switch (IntrinsicID) { | |||
10354 | case Intrinsic::thread_pointer: | |||
10355 | // Reads the thread pointer register, used for __builtin_thread_pointer. | |||
10356 | if (Subtarget.isPPC64()) | |||
10357 | return DAG.getRegister(PPC::X13, MVT::i64); | |||
10358 | return DAG.getRegister(PPC::R2, MVT::i32); | |||
10359 | ||||
10360 | case Intrinsic::ppc_mma_disassemble_acc: | |||
10361 | case Intrinsic::ppc_vsx_disassemble_pair: { | |||
10362 | int NumVecs = 2; | |||
10363 | SDValue WideVec = Op.getOperand(1); | |||
10364 | if (IntrinsicID == Intrinsic::ppc_mma_disassemble_acc) { | |||
10365 | NumVecs = 4; | |||
10366 | WideVec = DAG.getNode(PPCISD::XXMFACC, dl, MVT::v512i1, WideVec); | |||
10367 | } | |||
10368 | SmallVector<SDValue, 4> RetOps; | |||
10369 | for (int VecNo = 0; VecNo < NumVecs; VecNo++) { | |||
10370 | SDValue Extract = DAG.getNode( | |||
10371 | PPCISD::EXTRACT_VSX_REG, dl, MVT::v16i8, WideVec, | |||
10372 | DAG.getConstant(Subtarget.isLittleEndian() ? NumVecs - 1 - VecNo | |||
10373 | : VecNo, | |||
10374 | dl, getPointerTy(DAG.getDataLayout()))); | |||
10375 | RetOps.push_back(Extract); | |||
10376 | } | |||
10377 | return DAG.getMergeValues(RetOps, dl); | |||
10378 | } | |||
10379 | } | |||
10380 | ||||
10381 | // If this is a lowered altivec predicate compare, CompareOpc is set to the | |||
10382 | // opcode number of the comparison. | |||
10383 | int CompareOpc; | |||
10384 | bool isDot; | |||
10385 | if (!getVectorCompareInfo(Op, CompareOpc, isDot, Subtarget)) | |||
10386 | return SDValue(); // Don't custom lower most intrinsics. | |||
10387 | ||||
10388 | // If this is a non-dot comparison, make the VCMP node and we are done. | |||
10389 | if (!isDot) { | |||
10390 | SDValue Tmp = DAG.getNode(PPCISD::VCMP, dl, Op.getOperand(2).getValueType(), | |||
10391 | Op.getOperand(1), Op.getOperand(2), | |||
10392 | DAG.getConstant(CompareOpc, dl, MVT::i32)); | |||
10393 | return DAG.getNode(ISD::BITCAST, dl, Op.getValueType(), Tmp); | |||
10394 | } | |||
10395 | ||||
10396 | // Create the PPCISD altivec 'dot' comparison node. | |||
10397 | SDValue Ops[] = { | |||
10398 | Op.getOperand(2), // LHS | |||
10399 | Op.getOperand(3), // RHS | |||
10400 | DAG.getConstant(CompareOpc, dl, MVT::i32) | |||
10401 | }; | |||
10402 | EVT VTs[] = { Op.getOperand(2).getValueType(), MVT::Glue }; | |||
10403 | SDValue CompNode = DAG.getNode(PPCISD::VCMP_rec, dl, VTs, Ops); | |||
10404 | ||||
10405 | // Now that we have the comparison, emit a copy from the CR to a GPR. | |||
10406 | // This is flagged to the above dot comparison. | |||
10407 | SDValue Flags = DAG.getNode(PPCISD::MFOCRF, dl, MVT::i32, | |||
10408 | DAG.getRegister(PPC::CR6, MVT::i32), | |||
10409 | CompNode.getValue(1)); | |||
10410 | ||||
10411 | // Unpack the result based on how the target uses it. | |||
10412 | unsigned BitNo; // Bit # of CR6. | |||
10413 | bool InvertBit; // Invert result? | |||
10414 | switch (cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue()) { | |||
10415 | default: // Can't happen, don't crash on invalid number though. | |||
10416 | case 0: // Return the value of the EQ bit of CR6. | |||
10417 | BitNo = 0; InvertBit = false; | |||
10418 | break; | |||
10419 | case 1: // Return the inverted value of the EQ bit of CR6. | |||
10420 | BitNo = 0; InvertBit = true; | |||
10421 | break; | |||
10422 | case 2: // Return the value of the LT bit of CR6. | |||
10423 | BitNo = 2; InvertBit = false; | |||
10424 | break; | |||
10425 | case 3: // Return the inverted value of the LT bit of CR6. | |||
10426 | BitNo = 2; InvertBit = true; | |||
10427 | break; | |||
10428 | } | |||
10429 | ||||
10430 | // Shift the bit into the low position. | |||
10431 | Flags = DAG.getNode(ISD::SRL, dl, MVT::i32, Flags, | |||
10432 | DAG.getConstant(8 - (3 - BitNo), dl, MVT::i32)); | |||
10433 | // Isolate the bit. | |||
10434 | Flags = DAG.getNode(ISD::AND, dl, MVT::i32, Flags, | |||
10435 | DAG.getConstant(1, dl, MVT::i32)); | |||
10436 | ||||
10437 | // If we are supposed to, toggle the bit. | |||
10438 | if (InvertBit) | |||
10439 | Flags = DAG.getNode(ISD::XOR, dl, MVT::i32, Flags, | |||
10440 | DAG.getConstant(1, dl, MVT::i32)); | |||
10441 | return Flags; | |||
10442 | } | |||
10443 | ||||
10444 | SDValue PPCTargetLowering::LowerINTRINSIC_VOID(SDValue Op, | |||
10445 | SelectionDAG &DAG) const { | |||
10446 | // SelectionDAGBuilder::visitTargetIntrinsic may insert one extra chain to | |||
10447 | // the beginning of the argument list. | |||
10448 | int ArgStart = isa<ConstantSDNode>(Op.getOperand(0)) ? 0 : 1; | |||
10449 | SDLoc DL(Op); | |||
10450 | switch (cast<ConstantSDNode>(Op.getOperand(ArgStart))->getZExtValue()) { | |||
10451 | case Intrinsic::ppc_cfence: { | |||
10452 | assert(ArgStart == 1 && "llvm.ppc.cfence must carry a chain argument.")(static_cast <bool> (ArgStart == 1 && "llvm.ppc.cfence must carry a chain argument." ) ? void (0) : __assert_fail ("ArgStart == 1 && \"llvm.ppc.cfence must carry a chain argument.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10452, __extension__ __PRETTY_FUNCTION__)); | |||
10453 | assert(Subtarget.isPPC64() && "Only 64-bit is supported for now.")(static_cast <bool> (Subtarget.isPPC64() && "Only 64-bit is supported for now." ) ? void (0) : __assert_fail ("Subtarget.isPPC64() && \"Only 64-bit is supported for now.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10453, __extension__ __PRETTY_FUNCTION__)); | |||
10454 | return SDValue(DAG.getMachineNode(PPC::CFENCE8, DL, MVT::Other, | |||
10455 | DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, | |||
10456 | Op.getOperand(ArgStart + 1)), | |||
10457 | Op.getOperand(0)), | |||
10458 | 0); | |||
10459 | } | |||
10460 | default: | |||
10461 | break; | |||
10462 | } | |||
10463 | return SDValue(); | |||
10464 | } | |||
10465 | ||||
10466 | // Lower scalar BSWAP64 to xxbrd. | |||
10467 | SDValue PPCTargetLowering::LowerBSWAP(SDValue Op, SelectionDAG &DAG) const { | |||
10468 | SDLoc dl(Op); | |||
10469 | if (!Subtarget.isPPC64()) | |||
10470 | return Op; | |||
10471 | // MTVSRDD | |||
10472 | Op = DAG.getNode(ISD::BUILD_VECTOR, dl, MVT::v2i64, Op.getOperand(0), | |||
10473 | Op.getOperand(0)); | |||
10474 | // XXBRD | |||
10475 | Op = DAG.getNode(ISD::BSWAP, dl, MVT::v2i64, Op); | |||
10476 | // MFVSRD | |||
10477 | int VectorIndex = 0; | |||
10478 | if (Subtarget.isLittleEndian()) | |||
10479 | VectorIndex = 1; | |||
10480 | Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i64, Op, | |||
10481 | DAG.getTargetConstant(VectorIndex, dl, MVT::i32)); | |||
10482 | return Op; | |||
10483 | } | |||
10484 | ||||
10485 | // ATOMIC_CMP_SWAP for i8/i16 needs to zero-extend its input since it will be | |||
10486 | // compared to a value that is atomically loaded (atomic loads zero-extend). | |||
10487 | SDValue PPCTargetLowering::LowerATOMIC_CMP_SWAP(SDValue Op, | |||
10488 | SelectionDAG &DAG) const { | |||
10489 | assert(Op.getOpcode() == ISD::ATOMIC_CMP_SWAP &&(static_cast <bool> (Op.getOpcode() == ISD::ATOMIC_CMP_SWAP && "Expecting an atomic compare-and-swap here.") ? void (0) : __assert_fail ("Op.getOpcode() == ISD::ATOMIC_CMP_SWAP && \"Expecting an atomic compare-and-swap here.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10490, __extension__ __PRETTY_FUNCTION__)) | |||
10490 | "Expecting an atomic compare-and-swap here.")(static_cast <bool> (Op.getOpcode() == ISD::ATOMIC_CMP_SWAP && "Expecting an atomic compare-and-swap here.") ? void (0) : __assert_fail ("Op.getOpcode() == ISD::ATOMIC_CMP_SWAP && \"Expecting an atomic compare-and-swap here.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10490, __extension__ __PRETTY_FUNCTION__)); | |||
10491 | SDLoc dl(Op); | |||
10492 | auto *AtomicNode = cast<AtomicSDNode>(Op.getNode()); | |||
10493 | EVT MemVT = AtomicNode->getMemoryVT(); | |||
10494 | if (MemVT.getSizeInBits() >= 32) | |||
10495 | return Op; | |||
10496 | ||||
10497 | SDValue CmpOp = Op.getOperand(2); | |||
10498 | // If this is already correctly zero-extended, leave it alone. | |||
10499 | auto HighBits = APInt::getHighBitsSet(32, 32 - MemVT.getSizeInBits()); | |||
10500 | if (DAG.MaskedValueIsZero(CmpOp, HighBits)) | |||
10501 | return Op; | |||
10502 | ||||
10503 | // Clear the high bits of the compare operand. | |||
10504 | unsigned MaskVal = (1 << MemVT.getSizeInBits()) - 1; | |||
10505 | SDValue NewCmpOp = | |||
10506 | DAG.getNode(ISD::AND, dl, MVT::i32, CmpOp, | |||
10507 | DAG.getConstant(MaskVal, dl, MVT::i32)); | |||
10508 | ||||
10509 | // Replace the existing compare operand with the properly zero-extended one. | |||
10510 | SmallVector<SDValue, 4> Ops; | |||
10511 | for (int i = 0, e = AtomicNode->getNumOperands(); i < e; i++) | |||
10512 | Ops.push_back(AtomicNode->getOperand(i)); | |||
10513 | Ops[2] = NewCmpOp; | |||
10514 | MachineMemOperand *MMO = AtomicNode->getMemOperand(); | |||
10515 | SDVTList Tys = DAG.getVTList(MVT::i32, MVT::Other); | |||
10516 | auto NodeTy = | |||
10517 | (MemVT == MVT::i8) ? PPCISD::ATOMIC_CMP_SWAP_8 : PPCISD::ATOMIC_CMP_SWAP_16; | |||
10518 | return DAG.getMemIntrinsicNode(NodeTy, dl, Tys, Ops, MemVT, MMO); | |||
10519 | } | |||
10520 | ||||
10521 | SDValue PPCTargetLowering::LowerSCALAR_TO_VECTOR(SDValue Op, | |||
10522 | SelectionDAG &DAG) const { | |||
10523 | SDLoc dl(Op); | |||
10524 | // Create a stack slot that is 16-byte aligned. | |||
10525 | MachineFrameInfo &MFI = DAG.getMachineFunction().getFrameInfo(); | |||
10526 | int FrameIdx = MFI.CreateStackObject(16, Align(16), false); | |||
10527 | EVT PtrVT = getPointerTy(DAG.getDataLayout()); | |||
10528 | SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); | |||
10529 | ||||
10530 | // Store the input value into Value#0 of the stack slot. | |||
10531 | SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Op.getOperand(0), FIdx, | |||
10532 | MachinePointerInfo()); | |||
10533 | // Load it out. | |||
10534 | return DAG.getLoad(Op.getValueType(), dl, Store, FIdx, MachinePointerInfo()); | |||
10535 | } | |||
10536 | ||||
10537 | SDValue PPCTargetLowering::LowerINSERT_VECTOR_ELT(SDValue Op, | |||
10538 | SelectionDAG &DAG) const { | |||
10539 | assert(Op.getOpcode() == ISD::INSERT_VECTOR_ELT &&(static_cast <bool> (Op.getOpcode() == ISD::INSERT_VECTOR_ELT && "Should only be called for ISD::INSERT_VECTOR_ELT" ) ? void (0) : __assert_fail ("Op.getOpcode() == ISD::INSERT_VECTOR_ELT && \"Should only be called for ISD::INSERT_VECTOR_ELT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10540, __extension__ __PRETTY_FUNCTION__)) | |||
10540 | "Should only be called for ISD::INSERT_VECTOR_ELT")(static_cast <bool> (Op.getOpcode() == ISD::INSERT_VECTOR_ELT && "Should only be called for ISD::INSERT_VECTOR_ELT" ) ? void (0) : __assert_fail ("Op.getOpcode() == ISD::INSERT_VECTOR_ELT && \"Should only be called for ISD::INSERT_VECTOR_ELT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10540, __extension__ __PRETTY_FUNCTION__)); | |||
10541 | ||||
10542 | ConstantSDNode *C = dyn_cast<ConstantSDNode>(Op.getOperand(2)); | |||
10543 | ||||
10544 | EVT VT = Op.getValueType(); | |||
10545 | SDLoc dl(Op); | |||
10546 | SDValue V1 = Op.getOperand(0); | |||
10547 | SDValue V2 = Op.getOperand(1); | |||
10548 | SDValue V3 = Op.getOperand(2); | |||
10549 | ||||
10550 | if (VT == MVT::v2f64 && C) | |||
10551 | return Op; | |||
10552 | ||||
10553 | if (Subtarget.isISA3_1()) { | |||
10554 | if ((VT == MVT::v2i64 || VT == MVT::v2f64) && !Subtarget.isPPC64()) | |||
10555 | return SDValue(); | |||
10556 | // On P10, we have legal lowering for constant and variable indices for | |||
10557 | // integer vectors. | |||
10558 | if (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32 || | |||
10559 | VT == MVT::v2i64) | |||
10560 | return DAG.getNode(PPCISD::VECINSERT, dl, VT, V1, V2, V3); | |||
10561 | // For f32 and f64 vectors, we have legal lowering for variable indices. | |||
10562 | // For f32 we also have legal lowering when the element is loaded from | |||
10563 | // memory. | |||
10564 | if (VT == MVT::v4f32 || VT == MVT::v2f64) { | |||
10565 | if (!C || (VT == MVT::v4f32 && dyn_cast<LoadSDNode>(V2))) | |||
10566 | return DAG.getNode(PPCISD::VECINSERT, dl, VT, V1, V2, V3); | |||
10567 | return Op; | |||
10568 | } | |||
10569 | } | |||
10570 | ||||
10571 | // Before P10, we have legal lowering for constant indices but not for | |||
10572 | // variable ones. | |||
10573 | if (!C) | |||
10574 | return SDValue(); | |||
10575 | ||||
10576 | // We can use MTVSRZ + VECINSERT for v8i16 and v16i8 types. | |||
10577 | if (VT == MVT::v8i16 || VT == MVT::v16i8) { | |||
10578 | SDValue Mtvsrz = DAG.getNode(PPCISD::MTVSRZ, dl, VT, V2); | |||
10579 | unsigned BytesInEachElement = VT.getVectorElementType().getSizeInBits() / 8; | |||
10580 | unsigned InsertAtElement = C->getZExtValue(); | |||
10581 | unsigned InsertAtByte = InsertAtElement * BytesInEachElement; | |||
10582 | if (Subtarget.isLittleEndian()) { | |||
10583 | InsertAtByte = (16 - BytesInEachElement) - InsertAtByte; | |||
10584 | } | |||
10585 | return DAG.getNode(PPCISD::VECINSERT, dl, VT, V1, Mtvsrz, | |||
10586 | DAG.getConstant(InsertAtByte, dl, MVT::i32)); | |||
10587 | } | |||
10588 | return Op; | |||
10589 | } | |||
10590 | ||||
10591 | SDValue PPCTargetLowering::LowerVectorLoad(SDValue Op, | |||
10592 | SelectionDAG &DAG) const { | |||
10593 | SDLoc dl(Op); | |||
10594 | LoadSDNode *LN = cast<LoadSDNode>(Op.getNode()); | |||
10595 | SDValue LoadChain = LN->getChain(); | |||
10596 | SDValue BasePtr = LN->getBasePtr(); | |||
10597 | EVT VT = Op.getValueType(); | |||
10598 | ||||
10599 | if (VT != MVT::v256i1 && VT != MVT::v512i1) | |||
10600 | return Op; | |||
10601 | ||||
10602 | // Type v256i1 is used for pairs and v512i1 is used for accumulators. | |||
10603 | // Here we create 2 or 4 v16i8 loads to load the pair or accumulator value in | |||
10604 | // 2 or 4 vsx registers. | |||
10605 | assert((VT != MVT::v512i1 || Subtarget.hasMMA()) &&(static_cast <bool> ((VT != MVT::v512i1 || Subtarget.hasMMA ()) && "Type unsupported without MMA") ? void (0) : __assert_fail ("(VT != MVT::v512i1 || Subtarget.hasMMA()) && \"Type unsupported without MMA\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10606, __extension__ __PRETTY_FUNCTION__)) | |||
10606 | "Type unsupported without MMA")(static_cast <bool> ((VT != MVT::v512i1 || Subtarget.hasMMA ()) && "Type unsupported without MMA") ? void (0) : __assert_fail ("(VT != MVT::v512i1 || Subtarget.hasMMA()) && \"Type unsupported without MMA\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10606, __extension__ __PRETTY_FUNCTION__)); | |||
10607 | assert((VT != MVT::v256i1 || Subtarget.pairedVectorMemops()) &&(static_cast <bool> ((VT != MVT::v256i1 || Subtarget.pairedVectorMemops ()) && "Type unsupported without paired vector support" ) ? void (0) : __assert_fail ("(VT != MVT::v256i1 || Subtarget.pairedVectorMemops()) && \"Type unsupported without paired vector support\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10608, __extension__ __PRETTY_FUNCTION__)) | |||
10608 | "Type unsupported without paired vector support")(static_cast <bool> ((VT != MVT::v256i1 || Subtarget.pairedVectorMemops ()) && "Type unsupported without paired vector support" ) ? void (0) : __assert_fail ("(VT != MVT::v256i1 || Subtarget.pairedVectorMemops()) && \"Type unsupported without paired vector support\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10608, __extension__ __PRETTY_FUNCTION__)); | |||
10609 | Align Alignment = LN->getAlign(); | |||
10610 | SmallVector<SDValue, 4> Loads; | |||
10611 | SmallVector<SDValue, 4> LoadChains; | |||
10612 | unsigned NumVecs = VT.getSizeInBits() / 128; | |||
10613 | for (unsigned Idx = 0; Idx < NumVecs; ++Idx) { | |||
10614 | SDValue Load = | |||
10615 | DAG.getLoad(MVT::v16i8, dl, LoadChain, BasePtr, | |||
10616 | LN->getPointerInfo().getWithOffset(Idx * 16), | |||
10617 | commonAlignment(Alignment, Idx * 16), | |||
10618 | LN->getMemOperand()->getFlags(), LN->getAAInfo()); | |||
10619 | BasePtr = DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), BasePtr, | |||
10620 | DAG.getConstant(16, dl, BasePtr.getValueType())); | |||
10621 | Loads.push_back(Load); | |||
10622 | LoadChains.push_back(Load.getValue(1)); | |||
10623 | } | |||
10624 | if (Subtarget.isLittleEndian()) { | |||
10625 | std::reverse(Loads.begin(), Loads.end()); | |||
10626 | std::reverse(LoadChains.begin(), LoadChains.end()); | |||
10627 | } | |||
10628 | SDValue TF = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, LoadChains); | |||
10629 | SDValue Value = | |||
10630 | DAG.getNode(VT == MVT::v512i1 ? PPCISD::ACC_BUILD : PPCISD::PAIR_BUILD, | |||
10631 | dl, VT, Loads); | |||
10632 | SDValue RetOps[] = {Value, TF}; | |||
10633 | return DAG.getMergeValues(RetOps, dl); | |||
10634 | } | |||
10635 | ||||
10636 | SDValue PPCTargetLowering::LowerVectorStore(SDValue Op, | |||
10637 | SelectionDAG &DAG) const { | |||
10638 | SDLoc dl(Op); | |||
10639 | StoreSDNode *SN = cast<StoreSDNode>(Op.getNode()); | |||
10640 | SDValue StoreChain = SN->getChain(); | |||
10641 | SDValue BasePtr = SN->getBasePtr(); | |||
10642 | SDValue Value = SN->getValue(); | |||
10643 | EVT StoreVT = Value.getValueType(); | |||
10644 | ||||
10645 | if (StoreVT != MVT::v256i1 && StoreVT != MVT::v512i1) | |||
10646 | return Op; | |||
10647 | ||||
10648 | // Type v256i1 is used for pairs and v512i1 is used for accumulators. | |||
10649 | // Here we create 2 or 4 v16i8 stores to store the pair or accumulator | |||
10650 | // underlying registers individually. | |||
10651 | assert((StoreVT != MVT::v512i1 || Subtarget.hasMMA()) &&(static_cast <bool> ((StoreVT != MVT::v512i1 || Subtarget .hasMMA()) && "Type unsupported without MMA") ? void ( 0) : __assert_fail ("(StoreVT != MVT::v512i1 || Subtarget.hasMMA()) && \"Type unsupported without MMA\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10652, __extension__ __PRETTY_FUNCTION__)) | |||
10652 | "Type unsupported without MMA")(static_cast <bool> ((StoreVT != MVT::v512i1 || Subtarget .hasMMA()) && "Type unsupported without MMA") ? void ( 0) : __assert_fail ("(StoreVT != MVT::v512i1 || Subtarget.hasMMA()) && \"Type unsupported without MMA\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10652, __extension__ __PRETTY_FUNCTION__)); | |||
10653 | assert((StoreVT != MVT::v256i1 || Subtarget.pairedVectorMemops()) &&(static_cast <bool> ((StoreVT != MVT::v256i1 || Subtarget .pairedVectorMemops()) && "Type unsupported without paired vector support" ) ? void (0) : __assert_fail ("(StoreVT != MVT::v256i1 || Subtarget.pairedVectorMemops()) && \"Type unsupported without paired vector support\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10654, __extension__ __PRETTY_FUNCTION__)) | |||
10654 | "Type unsupported without paired vector support")(static_cast <bool> ((StoreVT != MVT::v256i1 || Subtarget .pairedVectorMemops()) && "Type unsupported without paired vector support" ) ? void (0) : __assert_fail ("(StoreVT != MVT::v256i1 || Subtarget.pairedVectorMemops()) && \"Type unsupported without paired vector support\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10654, __extension__ __PRETTY_FUNCTION__)); | |||
10655 | Align Alignment = SN->getAlign(); | |||
10656 | SmallVector<SDValue, 4> Stores; | |||
10657 | unsigned NumVecs = 2; | |||
10658 | if (StoreVT == MVT::v512i1) { | |||
10659 | Value = DAG.getNode(PPCISD::XXMFACC, dl, MVT::v512i1, Value); | |||
10660 | NumVecs = 4; | |||
10661 | } | |||
10662 | for (unsigned Idx = 0; Idx < NumVecs; ++Idx) { | |||
10663 | unsigned VecNum = Subtarget.isLittleEndian() ? NumVecs - 1 - Idx : Idx; | |||
10664 | SDValue Elt = DAG.getNode(PPCISD::EXTRACT_VSX_REG, dl, MVT::v16i8, Value, | |||
10665 | DAG.getConstant(VecNum, dl, getPointerTy(DAG.getDataLayout()))); | |||
10666 | SDValue Store = | |||
10667 | DAG.getStore(StoreChain, dl, Elt, BasePtr, | |||
10668 | SN->getPointerInfo().getWithOffset(Idx * 16), | |||
10669 | commonAlignment(Alignment, Idx * 16), | |||
10670 | SN->getMemOperand()->getFlags(), SN->getAAInfo()); | |||
10671 | BasePtr = DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), BasePtr, | |||
10672 | DAG.getConstant(16, dl, BasePtr.getValueType())); | |||
10673 | Stores.push_back(Store); | |||
10674 | } | |||
10675 | SDValue TF = DAG.getTokenFactor(dl, Stores); | |||
10676 | return TF; | |||
10677 | } | |||
10678 | ||||
10679 | SDValue PPCTargetLowering::LowerMUL(SDValue Op, SelectionDAG &DAG) const { | |||
10680 | SDLoc dl(Op); | |||
10681 | if (Op.getValueType() == MVT::v4i32) { | |||
10682 | SDValue LHS = Op.getOperand(0), RHS = Op.getOperand(1); | |||
10683 | ||||
10684 | SDValue Zero = getCanonicalConstSplat(0, 1, MVT::v4i32, DAG, dl); | |||
10685 | // +16 as shift amt. | |||
10686 | SDValue Neg16 = getCanonicalConstSplat(-16, 4, MVT::v4i32, DAG, dl); | |||
10687 | SDValue RHSSwap = // = vrlw RHS, 16 | |||
10688 | BuildIntrinsicOp(Intrinsic::ppc_altivec_vrlw, RHS, Neg16, DAG, dl); | |||
10689 | ||||
10690 | // Shrinkify inputs to v8i16. | |||
10691 | LHS = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, LHS); | |||
10692 | RHS = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, RHS); | |||
10693 | RHSSwap = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, RHSSwap); | |||
10694 | ||||
10695 | // Low parts multiplied together, generating 32-bit results (we ignore the | |||
10696 | // top parts). | |||
10697 | SDValue LoProd = BuildIntrinsicOp(Intrinsic::ppc_altivec_vmulouh, | |||
10698 | LHS, RHS, DAG, dl, MVT::v4i32); | |||
10699 | ||||
10700 | SDValue HiProd = BuildIntrinsicOp(Intrinsic::ppc_altivec_vmsumuhm, | |||
10701 | LHS, RHSSwap, Zero, DAG, dl, MVT::v4i32); | |||
10702 | // Shift the high parts up 16 bits. | |||
10703 | HiProd = BuildIntrinsicOp(Intrinsic::ppc_altivec_vslw, HiProd, | |||
10704 | Neg16, DAG, dl); | |||
10705 | return DAG.getNode(ISD::ADD, dl, MVT::v4i32, LoProd, HiProd); | |||
10706 | } else if (Op.getValueType() == MVT::v16i8) { | |||
10707 | SDValue LHS = Op.getOperand(0), RHS = Op.getOperand(1); | |||
10708 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
10709 | ||||
10710 | // Multiply the even 8-bit parts, producing 16-bit sums. | |||
10711 | SDValue EvenParts = BuildIntrinsicOp(Intrinsic::ppc_altivec_vmuleub, | |||
10712 | LHS, RHS, DAG, dl, MVT::v8i16); | |||
10713 | EvenParts = DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, EvenParts); | |||
10714 | ||||
10715 | // Multiply the odd 8-bit parts, producing 16-bit sums. | |||
10716 | SDValue OddParts = BuildIntrinsicOp(Intrinsic::ppc_altivec_vmuloub, | |||
10717 | LHS, RHS, DAG, dl, MVT::v8i16); | |||
10718 | OddParts = DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, OddParts); | |||
10719 | ||||
10720 | // Merge the results together. Because vmuleub and vmuloub are | |||
10721 | // instructions with a big-endian bias, we must reverse the | |||
10722 | // element numbering and reverse the meaning of "odd" and "even" | |||
10723 | // when generating little endian code. | |||
10724 | int Ops[16]; | |||
10725 | for (unsigned i = 0; i != 8; ++i) { | |||
10726 | if (isLittleEndian) { | |||
10727 | Ops[i*2 ] = 2*i; | |||
10728 | Ops[i*2+1] = 2*i+16; | |||
10729 | } else { | |||
10730 | Ops[i*2 ] = 2*i+1; | |||
10731 | Ops[i*2+1] = 2*i+1+16; | |||
10732 | } | |||
10733 | } | |||
10734 | if (isLittleEndian) | |||
10735 | return DAG.getVectorShuffle(MVT::v16i8, dl, OddParts, EvenParts, Ops); | |||
10736 | else | |||
10737 | return DAG.getVectorShuffle(MVT::v16i8, dl, EvenParts, OddParts, Ops); | |||
10738 | } else { | |||
10739 | llvm_unreachable("Unknown mul to lower!")::llvm::llvm_unreachable_internal("Unknown mul to lower!", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10739); | |||
10740 | } | |||
10741 | } | |||
10742 | ||||
10743 | SDValue PPCTargetLowering::LowerFP_ROUND(SDValue Op, SelectionDAG &DAG) const { | |||
10744 | bool IsStrict = Op->isStrictFPOpcode(); | |||
10745 | if (Op.getOperand(IsStrict ? 1 : 0).getValueType() == MVT::f128 && | |||
10746 | !Subtarget.hasP9Vector()) | |||
10747 | return SDValue(); | |||
10748 | ||||
10749 | return Op; | |||
10750 | } | |||
10751 | ||||
10752 | // Custom lowering for fpext vf32 to v2f64 | |||
10753 | SDValue PPCTargetLowering::LowerFP_EXTEND(SDValue Op, SelectionDAG &DAG) const { | |||
10754 | ||||
10755 | assert(Op.getOpcode() == ISD::FP_EXTEND &&(static_cast <bool> (Op.getOpcode() == ISD::FP_EXTEND && "Should only be called for ISD::FP_EXTEND") ? void (0) : __assert_fail ("Op.getOpcode() == ISD::FP_EXTEND && \"Should only be called for ISD::FP_EXTEND\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10756, __extension__ __PRETTY_FUNCTION__)) | |||
10756 | "Should only be called for ISD::FP_EXTEND")(static_cast <bool> (Op.getOpcode() == ISD::FP_EXTEND && "Should only be called for ISD::FP_EXTEND") ? void (0) : __assert_fail ("Op.getOpcode() == ISD::FP_EXTEND && \"Should only be called for ISD::FP_EXTEND\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10756, __extension__ __PRETTY_FUNCTION__)); | |||
10757 | ||||
10758 | // FIXME: handle extends from half precision float vectors on P9. | |||
10759 | // We only want to custom lower an extend from v2f32 to v2f64. | |||
10760 | if (Op.getValueType() != MVT::v2f64 || | |||
10761 | Op.getOperand(0).getValueType() != MVT::v2f32) | |||
10762 | return SDValue(); | |||
10763 | ||||
10764 | SDLoc dl(Op); | |||
10765 | SDValue Op0 = Op.getOperand(0); | |||
10766 | ||||
10767 | switch (Op0.getOpcode()) { | |||
10768 | default: | |||
10769 | return SDValue(); | |||
10770 | case ISD::EXTRACT_SUBVECTOR: { | |||
10771 | assert(Op0.getNumOperands() == 2 &&(static_cast <bool> (Op0.getNumOperands() == 2 && isa<ConstantSDNode>(Op0->getOperand(1)) && "Node should have 2 operands with second one being a constant!" ) ? void (0) : __assert_fail ("Op0.getNumOperands() == 2 && isa<ConstantSDNode>(Op0->getOperand(1)) && \"Node should have 2 operands with second one being a constant!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10773, __extension__ __PRETTY_FUNCTION__)) | |||
10772 | isa<ConstantSDNode>(Op0->getOperand(1)) &&(static_cast <bool> (Op0.getNumOperands() == 2 && isa<ConstantSDNode>(Op0->getOperand(1)) && "Node should have 2 operands with second one being a constant!" ) ? void (0) : __assert_fail ("Op0.getNumOperands() == 2 && isa<ConstantSDNode>(Op0->getOperand(1)) && \"Node should have 2 operands with second one being a constant!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10773, __extension__ __PRETTY_FUNCTION__)) | |||
10773 | "Node should have 2 operands with second one being a constant!")(static_cast <bool> (Op0.getNumOperands() == 2 && isa<ConstantSDNode>(Op0->getOperand(1)) && "Node should have 2 operands with second one being a constant!" ) ? void (0) : __assert_fail ("Op0.getNumOperands() == 2 && isa<ConstantSDNode>(Op0->getOperand(1)) && \"Node should have 2 operands with second one being a constant!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10773, __extension__ __PRETTY_FUNCTION__)); | |||
10774 | ||||
10775 | if (Op0.getOperand(0).getValueType() != MVT::v4f32) | |||
10776 | return SDValue(); | |||
10777 | ||||
10778 | // Custom lower is only done for high or low doubleword. | |||
10779 | int Idx = cast<ConstantSDNode>(Op0.getOperand(1))->getZExtValue(); | |||
10780 | if (Idx % 2 != 0) | |||
10781 | return SDValue(); | |||
10782 | ||||
10783 | // Since input is v4f32, at this point Idx is either 0 or 2. | |||
10784 | // Shift to get the doubleword position we want. | |||
10785 | int DWord = Idx >> 1; | |||
10786 | ||||
10787 | // High and low word positions are different on little endian. | |||
10788 | if (Subtarget.isLittleEndian()) | |||
10789 | DWord ^= 0x1; | |||
10790 | ||||
10791 | return DAG.getNode(PPCISD::FP_EXTEND_HALF, dl, MVT::v2f64, | |||
10792 | Op0.getOperand(0), DAG.getConstant(DWord, dl, MVT::i32)); | |||
10793 | } | |||
10794 | case ISD::FADD: | |||
10795 | case ISD::FMUL: | |||
10796 | case ISD::FSUB: { | |||
10797 | SDValue NewLoad[2]; | |||
10798 | for (unsigned i = 0, ie = Op0.getNumOperands(); i != ie; ++i) { | |||
10799 | // Ensure both input are loads. | |||
10800 | SDValue LdOp = Op0.getOperand(i); | |||
10801 | if (LdOp.getOpcode() != ISD::LOAD) | |||
10802 | return SDValue(); | |||
10803 | // Generate new load node. | |||
10804 | LoadSDNode *LD = cast<LoadSDNode>(LdOp); | |||
10805 | SDValue LoadOps[] = {LD->getChain(), LD->getBasePtr()}; | |||
10806 | NewLoad[i] = DAG.getMemIntrinsicNode( | |||
10807 | PPCISD::LD_VSX_LH, dl, DAG.getVTList(MVT::v4f32, MVT::Other), LoadOps, | |||
10808 | LD->getMemoryVT(), LD->getMemOperand()); | |||
10809 | } | |||
10810 | SDValue NewOp = | |||
10811 | DAG.getNode(Op0.getOpcode(), SDLoc(Op0), MVT::v4f32, NewLoad[0], | |||
10812 | NewLoad[1], Op0.getNode()->getFlags()); | |||
10813 | return DAG.getNode(PPCISD::FP_EXTEND_HALF, dl, MVT::v2f64, NewOp, | |||
10814 | DAG.getConstant(0, dl, MVT::i32)); | |||
10815 | } | |||
10816 | case ISD::LOAD: { | |||
10817 | LoadSDNode *LD = cast<LoadSDNode>(Op0); | |||
10818 | SDValue LoadOps[] = {LD->getChain(), LD->getBasePtr()}; | |||
10819 | SDValue NewLd = DAG.getMemIntrinsicNode( | |||
10820 | PPCISD::LD_VSX_LH, dl, DAG.getVTList(MVT::v4f32, MVT::Other), LoadOps, | |||
10821 | LD->getMemoryVT(), LD->getMemOperand()); | |||
10822 | return DAG.getNode(PPCISD::FP_EXTEND_HALF, dl, MVT::v2f64, NewLd, | |||
10823 | DAG.getConstant(0, dl, MVT::i32)); | |||
10824 | } | |||
10825 | } | |||
10826 | llvm_unreachable("ERROR:Should return for all cases within swtich.")::llvm::llvm_unreachable_internal("ERROR:Should return for all cases within swtich." , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10826); | |||
10827 | } | |||
10828 | ||||
10829 | /// LowerOperation - Provide custom lowering hooks for some operations. | |||
10830 | /// | |||
10831 | SDValue PPCTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const { | |||
10832 | switch (Op.getOpcode()) { | |||
10833 | default: llvm_unreachable("Wasn't expecting to be able to lower this!")::llvm::llvm_unreachable_internal("Wasn't expecting to be able to lower this!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10833); | |||
10834 | case ISD::ConstantPool: return LowerConstantPool(Op, DAG); | |||
10835 | case ISD::BlockAddress: return LowerBlockAddress(Op, DAG); | |||
10836 | case ISD::GlobalAddress: return LowerGlobalAddress(Op, DAG); | |||
10837 | case ISD::GlobalTLSAddress: return LowerGlobalTLSAddress(Op, DAG); | |||
10838 | case ISD::JumpTable: return LowerJumpTable(Op, DAG); | |||
10839 | case ISD::STRICT_FSETCC: | |||
10840 | case ISD::STRICT_FSETCCS: | |||
10841 | case ISD::SETCC: return LowerSETCC(Op, DAG); | |||
10842 | case ISD::INIT_TRAMPOLINE: return LowerINIT_TRAMPOLINE(Op, DAG); | |||
10843 | case ISD::ADJUST_TRAMPOLINE: return LowerADJUST_TRAMPOLINE(Op, DAG); | |||
10844 | ||||
10845 | case ISD::INLINEASM: | |||
10846 | case ISD::INLINEASM_BR: return LowerINLINEASM(Op, DAG); | |||
10847 | // Variable argument lowering. | |||
10848 | case ISD::VASTART: return LowerVASTART(Op, DAG); | |||
10849 | case ISD::VAARG: return LowerVAARG(Op, DAG); | |||
10850 | case ISD::VACOPY: return LowerVACOPY(Op, DAG); | |||
10851 | ||||
10852 | case ISD::STACKRESTORE: return LowerSTACKRESTORE(Op, DAG); | |||
10853 | case ISD::DYNAMIC_STACKALLOC: return LowerDYNAMIC_STACKALLOC(Op, DAG); | |||
10854 | case ISD::GET_DYNAMIC_AREA_OFFSET: | |||
10855 | return LowerGET_DYNAMIC_AREA_OFFSET(Op, DAG); | |||
10856 | ||||
10857 | // Exception handling lowering. | |||
10858 | case ISD::EH_DWARF_CFA: return LowerEH_DWARF_CFA(Op, DAG); | |||
10859 | case ISD::EH_SJLJ_SETJMP: return lowerEH_SJLJ_SETJMP(Op, DAG); | |||
10860 | case ISD::EH_SJLJ_LONGJMP: return lowerEH_SJLJ_LONGJMP(Op, DAG); | |||
10861 | ||||
10862 | case ISD::LOAD: return LowerLOAD(Op, DAG); | |||
10863 | case ISD::STORE: return LowerSTORE(Op, DAG); | |||
10864 | case ISD::TRUNCATE: return LowerTRUNCATE(Op, DAG); | |||
10865 | case ISD::SELECT_CC: return LowerSELECT_CC(Op, DAG); | |||
10866 | case ISD::STRICT_FP_TO_UINT: | |||
10867 | case ISD::STRICT_FP_TO_SINT: | |||
10868 | case ISD::FP_TO_UINT: | |||
10869 | case ISD::FP_TO_SINT: return LowerFP_TO_INT(Op, DAG, SDLoc(Op)); | |||
10870 | case ISD::STRICT_UINT_TO_FP: | |||
10871 | case ISD::STRICT_SINT_TO_FP: | |||
10872 | case ISD::UINT_TO_FP: | |||
10873 | case ISD::SINT_TO_FP: return LowerINT_TO_FP(Op, DAG); | |||
10874 | case ISD::FLT_ROUNDS_: return LowerFLT_ROUNDS_(Op, DAG); | |||
10875 | ||||
10876 | // Lower 64-bit shifts. | |||
10877 | case ISD::SHL_PARTS: return LowerSHL_PARTS(Op, DAG); | |||
10878 | case ISD::SRL_PARTS: return LowerSRL_PARTS(Op, DAG); | |||
10879 | case ISD::SRA_PARTS: return LowerSRA_PARTS(Op, DAG); | |||
10880 | ||||
10881 | case ISD::FSHL: return LowerFunnelShift(Op, DAG); | |||
10882 | case ISD::FSHR: return LowerFunnelShift(Op, DAG); | |||
10883 | ||||
10884 | // Vector-related lowering. | |||
10885 | case ISD::BUILD_VECTOR: return LowerBUILD_VECTOR(Op, DAG); | |||
10886 | case ISD::VECTOR_SHUFFLE: return LowerVECTOR_SHUFFLE(Op, DAG); | |||
10887 | case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG); | |||
10888 | case ISD::SCALAR_TO_VECTOR: return LowerSCALAR_TO_VECTOR(Op, DAG); | |||
10889 | case ISD::INSERT_VECTOR_ELT: return LowerINSERT_VECTOR_ELT(Op, DAG); | |||
10890 | case ISD::MUL: return LowerMUL(Op, DAG); | |||
10891 | case ISD::FP_EXTEND: return LowerFP_EXTEND(Op, DAG); | |||
10892 | case ISD::STRICT_FP_ROUND: | |||
10893 | case ISD::FP_ROUND: | |||
10894 | return LowerFP_ROUND(Op, DAG); | |||
10895 | case ISD::ROTL: return LowerROTL(Op, DAG); | |||
10896 | ||||
10897 | // For counter-based loop handling. | |||
10898 | case ISD::INTRINSIC_W_CHAIN: return SDValue(); | |||
10899 | ||||
10900 | case ISD::BITCAST: return LowerBITCAST(Op, DAG); | |||
10901 | ||||
10902 | // Frame & Return address. | |||
10903 | case ISD::RETURNADDR: return LowerRETURNADDR(Op, DAG); | |||
10904 | case ISD::FRAMEADDR: return LowerFRAMEADDR(Op, DAG); | |||
10905 | ||||
10906 | case ISD::INTRINSIC_VOID: | |||
10907 | return LowerINTRINSIC_VOID(Op, DAG); | |||
10908 | case ISD::BSWAP: | |||
10909 | return LowerBSWAP(Op, DAG); | |||
10910 | case ISD::ATOMIC_CMP_SWAP: | |||
10911 | return LowerATOMIC_CMP_SWAP(Op, DAG); | |||
10912 | } | |||
10913 | } | |||
10914 | ||||
10915 | void PPCTargetLowering::ReplaceNodeResults(SDNode *N, | |||
10916 | SmallVectorImpl<SDValue>&Results, | |||
10917 | SelectionDAG &DAG) const { | |||
10918 | SDLoc dl(N); | |||
10919 | switch (N->getOpcode()) { | |||
10920 | default: | |||
10921 | llvm_unreachable("Do not know how to custom type legalize this operation!")::llvm::llvm_unreachable_internal("Do not know how to custom type legalize this operation!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10921); | |||
10922 | case ISD::READCYCLECOUNTER: { | |||
10923 | SDVTList VTs = DAG.getVTList(MVT::i32, MVT::i32, MVT::Other); | |||
10924 | SDValue RTB = DAG.getNode(PPCISD::READ_TIME_BASE, dl, VTs, N->getOperand(0)); | |||
10925 | ||||
10926 | Results.push_back( | |||
10927 | DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, RTB, RTB.getValue(1))); | |||
10928 | Results.push_back(RTB.getValue(2)); | |||
10929 | break; | |||
10930 | } | |||
10931 | case ISD::INTRINSIC_W_CHAIN: { | |||
10932 | if (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue() != | |||
10933 | Intrinsic::loop_decrement) | |||
10934 | break; | |||
10935 | ||||
10936 | assert(N->getValueType(0) == MVT::i1 &&(static_cast <bool> (N->getValueType(0) == MVT::i1 && "Unexpected result type for CTR decrement intrinsic") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::i1 && \"Unexpected result type for CTR decrement intrinsic\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10937, __extension__ __PRETTY_FUNCTION__)) | |||
10937 | "Unexpected result type for CTR decrement intrinsic")(static_cast <bool> (N->getValueType(0) == MVT::i1 && "Unexpected result type for CTR decrement intrinsic") ? void (0) : __assert_fail ("N->getValueType(0) == MVT::i1 && \"Unexpected result type for CTR decrement intrinsic\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 10937, __extension__ __PRETTY_FUNCTION__)); | |||
10938 | EVT SVT = getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), | |||
10939 | N->getValueType(0)); | |||
10940 | SDVTList VTs = DAG.getVTList(SVT, MVT::Other); | |||
10941 | SDValue NewInt = DAG.getNode(N->getOpcode(), dl, VTs, N->getOperand(0), | |||
10942 | N->getOperand(1)); | |||
10943 | ||||
10944 | Results.push_back(DAG.getNode(ISD::TRUNCATE, dl, MVT::i1, NewInt)); | |||
10945 | Results.push_back(NewInt.getValue(1)); | |||
10946 | break; | |||
10947 | } | |||
10948 | case ISD::VAARG: { | |||
10949 | if (!Subtarget.isSVR4ABI() || Subtarget.isPPC64()) | |||
10950 | return; | |||
10951 | ||||
10952 | EVT VT = N->getValueType(0); | |||
10953 | ||||
10954 | if (VT == MVT::i64) { | |||
10955 | SDValue NewNode = LowerVAARG(SDValue(N, 1), DAG); | |||
10956 | ||||
10957 | Results.push_back(NewNode); | |||
10958 | Results.push_back(NewNode.getValue(1)); | |||
10959 | } | |||
10960 | return; | |||
10961 | } | |||
10962 | case ISD::STRICT_FP_TO_SINT: | |||
10963 | case ISD::STRICT_FP_TO_UINT: | |||
10964 | case ISD::FP_TO_SINT: | |||
10965 | case ISD::FP_TO_UINT: | |||
10966 | // LowerFP_TO_INT() can only handle f32 and f64. | |||
10967 | if (N->getOperand(N->isStrictFPOpcode() ? 1 : 0).getValueType() == | |||
10968 | MVT::ppcf128) | |||
10969 | return; | |||
10970 | Results.push_back(LowerFP_TO_INT(SDValue(N, 0), DAG, dl)); | |||
10971 | return; | |||
10972 | case ISD::TRUNCATE: { | |||
10973 | if (!N->getValueType(0).isVector()) | |||
10974 | return; | |||
10975 | SDValue Lowered = LowerTRUNCATEVector(SDValue(N, 0), DAG); | |||
10976 | if (Lowered) | |||
10977 | Results.push_back(Lowered); | |||
10978 | return; | |||
10979 | } | |||
10980 | case ISD::FSHL: | |||
10981 | case ISD::FSHR: | |||
10982 | // Don't handle funnel shifts here. | |||
10983 | return; | |||
10984 | case ISD::BITCAST: | |||
10985 | // Don't handle bitcast here. | |||
10986 | return; | |||
10987 | case ISD::FP_EXTEND: | |||
10988 | SDValue Lowered = LowerFP_EXTEND(SDValue(N, 0), DAG); | |||
10989 | if (Lowered) | |||
10990 | Results.push_back(Lowered); | |||
10991 | return; | |||
10992 | } | |||
10993 | } | |||
10994 | ||||
10995 | //===----------------------------------------------------------------------===// | |||
10996 | // Other Lowering Code | |||
10997 | //===----------------------------------------------------------------------===// | |||
10998 | ||||
10999 | static Instruction *callIntrinsic(IRBuilderBase &Builder, Intrinsic::ID Id) { | |||
11000 | Module *M = Builder.GetInsertBlock()->getParent()->getParent(); | |||
11001 | Function *Func = Intrinsic::getDeclaration(M, Id); | |||
11002 | return Builder.CreateCall(Func, {}); | |||
11003 | } | |||
11004 | ||||
11005 | // The mappings for emitLeading/TrailingFence is taken from | |||
11006 | // http://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html | |||
11007 | Instruction *PPCTargetLowering::emitLeadingFence(IRBuilderBase &Builder, | |||
11008 | Instruction *Inst, | |||
11009 | AtomicOrdering Ord) const { | |||
11010 | if (Ord == AtomicOrdering::SequentiallyConsistent) | |||
11011 | return callIntrinsic(Builder, Intrinsic::ppc_sync); | |||
11012 | if (isReleaseOrStronger(Ord)) | |||
11013 | return callIntrinsic(Builder, Intrinsic::ppc_lwsync); | |||
11014 | return nullptr; | |||
11015 | } | |||
11016 | ||||
11017 | Instruction *PPCTargetLowering::emitTrailingFence(IRBuilderBase &Builder, | |||
11018 | Instruction *Inst, | |||
11019 | AtomicOrdering Ord) const { | |||
11020 | if (Inst->hasAtomicLoad() && isAcquireOrStronger(Ord)) { | |||
11021 | // See http://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html and | |||
11022 | // http://www.rdrop.com/users/paulmck/scalability/paper/N2745r.2011.03.04a.html | |||
11023 | // and http://www.cl.cam.ac.uk/~pes20/cppppc/ for justification. | |||
11024 | if (isa<LoadInst>(Inst) && Subtarget.isPPC64()) | |||
11025 | return Builder.CreateCall( | |||
11026 | Intrinsic::getDeclaration( | |||
11027 | Builder.GetInsertBlock()->getParent()->getParent(), | |||
11028 | Intrinsic::ppc_cfence, {Inst->getType()}), | |||
11029 | {Inst}); | |||
11030 | // FIXME: Can use isync for rmw operation. | |||
11031 | return callIntrinsic(Builder, Intrinsic::ppc_lwsync); | |||
11032 | } | |||
11033 | return nullptr; | |||
11034 | } | |||
11035 | ||||
11036 | MachineBasicBlock * | |||
11037 | PPCTargetLowering::EmitAtomicBinary(MachineInstr &MI, MachineBasicBlock *BB, | |||
11038 | unsigned AtomicSize, | |||
11039 | unsigned BinOpcode, | |||
11040 | unsigned CmpOpcode, | |||
11041 | unsigned CmpPred) const { | |||
11042 | // This also handles ATOMIC_SWAP, indicated by BinOpcode==0. | |||
11043 | const TargetInstrInfo *TII = Subtarget.getInstrInfo(); | |||
11044 | ||||
11045 | auto LoadMnemonic = PPC::LDARX; | |||
11046 | auto StoreMnemonic = PPC::STDCX; | |||
11047 | switch (AtomicSize) { | |||
11048 | default: | |||
11049 | llvm_unreachable("Unexpected size of atomic entity")::llvm::llvm_unreachable_internal("Unexpected size of atomic entity" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11049); | |||
11050 | case 1: | |||
11051 | LoadMnemonic = PPC::LBARX; | |||
11052 | StoreMnemonic = PPC::STBCX; | |||
11053 | assert(Subtarget.hasPartwordAtomics() && "Call this only with size >=4")(static_cast <bool> (Subtarget.hasPartwordAtomics() && "Call this only with size >=4") ? void (0) : __assert_fail ("Subtarget.hasPartwordAtomics() && \"Call this only with size >=4\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11053, __extension__ __PRETTY_FUNCTION__)); | |||
11054 | break; | |||
11055 | case 2: | |||
11056 | LoadMnemonic = PPC::LHARX; | |||
11057 | StoreMnemonic = PPC::STHCX; | |||
11058 | assert(Subtarget.hasPartwordAtomics() && "Call this only with size >=4")(static_cast <bool> (Subtarget.hasPartwordAtomics() && "Call this only with size >=4") ? void (0) : __assert_fail ("Subtarget.hasPartwordAtomics() && \"Call this only with size >=4\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11058, __extension__ __PRETTY_FUNCTION__)); | |||
11059 | break; | |||
11060 | case 4: | |||
11061 | LoadMnemonic = PPC::LWARX; | |||
11062 | StoreMnemonic = PPC::STWCX; | |||
11063 | break; | |||
11064 | case 8: | |||
11065 | LoadMnemonic = PPC::LDARX; | |||
11066 | StoreMnemonic = PPC::STDCX; | |||
11067 | break; | |||
11068 | } | |||
11069 | ||||
11070 | const BasicBlock *LLVM_BB = BB->getBasicBlock(); | |||
11071 | MachineFunction *F = BB->getParent(); | |||
11072 | MachineFunction::iterator It = ++BB->getIterator(); | |||
11073 | ||||
11074 | Register dest = MI.getOperand(0).getReg(); | |||
11075 | Register ptrA = MI.getOperand(1).getReg(); | |||
11076 | Register ptrB = MI.getOperand(2).getReg(); | |||
11077 | Register incr = MI.getOperand(3).getReg(); | |||
11078 | DebugLoc dl = MI.getDebugLoc(); | |||
11079 | ||||
11080 | MachineBasicBlock *loopMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
11081 | MachineBasicBlock *loop2MBB = | |||
11082 | CmpOpcode ? F->CreateMachineBasicBlock(LLVM_BB) : nullptr; | |||
11083 | MachineBasicBlock *exitMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
11084 | F->insert(It, loopMBB); | |||
11085 | if (CmpOpcode) | |||
11086 | F->insert(It, loop2MBB); | |||
11087 | F->insert(It, exitMBB); | |||
11088 | exitMBB->splice(exitMBB->begin(), BB, | |||
11089 | std::next(MachineBasicBlock::iterator(MI)), BB->end()); | |||
11090 | exitMBB->transferSuccessorsAndUpdatePHIs(BB); | |||
11091 | ||||
11092 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
11093 | Register TmpReg = (!BinOpcode) ? incr : | |||
11094 | RegInfo.createVirtualRegister( AtomicSize == 8 ? &PPC::G8RCRegClass | |||
11095 | : &PPC::GPRCRegClass); | |||
11096 | ||||
11097 | // thisMBB: | |||
11098 | // ... | |||
11099 | // fallthrough --> loopMBB | |||
11100 | BB->addSuccessor(loopMBB); | |||
11101 | ||||
11102 | // loopMBB: | |||
11103 | // l[wd]arx dest, ptr | |||
11104 | // add r0, dest, incr | |||
11105 | // st[wd]cx. r0, ptr | |||
11106 | // bne- loopMBB | |||
11107 | // fallthrough --> exitMBB | |||
11108 | ||||
11109 | // For max/min... | |||
11110 | // loopMBB: | |||
11111 | // l[wd]arx dest, ptr | |||
11112 | // cmpl?[wd] incr, dest | |||
11113 | // bgt exitMBB | |||
11114 | // loop2MBB: | |||
11115 | // st[wd]cx. dest, ptr | |||
11116 | // bne- loopMBB | |||
11117 | // fallthrough --> exitMBB | |||
11118 | ||||
11119 | BB = loopMBB; | |||
11120 | BuildMI(BB, dl, TII->get(LoadMnemonic), dest) | |||
11121 | .addReg(ptrA).addReg(ptrB); | |||
11122 | if (BinOpcode) | |||
11123 | BuildMI(BB, dl, TII->get(BinOpcode), TmpReg).addReg(incr).addReg(dest); | |||
11124 | if (CmpOpcode) { | |||
11125 | // Signed comparisons of byte or halfword values must be sign-extended. | |||
11126 | if (CmpOpcode == PPC::CMPW && AtomicSize < 4) { | |||
11127 | Register ExtReg = RegInfo.createVirtualRegister(&PPC::GPRCRegClass); | |||
11128 | BuildMI(BB, dl, TII->get(AtomicSize == 1 ? PPC::EXTSB : PPC::EXTSH), | |||
11129 | ExtReg).addReg(dest); | |||
11130 | BuildMI(BB, dl, TII->get(CmpOpcode), PPC::CR0) | |||
11131 | .addReg(incr).addReg(ExtReg); | |||
11132 | } else | |||
11133 | BuildMI(BB, dl, TII->get(CmpOpcode), PPC::CR0) | |||
11134 | .addReg(incr).addReg(dest); | |||
11135 | ||||
11136 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
11137 | .addImm(CmpPred).addReg(PPC::CR0).addMBB(exitMBB); | |||
11138 | BB->addSuccessor(loop2MBB); | |||
11139 | BB->addSuccessor(exitMBB); | |||
11140 | BB = loop2MBB; | |||
11141 | } | |||
11142 | BuildMI(BB, dl, TII->get(StoreMnemonic)) | |||
11143 | .addReg(TmpReg).addReg(ptrA).addReg(ptrB); | |||
11144 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
11145 | .addImm(PPC::PRED_NE).addReg(PPC::CR0).addMBB(loopMBB); | |||
11146 | BB->addSuccessor(loopMBB); | |||
11147 | BB->addSuccessor(exitMBB); | |||
11148 | ||||
11149 | // exitMBB: | |||
11150 | // ... | |||
11151 | BB = exitMBB; | |||
11152 | return BB; | |||
11153 | } | |||
11154 | ||||
11155 | static bool isSignExtended(MachineInstr &MI, const PPCInstrInfo *TII) { | |||
11156 | switch(MI.getOpcode()) { | |||
11157 | default: | |||
11158 | return false; | |||
11159 | case PPC::COPY: | |||
11160 | return TII->isSignExtended(MI); | |||
11161 | case PPC::LHA: | |||
11162 | case PPC::LHA8: | |||
11163 | case PPC::LHAU: | |||
11164 | case PPC::LHAU8: | |||
11165 | case PPC::LHAUX: | |||
11166 | case PPC::LHAUX8: | |||
11167 | case PPC::LHAX: | |||
11168 | case PPC::LHAX8: | |||
11169 | case PPC::LWA: | |||
11170 | case PPC::LWAUX: | |||
11171 | case PPC::LWAX: | |||
11172 | case PPC::LWAX_32: | |||
11173 | case PPC::LWA_32: | |||
11174 | case PPC::PLHA: | |||
11175 | case PPC::PLHA8: | |||
11176 | case PPC::PLHA8pc: | |||
11177 | case PPC::PLHApc: | |||
11178 | case PPC::PLWA: | |||
11179 | case PPC::PLWA8: | |||
11180 | case PPC::PLWA8pc: | |||
11181 | case PPC::PLWApc: | |||
11182 | case PPC::EXTSB: | |||
11183 | case PPC::EXTSB8: | |||
11184 | case PPC::EXTSB8_32_64: | |||
11185 | case PPC::EXTSB8_rec: | |||
11186 | case PPC::EXTSB_rec: | |||
11187 | case PPC::EXTSH: | |||
11188 | case PPC::EXTSH8: | |||
11189 | case PPC::EXTSH8_32_64: | |||
11190 | case PPC::EXTSH8_rec: | |||
11191 | case PPC::EXTSH_rec: | |||
11192 | case PPC::EXTSW: | |||
11193 | case PPC::EXTSWSLI: | |||
11194 | case PPC::EXTSWSLI_32_64: | |||
11195 | case PPC::EXTSWSLI_32_64_rec: | |||
11196 | case PPC::EXTSWSLI_rec: | |||
11197 | case PPC::EXTSW_32: | |||
11198 | case PPC::EXTSW_32_64: | |||
11199 | case PPC::EXTSW_32_64_rec: | |||
11200 | case PPC::EXTSW_rec: | |||
11201 | case PPC::SRAW: | |||
11202 | case PPC::SRAWI: | |||
11203 | case PPC::SRAWI_rec: | |||
11204 | case PPC::SRAW_rec: | |||
11205 | return true; | |||
11206 | } | |||
11207 | return false; | |||
11208 | } | |||
11209 | ||||
11210 | MachineBasicBlock *PPCTargetLowering::EmitPartwordAtomicBinary( | |||
11211 | MachineInstr &MI, MachineBasicBlock *BB, | |||
11212 | bool is8bit, // operation | |||
11213 | unsigned BinOpcode, unsigned CmpOpcode, unsigned CmpPred) const { | |||
11214 | // This also handles ATOMIC_SWAP, indicated by BinOpcode==0. | |||
11215 | const PPCInstrInfo *TII = Subtarget.getInstrInfo(); | |||
11216 | ||||
11217 | // If this is a signed comparison and the value being compared is not known | |||
11218 | // to be sign extended, sign extend it here. | |||
11219 | DebugLoc dl = MI.getDebugLoc(); | |||
11220 | MachineFunction *F = BB->getParent(); | |||
11221 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
11222 | Register incr = MI.getOperand(3).getReg(); | |||
11223 | bool IsSignExtended = Register::isVirtualRegister(incr) && | |||
11224 | isSignExtended(*RegInfo.getVRegDef(incr), TII); | |||
11225 | ||||
11226 | if (CmpOpcode == PPC::CMPW && !IsSignExtended) { | |||
11227 | Register ValueReg = RegInfo.createVirtualRegister(&PPC::GPRCRegClass); | |||
11228 | BuildMI(*BB, MI, dl, TII->get(is8bit ? PPC::EXTSB : PPC::EXTSH), ValueReg) | |||
11229 | .addReg(MI.getOperand(3).getReg()); | |||
11230 | MI.getOperand(3).setReg(ValueReg); | |||
11231 | } | |||
11232 | // If we support part-word atomic mnemonics, just use them | |||
11233 | if (Subtarget.hasPartwordAtomics()) | |||
11234 | return EmitAtomicBinary(MI, BB, is8bit ? 1 : 2, BinOpcode, CmpOpcode, | |||
11235 | CmpPred); | |||
11236 | ||||
11237 | // In 64 bit mode we have to use 64 bits for addresses, even though the | |||
11238 | // lwarx/stwcx are 32 bits. With the 32-bit atomics we can use address | |||
11239 | // registers without caring whether they're 32 or 64, but here we're | |||
11240 | // doing actual arithmetic on the addresses. | |||
11241 | bool is64bit = Subtarget.isPPC64(); | |||
11242 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
11243 | unsigned ZeroReg = is64bit ? PPC::ZERO8 : PPC::ZERO; | |||
11244 | ||||
11245 | const BasicBlock *LLVM_BB = BB->getBasicBlock(); | |||
11246 | MachineFunction::iterator It = ++BB->getIterator(); | |||
11247 | ||||
11248 | Register dest = MI.getOperand(0).getReg(); | |||
11249 | Register ptrA = MI.getOperand(1).getReg(); | |||
11250 | Register ptrB = MI.getOperand(2).getReg(); | |||
11251 | ||||
11252 | MachineBasicBlock *loopMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
11253 | MachineBasicBlock *loop2MBB = | |||
11254 | CmpOpcode ? F->CreateMachineBasicBlock(LLVM_BB) : nullptr; | |||
11255 | MachineBasicBlock *exitMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
11256 | F->insert(It, loopMBB); | |||
11257 | if (CmpOpcode) | |||
11258 | F->insert(It, loop2MBB); | |||
11259 | F->insert(It, exitMBB); | |||
11260 | exitMBB->splice(exitMBB->begin(), BB, | |||
11261 | std::next(MachineBasicBlock::iterator(MI)), BB->end()); | |||
11262 | exitMBB->transferSuccessorsAndUpdatePHIs(BB); | |||
11263 | ||||
11264 | const TargetRegisterClass *RC = | |||
11265 | is64bit ? &PPC::G8RCRegClass : &PPC::GPRCRegClass; | |||
11266 | const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; | |||
11267 | ||||
11268 | Register PtrReg = RegInfo.createVirtualRegister(RC); | |||
11269 | Register Shift1Reg = RegInfo.createVirtualRegister(GPRC); | |||
11270 | Register ShiftReg = | |||
11271 | isLittleEndian ? Shift1Reg : RegInfo.createVirtualRegister(GPRC); | |||
11272 | Register Incr2Reg = RegInfo.createVirtualRegister(GPRC); | |||
11273 | Register MaskReg = RegInfo.createVirtualRegister(GPRC); | |||
11274 | Register Mask2Reg = RegInfo.createVirtualRegister(GPRC); | |||
11275 | Register Mask3Reg = RegInfo.createVirtualRegister(GPRC); | |||
11276 | Register Tmp2Reg = RegInfo.createVirtualRegister(GPRC); | |||
11277 | Register Tmp3Reg = RegInfo.createVirtualRegister(GPRC); | |||
11278 | Register Tmp4Reg = RegInfo.createVirtualRegister(GPRC); | |||
11279 | Register TmpDestReg = RegInfo.createVirtualRegister(GPRC); | |||
11280 | Register SrwDestReg = RegInfo.createVirtualRegister(GPRC); | |||
11281 | Register Ptr1Reg; | |||
11282 | Register TmpReg = | |||
11283 | (!BinOpcode) ? Incr2Reg : RegInfo.createVirtualRegister(GPRC); | |||
11284 | ||||
11285 | // thisMBB: | |||
11286 | // ... | |||
11287 | // fallthrough --> loopMBB | |||
11288 | BB->addSuccessor(loopMBB); | |||
11289 | ||||
11290 | // The 4-byte load must be aligned, while a char or short may be | |||
11291 | // anywhere in the word. Hence all this nasty bookkeeping code. | |||
11292 | // add ptr1, ptrA, ptrB [copy if ptrA==0] | |||
11293 | // rlwinm shift1, ptr1, 3, 27, 28 [3, 27, 27] | |||
11294 | // xori shift, shift1, 24 [16] | |||
11295 | // rlwinm ptr, ptr1, 0, 0, 29 | |||
11296 | // slw incr2, incr, shift | |||
11297 | // li mask2, 255 [li mask3, 0; ori mask2, mask3, 65535] | |||
11298 | // slw mask, mask2, shift | |||
11299 | // loopMBB: | |||
11300 | // lwarx tmpDest, ptr | |||
11301 | // add tmp, tmpDest, incr2 | |||
11302 | // andc tmp2, tmpDest, mask | |||
11303 | // and tmp3, tmp, mask | |||
11304 | // or tmp4, tmp3, tmp2 | |||
11305 | // stwcx. tmp4, ptr | |||
11306 | // bne- loopMBB | |||
11307 | // fallthrough --> exitMBB | |||
11308 | // srw SrwDest, tmpDest, shift | |||
11309 | // rlwinm SrwDest, SrwDest, 0, 24 [16], 31 | |||
11310 | if (ptrA != ZeroReg) { | |||
11311 | Ptr1Reg = RegInfo.createVirtualRegister(RC); | |||
11312 | BuildMI(BB, dl, TII->get(is64bit ? PPC::ADD8 : PPC::ADD4), Ptr1Reg) | |||
11313 | .addReg(ptrA) | |||
11314 | .addReg(ptrB); | |||
11315 | } else { | |||
11316 | Ptr1Reg = ptrB; | |||
11317 | } | |||
11318 | // We need use 32-bit subregister to avoid mismatch register class in 64-bit | |||
11319 | // mode. | |||
11320 | BuildMI(BB, dl, TII->get(PPC::RLWINM), Shift1Reg) | |||
11321 | .addReg(Ptr1Reg, 0, is64bit ? PPC::sub_32 : 0) | |||
11322 | .addImm(3) | |||
11323 | .addImm(27) | |||
11324 | .addImm(is8bit ? 28 : 27); | |||
11325 | if (!isLittleEndian) | |||
11326 | BuildMI(BB, dl, TII->get(PPC::XORI), ShiftReg) | |||
11327 | .addReg(Shift1Reg) | |||
11328 | .addImm(is8bit ? 24 : 16); | |||
11329 | if (is64bit) | |||
11330 | BuildMI(BB, dl, TII->get(PPC::RLDICR), PtrReg) | |||
11331 | .addReg(Ptr1Reg) | |||
11332 | .addImm(0) | |||
11333 | .addImm(61); | |||
11334 | else | |||
11335 | BuildMI(BB, dl, TII->get(PPC::RLWINM), PtrReg) | |||
11336 | .addReg(Ptr1Reg) | |||
11337 | .addImm(0) | |||
11338 | .addImm(0) | |||
11339 | .addImm(29); | |||
11340 | BuildMI(BB, dl, TII->get(PPC::SLW), Incr2Reg).addReg(incr).addReg(ShiftReg); | |||
11341 | if (is8bit) | |||
11342 | BuildMI(BB, dl, TII->get(PPC::LI), Mask2Reg).addImm(255); | |||
11343 | else { | |||
11344 | BuildMI(BB, dl, TII->get(PPC::LI), Mask3Reg).addImm(0); | |||
11345 | BuildMI(BB, dl, TII->get(PPC::ORI), Mask2Reg) | |||
11346 | .addReg(Mask3Reg) | |||
11347 | .addImm(65535); | |||
11348 | } | |||
11349 | BuildMI(BB, dl, TII->get(PPC::SLW), MaskReg) | |||
11350 | .addReg(Mask2Reg) | |||
11351 | .addReg(ShiftReg); | |||
11352 | ||||
11353 | BB = loopMBB; | |||
11354 | BuildMI(BB, dl, TII->get(PPC::LWARX), TmpDestReg) | |||
11355 | .addReg(ZeroReg) | |||
11356 | .addReg(PtrReg); | |||
11357 | if (BinOpcode) | |||
11358 | BuildMI(BB, dl, TII->get(BinOpcode), TmpReg) | |||
11359 | .addReg(Incr2Reg) | |||
11360 | .addReg(TmpDestReg); | |||
11361 | BuildMI(BB, dl, TII->get(PPC::ANDC), Tmp2Reg) | |||
11362 | .addReg(TmpDestReg) | |||
11363 | .addReg(MaskReg); | |||
11364 | BuildMI(BB, dl, TII->get(PPC::AND), Tmp3Reg).addReg(TmpReg).addReg(MaskReg); | |||
11365 | if (CmpOpcode) { | |||
11366 | // For unsigned comparisons, we can directly compare the shifted values. | |||
11367 | // For signed comparisons we shift and sign extend. | |||
11368 | Register SReg = RegInfo.createVirtualRegister(GPRC); | |||
11369 | BuildMI(BB, dl, TII->get(PPC::AND), SReg) | |||
11370 | .addReg(TmpDestReg) | |||
11371 | .addReg(MaskReg); | |||
11372 | unsigned ValueReg = SReg; | |||
11373 | unsigned CmpReg = Incr2Reg; | |||
11374 | if (CmpOpcode == PPC::CMPW) { | |||
11375 | ValueReg = RegInfo.createVirtualRegister(GPRC); | |||
11376 | BuildMI(BB, dl, TII->get(PPC::SRW), ValueReg) | |||
11377 | .addReg(SReg) | |||
11378 | .addReg(ShiftReg); | |||
11379 | Register ValueSReg = RegInfo.createVirtualRegister(GPRC); | |||
11380 | BuildMI(BB, dl, TII->get(is8bit ? PPC::EXTSB : PPC::EXTSH), ValueSReg) | |||
11381 | .addReg(ValueReg); | |||
11382 | ValueReg = ValueSReg; | |||
11383 | CmpReg = incr; | |||
11384 | } | |||
11385 | BuildMI(BB, dl, TII->get(CmpOpcode), PPC::CR0) | |||
11386 | .addReg(CmpReg) | |||
11387 | .addReg(ValueReg); | |||
11388 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
11389 | .addImm(CmpPred) | |||
11390 | .addReg(PPC::CR0) | |||
11391 | .addMBB(exitMBB); | |||
11392 | BB->addSuccessor(loop2MBB); | |||
11393 | BB->addSuccessor(exitMBB); | |||
11394 | BB = loop2MBB; | |||
11395 | } | |||
11396 | BuildMI(BB, dl, TII->get(PPC::OR), Tmp4Reg).addReg(Tmp3Reg).addReg(Tmp2Reg); | |||
11397 | BuildMI(BB, dl, TII->get(PPC::STWCX)) | |||
11398 | .addReg(Tmp4Reg) | |||
11399 | .addReg(ZeroReg) | |||
11400 | .addReg(PtrReg); | |||
11401 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
11402 | .addImm(PPC::PRED_NE) | |||
11403 | .addReg(PPC::CR0) | |||
11404 | .addMBB(loopMBB); | |||
11405 | BB->addSuccessor(loopMBB); | |||
11406 | BB->addSuccessor(exitMBB); | |||
11407 | ||||
11408 | // exitMBB: | |||
11409 | // ... | |||
11410 | BB = exitMBB; | |||
11411 | // Since the shift amount is not a constant, we need to clear | |||
11412 | // the upper bits with a separate RLWINM. | |||
11413 | BuildMI(*BB, BB->begin(), dl, TII->get(PPC::RLWINM), dest) | |||
11414 | .addReg(SrwDestReg) | |||
11415 | .addImm(0) | |||
11416 | .addImm(is8bit ? 24 : 16) | |||
11417 | .addImm(31); | |||
11418 | BuildMI(*BB, BB->begin(), dl, TII->get(PPC::SRW), SrwDestReg) | |||
11419 | .addReg(TmpDestReg) | |||
11420 | .addReg(ShiftReg); | |||
11421 | return BB; | |||
11422 | } | |||
11423 | ||||
11424 | llvm::MachineBasicBlock * | |||
11425 | PPCTargetLowering::emitEHSjLjSetJmp(MachineInstr &MI, | |||
11426 | MachineBasicBlock *MBB) const { | |||
11427 | DebugLoc DL = MI.getDebugLoc(); | |||
11428 | const TargetInstrInfo *TII = Subtarget.getInstrInfo(); | |||
11429 | const PPCRegisterInfo *TRI = Subtarget.getRegisterInfo(); | |||
11430 | ||||
11431 | MachineFunction *MF = MBB->getParent(); | |||
11432 | MachineRegisterInfo &MRI = MF->getRegInfo(); | |||
11433 | ||||
11434 | const BasicBlock *BB = MBB->getBasicBlock(); | |||
11435 | MachineFunction::iterator I = ++MBB->getIterator(); | |||
11436 | ||||
11437 | Register DstReg = MI.getOperand(0).getReg(); | |||
11438 | const TargetRegisterClass *RC = MRI.getRegClass(DstReg); | |||
11439 | assert(TRI->isTypeLegalForClass(*RC, MVT::i32) && "Invalid destination!")(static_cast <bool> (TRI->isTypeLegalForClass(*RC, MVT ::i32) && "Invalid destination!") ? void (0) : __assert_fail ("TRI->isTypeLegalForClass(*RC, MVT::i32) && \"Invalid destination!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11439, __extension__ __PRETTY_FUNCTION__)); | |||
11440 | Register mainDstReg = MRI.createVirtualRegister(RC); | |||
11441 | Register restoreDstReg = MRI.createVirtualRegister(RC); | |||
11442 | ||||
11443 | MVT PVT = getPointerTy(MF->getDataLayout()); | |||
11444 | assert((PVT == MVT::i64 || PVT == MVT::i32) &&(static_cast <bool> ((PVT == MVT::i64 || PVT == MVT::i32 ) && "Invalid Pointer Size!") ? void (0) : __assert_fail ("(PVT == MVT::i64 || PVT == MVT::i32) && \"Invalid Pointer Size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11445, __extension__ __PRETTY_FUNCTION__)) | |||
11445 | "Invalid Pointer Size!")(static_cast <bool> ((PVT == MVT::i64 || PVT == MVT::i32 ) && "Invalid Pointer Size!") ? void (0) : __assert_fail ("(PVT == MVT::i64 || PVT == MVT::i32) && \"Invalid Pointer Size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11445, __extension__ __PRETTY_FUNCTION__)); | |||
11446 | // For v = setjmp(buf), we generate | |||
11447 | // | |||
11448 | // thisMBB: | |||
11449 | // SjLjSetup mainMBB | |||
11450 | // bl mainMBB | |||
11451 | // v_restore = 1 | |||
11452 | // b sinkMBB | |||
11453 | // | |||
11454 | // mainMBB: | |||
11455 | // buf[LabelOffset] = LR | |||
11456 | // v_main = 0 | |||
11457 | // | |||
11458 | // sinkMBB: | |||
11459 | // v = phi(main, restore) | |||
11460 | // | |||
11461 | ||||
11462 | MachineBasicBlock *thisMBB = MBB; | |||
11463 | MachineBasicBlock *mainMBB = MF->CreateMachineBasicBlock(BB); | |||
11464 | MachineBasicBlock *sinkMBB = MF->CreateMachineBasicBlock(BB); | |||
11465 | MF->insert(I, mainMBB); | |||
11466 | MF->insert(I, sinkMBB); | |||
11467 | ||||
11468 | MachineInstrBuilder MIB; | |||
11469 | ||||
11470 | // Transfer the remainder of BB and its successor edges to sinkMBB. | |||
11471 | sinkMBB->splice(sinkMBB->begin(), MBB, | |||
11472 | std::next(MachineBasicBlock::iterator(MI)), MBB->end()); | |||
11473 | sinkMBB->transferSuccessorsAndUpdatePHIs(MBB); | |||
11474 | ||||
11475 | // Note that the structure of the jmp_buf used here is not compatible | |||
11476 | // with that used by libc, and is not designed to be. Specifically, it | |||
11477 | // stores only those 'reserved' registers that LLVM does not otherwise | |||
11478 | // understand how to spill. Also, by convention, by the time this | |||
11479 | // intrinsic is called, Clang has already stored the frame address in the | |||
11480 | // first slot of the buffer and stack address in the third. Following the | |||
11481 | // X86 target code, we'll store the jump address in the second slot. We also | |||
11482 | // need to save the TOC pointer (R2) to handle jumps between shared | |||
11483 | // libraries, and that will be stored in the fourth slot. The thread | |||
11484 | // identifier (R13) is not affected. | |||
11485 | ||||
11486 | // thisMBB: | |||
11487 | const int64_t LabelOffset = 1 * PVT.getStoreSize(); | |||
11488 | const int64_t TOCOffset = 3 * PVT.getStoreSize(); | |||
11489 | const int64_t BPOffset = 4 * PVT.getStoreSize(); | |||
11490 | ||||
11491 | // Prepare IP either in reg. | |||
11492 | const TargetRegisterClass *PtrRC = getRegClassFor(PVT); | |||
11493 | Register LabelReg = MRI.createVirtualRegister(PtrRC); | |||
11494 | Register BufReg = MI.getOperand(1).getReg(); | |||
11495 | ||||
11496 | if (Subtarget.is64BitELFABI()) { | |||
11497 | setUsesTOCBasePtr(*MBB->getParent()); | |||
11498 | MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::STD)) | |||
11499 | .addReg(PPC::X2) | |||
11500 | .addImm(TOCOffset) | |||
11501 | .addReg(BufReg) | |||
11502 | .cloneMemRefs(MI); | |||
11503 | } | |||
11504 | ||||
11505 | // Naked functions never have a base pointer, and so we use r1. For all | |||
11506 | // other functions, this decision must be delayed until during PEI. | |||
11507 | unsigned BaseReg; | |||
11508 | if (MF->getFunction().hasFnAttribute(Attribute::Naked)) | |||
11509 | BaseReg = Subtarget.isPPC64() ? PPC::X1 : PPC::R1; | |||
11510 | else | |||
11511 | BaseReg = Subtarget.isPPC64() ? PPC::BP8 : PPC::BP; | |||
11512 | ||||
11513 | MIB = BuildMI(*thisMBB, MI, DL, | |||
11514 | TII->get(Subtarget.isPPC64() ? PPC::STD : PPC::STW)) | |||
11515 | .addReg(BaseReg) | |||
11516 | .addImm(BPOffset) | |||
11517 | .addReg(BufReg) | |||
11518 | .cloneMemRefs(MI); | |||
11519 | ||||
11520 | // Setup | |||
11521 | MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::BCLalways)).addMBB(mainMBB); | |||
11522 | MIB.addRegMask(TRI->getNoPreservedMask()); | |||
11523 | ||||
11524 | BuildMI(*thisMBB, MI, DL, TII->get(PPC::LI), restoreDstReg).addImm(1); | |||
11525 | ||||
11526 | MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::EH_SjLj_Setup)) | |||
11527 | .addMBB(mainMBB); | |||
11528 | MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::B)).addMBB(sinkMBB); | |||
11529 | ||||
11530 | thisMBB->addSuccessor(mainMBB, BranchProbability::getZero()); | |||
11531 | thisMBB->addSuccessor(sinkMBB, BranchProbability::getOne()); | |||
11532 | ||||
11533 | // mainMBB: | |||
11534 | // mainDstReg = 0 | |||
11535 | MIB = | |||
11536 | BuildMI(mainMBB, DL, | |||
11537 | TII->get(Subtarget.isPPC64() ? PPC::MFLR8 : PPC::MFLR), LabelReg); | |||
11538 | ||||
11539 | // Store IP | |||
11540 | if (Subtarget.isPPC64()) { | |||
11541 | MIB = BuildMI(mainMBB, DL, TII->get(PPC::STD)) | |||
11542 | .addReg(LabelReg) | |||
11543 | .addImm(LabelOffset) | |||
11544 | .addReg(BufReg); | |||
11545 | } else { | |||
11546 | MIB = BuildMI(mainMBB, DL, TII->get(PPC::STW)) | |||
11547 | .addReg(LabelReg) | |||
11548 | .addImm(LabelOffset) | |||
11549 | .addReg(BufReg); | |||
11550 | } | |||
11551 | MIB.cloneMemRefs(MI); | |||
11552 | ||||
11553 | BuildMI(mainMBB, DL, TII->get(PPC::LI), mainDstReg).addImm(0); | |||
11554 | mainMBB->addSuccessor(sinkMBB); | |||
11555 | ||||
11556 | // sinkMBB: | |||
11557 | BuildMI(*sinkMBB, sinkMBB->begin(), DL, | |||
11558 | TII->get(PPC::PHI), DstReg) | |||
11559 | .addReg(mainDstReg).addMBB(mainMBB) | |||
11560 | .addReg(restoreDstReg).addMBB(thisMBB); | |||
11561 | ||||
11562 | MI.eraseFromParent(); | |||
11563 | return sinkMBB; | |||
11564 | } | |||
11565 | ||||
11566 | MachineBasicBlock * | |||
11567 | PPCTargetLowering::emitEHSjLjLongJmp(MachineInstr &MI, | |||
11568 | MachineBasicBlock *MBB) const { | |||
11569 | DebugLoc DL = MI.getDebugLoc(); | |||
11570 | const TargetInstrInfo *TII = Subtarget.getInstrInfo(); | |||
11571 | ||||
11572 | MachineFunction *MF = MBB->getParent(); | |||
11573 | MachineRegisterInfo &MRI = MF->getRegInfo(); | |||
11574 | ||||
11575 | MVT PVT = getPointerTy(MF->getDataLayout()); | |||
11576 | assert((PVT == MVT::i64 || PVT == MVT::i32) &&(static_cast <bool> ((PVT == MVT::i64 || PVT == MVT::i32 ) && "Invalid Pointer Size!") ? void (0) : __assert_fail ("(PVT == MVT::i64 || PVT == MVT::i32) && \"Invalid Pointer Size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11577, __extension__ __PRETTY_FUNCTION__)) | |||
11577 | "Invalid Pointer Size!")(static_cast <bool> ((PVT == MVT::i64 || PVT == MVT::i32 ) && "Invalid Pointer Size!") ? void (0) : __assert_fail ("(PVT == MVT::i64 || PVT == MVT::i32) && \"Invalid Pointer Size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11577, __extension__ __PRETTY_FUNCTION__)); | |||
11578 | ||||
11579 | const TargetRegisterClass *RC = | |||
11580 | (PVT == MVT::i64) ? &PPC::G8RCRegClass : &PPC::GPRCRegClass; | |||
11581 | Register Tmp = MRI.createVirtualRegister(RC); | |||
11582 | // Since FP is only updated here but NOT referenced, it's treated as GPR. | |||
11583 | unsigned FP = (PVT == MVT::i64) ? PPC::X31 : PPC::R31; | |||
11584 | unsigned SP = (PVT == MVT::i64) ? PPC::X1 : PPC::R1; | |||
11585 | unsigned BP = | |||
11586 | (PVT == MVT::i64) | |||
11587 | ? PPC::X30 | |||
11588 | : (Subtarget.isSVR4ABI() && isPositionIndependent() ? PPC::R29 | |||
11589 | : PPC::R30); | |||
11590 | ||||
11591 | MachineInstrBuilder MIB; | |||
11592 | ||||
11593 | const int64_t LabelOffset = 1 * PVT.getStoreSize(); | |||
11594 | const int64_t SPOffset = 2 * PVT.getStoreSize(); | |||
11595 | const int64_t TOCOffset = 3 * PVT.getStoreSize(); | |||
11596 | const int64_t BPOffset = 4 * PVT.getStoreSize(); | |||
11597 | ||||
11598 | Register BufReg = MI.getOperand(0).getReg(); | |||
11599 | ||||
11600 | // Reload FP (the jumped-to function may not have had a | |||
11601 | // frame pointer, and if so, then its r31 will be restored | |||
11602 | // as necessary). | |||
11603 | if (PVT == MVT::i64) { | |||
11604 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), FP) | |||
11605 | .addImm(0) | |||
11606 | .addReg(BufReg); | |||
11607 | } else { | |||
11608 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), FP) | |||
11609 | .addImm(0) | |||
11610 | .addReg(BufReg); | |||
11611 | } | |||
11612 | MIB.cloneMemRefs(MI); | |||
11613 | ||||
11614 | // Reload IP | |||
11615 | if (PVT == MVT::i64) { | |||
11616 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), Tmp) | |||
11617 | .addImm(LabelOffset) | |||
11618 | .addReg(BufReg); | |||
11619 | } else { | |||
11620 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), Tmp) | |||
11621 | .addImm(LabelOffset) | |||
11622 | .addReg(BufReg); | |||
11623 | } | |||
11624 | MIB.cloneMemRefs(MI); | |||
11625 | ||||
11626 | // Reload SP | |||
11627 | if (PVT == MVT::i64) { | |||
11628 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), SP) | |||
11629 | .addImm(SPOffset) | |||
11630 | .addReg(BufReg); | |||
11631 | } else { | |||
11632 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), SP) | |||
11633 | .addImm(SPOffset) | |||
11634 | .addReg(BufReg); | |||
11635 | } | |||
11636 | MIB.cloneMemRefs(MI); | |||
11637 | ||||
11638 | // Reload BP | |||
11639 | if (PVT == MVT::i64) { | |||
11640 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), BP) | |||
11641 | .addImm(BPOffset) | |||
11642 | .addReg(BufReg); | |||
11643 | } else { | |||
11644 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), BP) | |||
11645 | .addImm(BPOffset) | |||
11646 | .addReg(BufReg); | |||
11647 | } | |||
11648 | MIB.cloneMemRefs(MI); | |||
11649 | ||||
11650 | // Reload TOC | |||
11651 | if (PVT == MVT::i64 && Subtarget.isSVR4ABI()) { | |||
11652 | setUsesTOCBasePtr(*MBB->getParent()); | |||
11653 | MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), PPC::X2) | |||
11654 | .addImm(TOCOffset) | |||
11655 | .addReg(BufReg) | |||
11656 | .cloneMemRefs(MI); | |||
11657 | } | |||
11658 | ||||
11659 | // Jump | |||
11660 | BuildMI(*MBB, MI, DL, | |||
11661 | TII->get(PVT == MVT::i64 ? PPC::MTCTR8 : PPC::MTCTR)).addReg(Tmp); | |||
11662 | BuildMI(*MBB, MI, DL, TII->get(PVT == MVT::i64 ? PPC::BCTR8 : PPC::BCTR)); | |||
11663 | ||||
11664 | MI.eraseFromParent(); | |||
11665 | return MBB; | |||
11666 | } | |||
11667 | ||||
11668 | bool PPCTargetLowering::hasInlineStackProbe(MachineFunction &MF) const { | |||
11669 | // If the function specifically requests inline stack probes, emit them. | |||
11670 | if (MF.getFunction().hasFnAttribute("probe-stack")) | |||
11671 | return MF.getFunction().getFnAttribute("probe-stack").getValueAsString() == | |||
11672 | "inline-asm"; | |||
11673 | return false; | |||
11674 | } | |||
11675 | ||||
11676 | unsigned PPCTargetLowering::getStackProbeSize(MachineFunction &MF) const { | |||
11677 | const TargetFrameLowering *TFI = Subtarget.getFrameLowering(); | |||
11678 | unsigned StackAlign = TFI->getStackAlignment(); | |||
11679 | assert(StackAlign >= 1 && isPowerOf2_32(StackAlign) &&(static_cast <bool> (StackAlign >= 1 && isPowerOf2_32 (StackAlign) && "Unexpected stack alignment") ? void ( 0) : __assert_fail ("StackAlign >= 1 && isPowerOf2_32(StackAlign) && \"Unexpected stack alignment\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11680, __extension__ __PRETTY_FUNCTION__)) | |||
11680 | "Unexpected stack alignment")(static_cast <bool> (StackAlign >= 1 && isPowerOf2_32 (StackAlign) && "Unexpected stack alignment") ? void ( 0) : __assert_fail ("StackAlign >= 1 && isPowerOf2_32(StackAlign) && \"Unexpected stack alignment\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11680, __extension__ __PRETTY_FUNCTION__)); | |||
11681 | // The default stack probe size is 4096 if the function has no | |||
11682 | // stack-probe-size attribute. | |||
11683 | unsigned StackProbeSize = 4096; | |||
11684 | const Function &Fn = MF.getFunction(); | |||
11685 | if (Fn.hasFnAttribute("stack-probe-size")) | |||
11686 | Fn.getFnAttribute("stack-probe-size") | |||
11687 | .getValueAsString() | |||
11688 | .getAsInteger(0, StackProbeSize); | |||
11689 | // Round down to the stack alignment. | |||
11690 | StackProbeSize &= ~(StackAlign - 1); | |||
11691 | return StackProbeSize ? StackProbeSize : StackAlign; | |||
11692 | } | |||
11693 | ||||
11694 | // Lower dynamic stack allocation with probing. `emitProbedAlloca` is splitted | |||
11695 | // into three phases. In the first phase, it uses pseudo instruction | |||
11696 | // PREPARE_PROBED_ALLOCA to get the future result of actual FramePointer and | |||
11697 | // FinalStackPtr. In the second phase, it generates a loop for probing blocks. | |||
11698 | // At last, it uses pseudo instruction DYNAREAOFFSET to get the future result of | |||
11699 | // MaxCallFrameSize so that it can calculate correct data area pointer. | |||
11700 | MachineBasicBlock * | |||
11701 | PPCTargetLowering::emitProbedAlloca(MachineInstr &MI, | |||
11702 | MachineBasicBlock *MBB) const { | |||
11703 | const bool isPPC64 = Subtarget.isPPC64(); | |||
11704 | MachineFunction *MF = MBB->getParent(); | |||
11705 | const TargetInstrInfo *TII = Subtarget.getInstrInfo(); | |||
11706 | DebugLoc DL = MI.getDebugLoc(); | |||
11707 | const unsigned ProbeSize = getStackProbeSize(*MF); | |||
11708 | const BasicBlock *ProbedBB = MBB->getBasicBlock(); | |||
11709 | MachineRegisterInfo &MRI = MF->getRegInfo(); | |||
11710 | // The CFG of probing stack looks as | |||
11711 | // +-----+ | |||
11712 | // | MBB | | |||
11713 | // +--+--+ | |||
11714 | // | | |||
11715 | // +----v----+ | |||
11716 | // +--->+ TestMBB +---+ | |||
11717 | // | +----+----+ | | |||
11718 | // | | | | |||
11719 | // | +-----v----+ | | |||
11720 | // +---+ BlockMBB | | | |||
11721 | // +----------+ | | |||
11722 | // | | |||
11723 | // +---------+ | | |||
11724 | // | TailMBB +<--+ | |||
11725 | // +---------+ | |||
11726 | // In MBB, calculate previous frame pointer and final stack pointer. | |||
11727 | // In TestMBB, test if sp is equal to final stack pointer, if so, jump to | |||
11728 | // TailMBB. In BlockMBB, update the sp atomically and jump back to TestMBB. | |||
11729 | // TailMBB is spliced via \p MI. | |||
11730 | MachineBasicBlock *TestMBB = MF->CreateMachineBasicBlock(ProbedBB); | |||
11731 | MachineBasicBlock *TailMBB = MF->CreateMachineBasicBlock(ProbedBB); | |||
11732 | MachineBasicBlock *BlockMBB = MF->CreateMachineBasicBlock(ProbedBB); | |||
11733 | ||||
11734 | MachineFunction::iterator MBBIter = ++MBB->getIterator(); | |||
11735 | MF->insert(MBBIter, TestMBB); | |||
11736 | MF->insert(MBBIter, BlockMBB); | |||
11737 | MF->insert(MBBIter, TailMBB); | |||
11738 | ||||
11739 | const TargetRegisterClass *G8RC = &PPC::G8RCRegClass; | |||
11740 | const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; | |||
11741 | ||||
11742 | Register DstReg = MI.getOperand(0).getReg(); | |||
11743 | Register NegSizeReg = MI.getOperand(1).getReg(); | |||
11744 | Register SPReg = isPPC64 ? PPC::X1 : PPC::R1; | |||
11745 | Register FinalStackPtr = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11746 | Register FramePointer = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11747 | Register ActualNegSizeReg = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11748 | ||||
11749 | // Since value of NegSizeReg might be realigned in prologepilog, insert a | |||
11750 | // PREPARE_PROBED_ALLOCA pseudo instruction to get actual FramePointer and | |||
11751 | // NegSize. | |||
11752 | unsigned ProbeOpc; | |||
11753 | if (!MRI.hasOneNonDBGUse(NegSizeReg)) | |||
11754 | ProbeOpc = | |||
11755 | isPPC64 ? PPC::PREPARE_PROBED_ALLOCA_64 : PPC::PREPARE_PROBED_ALLOCA_32; | |||
11756 | else | |||
11757 | // By introducing PREPARE_PROBED_ALLOCA_NEGSIZE_OPT, ActualNegSizeReg | |||
11758 | // and NegSizeReg will be allocated in the same phyreg to avoid | |||
11759 | // redundant copy when NegSizeReg has only one use which is current MI and | |||
11760 | // will be replaced by PREPARE_PROBED_ALLOCA then. | |||
11761 | ProbeOpc = isPPC64 ? PPC::PREPARE_PROBED_ALLOCA_NEGSIZE_SAME_REG_64 | |||
11762 | : PPC::PREPARE_PROBED_ALLOCA_NEGSIZE_SAME_REG_32; | |||
11763 | BuildMI(*MBB, {MI}, DL, TII->get(ProbeOpc), FramePointer) | |||
11764 | .addDef(ActualNegSizeReg) | |||
11765 | .addReg(NegSizeReg) | |||
11766 | .add(MI.getOperand(2)) | |||
11767 | .add(MI.getOperand(3)); | |||
11768 | ||||
11769 | // Calculate final stack pointer, which equals to SP + ActualNegSize. | |||
11770 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::ADD8 : PPC::ADD4), | |||
11771 | FinalStackPtr) | |||
11772 | .addReg(SPReg) | |||
11773 | .addReg(ActualNegSizeReg); | |||
11774 | ||||
11775 | // Materialize a scratch register for update. | |||
11776 | int64_t NegProbeSize = -(int64_t)ProbeSize; | |||
11777 | assert(isInt<32>(NegProbeSize) && "Unhandled probe size!")(static_cast <bool> (isInt<32>(NegProbeSize) && "Unhandled probe size!") ? void (0) : __assert_fail ("isInt<32>(NegProbeSize) && \"Unhandled probe size!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 11777, __extension__ __PRETTY_FUNCTION__)); | |||
11778 | Register ScratchReg = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11779 | if (!isInt<16>(NegProbeSize)) { | |||
11780 | Register TempReg = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11781 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::LIS8 : PPC::LIS), TempReg) | |||
11782 | .addImm(NegProbeSize >> 16); | |||
11783 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::ORI8 : PPC::ORI), | |||
11784 | ScratchReg) | |||
11785 | .addReg(TempReg) | |||
11786 | .addImm(NegProbeSize & 0xFFFF); | |||
11787 | } else | |||
11788 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::LI8 : PPC::LI), ScratchReg) | |||
11789 | .addImm(NegProbeSize); | |||
11790 | ||||
11791 | { | |||
11792 | // Probing leading residual part. | |||
11793 | Register Div = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11794 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::DIVD : PPC::DIVW), Div) | |||
11795 | .addReg(ActualNegSizeReg) | |||
11796 | .addReg(ScratchReg); | |||
11797 | Register Mul = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11798 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::MULLD : PPC::MULLW), Mul) | |||
11799 | .addReg(Div) | |||
11800 | .addReg(ScratchReg); | |||
11801 | Register NegMod = MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11802 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::SUBF8 : PPC::SUBF), NegMod) | |||
11803 | .addReg(Mul) | |||
11804 | .addReg(ActualNegSizeReg); | |||
11805 | BuildMI(*MBB, {MI}, DL, TII->get(isPPC64 ? PPC::STDUX : PPC::STWUX), SPReg) | |||
11806 | .addReg(FramePointer) | |||
11807 | .addReg(SPReg) | |||
11808 | .addReg(NegMod); | |||
11809 | } | |||
11810 | ||||
11811 | { | |||
11812 | // Remaining part should be multiple of ProbeSize. | |||
11813 | Register CmpResult = MRI.createVirtualRegister(&PPC::CRRCRegClass); | |||
11814 | BuildMI(TestMBB, DL, TII->get(isPPC64 ? PPC::CMPD : PPC::CMPW), CmpResult) | |||
11815 | .addReg(SPReg) | |||
11816 | .addReg(FinalStackPtr); | |||
11817 | BuildMI(TestMBB, DL, TII->get(PPC::BCC)) | |||
11818 | .addImm(PPC::PRED_EQ) | |||
11819 | .addReg(CmpResult) | |||
11820 | .addMBB(TailMBB); | |||
11821 | TestMBB->addSuccessor(BlockMBB); | |||
11822 | TestMBB->addSuccessor(TailMBB); | |||
11823 | } | |||
11824 | ||||
11825 | { | |||
11826 | // Touch the block. | |||
11827 | // |P...|P...|P... | |||
11828 | BuildMI(BlockMBB, DL, TII->get(isPPC64 ? PPC::STDUX : PPC::STWUX), SPReg) | |||
11829 | .addReg(FramePointer) | |||
11830 | .addReg(SPReg) | |||
11831 | .addReg(ScratchReg); | |||
11832 | BuildMI(BlockMBB, DL, TII->get(PPC::B)).addMBB(TestMBB); | |||
11833 | BlockMBB->addSuccessor(TestMBB); | |||
11834 | } | |||
11835 | ||||
11836 | // Calculation of MaxCallFrameSize is deferred to prologepilog, use | |||
11837 | // DYNAREAOFFSET pseudo instruction to get the future result. | |||
11838 | Register MaxCallFrameSizeReg = | |||
11839 | MRI.createVirtualRegister(isPPC64 ? G8RC : GPRC); | |||
11840 | BuildMI(TailMBB, DL, | |||
11841 | TII->get(isPPC64 ? PPC::DYNAREAOFFSET8 : PPC::DYNAREAOFFSET), | |||
11842 | MaxCallFrameSizeReg) | |||
11843 | .add(MI.getOperand(2)) | |||
11844 | .add(MI.getOperand(3)); | |||
11845 | BuildMI(TailMBB, DL, TII->get(isPPC64 ? PPC::ADD8 : PPC::ADD4), DstReg) | |||
11846 | .addReg(SPReg) | |||
11847 | .addReg(MaxCallFrameSizeReg); | |||
11848 | ||||
11849 | // Splice instructions after MI to TailMBB. | |||
11850 | TailMBB->splice(TailMBB->end(), MBB, | |||
11851 | std::next(MachineBasicBlock::iterator(MI)), MBB->end()); | |||
11852 | TailMBB->transferSuccessorsAndUpdatePHIs(MBB); | |||
11853 | MBB->addSuccessor(TestMBB); | |||
11854 | ||||
11855 | // Delete the pseudo instruction. | |||
11856 | MI.eraseFromParent(); | |||
11857 | ||||
11858 | ++NumDynamicAllocaProbed; | |||
11859 | return TailMBB; | |||
11860 | } | |||
11861 | ||||
11862 | MachineBasicBlock * | |||
11863 | PPCTargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI, | |||
11864 | MachineBasicBlock *BB) const { | |||
11865 | if (MI.getOpcode() == TargetOpcode::STACKMAP || | |||
11866 | MI.getOpcode() == TargetOpcode::PATCHPOINT) { | |||
11867 | if (Subtarget.is64BitELFABI() && | |||
11868 | MI.getOpcode() == TargetOpcode::PATCHPOINT && | |||
11869 | !Subtarget.isUsingPCRelativeCalls()) { | |||
11870 | // Call lowering should have added an r2 operand to indicate a dependence | |||
11871 | // on the TOC base pointer value. It can't however, because there is no | |||
11872 | // way to mark the dependence as implicit there, and so the stackmap code | |||
11873 | // will confuse it with a regular operand. Instead, add the dependence | |||
11874 | // here. | |||
11875 | MI.addOperand(MachineOperand::CreateReg(PPC::X2, false, true)); | |||
11876 | } | |||
11877 | ||||
11878 | return emitPatchPoint(MI, BB); | |||
11879 | } | |||
11880 | ||||
11881 | if (MI.getOpcode() == PPC::EH_SjLj_SetJmp32 || | |||
11882 | MI.getOpcode() == PPC::EH_SjLj_SetJmp64) { | |||
11883 | return emitEHSjLjSetJmp(MI, BB); | |||
11884 | } else if (MI.getOpcode() == PPC::EH_SjLj_LongJmp32 || | |||
11885 | MI.getOpcode() == PPC::EH_SjLj_LongJmp64) { | |||
11886 | return emitEHSjLjLongJmp(MI, BB); | |||
11887 | } | |||
11888 | ||||
11889 | const TargetInstrInfo *TII = Subtarget.getInstrInfo(); | |||
11890 | ||||
11891 | // To "insert" these instructions we actually have to insert their | |||
11892 | // control-flow patterns. | |||
11893 | const BasicBlock *LLVM_BB = BB->getBasicBlock(); | |||
11894 | MachineFunction::iterator It = ++BB->getIterator(); | |||
11895 | ||||
11896 | MachineFunction *F = BB->getParent(); | |||
11897 | ||||
11898 | if (MI.getOpcode() == PPC::SELECT_CC_I4 || | |||
11899 | MI.getOpcode() == PPC::SELECT_CC_I8 || MI.getOpcode() == PPC::SELECT_I4 || | |||
11900 | MI.getOpcode() == PPC::SELECT_I8) { | |||
11901 | SmallVector<MachineOperand, 2> Cond; | |||
11902 | if (MI.getOpcode() == PPC::SELECT_CC_I4 || | |||
11903 | MI.getOpcode() == PPC::SELECT_CC_I8) | |||
11904 | Cond.push_back(MI.getOperand(4)); | |||
11905 | else | |||
11906 | Cond.push_back(MachineOperand::CreateImm(PPC::PRED_BIT_SET)); | |||
11907 | Cond.push_back(MI.getOperand(1)); | |||
11908 | ||||
11909 | DebugLoc dl = MI.getDebugLoc(); | |||
11910 | TII->insertSelect(*BB, MI, dl, MI.getOperand(0).getReg(), Cond, | |||
11911 | MI.getOperand(2).getReg(), MI.getOperand(3).getReg()); | |||
11912 | } else if (MI.getOpcode() == PPC::SELECT_CC_F4 || | |||
11913 | MI.getOpcode() == PPC::SELECT_CC_F8 || | |||
11914 | MI.getOpcode() == PPC::SELECT_CC_F16 || | |||
11915 | MI.getOpcode() == PPC::SELECT_CC_VRRC || | |||
11916 | MI.getOpcode() == PPC::SELECT_CC_VSFRC || | |||
11917 | MI.getOpcode() == PPC::SELECT_CC_VSSRC || | |||
11918 | MI.getOpcode() == PPC::SELECT_CC_VSRC || | |||
11919 | MI.getOpcode() == PPC::SELECT_CC_SPE4 || | |||
11920 | MI.getOpcode() == PPC::SELECT_CC_SPE || | |||
11921 | MI.getOpcode() == PPC::SELECT_F4 || | |||
11922 | MI.getOpcode() == PPC::SELECT_F8 || | |||
11923 | MI.getOpcode() == PPC::SELECT_F16 || | |||
11924 | MI.getOpcode() == PPC::SELECT_SPE || | |||
11925 | MI.getOpcode() == PPC::SELECT_SPE4 || | |||
11926 | MI.getOpcode() == PPC::SELECT_VRRC || | |||
11927 | MI.getOpcode() == PPC::SELECT_VSFRC || | |||
11928 | MI.getOpcode() == PPC::SELECT_VSSRC || | |||
11929 | MI.getOpcode() == PPC::SELECT_VSRC) { | |||
11930 | // The incoming instruction knows the destination vreg to set, the | |||
11931 | // condition code register to branch on, the true/false values to | |||
11932 | // select between, and a branch opcode to use. | |||
11933 | ||||
11934 | // thisMBB: | |||
11935 | // ... | |||
11936 | // TrueVal = ... | |||
11937 | // cmpTY ccX, r1, r2 | |||
11938 | // bCC copy1MBB | |||
11939 | // fallthrough --> copy0MBB | |||
11940 | MachineBasicBlock *thisMBB = BB; | |||
11941 | MachineBasicBlock *copy0MBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
11942 | MachineBasicBlock *sinkMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
11943 | DebugLoc dl = MI.getDebugLoc(); | |||
11944 | F->insert(It, copy0MBB); | |||
11945 | F->insert(It, sinkMBB); | |||
11946 | ||||
11947 | // Transfer the remainder of BB and its successor edges to sinkMBB. | |||
11948 | sinkMBB->splice(sinkMBB->begin(), BB, | |||
11949 | std::next(MachineBasicBlock::iterator(MI)), BB->end()); | |||
11950 | sinkMBB->transferSuccessorsAndUpdatePHIs(BB); | |||
11951 | ||||
11952 | // Next, add the true and fallthrough blocks as its successors. | |||
11953 | BB->addSuccessor(copy0MBB); | |||
11954 | BB->addSuccessor(sinkMBB); | |||
11955 | ||||
11956 | if (MI.getOpcode() == PPC::SELECT_I4 || MI.getOpcode() == PPC::SELECT_I8 || | |||
11957 | MI.getOpcode() == PPC::SELECT_F4 || MI.getOpcode() == PPC::SELECT_F8 || | |||
11958 | MI.getOpcode() == PPC::SELECT_F16 || | |||
11959 | MI.getOpcode() == PPC::SELECT_SPE4 || | |||
11960 | MI.getOpcode() == PPC::SELECT_SPE || | |||
11961 | MI.getOpcode() == PPC::SELECT_VRRC || | |||
11962 | MI.getOpcode() == PPC::SELECT_VSFRC || | |||
11963 | MI.getOpcode() == PPC::SELECT_VSSRC || | |||
11964 | MI.getOpcode() == PPC::SELECT_VSRC) { | |||
11965 | BuildMI(BB, dl, TII->get(PPC::BC)) | |||
11966 | .addReg(MI.getOperand(1).getReg()) | |||
11967 | .addMBB(sinkMBB); | |||
11968 | } else { | |||
11969 | unsigned SelectPred = MI.getOperand(4).getImm(); | |||
11970 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
11971 | .addImm(SelectPred) | |||
11972 | .addReg(MI.getOperand(1).getReg()) | |||
11973 | .addMBB(sinkMBB); | |||
11974 | } | |||
11975 | ||||
11976 | // copy0MBB: | |||
11977 | // %FalseValue = ... | |||
11978 | // # fallthrough to sinkMBB | |||
11979 | BB = copy0MBB; | |||
11980 | ||||
11981 | // Update machine-CFG edges | |||
11982 | BB->addSuccessor(sinkMBB); | |||
11983 | ||||
11984 | // sinkMBB: | |||
11985 | // %Result = phi [ %FalseValue, copy0MBB ], [ %TrueValue, thisMBB ] | |||
11986 | // ... | |||
11987 | BB = sinkMBB; | |||
11988 | BuildMI(*BB, BB->begin(), dl, TII->get(PPC::PHI), MI.getOperand(0).getReg()) | |||
11989 | .addReg(MI.getOperand(3).getReg()) | |||
11990 | .addMBB(copy0MBB) | |||
11991 | .addReg(MI.getOperand(2).getReg()) | |||
11992 | .addMBB(thisMBB); | |||
11993 | } else if (MI.getOpcode() == PPC::ReadTB) { | |||
11994 | // To read the 64-bit time-base register on a 32-bit target, we read the | |||
11995 | // two halves. Should the counter have wrapped while it was being read, we | |||
11996 | // need to try again. | |||
11997 | // ... | |||
11998 | // readLoop: | |||
11999 | // mfspr Rx,TBU # load from TBU | |||
12000 | // mfspr Ry,TB # load from TB | |||
12001 | // mfspr Rz,TBU # load from TBU | |||
12002 | // cmpw crX,Rx,Rz # check if 'old'='new' | |||
12003 | // bne readLoop # branch if they're not equal | |||
12004 | // ... | |||
12005 | ||||
12006 | MachineBasicBlock *readMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12007 | MachineBasicBlock *sinkMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12008 | DebugLoc dl = MI.getDebugLoc(); | |||
12009 | F->insert(It, readMBB); | |||
12010 | F->insert(It, sinkMBB); | |||
12011 | ||||
12012 | // Transfer the remainder of BB and its successor edges to sinkMBB. | |||
12013 | sinkMBB->splice(sinkMBB->begin(), BB, | |||
12014 | std::next(MachineBasicBlock::iterator(MI)), BB->end()); | |||
12015 | sinkMBB->transferSuccessorsAndUpdatePHIs(BB); | |||
12016 | ||||
12017 | BB->addSuccessor(readMBB); | |||
12018 | BB = readMBB; | |||
12019 | ||||
12020 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12021 | Register ReadAgainReg = RegInfo.createVirtualRegister(&PPC::GPRCRegClass); | |||
12022 | Register LoReg = MI.getOperand(0).getReg(); | |||
12023 | Register HiReg = MI.getOperand(1).getReg(); | |||
12024 | ||||
12025 | BuildMI(BB, dl, TII->get(PPC::MFSPR), HiReg).addImm(269); | |||
12026 | BuildMI(BB, dl, TII->get(PPC::MFSPR), LoReg).addImm(268); | |||
12027 | BuildMI(BB, dl, TII->get(PPC::MFSPR), ReadAgainReg).addImm(269); | |||
12028 | ||||
12029 | Register CmpReg = RegInfo.createVirtualRegister(&PPC::CRRCRegClass); | |||
12030 | ||||
12031 | BuildMI(BB, dl, TII->get(PPC::CMPW), CmpReg) | |||
12032 | .addReg(HiReg) | |||
12033 | .addReg(ReadAgainReg); | |||
12034 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
12035 | .addImm(PPC::PRED_NE) | |||
12036 | .addReg(CmpReg) | |||
12037 | .addMBB(readMBB); | |||
12038 | ||||
12039 | BB->addSuccessor(readMBB); | |||
12040 | BB->addSuccessor(sinkMBB); | |||
12041 | } else if (MI.getOpcode() == PPC::ATOMIC_LOAD_ADD_I8) | |||
12042 | BB = EmitPartwordAtomicBinary(MI, BB, true, PPC::ADD4); | |||
12043 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_ADD_I16) | |||
12044 | BB = EmitPartwordAtomicBinary(MI, BB, false, PPC::ADD4); | |||
12045 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_ADD_I32) | |||
12046 | BB = EmitAtomicBinary(MI, BB, 4, PPC::ADD4); | |||
12047 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_ADD_I64) | |||
12048 | BB = EmitAtomicBinary(MI, BB, 8, PPC::ADD8); | |||
12049 | ||||
12050 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_AND_I8) | |||
12051 | BB = EmitPartwordAtomicBinary(MI, BB, true, PPC::AND); | |||
12052 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_AND_I16) | |||
12053 | BB = EmitPartwordAtomicBinary(MI, BB, false, PPC::AND); | |||
12054 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_AND_I32) | |||
12055 | BB = EmitAtomicBinary(MI, BB, 4, PPC::AND); | |||
12056 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_AND_I64) | |||
12057 | BB = EmitAtomicBinary(MI, BB, 8, PPC::AND8); | |||
12058 | ||||
12059 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_OR_I8) | |||
12060 | BB = EmitPartwordAtomicBinary(MI, BB, true, PPC::OR); | |||
12061 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_OR_I16) | |||
12062 | BB = EmitPartwordAtomicBinary(MI, BB, false, PPC::OR); | |||
12063 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_OR_I32) | |||
12064 | BB = EmitAtomicBinary(MI, BB, 4, PPC::OR); | |||
12065 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_OR_I64) | |||
12066 | BB = EmitAtomicBinary(MI, BB, 8, PPC::OR8); | |||
12067 | ||||
12068 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_XOR_I8) | |||
12069 | BB = EmitPartwordAtomicBinary(MI, BB, true, PPC::XOR); | |||
12070 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_XOR_I16) | |||
12071 | BB = EmitPartwordAtomicBinary(MI, BB, false, PPC::XOR); | |||
12072 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_XOR_I32) | |||
12073 | BB = EmitAtomicBinary(MI, BB, 4, PPC::XOR); | |||
12074 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_XOR_I64) | |||
12075 | BB = EmitAtomicBinary(MI, BB, 8, PPC::XOR8); | |||
12076 | ||||
12077 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_NAND_I8) | |||
12078 | BB = EmitPartwordAtomicBinary(MI, BB, true, PPC::NAND); | |||
12079 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_NAND_I16) | |||
12080 | BB = EmitPartwordAtomicBinary(MI, BB, false, PPC::NAND); | |||
12081 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_NAND_I32) | |||
12082 | BB = EmitAtomicBinary(MI, BB, 4, PPC::NAND); | |||
12083 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_NAND_I64) | |||
12084 | BB = EmitAtomicBinary(MI, BB, 8, PPC::NAND8); | |||
12085 | ||||
12086 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_SUB_I8) | |||
12087 | BB = EmitPartwordAtomicBinary(MI, BB, true, PPC::SUBF); | |||
12088 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_SUB_I16) | |||
12089 | BB = EmitPartwordAtomicBinary(MI, BB, false, PPC::SUBF); | |||
12090 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_SUB_I32) | |||
12091 | BB = EmitAtomicBinary(MI, BB, 4, PPC::SUBF); | |||
12092 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_SUB_I64) | |||
12093 | BB = EmitAtomicBinary(MI, BB, 8, PPC::SUBF8); | |||
12094 | ||||
12095 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MIN_I8) | |||
12096 | BB = EmitPartwordAtomicBinary(MI, BB, true, 0, PPC::CMPW, PPC::PRED_GE); | |||
12097 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MIN_I16) | |||
12098 | BB = EmitPartwordAtomicBinary(MI, BB, false, 0, PPC::CMPW, PPC::PRED_GE); | |||
12099 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MIN_I32) | |||
12100 | BB = EmitAtomicBinary(MI, BB, 4, 0, PPC::CMPW, PPC::PRED_GE); | |||
12101 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MIN_I64) | |||
12102 | BB = EmitAtomicBinary(MI, BB, 8, 0, PPC::CMPD, PPC::PRED_GE); | |||
12103 | ||||
12104 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MAX_I8) | |||
12105 | BB = EmitPartwordAtomicBinary(MI, BB, true, 0, PPC::CMPW, PPC::PRED_LE); | |||
12106 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MAX_I16) | |||
12107 | BB = EmitPartwordAtomicBinary(MI, BB, false, 0, PPC::CMPW, PPC::PRED_LE); | |||
12108 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MAX_I32) | |||
12109 | BB = EmitAtomicBinary(MI, BB, 4, 0, PPC::CMPW, PPC::PRED_LE); | |||
12110 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_MAX_I64) | |||
12111 | BB = EmitAtomicBinary(MI, BB, 8, 0, PPC::CMPD, PPC::PRED_LE); | |||
12112 | ||||
12113 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMIN_I8) | |||
12114 | BB = EmitPartwordAtomicBinary(MI, BB, true, 0, PPC::CMPLW, PPC::PRED_GE); | |||
12115 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMIN_I16) | |||
12116 | BB = EmitPartwordAtomicBinary(MI, BB, false, 0, PPC::CMPLW, PPC::PRED_GE); | |||
12117 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMIN_I32) | |||
12118 | BB = EmitAtomicBinary(MI, BB, 4, 0, PPC::CMPLW, PPC::PRED_GE); | |||
12119 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMIN_I64) | |||
12120 | BB = EmitAtomicBinary(MI, BB, 8, 0, PPC::CMPLD, PPC::PRED_GE); | |||
12121 | ||||
12122 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMAX_I8) | |||
12123 | BB = EmitPartwordAtomicBinary(MI, BB, true, 0, PPC::CMPLW, PPC::PRED_LE); | |||
12124 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMAX_I16) | |||
12125 | BB = EmitPartwordAtomicBinary(MI, BB, false, 0, PPC::CMPLW, PPC::PRED_LE); | |||
12126 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMAX_I32) | |||
12127 | BB = EmitAtomicBinary(MI, BB, 4, 0, PPC::CMPLW, PPC::PRED_LE); | |||
12128 | else if (MI.getOpcode() == PPC::ATOMIC_LOAD_UMAX_I64) | |||
12129 | BB = EmitAtomicBinary(MI, BB, 8, 0, PPC::CMPLD, PPC::PRED_LE); | |||
12130 | ||||
12131 | else if (MI.getOpcode() == PPC::ATOMIC_SWAP_I8) | |||
12132 | BB = EmitPartwordAtomicBinary(MI, BB, true, 0); | |||
12133 | else if (MI.getOpcode() == PPC::ATOMIC_SWAP_I16) | |||
12134 | BB = EmitPartwordAtomicBinary(MI, BB, false, 0); | |||
12135 | else if (MI.getOpcode() == PPC::ATOMIC_SWAP_I32) | |||
12136 | BB = EmitAtomicBinary(MI, BB, 4, 0); | |||
12137 | else if (MI.getOpcode() == PPC::ATOMIC_SWAP_I64) | |||
12138 | BB = EmitAtomicBinary(MI, BB, 8, 0); | |||
12139 | else if (MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I32 || | |||
12140 | MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I64 || | |||
12141 | (Subtarget.hasPartwordAtomics() && | |||
12142 | MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I8) || | |||
12143 | (Subtarget.hasPartwordAtomics() && | |||
12144 | MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I16)) { | |||
12145 | bool is64bit = MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I64; | |||
12146 | ||||
12147 | auto LoadMnemonic = PPC::LDARX; | |||
12148 | auto StoreMnemonic = PPC::STDCX; | |||
12149 | switch (MI.getOpcode()) { | |||
12150 | default: | |||
12151 | llvm_unreachable("Compare and swap of unknown size")::llvm::llvm_unreachable_internal("Compare and swap of unknown size" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12151); | |||
12152 | case PPC::ATOMIC_CMP_SWAP_I8: | |||
12153 | LoadMnemonic = PPC::LBARX; | |||
12154 | StoreMnemonic = PPC::STBCX; | |||
12155 | assert(Subtarget.hasPartwordAtomics() && "No support partword atomics.")(static_cast <bool> (Subtarget.hasPartwordAtomics() && "No support partword atomics.") ? void (0) : __assert_fail ( "Subtarget.hasPartwordAtomics() && \"No support partword atomics.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12155, __extension__ __PRETTY_FUNCTION__)); | |||
12156 | break; | |||
12157 | case PPC::ATOMIC_CMP_SWAP_I16: | |||
12158 | LoadMnemonic = PPC::LHARX; | |||
12159 | StoreMnemonic = PPC::STHCX; | |||
12160 | assert(Subtarget.hasPartwordAtomics() && "No support partword atomics.")(static_cast <bool> (Subtarget.hasPartwordAtomics() && "No support partword atomics.") ? void (0) : __assert_fail ( "Subtarget.hasPartwordAtomics() && \"No support partword atomics.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12160, __extension__ __PRETTY_FUNCTION__)); | |||
12161 | break; | |||
12162 | case PPC::ATOMIC_CMP_SWAP_I32: | |||
12163 | LoadMnemonic = PPC::LWARX; | |||
12164 | StoreMnemonic = PPC::STWCX; | |||
12165 | break; | |||
12166 | case PPC::ATOMIC_CMP_SWAP_I64: | |||
12167 | LoadMnemonic = PPC::LDARX; | |||
12168 | StoreMnemonic = PPC::STDCX; | |||
12169 | break; | |||
12170 | } | |||
12171 | Register dest = MI.getOperand(0).getReg(); | |||
12172 | Register ptrA = MI.getOperand(1).getReg(); | |||
12173 | Register ptrB = MI.getOperand(2).getReg(); | |||
12174 | Register oldval = MI.getOperand(3).getReg(); | |||
12175 | Register newval = MI.getOperand(4).getReg(); | |||
12176 | DebugLoc dl = MI.getDebugLoc(); | |||
12177 | ||||
12178 | MachineBasicBlock *loop1MBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12179 | MachineBasicBlock *loop2MBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12180 | MachineBasicBlock *midMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12181 | MachineBasicBlock *exitMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12182 | F->insert(It, loop1MBB); | |||
12183 | F->insert(It, loop2MBB); | |||
12184 | F->insert(It, midMBB); | |||
12185 | F->insert(It, exitMBB); | |||
12186 | exitMBB->splice(exitMBB->begin(), BB, | |||
12187 | std::next(MachineBasicBlock::iterator(MI)), BB->end()); | |||
12188 | exitMBB->transferSuccessorsAndUpdatePHIs(BB); | |||
12189 | ||||
12190 | // thisMBB: | |||
12191 | // ... | |||
12192 | // fallthrough --> loopMBB | |||
12193 | BB->addSuccessor(loop1MBB); | |||
12194 | ||||
12195 | // loop1MBB: | |||
12196 | // l[bhwd]arx dest, ptr | |||
12197 | // cmp[wd] dest, oldval | |||
12198 | // bne- midMBB | |||
12199 | // loop2MBB: | |||
12200 | // st[bhwd]cx. newval, ptr | |||
12201 | // bne- loopMBB | |||
12202 | // b exitBB | |||
12203 | // midMBB: | |||
12204 | // st[bhwd]cx. dest, ptr | |||
12205 | // exitBB: | |||
12206 | BB = loop1MBB; | |||
12207 | BuildMI(BB, dl, TII->get(LoadMnemonic), dest).addReg(ptrA).addReg(ptrB); | |||
12208 | BuildMI(BB, dl, TII->get(is64bit ? PPC::CMPD : PPC::CMPW), PPC::CR0) | |||
12209 | .addReg(oldval) | |||
12210 | .addReg(dest); | |||
12211 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
12212 | .addImm(PPC::PRED_NE) | |||
12213 | .addReg(PPC::CR0) | |||
12214 | .addMBB(midMBB); | |||
12215 | BB->addSuccessor(loop2MBB); | |||
12216 | BB->addSuccessor(midMBB); | |||
12217 | ||||
12218 | BB = loop2MBB; | |||
12219 | BuildMI(BB, dl, TII->get(StoreMnemonic)) | |||
12220 | .addReg(newval) | |||
12221 | .addReg(ptrA) | |||
12222 | .addReg(ptrB); | |||
12223 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
12224 | .addImm(PPC::PRED_NE) | |||
12225 | .addReg(PPC::CR0) | |||
12226 | .addMBB(loop1MBB); | |||
12227 | BuildMI(BB, dl, TII->get(PPC::B)).addMBB(exitMBB); | |||
12228 | BB->addSuccessor(loop1MBB); | |||
12229 | BB->addSuccessor(exitMBB); | |||
12230 | ||||
12231 | BB = midMBB; | |||
12232 | BuildMI(BB, dl, TII->get(StoreMnemonic)) | |||
12233 | .addReg(dest) | |||
12234 | .addReg(ptrA) | |||
12235 | .addReg(ptrB); | |||
12236 | BB->addSuccessor(exitMBB); | |||
12237 | ||||
12238 | // exitMBB: | |||
12239 | // ... | |||
12240 | BB = exitMBB; | |||
12241 | } else if (MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I8 || | |||
12242 | MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I16) { | |||
12243 | // We must use 64-bit registers for addresses when targeting 64-bit, | |||
12244 | // since we're actually doing arithmetic on them. Other registers | |||
12245 | // can be 32-bit. | |||
12246 | bool is64bit = Subtarget.isPPC64(); | |||
12247 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
12248 | bool is8bit = MI.getOpcode() == PPC::ATOMIC_CMP_SWAP_I8; | |||
12249 | ||||
12250 | Register dest = MI.getOperand(0).getReg(); | |||
12251 | Register ptrA = MI.getOperand(1).getReg(); | |||
12252 | Register ptrB = MI.getOperand(2).getReg(); | |||
12253 | Register oldval = MI.getOperand(3).getReg(); | |||
12254 | Register newval = MI.getOperand(4).getReg(); | |||
12255 | DebugLoc dl = MI.getDebugLoc(); | |||
12256 | ||||
12257 | MachineBasicBlock *loop1MBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12258 | MachineBasicBlock *loop2MBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12259 | MachineBasicBlock *midMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12260 | MachineBasicBlock *exitMBB = F->CreateMachineBasicBlock(LLVM_BB); | |||
12261 | F->insert(It, loop1MBB); | |||
12262 | F->insert(It, loop2MBB); | |||
12263 | F->insert(It, midMBB); | |||
12264 | F->insert(It, exitMBB); | |||
12265 | exitMBB->splice(exitMBB->begin(), BB, | |||
12266 | std::next(MachineBasicBlock::iterator(MI)), BB->end()); | |||
12267 | exitMBB->transferSuccessorsAndUpdatePHIs(BB); | |||
12268 | ||||
12269 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12270 | const TargetRegisterClass *RC = | |||
12271 | is64bit ? &PPC::G8RCRegClass : &PPC::GPRCRegClass; | |||
12272 | const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; | |||
12273 | ||||
12274 | Register PtrReg = RegInfo.createVirtualRegister(RC); | |||
12275 | Register Shift1Reg = RegInfo.createVirtualRegister(GPRC); | |||
12276 | Register ShiftReg = | |||
12277 | isLittleEndian ? Shift1Reg : RegInfo.createVirtualRegister(GPRC); | |||
12278 | Register NewVal2Reg = RegInfo.createVirtualRegister(GPRC); | |||
12279 | Register NewVal3Reg = RegInfo.createVirtualRegister(GPRC); | |||
12280 | Register OldVal2Reg = RegInfo.createVirtualRegister(GPRC); | |||
12281 | Register OldVal3Reg = RegInfo.createVirtualRegister(GPRC); | |||
12282 | Register MaskReg = RegInfo.createVirtualRegister(GPRC); | |||
12283 | Register Mask2Reg = RegInfo.createVirtualRegister(GPRC); | |||
12284 | Register Mask3Reg = RegInfo.createVirtualRegister(GPRC); | |||
12285 | Register Tmp2Reg = RegInfo.createVirtualRegister(GPRC); | |||
12286 | Register Tmp4Reg = RegInfo.createVirtualRegister(GPRC); | |||
12287 | Register TmpDestReg = RegInfo.createVirtualRegister(GPRC); | |||
12288 | Register Ptr1Reg; | |||
12289 | Register TmpReg = RegInfo.createVirtualRegister(GPRC); | |||
12290 | Register ZeroReg = is64bit ? PPC::ZERO8 : PPC::ZERO; | |||
12291 | // thisMBB: | |||
12292 | // ... | |||
12293 | // fallthrough --> loopMBB | |||
12294 | BB->addSuccessor(loop1MBB); | |||
12295 | ||||
12296 | // The 4-byte load must be aligned, while a char or short may be | |||
12297 | // anywhere in the word. Hence all this nasty bookkeeping code. | |||
12298 | // add ptr1, ptrA, ptrB [copy if ptrA==0] | |||
12299 | // rlwinm shift1, ptr1, 3, 27, 28 [3, 27, 27] | |||
12300 | // xori shift, shift1, 24 [16] | |||
12301 | // rlwinm ptr, ptr1, 0, 0, 29 | |||
12302 | // slw newval2, newval, shift | |||
12303 | // slw oldval2, oldval,shift | |||
12304 | // li mask2, 255 [li mask3, 0; ori mask2, mask3, 65535] | |||
12305 | // slw mask, mask2, shift | |||
12306 | // and newval3, newval2, mask | |||
12307 | // and oldval3, oldval2, mask | |||
12308 | // loop1MBB: | |||
12309 | // lwarx tmpDest, ptr | |||
12310 | // and tmp, tmpDest, mask | |||
12311 | // cmpw tmp, oldval3 | |||
12312 | // bne- midMBB | |||
12313 | // loop2MBB: | |||
12314 | // andc tmp2, tmpDest, mask | |||
12315 | // or tmp4, tmp2, newval3 | |||
12316 | // stwcx. tmp4, ptr | |||
12317 | // bne- loop1MBB | |||
12318 | // b exitBB | |||
12319 | // midMBB: | |||
12320 | // stwcx. tmpDest, ptr | |||
12321 | // exitBB: | |||
12322 | // srw dest, tmpDest, shift | |||
12323 | if (ptrA != ZeroReg) { | |||
12324 | Ptr1Reg = RegInfo.createVirtualRegister(RC); | |||
12325 | BuildMI(BB, dl, TII->get(is64bit ? PPC::ADD8 : PPC::ADD4), Ptr1Reg) | |||
12326 | .addReg(ptrA) | |||
12327 | .addReg(ptrB); | |||
12328 | } else { | |||
12329 | Ptr1Reg = ptrB; | |||
12330 | } | |||
12331 | ||||
12332 | // We need use 32-bit subregister to avoid mismatch register class in 64-bit | |||
12333 | // mode. | |||
12334 | BuildMI(BB, dl, TII->get(PPC::RLWINM), Shift1Reg) | |||
12335 | .addReg(Ptr1Reg, 0, is64bit ? PPC::sub_32 : 0) | |||
12336 | .addImm(3) | |||
12337 | .addImm(27) | |||
12338 | .addImm(is8bit ? 28 : 27); | |||
12339 | if (!isLittleEndian) | |||
12340 | BuildMI(BB, dl, TII->get(PPC::XORI), ShiftReg) | |||
12341 | .addReg(Shift1Reg) | |||
12342 | .addImm(is8bit ? 24 : 16); | |||
12343 | if (is64bit) | |||
12344 | BuildMI(BB, dl, TII->get(PPC::RLDICR), PtrReg) | |||
12345 | .addReg(Ptr1Reg) | |||
12346 | .addImm(0) | |||
12347 | .addImm(61); | |||
12348 | else | |||
12349 | BuildMI(BB, dl, TII->get(PPC::RLWINM), PtrReg) | |||
12350 | .addReg(Ptr1Reg) | |||
12351 | .addImm(0) | |||
12352 | .addImm(0) | |||
12353 | .addImm(29); | |||
12354 | BuildMI(BB, dl, TII->get(PPC::SLW), NewVal2Reg) | |||
12355 | .addReg(newval) | |||
12356 | .addReg(ShiftReg); | |||
12357 | BuildMI(BB, dl, TII->get(PPC::SLW), OldVal2Reg) | |||
12358 | .addReg(oldval) | |||
12359 | .addReg(ShiftReg); | |||
12360 | if (is8bit) | |||
12361 | BuildMI(BB, dl, TII->get(PPC::LI), Mask2Reg).addImm(255); | |||
12362 | else { | |||
12363 | BuildMI(BB, dl, TII->get(PPC::LI), Mask3Reg).addImm(0); | |||
12364 | BuildMI(BB, dl, TII->get(PPC::ORI), Mask2Reg) | |||
12365 | .addReg(Mask3Reg) | |||
12366 | .addImm(65535); | |||
12367 | } | |||
12368 | BuildMI(BB, dl, TII->get(PPC::SLW), MaskReg) | |||
12369 | .addReg(Mask2Reg) | |||
12370 | .addReg(ShiftReg); | |||
12371 | BuildMI(BB, dl, TII->get(PPC::AND), NewVal3Reg) | |||
12372 | .addReg(NewVal2Reg) | |||
12373 | .addReg(MaskReg); | |||
12374 | BuildMI(BB, dl, TII->get(PPC::AND), OldVal3Reg) | |||
12375 | .addReg(OldVal2Reg) | |||
12376 | .addReg(MaskReg); | |||
12377 | ||||
12378 | BB = loop1MBB; | |||
12379 | BuildMI(BB, dl, TII->get(PPC::LWARX), TmpDestReg) | |||
12380 | .addReg(ZeroReg) | |||
12381 | .addReg(PtrReg); | |||
12382 | BuildMI(BB, dl, TII->get(PPC::AND), TmpReg) | |||
12383 | .addReg(TmpDestReg) | |||
12384 | .addReg(MaskReg); | |||
12385 | BuildMI(BB, dl, TII->get(PPC::CMPW), PPC::CR0) | |||
12386 | .addReg(TmpReg) | |||
12387 | .addReg(OldVal3Reg); | |||
12388 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
12389 | .addImm(PPC::PRED_NE) | |||
12390 | .addReg(PPC::CR0) | |||
12391 | .addMBB(midMBB); | |||
12392 | BB->addSuccessor(loop2MBB); | |||
12393 | BB->addSuccessor(midMBB); | |||
12394 | ||||
12395 | BB = loop2MBB; | |||
12396 | BuildMI(BB, dl, TII->get(PPC::ANDC), Tmp2Reg) | |||
12397 | .addReg(TmpDestReg) | |||
12398 | .addReg(MaskReg); | |||
12399 | BuildMI(BB, dl, TII->get(PPC::OR), Tmp4Reg) | |||
12400 | .addReg(Tmp2Reg) | |||
12401 | .addReg(NewVal3Reg); | |||
12402 | BuildMI(BB, dl, TII->get(PPC::STWCX)) | |||
12403 | .addReg(Tmp4Reg) | |||
12404 | .addReg(ZeroReg) | |||
12405 | .addReg(PtrReg); | |||
12406 | BuildMI(BB, dl, TII->get(PPC::BCC)) | |||
12407 | .addImm(PPC::PRED_NE) | |||
12408 | .addReg(PPC::CR0) | |||
12409 | .addMBB(loop1MBB); | |||
12410 | BuildMI(BB, dl, TII->get(PPC::B)).addMBB(exitMBB); | |||
12411 | BB->addSuccessor(loop1MBB); | |||
12412 | BB->addSuccessor(exitMBB); | |||
12413 | ||||
12414 | BB = midMBB; | |||
12415 | BuildMI(BB, dl, TII->get(PPC::STWCX)) | |||
12416 | .addReg(TmpDestReg) | |||
12417 | .addReg(ZeroReg) | |||
12418 | .addReg(PtrReg); | |||
12419 | BB->addSuccessor(exitMBB); | |||
12420 | ||||
12421 | // exitMBB: | |||
12422 | // ... | |||
12423 | BB = exitMBB; | |||
12424 | BuildMI(*BB, BB->begin(), dl, TII->get(PPC::SRW), dest) | |||
12425 | .addReg(TmpReg) | |||
12426 | .addReg(ShiftReg); | |||
12427 | } else if (MI.getOpcode() == PPC::FADDrtz) { | |||
12428 | // This pseudo performs an FADD with rounding mode temporarily forced | |||
12429 | // to round-to-zero. We emit this via custom inserter since the FPSCR | |||
12430 | // is not modeled at the SelectionDAG level. | |||
12431 | Register Dest = MI.getOperand(0).getReg(); | |||
12432 | Register Src1 = MI.getOperand(1).getReg(); | |||
12433 | Register Src2 = MI.getOperand(2).getReg(); | |||
12434 | DebugLoc dl = MI.getDebugLoc(); | |||
12435 | ||||
12436 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12437 | Register MFFSReg = RegInfo.createVirtualRegister(&PPC::F8RCRegClass); | |||
12438 | ||||
12439 | // Save FPSCR value. | |||
12440 | BuildMI(*BB, MI, dl, TII->get(PPC::MFFS), MFFSReg); | |||
12441 | ||||
12442 | // Set rounding mode to round-to-zero. | |||
12443 | BuildMI(*BB, MI, dl, TII->get(PPC::MTFSB1)) | |||
12444 | .addImm(31) | |||
12445 | .addReg(PPC::RM, RegState::ImplicitDefine); | |||
12446 | ||||
12447 | BuildMI(*BB, MI, dl, TII->get(PPC::MTFSB0)) | |||
12448 | .addImm(30) | |||
12449 | .addReg(PPC::RM, RegState::ImplicitDefine); | |||
12450 | ||||
12451 | // Perform addition. | |||
12452 | auto MIB = BuildMI(*BB, MI, dl, TII->get(PPC::FADD), Dest) | |||
12453 | .addReg(Src1) | |||
12454 | .addReg(Src2); | |||
12455 | if (MI.getFlag(MachineInstr::NoFPExcept)) | |||
12456 | MIB.setMIFlag(MachineInstr::NoFPExcept); | |||
12457 | ||||
12458 | // Restore FPSCR value. | |||
12459 | BuildMI(*BB, MI, dl, TII->get(PPC::MTFSFb)).addImm(1).addReg(MFFSReg); | |||
12460 | } else if (MI.getOpcode() == PPC::ANDI_rec_1_EQ_BIT || | |||
12461 | MI.getOpcode() == PPC::ANDI_rec_1_GT_BIT || | |||
12462 | MI.getOpcode() == PPC::ANDI_rec_1_EQ_BIT8 || | |||
12463 | MI.getOpcode() == PPC::ANDI_rec_1_GT_BIT8) { | |||
12464 | unsigned Opcode = (MI.getOpcode() == PPC::ANDI_rec_1_EQ_BIT8 || | |||
12465 | MI.getOpcode() == PPC::ANDI_rec_1_GT_BIT8) | |||
12466 | ? PPC::ANDI8_rec | |||
12467 | : PPC::ANDI_rec; | |||
12468 | bool IsEQ = (MI.getOpcode() == PPC::ANDI_rec_1_EQ_BIT || | |||
12469 | MI.getOpcode() == PPC::ANDI_rec_1_EQ_BIT8); | |||
12470 | ||||
12471 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12472 | Register Dest = RegInfo.createVirtualRegister( | |||
12473 | Opcode == PPC::ANDI_rec ? &PPC::GPRCRegClass : &PPC::G8RCRegClass); | |||
12474 | ||||
12475 | DebugLoc Dl = MI.getDebugLoc(); | |||
12476 | BuildMI(*BB, MI, Dl, TII->get(Opcode), Dest) | |||
12477 | .addReg(MI.getOperand(1).getReg()) | |||
12478 | .addImm(1); | |||
12479 | BuildMI(*BB, MI, Dl, TII->get(TargetOpcode::COPY), | |||
12480 | MI.getOperand(0).getReg()) | |||
12481 | .addReg(IsEQ ? PPC::CR0EQ : PPC::CR0GT); | |||
12482 | } else if (MI.getOpcode() == PPC::TCHECK_RET) { | |||
12483 | DebugLoc Dl = MI.getDebugLoc(); | |||
12484 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12485 | Register CRReg = RegInfo.createVirtualRegister(&PPC::CRRCRegClass); | |||
12486 | BuildMI(*BB, MI, Dl, TII->get(PPC::TCHECK), CRReg); | |||
12487 | BuildMI(*BB, MI, Dl, TII->get(TargetOpcode::COPY), | |||
12488 | MI.getOperand(0).getReg()) | |||
12489 | .addReg(CRReg); | |||
12490 | } else if (MI.getOpcode() == PPC::TBEGIN_RET) { | |||
12491 | DebugLoc Dl = MI.getDebugLoc(); | |||
12492 | unsigned Imm = MI.getOperand(1).getImm(); | |||
12493 | BuildMI(*BB, MI, Dl, TII->get(PPC::TBEGIN)).addImm(Imm); | |||
12494 | BuildMI(*BB, MI, Dl, TII->get(TargetOpcode::COPY), | |||
12495 | MI.getOperand(0).getReg()) | |||
12496 | .addReg(PPC::CR0EQ); | |||
12497 | } else if (MI.getOpcode() == PPC::SETRNDi) { | |||
12498 | DebugLoc dl = MI.getDebugLoc(); | |||
12499 | Register OldFPSCRReg = MI.getOperand(0).getReg(); | |||
12500 | ||||
12501 | // Save FPSCR value. | |||
12502 | BuildMI(*BB, MI, dl, TII->get(PPC::MFFS), OldFPSCRReg); | |||
12503 | ||||
12504 | // The floating point rounding mode is in the bits 62:63 of FPCSR, and has | |||
12505 | // the following settings: | |||
12506 | // 00 Round to nearest | |||
12507 | // 01 Round to 0 | |||
12508 | // 10 Round to +inf | |||
12509 | // 11 Round to -inf | |||
12510 | ||||
12511 | // When the operand is immediate, using the two least significant bits of | |||
12512 | // the immediate to set the bits 62:63 of FPSCR. | |||
12513 | unsigned Mode = MI.getOperand(1).getImm(); | |||
12514 | BuildMI(*BB, MI, dl, TII->get((Mode & 1) ? PPC::MTFSB1 : PPC::MTFSB0)) | |||
12515 | .addImm(31) | |||
12516 | .addReg(PPC::RM, RegState::ImplicitDefine); | |||
12517 | ||||
12518 | BuildMI(*BB, MI, dl, TII->get((Mode & 2) ? PPC::MTFSB1 : PPC::MTFSB0)) | |||
12519 | .addImm(30) | |||
12520 | .addReg(PPC::RM, RegState::ImplicitDefine); | |||
12521 | } else if (MI.getOpcode() == PPC::SETRND) { | |||
12522 | DebugLoc dl = MI.getDebugLoc(); | |||
12523 | ||||
12524 | // Copy register from F8RCRegClass::SrcReg to G8RCRegClass::DestReg | |||
12525 | // or copy register from G8RCRegClass::SrcReg to F8RCRegClass::DestReg. | |||
12526 | // If the target doesn't have DirectMove, we should use stack to do the | |||
12527 | // conversion, because the target doesn't have the instructions like mtvsrd | |||
12528 | // or mfvsrd to do this conversion directly. | |||
12529 | auto copyRegFromG8RCOrF8RC = [&] (unsigned DestReg, unsigned SrcReg) { | |||
12530 | if (Subtarget.hasDirectMove()) { | |||
12531 | BuildMI(*BB, MI, dl, TII->get(TargetOpcode::COPY), DestReg) | |||
12532 | .addReg(SrcReg); | |||
12533 | } else { | |||
12534 | // Use stack to do the register copy. | |||
12535 | unsigned StoreOp = PPC::STD, LoadOp = PPC::LFD; | |||
12536 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12537 | const TargetRegisterClass *RC = RegInfo.getRegClass(SrcReg); | |||
12538 | if (RC == &PPC::F8RCRegClass) { | |||
12539 | // Copy register from F8RCRegClass to G8RCRegclass. | |||
12540 | assert((RegInfo.getRegClass(DestReg) == &PPC::G8RCRegClass) &&(static_cast <bool> ((RegInfo.getRegClass(DestReg) == & PPC::G8RCRegClass) && "Unsupported RegClass.") ? void (0) : __assert_fail ("(RegInfo.getRegClass(DestReg) == &PPC::G8RCRegClass) && \"Unsupported RegClass.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12541, __extension__ __PRETTY_FUNCTION__)) | |||
12541 | "Unsupported RegClass.")(static_cast <bool> ((RegInfo.getRegClass(DestReg) == & PPC::G8RCRegClass) && "Unsupported RegClass.") ? void (0) : __assert_fail ("(RegInfo.getRegClass(DestReg) == &PPC::G8RCRegClass) && \"Unsupported RegClass.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12541, __extension__ __PRETTY_FUNCTION__)); | |||
12542 | ||||
12543 | StoreOp = PPC::STFD; | |||
12544 | LoadOp = PPC::LD; | |||
12545 | } else { | |||
12546 | // Copy register from G8RCRegClass to F8RCRegclass. | |||
12547 | assert((RegInfo.getRegClass(SrcReg) == &PPC::G8RCRegClass) &&(static_cast <bool> ((RegInfo.getRegClass(SrcReg) == & PPC::G8RCRegClass) && (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) && "Unsupported RegClass.") ? void (0) : __assert_fail ("(RegInfo.getRegClass(SrcReg) == &PPC::G8RCRegClass) && (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) && \"Unsupported RegClass.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12549, __extension__ __PRETTY_FUNCTION__)) | |||
12548 | (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) &&(static_cast <bool> ((RegInfo.getRegClass(SrcReg) == & PPC::G8RCRegClass) && (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) && "Unsupported RegClass.") ? void (0) : __assert_fail ("(RegInfo.getRegClass(SrcReg) == &PPC::G8RCRegClass) && (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) && \"Unsupported RegClass.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12549, __extension__ __PRETTY_FUNCTION__)) | |||
12549 | "Unsupported RegClass.")(static_cast <bool> ((RegInfo.getRegClass(SrcReg) == & PPC::G8RCRegClass) && (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) && "Unsupported RegClass.") ? void (0) : __assert_fail ("(RegInfo.getRegClass(SrcReg) == &PPC::G8RCRegClass) && (RegInfo.getRegClass(DestReg) == &PPC::F8RCRegClass) && \"Unsupported RegClass.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12549, __extension__ __PRETTY_FUNCTION__)); | |||
12550 | } | |||
12551 | ||||
12552 | MachineFrameInfo &MFI = F->getFrameInfo(); | |||
12553 | int FrameIdx = MFI.CreateStackObject(8, Align(8), false); | |||
12554 | ||||
12555 | MachineMemOperand *MMOStore = F->getMachineMemOperand( | |||
12556 | MachinePointerInfo::getFixedStack(*F, FrameIdx, 0), | |||
12557 | MachineMemOperand::MOStore, MFI.getObjectSize(FrameIdx), | |||
12558 | MFI.getObjectAlign(FrameIdx)); | |||
12559 | ||||
12560 | // Store the SrcReg into the stack. | |||
12561 | BuildMI(*BB, MI, dl, TII->get(StoreOp)) | |||
12562 | .addReg(SrcReg) | |||
12563 | .addImm(0) | |||
12564 | .addFrameIndex(FrameIdx) | |||
12565 | .addMemOperand(MMOStore); | |||
12566 | ||||
12567 | MachineMemOperand *MMOLoad = F->getMachineMemOperand( | |||
12568 | MachinePointerInfo::getFixedStack(*F, FrameIdx, 0), | |||
12569 | MachineMemOperand::MOLoad, MFI.getObjectSize(FrameIdx), | |||
12570 | MFI.getObjectAlign(FrameIdx)); | |||
12571 | ||||
12572 | // Load from the stack where SrcReg is stored, and save to DestReg, | |||
12573 | // so we have done the RegClass conversion from RegClass::SrcReg to | |||
12574 | // RegClass::DestReg. | |||
12575 | BuildMI(*BB, MI, dl, TII->get(LoadOp), DestReg) | |||
12576 | .addImm(0) | |||
12577 | .addFrameIndex(FrameIdx) | |||
12578 | .addMemOperand(MMOLoad); | |||
12579 | } | |||
12580 | }; | |||
12581 | ||||
12582 | Register OldFPSCRReg = MI.getOperand(0).getReg(); | |||
12583 | ||||
12584 | // Save FPSCR value. | |||
12585 | BuildMI(*BB, MI, dl, TII->get(PPC::MFFS), OldFPSCRReg); | |||
12586 | ||||
12587 | // When the operand is gprc register, use two least significant bits of the | |||
12588 | // register and mtfsf instruction to set the bits 62:63 of FPSCR. | |||
12589 | // | |||
12590 | // copy OldFPSCRTmpReg, OldFPSCRReg | |||
12591 | // (INSERT_SUBREG ExtSrcReg, (IMPLICIT_DEF ImDefReg), SrcOp, 1) | |||
12592 | // rldimi NewFPSCRTmpReg, ExtSrcReg, OldFPSCRReg, 0, 62 | |||
12593 | // copy NewFPSCRReg, NewFPSCRTmpReg | |||
12594 | // mtfsf 255, NewFPSCRReg | |||
12595 | MachineOperand SrcOp = MI.getOperand(1); | |||
12596 | MachineRegisterInfo &RegInfo = F->getRegInfo(); | |||
12597 | Register OldFPSCRTmpReg = RegInfo.createVirtualRegister(&PPC::G8RCRegClass); | |||
12598 | ||||
12599 | copyRegFromG8RCOrF8RC(OldFPSCRTmpReg, OldFPSCRReg); | |||
12600 | ||||
12601 | Register ImDefReg = RegInfo.createVirtualRegister(&PPC::G8RCRegClass); | |||
12602 | Register ExtSrcReg = RegInfo.createVirtualRegister(&PPC::G8RCRegClass); | |||
12603 | ||||
12604 | // The first operand of INSERT_SUBREG should be a register which has | |||
12605 | // subregisters, we only care about its RegClass, so we should use an | |||
12606 | // IMPLICIT_DEF register. | |||
12607 | BuildMI(*BB, MI, dl, TII->get(TargetOpcode::IMPLICIT_DEF), ImDefReg); | |||
12608 | BuildMI(*BB, MI, dl, TII->get(PPC::INSERT_SUBREG), ExtSrcReg) | |||
12609 | .addReg(ImDefReg) | |||
12610 | .add(SrcOp) | |||
12611 | .addImm(1); | |||
12612 | ||||
12613 | Register NewFPSCRTmpReg = RegInfo.createVirtualRegister(&PPC::G8RCRegClass); | |||
12614 | BuildMI(*BB, MI, dl, TII->get(PPC::RLDIMI), NewFPSCRTmpReg) | |||
12615 | .addReg(OldFPSCRTmpReg) | |||
12616 | .addReg(ExtSrcReg) | |||
12617 | .addImm(0) | |||
12618 | .addImm(62); | |||
12619 | ||||
12620 | Register NewFPSCRReg = RegInfo.createVirtualRegister(&PPC::F8RCRegClass); | |||
12621 | copyRegFromG8RCOrF8RC(NewFPSCRReg, NewFPSCRTmpReg); | |||
12622 | ||||
12623 | // The mask 255 means that put the 32:63 bits of NewFPSCRReg to the 32:63 | |||
12624 | // bits of FPSCR. | |||
12625 | BuildMI(*BB, MI, dl, TII->get(PPC::MTFSF)) | |||
12626 | .addImm(255) | |||
12627 | .addReg(NewFPSCRReg) | |||
12628 | .addImm(0) | |||
12629 | .addImm(0); | |||
12630 | } else if (MI.getOpcode() == PPC::SETFLM) { | |||
12631 | DebugLoc Dl = MI.getDebugLoc(); | |||
12632 | ||||
12633 | // Result of setflm is previous FPSCR content, so we need to save it first. | |||
12634 | Register OldFPSCRReg = MI.getOperand(0).getReg(); | |||
12635 | BuildMI(*BB, MI, Dl, TII->get(PPC::MFFS), OldFPSCRReg); | |||
12636 | ||||
12637 | // Put bits in 32:63 to FPSCR. | |||
12638 | Register NewFPSCRReg = MI.getOperand(1).getReg(); | |||
12639 | BuildMI(*BB, MI, Dl, TII->get(PPC::MTFSF)) | |||
12640 | .addImm(255) | |||
12641 | .addReg(NewFPSCRReg) | |||
12642 | .addImm(0) | |||
12643 | .addImm(0); | |||
12644 | } else if (MI.getOpcode() == PPC::PROBED_ALLOCA_32 || | |||
12645 | MI.getOpcode() == PPC::PROBED_ALLOCA_64) { | |||
12646 | return emitProbedAlloca(MI, BB); | |||
12647 | } else if (MI.getOpcode() == PPC::SPLIT_QUADWORD) { | |||
12648 | DebugLoc DL = MI.getDebugLoc(); | |||
12649 | Register Src = MI.getOperand(2).getReg(); | |||
12650 | Register Lo = MI.getOperand(0).getReg(); | |||
12651 | Register Hi = MI.getOperand(1).getReg(); | |||
12652 | BuildMI(*BB, MI, DL, TII->get(TargetOpcode::COPY)) | |||
12653 | .addDef(Lo) | |||
12654 | .addUse(Src, 0, PPC::sub_gp8_x1); | |||
12655 | BuildMI(*BB, MI, DL, TII->get(TargetOpcode::COPY)) | |||
12656 | .addDef(Hi) | |||
12657 | .addUse(Src, 0, PPC::sub_gp8_x0); | |||
12658 | } else { | |||
12659 | llvm_unreachable("Unexpected instr type to insert")::llvm::llvm_unreachable_internal("Unexpected instr type to insert" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12659); | |||
12660 | } | |||
12661 | ||||
12662 | MI.eraseFromParent(); // The pseudo instruction is gone now. | |||
12663 | return BB; | |||
12664 | } | |||
12665 | ||||
12666 | //===----------------------------------------------------------------------===// | |||
12667 | // Target Optimization Hooks | |||
12668 | //===----------------------------------------------------------------------===// | |||
12669 | ||||
12670 | static int getEstimateRefinementSteps(EVT VT, const PPCSubtarget &Subtarget) { | |||
12671 | // For the estimates, convergence is quadratic, so we essentially double the | |||
12672 | // number of digits correct after every iteration. For both FRE and FRSQRTE, | |||
12673 | // the minimum architected relative accuracy is 2^-5. When hasRecipPrec(), | |||
12674 | // this is 2^-14. IEEE float has 23 digits and double has 52 digits. | |||
12675 | int RefinementSteps = Subtarget.hasRecipPrec() ? 1 : 3; | |||
12676 | if (VT.getScalarType() == MVT::f64) | |||
12677 | RefinementSteps++; | |||
12678 | return RefinementSteps; | |||
12679 | } | |||
12680 | ||||
12681 | SDValue PPCTargetLowering::getSqrtInputTest(SDValue Op, SelectionDAG &DAG, | |||
12682 | const DenormalMode &Mode) const { | |||
12683 | // We only have VSX Vector Test for software Square Root. | |||
12684 | EVT VT = Op.getValueType(); | |||
12685 | if (!isTypeLegal(MVT::i1) || | |||
12686 | (VT != MVT::f64 && | |||
12687 | ((VT != MVT::v2f64 && VT != MVT::v4f32) || !Subtarget.hasVSX()))) | |||
12688 | return TargetLowering::getSqrtInputTest(Op, DAG, Mode); | |||
12689 | ||||
12690 | SDLoc DL(Op); | |||
12691 | // The output register of FTSQRT is CR field. | |||
12692 | SDValue FTSQRT = DAG.getNode(PPCISD::FTSQRT, DL, MVT::i32, Op); | |||
12693 | // ftsqrt BF,FRB | |||
12694 | // Let e_b be the unbiased exponent of the double-precision | |||
12695 | // floating-point operand in register FRB. | |||
12696 | // fe_flag is set to 1 if either of the following conditions occurs. | |||
12697 | // - The double-precision floating-point operand in register FRB is a zero, | |||
12698 | // a NaN, or an infinity, or a negative value. | |||
12699 | // - e_b is less than or equal to -970. | |||
12700 | // Otherwise fe_flag is set to 0. | |||
12701 | // Both VSX and non-VSX versions would set EQ bit in the CR if the number is | |||
12702 | // not eligible for iteration. (zero/negative/infinity/nan or unbiased | |||
12703 | // exponent is less than -970) | |||
12704 | SDValue SRIdxVal = DAG.getTargetConstant(PPC::sub_eq, DL, MVT::i32); | |||
12705 | return SDValue(DAG.getMachineNode(TargetOpcode::EXTRACT_SUBREG, DL, MVT::i1, | |||
12706 | FTSQRT, SRIdxVal), | |||
12707 | 0); | |||
12708 | } | |||
12709 | ||||
12710 | SDValue | |||
12711 | PPCTargetLowering::getSqrtResultForDenormInput(SDValue Op, | |||
12712 | SelectionDAG &DAG) const { | |||
12713 | // We only have VSX Vector Square Root. | |||
12714 | EVT VT = Op.getValueType(); | |||
12715 | if (VT != MVT::f64 && | |||
12716 | ((VT != MVT::v2f64 && VT != MVT::v4f32) || !Subtarget.hasVSX())) | |||
12717 | return TargetLowering::getSqrtResultForDenormInput(Op, DAG); | |||
12718 | ||||
12719 | return DAG.getNode(PPCISD::FSQRT, SDLoc(Op), VT, Op); | |||
12720 | } | |||
12721 | ||||
12722 | SDValue PPCTargetLowering::getSqrtEstimate(SDValue Operand, SelectionDAG &DAG, | |||
12723 | int Enabled, int &RefinementSteps, | |||
12724 | bool &UseOneConstNR, | |||
12725 | bool Reciprocal) const { | |||
12726 | EVT VT = Operand.getValueType(); | |||
12727 | if ((VT == MVT::f32 && Subtarget.hasFRSQRTES()) || | |||
12728 | (VT == MVT::f64 && Subtarget.hasFRSQRTE()) || | |||
12729 | (VT == MVT::v4f32 && Subtarget.hasAltivec()) || | |||
12730 | (VT == MVT::v2f64 && Subtarget.hasVSX())) { | |||
12731 | if (RefinementSteps == ReciprocalEstimate::Unspecified) | |||
12732 | RefinementSteps = getEstimateRefinementSteps(VT, Subtarget); | |||
12733 | ||||
12734 | // The Newton-Raphson computation with a single constant does not provide | |||
12735 | // enough accuracy on some CPUs. | |||
12736 | UseOneConstNR = !Subtarget.needsTwoConstNR(); | |||
12737 | return DAG.getNode(PPCISD::FRSQRTE, SDLoc(Operand), VT, Operand); | |||
12738 | } | |||
12739 | return SDValue(); | |||
12740 | } | |||
12741 | ||||
12742 | SDValue PPCTargetLowering::getRecipEstimate(SDValue Operand, SelectionDAG &DAG, | |||
12743 | int Enabled, | |||
12744 | int &RefinementSteps) const { | |||
12745 | EVT VT = Operand.getValueType(); | |||
12746 | if ((VT == MVT::f32 && Subtarget.hasFRES()) || | |||
12747 | (VT == MVT::f64 && Subtarget.hasFRE()) || | |||
12748 | (VT == MVT::v4f32 && Subtarget.hasAltivec()) || | |||
12749 | (VT == MVT::v2f64 && Subtarget.hasVSX())) { | |||
12750 | if (RefinementSteps == ReciprocalEstimate::Unspecified) | |||
12751 | RefinementSteps = getEstimateRefinementSteps(VT, Subtarget); | |||
12752 | return DAG.getNode(PPCISD::FRE, SDLoc(Operand), VT, Operand); | |||
12753 | } | |||
12754 | return SDValue(); | |||
12755 | } | |||
12756 | ||||
12757 | unsigned PPCTargetLowering::combineRepeatedFPDivisors() const { | |||
12758 | // Note: This functionality is used only when unsafe-fp-math is enabled, and | |||
12759 | // on cores with reciprocal estimates (which are used when unsafe-fp-math is | |||
12760 | // enabled for division), this functionality is redundant with the default | |||
12761 | // combiner logic (once the division -> reciprocal/multiply transformation | |||
12762 | // has taken place). As a result, this matters more for older cores than for | |||
12763 | // newer ones. | |||
12764 | ||||
12765 | // Combine multiple FDIVs with the same divisor into multiple FMULs by the | |||
12766 | // reciprocal if there are two or more FDIVs (for embedded cores with only | |||
12767 | // one FP pipeline) for three or more FDIVs (for generic OOO cores). | |||
12768 | switch (Subtarget.getCPUDirective()) { | |||
12769 | default: | |||
12770 | return 3; | |||
12771 | case PPC::DIR_440: | |||
12772 | case PPC::DIR_A2: | |||
12773 | case PPC::DIR_E500: | |||
12774 | case PPC::DIR_E500mc: | |||
12775 | case PPC::DIR_E5500: | |||
12776 | return 2; | |||
12777 | } | |||
12778 | } | |||
12779 | ||||
12780 | // isConsecutiveLSLoc needs to work even if all adds have not yet been | |||
12781 | // collapsed, and so we need to look through chains of them. | |||
12782 | static void getBaseWithConstantOffset(SDValue Loc, SDValue &Base, | |||
12783 | int64_t& Offset, SelectionDAG &DAG) { | |||
12784 | if (DAG.isBaseWithConstantOffset(Loc)) { | |||
12785 | Base = Loc.getOperand(0); | |||
12786 | Offset += cast<ConstantSDNode>(Loc.getOperand(1))->getSExtValue(); | |||
12787 | ||||
12788 | // The base might itself be a base plus an offset, and if so, accumulate | |||
12789 | // that as well. | |||
12790 | getBaseWithConstantOffset(Loc.getOperand(0), Base, Offset, DAG); | |||
12791 | } | |||
12792 | } | |||
12793 | ||||
12794 | static bool isConsecutiveLSLoc(SDValue Loc, EVT VT, LSBaseSDNode *Base, | |||
12795 | unsigned Bytes, int Dist, | |||
12796 | SelectionDAG &DAG) { | |||
12797 | if (VT.getSizeInBits() / 8 != Bytes) | |||
12798 | return false; | |||
12799 | ||||
12800 | SDValue BaseLoc = Base->getBasePtr(); | |||
12801 | if (Loc.getOpcode() == ISD::FrameIndex) { | |||
12802 | if (BaseLoc.getOpcode() != ISD::FrameIndex) | |||
12803 | return false; | |||
12804 | const MachineFrameInfo &MFI = DAG.getMachineFunction().getFrameInfo(); | |||
12805 | int FI = cast<FrameIndexSDNode>(Loc)->getIndex(); | |||
12806 | int BFI = cast<FrameIndexSDNode>(BaseLoc)->getIndex(); | |||
12807 | int FS = MFI.getObjectSize(FI); | |||
12808 | int BFS = MFI.getObjectSize(BFI); | |||
12809 | if (FS != BFS || FS != (int)Bytes) return false; | |||
12810 | return MFI.getObjectOffset(FI) == (MFI.getObjectOffset(BFI) + Dist*Bytes); | |||
12811 | } | |||
12812 | ||||
12813 | SDValue Base1 = Loc, Base2 = BaseLoc; | |||
12814 | int64_t Offset1 = 0, Offset2 = 0; | |||
12815 | getBaseWithConstantOffset(Loc, Base1, Offset1, DAG); | |||
12816 | getBaseWithConstantOffset(BaseLoc, Base2, Offset2, DAG); | |||
12817 | if (Base1 == Base2 && Offset1 == (Offset2 + Dist * Bytes)) | |||
12818 | return true; | |||
12819 | ||||
12820 | const TargetLowering &TLI = DAG.getTargetLoweringInfo(); | |||
12821 | const GlobalValue *GV1 = nullptr; | |||
12822 | const GlobalValue *GV2 = nullptr; | |||
12823 | Offset1 = 0; | |||
12824 | Offset2 = 0; | |||
12825 | bool isGA1 = TLI.isGAPlusOffset(Loc.getNode(), GV1, Offset1); | |||
12826 | bool isGA2 = TLI.isGAPlusOffset(BaseLoc.getNode(), GV2, Offset2); | |||
12827 | if (isGA1 && isGA2 && GV1 == GV2) | |||
12828 | return Offset1 == (Offset2 + Dist*Bytes); | |||
12829 | return false; | |||
12830 | } | |||
12831 | ||||
12832 | // Like SelectionDAG::isConsecutiveLoad, but also works for stores, and does | |||
12833 | // not enforce equality of the chain operands. | |||
12834 | static bool isConsecutiveLS(SDNode *N, LSBaseSDNode *Base, | |||
12835 | unsigned Bytes, int Dist, | |||
12836 | SelectionDAG &DAG) { | |||
12837 | if (LSBaseSDNode *LS = dyn_cast<LSBaseSDNode>(N)) { | |||
12838 | EVT VT = LS->getMemoryVT(); | |||
12839 | SDValue Loc = LS->getBasePtr(); | |||
12840 | return isConsecutiveLSLoc(Loc, VT, Base, Bytes, Dist, DAG); | |||
12841 | } | |||
12842 | ||||
12843 | if (N->getOpcode() == ISD::INTRINSIC_W_CHAIN) { | |||
12844 | EVT VT; | |||
12845 | switch (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue()) { | |||
12846 | default: return false; | |||
12847 | case Intrinsic::ppc_altivec_lvx: | |||
12848 | case Intrinsic::ppc_altivec_lvxl: | |||
12849 | case Intrinsic::ppc_vsx_lxvw4x: | |||
12850 | case Intrinsic::ppc_vsx_lxvw4x_be: | |||
12851 | VT = MVT::v4i32; | |||
12852 | break; | |||
12853 | case Intrinsic::ppc_vsx_lxvd2x: | |||
12854 | case Intrinsic::ppc_vsx_lxvd2x_be: | |||
12855 | VT = MVT::v2f64; | |||
12856 | break; | |||
12857 | case Intrinsic::ppc_altivec_lvebx: | |||
12858 | VT = MVT::i8; | |||
12859 | break; | |||
12860 | case Intrinsic::ppc_altivec_lvehx: | |||
12861 | VT = MVT::i16; | |||
12862 | break; | |||
12863 | case Intrinsic::ppc_altivec_lvewx: | |||
12864 | VT = MVT::i32; | |||
12865 | break; | |||
12866 | } | |||
12867 | ||||
12868 | return isConsecutiveLSLoc(N->getOperand(2), VT, Base, Bytes, Dist, DAG); | |||
12869 | } | |||
12870 | ||||
12871 | if (N->getOpcode() == ISD::INTRINSIC_VOID) { | |||
12872 | EVT VT; | |||
12873 | switch (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue()) { | |||
12874 | default: return false; | |||
12875 | case Intrinsic::ppc_altivec_stvx: | |||
12876 | case Intrinsic::ppc_altivec_stvxl: | |||
12877 | case Intrinsic::ppc_vsx_stxvw4x: | |||
12878 | VT = MVT::v4i32; | |||
12879 | break; | |||
12880 | case Intrinsic::ppc_vsx_stxvd2x: | |||
12881 | VT = MVT::v2f64; | |||
12882 | break; | |||
12883 | case Intrinsic::ppc_vsx_stxvw4x_be: | |||
12884 | VT = MVT::v4i32; | |||
12885 | break; | |||
12886 | case Intrinsic::ppc_vsx_stxvd2x_be: | |||
12887 | VT = MVT::v2f64; | |||
12888 | break; | |||
12889 | case Intrinsic::ppc_altivec_stvebx: | |||
12890 | VT = MVT::i8; | |||
12891 | break; | |||
12892 | case Intrinsic::ppc_altivec_stvehx: | |||
12893 | VT = MVT::i16; | |||
12894 | break; | |||
12895 | case Intrinsic::ppc_altivec_stvewx: | |||
12896 | VT = MVT::i32; | |||
12897 | break; | |||
12898 | } | |||
12899 | ||||
12900 | return isConsecutiveLSLoc(N->getOperand(3), VT, Base, Bytes, Dist, DAG); | |||
12901 | } | |||
12902 | ||||
12903 | return false; | |||
12904 | } | |||
12905 | ||||
12906 | // Return true is there is a nearyby consecutive load to the one provided | |||
12907 | // (regardless of alignment). We search up and down the chain, looking though | |||
12908 | // token factors and other loads (but nothing else). As a result, a true result | |||
12909 | // indicates that it is safe to create a new consecutive load adjacent to the | |||
12910 | // load provided. | |||
12911 | static bool findConsecutiveLoad(LoadSDNode *LD, SelectionDAG &DAG) { | |||
12912 | SDValue Chain = LD->getChain(); | |||
12913 | EVT VT = LD->getMemoryVT(); | |||
12914 | ||||
12915 | SmallSet<SDNode *, 16> LoadRoots; | |||
12916 | SmallVector<SDNode *, 8> Queue(1, Chain.getNode()); | |||
12917 | SmallSet<SDNode *, 16> Visited; | |||
12918 | ||||
12919 | // First, search up the chain, branching to follow all token-factor operands. | |||
12920 | // If we find a consecutive load, then we're done, otherwise, record all | |||
12921 | // nodes just above the top-level loads and token factors. | |||
12922 | while (!Queue.empty()) { | |||
12923 | SDNode *ChainNext = Queue.pop_back_val(); | |||
12924 | if (!Visited.insert(ChainNext).second) | |||
12925 | continue; | |||
12926 | ||||
12927 | if (MemSDNode *ChainLD = dyn_cast<MemSDNode>(ChainNext)) { | |||
12928 | if (isConsecutiveLS(ChainLD, LD, VT.getStoreSize(), 1, DAG)) | |||
12929 | return true; | |||
12930 | ||||
12931 | if (!Visited.count(ChainLD->getChain().getNode())) | |||
12932 | Queue.push_back(ChainLD->getChain().getNode()); | |||
12933 | } else if (ChainNext->getOpcode() == ISD::TokenFactor) { | |||
12934 | for (const SDUse &O : ChainNext->ops()) | |||
12935 | if (!Visited.count(O.getNode())) | |||
12936 | Queue.push_back(O.getNode()); | |||
12937 | } else | |||
12938 | LoadRoots.insert(ChainNext); | |||
12939 | } | |||
12940 | ||||
12941 | // Second, search down the chain, starting from the top-level nodes recorded | |||
12942 | // in the first phase. These top-level nodes are the nodes just above all | |||
12943 | // loads and token factors. Starting with their uses, recursively look though | |||
12944 | // all loads (just the chain uses) and token factors to find a consecutive | |||
12945 | // load. | |||
12946 | Visited.clear(); | |||
12947 | Queue.clear(); | |||
12948 | ||||
12949 | for (SmallSet<SDNode *, 16>::iterator I = LoadRoots.begin(), | |||
12950 | IE = LoadRoots.end(); I != IE; ++I) { | |||
12951 | Queue.push_back(*I); | |||
12952 | ||||
12953 | while (!Queue.empty()) { | |||
12954 | SDNode *LoadRoot = Queue.pop_back_val(); | |||
12955 | if (!Visited.insert(LoadRoot).second) | |||
12956 | continue; | |||
12957 | ||||
12958 | if (MemSDNode *ChainLD = dyn_cast<MemSDNode>(LoadRoot)) | |||
12959 | if (isConsecutiveLS(ChainLD, LD, VT.getStoreSize(), 1, DAG)) | |||
12960 | return true; | |||
12961 | ||||
12962 | for (SDNode::use_iterator UI = LoadRoot->use_begin(), | |||
12963 | UE = LoadRoot->use_end(); UI != UE; ++UI) | |||
12964 | if (((isa<MemSDNode>(*UI) && | |||
12965 | cast<MemSDNode>(*UI)->getChain().getNode() == LoadRoot) || | |||
12966 | UI->getOpcode() == ISD::TokenFactor) && !Visited.count(*UI)) | |||
12967 | Queue.push_back(*UI); | |||
12968 | } | |||
12969 | } | |||
12970 | ||||
12971 | return false; | |||
12972 | } | |||
12973 | ||||
12974 | /// This function is called when we have proved that a SETCC node can be replaced | |||
12975 | /// by subtraction (and other supporting instructions) so that the result of | |||
12976 | /// comparison is kept in a GPR instead of CR. This function is purely for | |||
12977 | /// codegen purposes and has some flags to guide the codegen process. | |||
12978 | static SDValue generateEquivalentSub(SDNode *N, int Size, bool Complement, | |||
12979 | bool Swap, SDLoc &DL, SelectionDAG &DAG) { | |||
12980 | assert(N->getOpcode() == ISD::SETCC && "ISD::SETCC Expected.")(static_cast <bool> (N->getOpcode() == ISD::SETCC && "ISD::SETCC Expected.") ? void (0) : __assert_fail ("N->getOpcode() == ISD::SETCC && \"ISD::SETCC Expected.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 12980, __extension__ __PRETTY_FUNCTION__)); | |||
12981 | ||||
12982 | // Zero extend the operands to the largest legal integer. Originally, they | |||
12983 | // must be of a strictly smaller size. | |||
12984 | auto Op0 = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i64, N->getOperand(0), | |||
12985 | DAG.getConstant(Size, DL, MVT::i32)); | |||
12986 | auto Op1 = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i64, N->getOperand(1), | |||
12987 | DAG.getConstant(Size, DL, MVT::i32)); | |||
12988 | ||||
12989 | // Swap if needed. Depends on the condition code. | |||
12990 | if (Swap) | |||
12991 | std::swap(Op0, Op1); | |||
12992 | ||||
12993 | // Subtract extended integers. | |||
12994 | auto SubNode = DAG.getNode(ISD::SUB, DL, MVT::i64, Op0, Op1); | |||
12995 | ||||
12996 | // Move the sign bit to the least significant position and zero out the rest. | |||
12997 | // Now the least significant bit carries the result of original comparison. | |||
12998 | auto Shifted = DAG.getNode(ISD::SRL, DL, MVT::i64, SubNode, | |||
12999 | DAG.getConstant(Size - 1, DL, MVT::i32)); | |||
13000 | auto Final = Shifted; | |||
13001 | ||||
13002 | // Complement the result if needed. Based on the condition code. | |||
13003 | if (Complement) | |||
13004 | Final = DAG.getNode(ISD::XOR, DL, MVT::i64, Shifted, | |||
13005 | DAG.getConstant(1, DL, MVT::i64)); | |||
13006 | ||||
13007 | return DAG.getNode(ISD::TRUNCATE, DL, MVT::i1, Final); | |||
13008 | } | |||
13009 | ||||
13010 | SDValue PPCTargetLowering::ConvertSETCCToSubtract(SDNode *N, | |||
13011 | DAGCombinerInfo &DCI) const { | |||
13012 | assert(N->getOpcode() == ISD::SETCC && "ISD::SETCC Expected.")(static_cast <bool> (N->getOpcode() == ISD::SETCC && "ISD::SETCC Expected.") ? void (0) : __assert_fail ("N->getOpcode() == ISD::SETCC && \"ISD::SETCC Expected.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13012, __extension__ __PRETTY_FUNCTION__)); | |||
13013 | ||||
13014 | SelectionDAG &DAG = DCI.DAG; | |||
13015 | SDLoc DL(N); | |||
13016 | ||||
13017 | // Size of integers being compared has a critical role in the following | |||
13018 | // analysis, so we prefer to do this when all types are legal. | |||
13019 | if (!DCI.isAfterLegalizeDAG()) | |||
13020 | return SDValue(); | |||
13021 | ||||
13022 | // If all users of SETCC extend its value to a legal integer type | |||
13023 | // then we replace SETCC with a subtraction | |||
13024 | for (SDNode::use_iterator UI = N->use_begin(), | |||
13025 | UE = N->use_end(); UI != UE; ++UI) { | |||
13026 | if (UI->getOpcode() != ISD::ZERO_EXTEND) | |||
13027 | return SDValue(); | |||
13028 | } | |||
13029 | ||||
13030 | ISD::CondCode CC = cast<CondCodeSDNode>(N->getOperand(2))->get(); | |||
13031 | auto OpSize = N->getOperand(0).getValueSizeInBits(); | |||
13032 | ||||
13033 | unsigned Size = DAG.getDataLayout().getLargestLegalIntTypeSizeInBits(); | |||
13034 | ||||
13035 | if (OpSize < Size) { | |||
13036 | switch (CC) { | |||
13037 | default: break; | |||
13038 | case ISD::SETULT: | |||
13039 | return generateEquivalentSub(N, Size, false, false, DL, DAG); | |||
13040 | case ISD::SETULE: | |||
13041 | return generateEquivalentSub(N, Size, true, true, DL, DAG); | |||
13042 | case ISD::SETUGT: | |||
13043 | return generateEquivalentSub(N, Size, false, true, DL, DAG); | |||
13044 | case ISD::SETUGE: | |||
13045 | return generateEquivalentSub(N, Size, true, false, DL, DAG); | |||
13046 | } | |||
13047 | } | |||
13048 | ||||
13049 | return SDValue(); | |||
13050 | } | |||
13051 | ||||
13052 | SDValue PPCTargetLowering::DAGCombineTruncBoolExt(SDNode *N, | |||
13053 | DAGCombinerInfo &DCI) const { | |||
13054 | SelectionDAG &DAG = DCI.DAG; | |||
13055 | SDLoc dl(N); | |||
13056 | ||||
13057 | assert(Subtarget.useCRBits() && "Expecting to be tracking CR bits")(static_cast <bool> (Subtarget.useCRBits() && "Expecting to be tracking CR bits" ) ? void (0) : __assert_fail ("Subtarget.useCRBits() && \"Expecting to be tracking CR bits\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13057, __extension__ __PRETTY_FUNCTION__)); | |||
13058 | // If we're tracking CR bits, we need to be careful that we don't have: | |||
13059 | // trunc(binary-ops(zext(x), zext(y))) | |||
13060 | // or | |||
13061 | // trunc(binary-ops(binary-ops(zext(x), zext(y)), ...) | |||
13062 | // such that we're unnecessarily moving things into GPRs when it would be | |||
13063 | // better to keep them in CR bits. | |||
13064 | ||||
13065 | // Note that trunc here can be an actual i1 trunc, or can be the effective | |||
13066 | // truncation that comes from a setcc or select_cc. | |||
13067 | if (N->getOpcode() == ISD::TRUNCATE && | |||
13068 | N->getValueType(0) != MVT::i1) | |||
13069 | return SDValue(); | |||
13070 | ||||
13071 | if (N->getOperand(0).getValueType() != MVT::i32 && | |||
13072 | N->getOperand(0).getValueType() != MVT::i64) | |||
13073 | return SDValue(); | |||
13074 | ||||
13075 | if (N->getOpcode() == ISD::SETCC || | |||
13076 | N->getOpcode() == ISD::SELECT_CC) { | |||
13077 | // If we're looking at a comparison, then we need to make sure that the | |||
13078 | // high bits (all except for the first) don't matter the result. | |||
13079 | ISD::CondCode CC = | |||
13080 | cast<CondCodeSDNode>(N->getOperand( | |||
13081 | N->getOpcode() == ISD::SETCC ? 2 : 4))->get(); | |||
13082 | unsigned OpBits = N->getOperand(0).getValueSizeInBits(); | |||
13083 | ||||
13084 | if (ISD::isSignedIntSetCC(CC)) { | |||
13085 | if (DAG.ComputeNumSignBits(N->getOperand(0)) != OpBits || | |||
13086 | DAG.ComputeNumSignBits(N->getOperand(1)) != OpBits) | |||
13087 | return SDValue(); | |||
13088 | } else if (ISD::isUnsignedIntSetCC(CC)) { | |||
13089 | if (!DAG.MaskedValueIsZero(N->getOperand(0), | |||
13090 | APInt::getHighBitsSet(OpBits, OpBits-1)) || | |||
13091 | !DAG.MaskedValueIsZero(N->getOperand(1), | |||
13092 | APInt::getHighBitsSet(OpBits, OpBits-1))) | |||
13093 | return (N->getOpcode() == ISD::SETCC ? ConvertSETCCToSubtract(N, DCI) | |||
13094 | : SDValue()); | |||
13095 | } else { | |||
13096 | // This is neither a signed nor an unsigned comparison, just make sure | |||
13097 | // that the high bits are equal. | |||
13098 | KnownBits Op1Known = DAG.computeKnownBits(N->getOperand(0)); | |||
13099 | KnownBits Op2Known = DAG.computeKnownBits(N->getOperand(1)); | |||
13100 | ||||
13101 | // We don't really care about what is known about the first bit (if | |||
13102 | // anything), so pretend that it is known zero for both to ensure they can | |||
13103 | // be compared as constants. | |||
13104 | Op1Known.Zero.setBit(0); Op1Known.One.clearBit(0); | |||
13105 | Op2Known.Zero.setBit(0); Op2Known.One.clearBit(0); | |||
13106 | ||||
13107 | if (!Op1Known.isConstant() || !Op2Known.isConstant() || | |||
13108 | Op1Known.getConstant() != Op2Known.getConstant()) | |||
13109 | return SDValue(); | |||
13110 | } | |||
13111 | } | |||
13112 | ||||
13113 | // We now know that the higher-order bits are irrelevant, we just need to | |||
13114 | // make sure that all of the intermediate operations are bit operations, and | |||
13115 | // all inputs are extensions. | |||
13116 | if (N->getOperand(0).getOpcode() != ISD::AND && | |||
13117 | N->getOperand(0).getOpcode() != ISD::OR && | |||
13118 | N->getOperand(0).getOpcode() != ISD::XOR && | |||
13119 | N->getOperand(0).getOpcode() != ISD::SELECT && | |||
13120 | N->getOperand(0).getOpcode() != ISD::SELECT_CC && | |||
13121 | N->getOperand(0).getOpcode() != ISD::TRUNCATE && | |||
13122 | N->getOperand(0).getOpcode() != ISD::SIGN_EXTEND && | |||
13123 | N->getOperand(0).getOpcode() != ISD::ZERO_EXTEND && | |||
13124 | N->getOperand(0).getOpcode() != ISD::ANY_EXTEND) | |||
13125 | return SDValue(); | |||
13126 | ||||
13127 | if ((N->getOpcode() == ISD::SETCC || N->getOpcode() == ISD::SELECT_CC) && | |||
13128 | N->getOperand(1).getOpcode() != ISD::AND && | |||
13129 | N->getOperand(1).getOpcode() != ISD::OR && | |||
13130 | N->getOperand(1).getOpcode() != ISD::XOR && | |||
13131 | N->getOperand(1).getOpcode() != ISD::SELECT && | |||
13132 | N->getOperand(1).getOpcode() != ISD::SELECT_CC && | |||
13133 | N->getOperand(1).getOpcode() != ISD::TRUNCATE && | |||
13134 | N->getOperand(1).getOpcode() != ISD::SIGN_EXTEND && | |||
13135 | N->getOperand(1).getOpcode() != ISD::ZERO_EXTEND && | |||
13136 | N->getOperand(1).getOpcode() != ISD::ANY_EXTEND) | |||
13137 | return SDValue(); | |||
13138 | ||||
13139 | SmallVector<SDValue, 4> Inputs; | |||
13140 | SmallVector<SDValue, 8> BinOps, PromOps; | |||
13141 | SmallPtrSet<SDNode *, 16> Visited; | |||
13142 | ||||
13143 | for (unsigned i = 0; i < 2; ++i) { | |||
13144 | if (((N->getOperand(i).getOpcode() == ISD::SIGN_EXTEND || | |||
13145 | N->getOperand(i).getOpcode() == ISD::ZERO_EXTEND || | |||
13146 | N->getOperand(i).getOpcode() == ISD::ANY_EXTEND) && | |||
13147 | N->getOperand(i).getOperand(0).getValueType() == MVT::i1) || | |||
13148 | isa<ConstantSDNode>(N->getOperand(i))) | |||
13149 | Inputs.push_back(N->getOperand(i)); | |||
13150 | else | |||
13151 | BinOps.push_back(N->getOperand(i)); | |||
13152 | ||||
13153 | if (N->getOpcode() == ISD::TRUNCATE) | |||
13154 | break; | |||
13155 | } | |||
13156 | ||||
13157 | // Visit all inputs, collect all binary operations (and, or, xor and | |||
13158 | // select) that are all fed by extensions. | |||
13159 | while (!BinOps.empty()) { | |||
13160 | SDValue BinOp = BinOps.pop_back_val(); | |||
13161 | ||||
13162 | if (!Visited.insert(BinOp.getNode()).second) | |||
13163 | continue; | |||
13164 | ||||
13165 | PromOps.push_back(BinOp); | |||
13166 | ||||
13167 | for (unsigned i = 0, ie = BinOp.getNumOperands(); i != ie; ++i) { | |||
13168 | // The condition of the select is not promoted. | |||
13169 | if (BinOp.getOpcode() == ISD::SELECT && i == 0) | |||
13170 | continue; | |||
13171 | if (BinOp.getOpcode() == ISD::SELECT_CC && i != 2 && i != 3) | |||
13172 | continue; | |||
13173 | ||||
13174 | if (((BinOp.getOperand(i).getOpcode() == ISD::SIGN_EXTEND || | |||
13175 | BinOp.getOperand(i).getOpcode() == ISD::ZERO_EXTEND || | |||
13176 | BinOp.getOperand(i).getOpcode() == ISD::ANY_EXTEND) && | |||
13177 | BinOp.getOperand(i).getOperand(0).getValueType() == MVT::i1) || | |||
13178 | isa<ConstantSDNode>(BinOp.getOperand(i))) { | |||
13179 | Inputs.push_back(BinOp.getOperand(i)); | |||
13180 | } else if (BinOp.getOperand(i).getOpcode() == ISD::AND || | |||
13181 | BinOp.getOperand(i).getOpcode() == ISD::OR || | |||
13182 | BinOp.getOperand(i).getOpcode() == ISD::XOR || | |||
13183 | BinOp.getOperand(i).getOpcode() == ISD::SELECT || | |||
13184 | BinOp.getOperand(i).getOpcode() == ISD::SELECT_CC || | |||
13185 | BinOp.getOperand(i).getOpcode() == ISD::TRUNCATE || | |||
13186 | BinOp.getOperand(i).getOpcode() == ISD::SIGN_EXTEND || | |||
13187 | BinOp.getOperand(i).getOpcode() == ISD::ZERO_EXTEND || | |||
13188 | BinOp.getOperand(i).getOpcode() == ISD::ANY_EXTEND) { | |||
13189 | BinOps.push_back(BinOp.getOperand(i)); | |||
13190 | } else { | |||
13191 | // We have an input that is not an extension or another binary | |||
13192 | // operation; we'll abort this transformation. | |||
13193 | return SDValue(); | |||
13194 | } | |||
13195 | } | |||
13196 | } | |||
13197 | ||||
13198 | // Make sure that this is a self-contained cluster of operations (which | |||
13199 | // is not quite the same thing as saying that everything has only one | |||
13200 | // use). | |||
13201 | for (unsigned i = 0, ie = Inputs.size(); i != ie; ++i) { | |||
13202 | if (isa<ConstantSDNode>(Inputs[i])) | |||
13203 | continue; | |||
13204 | ||||
13205 | for (SDNode::use_iterator UI = Inputs[i].getNode()->use_begin(), | |||
13206 | UE = Inputs[i].getNode()->use_end(); | |||
13207 | UI != UE; ++UI) { | |||
13208 | SDNode *User = *UI; | |||
13209 | if (User != N && !Visited.count(User)) | |||
13210 | return SDValue(); | |||
13211 | ||||
13212 | // Make sure that we're not going to promote the non-output-value | |||
13213 | // operand(s) or SELECT or SELECT_CC. | |||
13214 | // FIXME: Although we could sometimes handle this, and it does occur in | |||
13215 | // practice that one of the condition inputs to the select is also one of | |||
13216 | // the outputs, we currently can't deal with this. | |||
13217 | if (User->getOpcode() == ISD::SELECT) { | |||
13218 | if (User->getOperand(0) == Inputs[i]) | |||
13219 | return SDValue(); | |||
13220 | } else if (User->getOpcode() == ISD::SELECT_CC) { | |||
13221 | if (User->getOperand(0) == Inputs[i] || | |||
13222 | User->getOperand(1) == Inputs[i]) | |||
13223 | return SDValue(); | |||
13224 | } | |||
13225 | } | |||
13226 | } | |||
13227 | ||||
13228 | for (unsigned i = 0, ie = PromOps.size(); i != ie; ++i) { | |||
13229 | for (SDNode::use_iterator UI = PromOps[i].getNode()->use_begin(), | |||
13230 | UE = PromOps[i].getNode()->use_end(); | |||
13231 | UI != UE; ++UI) { | |||
13232 | SDNode *User = *UI; | |||
13233 | if (User != N && !Visited.count(User)) | |||
13234 | return SDValue(); | |||
13235 | ||||
13236 | // Make sure that we're not going to promote the non-output-value | |||
13237 | // operand(s) or SELECT or SELECT_CC. | |||
13238 | // FIXME: Although we could sometimes handle this, and it does occur in | |||
13239 | // practice that one of the condition inputs to the select is also one of | |||
13240 | // the outputs, we currently can't deal with this. | |||
13241 | if (User->getOpcode() == ISD::SELECT) { | |||
13242 | if (User->getOperand(0) == PromOps[i]) | |||
13243 | return SDValue(); | |||
13244 | } else if (User->getOpcode() == ISD::SELECT_CC) { | |||
13245 | if (User->getOperand(0) == PromOps[i] || | |||
13246 | User->getOperand(1) == PromOps[i]) | |||
13247 | return SDValue(); | |||
13248 | } | |||
13249 | } | |||
13250 | } | |||
13251 | ||||
13252 | // Replace all inputs with the extension operand. | |||
13253 | for (unsigned i = 0, ie = Inputs.size(); i != ie; ++i) { | |||
13254 | // Constants may have users outside the cluster of to-be-promoted nodes, | |||
13255 | // and so we need to replace those as we do the promotions. | |||
13256 | if (isa<ConstantSDNode>(Inputs[i])) | |||
13257 | continue; | |||
13258 | else | |||
13259 | DAG.ReplaceAllUsesOfValueWith(Inputs[i], Inputs[i].getOperand(0)); | |||
13260 | } | |||
13261 | ||||
13262 | std::list<HandleSDNode> PromOpHandles; | |||
13263 | for (auto &PromOp : PromOps) | |||
13264 | PromOpHandles.emplace_back(PromOp); | |||
13265 | ||||
13266 | // Replace all operations (these are all the same, but have a different | |||
13267 | // (i1) return type). DAG.getNode will validate that the types of | |||
13268 | // a binary operator match, so go through the list in reverse so that | |||
13269 | // we've likely promoted both operands first. Any intermediate truncations or | |||
13270 | // extensions disappear. | |||
13271 | while (!PromOpHandles.empty()) { | |||
13272 | SDValue PromOp = PromOpHandles.back().getValue(); | |||
13273 | PromOpHandles.pop_back(); | |||
13274 | ||||
13275 | if (PromOp.getOpcode() == ISD::TRUNCATE || | |||
13276 | PromOp.getOpcode() == ISD::SIGN_EXTEND || | |||
13277 | PromOp.getOpcode() == ISD::ZERO_EXTEND || | |||
13278 | PromOp.getOpcode() == ISD::ANY_EXTEND) { | |||
13279 | if (!isa<ConstantSDNode>(PromOp.getOperand(0)) && | |||
13280 | PromOp.getOperand(0).getValueType() != MVT::i1) { | |||
13281 | // The operand is not yet ready (see comment below). | |||
13282 | PromOpHandles.emplace_front(PromOp); | |||
13283 | continue; | |||
13284 | } | |||
13285 | ||||
13286 | SDValue RepValue = PromOp.getOperand(0); | |||
13287 | if (isa<ConstantSDNode>(RepValue)) | |||
13288 | RepValue = DAG.getNode(ISD::TRUNCATE, dl, MVT::i1, RepValue); | |||
13289 | ||||
13290 | DAG.ReplaceAllUsesOfValueWith(PromOp, RepValue); | |||
13291 | continue; | |||
13292 | } | |||
13293 | ||||
13294 | unsigned C; | |||
13295 | switch (PromOp.getOpcode()) { | |||
13296 | default: C = 0; break; | |||
13297 | case ISD::SELECT: C = 1; break; | |||
13298 | case ISD::SELECT_CC: C = 2; break; | |||
13299 | } | |||
13300 | ||||
13301 | if ((!isa<ConstantSDNode>(PromOp.getOperand(C)) && | |||
13302 | PromOp.getOperand(C).getValueType() != MVT::i1) || | |||
13303 | (!isa<ConstantSDNode>(PromOp.getOperand(C+1)) && | |||
13304 | PromOp.getOperand(C+1).getValueType() != MVT::i1)) { | |||
13305 | // The to-be-promoted operands of this node have not yet been | |||
13306 | // promoted (this should be rare because we're going through the | |||
13307 | // list backward, but if one of the operands has several users in | |||
13308 | // this cluster of to-be-promoted nodes, it is possible). | |||
13309 | PromOpHandles.emplace_front(PromOp); | |||
13310 | continue; | |||
13311 | } | |||
13312 | ||||
13313 | SmallVector<SDValue, 3> Ops(PromOp.getNode()->op_begin(), | |||
13314 | PromOp.getNode()->op_end()); | |||
13315 | ||||
13316 | // If there are any constant inputs, make sure they're replaced now. | |||
13317 | for (unsigned i = 0; i < 2; ++i) | |||
13318 | if (isa<ConstantSDNode>(Ops[C+i])) | |||
13319 | Ops[C+i] = DAG.getNode(ISD::TRUNCATE, dl, MVT::i1, Ops[C+i]); | |||
13320 | ||||
13321 | DAG.ReplaceAllUsesOfValueWith(PromOp, | |||
13322 | DAG.getNode(PromOp.getOpcode(), dl, MVT::i1, Ops)); | |||
13323 | } | |||
13324 | ||||
13325 | // Now we're left with the initial truncation itself. | |||
13326 | if (N->getOpcode() == ISD::TRUNCATE) | |||
13327 | return N->getOperand(0); | |||
13328 | ||||
13329 | // Otherwise, this is a comparison. The operands to be compared have just | |||
13330 | // changed type (to i1), but everything else is the same. | |||
13331 | return SDValue(N, 0); | |||
13332 | } | |||
13333 | ||||
13334 | SDValue PPCTargetLowering::DAGCombineExtBoolTrunc(SDNode *N, | |||
13335 | DAGCombinerInfo &DCI) const { | |||
13336 | SelectionDAG &DAG = DCI.DAG; | |||
13337 | SDLoc dl(N); | |||
13338 | ||||
13339 | // If we're tracking CR bits, we need to be careful that we don't have: | |||
13340 | // zext(binary-ops(trunc(x), trunc(y))) | |||
13341 | // or | |||
13342 | // zext(binary-ops(binary-ops(trunc(x), trunc(y)), ...) | |||
13343 | // such that we're unnecessarily moving things into CR bits that can more | |||
13344 | // efficiently stay in GPRs. Note that if we're not certain that the high | |||
13345 | // bits are set as required by the final extension, we still may need to do | |||
13346 | // some masking to get the proper behavior. | |||
13347 | ||||
13348 | // This same functionality is important on PPC64 when dealing with | |||
13349 | // 32-to-64-bit extensions; these occur often when 32-bit values are used as | |||
13350 | // the return values of functions. Because it is so similar, it is handled | |||
13351 | // here as well. | |||
13352 | ||||
13353 | if (N->getValueType(0) != MVT::i32 && | |||
13354 | N->getValueType(0) != MVT::i64) | |||
13355 | return SDValue(); | |||
13356 | ||||
13357 | if (!((N->getOperand(0).getValueType() == MVT::i1 && Subtarget.useCRBits()) || | |||
13358 | (N->getOperand(0).getValueType() == MVT::i32 && Subtarget.isPPC64()))) | |||
13359 | return SDValue(); | |||
13360 | ||||
13361 | if (N->getOperand(0).getOpcode() != ISD::AND && | |||
13362 | N->getOperand(0).getOpcode() != ISD::OR && | |||
13363 | N->getOperand(0).getOpcode() != ISD::XOR && | |||
13364 | N->getOperand(0).getOpcode() != ISD::SELECT && | |||
13365 | N->getOperand(0).getOpcode() != ISD::SELECT_CC) | |||
13366 | return SDValue(); | |||
13367 | ||||
13368 | SmallVector<SDValue, 4> Inputs; | |||
13369 | SmallVector<SDValue, 8> BinOps(1, N->getOperand(0)), PromOps; | |||
13370 | SmallPtrSet<SDNode *, 16> Visited; | |||
13371 | ||||
13372 | // Visit all inputs, collect all binary operations (and, or, xor and | |||
13373 | // select) that are all fed by truncations. | |||
13374 | while (!BinOps.empty()) { | |||
13375 | SDValue BinOp = BinOps.pop_back_val(); | |||
13376 | ||||
13377 | if (!Visited.insert(BinOp.getNode()).second) | |||
13378 | continue; | |||
13379 | ||||
13380 | PromOps.push_back(BinOp); | |||
13381 | ||||
13382 | for (unsigned i = 0, ie = BinOp.getNumOperands(); i != ie; ++i) { | |||
13383 | // The condition of the select is not promoted. | |||
13384 | if (BinOp.getOpcode() == ISD::SELECT && i == 0) | |||
13385 | continue; | |||
13386 | if (BinOp.getOpcode() == ISD::SELECT_CC && i != 2 && i != 3) | |||
13387 | continue; | |||
13388 | ||||
13389 | if (BinOp.getOperand(i).getOpcode() == ISD::TRUNCATE || | |||
13390 | isa<ConstantSDNode>(BinOp.getOperand(i))) { | |||
13391 | Inputs.push_back(BinOp.getOperand(i)); | |||
13392 | } else if (BinOp.getOperand(i).getOpcode() == ISD::AND || | |||
13393 | BinOp.getOperand(i).getOpcode() == ISD::OR || | |||
13394 | BinOp.getOperand(i).getOpcode() == ISD::XOR || | |||
13395 | BinOp.getOperand(i).getOpcode() == ISD::SELECT || | |||
13396 | BinOp.getOperand(i).getOpcode() == ISD::SELECT_CC) { | |||
13397 | BinOps.push_back(BinOp.getOperand(i)); | |||
13398 | } else { | |||
13399 | // We have an input that is not a truncation or another binary | |||
13400 | // operation; we'll abort this transformation. | |||
13401 | return SDValue(); | |||
13402 | } | |||
13403 | } | |||
13404 | } | |||
13405 | ||||
13406 | // The operands of a select that must be truncated when the select is | |||
13407 | // promoted because the operand is actually part of the to-be-promoted set. | |||
13408 | DenseMap<SDNode *, EVT> SelectTruncOp[2]; | |||
13409 | ||||
13410 | // Make sure that this is a self-contained cluster of operations (which | |||
13411 | // is not quite the same thing as saying that everything has only one | |||
13412 | // use). | |||
13413 | for (unsigned i = 0, ie = Inputs.size(); i != ie; ++i) { | |||
13414 | if (isa<ConstantSDNode>(Inputs[i])) | |||
13415 | continue; | |||
13416 | ||||
13417 | for (SDNode::use_iterator UI = Inputs[i].getNode()->use_begin(), | |||
13418 | UE = Inputs[i].getNode()->use_end(); | |||
13419 | UI != UE; ++UI) { | |||
13420 | SDNode *User = *UI; | |||
13421 | if (User != N && !Visited.count(User)) | |||
13422 | return SDValue(); | |||
13423 | ||||
13424 | // If we're going to promote the non-output-value operand(s) or SELECT or | |||
13425 | // SELECT_CC, record them for truncation. | |||
13426 | if (User->getOpcode() == ISD::SELECT) { | |||
13427 | if (User->getOperand(0) == Inputs[i]) | |||
13428 | SelectTruncOp[0].insert(std::make_pair(User, | |||
13429 | User->getOperand(0).getValueType())); | |||
13430 | } else if (User->getOpcode() == ISD::SELECT_CC) { | |||
13431 | if (User->getOperand(0) == Inputs[i]) | |||
13432 | SelectTruncOp[0].insert(std::make_pair(User, | |||
13433 | User->getOperand(0).getValueType())); | |||
13434 | if (User->getOperand(1) == Inputs[i]) | |||
13435 | SelectTruncOp[1].insert(std::make_pair(User, | |||
13436 | User->getOperand(1).getValueType())); | |||
13437 | } | |||
13438 | } | |||
13439 | } | |||
13440 | ||||
13441 | for (unsigned i = 0, ie = PromOps.size(); i != ie; ++i) { | |||
13442 | for (SDNode::use_iterator UI = PromOps[i].getNode()->use_begin(), | |||
13443 | UE = PromOps[i].getNode()->use_end(); | |||
13444 | UI != UE; ++UI) { | |||
13445 | SDNode *User = *UI; | |||
13446 | if (User != N && !Visited.count(User)) | |||
13447 | return SDValue(); | |||
13448 | ||||
13449 | // If we're going to promote the non-output-value operand(s) or SELECT or | |||
13450 | // SELECT_CC, record them for truncation. | |||
13451 | if (User->getOpcode() == ISD::SELECT) { | |||
13452 | if (User->getOperand(0) == PromOps[i]) | |||
13453 | SelectTruncOp[0].insert(std::make_pair(User, | |||
13454 | User->getOperand(0).getValueType())); | |||
13455 | } else if (User->getOpcode() == ISD::SELECT_CC) { | |||
13456 | if (User->getOperand(0) == PromOps[i]) | |||
13457 | SelectTruncOp[0].insert(std::make_pair(User, | |||
13458 | User->getOperand(0).getValueType())); | |||
13459 | if (User->getOperand(1) == PromOps[i]) | |||
13460 | SelectTruncOp[1].insert(std::make_pair(User, | |||
13461 | User->getOperand(1).getValueType())); | |||
13462 | } | |||
13463 | } | |||
13464 | } | |||
13465 | ||||
13466 | unsigned PromBits = N->getOperand(0).getValueSizeInBits(); | |||
13467 | bool ReallyNeedsExt = false; | |||
13468 | if (N->getOpcode() != ISD::ANY_EXTEND) { | |||
13469 | // If all of the inputs are not already sign/zero extended, then | |||
13470 | // we'll still need to do that at the end. | |||
13471 | for (unsigned i = 0, ie = Inputs.size(); i != ie; ++i) { | |||
13472 | if (isa<ConstantSDNode>(Inputs[i])) | |||
13473 | continue; | |||
13474 | ||||
13475 | unsigned OpBits = | |||
13476 | Inputs[i].getOperand(0).getValueSizeInBits(); | |||
13477 | assert(PromBits < OpBits && "Truncation not to a smaller bit count?")(static_cast <bool> (PromBits < OpBits && "Truncation not to a smaller bit count?" ) ? void (0) : __assert_fail ("PromBits < OpBits && \"Truncation not to a smaller bit count?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13477, __extension__ __PRETTY_FUNCTION__)); | |||
13478 | ||||
13479 | if ((N->getOpcode() == ISD::ZERO_EXTEND && | |||
13480 | !DAG.MaskedValueIsZero(Inputs[i].getOperand(0), | |||
13481 | APInt::getHighBitsSet(OpBits, | |||
13482 | OpBits-PromBits))) || | |||
13483 | (N->getOpcode() == ISD::SIGN_EXTEND && | |||
13484 | DAG.ComputeNumSignBits(Inputs[i].getOperand(0)) < | |||
13485 | (OpBits-(PromBits-1)))) { | |||
13486 | ReallyNeedsExt = true; | |||
13487 | break; | |||
13488 | } | |||
13489 | } | |||
13490 | } | |||
13491 | ||||
13492 | // Replace all inputs, either with the truncation operand, or a | |||
13493 | // truncation or extension to the final output type. | |||
13494 | for (unsigned i = 0, ie = Inputs.size(); i != ie; ++i) { | |||
13495 | // Constant inputs need to be replaced with the to-be-promoted nodes that | |||
13496 | // use them because they might have users outside of the cluster of | |||
13497 | // promoted nodes. | |||
13498 | if (isa<ConstantSDNode>(Inputs[i])) | |||
13499 | continue; | |||
13500 | ||||
13501 | SDValue InSrc = Inputs[i].getOperand(0); | |||
13502 | if (Inputs[i].getValueType() == N->getValueType(0)) | |||
13503 | DAG.ReplaceAllUsesOfValueWith(Inputs[i], InSrc); | |||
13504 | else if (N->getOpcode() == ISD::SIGN_EXTEND) | |||
13505 | DAG.ReplaceAllUsesOfValueWith(Inputs[i], | |||
13506 | DAG.getSExtOrTrunc(InSrc, dl, N->getValueType(0))); | |||
13507 | else if (N->getOpcode() == ISD::ZERO_EXTEND) | |||
13508 | DAG.ReplaceAllUsesOfValueWith(Inputs[i], | |||
13509 | DAG.getZExtOrTrunc(InSrc, dl, N->getValueType(0))); | |||
13510 | else | |||
13511 | DAG.ReplaceAllUsesOfValueWith(Inputs[i], | |||
13512 | DAG.getAnyExtOrTrunc(InSrc, dl, N->getValueType(0))); | |||
13513 | } | |||
13514 | ||||
13515 | std::list<HandleSDNode> PromOpHandles; | |||
13516 | for (auto &PromOp : PromOps) | |||
13517 | PromOpHandles.emplace_back(PromOp); | |||
13518 | ||||
13519 | // Replace all operations (these are all the same, but have a different | |||
13520 | // (promoted) return type). DAG.getNode will validate that the types of | |||
13521 | // a binary operator match, so go through the list in reverse so that | |||
13522 | // we've likely promoted both operands first. | |||
13523 | while (!PromOpHandles.empty()) { | |||
13524 | SDValue PromOp = PromOpHandles.back().getValue(); | |||
13525 | PromOpHandles.pop_back(); | |||
13526 | ||||
13527 | unsigned C; | |||
13528 | switch (PromOp.getOpcode()) { | |||
13529 | default: C = 0; break; | |||
13530 | case ISD::SELECT: C = 1; break; | |||
13531 | case ISD::SELECT_CC: C = 2; break; | |||
13532 | } | |||
13533 | ||||
13534 | if ((!isa<ConstantSDNode>(PromOp.getOperand(C)) && | |||
13535 | PromOp.getOperand(C).getValueType() != N->getValueType(0)) || | |||
13536 | (!isa<ConstantSDNode>(PromOp.getOperand(C+1)) && | |||
13537 | PromOp.getOperand(C+1).getValueType() != N->getValueType(0))) { | |||
13538 | // The to-be-promoted operands of this node have not yet been | |||
13539 | // promoted (this should be rare because we're going through the | |||
13540 | // list backward, but if one of the operands has several users in | |||
13541 | // this cluster of to-be-promoted nodes, it is possible). | |||
13542 | PromOpHandles.emplace_front(PromOp); | |||
13543 | continue; | |||
13544 | } | |||
13545 | ||||
13546 | // For SELECT and SELECT_CC nodes, we do a similar check for any | |||
13547 | // to-be-promoted comparison inputs. | |||
13548 | if (PromOp.getOpcode() == ISD::SELECT || | |||
13549 | PromOp.getOpcode() == ISD::SELECT_CC) { | |||
13550 | if ((SelectTruncOp[0].count(PromOp.getNode()) && | |||
13551 | PromOp.getOperand(0).getValueType() != N->getValueType(0)) || | |||
13552 | (SelectTruncOp[1].count(PromOp.getNode()) && | |||
13553 | PromOp.getOperand(1).getValueType() != N->getValueType(0))) { | |||
13554 | PromOpHandles.emplace_front(PromOp); | |||
13555 | continue; | |||
13556 | } | |||
13557 | } | |||
13558 | ||||
13559 | SmallVector<SDValue, 3> Ops(PromOp.getNode()->op_begin(), | |||
13560 | PromOp.getNode()->op_end()); | |||
13561 | ||||
13562 | // If this node has constant inputs, then they'll need to be promoted here. | |||
13563 | for (unsigned i = 0; i < 2; ++i) { | |||
13564 | if (!isa<ConstantSDNode>(Ops[C+i])) | |||
13565 | continue; | |||
13566 | if (Ops[C+i].getValueType() == N->getValueType(0)) | |||
13567 | continue; | |||
13568 | ||||
13569 | if (N->getOpcode() == ISD::SIGN_EXTEND) | |||
13570 | Ops[C+i] = DAG.getSExtOrTrunc(Ops[C+i], dl, N->getValueType(0)); | |||
13571 | else if (N->getOpcode() == ISD::ZERO_EXTEND) | |||
13572 | Ops[C+i] = DAG.getZExtOrTrunc(Ops[C+i], dl, N->getValueType(0)); | |||
13573 | else | |||
13574 | Ops[C+i] = DAG.getAnyExtOrTrunc(Ops[C+i], dl, N->getValueType(0)); | |||
13575 | } | |||
13576 | ||||
13577 | // If we've promoted the comparison inputs of a SELECT or SELECT_CC, | |||
13578 | // truncate them again to the original value type. | |||
13579 | if (PromOp.getOpcode() == ISD::SELECT || | |||
13580 | PromOp.getOpcode() == ISD::SELECT_CC) { | |||
13581 | auto SI0 = SelectTruncOp[0].find(PromOp.getNode()); | |||
13582 | if (SI0 != SelectTruncOp[0].end()) | |||
13583 | Ops[0] = DAG.getNode(ISD::TRUNCATE, dl, SI0->second, Ops[0]); | |||
13584 | auto SI1 = SelectTruncOp[1].find(PromOp.getNode()); | |||
13585 | if (SI1 != SelectTruncOp[1].end()) | |||
13586 | Ops[1] = DAG.getNode(ISD::TRUNCATE, dl, SI1->second, Ops[1]); | |||
13587 | } | |||
13588 | ||||
13589 | DAG.ReplaceAllUsesOfValueWith(PromOp, | |||
13590 | DAG.getNode(PromOp.getOpcode(), dl, N->getValueType(0), Ops)); | |||
13591 | } | |||
13592 | ||||
13593 | // Now we're left with the initial extension itself. | |||
13594 | if (!ReallyNeedsExt) | |||
13595 | return N->getOperand(0); | |||
13596 | ||||
13597 | // To zero extend, just mask off everything except for the first bit (in the | |||
13598 | // i1 case). | |||
13599 | if (N->getOpcode() == ISD::ZERO_EXTEND) | |||
13600 | return DAG.getNode(ISD::AND, dl, N->getValueType(0), N->getOperand(0), | |||
13601 | DAG.getConstant(APInt::getLowBitsSet( | |||
13602 | N->getValueSizeInBits(0), PromBits), | |||
13603 | dl, N->getValueType(0))); | |||
13604 | ||||
13605 | assert(N->getOpcode() == ISD::SIGN_EXTEND &&(static_cast <bool> (N->getOpcode() == ISD::SIGN_EXTEND && "Invalid extension type") ? void (0) : __assert_fail ("N->getOpcode() == ISD::SIGN_EXTEND && \"Invalid extension type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13606, __extension__ __PRETTY_FUNCTION__)) | |||
13606 | "Invalid extension type")(static_cast <bool> (N->getOpcode() == ISD::SIGN_EXTEND && "Invalid extension type") ? void (0) : __assert_fail ("N->getOpcode() == ISD::SIGN_EXTEND && \"Invalid extension type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13606, __extension__ __PRETTY_FUNCTION__)); | |||
13607 | EVT ShiftAmountTy = getShiftAmountTy(N->getValueType(0), DAG.getDataLayout()); | |||
13608 | SDValue ShiftCst = | |||
13609 | DAG.getConstant(N->getValueSizeInBits(0) - PromBits, dl, ShiftAmountTy); | |||
13610 | return DAG.getNode( | |||
13611 | ISD::SRA, dl, N->getValueType(0), | |||
13612 | DAG.getNode(ISD::SHL, dl, N->getValueType(0), N->getOperand(0), ShiftCst), | |||
13613 | ShiftCst); | |||
13614 | } | |||
13615 | ||||
13616 | SDValue PPCTargetLowering::combineSetCC(SDNode *N, | |||
13617 | DAGCombinerInfo &DCI) const { | |||
13618 | assert(N->getOpcode() == ISD::SETCC &&(static_cast <bool> (N->getOpcode() == ISD::SETCC && "Should be called with a SETCC node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::SETCC && \"Should be called with a SETCC node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13619, __extension__ __PRETTY_FUNCTION__)) | |||
13619 | "Should be called with a SETCC node")(static_cast <bool> (N->getOpcode() == ISD::SETCC && "Should be called with a SETCC node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::SETCC && \"Should be called with a SETCC node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13619, __extension__ __PRETTY_FUNCTION__)); | |||
13620 | ||||
13621 | ISD::CondCode CC = cast<CondCodeSDNode>(N->getOperand(2))->get(); | |||
13622 | if (CC == ISD::SETNE || CC == ISD::SETEQ) { | |||
13623 | SDValue LHS = N->getOperand(0); | |||
13624 | SDValue RHS = N->getOperand(1); | |||
13625 | ||||
13626 | // If there is a '0 - y' pattern, canonicalize the pattern to the RHS. | |||
13627 | if (LHS.getOpcode() == ISD::SUB && isNullConstant(LHS.getOperand(0)) && | |||
13628 | LHS.hasOneUse()) | |||
13629 | std::swap(LHS, RHS); | |||
13630 | ||||
13631 | // x == 0-y --> x+y == 0 | |||
13632 | // x != 0-y --> x+y != 0 | |||
13633 | if (RHS.getOpcode() == ISD::SUB && isNullConstant(RHS.getOperand(0)) && | |||
13634 | RHS.hasOneUse()) { | |||
13635 | SDLoc DL(N); | |||
13636 | SelectionDAG &DAG = DCI.DAG; | |||
13637 | EVT VT = N->getValueType(0); | |||
13638 | EVT OpVT = LHS.getValueType(); | |||
13639 | SDValue Add = DAG.getNode(ISD::ADD, DL, OpVT, LHS, RHS.getOperand(1)); | |||
13640 | return DAG.getSetCC(DL, VT, Add, DAG.getConstant(0, DL, OpVT), CC); | |||
13641 | } | |||
13642 | } | |||
13643 | ||||
13644 | return DAGCombineTruncBoolExt(N, DCI); | |||
13645 | } | |||
13646 | ||||
13647 | // Is this an extending load from an f32 to an f64? | |||
13648 | static bool isFPExtLoad(SDValue Op) { | |||
13649 | if (LoadSDNode *LD = dyn_cast<LoadSDNode>(Op.getNode())) | |||
13650 | return LD->getExtensionType() == ISD::EXTLOAD && | |||
13651 | Op.getValueType() == MVT::f64; | |||
13652 | return false; | |||
13653 | } | |||
13654 | ||||
13655 | /// Reduces the number of fp-to-int conversion when building a vector. | |||
13656 | /// | |||
13657 | /// If this vector is built out of floating to integer conversions, | |||
13658 | /// transform it to a vector built out of floating point values followed by a | |||
13659 | /// single floating to integer conversion of the vector. | |||
13660 | /// Namely (build_vector (fptosi $A), (fptosi $B), ...) | |||
13661 | /// becomes (fptosi (build_vector ($A, $B, ...))) | |||
13662 | SDValue PPCTargetLowering:: | |||
13663 | combineElementTruncationToVectorTruncation(SDNode *N, | |||
13664 | DAGCombinerInfo &DCI) const { | |||
13665 | assert(N->getOpcode() == ISD::BUILD_VECTOR &&(static_cast <bool> (N->getOpcode() == ISD::BUILD_VECTOR && "Should be called with a BUILD_VECTOR node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::BUILD_VECTOR && \"Should be called with a BUILD_VECTOR node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13666, __extension__ __PRETTY_FUNCTION__)) | |||
13666 | "Should be called with a BUILD_VECTOR node")(static_cast <bool> (N->getOpcode() == ISD::BUILD_VECTOR && "Should be called with a BUILD_VECTOR node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::BUILD_VECTOR && \"Should be called with a BUILD_VECTOR node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13666, __extension__ __PRETTY_FUNCTION__)); | |||
13667 | ||||
13668 | SelectionDAG &DAG = DCI.DAG; | |||
13669 | SDLoc dl(N); | |||
13670 | ||||
13671 | SDValue FirstInput = N->getOperand(0); | |||
13672 | assert(FirstInput.getOpcode() == PPCISD::MFVSR &&(static_cast <bool> (FirstInput.getOpcode() == PPCISD:: MFVSR && "The input operand must be an fp-to-int conversion." ) ? void (0) : __assert_fail ("FirstInput.getOpcode() == PPCISD::MFVSR && \"The input operand must be an fp-to-int conversion.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13673, __extension__ __PRETTY_FUNCTION__)) | |||
13673 | "The input operand must be an fp-to-int conversion.")(static_cast <bool> (FirstInput.getOpcode() == PPCISD:: MFVSR && "The input operand must be an fp-to-int conversion." ) ? void (0) : __assert_fail ("FirstInput.getOpcode() == PPCISD::MFVSR && \"The input operand must be an fp-to-int conversion.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13673, __extension__ __PRETTY_FUNCTION__)); | |||
13674 | ||||
13675 | // This combine happens after legalization so the fp_to_[su]i nodes are | |||
13676 | // already converted to PPCSISD nodes. | |||
13677 | unsigned FirstConversion = FirstInput.getOperand(0).getOpcode(); | |||
13678 | if (FirstConversion == PPCISD::FCTIDZ || | |||
13679 | FirstConversion == PPCISD::FCTIDUZ || | |||
13680 | FirstConversion == PPCISD::FCTIWZ || | |||
13681 | FirstConversion == PPCISD::FCTIWUZ) { | |||
13682 | bool IsSplat = true; | |||
13683 | bool Is32Bit = FirstConversion == PPCISD::FCTIWZ || | |||
13684 | FirstConversion == PPCISD::FCTIWUZ; | |||
13685 | EVT SrcVT = FirstInput.getOperand(0).getValueType(); | |||
13686 | SmallVector<SDValue, 4> Ops; | |||
13687 | EVT TargetVT = N->getValueType(0); | |||
13688 | for (int i = 0, e = N->getNumOperands(); i < e; ++i) { | |||
13689 | SDValue NextOp = N->getOperand(i); | |||
13690 | if (NextOp.getOpcode() != PPCISD::MFVSR) | |||
13691 | return SDValue(); | |||
13692 | unsigned NextConversion = NextOp.getOperand(0).getOpcode(); | |||
13693 | if (NextConversion != FirstConversion) | |||
13694 | return SDValue(); | |||
13695 | // If we are converting to 32-bit integers, we need to add an FP_ROUND. | |||
13696 | // This is not valid if the input was originally double precision. It is | |||
13697 | // also not profitable to do unless this is an extending load in which | |||
13698 | // case doing this combine will allow us to combine consecutive loads. | |||
13699 | if (Is32Bit && !isFPExtLoad(NextOp.getOperand(0).getOperand(0))) | |||
13700 | return SDValue(); | |||
13701 | if (N->getOperand(i) != FirstInput) | |||
13702 | IsSplat = false; | |||
13703 | } | |||
13704 | ||||
13705 | // If this is a splat, we leave it as-is since there will be only a single | |||
13706 | // fp-to-int conversion followed by a splat of the integer. This is better | |||
13707 | // for 32-bit and smaller ints and neutral for 64-bit ints. | |||
13708 | if (IsSplat) | |||
13709 | return SDValue(); | |||
13710 | ||||
13711 | // Now that we know we have the right type of node, get its operands | |||
13712 | for (int i = 0, e = N->getNumOperands(); i < e; ++i) { | |||
13713 | SDValue In = N->getOperand(i).getOperand(0); | |||
13714 | if (Is32Bit) { | |||
13715 | // For 32-bit values, we need to add an FP_ROUND node (if we made it | |||
13716 | // here, we know that all inputs are extending loads so this is safe). | |||
13717 | if (In.isUndef()) | |||
13718 | Ops.push_back(DAG.getUNDEF(SrcVT)); | |||
13719 | else { | |||
13720 | SDValue Trunc = DAG.getNode(ISD::FP_ROUND, dl, | |||
13721 | MVT::f32, In.getOperand(0), | |||
13722 | DAG.getIntPtrConstant(1, dl)); | |||
13723 | Ops.push_back(Trunc); | |||
13724 | } | |||
13725 | } else | |||
13726 | Ops.push_back(In.isUndef() ? DAG.getUNDEF(SrcVT) : In.getOperand(0)); | |||
13727 | } | |||
13728 | ||||
13729 | unsigned Opcode; | |||
13730 | if (FirstConversion == PPCISD::FCTIDZ || | |||
13731 | FirstConversion == PPCISD::FCTIWZ) | |||
13732 | Opcode = ISD::FP_TO_SINT; | |||
13733 | else | |||
13734 | Opcode = ISD::FP_TO_UINT; | |||
13735 | ||||
13736 | EVT NewVT = TargetVT == MVT::v2i64 ? MVT::v2f64 : MVT::v4f32; | |||
13737 | SDValue BV = DAG.getBuildVector(NewVT, dl, Ops); | |||
13738 | return DAG.getNode(Opcode, dl, TargetVT, BV); | |||
13739 | } | |||
13740 | return SDValue(); | |||
13741 | } | |||
13742 | ||||
13743 | /// Reduce the number of loads when building a vector. | |||
13744 | /// | |||
13745 | /// Building a vector out of multiple loads can be converted to a load | |||
13746 | /// of the vector type if the loads are consecutive. If the loads are | |||
13747 | /// consecutive but in descending order, a shuffle is added at the end | |||
13748 | /// to reorder the vector. | |||
13749 | static SDValue combineBVOfConsecutiveLoads(SDNode *N, SelectionDAG &DAG) { | |||
13750 | assert(N->getOpcode() == ISD::BUILD_VECTOR &&(static_cast <bool> (N->getOpcode() == ISD::BUILD_VECTOR && "Should be called with a BUILD_VECTOR node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::BUILD_VECTOR && \"Should be called with a BUILD_VECTOR node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13751, __extension__ __PRETTY_FUNCTION__)) | |||
13751 | "Should be called with a BUILD_VECTOR node")(static_cast <bool> (N->getOpcode() == ISD::BUILD_VECTOR && "Should be called with a BUILD_VECTOR node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::BUILD_VECTOR && \"Should be called with a BUILD_VECTOR node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13751, __extension__ __PRETTY_FUNCTION__)); | |||
13752 | ||||
13753 | SDLoc dl(N); | |||
13754 | ||||
13755 | // Return early for non byte-sized type, as they can't be consecutive. | |||
13756 | if (!N->getValueType(0).getVectorElementType().isByteSized()) | |||
13757 | return SDValue(); | |||
13758 | ||||
13759 | bool InputsAreConsecutiveLoads = true; | |||
13760 | bool InputsAreReverseConsecutive = true; | |||
13761 | unsigned ElemSize = N->getValueType(0).getScalarType().getStoreSize(); | |||
13762 | SDValue FirstInput = N->getOperand(0); | |||
13763 | bool IsRoundOfExtLoad = false; | |||
13764 | ||||
13765 | if (FirstInput.getOpcode() == ISD::FP_ROUND && | |||
13766 | FirstInput.getOperand(0).getOpcode() == ISD::LOAD) { | |||
13767 | LoadSDNode *LD = dyn_cast<LoadSDNode>(FirstInput.getOperand(0)); | |||
13768 | IsRoundOfExtLoad = LD->getExtensionType() == ISD::EXTLOAD; | |||
13769 | } | |||
13770 | // Not a build vector of (possibly fp_rounded) loads. | |||
13771 | if ((!IsRoundOfExtLoad && FirstInput.getOpcode() != ISD::LOAD) || | |||
13772 | N->getNumOperands() == 1) | |||
13773 | return SDValue(); | |||
13774 | ||||
13775 | for (int i = 1, e = N->getNumOperands(); i < e; ++i) { | |||
13776 | // If any inputs are fp_round(extload), they all must be. | |||
13777 | if (IsRoundOfExtLoad && N->getOperand(i).getOpcode() != ISD::FP_ROUND) | |||
13778 | return SDValue(); | |||
13779 | ||||
13780 | SDValue NextInput = IsRoundOfExtLoad ? N->getOperand(i).getOperand(0) : | |||
13781 | N->getOperand(i); | |||
13782 | if (NextInput.getOpcode() != ISD::LOAD) | |||
13783 | return SDValue(); | |||
13784 | ||||
13785 | SDValue PreviousInput = | |||
13786 | IsRoundOfExtLoad ? N->getOperand(i-1).getOperand(0) : N->getOperand(i-1); | |||
13787 | LoadSDNode *LD1 = dyn_cast<LoadSDNode>(PreviousInput); | |||
13788 | LoadSDNode *LD2 = dyn_cast<LoadSDNode>(NextInput); | |||
13789 | ||||
13790 | // If any inputs are fp_round(extload), they all must be. | |||
13791 | if (IsRoundOfExtLoad && LD2->getExtensionType() != ISD::EXTLOAD) | |||
13792 | return SDValue(); | |||
13793 | ||||
13794 | if (!isConsecutiveLS(LD2, LD1, ElemSize, 1, DAG)) | |||
13795 | InputsAreConsecutiveLoads = false; | |||
13796 | if (!isConsecutiveLS(LD1, LD2, ElemSize, 1, DAG)) | |||
13797 | InputsAreReverseConsecutive = false; | |||
13798 | ||||
13799 | // Exit early if the loads are neither consecutive nor reverse consecutive. | |||
13800 | if (!InputsAreConsecutiveLoads && !InputsAreReverseConsecutive) | |||
13801 | return SDValue(); | |||
13802 | } | |||
13803 | ||||
13804 | assert(!(InputsAreConsecutiveLoads && InputsAreReverseConsecutive) &&(static_cast <bool> (!(InputsAreConsecutiveLoads && InputsAreReverseConsecutive) && "The loads cannot be both consecutive and reverse consecutive." ) ? void (0) : __assert_fail ("!(InputsAreConsecutiveLoads && InputsAreReverseConsecutive) && \"The loads cannot be both consecutive and reverse consecutive.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13805, __extension__ __PRETTY_FUNCTION__)) | |||
13805 | "The loads cannot be both consecutive and reverse consecutive.")(static_cast <bool> (!(InputsAreConsecutiveLoads && InputsAreReverseConsecutive) && "The loads cannot be both consecutive and reverse consecutive." ) ? void (0) : __assert_fail ("!(InputsAreConsecutiveLoads && InputsAreReverseConsecutive) && \"The loads cannot be both consecutive and reverse consecutive.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13805, __extension__ __PRETTY_FUNCTION__)); | |||
13806 | ||||
13807 | SDValue FirstLoadOp = | |||
13808 | IsRoundOfExtLoad ? FirstInput.getOperand(0) : FirstInput; | |||
13809 | SDValue LastLoadOp = | |||
13810 | IsRoundOfExtLoad ? N->getOperand(N->getNumOperands()-1).getOperand(0) : | |||
13811 | N->getOperand(N->getNumOperands()-1); | |||
13812 | ||||
13813 | LoadSDNode *LD1 = dyn_cast<LoadSDNode>(FirstLoadOp); | |||
13814 | LoadSDNode *LDL = dyn_cast<LoadSDNode>(LastLoadOp); | |||
13815 | if (InputsAreConsecutiveLoads) { | |||
13816 | assert(LD1 && "Input needs to be a LoadSDNode.")(static_cast <bool> (LD1 && "Input needs to be a LoadSDNode." ) ? void (0) : __assert_fail ("LD1 && \"Input needs to be a LoadSDNode.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13816, __extension__ __PRETTY_FUNCTION__)); | |||
13817 | return DAG.getLoad(N->getValueType(0), dl, LD1->getChain(), | |||
13818 | LD1->getBasePtr(), LD1->getPointerInfo(), | |||
13819 | LD1->getAlignment()); | |||
13820 | } | |||
13821 | if (InputsAreReverseConsecutive) { | |||
13822 | assert(LDL && "Input needs to be a LoadSDNode.")(static_cast <bool> (LDL && "Input needs to be a LoadSDNode." ) ? void (0) : __assert_fail ("LDL && \"Input needs to be a LoadSDNode.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 13822, __extension__ __PRETTY_FUNCTION__)); | |||
13823 | SDValue Load = DAG.getLoad(N->getValueType(0), dl, LDL->getChain(), | |||
13824 | LDL->getBasePtr(), LDL->getPointerInfo(), | |||
13825 | LDL->getAlignment()); | |||
13826 | SmallVector<int, 16> Ops; | |||
13827 | for (int i = N->getNumOperands() - 1; i >= 0; i--) | |||
13828 | Ops.push_back(i); | |||
13829 | ||||
13830 | return DAG.getVectorShuffle(N->getValueType(0), dl, Load, | |||
13831 | DAG.getUNDEF(N->getValueType(0)), Ops); | |||
13832 | } | |||
13833 | return SDValue(); | |||
13834 | } | |||
13835 | ||||
13836 | // This function adds the required vector_shuffle needed to get | |||
13837 | // the elements of the vector extract in the correct position | |||
13838 | // as specified by the CorrectElems encoding. | |||
13839 | static SDValue addShuffleForVecExtend(SDNode *N, SelectionDAG &DAG, | |||
13840 | SDValue Input, uint64_t Elems, | |||
13841 | uint64_t CorrectElems) { | |||
13842 | SDLoc dl(N); | |||
13843 | ||||
13844 | unsigned NumElems = Input.getValueType().getVectorNumElements(); | |||
13845 | SmallVector<int, 16> ShuffleMask(NumElems, -1); | |||
13846 | ||||
13847 | // Knowing the element indices being extracted from the original | |||
13848 | // vector and the order in which they're being inserted, just put | |||
13849 | // them at element indices required for the instruction. | |||
13850 | for (unsigned i = 0; i < N->getNumOperands(); i++) { | |||
13851 | if (DAG.getDataLayout().isLittleEndian()) | |||
13852 | ShuffleMask[CorrectElems & 0xF] = Elems & 0xF; | |||
13853 | else | |||
13854 | ShuffleMask[(CorrectElems & 0xF0) >> 4] = (Elems & 0xF0) >> 4; | |||
13855 | CorrectElems = CorrectElems >> 8; | |||
13856 | Elems = Elems >> 8; | |||
13857 | } | |||
13858 | ||||
13859 | SDValue Shuffle = | |||
13860 | DAG.getVectorShuffle(Input.getValueType(), dl, Input, | |||
13861 | DAG.getUNDEF(Input.getValueType()), ShuffleMask); | |||
13862 | ||||
13863 | EVT VT = N->getValueType(0); | |||
13864 | SDValue Conv = DAG.getBitcast(VT, Shuffle); | |||
13865 | ||||
13866 | EVT ExtVT = EVT::getVectorVT(*DAG.getContext(), | |||
13867 | Input.getValueType().getVectorElementType(), | |||
13868 | VT.getVectorNumElements()); | |||
13869 | return DAG.getNode(ISD::SIGN_EXTEND_INREG, dl, VT, Conv, | |||
13870 | DAG.getValueType(ExtVT)); | |||
13871 | } | |||
13872 | ||||
13873 | // Look for build vector patterns where input operands come from sign | |||
13874 | // extended vector_extract elements of specific indices. If the correct indices | |||
13875 | // aren't used, add a vector shuffle to fix up the indices and create | |||
13876 | // SIGN_EXTEND_INREG node which selects the vector sign extend instructions | |||
13877 | // during instruction selection. | |||
13878 | static SDValue combineBVOfVecSExt(SDNode *N, SelectionDAG &DAG) { | |||
13879 | // This array encodes the indices that the vector sign extend instructions | |||
13880 | // extract from when extending from one type to another for both BE and LE. | |||
13881 | // The right nibble of each byte corresponds to the LE incides. | |||
13882 | // and the left nibble of each byte corresponds to the BE incides. | |||
13883 | // For example: 0x3074B8FC byte->word | |||
13884 | // For LE: the allowed indices are: 0x0,0x4,0x8,0xC | |||
13885 | // For BE: the allowed indices are: 0x3,0x7,0xB,0xF | |||
13886 | // For example: 0x000070F8 byte->double word | |||
13887 | // For LE: the allowed indices are: 0x0,0x8 | |||
13888 | // For BE: the allowed indices are: 0x7,0xF | |||
13889 | uint64_t TargetElems[] = { | |||
13890 | 0x3074B8FC, // b->w | |||
13891 | 0x000070F8, // b->d | |||
13892 | 0x10325476, // h->w | |||
13893 | 0x00003074, // h->d | |||
13894 | 0x00001032, // w->d | |||
13895 | }; | |||
13896 | ||||
13897 | uint64_t Elems = 0; | |||
13898 | int Index; | |||
13899 | SDValue Input; | |||
13900 | ||||
13901 | auto isSExtOfVecExtract = [&](SDValue Op) -> bool { | |||
13902 | if (!Op) | |||
13903 | return false; | |||
13904 | if (Op.getOpcode() != ISD::SIGN_EXTEND && | |||
13905 | Op.getOpcode() != ISD::SIGN_EXTEND_INREG) | |||
13906 | return false; | |||
13907 | ||||
13908 | // A SIGN_EXTEND_INREG might be fed by an ANY_EXTEND to produce a value | |||
13909 | // of the right width. | |||
13910 | SDValue Extract = Op.getOperand(0); | |||
13911 | if (Extract.getOpcode() == ISD::ANY_EXTEND) | |||
13912 | Extract = Extract.getOperand(0); | |||
13913 | if (Extract.getOpcode() != ISD::EXTRACT_VECTOR_ELT) | |||
13914 | return false; | |||
13915 | ||||
13916 | ConstantSDNode *ExtOp = dyn_cast<ConstantSDNode>(Extract.getOperand(1)); | |||
13917 | if (!ExtOp) | |||
13918 | return false; | |||
13919 | ||||
13920 | Index = ExtOp->getZExtValue(); | |||
13921 | if (Input && Input != Extract.getOperand(0)) | |||
13922 | return false; | |||
13923 | ||||
13924 | if (!Input) | |||
13925 | Input = Extract.getOperand(0); | |||
13926 | ||||
13927 | Elems = Elems << 8; | |||
13928 | Index = DAG.getDataLayout().isLittleEndian() ? Index : Index << 4; | |||
13929 | Elems |= Index; | |||
13930 | ||||
13931 | return true; | |||
13932 | }; | |||
13933 | ||||
13934 | // If the build vector operands aren't sign extended vector extracts, | |||
13935 | // of the same input vector, then return. | |||
13936 | for (unsigned i = 0; i < N->getNumOperands(); i++) { | |||
13937 | if (!isSExtOfVecExtract(N->getOperand(i))) { | |||
13938 | return SDValue(); | |||
13939 | } | |||
13940 | } | |||
13941 | ||||
13942 | // If the vector extract indicies are not correct, add the appropriate | |||
13943 | // vector_shuffle. | |||
13944 | int TgtElemArrayIdx; | |||
13945 | int InputSize = Input.getValueType().getScalarSizeInBits(); | |||
13946 | int OutputSize = N->getValueType(0).getScalarSizeInBits(); | |||
13947 | if (InputSize + OutputSize == 40) | |||
13948 | TgtElemArrayIdx = 0; | |||
13949 | else if (InputSize + OutputSize == 72) | |||
13950 | TgtElemArrayIdx = 1; | |||
13951 | else if (InputSize + OutputSize == 48) | |||
13952 | TgtElemArrayIdx = 2; | |||
13953 | else if (InputSize + OutputSize == 80) | |||
13954 | TgtElemArrayIdx = 3; | |||
13955 | else if (InputSize + OutputSize == 96) | |||
13956 | TgtElemArrayIdx = 4; | |||
13957 | else | |||
13958 | return SDValue(); | |||
13959 | ||||
13960 | uint64_t CorrectElems = TargetElems[TgtElemArrayIdx]; | |||
13961 | CorrectElems = DAG.getDataLayout().isLittleEndian() | |||
13962 | ? CorrectElems & 0x0F0F0F0F0F0F0F0F | |||
13963 | : CorrectElems & 0xF0F0F0F0F0F0F0F0; | |||
13964 | if (Elems != CorrectElems) { | |||
13965 | return addShuffleForVecExtend(N, DAG, Input, Elems, CorrectElems); | |||
13966 | } | |||
13967 | ||||
13968 | // Regular lowering will catch cases where a shuffle is not needed. | |||
13969 | return SDValue(); | |||
13970 | } | |||
13971 | ||||
13972 | // Look for the pattern of a load from a narrow width to i128, feeding | |||
13973 | // into a BUILD_VECTOR of v1i128. Replace this sequence with a PPCISD node | |||
13974 | // (LXVRZX). This node represents a zero extending load that will be matched | |||
13975 | // to the Load VSX Vector Rightmost instructions. | |||
13976 | static SDValue combineBVZEXTLOAD(SDNode *N, SelectionDAG &DAG) { | |||
13977 | SDLoc DL(N); | |||
13978 | ||||
13979 | // This combine is only eligible for a BUILD_VECTOR of v1i128. | |||
13980 | if (N->getValueType(0) != MVT::v1i128) | |||
13981 | return SDValue(); | |||
13982 | ||||
13983 | SDValue Operand = N->getOperand(0); | |||
13984 | // Proceed with the transformation if the operand to the BUILD_VECTOR | |||
13985 | // is a load instruction. | |||
13986 | if (Operand.getOpcode() != ISD::LOAD) | |||
13987 | return SDValue(); | |||
13988 | ||||
13989 | auto *LD = cast<LoadSDNode>(Operand); | |||
13990 | EVT MemoryType = LD->getMemoryVT(); | |||
13991 | ||||
13992 | // This transformation is only valid if the we are loading either a byte, | |||
13993 | // halfword, word, or doubleword. | |||
13994 | bool ValidLDType = MemoryType == MVT::i8 || MemoryType == MVT::i16 || | |||
13995 | MemoryType == MVT::i32 || MemoryType == MVT::i64; | |||
13996 | ||||
13997 | // Ensure that the load from the narrow width is being zero extended to i128. | |||
13998 | if (!ValidLDType || | |||
13999 | (LD->getExtensionType() != ISD::ZEXTLOAD && | |||
14000 | LD->getExtensionType() != ISD::EXTLOAD)) | |||
14001 | return SDValue(); | |||
14002 | ||||
14003 | SDValue LoadOps[] = { | |||
14004 | LD->getChain(), LD->getBasePtr(), | |||
14005 | DAG.getIntPtrConstant(MemoryType.getScalarSizeInBits(), DL)}; | |||
14006 | ||||
14007 | return DAG.getMemIntrinsicNode(PPCISD::LXVRZX, DL, | |||
14008 | DAG.getVTList(MVT::v1i128, MVT::Other), | |||
14009 | LoadOps, MemoryType, LD->getMemOperand()); | |||
14010 | } | |||
14011 | ||||
14012 | SDValue PPCTargetLowering::DAGCombineBuildVector(SDNode *N, | |||
14013 | DAGCombinerInfo &DCI) const { | |||
14014 | assert(N->getOpcode() == ISD::BUILD_VECTOR &&(static_cast <bool> (N->getOpcode() == ISD::BUILD_VECTOR && "Should be called with a BUILD_VECTOR node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::BUILD_VECTOR && \"Should be called with a BUILD_VECTOR node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14015, __extension__ __PRETTY_FUNCTION__)) | |||
14015 | "Should be called with a BUILD_VECTOR node")(static_cast <bool> (N->getOpcode() == ISD::BUILD_VECTOR && "Should be called with a BUILD_VECTOR node") ? void (0) : __assert_fail ("N->getOpcode() == ISD::BUILD_VECTOR && \"Should be called with a BUILD_VECTOR node\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14015, __extension__ __PRETTY_FUNCTION__)); | |||
14016 | ||||
14017 | SelectionDAG &DAG = DCI.DAG; | |||
14018 | SDLoc dl(N); | |||
14019 | ||||
14020 | if (!Subtarget.hasVSX()) | |||
14021 | return SDValue(); | |||
14022 | ||||
14023 | // The target independent DAG combiner will leave a build_vector of | |||
14024 | // float-to-int conversions intact. We can generate MUCH better code for | |||
14025 | // a float-to-int conversion of a vector of floats. | |||
14026 | SDValue FirstInput = N->getOperand(0); | |||
14027 | if (FirstInput.getOpcode() == PPCISD::MFVSR) { | |||
14028 | SDValue Reduced = combineElementTruncationToVectorTruncation(N, DCI); | |||
14029 | if (Reduced) | |||
14030 | return Reduced; | |||
14031 | } | |||
14032 | ||||
14033 | // If we're building a vector out of consecutive loads, just load that | |||
14034 | // vector type. | |||
14035 | SDValue Reduced = combineBVOfConsecutiveLoads(N, DAG); | |||
14036 | if (Reduced) | |||
14037 | return Reduced; | |||
14038 | ||||
14039 | // If we're building a vector out of extended elements from another vector | |||
14040 | // we have P9 vector integer extend instructions. The code assumes legal | |||
14041 | // input types (i.e. it can't handle things like v4i16) so do not run before | |||
14042 | // legalization. | |||
14043 | if (Subtarget.hasP9Altivec() && !DCI.isBeforeLegalize()) { | |||
14044 | Reduced = combineBVOfVecSExt(N, DAG); | |||
14045 | if (Reduced) | |||
14046 | return Reduced; | |||
14047 | } | |||
14048 | ||||
14049 | // On Power10, the Load VSX Vector Rightmost instructions can be utilized | |||
14050 | // if this is a BUILD_VECTOR of v1i128, and if the operand to the BUILD_VECTOR | |||
14051 | // is a load from <valid narrow width> to i128. | |||
14052 | if (Subtarget.isISA3_1()) { | |||
14053 | SDValue BVOfZLoad = combineBVZEXTLOAD(N, DAG); | |||
14054 | if (BVOfZLoad) | |||
14055 | return BVOfZLoad; | |||
14056 | } | |||
14057 | ||||
14058 | if (N->getValueType(0) != MVT::v2f64) | |||
14059 | return SDValue(); | |||
14060 | ||||
14061 | // Looking for: | |||
14062 | // (build_vector ([su]int_to_fp (extractelt 0)), [su]int_to_fp (extractelt 1)) | |||
14063 | if (FirstInput.getOpcode() != ISD::SINT_TO_FP && | |||
14064 | FirstInput.getOpcode() != ISD::UINT_TO_FP) | |||
14065 | return SDValue(); | |||
14066 | if (N->getOperand(1).getOpcode() != ISD::SINT_TO_FP && | |||
14067 | N->getOperand(1).getOpcode() != ISD::UINT_TO_FP) | |||
14068 | return SDValue(); | |||
14069 | if (FirstInput.getOpcode() != N->getOperand(1).getOpcode()) | |||
14070 | return SDValue(); | |||
14071 | ||||
14072 | SDValue Ext1 = FirstInput.getOperand(0); | |||
14073 | SDValue Ext2 = N->getOperand(1).getOperand(0); | |||
14074 | if(Ext1.getOpcode() != ISD::EXTRACT_VECTOR_ELT || | |||
14075 | Ext2.getOpcode() != ISD::EXTRACT_VECTOR_ELT) | |||
14076 | return SDValue(); | |||
14077 | ||||
14078 | ConstantSDNode *Ext1Op = dyn_cast<ConstantSDNode>(Ext1.getOperand(1)); | |||
14079 | ConstantSDNode *Ext2Op = dyn_cast<ConstantSDNode>(Ext2.getOperand(1)); | |||
14080 | if (!Ext1Op || !Ext2Op) | |||
14081 | return SDValue(); | |||
14082 | if (Ext1.getOperand(0).getValueType() != MVT::v4i32 || | |||
14083 | Ext1.getOperand(0) != Ext2.getOperand(0)) | |||
14084 | return SDValue(); | |||
14085 | ||||
14086 | int FirstElem = Ext1Op->getZExtValue(); | |||
14087 | int SecondElem = Ext2Op->getZExtValue(); | |||
14088 | int SubvecIdx; | |||
14089 | if (FirstElem == 0 && SecondElem == 1) | |||
14090 | SubvecIdx = Subtarget.isLittleEndian() ? 1 : 0; | |||
14091 | else if (FirstElem == 2 && SecondElem == 3) | |||
14092 | SubvecIdx = Subtarget.isLittleEndian() ? 0 : 1; | |||
14093 | else | |||
14094 | return SDValue(); | |||
14095 | ||||
14096 | SDValue SrcVec = Ext1.getOperand(0); | |||
14097 | auto NodeType = (N->getOperand(1).getOpcode() == ISD::SINT_TO_FP) ? | |||
14098 | PPCISD::SINT_VEC_TO_FP : PPCISD::UINT_VEC_TO_FP; | |||
14099 | return DAG.getNode(NodeType, dl, MVT::v2f64, | |||
14100 | SrcVec, DAG.getIntPtrConstant(SubvecIdx, dl)); | |||
14101 | } | |||
14102 | ||||
14103 | SDValue PPCTargetLowering::combineFPToIntToFP(SDNode *N, | |||
14104 | DAGCombinerInfo &DCI) const { | |||
14105 | assert((N->getOpcode() == ISD::SINT_TO_FP ||(static_cast <bool> ((N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) && "Need an int -> FP conversion node here" ) ? void (0) : __assert_fail ("(N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) && \"Need an int -> FP conversion node here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14107, __extension__ __PRETTY_FUNCTION__)) | |||
14106 | N->getOpcode() == ISD::UINT_TO_FP) &&(static_cast <bool> ((N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) && "Need an int -> FP conversion node here" ) ? void (0) : __assert_fail ("(N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) && \"Need an int -> FP conversion node here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14107, __extension__ __PRETTY_FUNCTION__)) | |||
14107 | "Need an int -> FP conversion node here")(static_cast <bool> ((N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) && "Need an int -> FP conversion node here" ) ? void (0) : __assert_fail ("(N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) && \"Need an int -> FP conversion node here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14107, __extension__ __PRETTY_FUNCTION__)); | |||
14108 | ||||
14109 | if (useSoftFloat() || !Subtarget.has64BitSupport()) | |||
14110 | return SDValue(); | |||
14111 | ||||
14112 | SelectionDAG &DAG = DCI.DAG; | |||
14113 | SDLoc dl(N); | |||
14114 | SDValue Op(N, 0); | |||
14115 | ||||
14116 | // Don't handle ppc_fp128 here or conversions that are out-of-range capable | |||
14117 | // from the hardware. | |||
14118 | if (Op.getValueType() != MVT::f32 && Op.getValueType() != MVT::f64) | |||
14119 | return SDValue(); | |||
14120 | if (!Op.getOperand(0).getValueType().isSimple()) | |||
14121 | return SDValue(); | |||
14122 | if (Op.getOperand(0).getValueType().getSimpleVT() <= MVT(MVT::i1) || | |||
14123 | Op.getOperand(0).getValueType().getSimpleVT() > MVT(MVT::i64)) | |||
14124 | return SDValue(); | |||
14125 | ||||
14126 | SDValue FirstOperand(Op.getOperand(0)); | |||
14127 | bool SubWordLoad = FirstOperand.getOpcode() == ISD::LOAD && | |||
14128 | (FirstOperand.getValueType() == MVT::i8 || | |||
14129 | FirstOperand.getValueType() == MVT::i16); | |||
14130 | if (Subtarget.hasP9Vector() && Subtarget.hasP9Altivec() && SubWordLoad) { | |||
14131 | bool Signed = N->getOpcode() == ISD::SINT_TO_FP; | |||
14132 | bool DstDouble = Op.getValueType() == MVT::f64; | |||
14133 | unsigned ConvOp = Signed ? | |||
14134 | (DstDouble ? PPCISD::FCFID : PPCISD::FCFIDS) : | |||
14135 | (DstDouble ? PPCISD::FCFIDU : PPCISD::FCFIDUS); | |||
14136 | SDValue WidthConst = | |||
14137 | DAG.getIntPtrConstant(FirstOperand.getValueType() == MVT::i8 ? 1 : 2, | |||
14138 | dl, false); | |||
14139 | LoadSDNode *LDN = cast<LoadSDNode>(FirstOperand.getNode()); | |||
14140 | SDValue Ops[] = { LDN->getChain(), LDN->getBasePtr(), WidthConst }; | |||
14141 | SDValue Ld = DAG.getMemIntrinsicNode(PPCISD::LXSIZX, dl, | |||
14142 | DAG.getVTList(MVT::f64, MVT::Other), | |||
14143 | Ops, MVT::i8, LDN->getMemOperand()); | |||
14144 | ||||
14145 | // For signed conversion, we need to sign-extend the value in the VSR | |||
14146 | if (Signed) { | |||
14147 | SDValue ExtOps[] = { Ld, WidthConst }; | |||
14148 | SDValue Ext = DAG.getNode(PPCISD::VEXTS, dl, MVT::f64, ExtOps); | |||
14149 | return DAG.getNode(ConvOp, dl, DstDouble ? MVT::f64 : MVT::f32, Ext); | |||
14150 | } else | |||
14151 | return DAG.getNode(ConvOp, dl, DstDouble ? MVT::f64 : MVT::f32, Ld); | |||
14152 | } | |||
14153 | ||||
14154 | ||||
14155 | // For i32 intermediate values, unfortunately, the conversion functions | |||
14156 | // leave the upper 32 bits of the value are undefined. Within the set of | |||
14157 | // scalar instructions, we have no method for zero- or sign-extending the | |||
14158 | // value. Thus, we cannot handle i32 intermediate values here. | |||
14159 | if (Op.getOperand(0).getValueType() == MVT::i32) | |||
14160 | return SDValue(); | |||
14161 | ||||
14162 | assert((Op.getOpcode() == ISD::SINT_TO_FP || Subtarget.hasFPCVT()) &&(static_cast <bool> ((Op.getOpcode() == ISD::SINT_TO_FP || Subtarget.hasFPCVT()) && "UINT_TO_FP is supported only with FPCVT" ) ? void (0) : __assert_fail ("(Op.getOpcode() == ISD::SINT_TO_FP || Subtarget.hasFPCVT()) && \"UINT_TO_FP is supported only with FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14163, __extension__ __PRETTY_FUNCTION__)) | |||
14163 | "UINT_TO_FP is supported only with FPCVT")(static_cast <bool> ((Op.getOpcode() == ISD::SINT_TO_FP || Subtarget.hasFPCVT()) && "UINT_TO_FP is supported only with FPCVT" ) ? void (0) : __assert_fail ("(Op.getOpcode() == ISD::SINT_TO_FP || Subtarget.hasFPCVT()) && \"UINT_TO_FP is supported only with FPCVT\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14163, __extension__ __PRETTY_FUNCTION__)); | |||
14164 | ||||
14165 | // If we have FCFIDS, then use it when converting to single-precision. | |||
14166 | // Otherwise, convert to double-precision and then round. | |||
14167 | unsigned FCFOp = (Subtarget.hasFPCVT() && Op.getValueType() == MVT::f32) | |||
14168 | ? (Op.getOpcode() == ISD::UINT_TO_FP ? PPCISD::FCFIDUS | |||
14169 | : PPCISD::FCFIDS) | |||
14170 | : (Op.getOpcode() == ISD::UINT_TO_FP ? PPCISD::FCFIDU | |||
14171 | : PPCISD::FCFID); | |||
14172 | MVT FCFTy = (Subtarget.hasFPCVT() && Op.getValueType() == MVT::f32) | |||
14173 | ? MVT::f32 | |||
14174 | : MVT::f64; | |||
14175 | ||||
14176 | // If we're converting from a float, to an int, and back to a float again, | |||
14177 | // then we don't need the store/load pair at all. | |||
14178 | if ((Op.getOperand(0).getOpcode() == ISD::FP_TO_UINT && | |||
14179 | Subtarget.hasFPCVT()) || | |||
14180 | (Op.getOperand(0).getOpcode() == ISD::FP_TO_SINT)) { | |||
14181 | SDValue Src = Op.getOperand(0).getOperand(0); | |||
14182 | if (Src.getValueType() == MVT::f32) { | |||
14183 | Src = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Src); | |||
14184 | DCI.AddToWorklist(Src.getNode()); | |||
14185 | } else if (Src.getValueType() != MVT::f64) { | |||
14186 | // Make sure that we don't pick up a ppc_fp128 source value. | |||
14187 | return SDValue(); | |||
14188 | } | |||
14189 | ||||
14190 | unsigned FCTOp = | |||
14191 | Op.getOperand(0).getOpcode() == ISD::FP_TO_SINT ? PPCISD::FCTIDZ : | |||
14192 | PPCISD::FCTIDUZ; | |||
14193 | ||||
14194 | SDValue Tmp = DAG.getNode(FCTOp, dl, MVT::f64, Src); | |||
14195 | SDValue FP = DAG.getNode(FCFOp, dl, FCFTy, Tmp); | |||
14196 | ||||
14197 | if (Op.getValueType() == MVT::f32 && !Subtarget.hasFPCVT()) { | |||
14198 | FP = DAG.getNode(ISD::FP_ROUND, dl, | |||
14199 | MVT::f32, FP, DAG.getIntPtrConstant(0, dl)); | |||
14200 | DCI.AddToWorklist(FP.getNode()); | |||
14201 | } | |||
14202 | ||||
14203 | return FP; | |||
14204 | } | |||
14205 | ||||
14206 | return SDValue(); | |||
14207 | } | |||
14208 | ||||
14209 | // expandVSXLoadForLE - Convert VSX loads (which may be intrinsics for | |||
14210 | // builtins) into loads with swaps. | |||
14211 | SDValue PPCTargetLowering::expandVSXLoadForLE(SDNode *N, | |||
14212 | DAGCombinerInfo &DCI) const { | |||
14213 | SelectionDAG &DAG = DCI.DAG; | |||
14214 | SDLoc dl(N); | |||
14215 | SDValue Chain; | |||
14216 | SDValue Base; | |||
14217 | MachineMemOperand *MMO; | |||
14218 | ||||
14219 | switch (N->getOpcode()) { | |||
14220 | default: | |||
14221 | llvm_unreachable("Unexpected opcode for little endian VSX load")::llvm::llvm_unreachable_internal("Unexpected opcode for little endian VSX load" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14221); | |||
14222 | case ISD::LOAD: { | |||
14223 | LoadSDNode *LD = cast<LoadSDNode>(N); | |||
14224 | Chain = LD->getChain(); | |||
14225 | Base = LD->getBasePtr(); | |||
14226 | MMO = LD->getMemOperand(); | |||
14227 | // If the MMO suggests this isn't a load of a full vector, leave | |||
14228 | // things alone. For a built-in, we have to make the change for | |||
14229 | // correctness, so if there is a size problem that will be a bug. | |||
14230 | if (MMO->getSize() < 16) | |||
14231 | return SDValue(); | |||
14232 | break; | |||
14233 | } | |||
14234 | case ISD::INTRINSIC_W_CHAIN: { | |||
14235 | MemIntrinsicSDNode *Intrin = cast<MemIntrinsicSDNode>(N); | |||
14236 | Chain = Intrin->getChain(); | |||
14237 | // Similarly to the store case below, Intrin->getBasePtr() doesn't get | |||
14238 | // us what we want. Get operand 2 instead. | |||
14239 | Base = Intrin->getOperand(2); | |||
14240 | MMO = Intrin->getMemOperand(); | |||
14241 | break; | |||
14242 | } | |||
14243 | } | |||
14244 | ||||
14245 | MVT VecTy = N->getValueType(0).getSimpleVT(); | |||
14246 | ||||
14247 | // Do not expand to PPCISD::LXVD2X + PPCISD::XXSWAPD when the load is | |||
14248 | // aligned and the type is a vector with elements up to 4 bytes | |||
14249 | if (Subtarget.needsSwapsForVSXMemOps() && MMO->getAlign() >= Align(16) && | |||
14250 | VecTy.getScalarSizeInBits() <= 32) { | |||
14251 | return SDValue(); | |||
14252 | } | |||
14253 | ||||
14254 | SDValue LoadOps[] = { Chain, Base }; | |||
14255 | SDValue Load = DAG.getMemIntrinsicNode(PPCISD::LXVD2X, dl, | |||
14256 | DAG.getVTList(MVT::v2f64, MVT::Other), | |||
14257 | LoadOps, MVT::v2f64, MMO); | |||
14258 | ||||
14259 | DCI.AddToWorklist(Load.getNode()); | |||
14260 | Chain = Load.getValue(1); | |||
14261 | SDValue Swap = DAG.getNode( | |||
14262 | PPCISD::XXSWAPD, dl, DAG.getVTList(MVT::v2f64, MVT::Other), Chain, Load); | |||
14263 | DCI.AddToWorklist(Swap.getNode()); | |||
14264 | ||||
14265 | // Add a bitcast if the resulting load type doesn't match v2f64. | |||
14266 | if (VecTy != MVT::v2f64) { | |||
14267 | SDValue N = DAG.getNode(ISD::BITCAST, dl, VecTy, Swap); | |||
14268 | DCI.AddToWorklist(N.getNode()); | |||
14269 | // Package {bitcast value, swap's chain} to match Load's shape. | |||
14270 | return DAG.getNode(ISD::MERGE_VALUES, dl, DAG.getVTList(VecTy, MVT::Other), | |||
14271 | N, Swap.getValue(1)); | |||
14272 | } | |||
14273 | ||||
14274 | return Swap; | |||
14275 | } | |||
14276 | ||||
14277 | // expandVSXStoreForLE - Convert VSX stores (which may be intrinsics for | |||
14278 | // builtins) into stores with swaps. | |||
14279 | SDValue PPCTargetLowering::expandVSXStoreForLE(SDNode *N, | |||
14280 | DAGCombinerInfo &DCI) const { | |||
14281 | SelectionDAG &DAG = DCI.DAG; | |||
14282 | SDLoc dl(N); | |||
14283 | SDValue Chain; | |||
14284 | SDValue Base; | |||
14285 | unsigned SrcOpnd; | |||
14286 | MachineMemOperand *MMO; | |||
14287 | ||||
14288 | switch (N->getOpcode()) { | |||
14289 | default: | |||
14290 | llvm_unreachable("Unexpected opcode for little endian VSX store")::llvm::llvm_unreachable_internal("Unexpected opcode for little endian VSX store" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14290); | |||
14291 | case ISD::STORE: { | |||
14292 | StoreSDNode *ST = cast<StoreSDNode>(N); | |||
14293 | Chain = ST->getChain(); | |||
14294 | Base = ST->getBasePtr(); | |||
14295 | MMO = ST->getMemOperand(); | |||
14296 | SrcOpnd = 1; | |||
14297 | // If the MMO suggests this isn't a store of a full vector, leave | |||
14298 | // things alone. For a built-in, we have to make the change for | |||
14299 | // correctness, so if there is a size problem that will be a bug. | |||
14300 | if (MMO->getSize() < 16) | |||
14301 | return SDValue(); | |||
14302 | break; | |||
14303 | } | |||
14304 | case ISD::INTRINSIC_VOID: { | |||
14305 | MemIntrinsicSDNode *Intrin = cast<MemIntrinsicSDNode>(N); | |||
14306 | Chain = Intrin->getChain(); | |||
14307 | // Intrin->getBasePtr() oddly does not get what we want. | |||
14308 | Base = Intrin->getOperand(3); | |||
14309 | MMO = Intrin->getMemOperand(); | |||
14310 | SrcOpnd = 2; | |||
14311 | break; | |||
14312 | } | |||
14313 | } | |||
14314 | ||||
14315 | SDValue Src = N->getOperand(SrcOpnd); | |||
14316 | MVT VecTy = Src.getValueType().getSimpleVT(); | |||
14317 | ||||
14318 | // Do not expand to PPCISD::XXSWAPD and PPCISD::STXVD2X when the load is | |||
14319 | // aligned and the type is a vector with elements up to 4 bytes | |||
14320 | if (Subtarget.needsSwapsForVSXMemOps() && MMO->getAlign() >= Align(16) && | |||
14321 | VecTy.getScalarSizeInBits() <= 32) { | |||
14322 | return SDValue(); | |||
14323 | } | |||
14324 | ||||
14325 | // All stores are done as v2f64 and possible bit cast. | |||
14326 | if (VecTy != MVT::v2f64) { | |||
14327 | Src = DAG.getNode(ISD::BITCAST, dl, MVT::v2f64, Src); | |||
14328 | DCI.AddToWorklist(Src.getNode()); | |||
14329 | } | |||
14330 | ||||
14331 | SDValue Swap = DAG.getNode(PPCISD::XXSWAPD, dl, | |||
14332 | DAG.getVTList(MVT::v2f64, MVT::Other), Chain, Src); | |||
14333 | DCI.AddToWorklist(Swap.getNode()); | |||
14334 | Chain = Swap.getValue(1); | |||
14335 | SDValue StoreOps[] = { Chain, Swap, Base }; | |||
14336 | SDValue Store = DAG.getMemIntrinsicNode(PPCISD::STXVD2X, dl, | |||
14337 | DAG.getVTList(MVT::Other), | |||
14338 | StoreOps, VecTy, MMO); | |||
14339 | DCI.AddToWorklist(Store.getNode()); | |||
14340 | return Store; | |||
14341 | } | |||
14342 | ||||
14343 | // Handle DAG combine for STORE (FP_TO_INT F). | |||
14344 | SDValue PPCTargetLowering::combineStoreFPToInt(SDNode *N, | |||
14345 | DAGCombinerInfo &DCI) const { | |||
14346 | ||||
14347 | SelectionDAG &DAG = DCI.DAG; | |||
14348 | SDLoc dl(N); | |||
14349 | unsigned Opcode = N->getOperand(1).getOpcode(); | |||
14350 | ||||
14351 | assert((Opcode == ISD::FP_TO_SINT || Opcode == ISD::FP_TO_UINT)(static_cast <bool> ((Opcode == ISD::FP_TO_SINT || Opcode == ISD::FP_TO_UINT) && "Not a FP_TO_INT Instruction!" ) ? void (0) : __assert_fail ("(Opcode == ISD::FP_TO_SINT || Opcode == ISD::FP_TO_UINT) && \"Not a FP_TO_INT Instruction!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14352, __extension__ __PRETTY_FUNCTION__)) | |||
14352 | && "Not a FP_TO_INT Instruction!")(static_cast <bool> ((Opcode == ISD::FP_TO_SINT || Opcode == ISD::FP_TO_UINT) && "Not a FP_TO_INT Instruction!" ) ? void (0) : __assert_fail ("(Opcode == ISD::FP_TO_SINT || Opcode == ISD::FP_TO_UINT) && \"Not a FP_TO_INT Instruction!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14352, __extension__ __PRETTY_FUNCTION__)); | |||
14353 | ||||
14354 | SDValue Val = N->getOperand(1).getOperand(0); | |||
14355 | EVT Op1VT = N->getOperand(1).getValueType(); | |||
14356 | EVT ResVT = Val.getValueType(); | |||
14357 | ||||
14358 | if (!isTypeLegal(ResVT)) | |||
14359 | return SDValue(); | |||
14360 | ||||
14361 | // Only perform combine for conversion to i64/i32 or power9 i16/i8. | |||
14362 | bool ValidTypeForStoreFltAsInt = | |||
14363 | (Op1VT == MVT::i32 || Op1VT == MVT::i64 || | |||
14364 | (Subtarget.hasP9Vector() && (Op1VT == MVT::i16 || Op1VT == MVT::i8))); | |||
14365 | ||||
14366 | if (ResVT == MVT::f128 && !Subtarget.hasP9Vector()) | |||
14367 | return SDValue(); | |||
14368 | ||||
14369 | if (ResVT == MVT::ppcf128 || !Subtarget.hasP8Vector() || | |||
14370 | cast<StoreSDNode>(N)->isTruncatingStore() || !ValidTypeForStoreFltAsInt) | |||
14371 | return SDValue(); | |||
14372 | ||||
14373 | // Extend f32 values to f64 | |||
14374 | if (ResVT.getScalarSizeInBits() == 32) { | |||
14375 | Val = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Val); | |||
14376 | DCI.AddToWorklist(Val.getNode()); | |||
14377 | } | |||
14378 | ||||
14379 | // Set signed or unsigned conversion opcode. | |||
14380 | unsigned ConvOpcode = (Opcode == ISD::FP_TO_SINT) ? | |||
14381 | PPCISD::FP_TO_SINT_IN_VSR : | |||
14382 | PPCISD::FP_TO_UINT_IN_VSR; | |||
14383 | ||||
14384 | Val = DAG.getNode(ConvOpcode, | |||
14385 | dl, ResVT == MVT::f128 ? MVT::f128 : MVT::f64, Val); | |||
14386 | DCI.AddToWorklist(Val.getNode()); | |||
14387 | ||||
14388 | // Set number of bytes being converted. | |||
14389 | unsigned ByteSize = Op1VT.getScalarSizeInBits() / 8; | |||
14390 | SDValue Ops[] = { N->getOperand(0), Val, N->getOperand(2), | |||
14391 | DAG.getIntPtrConstant(ByteSize, dl, false), | |||
14392 | DAG.getValueType(Op1VT) }; | |||
14393 | ||||
14394 | Val = DAG.getMemIntrinsicNode(PPCISD::ST_VSR_SCAL_INT, dl, | |||
14395 | DAG.getVTList(MVT::Other), Ops, | |||
14396 | cast<StoreSDNode>(N)->getMemoryVT(), | |||
14397 | cast<StoreSDNode>(N)->getMemOperand()); | |||
14398 | ||||
14399 | DCI.AddToWorklist(Val.getNode()); | |||
14400 | return Val; | |||
14401 | } | |||
14402 | ||||
14403 | static bool isAlternatingShuffMask(const ArrayRef<int> &Mask, int NumElts) { | |||
14404 | // Check that the source of the element keeps flipping | |||
14405 | // (i.e. Mask[i] < NumElts -> Mask[i+i] >= NumElts). | |||
14406 | bool PrevElemFromFirstVec = Mask[0] < NumElts; | |||
14407 | for (int i = 1, e = Mask.size(); i < e; i++) { | |||
14408 | if (PrevElemFromFirstVec && Mask[i] < NumElts) | |||
14409 | return false; | |||
14410 | if (!PrevElemFromFirstVec && Mask[i] >= NumElts) | |||
14411 | return false; | |||
14412 | PrevElemFromFirstVec = !PrevElemFromFirstVec; | |||
14413 | } | |||
14414 | return true; | |||
14415 | } | |||
14416 | ||||
14417 | static bool isSplatBV(SDValue Op) { | |||
14418 | if (Op.getOpcode() != ISD::BUILD_VECTOR) | |||
14419 | return false; | |||
14420 | SDValue FirstOp; | |||
14421 | ||||
14422 | // Find first non-undef input. | |||
14423 | for (int i = 0, e = Op.getNumOperands(); i < e; i++) { | |||
14424 | FirstOp = Op.getOperand(i); | |||
14425 | if (!FirstOp.isUndef()) | |||
14426 | break; | |||
14427 | } | |||
14428 | ||||
14429 | // All inputs are undef or the same as the first non-undef input. | |||
14430 | for (int i = 1, e = Op.getNumOperands(); i < e; i++) | |||
14431 | if (Op.getOperand(i) != FirstOp && !Op.getOperand(i).isUndef()) | |||
14432 | return false; | |||
14433 | return true; | |||
14434 | } | |||
14435 | ||||
14436 | static SDValue isScalarToVec(SDValue Op) { | |||
14437 | if (Op.getOpcode() == ISD::SCALAR_TO_VECTOR) | |||
14438 | return Op; | |||
14439 | if (Op.getOpcode() != ISD::BITCAST) | |||
14440 | return SDValue(); | |||
14441 | Op = Op.getOperand(0); | |||
14442 | if (Op.getOpcode() == ISD::SCALAR_TO_VECTOR) | |||
14443 | return Op; | |||
14444 | return SDValue(); | |||
14445 | } | |||
14446 | ||||
14447 | // Fix up the shuffle mask to account for the fact that the result of | |||
14448 | // scalar_to_vector is not in lane zero. This just takes all values in | |||
14449 | // the ranges specified by the min/max indices and adds the number of | |||
14450 | // elements required to ensure each element comes from the respective | |||
14451 | // position in the valid lane. | |||
14452 | // On little endian, that's just the corresponding element in the other | |||
14453 | // half of the vector. On big endian, it is in the same half but right | |||
14454 | // justified rather than left justified in that half. | |||
14455 | static void fixupShuffleMaskForPermutedSToV(SmallVectorImpl<int> &ShuffV, | |||
14456 | int LHSMaxIdx, int RHSMinIdx, | |||
14457 | int RHSMaxIdx, int HalfVec, | |||
14458 | unsigned ValidLaneWidth, | |||
14459 | const PPCSubtarget &Subtarget) { | |||
14460 | for (int i = 0, e = ShuffV.size(); i < e; i++) { | |||
14461 | int Idx = ShuffV[i]; | |||
14462 | if ((Idx >= 0 && Idx < LHSMaxIdx) || (Idx >= RHSMinIdx && Idx < RHSMaxIdx)) | |||
14463 | ShuffV[i] += | |||
14464 | Subtarget.isLittleEndian() ? HalfVec : HalfVec - ValidLaneWidth; | |||
14465 | } | |||
14466 | } | |||
14467 | ||||
14468 | // Replace a SCALAR_TO_VECTOR with a SCALAR_TO_VECTOR_PERMUTED except if | |||
14469 | // the original is: | |||
14470 | // (<n x Ty> (scalar_to_vector (Ty (extract_elt <n x Ty> %a, C)))) | |||
14471 | // In such a case, just change the shuffle mask to extract the element | |||
14472 | // from the permuted index. | |||
14473 | static SDValue getSToVPermuted(SDValue OrigSToV, SelectionDAG &DAG, | |||
14474 | const PPCSubtarget &Subtarget) { | |||
14475 | SDLoc dl(OrigSToV); | |||
14476 | EVT VT = OrigSToV.getValueType(); | |||
14477 | assert(OrigSToV.getOpcode() == ISD::SCALAR_TO_VECTOR &&(static_cast <bool> (OrigSToV.getOpcode() == ISD::SCALAR_TO_VECTOR && "Expecting a SCALAR_TO_VECTOR here") ? void (0) : __assert_fail ("OrigSToV.getOpcode() == ISD::SCALAR_TO_VECTOR && \"Expecting a SCALAR_TO_VECTOR here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14478, __extension__ __PRETTY_FUNCTION__)) | |||
14478 | "Expecting a SCALAR_TO_VECTOR here")(static_cast <bool> (OrigSToV.getOpcode() == ISD::SCALAR_TO_VECTOR && "Expecting a SCALAR_TO_VECTOR here") ? void (0) : __assert_fail ("OrigSToV.getOpcode() == ISD::SCALAR_TO_VECTOR && \"Expecting a SCALAR_TO_VECTOR here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14478, __extension__ __PRETTY_FUNCTION__)); | |||
14479 | SDValue Input = OrigSToV.getOperand(0); | |||
14480 | ||||
14481 | if (Input.getOpcode() == ISD::EXTRACT_VECTOR_ELT) { | |||
14482 | ConstantSDNode *Idx = dyn_cast<ConstantSDNode>(Input.getOperand(1)); | |||
14483 | SDValue OrigVector = Input.getOperand(0); | |||
14484 | ||||
14485 | // Can't handle non-const element indices or different vector types | |||
14486 | // for the input to the extract and the output of the scalar_to_vector. | |||
14487 | if (Idx && VT == OrigVector.getValueType()) { | |||
14488 | unsigned NumElts = VT.getVectorNumElements(); | |||
14489 | assert((static_cast <bool> (NumElts > 1 && "Cannot produce a permuted scalar_to_vector for one element vector" ) ? void (0) : __assert_fail ("NumElts > 1 && \"Cannot produce a permuted scalar_to_vector for one element vector\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14491, __extension__ __PRETTY_FUNCTION__)) | |||
14490 | NumElts > 1 &&(static_cast <bool> (NumElts > 1 && "Cannot produce a permuted scalar_to_vector for one element vector" ) ? void (0) : __assert_fail ("NumElts > 1 && \"Cannot produce a permuted scalar_to_vector for one element vector\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14491, __extension__ __PRETTY_FUNCTION__)) | |||
14491 | "Cannot produce a permuted scalar_to_vector for one element vector")(static_cast <bool> (NumElts > 1 && "Cannot produce a permuted scalar_to_vector for one element vector" ) ? void (0) : __assert_fail ("NumElts > 1 && \"Cannot produce a permuted scalar_to_vector for one element vector\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14491, __extension__ __PRETTY_FUNCTION__)); | |||
14492 | SmallVector<int, 16> NewMask(NumElts, -1); | |||
14493 | unsigned ResultInElt = NumElts / 2; | |||
14494 | ResultInElt -= Subtarget.isLittleEndian() ? 0 : 1; | |||
14495 | NewMask[ResultInElt] = Idx->getZExtValue(); | |||
14496 | return DAG.getVectorShuffle(VT, dl, OrigVector, OrigVector, NewMask); | |||
14497 | } | |||
14498 | } | |||
14499 | return DAG.getNode(PPCISD::SCALAR_TO_VECTOR_PERMUTED, dl, VT, | |||
14500 | OrigSToV.getOperand(0)); | |||
14501 | } | |||
14502 | ||||
14503 | // On little endian subtargets, combine shuffles such as: | |||
14504 | // vector_shuffle<16,1,17,3,18,5,19,7,20,9,21,11,22,13,23,15>, <zero>, %b | |||
14505 | // into: | |||
14506 | // vector_shuffle<16,0,17,1,18,2,19,3,20,4,21,5,22,6,23,7>, <zero>, %b | |||
14507 | // because the latter can be matched to a single instruction merge. | |||
14508 | // Furthermore, SCALAR_TO_VECTOR on little endian always involves a permute | |||
14509 | // to put the value into element zero. Adjust the shuffle mask so that the | |||
14510 | // vector can remain in permuted form (to prevent a swap prior to a shuffle). | |||
14511 | // On big endian targets, this is still useful for SCALAR_TO_VECTOR | |||
14512 | // nodes with elements smaller than doubleword because all the ways | |||
14513 | // of getting scalar data into a vector register put the value in the | |||
14514 | // rightmost element of the left half of the vector. | |||
14515 | SDValue PPCTargetLowering::combineVectorShuffle(ShuffleVectorSDNode *SVN, | |||
14516 | SelectionDAG &DAG) const { | |||
14517 | SDValue LHS = SVN->getOperand(0); | |||
14518 | SDValue RHS = SVN->getOperand(1); | |||
14519 | auto Mask = SVN->getMask(); | |||
14520 | int NumElts = LHS.getValueType().getVectorNumElements(); | |||
14521 | SDValue Res(SVN, 0); | |||
14522 | SDLoc dl(SVN); | |||
14523 | bool IsLittleEndian = Subtarget.isLittleEndian(); | |||
14524 | ||||
14525 | // On big endian targets this is only useful for subtargets with direct moves. | |||
14526 | // On little endian targets it would be useful for all subtargets with VSX. | |||
14527 | // However adding special handling for LE subtargets without direct moves | |||
14528 | // would be wasted effort since the minimum arch for LE is ISA 2.07 (Power8) | |||
14529 | // which includes direct moves. | |||
14530 | if (!Subtarget.hasDirectMove()) | |||
14531 | return Res; | |||
14532 | ||||
14533 | // If this is not a shuffle of a shuffle and the first element comes from | |||
14534 | // the second vector, canonicalize to the commuted form. This will make it | |||
14535 | // more likely to match one of the single instruction patterns. | |||
14536 | if (Mask[0] >= NumElts && LHS.getOpcode() != ISD::VECTOR_SHUFFLE && | |||
14537 | RHS.getOpcode() != ISD::VECTOR_SHUFFLE) { | |||
14538 | std::swap(LHS, RHS); | |||
14539 | Res = DAG.getCommutedVectorShuffle(*SVN); | |||
14540 | Mask = cast<ShuffleVectorSDNode>(Res)->getMask(); | |||
14541 | } | |||
14542 | ||||
14543 | // Adjust the shuffle mask if either input vector comes from a | |||
14544 | // SCALAR_TO_VECTOR and keep the respective input vector in permuted | |||
14545 | // form (to prevent the need for a swap). | |||
14546 | SmallVector<int, 16> ShuffV(Mask.begin(), Mask.end()); | |||
14547 | SDValue SToVLHS = isScalarToVec(LHS); | |||
14548 | SDValue SToVRHS = isScalarToVec(RHS); | |||
14549 | if (SToVLHS || SToVRHS) { | |||
14550 | int NumEltsIn = SToVLHS ? SToVLHS.getValueType().getVectorNumElements() | |||
14551 | : SToVRHS.getValueType().getVectorNumElements(); | |||
14552 | int NumEltsOut = ShuffV.size(); | |||
14553 | // The width of the "valid lane" (i.e. the lane that contains the value that | |||
14554 | // is vectorized) needs to be expressed in terms of the number of elements | |||
14555 | // of the shuffle. It is thereby the ratio of the values before and after | |||
14556 | // any bitcast. | |||
14557 | unsigned ValidLaneWidth = | |||
14558 | SToVLHS ? SToVLHS.getValueType().getScalarSizeInBits() / | |||
14559 | LHS.getValueType().getScalarSizeInBits() | |||
14560 | : SToVRHS.getValueType().getScalarSizeInBits() / | |||
14561 | RHS.getValueType().getScalarSizeInBits(); | |||
14562 | ||||
14563 | // Initially assume that neither input is permuted. These will be adjusted | |||
14564 | // accordingly if either input is. | |||
14565 | int LHSMaxIdx = -1; | |||
14566 | int RHSMinIdx = -1; | |||
14567 | int RHSMaxIdx = -1; | |||
14568 | int HalfVec = LHS.getValueType().getVectorNumElements() / 2; | |||
14569 | ||||
14570 | // Get the permuted scalar to vector nodes for the source(s) that come from | |||
14571 | // ISD::SCALAR_TO_VECTOR. | |||
14572 | // On big endian systems, this only makes sense for element sizes smaller | |||
14573 | // than 64 bits since for 64-bit elements, all instructions already put | |||
14574 | // the value into element zero. Since scalar size of LHS and RHS may differ | |||
14575 | // after isScalarToVec, this should be checked using their own sizes. | |||
14576 | if (SToVLHS) { | |||
14577 | if (!IsLittleEndian && SToVLHS.getValueType().getScalarSizeInBits() >= 64) | |||
14578 | return Res; | |||
14579 | // Set up the values for the shuffle vector fixup. | |||
14580 | LHSMaxIdx = NumEltsOut / NumEltsIn; | |||
14581 | SToVLHS = getSToVPermuted(SToVLHS, DAG, Subtarget); | |||
14582 | if (SToVLHS.getValueType() != LHS.getValueType()) | |||
14583 | SToVLHS = DAG.getBitcast(LHS.getValueType(), SToVLHS); | |||
14584 | LHS = SToVLHS; | |||
14585 | } | |||
14586 | if (SToVRHS) { | |||
14587 | if (!IsLittleEndian && SToVRHS.getValueType().getScalarSizeInBits() >= 64) | |||
14588 | return Res; | |||
14589 | RHSMinIdx = NumEltsOut; | |||
14590 | RHSMaxIdx = NumEltsOut / NumEltsIn + RHSMinIdx; | |||
14591 | SToVRHS = getSToVPermuted(SToVRHS, DAG, Subtarget); | |||
14592 | if (SToVRHS.getValueType() != RHS.getValueType()) | |||
14593 | SToVRHS = DAG.getBitcast(RHS.getValueType(), SToVRHS); | |||
14594 | RHS = SToVRHS; | |||
14595 | } | |||
14596 | ||||
14597 | // Fix up the shuffle mask to reflect where the desired element actually is. | |||
14598 | // The minimum and maximum indices that correspond to element zero for both | |||
14599 | // the LHS and RHS are computed and will control which shuffle mask entries | |||
14600 | // are to be changed. For example, if the RHS is permuted, any shuffle mask | |||
14601 | // entries in the range [RHSMinIdx,RHSMaxIdx) will be adjusted. | |||
14602 | fixupShuffleMaskForPermutedSToV(ShuffV, LHSMaxIdx, RHSMinIdx, RHSMaxIdx, | |||
14603 | HalfVec, ValidLaneWidth, Subtarget); | |||
14604 | Res = DAG.getVectorShuffle(SVN->getValueType(0), dl, LHS, RHS, ShuffV); | |||
14605 | ||||
14606 | // We may have simplified away the shuffle. We won't be able to do anything | |||
14607 | // further with it here. | |||
14608 | if (!isa<ShuffleVectorSDNode>(Res)) | |||
14609 | return Res; | |||
14610 | Mask = cast<ShuffleVectorSDNode>(Res)->getMask(); | |||
14611 | } | |||
14612 | ||||
14613 | SDValue TheSplat = IsLittleEndian ? RHS : LHS; | |||
14614 | // The common case after we commuted the shuffle is that the RHS is a splat | |||
14615 | // and we have elements coming in from the splat at indices that are not | |||
14616 | // conducive to using a merge. | |||
14617 | // Example: | |||
14618 | // vector_shuffle<0,17,1,19,2,21,3,23,4,25,5,27,6,29,7,31> t1, <zero> | |||
14619 | if (!isSplatBV(TheSplat)) | |||
14620 | return Res; | |||
14621 | ||||
14622 | // We are looking for a mask such that all even elements are from | |||
14623 | // one vector and all odd elements from the other. | |||
14624 | if (!isAlternatingShuffMask(Mask, NumElts)) | |||
14625 | return Res; | |||
14626 | ||||
14627 | // Adjust the mask so we are pulling in the same index from the splat | |||
14628 | // as the index from the interesting vector in consecutive elements. | |||
14629 | if (IsLittleEndian) { | |||
14630 | // Example (even elements from first vector): | |||
14631 | // vector_shuffle<0,16,1,17,2,18,3,19,4,20,5,21,6,22,7,23> t1, <zero> | |||
14632 | if (Mask[0] < NumElts) | |||
14633 | for (int i = 1, e = Mask.size(); i < e; i += 2) | |||
14634 | ShuffV[i] = (ShuffV[i - 1] + NumElts); | |||
14635 | // Example (odd elements from first vector): | |||
14636 | // vector_shuffle<16,0,17,1,18,2,19,3,20,4,21,5,22,6,23,7> t1, <zero> | |||
14637 | else | |||
14638 | for (int i = 0, e = Mask.size(); i < e; i += 2) | |||
14639 | ShuffV[i] = (ShuffV[i + 1] + NumElts); | |||
14640 | } else { | |||
14641 | // Example (even elements from first vector): | |||
14642 | // vector_shuffle<0,16,1,17,2,18,3,19,4,20,5,21,6,22,7,23> <zero>, t1 | |||
14643 | if (Mask[0] < NumElts) | |||
14644 | for (int i = 0, e = Mask.size(); i < e; i += 2) | |||
14645 | ShuffV[i] = ShuffV[i + 1] - NumElts; | |||
14646 | // Example (odd elements from first vector): | |||
14647 | // vector_shuffle<16,0,17,1,18,2,19,3,20,4,21,5,22,6,23,7> <zero>, t1 | |||
14648 | else | |||
14649 | for (int i = 1, e = Mask.size(); i < e; i += 2) | |||
14650 | ShuffV[i] = ShuffV[i - 1] - NumElts; | |||
14651 | } | |||
14652 | ||||
14653 | // If the RHS has undefs, we need to remove them since we may have created | |||
14654 | // a shuffle that adds those instead of the splat value. | |||
14655 | SDValue SplatVal = | |||
14656 | cast<BuildVectorSDNode>(TheSplat.getNode())->getSplatValue(); | |||
14657 | TheSplat = DAG.getSplatBuildVector(TheSplat.getValueType(), dl, SplatVal); | |||
14658 | ||||
14659 | if (IsLittleEndian) | |||
14660 | RHS = TheSplat; | |||
14661 | else | |||
14662 | LHS = TheSplat; | |||
14663 | return DAG.getVectorShuffle(SVN->getValueType(0), dl, LHS, RHS, ShuffV); | |||
14664 | } | |||
14665 | ||||
14666 | SDValue PPCTargetLowering::combineVReverseMemOP(ShuffleVectorSDNode *SVN, | |||
14667 | LSBaseSDNode *LSBase, | |||
14668 | DAGCombinerInfo &DCI) const { | |||
14669 | assert((ISD::isNormalLoad(LSBase) || ISD::isNormalStore(LSBase)) &&(static_cast <bool> ((ISD::isNormalLoad(LSBase) || ISD:: isNormalStore(LSBase)) && "Not a reverse memop pattern!" ) ? void (0) : __assert_fail ("(ISD::isNormalLoad(LSBase) || ISD::isNormalStore(LSBase)) && \"Not a reverse memop pattern!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14670, __extension__ __PRETTY_FUNCTION__)) | |||
14670 | "Not a reverse memop pattern!")(static_cast <bool> ((ISD::isNormalLoad(LSBase) || ISD:: isNormalStore(LSBase)) && "Not a reverse memop pattern!" ) ? void (0) : __assert_fail ("(ISD::isNormalLoad(LSBase) || ISD::isNormalStore(LSBase)) && \"Not a reverse memop pattern!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14670, __extension__ __PRETTY_FUNCTION__)); | |||
14671 | ||||
14672 | auto IsElementReverse = [](const ShuffleVectorSDNode *SVN) -> bool { | |||
14673 | auto Mask = SVN->getMask(); | |||
14674 | int i = 0; | |||
14675 | auto I = Mask.rbegin(); | |||
14676 | auto E = Mask.rend(); | |||
14677 | ||||
14678 | for (; I != E; ++I) { | |||
14679 | if (*I != i) | |||
14680 | return false; | |||
14681 | i++; | |||
14682 | } | |||
14683 | return true; | |||
14684 | }; | |||
14685 | ||||
14686 | SelectionDAG &DAG = DCI.DAG; | |||
14687 | EVT VT = SVN->getValueType(0); | |||
14688 | ||||
14689 | if (!isTypeLegal(VT) || !Subtarget.isLittleEndian() || !Subtarget.hasVSX()) | |||
14690 | return SDValue(); | |||
14691 | ||||
14692 | // Before P9, we have PPCVSXSwapRemoval pass to hack the element order. | |||
14693 | // See comment in PPCVSXSwapRemoval.cpp. | |||
14694 | // It is conflict with PPCVSXSwapRemoval opt. So we don't do it. | |||
14695 | if (!Subtarget.hasP9Vector()) | |||
14696 | return SDValue(); | |||
14697 | ||||
14698 | if(!IsElementReverse(SVN)) | |||
14699 | return SDValue(); | |||
14700 | ||||
14701 | if (LSBase->getOpcode() == ISD::LOAD) { | |||
14702 | // If the load return value 0 has more than one user except the | |||
14703 | // shufflevector instruction, it is not profitable to replace the | |||
14704 | // shufflevector with a reverse load. | |||
14705 | for (SDNode::use_iterator UI = LSBase->use_begin(), UE = LSBase->use_end(); | |||
14706 | UI != UE; ++UI) | |||
14707 | if (UI.getUse().getResNo() == 0 && UI->getOpcode() != ISD::VECTOR_SHUFFLE) | |||
14708 | return SDValue(); | |||
14709 | ||||
14710 | SDLoc dl(LSBase); | |||
14711 | SDValue LoadOps[] = {LSBase->getChain(), LSBase->getBasePtr()}; | |||
14712 | return DAG.getMemIntrinsicNode( | |||
14713 | PPCISD::LOAD_VEC_BE, dl, DAG.getVTList(VT, MVT::Other), LoadOps, | |||
14714 | LSBase->getMemoryVT(), LSBase->getMemOperand()); | |||
14715 | } | |||
14716 | ||||
14717 | if (LSBase->getOpcode() == ISD::STORE) { | |||
14718 | // If there are other uses of the shuffle, the swap cannot be avoided. | |||
14719 | // Forcing the use of an X-Form (since swapped stores only have | |||
14720 | // X-Forms) without removing the swap is unprofitable. | |||
14721 | if (!SVN->hasOneUse()) | |||
14722 | return SDValue(); | |||
14723 | ||||
14724 | SDLoc dl(LSBase); | |||
14725 | SDValue StoreOps[] = {LSBase->getChain(), SVN->getOperand(0), | |||
14726 | LSBase->getBasePtr()}; | |||
14727 | return DAG.getMemIntrinsicNode( | |||
14728 | PPCISD::STORE_VEC_BE, dl, DAG.getVTList(MVT::Other), StoreOps, | |||
14729 | LSBase->getMemoryVT(), LSBase->getMemOperand()); | |||
14730 | } | |||
14731 | ||||
14732 | llvm_unreachable("Expected a load or store node here")::llvm::llvm_unreachable_internal("Expected a load or store node here" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14732); | |||
14733 | } | |||
14734 | ||||
14735 | SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, | |||
14736 | DAGCombinerInfo &DCI) const { | |||
14737 | SelectionDAG &DAG = DCI.DAG; | |||
14738 | SDLoc dl(N); | |||
14739 | switch (N->getOpcode()) { | |||
14740 | default: break; | |||
14741 | case ISD::ADD: | |||
14742 | return combineADD(N, DCI); | |||
14743 | case ISD::SHL: | |||
14744 | return combineSHL(N, DCI); | |||
14745 | case ISD::SRA: | |||
14746 | return combineSRA(N, DCI); | |||
14747 | case ISD::SRL: | |||
14748 | return combineSRL(N, DCI); | |||
14749 | case ISD::MUL: | |||
14750 | return combineMUL(N, DCI); | |||
14751 | case ISD::FMA: | |||
14752 | case PPCISD::FNMSUB: | |||
14753 | return combineFMALike(N, DCI); | |||
14754 | case PPCISD::SHL: | |||
14755 | if (isNullConstant(N->getOperand(0))) // 0 << V -> 0. | |||
14756 | return N->getOperand(0); | |||
14757 | break; | |||
14758 | case PPCISD::SRL: | |||
14759 | if (isNullConstant(N->getOperand(0))) // 0 >>u V -> 0. | |||
14760 | return N->getOperand(0); | |||
14761 | break; | |||
14762 | case PPCISD::SRA: | |||
14763 | if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(N->getOperand(0))) { | |||
14764 | if (C->isNullValue() || // 0 >>s V -> 0. | |||
14765 | C->isAllOnesValue()) // -1 >>s V -> -1. | |||
14766 | return N->getOperand(0); | |||
14767 | } | |||
14768 | break; | |||
14769 | case ISD::SIGN_EXTEND: | |||
14770 | case ISD::ZERO_EXTEND: | |||
14771 | case ISD::ANY_EXTEND: | |||
14772 | return DAGCombineExtBoolTrunc(N, DCI); | |||
14773 | case ISD::TRUNCATE: | |||
14774 | return combineTRUNCATE(N, DCI); | |||
14775 | case ISD::SETCC: | |||
14776 | if (SDValue CSCC = combineSetCC(N, DCI)) | |||
14777 | return CSCC; | |||
14778 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
14779 | case ISD::SELECT_CC: | |||
14780 | return DAGCombineTruncBoolExt(N, DCI); | |||
14781 | case ISD::SINT_TO_FP: | |||
14782 | case ISD::UINT_TO_FP: | |||
14783 | return combineFPToIntToFP(N, DCI); | |||
14784 | case ISD::VECTOR_SHUFFLE: | |||
14785 | if (ISD::isNormalLoad(N->getOperand(0).getNode())) { | |||
14786 | LSBaseSDNode* LSBase = cast<LSBaseSDNode>(N->getOperand(0)); | |||
14787 | return combineVReverseMemOP(cast<ShuffleVectorSDNode>(N), LSBase, DCI); | |||
14788 | } | |||
14789 | return combineVectorShuffle(cast<ShuffleVectorSDNode>(N), DCI.DAG); | |||
14790 | case ISD::STORE: { | |||
14791 | ||||
14792 | EVT Op1VT = N->getOperand(1).getValueType(); | |||
14793 | unsigned Opcode = N->getOperand(1).getOpcode(); | |||
14794 | ||||
14795 | if (Opcode == ISD::FP_TO_SINT || Opcode == ISD::FP_TO_UINT) { | |||
14796 | SDValue Val= combineStoreFPToInt(N, DCI); | |||
14797 | if (Val) | |||
14798 | return Val; | |||
14799 | } | |||
14800 | ||||
14801 | if (Opcode == ISD::VECTOR_SHUFFLE && ISD::isNormalStore(N)) { | |||
14802 | ShuffleVectorSDNode *SVN = cast<ShuffleVectorSDNode>(N->getOperand(1)); | |||
14803 | SDValue Val= combineVReverseMemOP(SVN, cast<LSBaseSDNode>(N), DCI); | |||
14804 | if (Val) | |||
14805 | return Val; | |||
14806 | } | |||
14807 | ||||
14808 | // Turn STORE (BSWAP) -> sthbrx/stwbrx. | |||
14809 | if (cast<StoreSDNode>(N)->isUnindexed() && Opcode == ISD::BSWAP && | |||
14810 | N->getOperand(1).getNode()->hasOneUse() && | |||
14811 | (Op1VT == MVT::i32 || Op1VT == MVT::i16 || | |||
14812 | (Subtarget.hasLDBRX() && Subtarget.isPPC64() && Op1VT == MVT::i64))) { | |||
14813 | ||||
14814 | // STBRX can only handle simple types and it makes no sense to store less | |||
14815 | // two bytes in byte-reversed order. | |||
14816 | EVT mVT = cast<StoreSDNode>(N)->getMemoryVT(); | |||
14817 | if (mVT.isExtended() || mVT.getSizeInBits() < 16) | |||
14818 | break; | |||
14819 | ||||
14820 | SDValue BSwapOp = N->getOperand(1).getOperand(0); | |||
14821 | // Do an any-extend to 32-bits if this is a half-word input. | |||
14822 | if (BSwapOp.getValueType() == MVT::i16) | |||
14823 | BSwapOp = DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i32, BSwapOp); | |||
14824 | ||||
14825 | // If the type of BSWAP operand is wider than stored memory width | |||
14826 | // it need to be shifted to the right side before STBRX. | |||
14827 | if (Op1VT.bitsGT(mVT)) { | |||
14828 | int Shift = Op1VT.getSizeInBits() - mVT.getSizeInBits(); | |||
14829 | BSwapOp = DAG.getNode(ISD::SRL, dl, Op1VT, BSwapOp, | |||
14830 | DAG.getConstant(Shift, dl, MVT::i32)); | |||
14831 | // Need to truncate if this is a bswap of i64 stored as i32/i16. | |||
14832 | if (Op1VT == MVT::i64) | |||
14833 | BSwapOp = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, BSwapOp); | |||
14834 | } | |||
14835 | ||||
14836 | SDValue Ops[] = { | |||
14837 | N->getOperand(0), BSwapOp, N->getOperand(2), DAG.getValueType(mVT) | |||
14838 | }; | |||
14839 | return | |||
14840 | DAG.getMemIntrinsicNode(PPCISD::STBRX, dl, DAG.getVTList(MVT::Other), | |||
14841 | Ops, cast<StoreSDNode>(N)->getMemoryVT(), | |||
14842 | cast<StoreSDNode>(N)->getMemOperand()); | |||
14843 | } | |||
14844 | ||||
14845 | // STORE Constant:i32<0> -> STORE<trunc to i32> Constant:i64<0> | |||
14846 | // So it can increase the chance of CSE constant construction. | |||
14847 | if (Subtarget.isPPC64() && !DCI.isBeforeLegalize() && | |||
14848 | isa<ConstantSDNode>(N->getOperand(1)) && Op1VT == MVT::i32) { | |||
14849 | // Need to sign-extended to 64-bits to handle negative values. | |||
14850 | EVT MemVT = cast<StoreSDNode>(N)->getMemoryVT(); | |||
14851 | uint64_t Val64 = SignExtend64(N->getConstantOperandVal(1), | |||
14852 | MemVT.getSizeInBits()); | |||
14853 | SDValue Const64 = DAG.getConstant(Val64, dl, MVT::i64); | |||
14854 | ||||
14855 | // DAG.getTruncStore() can't be used here because it doesn't accept | |||
14856 | // the general (base + offset) addressing mode. | |||
14857 | // So we use UpdateNodeOperands and setTruncatingStore instead. | |||
14858 | DAG.UpdateNodeOperands(N, N->getOperand(0), Const64, N->getOperand(2), | |||
14859 | N->getOperand(3)); | |||
14860 | cast<StoreSDNode>(N)->setTruncatingStore(true); | |||
14861 | return SDValue(N, 0); | |||
14862 | } | |||
14863 | ||||
14864 | // For little endian, VSX stores require generating xxswapd/lxvd2x. | |||
14865 | // Not needed on ISA 3.0 based CPUs since we have a non-permuting store. | |||
14866 | if (Op1VT.isSimple()) { | |||
14867 | MVT StoreVT = Op1VT.getSimpleVT(); | |||
14868 | if (Subtarget.needsSwapsForVSXMemOps() && | |||
14869 | (StoreVT == MVT::v2f64 || StoreVT == MVT::v2i64 || | |||
14870 | StoreVT == MVT::v4f32 || StoreVT == MVT::v4i32)) | |||
14871 | return expandVSXStoreForLE(N, DCI); | |||
14872 | } | |||
14873 | break; | |||
14874 | } | |||
14875 | case ISD::LOAD: { | |||
14876 | LoadSDNode *LD = cast<LoadSDNode>(N); | |||
14877 | EVT VT = LD->getValueType(0); | |||
14878 | ||||
14879 | // For little endian, VSX loads require generating lxvd2x/xxswapd. | |||
14880 | // Not needed on ISA 3.0 based CPUs since we have a non-permuting load. | |||
14881 | if (VT.isSimple()) { | |||
14882 | MVT LoadVT = VT.getSimpleVT(); | |||
14883 | if (Subtarget.needsSwapsForVSXMemOps() && | |||
14884 | (LoadVT == MVT::v2f64 || LoadVT == MVT::v2i64 || | |||
14885 | LoadVT == MVT::v4f32 || LoadVT == MVT::v4i32)) | |||
14886 | return expandVSXLoadForLE(N, DCI); | |||
14887 | } | |||
14888 | ||||
14889 | // We sometimes end up with a 64-bit integer load, from which we extract | |||
14890 | // two single-precision floating-point numbers. This happens with | |||
14891 | // std::complex<float>, and other similar structures, because of the way we | |||
14892 | // canonicalize structure copies. However, if we lack direct moves, | |||
14893 | // then the final bitcasts from the extracted integer values to the | |||
14894 | // floating-point numbers turn into store/load pairs. Even with direct moves, | |||
14895 | // just loading the two floating-point numbers is likely better. | |||
14896 | auto ReplaceTwoFloatLoad = [&]() { | |||
14897 | if (VT != MVT::i64) | |||
14898 | return false; | |||
14899 | ||||
14900 | if (LD->getExtensionType() != ISD::NON_EXTLOAD || | |||
14901 | LD->isVolatile()) | |||
14902 | return false; | |||
14903 | ||||
14904 | // We're looking for a sequence like this: | |||
14905 | // t13: i64,ch = load<LD8[%ref.tmp]> t0, t6, undef:i64 | |||
14906 | // t16: i64 = srl t13, Constant:i32<32> | |||
14907 | // t17: i32 = truncate t16 | |||
14908 | // t18: f32 = bitcast t17 | |||
14909 | // t19: i32 = truncate t13 | |||
14910 | // t20: f32 = bitcast t19 | |||
14911 | ||||
14912 | if (!LD->hasNUsesOfValue(2, 0)) | |||
14913 | return false; | |||
14914 | ||||
14915 | auto UI = LD->use_begin(); | |||
14916 | while (UI.getUse().getResNo() != 0) ++UI; | |||
14917 | SDNode *Trunc = *UI++; | |||
14918 | while (UI.getUse().getResNo() != 0) ++UI; | |||
14919 | SDNode *RightShift = *UI; | |||
14920 | if (Trunc->getOpcode() != ISD::TRUNCATE) | |||
14921 | std::swap(Trunc, RightShift); | |||
14922 | ||||
14923 | if (Trunc->getOpcode() != ISD::TRUNCATE || | |||
14924 | Trunc->getValueType(0) != MVT::i32 || | |||
14925 | !Trunc->hasOneUse()) | |||
14926 | return false; | |||
14927 | if (RightShift->getOpcode() != ISD::SRL || | |||
14928 | !isa<ConstantSDNode>(RightShift->getOperand(1)) || | |||
14929 | RightShift->getConstantOperandVal(1) != 32 || | |||
14930 | !RightShift->hasOneUse()) | |||
14931 | return false; | |||
14932 | ||||
14933 | SDNode *Trunc2 = *RightShift->use_begin(); | |||
14934 | if (Trunc2->getOpcode() != ISD::TRUNCATE || | |||
14935 | Trunc2->getValueType(0) != MVT::i32 || | |||
14936 | !Trunc2->hasOneUse()) | |||
14937 | return false; | |||
14938 | ||||
14939 | SDNode *Bitcast = *Trunc->use_begin(); | |||
14940 | SDNode *Bitcast2 = *Trunc2->use_begin(); | |||
14941 | ||||
14942 | if (Bitcast->getOpcode() != ISD::BITCAST || | |||
14943 | Bitcast->getValueType(0) != MVT::f32) | |||
14944 | return false; | |||
14945 | if (Bitcast2->getOpcode() != ISD::BITCAST || | |||
14946 | Bitcast2->getValueType(0) != MVT::f32) | |||
14947 | return false; | |||
14948 | ||||
14949 | if (Subtarget.isLittleEndian()) | |||
14950 | std::swap(Bitcast, Bitcast2); | |||
14951 | ||||
14952 | // Bitcast has the second float (in memory-layout order) and Bitcast2 | |||
14953 | // has the first one. | |||
14954 | ||||
14955 | SDValue BasePtr = LD->getBasePtr(); | |||
14956 | if (LD->isIndexed()) { | |||
14957 | assert(LD->getAddressingMode() == ISD::PRE_INC &&(static_cast <bool> (LD->getAddressingMode() == ISD:: PRE_INC && "Non-pre-inc AM on PPC?") ? void (0) : __assert_fail ("LD->getAddressingMode() == ISD::PRE_INC && \"Non-pre-inc AM on PPC?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14958, __extension__ __PRETTY_FUNCTION__)) | |||
14958 | "Non-pre-inc AM on PPC?")(static_cast <bool> (LD->getAddressingMode() == ISD:: PRE_INC && "Non-pre-inc AM on PPC?") ? void (0) : __assert_fail ("LD->getAddressingMode() == ISD::PRE_INC && \"Non-pre-inc AM on PPC?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 14958, __extension__ __PRETTY_FUNCTION__)); | |||
14959 | BasePtr = | |||
14960 | DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), BasePtr, | |||
14961 | LD->getOffset()); | |||
14962 | } | |||
14963 | ||||
14964 | auto MMOFlags = | |||
14965 | LD->getMemOperand()->getFlags() & ~MachineMemOperand::MOVolatile; | |||
14966 | SDValue FloatLoad = DAG.getLoad(MVT::f32, dl, LD->getChain(), BasePtr, | |||
14967 | LD->getPointerInfo(), LD->getAlignment(), | |||
14968 | MMOFlags, LD->getAAInfo()); | |||
14969 | SDValue AddPtr = | |||
14970 | DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), | |||
14971 | BasePtr, DAG.getIntPtrConstant(4, dl)); | |||
14972 | SDValue FloatLoad2 = DAG.getLoad( | |||
14973 | MVT::f32, dl, SDValue(FloatLoad.getNode(), 1), AddPtr, | |||
14974 | LD->getPointerInfo().getWithOffset(4), | |||
14975 | MinAlign(LD->getAlignment(), 4), MMOFlags, LD->getAAInfo()); | |||
14976 | ||||
14977 | if (LD->isIndexed()) { | |||
14978 | // Note that DAGCombine should re-form any pre-increment load(s) from | |||
14979 | // what is produced here if that makes sense. | |||
14980 | DAG.ReplaceAllUsesOfValueWith(SDValue(LD, 1), BasePtr); | |||
14981 | } | |||
14982 | ||||
14983 | DCI.CombineTo(Bitcast2, FloatLoad); | |||
14984 | DCI.CombineTo(Bitcast, FloatLoad2); | |||
14985 | ||||
14986 | DAG.ReplaceAllUsesOfValueWith(SDValue(LD, LD->isIndexed() ? 2 : 1), | |||
14987 | SDValue(FloatLoad2.getNode(), 1)); | |||
14988 | return true; | |||
14989 | }; | |||
14990 | ||||
14991 | if (ReplaceTwoFloatLoad()) | |||
14992 | return SDValue(N, 0); | |||
14993 | ||||
14994 | EVT MemVT = LD->getMemoryVT(); | |||
14995 | Type *Ty = MemVT.getTypeForEVT(*DAG.getContext()); | |||
14996 | Align ABIAlignment = DAG.getDataLayout().getABITypeAlign(Ty); | |||
14997 | if (LD->isUnindexed() && VT.isVector() && | |||
14998 | ((Subtarget.hasAltivec() && ISD::isNON_EXTLoad(N) && | |||
14999 | // P8 and later hardware should just use LOAD. | |||
15000 | !Subtarget.hasP8Vector() && | |||
15001 | (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32 || | |||
15002 | VT == MVT::v4f32))) && | |||
15003 | LD->getAlign() < ABIAlignment) { | |||
15004 | // This is a type-legal unaligned Altivec load. | |||
15005 | SDValue Chain = LD->getChain(); | |||
15006 | SDValue Ptr = LD->getBasePtr(); | |||
15007 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
15008 | ||||
15009 | // This implements the loading of unaligned vectors as described in | |||
15010 | // the venerable Apple Velocity Engine overview. Specifically: | |||
15011 | // https://developer.apple.com/hardwaredrivers/ve/alignment.html | |||
15012 | // https://developer.apple.com/hardwaredrivers/ve/code_optimization.html | |||
15013 | // | |||
15014 | // The general idea is to expand a sequence of one or more unaligned | |||
15015 | // loads into an alignment-based permutation-control instruction (lvsl | |||
15016 | // or lvsr), a series of regular vector loads (which always truncate | |||
15017 | // their input address to an aligned address), and a series of | |||
15018 | // permutations. The results of these permutations are the requested | |||
15019 | // loaded values. The trick is that the last "extra" load is not taken | |||
15020 | // from the address you might suspect (sizeof(vector) bytes after the | |||
15021 | // last requested load), but rather sizeof(vector) - 1 bytes after the | |||
15022 | // last requested vector. The point of this is to avoid a page fault if | |||
15023 | // the base address happened to be aligned. This works because if the | |||
15024 | // base address is aligned, then adding less than a full vector length | |||
15025 | // will cause the last vector in the sequence to be (re)loaded. | |||
15026 | // Otherwise, the next vector will be fetched as you might suspect was | |||
15027 | // necessary. | |||
15028 | ||||
15029 | // We might be able to reuse the permutation generation from | |||
15030 | // a different base address offset from this one by an aligned amount. | |||
15031 | // The INTRINSIC_WO_CHAIN DAG combine will attempt to perform this | |||
15032 | // optimization later. | |||
15033 | Intrinsic::ID Intr, IntrLD, IntrPerm; | |||
15034 | MVT PermCntlTy, PermTy, LDTy; | |||
15035 | Intr = isLittleEndian ? Intrinsic::ppc_altivec_lvsr | |||
15036 | : Intrinsic::ppc_altivec_lvsl; | |||
15037 | IntrLD = Intrinsic::ppc_altivec_lvx; | |||
15038 | IntrPerm = Intrinsic::ppc_altivec_vperm; | |||
15039 | PermCntlTy = MVT::v16i8; | |||
15040 | PermTy = MVT::v4i32; | |||
15041 | LDTy = MVT::v4i32; | |||
15042 | ||||
15043 | SDValue PermCntl = BuildIntrinsicOp(Intr, Ptr, DAG, dl, PermCntlTy); | |||
15044 | ||||
15045 | // Create the new MMO for the new base load. It is like the original MMO, | |||
15046 | // but represents an area in memory almost twice the vector size centered | |||
15047 | // on the original address. If the address is unaligned, we might start | |||
15048 | // reading up to (sizeof(vector)-1) bytes below the address of the | |||
15049 | // original unaligned load. | |||
15050 | MachineFunction &MF = DAG.getMachineFunction(); | |||
15051 | MachineMemOperand *BaseMMO = | |||
15052 | MF.getMachineMemOperand(LD->getMemOperand(), | |||
15053 | -(long)MemVT.getStoreSize()+1, | |||
15054 | 2*MemVT.getStoreSize()-1); | |||
15055 | ||||
15056 | // Create the new base load. | |||
15057 | SDValue LDXIntID = | |||
15058 | DAG.getTargetConstant(IntrLD, dl, getPointerTy(MF.getDataLayout())); | |||
15059 | SDValue BaseLoadOps[] = { Chain, LDXIntID, Ptr }; | |||
15060 | SDValue BaseLoad = | |||
15061 | DAG.getMemIntrinsicNode(ISD::INTRINSIC_W_CHAIN, dl, | |||
15062 | DAG.getVTList(PermTy, MVT::Other), | |||
15063 | BaseLoadOps, LDTy, BaseMMO); | |||
15064 | ||||
15065 | // Note that the value of IncOffset (which is provided to the next | |||
15066 | // load's pointer info offset value, and thus used to calculate the | |||
15067 | // alignment), and the value of IncValue (which is actually used to | |||
15068 | // increment the pointer value) are different! This is because we | |||
15069 | // require the next load to appear to be aligned, even though it | |||
15070 | // is actually offset from the base pointer by a lesser amount. | |||
15071 | int IncOffset = VT.getSizeInBits() / 8; | |||
15072 | int IncValue = IncOffset; | |||
15073 | ||||
15074 | // Walk (both up and down) the chain looking for another load at the real | |||
15075 | // (aligned) offset (the alignment of the other load does not matter in | |||
15076 | // this case). If found, then do not use the offset reduction trick, as | |||
15077 | // that will prevent the loads from being later combined (as they would | |||
15078 | // otherwise be duplicates). | |||
15079 | if (!findConsecutiveLoad(LD, DAG)) | |||
15080 | --IncValue; | |||
15081 | ||||
15082 | SDValue Increment = | |||
15083 | DAG.getConstant(IncValue, dl, getPointerTy(MF.getDataLayout())); | |||
15084 | Ptr = DAG.getNode(ISD::ADD, dl, Ptr.getValueType(), Ptr, Increment); | |||
15085 | ||||
15086 | MachineMemOperand *ExtraMMO = | |||
15087 | MF.getMachineMemOperand(LD->getMemOperand(), | |||
15088 | 1, 2*MemVT.getStoreSize()-1); | |||
15089 | SDValue ExtraLoadOps[] = { Chain, LDXIntID, Ptr }; | |||
15090 | SDValue ExtraLoad = | |||
15091 | DAG.getMemIntrinsicNode(ISD::INTRINSIC_W_CHAIN, dl, | |||
15092 | DAG.getVTList(PermTy, MVT::Other), | |||
15093 | ExtraLoadOps, LDTy, ExtraMMO); | |||
15094 | ||||
15095 | SDValue TF = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, | |||
15096 | BaseLoad.getValue(1), ExtraLoad.getValue(1)); | |||
15097 | ||||
15098 | // Because vperm has a big-endian bias, we must reverse the order | |||
15099 | // of the input vectors and complement the permute control vector | |||
15100 | // when generating little endian code. We have already handled the | |||
15101 | // latter by using lvsr instead of lvsl, so just reverse BaseLoad | |||
15102 | // and ExtraLoad here. | |||
15103 | SDValue Perm; | |||
15104 | if (isLittleEndian) | |||
15105 | Perm = BuildIntrinsicOp(IntrPerm, | |||
15106 | ExtraLoad, BaseLoad, PermCntl, DAG, dl); | |||
15107 | else | |||
15108 | Perm = BuildIntrinsicOp(IntrPerm, | |||
15109 | BaseLoad, ExtraLoad, PermCntl, DAG, dl); | |||
15110 | ||||
15111 | if (VT != PermTy) | |||
15112 | Perm = Subtarget.hasAltivec() | |||
15113 | ? DAG.getNode(ISD::BITCAST, dl, VT, Perm) | |||
15114 | : DAG.getNode(ISD::FP_ROUND, dl, VT, Perm, | |||
15115 | DAG.getTargetConstant(1, dl, MVT::i64)); | |||
15116 | // second argument is 1 because this rounding | |||
15117 | // is always exact. | |||
15118 | ||||
15119 | // The output of the permutation is our loaded result, the TokenFactor is | |||
15120 | // our new chain. | |||
15121 | DCI.CombineTo(N, Perm, TF); | |||
15122 | return SDValue(N, 0); | |||
15123 | } | |||
15124 | } | |||
15125 | break; | |||
15126 | case ISD::INTRINSIC_WO_CHAIN: { | |||
15127 | bool isLittleEndian = Subtarget.isLittleEndian(); | |||
15128 | unsigned IID = cast<ConstantSDNode>(N->getOperand(0))->getZExtValue(); | |||
15129 | Intrinsic::ID Intr = (isLittleEndian ? Intrinsic::ppc_altivec_lvsr | |||
15130 | : Intrinsic::ppc_altivec_lvsl); | |||
15131 | if (IID == Intr && N->getOperand(1)->getOpcode() == ISD::ADD) { | |||
15132 | SDValue Add = N->getOperand(1); | |||
15133 | ||||
15134 | int Bits = 4 /* 16 byte alignment */; | |||
15135 | ||||
15136 | if (DAG.MaskedValueIsZero(Add->getOperand(1), | |||
15137 | APInt::getAllOnesValue(Bits /* alignment */) | |||
15138 | .zext(Add.getScalarValueSizeInBits()))) { | |||
15139 | SDNode *BasePtr = Add->getOperand(0).getNode(); | |||
15140 | for (SDNode::use_iterator UI = BasePtr->use_begin(), | |||
15141 | UE = BasePtr->use_end(); | |||
15142 | UI != UE; ++UI) { | |||
15143 | if (UI->getOpcode() == ISD::INTRINSIC_WO_CHAIN && | |||
15144 | cast<ConstantSDNode>(UI->getOperand(0))->getZExtValue() == | |||
15145 | IID) { | |||
15146 | // We've found another LVSL/LVSR, and this address is an aligned | |||
15147 | // multiple of that one. The results will be the same, so use the | |||
15148 | // one we've just found instead. | |||
15149 | ||||
15150 | return SDValue(*UI, 0); | |||
15151 | } | |||
15152 | } | |||
15153 | } | |||
15154 | ||||
15155 | if (isa<ConstantSDNode>(Add->getOperand(1))) { | |||
15156 | SDNode *BasePtr = Add->getOperand(0).getNode(); | |||
15157 | for (SDNode::use_iterator UI = BasePtr->use_begin(), | |||
15158 | UE = BasePtr->use_end(); UI != UE; ++UI) { | |||
15159 | if (UI->getOpcode() == ISD::ADD && | |||
15160 | isa<ConstantSDNode>(UI->getOperand(1)) && | |||
15161 | (cast<ConstantSDNode>(Add->getOperand(1))->getZExtValue() - | |||
15162 | cast<ConstantSDNode>(UI->getOperand(1))->getZExtValue()) % | |||
15163 | (1ULL << Bits) == 0) { | |||
15164 | SDNode *OtherAdd = *UI; | |||
15165 | for (SDNode::use_iterator VI = OtherAdd->use_begin(), | |||
15166 | VE = OtherAdd->use_end(); VI != VE; ++VI) { | |||
15167 | if (VI->getOpcode() == ISD::INTRINSIC_WO_CHAIN && | |||
15168 | cast<ConstantSDNode>(VI->getOperand(0))->getZExtValue() == IID) { | |||
15169 | return SDValue(*VI, 0); | |||
15170 | } | |||
15171 | } | |||
15172 | } | |||
15173 | } | |||
15174 | } | |||
15175 | } | |||
15176 | ||||
15177 | // Combine vmaxsw/h/b(a, a's negation) to abs(a) | |||
15178 | // Expose the vabsduw/h/b opportunity for down stream | |||
15179 | if (!DCI.isAfterLegalizeDAG() && Subtarget.hasP9Altivec() && | |||
15180 | (IID == Intrinsic::ppc_altivec_vmaxsw || | |||
15181 | IID == Intrinsic::ppc_altivec_vmaxsh || | |||
15182 | IID == Intrinsic::ppc_altivec_vmaxsb)) { | |||
15183 | SDValue V1 = N->getOperand(1); | |||
15184 | SDValue V2 = N->getOperand(2); | |||
15185 | if ((V1.getSimpleValueType() == MVT::v4i32 || | |||
15186 | V1.getSimpleValueType() == MVT::v8i16 || | |||
15187 | V1.getSimpleValueType() == MVT::v16i8) && | |||
15188 | V1.getSimpleValueType() == V2.getSimpleValueType()) { | |||
15189 | // (0-a, a) | |||
15190 | if (V1.getOpcode() == ISD::SUB && | |||
15191 | ISD::isBuildVectorAllZeros(V1.getOperand(0).getNode()) && | |||
15192 | V1.getOperand(1) == V2) { | |||
15193 | return DAG.getNode(ISD::ABS, dl, V2.getValueType(), V2); | |||
15194 | } | |||
15195 | // (a, 0-a) | |||
15196 | if (V2.getOpcode() == ISD::SUB && | |||
15197 | ISD::isBuildVectorAllZeros(V2.getOperand(0).getNode()) && | |||
15198 | V2.getOperand(1) == V1) { | |||
15199 | return DAG.getNode(ISD::ABS, dl, V1.getValueType(), V1); | |||
15200 | } | |||
15201 | // (x-y, y-x) | |||
15202 | if (V1.getOpcode() == ISD::SUB && V2.getOpcode() == ISD::SUB && | |||
15203 | V1.getOperand(0) == V2.getOperand(1) && | |||
15204 | V1.getOperand(1) == V2.getOperand(0)) { | |||
15205 | return DAG.getNode(ISD::ABS, dl, V1.getValueType(), V1); | |||
15206 | } | |||
15207 | } | |||
15208 | } | |||
15209 | } | |||
15210 | ||||
15211 | break; | |||
15212 | case ISD::INTRINSIC_W_CHAIN: | |||
15213 | // For little endian, VSX loads require generating lxvd2x/xxswapd. | |||
15214 | // Not needed on ISA 3.0 based CPUs since we have a non-permuting load. | |||
15215 | if (Subtarget.needsSwapsForVSXMemOps()) { | |||
15216 | switch (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue()) { | |||
15217 | default: | |||
15218 | break; | |||
15219 | case Intrinsic::ppc_vsx_lxvw4x: | |||
15220 | case Intrinsic::ppc_vsx_lxvd2x: | |||
15221 | return expandVSXLoadForLE(N, DCI); | |||
15222 | } | |||
15223 | } | |||
15224 | break; | |||
15225 | case ISD::INTRINSIC_VOID: | |||
15226 | // For little endian, VSX stores require generating xxswapd/stxvd2x. | |||
15227 | // Not needed on ISA 3.0 based CPUs since we have a non-permuting store. | |||
15228 | if (Subtarget.needsSwapsForVSXMemOps()) { | |||
15229 | switch (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue()) { | |||
15230 | default: | |||
15231 | break; | |||
15232 | case Intrinsic::ppc_vsx_stxvw4x: | |||
15233 | case Intrinsic::ppc_vsx_stxvd2x: | |||
15234 | return expandVSXStoreForLE(N, DCI); | |||
15235 | } | |||
15236 | } | |||
15237 | break; | |||
15238 | case ISD::BSWAP: { | |||
15239 | // Turn BSWAP (LOAD) -> lhbrx/lwbrx. | |||
15240 | // For subtargets without LDBRX, we can still do better than the default | |||
15241 | // expansion even for 64-bit BSWAP (LOAD). | |||
15242 | bool Is64BitBswapOn64BitTgt = | |||
15243 | Subtarget.isPPC64() && N->getValueType(0) == MVT::i64; | |||
15244 | bool IsSingleUseNormalLd = ISD::isNormalLoad(N->getOperand(0).getNode()) && | |||
15245 | N->getOperand(0).hasOneUse(); | |||
15246 | if (IsSingleUseNormalLd && | |||
15247 | (N->getValueType(0) == MVT::i32 || N->getValueType(0) == MVT::i16 || | |||
15248 | (Subtarget.hasLDBRX() && Is64BitBswapOn64BitTgt))) { | |||
15249 | SDValue Load = N->getOperand(0); | |||
15250 | LoadSDNode *LD = cast<LoadSDNode>(Load); | |||
15251 | // Create the byte-swapping load. | |||
15252 | SDValue Ops[] = { | |||
15253 | LD->getChain(), // Chain | |||
15254 | LD->getBasePtr(), // Ptr | |||
15255 | DAG.getValueType(N->getValueType(0)) // VT | |||
15256 | }; | |||
15257 | SDValue BSLoad = | |||
15258 | DAG.getMemIntrinsicNode(PPCISD::LBRX, dl, | |||
15259 | DAG.getVTList(N->getValueType(0) == MVT::i64 ? | |||
15260 | MVT::i64 : MVT::i32, MVT::Other), | |||
15261 | Ops, LD->getMemoryVT(), LD->getMemOperand()); | |||
15262 | ||||
15263 | // If this is an i16 load, insert the truncate. | |||
15264 | SDValue ResVal = BSLoad; | |||
15265 | if (N->getValueType(0) == MVT::i16) | |||
15266 | ResVal = DAG.getNode(ISD::TRUNCATE, dl, MVT::i16, BSLoad); | |||
15267 | ||||
15268 | // First, combine the bswap away. This makes the value produced by the | |||
15269 | // load dead. | |||
15270 | DCI.CombineTo(N, ResVal); | |||
15271 | ||||
15272 | // Next, combine the load away, we give it a bogus result value but a real | |||
15273 | // chain result. The result value is dead because the bswap is dead. | |||
15274 | DCI.CombineTo(Load.getNode(), ResVal, BSLoad.getValue(1)); | |||
15275 | ||||
15276 | // Return N so it doesn't get rechecked! | |||
15277 | return SDValue(N, 0); | |||
15278 | } | |||
15279 | // Convert this to two 32-bit bswap loads and a BUILD_PAIR. Do this only | |||
15280 | // before legalization so that the BUILD_PAIR is handled correctly. | |||
15281 | if (!DCI.isBeforeLegalize() || !Is64BitBswapOn64BitTgt || | |||
15282 | !IsSingleUseNormalLd) | |||
15283 | return SDValue(); | |||
15284 | LoadSDNode *LD = cast<LoadSDNode>(N->getOperand(0)); | |||
15285 | ||||
15286 | // Can't split volatile or atomic loads. | |||
15287 | if (!LD->isSimple()) | |||
15288 | return SDValue(); | |||
15289 | SDValue BasePtr = LD->getBasePtr(); | |||
15290 | SDValue Lo = DAG.getLoad(MVT::i32, dl, LD->getChain(), BasePtr, | |||
15291 | LD->getPointerInfo(), LD->getAlignment()); | |||
15292 | Lo = DAG.getNode(ISD::BSWAP, dl, MVT::i32, Lo); | |||
15293 | BasePtr = DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), BasePtr, | |||
15294 | DAG.getIntPtrConstant(4, dl)); | |||
15295 | MachineMemOperand *NewMMO = DAG.getMachineFunction().getMachineMemOperand( | |||
15296 | LD->getMemOperand(), 4, 4); | |||
15297 | SDValue Hi = DAG.getLoad(MVT::i32, dl, LD->getChain(), BasePtr, NewMMO); | |||
15298 | Hi = DAG.getNode(ISD::BSWAP, dl, MVT::i32, Hi); | |||
15299 | SDValue Res; | |||
15300 | if (Subtarget.isLittleEndian()) | |||
15301 | Res = DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Hi, Lo); | |||
15302 | else | |||
15303 | Res = DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi); | |||
15304 | SDValue TF = | |||
15305 | DAG.getNode(ISD::TokenFactor, dl, MVT::Other, | |||
15306 | Hi.getOperand(0).getValue(1), Lo.getOperand(0).getValue(1)); | |||
15307 | DAG.ReplaceAllUsesOfValueWith(SDValue(LD, 1), TF); | |||
15308 | return Res; | |||
15309 | } | |||
15310 | case PPCISD::VCMP: | |||
15311 | // If a VCMP_rec node already exists with exactly the same operands as this | |||
15312 | // node, use its result instead of this node (VCMP_rec computes both a CR6 | |||
15313 | // and a normal output). | |||
15314 | // | |||
15315 | if (!N->getOperand(0).hasOneUse() && | |||
15316 | !N->getOperand(1).hasOneUse() && | |||
15317 | !N->getOperand(2).hasOneUse()) { | |||
15318 | ||||
15319 | // Scan all of the users of the LHS, looking for VCMP_rec's that match. | |||
15320 | SDNode *VCMPrecNode = nullptr; | |||
15321 | ||||
15322 | SDNode *LHSN = N->getOperand(0).getNode(); | |||
15323 | for (SDNode::use_iterator UI = LHSN->use_begin(), E = LHSN->use_end(); | |||
15324 | UI != E; ++UI) | |||
15325 | if (UI->getOpcode() == PPCISD::VCMP_rec && | |||
15326 | UI->getOperand(1) == N->getOperand(1) && | |||
15327 | UI->getOperand(2) == N->getOperand(2) && | |||
15328 | UI->getOperand(0) == N->getOperand(0)) { | |||
15329 | VCMPrecNode = *UI; | |||
15330 | break; | |||
15331 | } | |||
15332 | ||||
15333 | // If there is no VCMP_rec node, or if the flag value has a single use, | |||
15334 | // don't transform this. | |||
15335 | if (!VCMPrecNode || VCMPrecNode->hasNUsesOfValue(0, 1)) | |||
15336 | break; | |||
15337 | ||||
15338 | // Look at the (necessarily single) use of the flag value. If it has a | |||
15339 | // chain, this transformation is more complex. Note that multiple things | |||
15340 | // could use the value result, which we should ignore. | |||
15341 | SDNode *FlagUser = nullptr; | |||
15342 | for (SDNode::use_iterator UI = VCMPrecNode->use_begin(); | |||
15343 | FlagUser == nullptr; ++UI) { | |||
15344 | assert(UI != VCMPrecNode->use_end() && "Didn't find user!")(static_cast <bool> (UI != VCMPrecNode->use_end() && "Didn't find user!") ? void (0) : __assert_fail ("UI != VCMPrecNode->use_end() && \"Didn't find user!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15344, __extension__ __PRETTY_FUNCTION__)); | |||
15345 | SDNode *User = *UI; | |||
15346 | for (unsigned i = 0, e = User->getNumOperands(); i != e; ++i) { | |||
15347 | if (User->getOperand(i) == SDValue(VCMPrecNode, 1)) { | |||
15348 | FlagUser = User; | |||
15349 | break; | |||
15350 | } | |||
15351 | } | |||
15352 | } | |||
15353 | ||||
15354 | // If the user is a MFOCRF instruction, we know this is safe. | |||
15355 | // Otherwise we give up for right now. | |||
15356 | if (FlagUser->getOpcode() == PPCISD::MFOCRF) | |||
15357 | return SDValue(VCMPrecNode, 0); | |||
15358 | } | |||
15359 | break; | |||
15360 | case ISD::BRCOND: { | |||
15361 | SDValue Cond = N->getOperand(1); | |||
15362 | SDValue Target = N->getOperand(2); | |||
15363 | ||||
15364 | if (Cond.getOpcode() == ISD::INTRINSIC_W_CHAIN && | |||
15365 | cast<ConstantSDNode>(Cond.getOperand(1))->getZExtValue() == | |||
15366 | Intrinsic::loop_decrement) { | |||
15367 | ||||
15368 | // We now need to make the intrinsic dead (it cannot be instruction | |||
15369 | // selected). | |||
15370 | DAG.ReplaceAllUsesOfValueWith(Cond.getValue(1), Cond.getOperand(0)); | |||
15371 | assert(Cond.getNode()->hasOneUse() &&(static_cast <bool> (Cond.getNode()->hasOneUse() && "Counter decrement has more than one use") ? void (0) : __assert_fail ("Cond.getNode()->hasOneUse() && \"Counter decrement has more than one use\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15372, __extension__ __PRETTY_FUNCTION__)) | |||
15372 | "Counter decrement has more than one use")(static_cast <bool> (Cond.getNode()->hasOneUse() && "Counter decrement has more than one use") ? void (0) : __assert_fail ("Cond.getNode()->hasOneUse() && \"Counter decrement has more than one use\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15372, __extension__ __PRETTY_FUNCTION__)); | |||
15373 | ||||
15374 | return DAG.getNode(PPCISD::BDNZ, dl, MVT::Other, | |||
15375 | N->getOperand(0), Target); | |||
15376 | } | |||
15377 | } | |||
15378 | break; | |||
15379 | case ISD::BR_CC: { | |||
15380 | // If this is a branch on an altivec predicate comparison, lower this so | |||
15381 | // that we don't have to do a MFOCRF: instead, branch directly on CR6. This | |||
15382 | // lowering is done pre-legalize, because the legalizer lowers the predicate | |||
15383 | // compare down to code that is difficult to reassemble. | |||
15384 | ISD::CondCode CC = cast<CondCodeSDNode>(N->getOperand(1))->get(); | |||
15385 | SDValue LHS = N->getOperand(2), RHS = N->getOperand(3); | |||
15386 | ||||
15387 | // Sometimes the promoted value of the intrinsic is ANDed by some non-zero | |||
15388 | // value. If so, pass-through the AND to get to the intrinsic. | |||
15389 | if (LHS.getOpcode() == ISD::AND && | |||
15390 | LHS.getOperand(0).getOpcode() == ISD::INTRINSIC_W_CHAIN && | |||
15391 | cast<ConstantSDNode>(LHS.getOperand(0).getOperand(1))->getZExtValue() == | |||
15392 | Intrinsic::loop_decrement && | |||
15393 | isa<ConstantSDNode>(LHS.getOperand(1)) && | |||
15394 | !isNullConstant(LHS.getOperand(1))) | |||
15395 | LHS = LHS.getOperand(0); | |||
15396 | ||||
15397 | if (LHS.getOpcode() == ISD::INTRINSIC_W_CHAIN && | |||
15398 | cast<ConstantSDNode>(LHS.getOperand(1))->getZExtValue() == | |||
15399 | Intrinsic::loop_decrement && | |||
15400 | isa<ConstantSDNode>(RHS)) { | |||
15401 | assert((CC == ISD::SETEQ || CC == ISD::SETNE) &&(static_cast <bool> ((CC == ISD::SETEQ || CC == ISD::SETNE ) && "Counter decrement comparison is not EQ or NE") ? void (0) : __assert_fail ("(CC == ISD::SETEQ || CC == ISD::SETNE) && \"Counter decrement comparison is not EQ or NE\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15402, __extension__ __PRETTY_FUNCTION__)) | |||
15402 | "Counter decrement comparison is not EQ or NE")(static_cast <bool> ((CC == ISD::SETEQ || CC == ISD::SETNE ) && "Counter decrement comparison is not EQ or NE") ? void (0) : __assert_fail ("(CC == ISD::SETEQ || CC == ISD::SETNE) && \"Counter decrement comparison is not EQ or NE\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15402, __extension__ __PRETTY_FUNCTION__)); | |||
15403 | ||||
15404 | unsigned Val = cast<ConstantSDNode>(RHS)->getZExtValue(); | |||
15405 | bool isBDNZ = (CC == ISD::SETEQ && Val) || | |||
15406 | (CC == ISD::SETNE && !Val); | |||
15407 | ||||
15408 | // We now need to make the intrinsic dead (it cannot be instruction | |||
15409 | // selected). | |||
15410 | DAG.ReplaceAllUsesOfValueWith(LHS.getValue(1), LHS.getOperand(0)); | |||
15411 | assert(LHS.getNode()->hasOneUse() &&(static_cast <bool> (LHS.getNode()->hasOneUse() && "Counter decrement has more than one use") ? void (0) : __assert_fail ("LHS.getNode()->hasOneUse() && \"Counter decrement has more than one use\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15412, __extension__ __PRETTY_FUNCTION__)) | |||
15412 | "Counter decrement has more than one use")(static_cast <bool> (LHS.getNode()->hasOneUse() && "Counter decrement has more than one use") ? void (0) : __assert_fail ("LHS.getNode()->hasOneUse() && \"Counter decrement has more than one use\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15412, __extension__ __PRETTY_FUNCTION__)); | |||
15413 | ||||
15414 | return DAG.getNode(isBDNZ ? PPCISD::BDNZ : PPCISD::BDZ, dl, MVT::Other, | |||
15415 | N->getOperand(0), N->getOperand(4)); | |||
15416 | } | |||
15417 | ||||
15418 | int CompareOpc; | |||
15419 | bool isDot; | |||
15420 | ||||
15421 | if (LHS.getOpcode() == ISD::INTRINSIC_WO_CHAIN && | |||
15422 | isa<ConstantSDNode>(RHS) && (CC == ISD::SETEQ || CC == ISD::SETNE) && | |||
15423 | getVectorCompareInfo(LHS, CompareOpc, isDot, Subtarget)) { | |||
15424 | assert(isDot && "Can't compare against a vector result!")(static_cast <bool> (isDot && "Can't compare against a vector result!" ) ? void (0) : __assert_fail ("isDot && \"Can't compare against a vector result!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15424, __extension__ __PRETTY_FUNCTION__)); | |||
15425 | ||||
15426 | // If this is a comparison against something other than 0/1, then we know | |||
15427 | // that the condition is never/always true. | |||
15428 | unsigned Val = cast<ConstantSDNode>(RHS)->getZExtValue(); | |||
15429 | if (Val != 0 && Val != 1) { | |||
15430 | if (CC == ISD::SETEQ) // Cond never true, remove branch. | |||
15431 | return N->getOperand(0); | |||
15432 | // Always !=, turn it into an unconditional branch. | |||
15433 | return DAG.getNode(ISD::BR, dl, MVT::Other, | |||
15434 | N->getOperand(0), N->getOperand(4)); | |||
15435 | } | |||
15436 | ||||
15437 | bool BranchOnWhenPredTrue = (CC == ISD::SETEQ) ^ (Val == 0); | |||
15438 | ||||
15439 | // Create the PPCISD altivec 'dot' comparison node. | |||
15440 | SDValue Ops[] = { | |||
15441 | LHS.getOperand(2), // LHS of compare | |||
15442 | LHS.getOperand(3), // RHS of compare | |||
15443 | DAG.getConstant(CompareOpc, dl, MVT::i32) | |||
15444 | }; | |||
15445 | EVT VTs[] = { LHS.getOperand(2).getValueType(), MVT::Glue }; | |||
15446 | SDValue CompNode = DAG.getNode(PPCISD::VCMP_rec, dl, VTs, Ops); | |||
15447 | ||||
15448 | // Unpack the result based on how the target uses it. | |||
15449 | PPC::Predicate CompOpc; | |||
15450 | switch (cast<ConstantSDNode>(LHS.getOperand(1))->getZExtValue()) { | |||
15451 | default: // Can't happen, don't crash on invalid number though. | |||
15452 | case 0: // Branch on the value of the EQ bit of CR6. | |||
15453 | CompOpc = BranchOnWhenPredTrue ? PPC::PRED_EQ : PPC::PRED_NE; | |||
15454 | break; | |||
15455 | case 1: // Branch on the inverted value of the EQ bit of CR6. | |||
15456 | CompOpc = BranchOnWhenPredTrue ? PPC::PRED_NE : PPC::PRED_EQ; | |||
15457 | break; | |||
15458 | case 2: // Branch on the value of the LT bit of CR6. | |||
15459 | CompOpc = BranchOnWhenPredTrue ? PPC::PRED_LT : PPC::PRED_GE; | |||
15460 | break; | |||
15461 | case 3: // Branch on the inverted value of the LT bit of CR6. | |||
15462 | CompOpc = BranchOnWhenPredTrue ? PPC::PRED_GE : PPC::PRED_LT; | |||
15463 | break; | |||
15464 | } | |||
15465 | ||||
15466 | return DAG.getNode(PPCISD::COND_BRANCH, dl, MVT::Other, N->getOperand(0), | |||
15467 | DAG.getConstant(CompOpc, dl, MVT::i32), | |||
15468 | DAG.getRegister(PPC::CR6, MVT::i32), | |||
15469 | N->getOperand(4), CompNode.getValue(1)); | |||
15470 | } | |||
15471 | break; | |||
15472 | } | |||
15473 | case ISD::BUILD_VECTOR: | |||
15474 | return DAGCombineBuildVector(N, DCI); | |||
15475 | case ISD::ABS: | |||
15476 | return combineABS(N, DCI); | |||
15477 | case ISD::VSELECT: | |||
15478 | return combineVSelect(N, DCI); | |||
15479 | } | |||
15480 | ||||
15481 | return SDValue(); | |||
15482 | } | |||
15483 | ||||
15484 | SDValue | |||
15485 | PPCTargetLowering::BuildSDIVPow2(SDNode *N, const APInt &Divisor, | |||
15486 | SelectionDAG &DAG, | |||
15487 | SmallVectorImpl<SDNode *> &Created) const { | |||
15488 | // fold (sdiv X, pow2) | |||
15489 | EVT VT = N->getValueType(0); | |||
15490 | if (VT == MVT::i64 && !Subtarget.isPPC64()) | |||
15491 | return SDValue(); | |||
15492 | if ((VT != MVT::i32 && VT != MVT::i64) || | |||
15493 | !(Divisor.isPowerOf2() || (-Divisor).isPowerOf2())) | |||
15494 | return SDValue(); | |||
15495 | ||||
15496 | SDLoc DL(N); | |||
15497 | SDValue N0 = N->getOperand(0); | |||
15498 | ||||
15499 | bool IsNegPow2 = (-Divisor).isPowerOf2(); | |||
15500 | unsigned Lg2 = (IsNegPow2 ? -Divisor : Divisor).countTrailingZeros(); | |||
15501 | SDValue ShiftAmt = DAG.getConstant(Lg2, DL, VT); | |||
15502 | ||||
15503 | SDValue Op = DAG.getNode(PPCISD::SRA_ADDZE, DL, VT, N0, ShiftAmt); | |||
15504 | Created.push_back(Op.getNode()); | |||
15505 | ||||
15506 | if (IsNegPow2) { | |||
15507 | Op = DAG.getNode(ISD::SUB, DL, VT, DAG.getConstant(0, DL, VT), Op); | |||
15508 | Created.push_back(Op.getNode()); | |||
15509 | } | |||
15510 | ||||
15511 | return Op; | |||
15512 | } | |||
15513 | ||||
15514 | //===----------------------------------------------------------------------===// | |||
15515 | // Inline Assembly Support | |||
15516 | //===----------------------------------------------------------------------===// | |||
15517 | ||||
15518 | void PPCTargetLowering::computeKnownBitsForTargetNode(const SDValue Op, | |||
15519 | KnownBits &Known, | |||
15520 | const APInt &DemandedElts, | |||
15521 | const SelectionDAG &DAG, | |||
15522 | unsigned Depth) const { | |||
15523 | Known.resetAll(); | |||
15524 | switch (Op.getOpcode()) { | |||
15525 | default: break; | |||
15526 | case PPCISD::LBRX: { | |||
15527 | // lhbrx is known to have the top bits cleared out. | |||
15528 | if (cast<VTSDNode>(Op.getOperand(2))->getVT() == MVT::i16) | |||
15529 | Known.Zero = 0xFFFF0000; | |||
15530 | break; | |||
15531 | } | |||
15532 | case ISD::INTRINSIC_WO_CHAIN: { | |||
15533 | switch (cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue()) { | |||
15534 | default: break; | |||
15535 | case Intrinsic::ppc_altivec_vcmpbfp_p: | |||
15536 | case Intrinsic::ppc_altivec_vcmpeqfp_p: | |||
15537 | case Intrinsic::ppc_altivec_vcmpequb_p: | |||
15538 | case Intrinsic::ppc_altivec_vcmpequh_p: | |||
15539 | case Intrinsic::ppc_altivec_vcmpequw_p: | |||
15540 | case Intrinsic::ppc_altivec_vcmpequd_p: | |||
15541 | case Intrinsic::ppc_altivec_vcmpequq_p: | |||
15542 | case Intrinsic::ppc_altivec_vcmpgefp_p: | |||
15543 | case Intrinsic::ppc_altivec_vcmpgtfp_p: | |||
15544 | case Intrinsic::ppc_altivec_vcmpgtsb_p: | |||
15545 | case Intrinsic::ppc_altivec_vcmpgtsh_p: | |||
15546 | case Intrinsic::ppc_altivec_vcmpgtsw_p: | |||
15547 | case Intrinsic::ppc_altivec_vcmpgtsd_p: | |||
15548 | case Intrinsic::ppc_altivec_vcmpgtsq_p: | |||
15549 | case Intrinsic::ppc_altivec_vcmpgtub_p: | |||
15550 | case Intrinsic::ppc_altivec_vcmpgtuh_p: | |||
15551 | case Intrinsic::ppc_altivec_vcmpgtuw_p: | |||
15552 | case Intrinsic::ppc_altivec_vcmpgtud_p: | |||
15553 | case Intrinsic::ppc_altivec_vcmpgtuq_p: | |||
15554 | Known.Zero = ~1U; // All bits but the low one are known to be zero. | |||
15555 | break; | |||
15556 | } | |||
15557 | } | |||
15558 | } | |||
15559 | } | |||
15560 | ||||
15561 | Align PPCTargetLowering::getPrefLoopAlignment(MachineLoop *ML) const { | |||
15562 | switch (Subtarget.getCPUDirective()) { | |||
15563 | default: break; | |||
15564 | case PPC::DIR_970: | |||
15565 | case PPC::DIR_PWR4: | |||
15566 | case PPC::DIR_PWR5: | |||
15567 | case PPC::DIR_PWR5X: | |||
15568 | case PPC::DIR_PWR6: | |||
15569 | case PPC::DIR_PWR6X: | |||
15570 | case PPC::DIR_PWR7: | |||
15571 | case PPC::DIR_PWR8: | |||
15572 | case PPC::DIR_PWR9: | |||
15573 | case PPC::DIR_PWR10: | |||
15574 | case PPC::DIR_PWR_FUTURE: { | |||
15575 | if (!ML) | |||
15576 | break; | |||
15577 | ||||
15578 | if (!DisableInnermostLoopAlign32) { | |||
15579 | // If the nested loop is an innermost loop, prefer to a 32-byte alignment, | |||
15580 | // so that we can decrease cache misses and branch-prediction misses. | |||
15581 | // Actual alignment of the loop will depend on the hotness check and other | |||
15582 | // logic in alignBlocks. | |||
15583 | if (ML->getLoopDepth() > 1 && ML->getSubLoops().empty()) | |||
15584 | return Align(32); | |||
15585 | } | |||
15586 | ||||
15587 | const PPCInstrInfo *TII = Subtarget.getInstrInfo(); | |||
15588 | ||||
15589 | // For small loops (between 5 and 8 instructions), align to a 32-byte | |||
15590 | // boundary so that the entire loop fits in one instruction-cache line. | |||
15591 | uint64_t LoopSize = 0; | |||
15592 | for (auto I = ML->block_begin(), IE = ML->block_end(); I != IE; ++I) | |||
15593 | for (auto J = (*I)->begin(), JE = (*I)->end(); J != JE; ++J) { | |||
15594 | LoopSize += TII->getInstSizeInBytes(*J); | |||
15595 | if (LoopSize > 32) | |||
15596 | break; | |||
15597 | } | |||
15598 | ||||
15599 | if (LoopSize > 16 && LoopSize <= 32) | |||
15600 | return Align(32); | |||
15601 | ||||
15602 | break; | |||
15603 | } | |||
15604 | } | |||
15605 | ||||
15606 | return TargetLowering::getPrefLoopAlignment(ML); | |||
15607 | } | |||
15608 | ||||
15609 | /// getConstraintType - Given a constraint, return the type of | |||
15610 | /// constraint it is for this target. | |||
15611 | PPCTargetLowering::ConstraintType | |||
15612 | PPCTargetLowering::getConstraintType(StringRef Constraint) const { | |||
15613 | if (Constraint.size() == 1) { | |||
15614 | switch (Constraint[0]) { | |||
15615 | default: break; | |||
15616 | case 'b': | |||
15617 | case 'r': | |||
15618 | case 'f': | |||
15619 | case 'd': | |||
15620 | case 'v': | |||
15621 | case 'y': | |||
15622 | return C_RegisterClass; | |||
15623 | case 'Z': | |||
15624 | // FIXME: While Z does indicate a memory constraint, it specifically | |||
15625 | // indicates an r+r address (used in conjunction with the 'y' modifier | |||
15626 | // in the replacement string). Currently, we're forcing the base | |||
15627 | // register to be r0 in the asm printer (which is interpreted as zero) | |||
15628 | // and forming the complete address in the second register. This is | |||
15629 | // suboptimal. | |||
15630 | return C_Memory; | |||
15631 | } | |||
15632 | } else if (Constraint == "wc") { // individual CR bits. | |||
15633 | return C_RegisterClass; | |||
15634 | } else if (Constraint == "wa" || Constraint == "wd" || | |||
15635 | Constraint == "wf" || Constraint == "ws" || | |||
15636 | Constraint == "wi" || Constraint == "ww") { | |||
15637 | return C_RegisterClass; // VSX registers. | |||
15638 | } | |||
15639 | return TargetLowering::getConstraintType(Constraint); | |||
15640 | } | |||
15641 | ||||
15642 | /// Examine constraint type and operand type and determine a weight value. | |||
15643 | /// This object must already have been set up with the operand type | |||
15644 | /// and the current alternative constraint selected. | |||
15645 | TargetLowering::ConstraintWeight | |||
15646 | PPCTargetLowering::getSingleConstraintMatchWeight( | |||
15647 | AsmOperandInfo &info, const char *constraint) const { | |||
15648 | ConstraintWeight weight = CW_Invalid; | |||
15649 | Value *CallOperandVal = info.CallOperandVal; | |||
15650 | // If we don't have a value, we can't do a match, | |||
15651 | // but allow it at the lowest weight. | |||
15652 | if (!CallOperandVal) | |||
15653 | return CW_Default; | |||
15654 | Type *type = CallOperandVal->getType(); | |||
15655 | ||||
15656 | // Look at the constraint type. | |||
15657 | if (StringRef(constraint) == "wc" && type->isIntegerTy(1)) | |||
15658 | return CW_Register; // an individual CR bit. | |||
15659 | else if ((StringRef(constraint) == "wa" || | |||
15660 | StringRef(constraint) == "wd" || | |||
15661 | StringRef(constraint) == "wf") && | |||
15662 | type->isVectorTy()) | |||
15663 | return CW_Register; | |||
15664 | else if (StringRef(constraint) == "wi" && type->isIntegerTy(64)) | |||
15665 | return CW_Register; // just hold 64-bit integers data. | |||
15666 | else if (StringRef(constraint) == "ws" && type->isDoubleTy()) | |||
15667 | return CW_Register; | |||
15668 | else if (StringRef(constraint) == "ww" && type->isFloatTy()) | |||
15669 | return CW_Register; | |||
15670 | ||||
15671 | switch (*constraint) { | |||
15672 | default: | |||
15673 | weight = TargetLowering::getSingleConstraintMatchWeight(info, constraint); | |||
15674 | break; | |||
15675 | case 'b': | |||
15676 | if (type->isIntegerTy()) | |||
15677 | weight = CW_Register; | |||
15678 | break; | |||
15679 | case 'f': | |||
15680 | if (type->isFloatTy()) | |||
15681 | weight = CW_Register; | |||
15682 | break; | |||
15683 | case 'd': | |||
15684 | if (type->isDoubleTy()) | |||
15685 | weight = CW_Register; | |||
15686 | break; | |||
15687 | case 'v': | |||
15688 | if (type->isVectorTy()) | |||
15689 | weight = CW_Register; | |||
15690 | break; | |||
15691 | case 'y': | |||
15692 | weight = CW_Register; | |||
15693 | break; | |||
15694 | case 'Z': | |||
15695 | weight = CW_Memory; | |||
15696 | break; | |||
15697 | } | |||
15698 | return weight; | |||
15699 | } | |||
15700 | ||||
15701 | std::pair<unsigned, const TargetRegisterClass *> | |||
15702 | PPCTargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI, | |||
15703 | StringRef Constraint, | |||
15704 | MVT VT) const { | |||
15705 | if (Constraint.size() == 1) { | |||
15706 | // GCC RS6000 Constraint Letters | |||
15707 | switch (Constraint[0]) { | |||
15708 | case 'b': // R1-R31 | |||
15709 | if (VT == MVT::i64 && Subtarget.isPPC64()) | |||
15710 | return std::make_pair(0U, &PPC::G8RC_NOX0RegClass); | |||
15711 | return std::make_pair(0U, &PPC::GPRC_NOR0RegClass); | |||
15712 | case 'r': // R0-R31 | |||
15713 | if (VT == MVT::i64 && Subtarget.isPPC64()) | |||
15714 | return std::make_pair(0U, &PPC::G8RCRegClass); | |||
15715 | return std::make_pair(0U, &PPC::GPRCRegClass); | |||
15716 | // 'd' and 'f' constraints are both defined to be "the floating point | |||
15717 | // registers", where one is for 32-bit and the other for 64-bit. We don't | |||
15718 | // really care overly much here so just give them all the same reg classes. | |||
15719 | case 'd': | |||
15720 | case 'f': | |||
15721 | if (Subtarget.hasSPE()) { | |||
15722 | if (VT == MVT::f32 || VT == MVT::i32) | |||
15723 | return std::make_pair(0U, &PPC::GPRCRegClass); | |||
15724 | if (VT == MVT::f64 || VT == MVT::i64) | |||
15725 | return std::make_pair(0U, &PPC::SPERCRegClass); | |||
15726 | } else { | |||
15727 | if (VT == MVT::f32 || VT == MVT::i32) | |||
15728 | return std::make_pair(0U, &PPC::F4RCRegClass); | |||
15729 | if (VT == MVT::f64 || VT == MVT::i64) | |||
15730 | return std::make_pair(0U, &PPC::F8RCRegClass); | |||
15731 | } | |||
15732 | break; | |||
15733 | case 'v': | |||
15734 | if (Subtarget.hasAltivec()) | |||
15735 | return std::make_pair(0U, &PPC::VRRCRegClass); | |||
15736 | break; | |||
15737 | case 'y': // crrc | |||
15738 | return std::make_pair(0U, &PPC::CRRCRegClass); | |||
15739 | } | |||
15740 | } else if (Constraint == "wc" && Subtarget.useCRBits()) { | |||
15741 | // An individual CR bit. | |||
15742 | return std::make_pair(0U, &PPC::CRBITRCRegClass); | |||
15743 | } else if ((Constraint == "wa" || Constraint == "wd" || | |||
15744 | Constraint == "wf" || Constraint == "wi") && | |||
15745 | Subtarget.hasVSX()) { | |||
15746 | // A VSX register for either a scalar (FP) or vector. There is no | |||
15747 | // support for single precision scalars on subtargets prior to Power8. | |||
15748 | if (VT.isVector()) | |||
15749 | return std::make_pair(0U, &PPC::VSRCRegClass); | |||
15750 | if (VT == MVT::f32 && Subtarget.hasP8Vector()) | |||
15751 | return std::make_pair(0U, &PPC::VSSRCRegClass); | |||
15752 | return std::make_pair(0U, &PPC::VSFRCRegClass); | |||
15753 | } else if ((Constraint == "ws" || Constraint == "ww") && Subtarget.hasVSX()) { | |||
15754 | if (VT == MVT::f32 && Subtarget.hasP8Vector()) | |||
15755 | return std::make_pair(0U, &PPC::VSSRCRegClass); | |||
15756 | else | |||
15757 | return std::make_pair(0U, &PPC::VSFRCRegClass); | |||
15758 | } else if (Constraint == "lr") { | |||
15759 | if (VT == MVT::i64) | |||
15760 | return std::make_pair(0U, &PPC::LR8RCRegClass); | |||
15761 | else | |||
15762 | return std::make_pair(0U, &PPC::LRRCRegClass); | |||
15763 | } | |||
15764 | ||||
15765 | // Handle special cases of physical registers that are not properly handled | |||
15766 | // by the base class. | |||
15767 | if (Constraint[0] == '{' && Constraint[Constraint.size() - 1] == '}') { | |||
15768 | // If we name a VSX register, we can't defer to the base class because it | |||
15769 | // will not recognize the correct register (their names will be VSL{0-31} | |||
15770 | // and V{0-31} so they won't match). So we match them here. | |||
15771 | if (Constraint.size() > 3 && Constraint[1] == 'v' && Constraint[2] == 's') { | |||
15772 | int VSNum = atoi(Constraint.data() + 3); | |||
15773 | assert(VSNum >= 0 && VSNum <= 63 &&(static_cast <bool> (VSNum >= 0 && VSNum <= 63 && "Attempted to access a vsr out of range") ? void (0) : __assert_fail ("VSNum >= 0 && VSNum <= 63 && \"Attempted to access a vsr out of range\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15774, __extension__ __PRETTY_FUNCTION__)) | |||
15774 | "Attempted to access a vsr out of range")(static_cast <bool> (VSNum >= 0 && VSNum <= 63 && "Attempted to access a vsr out of range") ? void (0) : __assert_fail ("VSNum >= 0 && VSNum <= 63 && \"Attempted to access a vsr out of range\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15774, __extension__ __PRETTY_FUNCTION__)); | |||
15775 | if (VSNum < 32) | |||
15776 | return std::make_pair(PPC::VSL0 + VSNum, &PPC::VSRCRegClass); | |||
15777 | return std::make_pair(PPC::V0 + VSNum - 32, &PPC::VSRCRegClass); | |||
15778 | } | |||
15779 | ||||
15780 | // For float registers, we can't defer to the base class as it will match | |||
15781 | // the SPILLTOVSRRC class. | |||
15782 | if (Constraint.size() > 3 && Constraint[1] == 'f') { | |||
15783 | int RegNum = atoi(Constraint.data() + 2); | |||
15784 | if (RegNum > 31 || RegNum < 0) | |||
15785 | report_fatal_error("Invalid floating point register number"); | |||
15786 | if (VT == MVT::f32 || VT == MVT::i32) | |||
15787 | return Subtarget.hasSPE() | |||
15788 | ? std::make_pair(PPC::R0 + RegNum, &PPC::GPRCRegClass) | |||
15789 | : std::make_pair(PPC::F0 + RegNum, &PPC::F4RCRegClass); | |||
15790 | if (VT == MVT::f64 || VT == MVT::i64) | |||
15791 | return Subtarget.hasSPE() | |||
15792 | ? std::make_pair(PPC::S0 + RegNum, &PPC::SPERCRegClass) | |||
15793 | : std::make_pair(PPC::F0 + RegNum, &PPC::F8RCRegClass); | |||
15794 | } | |||
15795 | } | |||
15796 | ||||
15797 | std::pair<unsigned, const TargetRegisterClass *> R = | |||
15798 | TargetLowering::getRegForInlineAsmConstraint(TRI, Constraint, VT); | |||
15799 | ||||
15800 | // r[0-9]+ are used, on PPC64, to refer to the corresponding 64-bit registers | |||
15801 | // (which we call X[0-9]+). If a 64-bit value has been requested, and a | |||
15802 | // 32-bit GPR has been selected, then 'upgrade' it to the 64-bit parent | |||
15803 | // register. | |||
15804 | // FIXME: If TargetLowering::getRegForInlineAsmConstraint could somehow use | |||
15805 | // the AsmName field from *RegisterInfo.td, then this would not be necessary. | |||
15806 | if (R.first && VT == MVT::i64 && Subtarget.isPPC64() && | |||
15807 | PPC::GPRCRegClass.contains(R.first)) | |||
15808 | return std::make_pair(TRI->getMatchingSuperReg(R.first, | |||
15809 | PPC::sub_32, &PPC::G8RCRegClass), | |||
15810 | &PPC::G8RCRegClass); | |||
15811 | ||||
15812 | // GCC accepts 'cc' as an alias for 'cr0', and we need to do the same. | |||
15813 | if (!R.second && StringRef("{cc}").equals_insensitive(Constraint)) { | |||
15814 | R.first = PPC::CR0; | |||
15815 | R.second = &PPC::CRRCRegClass; | |||
15816 | } | |||
15817 | // FIXME: This warning should ideally be emitted in the front end. | |||
15818 | const auto &TM = getTargetMachine(); | |||
15819 | if (Subtarget.isAIXABI() && !TM.getAIXExtendedAltivecABI()) { | |||
15820 | if (((R.first >= PPC::V20 && R.first <= PPC::V31) || | |||
15821 | (R.first >= PPC::VF20 && R.first <= PPC::VF31)) && | |||
15822 | (R.second == &PPC::VSRCRegClass || R.second == &PPC::VSFRCRegClass)) | |||
15823 | errs() << "warning: vector registers 20 to 32 are reserved in the " | |||
15824 | "default AIX AltiVec ABI and cannot be used\n"; | |||
15825 | } | |||
15826 | ||||
15827 | return R; | |||
15828 | } | |||
15829 | ||||
15830 | /// LowerAsmOperandForConstraint - Lower the specified operand into the Ops | |||
15831 | /// vector. If it is invalid, don't add anything to Ops. | |||
15832 | void PPCTargetLowering::LowerAsmOperandForConstraint(SDValue Op, | |||
15833 | std::string &Constraint, | |||
15834 | std::vector<SDValue>&Ops, | |||
15835 | SelectionDAG &DAG) const { | |||
15836 | SDValue Result; | |||
15837 | ||||
15838 | // Only support length 1 constraints. | |||
15839 | if (Constraint.length() > 1) return; | |||
15840 | ||||
15841 | char Letter = Constraint[0]; | |||
15842 | switch (Letter) { | |||
15843 | default: break; | |||
15844 | case 'I': | |||
15845 | case 'J': | |||
15846 | case 'K': | |||
15847 | case 'L': | |||
15848 | case 'M': | |||
15849 | case 'N': | |||
15850 | case 'O': | |||
15851 | case 'P': { | |||
15852 | ConstantSDNode *CST = dyn_cast<ConstantSDNode>(Op); | |||
15853 | if (!CST) return; // Must be an immediate to match. | |||
15854 | SDLoc dl(Op); | |||
15855 | int64_t Value = CST->getSExtValue(); | |||
15856 | EVT TCVT = MVT::i64; // All constants taken to be 64 bits so that negative | |||
15857 | // numbers are printed as such. | |||
15858 | switch (Letter) { | |||
15859 | default: llvm_unreachable("Unknown constraint letter!")::llvm::llvm_unreachable_internal("Unknown constraint letter!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 15859); | |||
15860 | case 'I': // "I" is a signed 16-bit constant. | |||
15861 | if (isInt<16>(Value)) | |||
15862 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15863 | break; | |||
15864 | case 'J': // "J" is a constant with only the high-order 16 bits nonzero. | |||
15865 | if (isShiftedUInt<16, 16>(Value)) | |||
15866 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15867 | break; | |||
15868 | case 'L': // "L" is a signed 16-bit constant shifted left 16 bits. | |||
15869 | if (isShiftedInt<16, 16>(Value)) | |||
15870 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15871 | break; | |||
15872 | case 'K': // "K" is a constant with only the low-order 16 bits nonzero. | |||
15873 | if (isUInt<16>(Value)) | |||
15874 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15875 | break; | |||
15876 | case 'M': // "M" is a constant that is greater than 31. | |||
15877 | if (Value > 31) | |||
15878 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15879 | break; | |||
15880 | case 'N': // "N" is a positive constant that is an exact power of two. | |||
15881 | if (Value > 0 && isPowerOf2_64(Value)) | |||
15882 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15883 | break; | |||
15884 | case 'O': // "O" is the constant zero. | |||
15885 | if (Value == 0) | |||
15886 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15887 | break; | |||
15888 | case 'P': // "P" is a constant whose negation is a signed 16-bit constant. | |||
15889 | if (isInt<16>(-Value)) | |||
15890 | Result = DAG.getTargetConstant(Value, dl, TCVT); | |||
15891 | break; | |||
15892 | } | |||
15893 | break; | |||
15894 | } | |||
15895 | } | |||
15896 | ||||
15897 | if (Result.getNode()) { | |||
15898 | Ops.push_back(Result); | |||
15899 | return; | |||
15900 | } | |||
15901 | ||||
15902 | // Handle standard constraint letters. | |||
15903 | TargetLowering::LowerAsmOperandForConstraint(Op, Constraint, Ops, DAG); | |||
15904 | } | |||
15905 | ||||
15906 | // isLegalAddressingMode - Return true if the addressing mode represented | |||
15907 | // by AM is legal for this target, for a load/store of the specified type. | |||
15908 | bool PPCTargetLowering::isLegalAddressingMode(const DataLayout &DL, | |||
15909 | const AddrMode &AM, Type *Ty, | |||
15910 | unsigned AS, | |||
15911 | Instruction *I) const { | |||
15912 | // Vector type r+i form is supported since power9 as DQ form. We don't check | |||
15913 | // the offset matching DQ form requirement(off % 16 == 0), because on PowerPC, | |||
15914 | // imm form is preferred and the offset can be adjusted to use imm form later | |||
15915 | // in pass PPCLoopInstrFormPrep. Also in LSR, for one LSRUse, it uses min and | |||
15916 | // max offset to check legal addressing mode, we should be a little aggressive | |||
15917 | // to contain other offsets for that LSRUse. | |||
15918 | if (Ty->isVectorTy() && AM.BaseOffs != 0 && !Subtarget.hasP9Vector()) | |||
15919 | return false; | |||
15920 | ||||
15921 | // PPC allows a sign-extended 16-bit immediate field. | |||
15922 | if (AM.BaseOffs <= -(1LL << 16) || AM.BaseOffs >= (1LL << 16)-1) | |||
15923 | return false; | |||
15924 | ||||
15925 | // No global is ever allowed as a base. | |||
15926 | if (AM.BaseGV) | |||
15927 | return false; | |||
15928 | ||||
15929 | // PPC only support r+r, | |||
15930 | switch (AM.Scale) { | |||
15931 | case 0: // "r+i" or just "i", depending on HasBaseReg. | |||
15932 | break; | |||
15933 | case 1: | |||
15934 | if (AM.HasBaseReg && AM.BaseOffs) // "r+r+i" is not allowed. | |||
15935 | return false; | |||
15936 | // Otherwise we have r+r or r+i. | |||
15937 | break; | |||
15938 | case 2: | |||
15939 | if (AM.HasBaseReg || AM.BaseOffs) // 2*r+r or 2*r+i is not allowed. | |||
15940 | return false; | |||
15941 | // Allow 2*r as r+r. | |||
15942 | break; | |||
15943 | default: | |||
15944 | // No other scales are supported. | |||
15945 | return false; | |||
15946 | } | |||
15947 | ||||
15948 | return true; | |||
15949 | } | |||
15950 | ||||
15951 | SDValue PPCTargetLowering::LowerRETURNADDR(SDValue Op, | |||
15952 | SelectionDAG &DAG) const { | |||
15953 | MachineFunction &MF = DAG.getMachineFunction(); | |||
15954 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
15955 | MFI.setReturnAddressIsTaken(true); | |||
15956 | ||||
15957 | if (verifyReturnAddressArgumentIsConstant(Op, DAG)) | |||
15958 | return SDValue(); | |||
15959 | ||||
15960 | SDLoc dl(Op); | |||
15961 | unsigned Depth = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue(); | |||
15962 | ||||
15963 | // Make sure the function does not optimize away the store of the RA to | |||
15964 | // the stack. | |||
15965 | PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); | |||
15966 | FuncInfo->setLRStoreRequired(); | |||
15967 | bool isPPC64 = Subtarget.isPPC64(); | |||
15968 | auto PtrVT = getPointerTy(MF.getDataLayout()); | |||
15969 | ||||
15970 | if (Depth > 0) { | |||
15971 | // The link register (return address) is saved in the caller's frame | |||
15972 | // not the callee's stack frame. So we must get the caller's frame | |||
15973 | // address and load the return address at the LR offset from there. | |||
15974 | SDValue FrameAddr = | |||
15975 | DAG.getLoad(Op.getValueType(), dl, DAG.getEntryNode(), | |||
15976 | LowerFRAMEADDR(Op, DAG), MachinePointerInfo()); | |||
15977 | SDValue Offset = | |||
15978 | DAG.getConstant(Subtarget.getFrameLowering()->getReturnSaveOffset(), dl, | |||
15979 | isPPC64 ? MVT::i64 : MVT::i32); | |||
15980 | return DAG.getLoad(PtrVT, dl, DAG.getEntryNode(), | |||
15981 | DAG.getNode(ISD::ADD, dl, PtrVT, FrameAddr, Offset), | |||
15982 | MachinePointerInfo()); | |||
15983 | } | |||
15984 | ||||
15985 | // Just load the return address off the stack. | |||
15986 | SDValue RetAddrFI = getReturnAddrFrameIndex(DAG); | |||
15987 | return DAG.getLoad(PtrVT, dl, DAG.getEntryNode(), RetAddrFI, | |||
15988 | MachinePointerInfo()); | |||
15989 | } | |||
15990 | ||||
15991 | SDValue PPCTargetLowering::LowerFRAMEADDR(SDValue Op, | |||
15992 | SelectionDAG &DAG) const { | |||
15993 | SDLoc dl(Op); | |||
15994 | unsigned Depth = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue(); | |||
15995 | ||||
15996 | MachineFunction &MF = DAG.getMachineFunction(); | |||
15997 | MachineFrameInfo &MFI = MF.getFrameInfo(); | |||
15998 | MFI.setFrameAddressIsTaken(true); | |||
15999 | ||||
16000 | EVT PtrVT = getPointerTy(MF.getDataLayout()); | |||
16001 | bool isPPC64 = PtrVT == MVT::i64; | |||
16002 | ||||
16003 | // Naked functions never have a frame pointer, and so we use r1. For all | |||
16004 | // other functions, this decision must be delayed until during PEI. | |||
16005 | unsigned FrameReg; | |||
16006 | if (MF.getFunction().hasFnAttribute(Attribute::Naked)) | |||
16007 | FrameReg = isPPC64 ? PPC::X1 : PPC::R1; | |||
16008 | else | |||
16009 | FrameReg = isPPC64 ? PPC::FP8 : PPC::FP; | |||
16010 | ||||
16011 | SDValue FrameAddr = DAG.getCopyFromReg(DAG.getEntryNode(), dl, FrameReg, | |||
16012 | PtrVT); | |||
16013 | while (Depth--) | |||
16014 | FrameAddr = DAG.getLoad(Op.getValueType(), dl, DAG.getEntryNode(), | |||
16015 | FrameAddr, MachinePointerInfo()); | |||
16016 | return FrameAddr; | |||
16017 | } | |||
16018 | ||||
16019 | // FIXME? Maybe this could be a TableGen attribute on some registers and | |||
16020 | // this table could be generated automatically from RegInfo. | |||
16021 | Register PPCTargetLowering::getRegisterByName(const char* RegName, LLT VT, | |||
16022 | const MachineFunction &MF) const { | |||
16023 | bool isPPC64 = Subtarget.isPPC64(); | |||
16024 | ||||
16025 | bool is64Bit = isPPC64 && VT == LLT::scalar(64); | |||
16026 | if (!is64Bit && VT != LLT::scalar(32)) | |||
16027 | report_fatal_error("Invalid register global variable type"); | |||
16028 | ||||
16029 | Register Reg = StringSwitch<Register>(RegName) | |||
16030 | .Case("r1", is64Bit ? PPC::X1 : PPC::R1) | |||
16031 | .Case("r2", isPPC64 ? Register() : PPC::R2) | |||
16032 | .Case("r13", (is64Bit ? PPC::X13 : PPC::R13)) | |||
16033 | .Default(Register()); | |||
16034 | ||||
16035 | if (Reg) | |||
16036 | return Reg; | |||
16037 | report_fatal_error("Invalid register name global variable"); | |||
16038 | } | |||
16039 | ||||
16040 | bool PPCTargetLowering::isAccessedAsGotIndirect(SDValue GA) const { | |||
16041 | // 32-bit SVR4 ABI access everything as got-indirect. | |||
16042 | if (Subtarget.is32BitELFABI()) | |||
16043 | return true; | |||
16044 | ||||
16045 | // AIX accesses everything indirectly through the TOC, which is similar to | |||
16046 | // the GOT. | |||
16047 | if (Subtarget.isAIXABI()) | |||
16048 | return true; | |||
16049 | ||||
16050 | CodeModel::Model CModel = getTargetMachine().getCodeModel(); | |||
16051 | // If it is small or large code model, module locals are accessed | |||
16052 | // indirectly by loading their address from .toc/.got. | |||
16053 | if (CModel == CodeModel::Small || CModel == CodeModel::Large) | |||
16054 | return true; | |||
16055 | ||||
16056 | // JumpTable and BlockAddress are accessed as got-indirect. | |||
16057 | if (isa<JumpTableSDNode>(GA) || isa<BlockAddressSDNode>(GA)) | |||
16058 | return true; | |||
16059 | ||||
16060 | if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(GA)) | |||
16061 | return Subtarget.isGVIndirectSymbol(G->getGlobal()); | |||
16062 | ||||
16063 | return false; | |||
16064 | } | |||
16065 | ||||
16066 | bool | |||
16067 | PPCTargetLowering::isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const { | |||
16068 | // The PowerPC target isn't yet aware of offsets. | |||
16069 | return false; | |||
16070 | } | |||
16071 | ||||
16072 | bool PPCTargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info, | |||
16073 | const CallInst &I, | |||
16074 | MachineFunction &MF, | |||
16075 | unsigned Intrinsic) const { | |||
16076 | switch (Intrinsic) { | |||
16077 | case Intrinsic::ppc_atomicrmw_xchg_i128: | |||
16078 | case Intrinsic::ppc_atomicrmw_add_i128: | |||
16079 | case Intrinsic::ppc_atomicrmw_sub_i128: | |||
16080 | case Intrinsic::ppc_atomicrmw_nand_i128: | |||
16081 | case Intrinsic::ppc_atomicrmw_and_i128: | |||
16082 | case Intrinsic::ppc_atomicrmw_or_i128: | |||
16083 | case Intrinsic::ppc_atomicrmw_xor_i128: | |||
16084 | case Intrinsic::ppc_cmpxchg_i128: | |||
16085 | Info.opc = ISD::INTRINSIC_W_CHAIN; | |||
16086 | Info.memVT = MVT::i128; | |||
16087 | Info.ptrVal = I.getArgOperand(0); | |||
16088 | Info.offset = 0; | |||
16089 | Info.align = Align(16); | |||
16090 | Info.flags = MachineMemOperand::MOLoad | MachineMemOperand::MOStore | | |||
16091 | MachineMemOperand::MOVolatile; | |||
16092 | return true; | |||
16093 | case Intrinsic::ppc_altivec_lvx: | |||
16094 | case Intrinsic::ppc_altivec_lvxl: | |||
16095 | case Intrinsic::ppc_altivec_lvebx: | |||
16096 | case Intrinsic::ppc_altivec_lvehx: | |||
16097 | case Intrinsic::ppc_altivec_lvewx: | |||
16098 | case Intrinsic::ppc_vsx_lxvd2x: | |||
16099 | case Intrinsic::ppc_vsx_lxvw4x: | |||
16100 | case Intrinsic::ppc_vsx_lxvd2x_be: | |||
16101 | case Intrinsic::ppc_vsx_lxvw4x_be: | |||
16102 | case Intrinsic::ppc_vsx_lxvl: | |||
16103 | case Intrinsic::ppc_vsx_lxvll: { | |||
16104 | EVT VT; | |||
16105 | switch (Intrinsic) { | |||
16106 | case Intrinsic::ppc_altivec_lvebx: | |||
16107 | VT = MVT::i8; | |||
16108 | break; | |||
16109 | case Intrinsic::ppc_altivec_lvehx: | |||
16110 | VT = MVT::i16; | |||
16111 | break; | |||
16112 | case Intrinsic::ppc_altivec_lvewx: | |||
16113 | VT = MVT::i32; | |||
16114 | break; | |||
16115 | case Intrinsic::ppc_vsx_lxvd2x: | |||
16116 | case Intrinsic::ppc_vsx_lxvd2x_be: | |||
16117 | VT = MVT::v2f64; | |||
16118 | break; | |||
16119 | default: | |||
16120 | VT = MVT::v4i32; | |||
16121 | break; | |||
16122 | } | |||
16123 | ||||
16124 | Info.opc = ISD::INTRINSIC_W_CHAIN; | |||
16125 | Info.memVT = VT; | |||
16126 | Info.ptrVal = I.getArgOperand(0); | |||
16127 | Info.offset = -VT.getStoreSize()+1; | |||
16128 | Info.size = 2*VT.getStoreSize()-1; | |||
16129 | Info.align = Align(1); | |||
16130 | Info.flags = MachineMemOperand::MOLoad; | |||
16131 | return true; | |||
16132 | } | |||
16133 | case Intrinsic::ppc_altivec_stvx: | |||
16134 | case Intrinsic::ppc_altivec_stvxl: | |||
16135 | case Intrinsic::ppc_altivec_stvebx: | |||
16136 | case Intrinsic::ppc_altivec_stvehx: | |||
16137 | case Intrinsic::ppc_altivec_stvewx: | |||
16138 | case Intrinsic::ppc_vsx_stxvd2x: | |||
16139 | case Intrinsic::ppc_vsx_stxvw4x: | |||
16140 | case Intrinsic::ppc_vsx_stxvd2x_be: | |||
16141 | case Intrinsic::ppc_vsx_stxvw4x_be: | |||
16142 | case Intrinsic::ppc_vsx_stxvl: | |||
16143 | case Intrinsic::ppc_vsx_stxvll: { | |||
16144 | EVT VT; | |||
16145 | switch (Intrinsic) { | |||
16146 | case Intrinsic::ppc_altivec_stvebx: | |||
16147 | VT = MVT::i8; | |||
16148 | break; | |||
16149 | case Intrinsic::ppc_altivec_stvehx: | |||
16150 | VT = MVT::i16; | |||
16151 | break; | |||
16152 | case Intrinsic::ppc_altivec_stvewx: | |||
16153 | VT = MVT::i32; | |||
16154 | break; | |||
16155 | case Intrinsic::ppc_vsx_stxvd2x: | |||
16156 | case Intrinsic::ppc_vsx_stxvd2x_be: | |||
16157 | VT = MVT::v2f64; | |||
16158 | break; | |||
16159 | default: | |||
16160 | VT = MVT::v4i32; | |||
16161 | break; | |||
16162 | } | |||
16163 | ||||
16164 | Info.opc = ISD::INTRINSIC_VOID; | |||
16165 | Info.memVT = VT; | |||
16166 | Info.ptrVal = I.getArgOperand(1); | |||
16167 | Info.offset = -VT.getStoreSize()+1; | |||
16168 | Info.size = 2*VT.getStoreSize()-1; | |||
16169 | Info.align = Align(1); | |||
16170 | Info.flags = MachineMemOperand::MOStore; | |||
16171 | return true; | |||
16172 | } | |||
16173 | default: | |||
16174 | break; | |||
16175 | } | |||
16176 | ||||
16177 | return false; | |||
16178 | } | |||
16179 | ||||
16180 | /// It returns EVT::Other if the type should be determined using generic | |||
16181 | /// target-independent logic. | |||
16182 | EVT PPCTargetLowering::getOptimalMemOpType( | |||
16183 | const MemOp &Op, const AttributeList &FuncAttributes) const { | |||
16184 | if (getTargetMachine().getOptLevel() != CodeGenOpt::None) { | |||
16185 | // We should use Altivec/VSX loads and stores when available. For unaligned | |||
16186 | // addresses, unaligned VSX loads are only fast starting with the P8. | |||
16187 | if (Subtarget.hasAltivec() && Op.size() >= 16 && | |||
16188 | (Op.isAligned(Align(16)) || | |||
16189 | ((Op.isMemset() && Subtarget.hasVSX()) || Subtarget.hasP8Vector()))) | |||
16190 | return MVT::v4i32; | |||
16191 | } | |||
16192 | ||||
16193 | if (Subtarget.isPPC64()) { | |||
16194 | return MVT::i64; | |||
16195 | } | |||
16196 | ||||
16197 | return MVT::i32; | |||
16198 | } | |||
16199 | ||||
16200 | /// Returns true if it is beneficial to convert a load of a constant | |||
16201 | /// to just the constant itself. | |||
16202 | bool PPCTargetLowering::shouldConvertConstantLoadToIntImm(const APInt &Imm, | |||
16203 | Type *Ty) const { | |||
16204 | assert(Ty->isIntegerTy())(static_cast <bool> (Ty->isIntegerTy()) ? void (0) : __assert_fail ("Ty->isIntegerTy()", "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 16204, __extension__ __PRETTY_FUNCTION__)); | |||
16205 | ||||
16206 | unsigned BitSize = Ty->getPrimitiveSizeInBits(); | |||
16207 | return !(BitSize == 0 || BitSize > 64); | |||
16208 | } | |||
16209 | ||||
16210 | bool PPCTargetLowering::isTruncateFree(Type *Ty1, Type *Ty2) const { | |||
16211 | if (!Ty1->isIntegerTy() || !Ty2->isIntegerTy()) | |||
16212 | return false; | |||
16213 | unsigned NumBits1 = Ty1->getPrimitiveSizeInBits(); | |||
16214 | unsigned NumBits2 = Ty2->getPrimitiveSizeInBits(); | |||
16215 | return NumBits1 == 64 && NumBits2 == 32; | |||
16216 | } | |||
16217 | ||||
16218 | bool PPCTargetLowering::isTruncateFree(EVT VT1, EVT VT2) const { | |||
16219 | if (!VT1.isInteger() || !VT2.isInteger()) | |||
16220 | return false; | |||
16221 | unsigned NumBits1 = VT1.getSizeInBits(); | |||
16222 | unsigned NumBits2 = VT2.getSizeInBits(); | |||
16223 | return NumBits1 == 64 && NumBits2 == 32; | |||
16224 | } | |||
16225 | ||||
16226 | bool PPCTargetLowering::isZExtFree(SDValue Val, EVT VT2) const { | |||
16227 | // Generally speaking, zexts are not free, but they are free when they can be | |||
16228 | // folded with other operations. | |||
16229 | if (LoadSDNode *LD = dyn_cast<LoadSDNode>(Val)) { | |||
16230 | EVT MemVT = LD->getMemoryVT(); | |||
16231 | if ((MemVT == MVT::i1 || MemVT == MVT::i8 || MemVT == MVT::i16 || | |||
16232 | (Subtarget.isPPC64() && MemVT == MVT::i32)) && | |||
16233 | (LD->getExtensionType() == ISD::NON_EXTLOAD || | |||
16234 | LD->getExtensionType() == ISD::ZEXTLOAD)) | |||
16235 | return true; | |||
16236 | } | |||
16237 | ||||
16238 | // FIXME: Add other cases... | |||
16239 | // - 32-bit shifts with a zext to i64 | |||
16240 | // - zext after ctlz, bswap, etc. | |||
16241 | // - zext after and by a constant mask | |||
16242 | ||||
16243 | return TargetLowering::isZExtFree(Val, VT2); | |||
16244 | } | |||
16245 | ||||
16246 | bool PPCTargetLowering::isFPExtFree(EVT DestVT, EVT SrcVT) const { | |||
16247 | assert(DestVT.isFloatingPoint() && SrcVT.isFloatingPoint() &&(static_cast <bool> (DestVT.isFloatingPoint() && SrcVT.isFloatingPoint() && "invalid fpext types") ? void (0) : __assert_fail ("DestVT.isFloatingPoint() && SrcVT.isFloatingPoint() && \"invalid fpext types\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 16248, __extension__ __PRETTY_FUNCTION__)) | |||
16248 | "invalid fpext types")(static_cast <bool> (DestVT.isFloatingPoint() && SrcVT.isFloatingPoint() && "invalid fpext types") ? void (0) : __assert_fail ("DestVT.isFloatingPoint() && SrcVT.isFloatingPoint() && \"invalid fpext types\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 16248, __extension__ __PRETTY_FUNCTION__)); | |||
16249 | // Extending to float128 is not free. | |||
16250 | if (DestVT == MVT::f128) | |||
16251 | return false; | |||
16252 | return true; | |||
16253 | } | |||
16254 | ||||
16255 | bool PPCTargetLowering::isLegalICmpImmediate(int64_t Imm) const { | |||
16256 | return isInt<16>(Imm) || isUInt<16>(Imm); | |||
16257 | } | |||
16258 | ||||
16259 | bool PPCTargetLowering::isLegalAddImmediate(int64_t Imm) const { | |||
16260 | return isInt<16>(Imm) || isUInt<16>(Imm); | |||
16261 | } | |||
16262 | ||||
16263 | bool PPCTargetLowering::allowsMisalignedMemoryAccesses(EVT VT, unsigned, Align, | |||
16264 | MachineMemOperand::Flags, | |||
16265 | bool *Fast) const { | |||
16266 | if (DisablePPCUnaligned) | |||
16267 | return false; | |||
16268 | ||||
16269 | // PowerPC supports unaligned memory access for simple non-vector types. | |||
16270 | // Although accessing unaligned addresses is not as efficient as accessing | |||
16271 | // aligned addresses, it is generally more efficient than manual expansion, | |||
16272 | // and generally only traps for software emulation when crossing page | |||
16273 | // boundaries. | |||
16274 | ||||
16275 | if (!VT.isSimple()) | |||
16276 | return false; | |||
16277 | ||||
16278 | if (VT.isFloatingPoint() && !VT.isVector() && | |||
16279 | !Subtarget.allowsUnalignedFPAccess()) | |||
16280 | return false; | |||
16281 | ||||
16282 | if (VT.getSimpleVT().isVector()) { | |||
16283 | if (Subtarget.hasVSX()) { | |||
16284 | if (VT != MVT::v2f64 && VT != MVT::v2i64 && | |||
16285 | VT != MVT::v4f32 && VT != MVT::v4i32) | |||
16286 | return false; | |||
16287 | } else { | |||
16288 | return false; | |||
16289 | } | |||
16290 | } | |||
16291 | ||||
16292 | if (VT == MVT::ppcf128) | |||
16293 | return false; | |||
16294 | ||||
16295 | if (Fast) | |||
16296 | *Fast = true; | |||
16297 | ||||
16298 | return true; | |||
16299 | } | |||
16300 | ||||
16301 | bool PPCTargetLowering::decomposeMulByConstant(LLVMContext &Context, EVT VT, | |||
16302 | SDValue C) const { | |||
16303 | // Check integral scalar types. | |||
16304 | if (!VT.isScalarInteger()) | |||
16305 | return false; | |||
16306 | if (auto *ConstNode = dyn_cast<ConstantSDNode>(C.getNode())) { | |||
16307 | if (!ConstNode->getAPIntValue().isSignedIntN(64)) | |||
16308 | return false; | |||
16309 | // This transformation will generate >= 2 operations. But the following | |||
16310 | // cases will generate <= 2 instructions during ISEL. So exclude them. | |||
16311 | // 1. If the constant multiplier fits 16 bits, it can be handled by one | |||
16312 | // HW instruction, ie. MULLI | |||
16313 | // 2. If the multiplier after shifted fits 16 bits, an extra shift | |||
16314 | // instruction is needed than case 1, ie. MULLI and RLDICR | |||
16315 | int64_t Imm = ConstNode->getSExtValue(); | |||
16316 | unsigned Shift = countTrailingZeros<uint64_t>(Imm); | |||
16317 | Imm >>= Shift; | |||
16318 | if (isInt<16>(Imm)) | |||
16319 | return false; | |||
16320 | uint64_t UImm = static_cast<uint64_t>(Imm); | |||
16321 | if (isPowerOf2_64(UImm + 1) || isPowerOf2_64(UImm - 1) || | |||
16322 | isPowerOf2_64(1 - UImm) || isPowerOf2_64(-1 - UImm)) | |||
16323 | return true; | |||
16324 | } | |||
16325 | return false; | |||
16326 | } | |||
16327 | ||||
16328 | bool PPCTargetLowering::isFMAFasterThanFMulAndFAdd(const MachineFunction &MF, | |||
16329 | EVT VT) const { | |||
16330 | return isFMAFasterThanFMulAndFAdd( | |||
16331 | MF.getFunction(), VT.getTypeForEVT(MF.getFunction().getContext())); | |||
16332 | } | |||
16333 | ||||
16334 | bool PPCTargetLowering::isFMAFasterThanFMulAndFAdd(const Function &F, | |||
16335 | Type *Ty) const { | |||
16336 | switch (Ty->getScalarType()->getTypeID()) { | |||
16337 | case Type::FloatTyID: | |||
16338 | case Type::DoubleTyID: | |||
16339 | return true; | |||
16340 | case Type::FP128TyID: | |||
16341 | return Subtarget.hasP9Vector(); | |||
16342 | default: | |||
16343 | return false; | |||
16344 | } | |||
16345 | } | |||
16346 | ||||
16347 | // FIXME: add more patterns which are not profitable to hoist. | |||
16348 | bool PPCTargetLowering::isProfitableToHoist(Instruction *I) const { | |||
16349 | if (!I->hasOneUse()) | |||
16350 | return true; | |||
16351 | ||||
16352 | Instruction *User = I->user_back(); | |||
16353 | assert(User && "A single use instruction with no uses.")(static_cast <bool> (User && "A single use instruction with no uses." ) ? void (0) : __assert_fail ("User && \"A single use instruction with no uses.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 16353, __extension__ __PRETTY_FUNCTION__)); | |||
16354 | ||||
16355 | switch (I->getOpcode()) { | |||
16356 | case Instruction::FMul: { | |||
16357 | // Don't break FMA, PowerPC prefers FMA. | |||
16358 | if (User->getOpcode() != Instruction::FSub && | |||
16359 | User->getOpcode() != Instruction::FAdd) | |||
16360 | return true; | |||
16361 | ||||
16362 | const TargetOptions &Options = getTargetMachine().Options; | |||
16363 | const Function *F = I->getFunction(); | |||
16364 | const DataLayout &DL = F->getParent()->getDataLayout(); | |||
16365 | Type *Ty = User->getOperand(0)->getType(); | |||
16366 | ||||
16367 | return !( | |||
16368 | isFMAFasterThanFMulAndFAdd(*F, Ty) && | |||
16369 | isOperationLegalOrCustom(ISD::FMA, getValueType(DL, Ty)) && | |||
16370 | (Options.AllowFPOpFusion == FPOpFusion::Fast || Options.UnsafeFPMath)); | |||
16371 | } | |||
16372 | case Instruction::Load: { | |||
16373 | // Don't break "store (load float*)" pattern, this pattern will be combined | |||
16374 | // to "store (load int32)" in later InstCombine pass. See function | |||
16375 | // combineLoadToOperationType. On PowerPC, loading a float point takes more | |||
16376 | // cycles than loading a 32 bit integer. | |||
16377 | LoadInst *LI = cast<LoadInst>(I); | |||
16378 | // For the loads that combineLoadToOperationType does nothing, like | |||
16379 | // ordered load, it should be profitable to hoist them. | |||
16380 | // For swifterror load, it can only be used for pointer to pointer type, so | |||
16381 | // later type check should get rid of this case. | |||
16382 | if (!LI->isUnordered()) | |||
16383 | return true; | |||
16384 | ||||
16385 | if (User->getOpcode() != Instruction::Store) | |||
16386 | return true; | |||
16387 | ||||
16388 | if (I->getType()->getTypeID() != Type::FloatTyID) | |||
16389 | return true; | |||
16390 | ||||
16391 | return false; | |||
16392 | } | |||
16393 | default: | |||
16394 | return true; | |||
16395 | } | |||
16396 | return true; | |||
16397 | } | |||
16398 | ||||
16399 | const MCPhysReg * | |||
16400 | PPCTargetLowering::getScratchRegisters(CallingConv::ID) const { | |||
16401 | // LR is a callee-save register, but we must treat it as clobbered by any call | |||
16402 | // site. Hence we include LR in the scratch registers, which are in turn added | |||
16403 | // as implicit-defs for stackmaps and patchpoints. The same reasoning applies | |||
16404 | // to CTR, which is used by any indirect call. | |||
16405 | static const MCPhysReg ScratchRegs[] = { | |||
16406 | PPC::X12, PPC::LR8, PPC::CTR8, 0 | |||
16407 | }; | |||
16408 | ||||
16409 | return ScratchRegs; | |||
16410 | } | |||
16411 | ||||
16412 | Register PPCTargetLowering::getExceptionPointerRegister( | |||
16413 | const Constant *PersonalityFn) const { | |||
16414 | return Subtarget.isPPC64() ? PPC::X3 : PPC::R3; | |||
16415 | } | |||
16416 | ||||
16417 | Register PPCTargetLowering::getExceptionSelectorRegister( | |||
16418 | const Constant *PersonalityFn) const { | |||
16419 | return Subtarget.isPPC64() ? PPC::X4 : PPC::R4; | |||
16420 | } | |||
16421 | ||||
16422 | bool | |||
16423 | PPCTargetLowering::shouldExpandBuildVectorWithShuffles( | |||
16424 | EVT VT , unsigned DefinedValues) const { | |||
16425 | if (VT == MVT::v2i64) | |||
16426 | return Subtarget.hasDirectMove(); // Don't need stack ops with direct moves | |||
16427 | ||||
16428 | if (Subtarget.hasVSX()) | |||
16429 | return true; | |||
16430 | ||||
16431 | return TargetLowering::shouldExpandBuildVectorWithShuffles(VT, DefinedValues); | |||
16432 | } | |||
16433 | ||||
16434 | Sched::Preference PPCTargetLowering::getSchedulingPreference(SDNode *N) const { | |||
16435 | if (DisableILPPref || Subtarget.enableMachineScheduler()) | |||
16436 | return TargetLowering::getSchedulingPreference(N); | |||
16437 | ||||
16438 | return Sched::ILP; | |||
16439 | } | |||
16440 | ||||
16441 | // Create a fast isel object. | |||
16442 | FastISel * | |||
16443 | PPCTargetLowering::createFastISel(FunctionLoweringInfo &FuncInfo, | |||
16444 | const TargetLibraryInfo *LibInfo) const { | |||
16445 | return PPC::createFastISel(FuncInfo, LibInfo); | |||
16446 | } | |||
16447 | ||||
16448 | // 'Inverted' means the FMA opcode after negating one multiplicand. | |||
16449 | // For example, (fma -a b c) = (fnmsub a b c) | |||
16450 | static unsigned invertFMAOpcode(unsigned Opc) { | |||
16451 | switch (Opc) { | |||
16452 | default: | |||
16453 | llvm_unreachable("Invalid FMA opcode for PowerPC!")::llvm::llvm_unreachable_internal("Invalid FMA opcode for PowerPC!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 16453); | |||
16454 | case ISD::FMA: | |||
16455 | return PPCISD::FNMSUB; | |||
16456 | case PPCISD::FNMSUB: | |||
16457 | return ISD::FMA; | |||
16458 | } | |||
16459 | } | |||
16460 | ||||
16461 | SDValue PPCTargetLowering::getNegatedExpression(SDValue Op, SelectionDAG &DAG, | |||
16462 | bool LegalOps, bool OptForSize, | |||
16463 | NegatibleCost &Cost, | |||
16464 | unsigned Depth) const { | |||
16465 | if (Depth > SelectionDAG::MaxRecursionDepth) | |||
16466 | return SDValue(); | |||
16467 | ||||
16468 | unsigned Opc = Op.getOpcode(); | |||
16469 | EVT VT = Op.getValueType(); | |||
16470 | SDNodeFlags Flags = Op.getNode()->getFlags(); | |||
16471 | ||||
16472 | switch (Opc) { | |||
16473 | case PPCISD::FNMSUB: | |||
16474 | if (!Op.hasOneUse() || !isTypeLegal(VT)) | |||
16475 | break; | |||
16476 | ||||
16477 | const TargetOptions &Options = getTargetMachine().Options; | |||
16478 | SDValue N0 = Op.getOperand(0); | |||
16479 | SDValue N1 = Op.getOperand(1); | |||
16480 | SDValue N2 = Op.getOperand(2); | |||
16481 | SDLoc Loc(Op); | |||
16482 | ||||
16483 | NegatibleCost N2Cost = NegatibleCost::Expensive; | |||
16484 | SDValue NegN2 = | |||
16485 | getNegatedExpression(N2, DAG, LegalOps, OptForSize, N2Cost, Depth + 1); | |||
16486 | ||||
16487 | if (!NegN2) | |||
16488 | return SDValue(); | |||
16489 | ||||
16490 | // (fneg (fnmsub a b c)) => (fnmsub (fneg a) b (fneg c)) | |||
16491 | // (fneg (fnmsub a b c)) => (fnmsub a (fneg b) (fneg c)) | |||
16492 | // These transformations may change sign of zeroes. For example, | |||
16493 | // -(-ab-(-c))=-0 while -(-(ab-c))=+0 when a=b=c=1. | |||
16494 | if (Flags.hasNoSignedZeros() || Options.NoSignedZerosFPMath) { | |||
16495 | // Try and choose the cheaper one to negate. | |||
16496 | NegatibleCost N0Cost = NegatibleCost::Expensive; | |||
16497 | SDValue NegN0 = getNegatedExpression(N0, DAG, LegalOps, OptForSize, | |||
16498 | N0Cost, Depth + 1); | |||
16499 | ||||
16500 | NegatibleCost N1Cost = NegatibleCost::Expensive; | |||
16501 | SDValue NegN1 = getNegatedExpression(N1, DAG, LegalOps, OptForSize, | |||
16502 | N1Cost, Depth + 1); | |||
16503 | ||||
16504 | if (NegN0 && N0Cost <= N1Cost) { | |||
16505 | Cost = std::min(N0Cost, N2Cost); | |||
16506 | return DAG.getNode(Opc, Loc, VT, NegN0, N1, NegN2, Flags); | |||
16507 | } else if (NegN1) { | |||
16508 | Cost = std::min(N1Cost, N2Cost); | |||
16509 | return DAG.getNode(Opc, Loc, VT, N0, NegN1, NegN2, Flags); | |||
16510 | } | |||
16511 | } | |||
16512 | ||||
16513 | // (fneg (fnmsub a b c)) => (fma a b (fneg c)) | |||
16514 | if (isOperationLegal(ISD::FMA, VT)) { | |||
16515 | Cost = N2Cost; | |||
16516 | return DAG.getNode(ISD::FMA, Loc, VT, N0, N1, NegN2, Flags); | |||
16517 | } | |||
16518 | ||||
16519 | break; | |||
16520 | } | |||
16521 | ||||
16522 | return TargetLowering::getNegatedExpression(Op, DAG, LegalOps, OptForSize, | |||
16523 | Cost, Depth); | |||
16524 | } | |||
16525 | ||||
16526 | // Override to enable LOAD_STACK_GUARD lowering on Linux. | |||
16527 | bool PPCTargetLowering::useLoadStackGuardNode() const { | |||
16528 | if (!Subtarget.isTargetLinux()) | |||
16529 | return TargetLowering::useLoadStackGuardNode(); | |||
16530 | return true; | |||
16531 | } | |||
16532 | ||||
16533 | // Override to disable global variable loading on Linux and insert AIX canary | |||
16534 | // word declaration. | |||
16535 | void PPCTargetLowering::insertSSPDeclarations(Module &M) const { | |||
16536 | if (Subtarget.isAIXABI()) { | |||
16537 | M.getOrInsertGlobal(AIXSSPCanaryWordName, | |||
16538 | Type::getInt8PtrTy(M.getContext())); | |||
16539 | return; | |||
16540 | } | |||
16541 | if (!Subtarget.isTargetLinux()) | |||
16542 | return TargetLowering::insertSSPDeclarations(M); | |||
16543 | } | |||
16544 | ||||
16545 | Value *PPCTargetLowering::getSDagStackGuard(const Module &M) const { | |||
16546 | if (Subtarget.isAIXABI()) | |||
16547 | return M.getGlobalVariable(AIXSSPCanaryWordName); | |||
16548 | return TargetLowering::getSDagStackGuard(M); | |||
16549 | } | |||
16550 | ||||
16551 | bool PPCTargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT, | |||
16552 | bool ForCodeSize) const { | |||
16553 | if (!VT.isSimple() || !Subtarget.hasVSX()) | |||
16554 | return false; | |||
16555 | ||||
16556 | switch(VT.getSimpleVT().SimpleTy) { | |||
16557 | default: | |||
16558 | // For FP types that are currently not supported by PPC backend, return | |||
16559 | // false. Examples: f16, f80. | |||
16560 | return false; | |||
16561 | case MVT::f32: | |||
16562 | case MVT::f64: | |||
16563 | if (Subtarget.hasPrefixInstrs()) { | |||
16564 | // we can materialize all immediatess via XXSPLTI32DX and XXSPLTIDP. | |||
16565 | return true; | |||
16566 | } | |||
16567 | LLVM_FALLTHROUGH[[gnu::fallthrough]]; | |||
16568 | case MVT::ppcf128: | |||
16569 | return Imm.isPosZero(); | |||
16570 | } | |||
16571 | } | |||
16572 | ||||
16573 | // For vector shift operation op, fold | |||
16574 | // (op x, (and y, ((1 << numbits(x)) - 1))) -> (target op x, y) | |||
16575 | static SDValue stripModuloOnShift(const TargetLowering &TLI, SDNode *N, | |||
16576 | SelectionDAG &DAG) { | |||
16577 | SDValue N0 = N->getOperand(0); | |||
16578 | SDValue N1 = N->getOperand(1); | |||
16579 | EVT VT = N0.getValueType(); | |||
16580 | unsigned OpSizeInBits = VT.getScalarSizeInBits(); | |||
16581 | unsigned Opcode = N->getOpcode(); | |||
16582 | unsigned TargetOpcode; | |||
16583 | ||||
16584 | switch (Opcode) { | |||
16585 | default: | |||
16586 | llvm_unreachable("Unexpected shift operation")::llvm::llvm_unreachable_internal("Unexpected shift operation" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 16586); | |||
16587 | case ISD::SHL: | |||
16588 | TargetOpcode = PPCISD::SHL; | |||
16589 | break; | |||
16590 | case ISD::SRL: | |||
16591 | TargetOpcode = PPCISD::SRL; | |||
16592 | break; | |||
16593 | case ISD::SRA: | |||
16594 | TargetOpcode = PPCISD::SRA; | |||
16595 | break; | |||
16596 | } | |||
16597 | ||||
16598 | if (VT.isVector() && TLI.isOperationLegal(Opcode, VT) && | |||
16599 | N1->getOpcode() == ISD::AND) | |||
16600 | if (ConstantSDNode *Mask = isConstOrConstSplat(N1->getOperand(1))) | |||
16601 | if (Mask->getZExtValue() == OpSizeInBits - 1) | |||
16602 | return DAG.getNode(TargetOpcode, SDLoc(N), VT, N0, N1->getOperand(0)); | |||
16603 | ||||
16604 | return SDValue(); | |||
16605 | } | |||
16606 | ||||
16607 | SDValue PPCTargetLowering::combineSHL(SDNode *N, DAGCombinerInfo &DCI) const { | |||
16608 | if (auto Value = stripModuloOnShift(*this, N, DCI.DAG)) | |||
16609 | return Value; | |||
16610 | ||||
16611 | SDValue N0 = N->getOperand(0); | |||
16612 | ConstantSDNode *CN1 = dyn_cast<ConstantSDNode>(N->getOperand(1)); | |||
16613 | if (!Subtarget.isISA3_0() || !Subtarget.isPPC64() || | |||
16614 | N0.getOpcode() != ISD::SIGN_EXTEND || | |||
16615 | N0.getOperand(0).getValueType() != MVT::i32 || CN1 == nullptr || | |||
16616 | N->getValueType(0) != MVT::i64) | |||
16617 | return SDValue(); | |||
16618 | ||||
16619 | // We can't save an operation here if the value is already extended, and | |||
16620 | // the existing shift is easier to combine. | |||
16621 | SDValue ExtsSrc = N0.getOperand(0); | |||
16622 | if (ExtsSrc.getOpcode() == ISD::TRUNCATE && | |||
16623 | ExtsSrc.getOperand(0).getOpcode() == ISD::AssertSext) | |||
16624 | return SDValue(); | |||
16625 | ||||
16626 | SDLoc DL(N0); | |||
16627 | SDValue ShiftBy = SDValue(CN1, 0); | |||
16628 | // We want the shift amount to be i32 on the extswli, but the shift could | |||
16629 | // have an i64. | |||
16630 | if (ShiftBy.getValueType() == MVT::i64) | |||
16631 | ShiftBy = DCI.DAG.getConstant(CN1->getZExtValue(), DL, MVT::i32); | |||
16632 | ||||
16633 | return DCI.DAG.getNode(PPCISD::EXTSWSLI, DL, MVT::i64, N0->getOperand(0), | |||
16634 | ShiftBy); | |||
16635 | } | |||
16636 | ||||
16637 | SDValue PPCTargetLowering::combineSRA(SDNode *N, DAGCombinerInfo &DCI) const { | |||
16638 | if (auto Value = stripModuloOnShift(*this, N, DCI.DAG)) | |||
16639 | return Value; | |||
16640 | ||||
16641 | return SDValue(); | |||
16642 | } | |||
16643 | ||||
16644 | SDValue PPCTargetLowering::combineSRL(SDNode *N, DAGCombinerInfo &DCI) const { | |||
16645 | if (auto Value = stripModuloOnShift(*this, N, DCI.DAG)) | |||
16646 | return Value; | |||
16647 | ||||
16648 | return SDValue(); | |||
16649 | } | |||
16650 | ||||
16651 | // Transform (add X, (zext(setne Z, C))) -> (addze X, (addic (addi Z, -C), -1)) | |||
16652 | // Transform (add X, (zext(sete Z, C))) -> (addze X, (subfic (addi Z, -C), 0)) | |||
16653 | // When C is zero, the equation (addi Z, -C) can be simplified to Z | |||
16654 | // Requirement: -C in [-32768, 32767], X and Z are MVT::i64 types | |||
16655 | static SDValue combineADDToADDZE(SDNode *N, SelectionDAG &DAG, | |||
16656 | const PPCSubtarget &Subtarget) { | |||
16657 | if (!Subtarget.isPPC64()) | |||
16658 | return SDValue(); | |||
16659 | ||||
16660 | SDValue LHS = N->getOperand(0); | |||
16661 | SDValue RHS = N->getOperand(1); | |||
16662 | ||||
16663 | auto isZextOfCompareWithConstant = [](SDValue Op) { | |||
16664 | if (Op.getOpcode() != ISD::ZERO_EXTEND || !Op.hasOneUse() || | |||
16665 | Op.getValueType() != MVT::i64) | |||
16666 | return false; | |||
16667 | ||||
16668 | SDValue Cmp = Op.getOperand(0); | |||
16669 | if (Cmp.getOpcode() != ISD::SETCC || !Cmp.hasOneUse() || | |||
16670 | Cmp.getOperand(0).getValueType() != MVT::i64) | |||
16671 | return false; | |||
16672 | ||||
16673 | if (auto *Constant = dyn_cast<ConstantSDNode>(Cmp.getOperand(1))) { | |||
16674 | int64_t NegConstant = 0 - Constant->getSExtValue(); | |||
16675 | // Due to the limitations of the addi instruction, | |||
16676 | // -C is required to be [-32768, 32767]. | |||
16677 | return isInt<16>(NegConstant); | |||
16678 | } | |||
16679 | ||||
16680 | return false; | |||
16681 | }; | |||
16682 | ||||
16683 | bool LHSHasPattern = isZextOfCompareWithConstant(LHS); | |||
16684 | bool RHSHasPattern = isZextOfCompareWithConstant(RHS); | |||
16685 | ||||
16686 | // If there is a pattern, canonicalize a zext operand to the RHS. | |||
16687 | if (LHSHasPattern && !RHSHasPattern) | |||
16688 | std::swap(LHS, RHS); | |||
16689 | else if (!LHSHasPattern && !RHSHasPattern) | |||
16690 | return SDValue(); | |||
16691 | ||||
16692 | SDLoc DL(N); | |||
16693 | SDVTList VTs = DAG.getVTList(MVT::i64, MVT::Glue); | |||
16694 | SDValue Cmp = RHS.getOperand(0); | |||
16695 | SDValue Z = Cmp.getOperand(0); | |||
16696 | auto *Constant = cast<ConstantSDNode>(Cmp.getOperand(1)); | |||
16697 | int64_t NegConstant = 0 - Constant->getSExtValue(); | |||
16698 | ||||
16699 | switch(cast<CondCodeSDNode>(Cmp.getOperand(2))->get()) { | |||
16700 | default: break; | |||
16701 | case ISD::SETNE: { | |||
16702 | // when C == 0 | |||
16703 | // --> addze X, (addic Z, -1).carry | |||
16704 | // / | |||
16705 | // add X, (zext(setne Z, C))-- | |||
16706 | // \ when -32768 <= -C <= 32767 && C != 0 | |||
16707 | // --> addze X, (addic (addi Z, -C), -1).carry | |||
16708 | SDValue Add = DAG.getNode(ISD::ADD, DL, MVT::i64, Z, | |||
16709 | DAG.getConstant(NegConstant, DL, MVT::i64)); | |||
16710 | SDValue AddOrZ = NegConstant != 0 ? Add : Z; | |||
16711 | SDValue Addc = DAG.getNode(ISD::ADDC, DL, DAG.getVTList(MVT::i64, MVT::Glue), | |||
16712 | AddOrZ, DAG.getConstant(-1ULL, DL, MVT::i64)); | |||
16713 | return DAG.getNode(ISD::ADDE, DL, VTs, LHS, DAG.getConstant(0, DL, MVT::i64), | |||
16714 | SDValue(Addc.getNode(), 1)); | |||
16715 | } | |||
16716 | case ISD::SETEQ: { | |||
16717 | // when C == 0 | |||
16718 | // --> addze X, (subfic Z, 0).carry | |||
16719 | // / | |||
16720 | // add X, (zext(sete Z, C))-- | |||
16721 | // \ when -32768 <= -C <= 32767 && C != 0 | |||
16722 | // --> addze X, (subfic (addi Z, -C), 0).carry | |||
16723 | SDValue Add = DAG.getNode(ISD::ADD, DL, MVT::i64, Z, | |||
16724 | DAG.getConstant(NegConstant, DL, MVT::i64)); | |||
16725 | SDValue AddOrZ = NegConstant != 0 ? Add : Z; | |||
16726 | SDValue Subc = DAG.getNode(ISD::SUBC, DL, DAG.getVTList(MVT::i64, MVT::Glue), | |||
16727 | DAG.getConstant(0, DL, MVT::i64), AddOrZ); | |||
16728 | return DAG.getNode(ISD::ADDE, DL, VTs, LHS, DAG.getConstant(0, DL, MVT::i64), | |||
16729 | SDValue(Subc.getNode(), 1)); | |||
16730 | } | |||
16731 | } | |||
16732 | ||||
16733 | return SDValue(); | |||
16734 | } | |||
16735 | ||||
16736 | // Transform | |||
16737 | // (add C1, (MAT_PCREL_ADDR GlobalAddr+C2)) to | |||
16738 | // (MAT_PCREL_ADDR GlobalAddr+(C1+C2)) | |||
16739 | // In this case both C1 and C2 must be known constants. | |||
16740 | // C1+C2 must fit into a 34 bit signed integer. | |||
16741 | static SDValue combineADDToMAT_PCREL_ADDR(SDNode *N, SelectionDAG &DAG, | |||
16742 | const PPCSubtarget &Subtarget) { | |||
16743 | if (!Subtarget.isUsingPCRelativeCalls()) | |||
16744 | return SDValue(); | |||
16745 | ||||
16746 | // Check both Operand 0 and Operand 1 of the ADD node for the PCRel node. | |||
16747 | // If we find that node try to cast the Global Address and the Constant. | |||
16748 | SDValue LHS = N->getOperand(0); | |||
16749 | SDValue RHS = N->getOperand(1); | |||
16750 | ||||
16751 | if (LHS.getOpcode() != PPCISD::MAT_PCREL_ADDR) | |||
16752 | std::swap(LHS, RHS); | |||
16753 | ||||
16754 | if (LHS.getOpcode() != PPCISD::MAT_PCREL_ADDR) | |||
16755 | return SDValue(); | |||
16756 | ||||
16757 | // Operand zero of PPCISD::MAT_PCREL_ADDR is the GA node. | |||
16758 | GlobalAddressSDNode *GSDN = dyn_cast<GlobalAddressSDNode>(LHS.getOperand(0)); | |||
16759 | ConstantSDNode* ConstNode = dyn_cast<ConstantSDNode>(RHS); | |||
16760 | ||||
16761 | // Check that both casts succeeded. | |||
16762 | if (!GSDN || !ConstNode) | |||
16763 | return SDValue(); | |||
16764 | ||||
16765 | int64_t NewOffset = GSDN->getOffset() + ConstNode->getSExtValue(); | |||
16766 | SDLoc DL(GSDN); | |||
16767 | ||||
16768 | // The signed int offset needs to fit in 34 bits. | |||
16769 | if (!isInt<34>(NewOffset)) | |||
16770 | return SDValue(); | |||
16771 | ||||
16772 | // The new global address is a copy of the old global address except | |||
16773 | // that it has the updated Offset. | |||
16774 | SDValue GA = | |||
16775 | DAG.getTargetGlobalAddress(GSDN->getGlobal(), DL, GSDN->getValueType(0), | |||
16776 | NewOffset, GSDN->getTargetFlags()); | |||
16777 | SDValue MatPCRel = | |||
16778 | DAG.getNode(PPCISD::MAT_PCREL_ADDR, DL, GSDN->getValueType(0), GA); | |||
16779 | return MatPCRel; | |||
16780 | } | |||
16781 | ||||
16782 | SDValue PPCTargetLowering::combineADD(SDNode *N, DAGCombinerInfo &DCI) const { | |||
16783 | if (auto Value = combineADDToADDZE(N, DCI.DAG, Subtarget)) | |||
16784 | return Value; | |||
16785 | ||||
16786 | if (auto Value = combineADDToMAT_PCREL_ADDR(N, DCI.DAG, Subtarget)) | |||
16787 | return Value; | |||
16788 | ||||
16789 | return SDValue(); | |||
16790 | } | |||
16791 | ||||
16792 | // Detect TRUNCATE operations on bitcasts of float128 values. | |||
16793 | // What we are looking for here is the situtation where we extract a subset | |||
16794 | // of bits from a 128 bit float. | |||
16795 | // This can be of two forms: | |||
16796 | // 1) BITCAST of f128 feeding TRUNCATE | |||
16797 | // 2) BITCAST of f128 feeding SRL (a shift) feeding TRUNCATE | |||
16798 | // The reason this is required is because we do not have a legal i128 type | |||
16799 | // and so we want to prevent having to store the f128 and then reload part | |||
16800 | // of it. | |||
16801 | SDValue PPCTargetLowering::combineTRUNCATE(SDNode *N, | |||
16802 | DAGCombinerInfo &DCI) const { | |||
16803 | // If we are using CRBits then try that first. | |||
16804 | if (Subtarget.useCRBits()) { | |||
16805 | // Check if CRBits did anything and return that if it did. | |||
16806 | if (SDValue CRTruncValue = DAGCombineTruncBoolExt(N, DCI)) | |||
16807 | return CRTruncValue; | |||
16808 | } | |||
16809 | ||||
16810 | SDLoc dl(N); | |||
16811 | SDValue Op0 = N->getOperand(0); | |||
16812 | ||||
16813 | // fold (truncate (abs (sub (zext a), (zext b)))) -> (vabsd a, b) | |||
16814 | if (Subtarget.hasP9Altivec() && Op0.getOpcode() == ISD::ABS) { | |||
16815 | EVT VT = N->getValueType(0); | |||
16816 | if (VT != MVT::v4i32 && VT != MVT::v8i16 && VT != MVT::v16i8) | |||
16817 | return SDValue(); | |||
16818 | SDValue Sub = Op0.getOperand(0); | |||
16819 | if (Sub.getOpcode() == ISD::SUB) { | |||
16820 | SDValue SubOp0 = Sub.getOperand(0); | |||
16821 | SDValue SubOp1 = Sub.getOperand(1); | |||
16822 | if ((SubOp0.getOpcode() == ISD::ZERO_EXTEND) && | |||
16823 | (SubOp1.getOpcode() == ISD::ZERO_EXTEND)) { | |||
16824 | return DCI.DAG.getNode(PPCISD::VABSD, dl, VT, SubOp0.getOperand(0), | |||
16825 | SubOp1.getOperand(0), | |||
16826 | DCI.DAG.getTargetConstant(0, dl, MVT::i32)); | |||
16827 | } | |||
16828 | } | |||
16829 | } | |||
16830 | ||||
16831 | // Looking for a truncate of i128 to i64. | |||
16832 | if (Op0.getValueType() != MVT::i128 || N->getValueType(0) != MVT::i64) | |||
16833 | return SDValue(); | |||
16834 | ||||
16835 | int EltToExtract = DCI.DAG.getDataLayout().isBigEndian() ? 1 : 0; | |||
16836 | ||||
16837 | // SRL feeding TRUNCATE. | |||
16838 | if (Op0.getOpcode() == ISD::SRL) { | |||
16839 | ConstantSDNode *ConstNode = dyn_cast<ConstantSDNode>(Op0.getOperand(1)); | |||
16840 | // The right shift has to be by 64 bits. | |||
16841 | if (!ConstNode || ConstNode->getZExtValue() != 64) | |||
16842 | return SDValue(); | |||
16843 | ||||
16844 | // Switch the element number to extract. | |||
16845 | EltToExtract = EltToExtract ? 0 : 1; | |||
16846 | // Update Op0 past the SRL. | |||
16847 | Op0 = Op0.getOperand(0); | |||
16848 | } | |||
16849 | ||||
16850 | // BITCAST feeding a TRUNCATE possibly via SRL. | |||
16851 | if (Op0.getOpcode() == ISD::BITCAST && | |||
16852 | Op0.getValueType() == MVT::i128 && | |||
16853 | Op0.getOperand(0).getValueType() == MVT::f128) { | |||
16854 | SDValue Bitcast = DCI.DAG.getBitcast(MVT::v2i64, Op0.getOperand(0)); | |||
16855 | return DCI.DAG.getNode( | |||
16856 | ISD::EXTRACT_VECTOR_ELT, dl, MVT::i64, Bitcast, | |||
16857 | DCI.DAG.getTargetConstant(EltToExtract, dl, MVT::i32)); | |||
16858 | } | |||
16859 | return SDValue(); | |||
16860 | } | |||
16861 | ||||
16862 | SDValue PPCTargetLowering::combineMUL(SDNode *N, DAGCombinerInfo &DCI) const { | |||
16863 | SelectionDAG &DAG = DCI.DAG; | |||
16864 | ||||
16865 | ConstantSDNode *ConstOpOrElement = isConstOrConstSplat(N->getOperand(1)); | |||
16866 | if (!ConstOpOrElement) | |||
16867 | return SDValue(); | |||
16868 | ||||
16869 | // An imul is usually smaller than the alternative sequence for legal type. | |||
16870 | if (DAG.getMachineFunction().getFunction().hasMinSize() && | |||
16871 | isOperationLegal(ISD::MUL, N->getValueType(0))) | |||
16872 | return SDValue(); | |||
16873 | ||||
16874 | auto IsProfitable = [this](bool IsNeg, bool IsAddOne, EVT VT) -> bool { | |||
16875 | switch (this->Subtarget.getCPUDirective()) { | |||
16876 | default: | |||
16877 | // TODO: enhance the condition for subtarget before pwr8 | |||
16878 | return false; | |||
16879 | case PPC::DIR_PWR8: | |||
16880 | // type mul add shl | |||
16881 | // scalar 4 1 1 | |||
16882 | // vector 7 2 2 | |||
16883 | return true; | |||
16884 | case PPC::DIR_PWR9: | |||
16885 | case PPC::DIR_PWR10: | |||
16886 | case PPC::DIR_PWR_FUTURE: | |||
16887 | // type mul add shl | |||
16888 | // scalar 5 2 2 | |||
16889 | // vector 7 2 2 | |||
16890 | ||||
16891 | // The cycle RATIO of related operations are showed as a table above. | |||
16892 | // Because mul is 5(scalar)/7(vector), add/sub/shl are all 2 for both | |||
16893 | // scalar and vector type. For 2 instrs patterns, add/sub + shl | |||
16894 | // are 4, it is always profitable; but for 3 instrs patterns | |||
16895 | // (mul x, -(2^N + 1)) => -(add (shl x, N), x), sub + add + shl are 6. | |||
16896 | // So we should only do it for vector type. | |||
16897 | return IsAddOne && IsNeg ? VT.isVector() : true; | |||
16898 | } | |||
16899 | }; | |||
16900 | ||||
16901 | EVT VT = N->getValueType(0); | |||
16902 | SDLoc DL(N); | |||
16903 | ||||
16904 | const APInt &MulAmt = ConstOpOrElement->getAPIntValue(); | |||
16905 | bool IsNeg = MulAmt.isNegative(); | |||
16906 | APInt MulAmtAbs = MulAmt.abs(); | |||
16907 | ||||
16908 | if ((MulAmtAbs - 1).isPowerOf2()) { | |||
16909 | // (mul x, 2^N + 1) => (add (shl x, N), x) | |||
16910 | // (mul x, -(2^N + 1)) => -(add (shl x, N), x) | |||
16911 | ||||
16912 | if (!IsProfitable(IsNeg, true, VT)) | |||
16913 | return SDValue(); | |||
16914 | ||||
16915 | SDValue Op0 = N->getOperand(0); | |||
16916 | SDValue Op1 = | |||
16917 | DAG.getNode(ISD::SHL, DL, VT, N->getOperand(0), | |||
16918 | DAG.getConstant((MulAmtAbs - 1).logBase2(), DL, VT)); | |||
16919 | SDValue Res = DAG.getNode(ISD::ADD, DL, VT, Op0, Op1); | |||
16920 | ||||
16921 | if (!IsNeg) | |||
16922 | return Res; | |||
16923 | ||||
16924 | return DAG.getNode(ISD::SUB, DL, VT, DAG.getConstant(0, DL, VT), Res); | |||
16925 | } else if ((MulAmtAbs + 1).isPowerOf2()) { | |||
16926 | // (mul x, 2^N - 1) => (sub (shl x, N), x) | |||
16927 | // (mul x, -(2^N - 1)) => (sub x, (shl x, N)) | |||
16928 | ||||
16929 | if (!IsProfitable(IsNeg, false, VT)) | |||
16930 | return SDValue(); | |||
16931 | ||||
16932 | SDValue Op0 = N->getOperand(0); | |||
16933 | SDValue Op1 = | |||
16934 | DAG.getNode(ISD::SHL, DL, VT, N->getOperand(0), | |||
16935 | DAG.getConstant((MulAmtAbs + 1).logBase2(), DL, VT)); | |||
16936 | ||||
16937 | if (!IsNeg) | |||
16938 | return DAG.getNode(ISD::SUB, DL, VT, Op1, Op0); | |||
16939 | else | |||
16940 | return DAG.getNode(ISD::SUB, DL, VT, Op0, Op1); | |||
16941 | ||||
16942 | } else { | |||
16943 | return SDValue(); | |||
16944 | } | |||
16945 | } | |||
16946 | ||||
16947 | // Combine fma-like op (like fnmsub) with fnegs to appropriate op. Do this | |||
16948 | // in combiner since we need to check SD flags and other subtarget features. | |||
16949 | SDValue PPCTargetLowering::combineFMALike(SDNode *N, | |||
16950 | DAGCombinerInfo &DCI) const { | |||
16951 | SDValue N0 = N->getOperand(0); | |||
16952 | SDValue N1 = N->getOperand(1); | |||
16953 | SDValue N2 = N->getOperand(2); | |||
16954 | SDNodeFlags Flags = N->getFlags(); | |||
16955 | EVT VT = N->getValueType(0); | |||
16956 | SelectionDAG &DAG = DCI.DAG; | |||
16957 | const TargetOptions &Options = getTargetMachine().Options; | |||
16958 | unsigned Opc = N->getOpcode(); | |||
16959 | bool CodeSize = DAG.getMachineFunction().getFunction().hasOptSize(); | |||
16960 | bool LegalOps = !DCI.isBeforeLegalizeOps(); | |||
16961 | SDLoc Loc(N); | |||
16962 | ||||
16963 | if (!isOperationLegal(ISD::FMA, VT)) | |||
16964 | return SDValue(); | |||
16965 | ||||
16966 | // Allowing transformation to FNMSUB may change sign of zeroes when ab-c=0 | |||
16967 | // since (fnmsub a b c)=-0 while c-ab=+0. | |||
16968 | if (!Flags.hasNoSignedZeros() && !Options.NoSignedZerosFPMath) | |||
16969 | return SDValue(); | |||
16970 | ||||
16971 | // (fma (fneg a) b c) => (fnmsub a b c) | |||
16972 | // (fnmsub (fneg a) b c) => (fma a b c) | |||
16973 | if (SDValue NegN0 = getCheaperNegatedExpression(N0, DAG, LegalOps, CodeSize)) | |||
16974 | return DAG.getNode(invertFMAOpcode(Opc), Loc, VT, NegN0, N1, N2, Flags); | |||
16975 | ||||
16976 | // (fma a (fneg b) c) => (fnmsub a b c) | |||
16977 | // (fnmsub a (fneg b) c) => (fma a b c) | |||
16978 | if (SDValue NegN1 = getCheaperNegatedExpression(N1, DAG, LegalOps, CodeSize)) | |||
16979 | return DAG.getNode(invertFMAOpcode(Opc), Loc, VT, N0, NegN1, N2, Flags); | |||
16980 | ||||
16981 | return SDValue(); | |||
16982 | } | |||
16983 | ||||
16984 | bool PPCTargetLowering::mayBeEmittedAsTailCall(const CallInst *CI) const { | |||
16985 | // Only duplicate to increase tail-calls for the 64bit SysV ABIs. | |||
16986 | if (!Subtarget.is64BitELFABI()) | |||
16987 | return false; | |||
16988 | ||||
16989 | // If not a tail call then no need to proceed. | |||
16990 | if (!CI->isTailCall()) | |||
16991 | return false; | |||
16992 | ||||
16993 | // If sibling calls have been disabled and tail-calls aren't guaranteed | |||
16994 | // there is no reason to duplicate. | |||
16995 | auto &TM = getTargetMachine(); | |||
16996 | if (!TM.Options.GuaranteedTailCallOpt && DisableSCO) | |||
16997 | return false; | |||
16998 | ||||
16999 | // Can't tail call a function called indirectly, or if it has variadic args. | |||
17000 | const Function *Callee = CI->getCalledFunction(); | |||
17001 | if (!Callee || Callee->isVarArg()) | |||
17002 | return false; | |||
17003 | ||||
17004 | // Make sure the callee and caller calling conventions are eligible for tco. | |||
17005 | const Function *Caller = CI->getParent()->getParent(); | |||
17006 | if (!areCallingConvEligibleForTCO_64SVR4(Caller->getCallingConv(), | |||
17007 | CI->getCallingConv())) | |||
17008 | return false; | |||
17009 | ||||
17010 | // If the function is local then we have a good chance at tail-calling it | |||
17011 | return getTargetMachine().shouldAssumeDSOLocal(*Caller->getParent(), Callee); | |||
17012 | } | |||
17013 | ||||
17014 | bool PPCTargetLowering::hasBitPreservingFPLogic(EVT VT) const { | |||
17015 | if (!Subtarget.hasVSX()) | |||
17016 | return false; | |||
17017 | if (Subtarget.hasP9Vector() && VT == MVT::f128) | |||
17018 | return true; | |||
17019 | return VT == MVT::f32 || VT == MVT::f64 || | |||
17020 | VT == MVT::v4f32 || VT == MVT::v2f64; | |||
17021 | } | |||
17022 | ||||
17023 | bool PPCTargetLowering:: | |||
17024 | isMaskAndCmp0FoldingBeneficial(const Instruction &AndI) const { | |||
17025 | const Value *Mask = AndI.getOperand(1); | |||
17026 | // If the mask is suitable for andi. or andis. we should sink the and. | |||
17027 | if (const ConstantInt *CI = dyn_cast<ConstantInt>(Mask)) { | |||
17028 | // Can't handle constants wider than 64-bits. | |||
17029 | if (CI->getBitWidth() > 64) | |||
17030 | return false; | |||
17031 | int64_t ConstVal = CI->getZExtValue(); | |||
17032 | return isUInt<16>(ConstVal) || | |||
17033 | (isUInt<16>(ConstVal >> 16) && !(ConstVal & 0xFFFF)); | |||
17034 | } | |||
17035 | ||||
17036 | // For non-constant masks, we can always use the record-form and. | |||
17037 | return true; | |||
17038 | } | |||
17039 | ||||
17040 | // Transform (abs (sub (zext a), (zext b))) to (vabsd a b 0) | |||
17041 | // Transform (abs (sub (zext a), (zext_invec b))) to (vabsd a b 0) | |||
17042 | // Transform (abs (sub (zext_invec a), (zext_invec b))) to (vabsd a b 0) | |||
17043 | // Transform (abs (sub (zext_invec a), (zext b))) to (vabsd a b 0) | |||
17044 | // Transform (abs (sub a, b) to (vabsd a b 1)) if a & b of type v4i32 | |||
17045 | SDValue PPCTargetLowering::combineABS(SDNode *N, DAGCombinerInfo &DCI) const { | |||
17046 | assert((N->getOpcode() == ISD::ABS) && "Need ABS node here")(static_cast <bool> ((N->getOpcode() == ISD::ABS) && "Need ABS node here") ? void (0) : __assert_fail ("(N->getOpcode() == ISD::ABS) && \"Need ABS node here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17046, __extension__ __PRETTY_FUNCTION__)); | |||
17047 | assert(Subtarget.hasP9Altivec() &&(static_cast <bool> (Subtarget.hasP9Altivec() && "Only combine this when P9 altivec supported!") ? void (0) : __assert_fail ("Subtarget.hasP9Altivec() && \"Only combine this when P9 altivec supported!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17048, __extension__ __PRETTY_FUNCTION__)) | |||
17048 | "Only combine this when P9 altivec supported!")(static_cast <bool> (Subtarget.hasP9Altivec() && "Only combine this when P9 altivec supported!") ? void (0) : __assert_fail ("Subtarget.hasP9Altivec() && \"Only combine this when P9 altivec supported!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17048, __extension__ __PRETTY_FUNCTION__)); | |||
17049 | EVT VT = N->getValueType(0); | |||
17050 | if (VT != MVT::v4i32 && VT != MVT::v8i16 && VT != MVT::v16i8) | |||
17051 | return SDValue(); | |||
17052 | ||||
17053 | SelectionDAG &DAG = DCI.DAG; | |||
17054 | SDLoc dl(N); | |||
17055 | if (N->getOperand(0).getOpcode() == ISD::SUB) { | |||
17056 | // Even for signed integers, if it's known to be positive (as signed | |||
17057 | // integer) due to zero-extended inputs. | |||
17058 | unsigned SubOpcd0 = N->getOperand(0)->getOperand(0).getOpcode(); | |||
17059 | unsigned SubOpcd1 = N->getOperand(0)->getOperand(1).getOpcode(); | |||
17060 | if ((SubOpcd0 == ISD::ZERO_EXTEND || | |||
17061 | SubOpcd0 == ISD::ZERO_EXTEND_VECTOR_INREG) && | |||
17062 | (SubOpcd1 == ISD::ZERO_EXTEND || | |||
17063 | SubOpcd1 == ISD::ZERO_EXTEND_VECTOR_INREG)) { | |||
17064 | return DAG.getNode(PPCISD::VABSD, dl, N->getOperand(0).getValueType(), | |||
17065 | N->getOperand(0)->getOperand(0), | |||
17066 | N->getOperand(0)->getOperand(1), | |||
17067 | DAG.getTargetConstant(0, dl, MVT::i32)); | |||
17068 | } | |||
17069 | ||||
17070 | // For type v4i32, it can be optimized with xvnegsp + vabsduw | |||
17071 | if (N->getOperand(0).getValueType() == MVT::v4i32 && | |||
17072 | N->getOperand(0).hasOneUse()) { | |||
17073 | return DAG.getNode(PPCISD::VABSD, dl, N->getOperand(0).getValueType(), | |||
17074 | N->getOperand(0)->getOperand(0), | |||
17075 | N->getOperand(0)->getOperand(1), | |||
17076 | DAG.getTargetConstant(1, dl, MVT::i32)); | |||
17077 | } | |||
17078 | } | |||
17079 | ||||
17080 | return SDValue(); | |||
17081 | } | |||
17082 | ||||
17083 | // For type v4i32/v8ii16/v16i8, transform | |||
17084 | // from (vselect (setcc a, b, setugt), (sub a, b), (sub b, a)) to (vabsd a, b) | |||
17085 | // from (vselect (setcc a, b, setuge), (sub a, b), (sub b, a)) to (vabsd a, b) | |||
17086 | // from (vselect (setcc a, b, setult), (sub b, a), (sub a, b)) to (vabsd a, b) | |||
17087 | // from (vselect (setcc a, b, setule), (sub b, a), (sub a, b)) to (vabsd a, b) | |||
17088 | SDValue PPCTargetLowering::combineVSelect(SDNode *N, | |||
17089 | DAGCombinerInfo &DCI) const { | |||
17090 | assert((N->getOpcode() == ISD::VSELECT) && "Need VSELECT node here")(static_cast <bool> ((N->getOpcode() == ISD::VSELECT ) && "Need VSELECT node here") ? void (0) : __assert_fail ("(N->getOpcode() == ISD::VSELECT) && \"Need VSELECT node here\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17090, __extension__ __PRETTY_FUNCTION__)); | |||
17091 | assert(Subtarget.hasP9Altivec() &&(static_cast <bool> (Subtarget.hasP9Altivec() && "Only combine this when P9 altivec supported!") ? void (0) : __assert_fail ("Subtarget.hasP9Altivec() && \"Only combine this when P9 altivec supported!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17092, __extension__ __PRETTY_FUNCTION__)) | |||
17092 | "Only combine this when P9 altivec supported!")(static_cast <bool> (Subtarget.hasP9Altivec() && "Only combine this when P9 altivec supported!") ? void (0) : __assert_fail ("Subtarget.hasP9Altivec() && \"Only combine this when P9 altivec supported!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17092, __extension__ __PRETTY_FUNCTION__)); | |||
17093 | ||||
17094 | SelectionDAG &DAG = DCI.DAG; | |||
17095 | SDLoc dl(N); | |||
17096 | SDValue Cond = N->getOperand(0); | |||
17097 | SDValue TrueOpnd = N->getOperand(1); | |||
17098 | SDValue FalseOpnd = N->getOperand(2); | |||
17099 | EVT VT = N->getOperand(1).getValueType(); | |||
17100 | ||||
17101 | if (Cond.getOpcode() != ISD::SETCC || TrueOpnd.getOpcode() != ISD::SUB || | |||
17102 | FalseOpnd.getOpcode() != ISD::SUB) | |||
17103 | return SDValue(); | |||
17104 | ||||
17105 | // ABSD only available for type v4i32/v8i16/v16i8 | |||
17106 | if (VT != MVT::v4i32 && VT != MVT::v8i16 && VT != MVT::v16i8) | |||
17107 | return SDValue(); | |||
17108 | ||||
17109 | // At least to save one more dependent computation | |||
17110 | if (!(Cond.hasOneUse() || TrueOpnd.hasOneUse() || FalseOpnd.hasOneUse())) | |||
17111 | return SDValue(); | |||
17112 | ||||
17113 | ISD::CondCode CC = cast<CondCodeSDNode>(Cond.getOperand(2))->get(); | |||
17114 | ||||
17115 | // Can only handle unsigned comparison here | |||
17116 | switch (CC) { | |||
17117 | default: | |||
17118 | return SDValue(); | |||
17119 | case ISD::SETUGT: | |||
17120 | case ISD::SETUGE: | |||
17121 | break; | |||
17122 | case ISD::SETULT: | |||
17123 | case ISD::SETULE: | |||
17124 | std::swap(TrueOpnd, FalseOpnd); | |||
17125 | break; | |||
17126 | } | |||
17127 | ||||
17128 | SDValue CmpOpnd1 = Cond.getOperand(0); | |||
17129 | SDValue CmpOpnd2 = Cond.getOperand(1); | |||
17130 | ||||
17131 | // SETCC CmpOpnd1 CmpOpnd2 cond | |||
17132 | // TrueOpnd = CmpOpnd1 - CmpOpnd2 | |||
17133 | // FalseOpnd = CmpOpnd2 - CmpOpnd1 | |||
17134 | if (TrueOpnd.getOperand(0) == CmpOpnd1 && | |||
17135 | TrueOpnd.getOperand(1) == CmpOpnd2 && | |||
17136 | FalseOpnd.getOperand(0) == CmpOpnd2 && | |||
17137 | FalseOpnd.getOperand(1) == CmpOpnd1) { | |||
17138 | return DAG.getNode(PPCISD::VABSD, dl, N->getOperand(1).getValueType(), | |||
17139 | CmpOpnd1, CmpOpnd2, | |||
17140 | DAG.getTargetConstant(0, dl, MVT::i32)); | |||
17141 | } | |||
17142 | ||||
17143 | return SDValue(); | |||
17144 | } | |||
17145 | ||||
17146 | /// getAddrModeForFlags - Based on the set of address flags, select the most | |||
17147 | /// optimal instruction format to match by. | |||
17148 | PPC::AddrMode PPCTargetLowering::getAddrModeForFlags(unsigned Flags) const { | |||
17149 | // This is not a node we should be handling here. | |||
17150 | if (Flags == PPC::MOF_None) | |||
17151 | return PPC::AM_None; | |||
17152 | // Unaligned D-Forms are tried first, followed by the aligned D-Forms. | |||
17153 | for (auto FlagSet : AddrModesMap.at(PPC::AM_DForm)) | |||
17154 | if ((Flags & FlagSet) == FlagSet) | |||
17155 | return PPC::AM_DForm; | |||
17156 | for (auto FlagSet : AddrModesMap.at(PPC::AM_DSForm)) | |||
17157 | if ((Flags & FlagSet) == FlagSet) | |||
17158 | return PPC::AM_DSForm; | |||
17159 | for (auto FlagSet : AddrModesMap.at(PPC::AM_DQForm)) | |||
17160 | if ((Flags & FlagSet) == FlagSet) | |||
17161 | return PPC::AM_DQForm; | |||
17162 | // If no other forms are selected, return an X-Form as it is the most | |||
17163 | // general addressing mode. | |||
17164 | return PPC::AM_XForm; | |||
17165 | } | |||
17166 | ||||
17167 | /// Set alignment flags based on whether or not the Frame Index is aligned. | |||
17168 | /// Utilized when computing flags for address computation when selecting | |||
17169 | /// load and store instructions. | |||
17170 | static void setAlignFlagsForFI(SDValue N, unsigned &FlagSet, | |||
17171 | SelectionDAG &DAG) { | |||
17172 | bool IsAdd = ((N.getOpcode() == ISD::ADD) || (N.getOpcode() == ISD::OR)); | |||
17173 | FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(IsAdd ? N.getOperand(0) : N); | |||
17174 | if (!FI) | |||
17175 | return; | |||
17176 | const MachineFrameInfo &MFI = DAG.getMachineFunction().getFrameInfo(); | |||
17177 | unsigned FrameIndexAlign = MFI.getObjectAlign(FI->getIndex()).value(); | |||
17178 | // If this is (add $FI, $S16Imm), the alignment flags are already set | |||
17179 | // based on the immediate. We just need to clear the alignment flags | |||
17180 | // if the FI alignment is weaker. | |||
17181 | if ((FrameIndexAlign % 4) != 0) | |||
17182 | FlagSet &= ~PPC::MOF_RPlusSImm16Mult4; | |||
17183 | if ((FrameIndexAlign % 16) != 0) | |||
17184 | FlagSet &= ~PPC::MOF_RPlusSImm16Mult16; | |||
17185 | // If the address is a plain FrameIndex, set alignment flags based on | |||
17186 | // FI alignment. | |||
17187 | if (!IsAdd) { | |||
17188 | if ((FrameIndexAlign % 4) == 0) | |||
17189 | FlagSet |= PPC::MOF_RPlusSImm16Mult4; | |||
17190 | if ((FrameIndexAlign % 16) == 0) | |||
17191 | FlagSet |= PPC::MOF_RPlusSImm16Mult16; | |||
17192 | } | |||
17193 | } | |||
17194 | ||||
17195 | /// Given a node, compute flags that are used for address computation when | |||
17196 | /// selecting load and store instructions. The flags computed are stored in | |||
17197 | /// FlagSet. This function takes into account whether the node is a constant, | |||
17198 | /// an ADD, OR, or a constant, and computes the address flags accordingly. | |||
17199 | static void computeFlagsForAddressComputation(SDValue N, unsigned &FlagSet, | |||
17200 | SelectionDAG &DAG) { | |||
17201 | // Set the alignment flags for the node depending on if the node is | |||
17202 | // 4-byte or 16-byte aligned. | |||
17203 | auto SetAlignFlagsForImm = [&](uint64_t Imm) { | |||
17204 | if ((Imm & 0x3) == 0) | |||
17205 | FlagSet |= PPC::MOF_RPlusSImm16Mult4; | |||
17206 | if ((Imm & 0xf) == 0) | |||
17207 | FlagSet |= PPC::MOF_RPlusSImm16Mult16; | |||
17208 | }; | |||
17209 | ||||
17210 | if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N)) { | |||
17211 | // All 32-bit constants can be computed as LIS + Disp. | |||
17212 | const APInt &ConstImm = CN->getAPIntValue(); | |||
17213 | if (ConstImm.isSignedIntN(32)) { // Flag to handle 32-bit constants. | |||
17214 | FlagSet |= PPC::MOF_AddrIsSImm32; | |||
17215 | SetAlignFlagsForImm(ConstImm.getZExtValue()); | |||
17216 | setAlignFlagsForFI(N, FlagSet, DAG); | |||
17217 | } | |||
17218 | if (ConstImm.isSignedIntN(34)) // Flag to handle 34-bit constants. | |||
17219 | FlagSet |= PPC::MOF_RPlusSImm34; | |||
17220 | else // Let constant materialization handle large constants. | |||
17221 | FlagSet |= PPC::MOF_NotAddNorCst; | |||
17222 | } else if (N.getOpcode() == ISD::ADD || provablyDisjointOr(DAG, N)) { | |||
17223 | // This address can be represented as an addition of: | |||
17224 | // - Register + Imm16 (possibly a multiple of 4/16) | |||
17225 | // - Register + Imm34 | |||
17226 | // - Register + PPCISD::Lo | |||
17227 | // - Register + Register | |||
17228 | // In any case, we won't have to match this as Base + Zero. | |||
17229 | SDValue RHS = N.getOperand(1); | |||
17230 | if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(RHS)) { | |||
17231 | const APInt &ConstImm = CN->getAPIntValue(); | |||
17232 | if (ConstImm.isSignedIntN(16)) { | |||
17233 | FlagSet |= PPC::MOF_RPlusSImm16; // Signed 16-bit immediates. | |||
17234 | SetAlignFlagsForImm(ConstImm.getZExtValue()); | |||
17235 | setAlignFlagsForFI(N, FlagSet, DAG); | |||
17236 | } | |||
17237 | if (ConstImm.isSignedIntN(34)) | |||
17238 | FlagSet |= PPC::MOF_RPlusSImm34; // Signed 34-bit immediates. | |||
17239 | else | |||
17240 | FlagSet |= PPC::MOF_RPlusR; // Register. | |||
17241 | } else if (RHS.getOpcode() == PPCISD::Lo && | |||
17242 | !cast<ConstantSDNode>(RHS.getOperand(1))->getZExtValue()) | |||
17243 | FlagSet |= PPC::MOF_RPlusLo; // PPCISD::Lo. | |||
17244 | else | |||
17245 | FlagSet |= PPC::MOF_RPlusR; | |||
17246 | } else { // The address computation is not a constant or an addition. | |||
17247 | setAlignFlagsForFI(N, FlagSet, DAG); | |||
17248 | FlagSet |= PPC::MOF_NotAddNorCst; | |||
17249 | } | |||
17250 | } | |||
17251 | ||||
17252 | /// computeMOFlags - Given a node N and it's Parent (a MemSDNode), compute | |||
17253 | /// the address flags of the load/store instruction that is to be matched. | |||
17254 | unsigned PPCTargetLowering::computeMOFlags(const SDNode *Parent, SDValue N, | |||
17255 | SelectionDAG &DAG) const { | |||
17256 | unsigned FlagSet = PPC::MOF_None; | |||
17257 | ||||
17258 | // Compute subtarget flags. | |||
17259 | if (!Subtarget.hasP9Vector()) | |||
17260 | FlagSet |= PPC::MOF_SubtargetBeforeP9; | |||
17261 | else { | |||
17262 | FlagSet |= PPC::MOF_SubtargetP9; | |||
17263 | if (Subtarget.hasPrefixInstrs()) | |||
17264 | FlagSet |= PPC::MOF_SubtargetP10; | |||
17265 | } | |||
17266 | if (Subtarget.hasSPE()) | |||
17267 | FlagSet |= PPC::MOF_SubtargetSPE; | |||
17268 | ||||
17269 | // Mark this as something we don't want to handle here if it is atomic | |||
17270 | // or pre-increment instruction. | |||
17271 | if (const LSBaseSDNode *LSB = dyn_cast<LSBaseSDNode>(Parent)) | |||
17272 | if (LSB->isIndexed()) | |||
17273 | return PPC::MOF_None; | |||
17274 | ||||
17275 | // Compute in-memory type flags. This is based on if there are scalars, | |||
17276 | // floats or vectors. | |||
17277 | const MemSDNode *MN = dyn_cast<MemSDNode>(Parent); | |||
17278 | assert(MN && "Parent should be a MemSDNode!")(static_cast <bool> (MN && "Parent should be a MemSDNode!" ) ? void (0) : __assert_fail ("MN && \"Parent should be a MemSDNode!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17278, __extension__ __PRETTY_FUNCTION__)); | |||
17279 | EVT MemVT = MN->getMemoryVT(); | |||
17280 | unsigned Size = MemVT.getSizeInBits(); | |||
17281 | if (MemVT.isScalarInteger()) { | |||
17282 | assert(Size <= 64 && "Not expecting scalar integers larger than 8 bytes!")(static_cast <bool> (Size <= 64 && "Not expecting scalar integers larger than 8 bytes!" ) ? void (0) : __assert_fail ("Size <= 64 && \"Not expecting scalar integers larger than 8 bytes!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17282, __extension__ __PRETTY_FUNCTION__)); | |||
17283 | if (Size < 32) | |||
17284 | FlagSet |= PPC::MOF_SubWordInt; | |||
17285 | else if (Size == 32) | |||
17286 | FlagSet |= PPC::MOF_WordInt; | |||
17287 | else | |||
17288 | FlagSet |= PPC::MOF_DoubleWordInt; | |||
17289 | } else if (MemVT.isVector() && !MemVT.isFloatingPoint()) { // Integer vectors. | |||
17290 | if (Size == 128) | |||
17291 | FlagSet |= PPC::MOF_Vector; | |||
17292 | else if (Size == 256) | |||
17293 | FlagSet |= PPC::MOF_Vector256; | |||
17294 | else | |||
17295 | llvm_unreachable("Not expecting illegal vectors!")::llvm::llvm_unreachable_internal("Not expecting illegal vectors!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17295); | |||
17296 | } else { // Floating point type: can be scalar, f128 or vector types. | |||
17297 | if (Size == 32 || Size == 64) | |||
17298 | FlagSet |= PPC::MOF_ScalarFloat; | |||
17299 | else if (MemVT == MVT::f128 || MemVT.isVector()) | |||
17300 | FlagSet |= PPC::MOF_Vector; | |||
17301 | else | |||
17302 | llvm_unreachable("Not expecting illegal scalar floats!")::llvm::llvm_unreachable_internal("Not expecting illegal scalar floats!" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17302); | |||
17303 | } | |||
17304 | ||||
17305 | // Compute flags for address computation. | |||
17306 | computeFlagsForAddressComputation(N, FlagSet, DAG); | |||
17307 | ||||
17308 | // Compute type extension flags. | |||
17309 | if (const LoadSDNode *LN = dyn_cast<LoadSDNode>(Parent)) { | |||
17310 | switch (LN->getExtensionType()) { | |||
17311 | case ISD::SEXTLOAD: | |||
17312 | FlagSet |= PPC::MOF_SExt; | |||
17313 | break; | |||
17314 | case ISD::EXTLOAD: | |||
17315 | case ISD::ZEXTLOAD: | |||
17316 | FlagSet |= PPC::MOF_ZExt; | |||
17317 | break; | |||
17318 | case ISD::NON_EXTLOAD: | |||
17319 | FlagSet |= PPC::MOF_NoExt; | |||
17320 | break; | |||
17321 | } | |||
17322 | } else | |||
17323 | FlagSet |= PPC::MOF_NoExt; | |||
17324 | ||||
17325 | // For integers, no extension is the same as zero extension. | |||
17326 | // We set the extension mode to zero extension so we don't have | |||
17327 | // to add separate entries in AddrModesMap for loads and stores. | |||
17328 | if (MemVT.isScalarInteger() && (FlagSet & PPC::MOF_NoExt)) { | |||
17329 | FlagSet |= PPC::MOF_ZExt; | |||
17330 | FlagSet &= ~PPC::MOF_NoExt; | |||
17331 | } | |||
17332 | ||||
17333 | // If we don't have prefixed instructions, 34-bit constants should be | |||
17334 | // treated as PPC::MOF_NotAddNorCst so they can match D-Forms. | |||
17335 | bool IsNonP1034BitConst = | |||
17336 | ((PPC::MOF_RPlusSImm34 | PPC::MOF_AddrIsSImm32 | PPC::MOF_SubtargetP10) & | |||
17337 | FlagSet) == PPC::MOF_RPlusSImm34; | |||
17338 | if (N.getOpcode() != ISD::ADD && N.getOpcode() != ISD::OR && | |||
17339 | IsNonP1034BitConst) | |||
17340 | FlagSet |= PPC::MOF_NotAddNorCst; | |||
17341 | ||||
17342 | return FlagSet; | |||
17343 | } | |||
17344 | ||||
17345 | /// SelectForceXFormMode - Given the specified address, force it to be | |||
17346 | /// represented as an indexed [r+r] operation (an XForm instruction). | |||
17347 | PPC::AddrMode PPCTargetLowering::SelectForceXFormMode(SDValue N, SDValue &Disp, | |||
17348 | SDValue &Base, | |||
17349 | SelectionDAG &DAG) const { | |||
17350 | ||||
17351 | PPC::AddrMode Mode = PPC::AM_XForm; | |||
17352 | int16_t ForceXFormImm = 0; | |||
17353 | if (provablyDisjointOr(DAG, N) && | |||
17354 | !isIntS16Immediate(N.getOperand(1), ForceXFormImm)) { | |||
17355 | Disp = N.getOperand(0); | |||
17356 | Base = N.getOperand(1); | |||
17357 | return Mode; | |||
17358 | } | |||
17359 | ||||
17360 | // If the address is the result of an add, we will utilize the fact that the | |||
17361 | // address calculation includes an implicit add. However, we can reduce | |||
17362 | // register pressure if we do not materialize a constant just for use as the | |||
17363 | // index register. We only get rid of the add if it is not an add of a | |||
17364 | // value and a 16-bit signed constant and both have a single use. | |||
17365 | if (N.getOpcode() == ISD::ADD && | |||
17366 | (!isIntS16Immediate(N.getOperand(1), ForceXFormImm) || | |||
17367 | !N.getOperand(1).hasOneUse() || !N.getOperand(0).hasOneUse())) { | |||
17368 | Disp = N.getOperand(0); | |||
17369 | Base = N.getOperand(1); | |||
17370 | return Mode; | |||
17371 | } | |||
17372 | ||||
17373 | // Otherwise, use R0 as the base register. | |||
17374 | Disp = DAG.getRegister(Subtarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, | |||
17375 | N.getValueType()); | |||
17376 | Base = N; | |||
17377 | ||||
17378 | return Mode; | |||
17379 | } | |||
17380 | ||||
17381 | // If we happen to match to an aligned D-Form, check if the Frame Index is | |||
17382 | // adequately aligned. If it is not, reset the mode to match to X-Form. | |||
17383 | static void setXFormForUnalignedFI(SDValue N, unsigned Flags, | |||
17384 | PPC::AddrMode &Mode) { | |||
17385 | if (!isa<FrameIndexSDNode>(N)) | |||
17386 | return; | |||
17387 | if ((Mode == PPC::AM_DSForm && !(Flags & PPC::MOF_RPlusSImm16Mult4)) || | |||
17388 | (Mode == PPC::AM_DQForm && !(Flags & PPC::MOF_RPlusSImm16Mult16))) | |||
17389 | Mode = PPC::AM_XForm; | |||
17390 | } | |||
17391 | ||||
17392 | /// SelectOptimalAddrMode - Based on a node N and it's Parent (a MemSDNode), | |||
17393 | /// compute the address flags of the node, get the optimal address mode based | |||
17394 | /// on the flags, and set the Base and Disp based on the address mode. | |||
17395 | PPC::AddrMode PPCTargetLowering::SelectOptimalAddrMode(const SDNode *Parent, | |||
17396 | SDValue N, SDValue &Disp, | |||
17397 | SDValue &Base, | |||
17398 | SelectionDAG &DAG, | |||
17399 | MaybeAlign Align) const { | |||
17400 | SDLoc DL(Parent); | |||
17401 | ||||
17402 | // Compute the address flags. | |||
17403 | unsigned Flags = computeMOFlags(Parent, N, DAG); | |||
17404 | ||||
17405 | // Get the optimal address mode based on the Flags. | |||
17406 | PPC::AddrMode Mode = getAddrModeForFlags(Flags); | |||
17407 | ||||
17408 | // If the address mode is DS-Form or DQ-Form, check if the FI is aligned. | |||
17409 | // Select an X-Form load if it is not. | |||
17410 | setXFormForUnalignedFI(N, Flags, Mode); | |||
17411 | ||||
17412 | // Set Base and Disp accordingly depending on the address mode. | |||
17413 | switch (Mode) { | |||
| ||||
17414 | case PPC::AM_DForm: | |||
17415 | case PPC::AM_DSForm: | |||
17416 | case PPC::AM_DQForm: { | |||
17417 | // This is a register plus a 16-bit immediate. The base will be the | |||
17418 | // register and the displacement will be the immediate unless it | |||
17419 | // isn't sufficiently aligned. | |||
17420 | if (Flags & PPC::MOF_RPlusSImm16) { | |||
17421 | SDValue Op0 = N.getOperand(0); | |||
17422 | SDValue Op1 = N.getOperand(1); | |||
17423 | int16_t Imm = cast<ConstantSDNode>(Op1)->getAPIntValue().getZExtValue(); | |||
17424 | if (!Align || isAligned(*Align, Imm)) { | |||
17425 | Disp = DAG.getTargetConstant(Imm, DL, N.getValueType()); | |||
17426 | Base = Op0; | |||
17427 | if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Op0)) { | |||
17428 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
17429 | fixupFuncForFI(DAG, FI->getIndex(), N.getValueType()); | |||
17430 | } | |||
17431 | break; | |||
17432 | } | |||
17433 | } | |||
17434 | // This is a register plus the @lo relocation. The base is the register | |||
17435 | // and the displacement is the global address. | |||
17436 | else if (Flags & PPC::MOF_RPlusLo) { | |||
17437 | Disp = N.getOperand(1).getOperand(0); // The global address. | |||
17438 | assert(Disp.getOpcode() == ISD::TargetGlobalAddress ||(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17441, __extension__ __PRETTY_FUNCTION__)) | |||
17439 | Disp.getOpcode() == ISD::TargetGlobalTLSAddress ||(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17441, __extension__ __PRETTY_FUNCTION__)) | |||
17440 | Disp.getOpcode() == ISD::TargetConstantPool ||(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17441, __extension__ __PRETTY_FUNCTION__)) | |||
17441 | Disp.getOpcode() == ISD::TargetJumpTable)(static_cast <bool> (Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode () == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable ) ? void (0) : __assert_fail ("Disp.getOpcode() == ISD::TargetGlobalAddress || Disp.getOpcode() == ISD::TargetGlobalTLSAddress || Disp.getOpcode() == ISD::TargetConstantPool || Disp.getOpcode() == ISD::TargetJumpTable" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17441, __extension__ __PRETTY_FUNCTION__)); | |||
17442 | Base = N.getOperand(0); | |||
17443 | break; | |||
17444 | } | |||
17445 | // This is a constant address at most 32 bits. The base will be | |||
17446 | // zero or load-immediate-shifted and the displacement will be | |||
17447 | // the low 16 bits of the address. | |||
17448 | else if (Flags & PPC::MOF_AddrIsSImm32) { | |||
17449 | auto *CN = cast<ConstantSDNode>(N); | |||
17450 | EVT CNType = CN->getValueType(0); | |||
17451 | uint64_t CNImm = CN->getZExtValue(); | |||
17452 | // If this address fits entirely in a 16-bit sext immediate field, codegen | |||
17453 | // this as "d, 0". | |||
17454 | int16_t Imm; | |||
17455 | if (isIntS16Immediate(CN, Imm) && (!Align || isAligned(*Align, Imm))) { | |||
17456 | Disp = DAG.getTargetConstant(Imm, DL, CNType); | |||
17457 | Base = DAG.getRegister(Subtarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, | |||
17458 | CNType); | |||
17459 | break; | |||
17460 | } | |||
17461 | // Handle 32-bit sext immediate with LIS + Addr mode. | |||
17462 | if ((CNType == MVT::i32 || isInt<32>(CNImm)) && | |||
17463 | (!Align || isAligned(*Align, CNImm))) { | |||
17464 | int32_t Addr = (int32_t)CNImm; | |||
17465 | // Otherwise, break this down into LIS + Disp. | |||
17466 | Disp = DAG.getTargetConstant((int16_t)Addr, DL, MVT::i32); | |||
17467 | Base = | |||
17468 | DAG.getTargetConstant((Addr - (int16_t)Addr) >> 16, DL, MVT::i32); | |||
17469 | uint32_t LIS = CNType == MVT::i32 ? PPC::LIS : PPC::LIS8; | |||
17470 | Base = SDValue(DAG.getMachineNode(LIS, DL, CNType, Base), 0); | |||
17471 | break; | |||
17472 | } | |||
17473 | } | |||
17474 | // Otherwise, the PPC:MOF_NotAdd flag is set. Load/Store is Non-foldable. | |||
17475 | Disp = DAG.getTargetConstant(0, DL, getPointerTy(DAG.getDataLayout())); | |||
17476 | if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N)) { | |||
17477 | Base = DAG.getTargetFrameIndex(FI->getIndex(), N.getValueType()); | |||
17478 | fixupFuncForFI(DAG, FI->getIndex(), N.getValueType()); | |||
17479 | } else | |||
17480 | Base = N; | |||
17481 | break; | |||
17482 | } | |||
17483 | case PPC::AM_None: | |||
17484 | break; | |||
17485 | default: { // By default, X-Form is always available to be selected. | |||
17486 | // When a frame index is not aligned, we also match by XForm. | |||
17487 | FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N); | |||
17488 | Base = FI ? N : N.getOperand(1); | |||
17489 | Disp = FI ? DAG.getRegister(Subtarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, | |||
17490 | N.getValueType()) | |||
17491 | : N.getOperand(0); | |||
17492 | break; | |||
17493 | } | |||
17494 | } | |||
17495 | return Mode; | |||
17496 | } | |||
17497 | ||||
17498 | CCAssignFn *PPCTargetLowering::ccAssignFnForCall(CallingConv::ID CC, | |||
17499 | bool Return, | |||
17500 | bool IsVarArg) const { | |||
17501 | switch (CC) { | |||
17502 | case CallingConv::Cold: | |||
17503 | return (Return ? RetCC_PPC_Cold : CC_PPC64_ELF_FIS); | |||
17504 | default: | |||
17505 | return CC_PPC64_ELF_FIS; | |||
17506 | } | |||
17507 | } | |||
17508 | ||||
17509 | TargetLowering::AtomicExpansionKind | |||
17510 | PPCTargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const { | |||
17511 | unsigned Size = AI->getType()->getPrimitiveSizeInBits(); | |||
17512 | if (EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() && Size == 128) | |||
17513 | return AtomicExpansionKind::MaskedIntrinsic; | |||
17514 | return TargetLowering::shouldExpandAtomicRMWInIR(AI); | |||
17515 | } | |||
17516 | ||||
17517 | TargetLowering::AtomicExpansionKind | |||
17518 | PPCTargetLowering::shouldExpandAtomicCmpXchgInIR(AtomicCmpXchgInst *AI) const { | |||
17519 | unsigned Size = AI->getPointerOperand() | |||
17520 | ->getType() | |||
17521 | ->getPointerElementType() | |||
17522 | ->getPrimitiveSizeInBits(); | |||
17523 | if (EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() && Size == 128) | |||
17524 | return AtomicExpansionKind::MaskedIntrinsic; | |||
17525 | return TargetLowering::shouldExpandAtomicCmpXchgInIR(AI); | |||
17526 | } | |||
17527 | ||||
17528 | static Intrinsic::ID | |||
17529 | getIntrinsicForAtomicRMWBinOp128(AtomicRMWInst::BinOp BinOp) { | |||
17530 | switch (BinOp) { | |||
17531 | default: | |||
17532 | llvm_unreachable("Unexpected AtomicRMW BinOp")::llvm::llvm_unreachable_internal("Unexpected AtomicRMW BinOp" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17532); | |||
17533 | case AtomicRMWInst::Xchg: | |||
17534 | return Intrinsic::ppc_atomicrmw_xchg_i128; | |||
17535 | case AtomicRMWInst::Add: | |||
17536 | return Intrinsic::ppc_atomicrmw_add_i128; | |||
17537 | case AtomicRMWInst::Sub: | |||
17538 | return Intrinsic::ppc_atomicrmw_sub_i128; | |||
17539 | case AtomicRMWInst::And: | |||
17540 | return Intrinsic::ppc_atomicrmw_and_i128; | |||
17541 | case AtomicRMWInst::Or: | |||
17542 | return Intrinsic::ppc_atomicrmw_or_i128; | |||
17543 | case AtomicRMWInst::Xor: | |||
17544 | return Intrinsic::ppc_atomicrmw_xor_i128; | |||
17545 | case AtomicRMWInst::Nand: | |||
17546 | return Intrinsic::ppc_atomicrmw_nand_i128; | |||
17547 | } | |||
17548 | } | |||
17549 | ||||
17550 | Value *PPCTargetLowering::emitMaskedAtomicRMWIntrinsic( | |||
17551 | IRBuilderBase &Builder, AtomicRMWInst *AI, Value *AlignedAddr, Value *Incr, | |||
17552 | Value *Mask, Value *ShiftAmt, AtomicOrdering Ord) const { | |||
17553 | assert(EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() &&(static_cast <bool> (EnableQuadwordAtomics && Subtarget .hasQuadwordAtomics() && "Only support quadword now") ? void (0) : __assert_fail ("EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() && \"Only support quadword now\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17554, __extension__ __PRETTY_FUNCTION__)) | |||
17554 | "Only support quadword now")(static_cast <bool> (EnableQuadwordAtomics && Subtarget .hasQuadwordAtomics() && "Only support quadword now") ? void (0) : __assert_fail ("EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() && \"Only support quadword now\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17554, __extension__ __PRETTY_FUNCTION__)); | |||
17555 | Module *M = Builder.GetInsertBlock()->getParent()->getParent(); | |||
17556 | Type *ValTy = cast<PointerType>(AlignedAddr->getType())->getElementType(); | |||
17557 | assert(ValTy->getPrimitiveSizeInBits() == 128)(static_cast <bool> (ValTy->getPrimitiveSizeInBits() == 128) ? void (0) : __assert_fail ("ValTy->getPrimitiveSizeInBits() == 128" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17557, __extension__ __PRETTY_FUNCTION__)); | |||
17558 | Function *RMW = Intrinsic::getDeclaration( | |||
17559 | M, getIntrinsicForAtomicRMWBinOp128(AI->getOperation())); | |||
17560 | Type *Int64Ty = Type::getInt64Ty(M->getContext()); | |||
17561 | Value *IncrLo = Builder.CreateTrunc(Incr, Int64Ty, "incr_lo"); | |||
17562 | Value *IncrHi = | |||
17563 | Builder.CreateTrunc(Builder.CreateLShr(Incr, 64), Int64Ty, "incr_hi"); | |||
17564 | Value *Addr = | |||
17565 | Builder.CreateBitCast(AlignedAddr, Type::getInt8PtrTy(M->getContext())); | |||
17566 | Value *LoHi = Builder.CreateCall(RMW, {Addr, IncrLo, IncrHi}); | |||
17567 | Value *Lo = Builder.CreateExtractValue(LoHi, 0, "lo"); | |||
17568 | Value *Hi = Builder.CreateExtractValue(LoHi, 1, "hi"); | |||
17569 | Lo = Builder.CreateZExt(Lo, ValTy, "lo64"); | |||
17570 | Hi = Builder.CreateZExt(Hi, ValTy, "hi64"); | |||
17571 | return Builder.CreateOr( | |||
17572 | Lo, Builder.CreateShl(Hi, ConstantInt::get(ValTy, 64)), "val64"); | |||
17573 | } | |||
17574 | ||||
17575 | Value *PPCTargetLowering::emitMaskedAtomicCmpXchgIntrinsic( | |||
17576 | IRBuilderBase &Builder, AtomicCmpXchgInst *CI, Value *AlignedAddr, | |||
17577 | Value *CmpVal, Value *NewVal, Value *Mask, AtomicOrdering Ord) const { | |||
17578 | assert(EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() &&(static_cast <bool> (EnableQuadwordAtomics && Subtarget .hasQuadwordAtomics() && "Only support quadword now") ? void (0) : __assert_fail ("EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() && \"Only support quadword now\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17579, __extension__ __PRETTY_FUNCTION__)) | |||
17579 | "Only support quadword now")(static_cast <bool> (EnableQuadwordAtomics && Subtarget .hasQuadwordAtomics() && "Only support quadword now") ? void (0) : __assert_fail ("EnableQuadwordAtomics && Subtarget.hasQuadwordAtomics() && \"Only support quadword now\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17579, __extension__ __PRETTY_FUNCTION__)); | |||
17580 | Module *M = Builder.GetInsertBlock()->getParent()->getParent(); | |||
17581 | Type *ValTy = cast<PointerType>(AlignedAddr->getType())->getElementType(); | |||
17582 | assert(ValTy->getPrimitiveSizeInBits() == 128)(static_cast <bool> (ValTy->getPrimitiveSizeInBits() == 128) ? void (0) : __assert_fail ("ValTy->getPrimitiveSizeInBits() == 128" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/lib/Target/PowerPC/PPCISelLowering.cpp" , 17582, __extension__ __PRETTY_FUNCTION__)); | |||
17583 | Function *IntCmpXchg = | |||
17584 | Intrinsic::getDeclaration(M, Intrinsic::ppc_cmpxchg_i128); | |||
17585 | Type *Int64Ty = Type::getInt64Ty(M->getContext()); | |||
17586 | Value *CmpLo = Builder.CreateTrunc(CmpVal, Int64Ty, "cmp_lo"); | |||
17587 | Value *CmpHi = | |||
17588 | Builder.CreateTrunc(Builder.CreateLShr(CmpVal, 64), Int64Ty, "cmp_hi"); | |||
17589 | Value *NewLo = Builder.CreateTrunc(NewVal, Int64Ty, "new_lo"); | |||
17590 | Value *NewHi = | |||
17591 | Builder.CreateTrunc(Builder.CreateLShr(NewVal, 64), Int64Ty, "new_hi"); | |||
17592 | Value *Addr = | |||
17593 | Builder.CreateBitCast(AlignedAddr, Type::getInt8PtrTy(M->getContext())); | |||
17594 | emitLeadingFence(Builder, CI, Ord); | |||
17595 | Value *LoHi = | |||
17596 | Builder.CreateCall(IntCmpXchg, {Addr, CmpLo, CmpHi, NewLo, NewHi}); | |||
17597 | emitTrailingFence(Builder, CI, Ord); | |||
17598 | Value *Lo = Builder.CreateExtractValue(LoHi, 0, "lo"); | |||
17599 | Value *Hi = Builder.CreateExtractValue(LoHi, 1, "hi"); | |||
17600 | Lo = Builder.CreateZExt(Lo, ValTy, "lo64"); | |||
17601 | Hi = Builder.CreateZExt(Hi, ValTy, "hi64"); | |||
17602 | return Builder.CreateOr( | |||
17603 | Lo, Builder.CreateShl(Hi, ConstantInt::get(ValTy, 64)), "val64"); | |||
17604 | } |
1 | //===- llvm/Support/Casting.h - Allow flexible, checked, casts --*- C++ -*-===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | // |
9 | // This file defines the isa<X>(), cast<X>(), dyn_cast<X>(), cast_or_null<X>(), |
10 | // and dyn_cast_or_null<X>() templates. |
11 | // |
12 | //===----------------------------------------------------------------------===// |
13 | |
14 | #ifndef LLVM_SUPPORT_CASTING_H |
15 | #define LLVM_SUPPORT_CASTING_H |
16 | |
17 | #include "llvm/Support/Compiler.h" |
18 | #include "llvm/Support/type_traits.h" |
19 | #include <cassert> |
20 | #include <memory> |
21 | #include <type_traits> |
22 | |
23 | namespace llvm { |
24 | |
25 | //===----------------------------------------------------------------------===// |
26 | // isa<x> Support Templates |
27 | //===----------------------------------------------------------------------===// |
28 | |
29 | // Define a template that can be specialized by smart pointers to reflect the |
30 | // fact that they are automatically dereferenced, and are not involved with the |
31 | // template selection process... the default implementation is a noop. |
32 | // |
33 | template<typename From> struct simplify_type { |
34 | using SimpleType = From; // The real type this represents... |
35 | |
36 | // An accessor to get the real value... |
37 | static SimpleType &getSimplifiedValue(From &Val) { return Val; } |
38 | }; |
39 | |
40 | template<typename From> struct simplify_type<const From> { |
41 | using NonConstSimpleType = typename simplify_type<From>::SimpleType; |
42 | using SimpleType = |
43 | typename add_const_past_pointer<NonConstSimpleType>::type; |
44 | using RetType = |
45 | typename add_lvalue_reference_if_not_pointer<SimpleType>::type; |
46 | |
47 | static RetType getSimplifiedValue(const From& Val) { |
48 | return simplify_type<From>::getSimplifiedValue(const_cast<From&>(Val)); |
49 | } |
50 | }; |
51 | |
52 | // The core of the implementation of isa<X> is here; To and From should be |
53 | // the names of classes. This template can be specialized to customize the |
54 | // implementation of isa<> without rewriting it from scratch. |
55 | template <typename To, typename From, typename Enabler = void> |
56 | struct isa_impl { |
57 | static inline bool doit(const From &Val) { |
58 | return To::classof(&Val); |
59 | } |
60 | }; |
61 | |
62 | /// Always allow upcasts, and perform no dynamic check for them. |
63 | template <typename To, typename From> |
64 | struct isa_impl<To, From, std::enable_if_t<std::is_base_of<To, From>::value>> { |
65 | static inline bool doit(const From &) { return true; } |
66 | }; |
67 | |
68 | template <typename To, typename From> struct isa_impl_cl { |
69 | static inline bool doit(const From &Val) { |
70 | return isa_impl<To, From>::doit(Val); |
71 | } |
72 | }; |
73 | |
74 | template <typename To, typename From> struct isa_impl_cl<To, const From> { |
75 | static inline bool doit(const From &Val) { |
76 | return isa_impl<To, From>::doit(Val); |
77 | } |
78 | }; |
79 | |
80 | template <typename To, typename From> |
81 | struct isa_impl_cl<To, const std::unique_ptr<From>> { |
82 | static inline bool doit(const std::unique_ptr<From> &Val) { |
83 | assert(Val && "isa<> used on a null pointer")(static_cast <bool> (Val && "isa<> used on a null pointer" ) ? void (0) : __assert_fail ("Val && \"isa<> used on a null pointer\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 83, __extension__ __PRETTY_FUNCTION__)); |
84 | return isa_impl_cl<To, From>::doit(*Val); |
85 | } |
86 | }; |
87 | |
88 | template <typename To, typename From> struct isa_impl_cl<To, From*> { |
89 | static inline bool doit(const From *Val) { |
90 | assert(Val && "isa<> used on a null pointer")(static_cast <bool> (Val && "isa<> used on a null pointer" ) ? void (0) : __assert_fail ("Val && \"isa<> used on a null pointer\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 90, __extension__ __PRETTY_FUNCTION__)); |
91 | return isa_impl<To, From>::doit(*Val); |
92 | } |
93 | }; |
94 | |
95 | template <typename To, typename From> struct isa_impl_cl<To, From*const> { |
96 | static inline bool doit(const From *Val) { |
97 | assert(Val && "isa<> used on a null pointer")(static_cast <bool> (Val && "isa<> used on a null pointer" ) ? void (0) : __assert_fail ("Val && \"isa<> used on a null pointer\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 97, __extension__ __PRETTY_FUNCTION__)); |
98 | return isa_impl<To, From>::doit(*Val); |
99 | } |
100 | }; |
101 | |
102 | template <typename To, typename From> struct isa_impl_cl<To, const From*> { |
103 | static inline bool doit(const From *Val) { |
104 | assert(Val && "isa<> used on a null pointer")(static_cast <bool> (Val && "isa<> used on a null pointer" ) ? void (0) : __assert_fail ("Val && \"isa<> used on a null pointer\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 104, __extension__ __PRETTY_FUNCTION__)); |
105 | return isa_impl<To, From>::doit(*Val); |
106 | } |
107 | }; |
108 | |
109 | template <typename To, typename From> struct isa_impl_cl<To, const From*const> { |
110 | static inline bool doit(const From *Val) { |
111 | assert(Val && "isa<> used on a null pointer")(static_cast <bool> (Val && "isa<> used on a null pointer" ) ? void (0) : __assert_fail ("Val && \"isa<> used on a null pointer\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 111, __extension__ __PRETTY_FUNCTION__)); |
112 | return isa_impl<To, From>::doit(*Val); |
113 | } |
114 | }; |
115 | |
116 | template<typename To, typename From, typename SimpleFrom> |
117 | struct isa_impl_wrap { |
118 | // When From != SimplifiedType, we can simplify the type some more by using |
119 | // the simplify_type template. |
120 | static bool doit(const From &Val) { |
121 | return isa_impl_wrap<To, SimpleFrom, |
122 | typename simplify_type<SimpleFrom>::SimpleType>::doit( |
123 | simplify_type<const From>::getSimplifiedValue(Val)); |
124 | } |
125 | }; |
126 | |
127 | template<typename To, typename FromTy> |
128 | struct isa_impl_wrap<To, FromTy, FromTy> { |
129 | // When From == SimpleType, we are as simple as we are going to get. |
130 | static bool doit(const FromTy &Val) { |
131 | return isa_impl_cl<To,FromTy>::doit(Val); |
132 | } |
133 | }; |
134 | |
135 | // isa<X> - Return true if the parameter to the template is an instance of one |
136 | // of the template type arguments. Used like this: |
137 | // |
138 | // if (isa<Type>(myVal)) { ... } |
139 | // if (isa<Type0, Type1, Type2>(myVal)) { ... } |
140 | // |
141 | template <class X, class Y> LLVM_NODISCARD[[clang::warn_unused_result]] inline bool isa(const Y &Val) { |
142 | return isa_impl_wrap<X, const Y, |
143 | typename simplify_type<const Y>::SimpleType>::doit(Val); |
144 | } |
145 | |
146 | template <typename First, typename Second, typename... Rest, typename Y> |
147 | LLVM_NODISCARD[[clang::warn_unused_result]] inline bool isa(const Y &Val) { |
148 | return isa<First>(Val) || isa<Second, Rest...>(Val); |
149 | } |
150 | |
151 | // isa_and_nonnull<X> - Functionally identical to isa, except that a null value |
152 | // is accepted. |
153 | // |
154 | template <typename... X, class Y> |
155 | LLVM_NODISCARD[[clang::warn_unused_result]] inline bool isa_and_nonnull(const Y &Val) { |
156 | if (!Val) |
157 | return false; |
158 | return isa<X...>(Val); |
159 | } |
160 | |
161 | //===----------------------------------------------------------------------===// |
162 | // cast<x> Support Templates |
163 | //===----------------------------------------------------------------------===// |
164 | |
165 | template<class To, class From> struct cast_retty; |
166 | |
167 | // Calculate what type the 'cast' function should return, based on a requested |
168 | // type of To and a source type of From. |
169 | template<class To, class From> struct cast_retty_impl { |
170 | using ret_type = To &; // Normal case, return Ty& |
171 | }; |
172 | template<class To, class From> struct cast_retty_impl<To, const From> { |
173 | using ret_type = const To &; // Normal case, return Ty& |
174 | }; |
175 | |
176 | template<class To, class From> struct cast_retty_impl<To, From*> { |
177 | using ret_type = To *; // Pointer arg case, return Ty* |
178 | }; |
179 | |
180 | template<class To, class From> struct cast_retty_impl<To, const From*> { |
181 | using ret_type = const To *; // Constant pointer arg case, return const Ty* |
182 | }; |
183 | |
184 | template<class To, class From> struct cast_retty_impl<To, const From*const> { |
185 | using ret_type = const To *; // Constant pointer arg case, return const Ty* |
186 | }; |
187 | |
188 | template <class To, class From> |
189 | struct cast_retty_impl<To, std::unique_ptr<From>> { |
190 | private: |
191 | using PointerType = typename cast_retty_impl<To, From *>::ret_type; |
192 | using ResultType = std::remove_pointer_t<PointerType>; |
193 | |
194 | public: |
195 | using ret_type = std::unique_ptr<ResultType>; |
196 | }; |
197 | |
198 | template<class To, class From, class SimpleFrom> |
199 | struct cast_retty_wrap { |
200 | // When the simplified type and the from type are not the same, use the type |
201 | // simplifier to reduce the type, then reuse cast_retty_impl to get the |
202 | // resultant type. |
203 | using ret_type = typename cast_retty<To, SimpleFrom>::ret_type; |
204 | }; |
205 | |
206 | template<class To, class FromTy> |
207 | struct cast_retty_wrap<To, FromTy, FromTy> { |
208 | // When the simplified type is equal to the from type, use it directly. |
209 | using ret_type = typename cast_retty_impl<To,FromTy>::ret_type; |
210 | }; |
211 | |
212 | template<class To, class From> |
213 | struct cast_retty { |
214 | using ret_type = typename cast_retty_wrap< |
215 | To, From, typename simplify_type<From>::SimpleType>::ret_type; |
216 | }; |
217 | |
218 | // Ensure the non-simple values are converted using the simplify_type template |
219 | // that may be specialized by smart pointers... |
220 | // |
221 | template<class To, class From, class SimpleFrom> struct cast_convert_val { |
222 | // This is not a simple type, use the template to simplify it... |
223 | static typename cast_retty<To, From>::ret_type doit(From &Val) { |
224 | return cast_convert_val<To, SimpleFrom, |
225 | typename simplify_type<SimpleFrom>::SimpleType>::doit( |
226 | simplify_type<From>::getSimplifiedValue(Val)); |
227 | } |
228 | }; |
229 | |
230 | template<class To, class FromTy> struct cast_convert_val<To,FromTy,FromTy> { |
231 | // This _is_ a simple type, just cast it. |
232 | static typename cast_retty<To, FromTy>::ret_type doit(const FromTy &Val) { |
233 | typename cast_retty<To, FromTy>::ret_type Res2 |
234 | = (typename cast_retty<To, FromTy>::ret_type)const_cast<FromTy&>(Val); |
235 | return Res2; |
236 | } |
237 | }; |
238 | |
239 | template <class X> struct is_simple_type { |
240 | static const bool value = |
241 | std::is_same<X, typename simplify_type<X>::SimpleType>::value; |
242 | }; |
243 | |
244 | // cast<X> - Return the argument parameter cast to the specified type. This |
245 | // casting operator asserts that the type is correct, so it does not return null |
246 | // on failure. It does not allow a null argument (use cast_or_null for that). |
247 | // It is typically used like this: |
248 | // |
249 | // cast<Instruction>(myVal)->getParent() |
250 | // |
251 | template <class X, class Y> |
252 | inline std::enable_if_t<!is_simple_type<Y>::value, |
253 | typename cast_retty<X, const Y>::ret_type> |
254 | cast(const Y &Val) { |
255 | assert(isa<X>(Val) && "cast<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val) && "cast<Ty>() argument of incompatible type!" ) ? void (0) : __assert_fail ("isa<X>(Val) && \"cast<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 255, __extension__ __PRETTY_FUNCTION__)); |
256 | return cast_convert_val< |
257 | X, const Y, typename simplify_type<const Y>::SimpleType>::doit(Val); |
258 | } |
259 | |
260 | template <class X, class Y> |
261 | inline typename cast_retty<X, Y>::ret_type cast(Y &Val) { |
262 | assert(isa<X>(Val) && "cast<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val) && "cast<Ty>() argument of incompatible type!" ) ? void (0) : __assert_fail ("isa<X>(Val) && \"cast<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 262, __extension__ __PRETTY_FUNCTION__)); |
263 | return cast_convert_val<X, Y, |
264 | typename simplify_type<Y>::SimpleType>::doit(Val); |
265 | } |
266 | |
267 | template <class X, class Y> |
268 | inline typename cast_retty<X, Y *>::ret_type cast(Y *Val) { |
269 | assert(isa<X>(Val) && "cast<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val) && "cast<Ty>() argument of incompatible type!" ) ? void (0) : __assert_fail ("isa<X>(Val) && \"cast<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 269, __extension__ __PRETTY_FUNCTION__)); |
270 | return cast_convert_val<X, Y*, |
271 | typename simplify_type<Y*>::SimpleType>::doit(Val); |
272 | } |
273 | |
274 | template <class X, class Y> |
275 | inline typename cast_retty<X, std::unique_ptr<Y>>::ret_type |
276 | cast(std::unique_ptr<Y> &&Val) { |
277 | assert(isa<X>(Val.get()) && "cast<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val.get()) && "cast<Ty>() argument of incompatible type!") ? void (0 ) : __assert_fail ("isa<X>(Val.get()) && \"cast<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 277, __extension__ __PRETTY_FUNCTION__)); |
278 | using ret_type = typename cast_retty<X, std::unique_ptr<Y>>::ret_type; |
279 | return ret_type( |
280 | cast_convert_val<X, Y *, typename simplify_type<Y *>::SimpleType>::doit( |
281 | Val.release())); |
282 | } |
283 | |
284 | // cast_or_null<X> - Functionally identical to cast, except that a null value is |
285 | // accepted. |
286 | // |
287 | template <class X, class Y> |
288 | LLVM_NODISCARD[[clang::warn_unused_result]] inline std::enable_if_t< |
289 | !is_simple_type<Y>::value, typename cast_retty<X, const Y>::ret_type> |
290 | cast_or_null(const Y &Val) { |
291 | if (!Val) |
292 | return nullptr; |
293 | assert(isa<X>(Val) && "cast_or_null<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val) && "cast_or_null<Ty>() argument of incompatible type!" ) ? void (0) : __assert_fail ("isa<X>(Val) && \"cast_or_null<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 293, __extension__ __PRETTY_FUNCTION__)); |
294 | return cast<X>(Val); |
295 | } |
296 | |
297 | template <class X, class Y> |
298 | LLVM_NODISCARD[[clang::warn_unused_result]] inline std::enable_if_t<!is_simple_type<Y>::value, |
299 | typename cast_retty<X, Y>::ret_type> |
300 | cast_or_null(Y &Val) { |
301 | if (!Val) |
302 | return nullptr; |
303 | assert(isa<X>(Val) && "cast_or_null<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val) && "cast_or_null<Ty>() argument of incompatible type!" ) ? void (0) : __assert_fail ("isa<X>(Val) && \"cast_or_null<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 303, __extension__ __PRETTY_FUNCTION__)); |
304 | return cast<X>(Val); |
305 | } |
306 | |
307 | template <class X, class Y> |
308 | LLVM_NODISCARD[[clang::warn_unused_result]] inline typename cast_retty<X, Y *>::ret_type |
309 | cast_or_null(Y *Val) { |
310 | if (!Val) return nullptr; |
311 | assert(isa<X>(Val) && "cast_or_null<Ty>() argument of incompatible type!")(static_cast <bool> (isa<X>(Val) && "cast_or_null<Ty>() argument of incompatible type!" ) ? void (0) : __assert_fail ("isa<X>(Val) && \"cast_or_null<Ty>() argument of incompatible type!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Support/Casting.h" , 311, __extension__ __PRETTY_FUNCTION__)); |
312 | return cast<X>(Val); |
313 | } |
314 | |
315 | template <class X, class Y> |
316 | inline typename cast_retty<X, std::unique_ptr<Y>>::ret_type |
317 | cast_or_null(std::unique_ptr<Y> &&Val) { |
318 | if (!Val) |
319 | return nullptr; |
320 | return cast<X>(std::move(Val)); |
321 | } |
322 | |
323 | // dyn_cast<X> - Return the argument parameter cast to the specified type. This |
324 | // casting operator returns null if the argument is of the wrong type, so it can |
325 | // be used to test for a type as well as cast if successful. This should be |
326 | // used in the context of an if statement like this: |
327 | // |
328 | // if (const Instruction *I = dyn_cast<Instruction>(myVal)) { ... } |
329 | // |
330 | |
331 | template <class X, class Y> |
332 | LLVM_NODISCARD[[clang::warn_unused_result]] inline std::enable_if_t< |
333 | !is_simple_type<Y>::value, typename cast_retty<X, const Y>::ret_type> |
334 | dyn_cast(const Y &Val) { |
335 | return isa<X>(Val) ? cast<X>(Val) : nullptr; |
336 | } |
337 | |
338 | template <class X, class Y> |
339 | LLVM_NODISCARD[[clang::warn_unused_result]] inline typename cast_retty<X, Y>::ret_type dyn_cast(Y &Val) { |
340 | return isa<X>(Val) ? cast<X>(Val) : nullptr; |
341 | } |
342 | |
343 | template <class X, class Y> |
344 | LLVM_NODISCARD[[clang::warn_unused_result]] inline typename cast_retty<X, Y *>::ret_type dyn_cast(Y *Val) { |
345 | return isa<X>(Val) ? cast<X>(Val) : nullptr; |
346 | } |
347 | |
348 | // dyn_cast_or_null<X> - Functionally identical to dyn_cast, except that a null |
349 | // value is accepted. |
350 | // |
351 | template <class X, class Y> |
352 | LLVM_NODISCARD[[clang::warn_unused_result]] inline std::enable_if_t< |
353 | !is_simple_type<Y>::value, typename cast_retty<X, const Y>::ret_type> |
354 | dyn_cast_or_null(const Y &Val) { |
355 | return (Val && isa<X>(Val)) ? cast<X>(Val) : nullptr; |
356 | } |
357 | |
358 | template <class X, class Y> |
359 | LLVM_NODISCARD[[clang::warn_unused_result]] inline std::enable_if_t<!is_simple_type<Y>::value, |
360 | typename cast_retty<X, Y>::ret_type> |
361 | dyn_cast_or_null(Y &Val) { |
362 | return (Val && isa<X>(Val)) ? cast<X>(Val) : nullptr; |
363 | } |
364 | |
365 | template <class X, class Y> |
366 | LLVM_NODISCARD[[clang::warn_unused_result]] inline typename cast_retty<X, Y *>::ret_type |
367 | dyn_cast_or_null(Y *Val) { |
368 | return (Val && isa<X>(Val)) ? cast<X>(Val) : nullptr; |
369 | } |
370 | |
371 | // unique_dyn_cast<X> - Given a unique_ptr<Y>, try to return a unique_ptr<X>, |
372 | // taking ownership of the input pointer iff isa<X>(Val) is true. If the |
373 | // cast is successful, From refers to nullptr on exit and the casted value |
374 | // is returned. If the cast is unsuccessful, the function returns nullptr |
375 | // and From is unchanged. |
376 | template <class X, class Y> |
377 | LLVM_NODISCARD[[clang::warn_unused_result]] inline auto unique_dyn_cast(std::unique_ptr<Y> &Val) |
378 | -> decltype(cast<X>(Val)) { |
379 | if (!isa<X>(Val)) |
380 | return nullptr; |
381 | return cast<X>(std::move(Val)); |
382 | } |
383 | |
384 | template <class X, class Y> |
385 | LLVM_NODISCARD[[clang::warn_unused_result]] inline auto unique_dyn_cast(std::unique_ptr<Y> &&Val) { |
386 | return unique_dyn_cast<X, Y>(Val); |
387 | } |
388 | |
389 | // dyn_cast_or_null<X> - Functionally identical to unique_dyn_cast, except that |
390 | // a null value is accepted. |
391 | template <class X, class Y> |
392 | LLVM_NODISCARD[[clang::warn_unused_result]] inline auto unique_dyn_cast_or_null(std::unique_ptr<Y> &Val) |
393 | -> decltype(cast<X>(Val)) { |
394 | if (!Val) |
395 | return nullptr; |
396 | return unique_dyn_cast<X, Y>(Val); |
397 | } |
398 | |
399 | template <class X, class Y> |
400 | LLVM_NODISCARD[[clang::warn_unused_result]] inline auto unique_dyn_cast_or_null(std::unique_ptr<Y> &&Val) { |
401 | return unique_dyn_cast_or_null<X, Y>(Val); |
402 | } |
403 | |
404 | } // end namespace llvm |
405 | |
406 | #endif // LLVM_SUPPORT_CASTING_H |
1 | //===- llvm/CodeGen/SelectionDAGNodes.h - SelectionDAG Nodes ----*- C++ -*-===// | |||
2 | // | |||
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. | |||
4 | // See https://llvm.org/LICENSE.txt for license information. | |||
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception | |||
6 | // | |||
7 | //===----------------------------------------------------------------------===// | |||
8 | // | |||
9 | // This file declares the SDNode class and derived classes, which are used to | |||
10 | // represent the nodes and operations present in a SelectionDAG. These nodes | |||
11 | // and operations are machine code level operations, with some similarities to | |||
12 | // the GCC RTL representation. | |||
13 | // | |||
14 | // Clients should include the SelectionDAG.h file instead of this file directly. | |||
15 | // | |||
16 | //===----------------------------------------------------------------------===// | |||
17 | ||||
18 | #ifndef LLVM_CODEGEN_SELECTIONDAGNODES_H | |||
19 | #define LLVM_CODEGEN_SELECTIONDAGNODES_H | |||
20 | ||||
21 | #include "llvm/ADT/APFloat.h" | |||
22 | #include "llvm/ADT/ArrayRef.h" | |||
23 | #include "llvm/ADT/BitVector.h" | |||
24 | #include "llvm/ADT/FoldingSet.h" | |||
25 | #include "llvm/ADT/GraphTraits.h" | |||
26 | #include "llvm/ADT/SmallPtrSet.h" | |||
27 | #include "llvm/ADT/SmallVector.h" | |||
28 | #include "llvm/ADT/ilist_node.h" | |||
29 | #include "llvm/ADT/iterator.h" | |||
30 | #include "llvm/ADT/iterator_range.h" | |||
31 | #include "llvm/CodeGen/ISDOpcodes.h" | |||
32 | #include "llvm/CodeGen/MachineMemOperand.h" | |||
33 | #include "llvm/CodeGen/Register.h" | |||
34 | #include "llvm/CodeGen/ValueTypes.h" | |||
35 | #include "llvm/IR/Constants.h" | |||
36 | #include "llvm/IR/DebugLoc.h" | |||
37 | #include "llvm/IR/Instruction.h" | |||
38 | #include "llvm/IR/Instructions.h" | |||
39 | #include "llvm/IR/Metadata.h" | |||
40 | #include "llvm/IR/Operator.h" | |||
41 | #include "llvm/Support/AlignOf.h" | |||
42 | #include "llvm/Support/AtomicOrdering.h" | |||
43 | #include "llvm/Support/Casting.h" | |||
44 | #include "llvm/Support/ErrorHandling.h" | |||
45 | #include "llvm/Support/MachineValueType.h" | |||
46 | #include "llvm/Support/TypeSize.h" | |||
47 | #include <algorithm> | |||
48 | #include <cassert> | |||
49 | #include <climits> | |||
50 | #include <cstddef> | |||
51 | #include <cstdint> | |||
52 | #include <cstring> | |||
53 | #include <iterator> | |||
54 | #include <string> | |||
55 | #include <tuple> | |||
56 | ||||
57 | namespace llvm { | |||
58 | ||||
59 | class APInt; | |||
60 | class Constant; | |||
61 | template <typename T> struct DenseMapInfo; | |||
62 | class GlobalValue; | |||
63 | class MachineBasicBlock; | |||
64 | class MachineConstantPoolValue; | |||
65 | class MCSymbol; | |||
66 | class raw_ostream; | |||
67 | class SDNode; | |||
68 | class SelectionDAG; | |||
69 | class Type; | |||
70 | class Value; | |||
71 | ||||
72 | void checkForCycles(const SDNode *N, const SelectionDAG *DAG = nullptr, | |||
73 | bool force = false); | |||
74 | ||||
75 | /// This represents a list of ValueType's that has been intern'd by | |||
76 | /// a SelectionDAG. Instances of this simple value class are returned by | |||
77 | /// SelectionDAG::getVTList(...). | |||
78 | /// | |||
79 | struct SDVTList { | |||
80 | const EVT *VTs; | |||
81 | unsigned int NumVTs; | |||
82 | }; | |||
83 | ||||
84 | namespace ISD { | |||
85 | ||||
86 | /// Node predicates | |||
87 | ||||
88 | /// If N is a BUILD_VECTOR or SPLAT_VECTOR node whose elements are all the | |||
89 | /// same constant or undefined, return true and return the constant value in | |||
90 | /// \p SplatValue. | |||
91 | bool isConstantSplatVector(const SDNode *N, APInt &SplatValue); | |||
92 | ||||
93 | /// Return true if the specified node is a BUILD_VECTOR or SPLAT_VECTOR where | |||
94 | /// all of the elements are ~0 or undef. If \p BuildVectorOnly is set to | |||
95 | /// true, it only checks BUILD_VECTOR. | |||
96 | bool isConstantSplatVectorAllOnes(const SDNode *N, | |||
97 | bool BuildVectorOnly = false); | |||
98 | ||||
99 | /// Return true if the specified node is a BUILD_VECTOR or SPLAT_VECTOR where | |||
100 | /// all of the elements are 0 or undef. If \p BuildVectorOnly is set to true, it | |||
101 | /// only checks BUILD_VECTOR. | |||
102 | bool isConstantSplatVectorAllZeros(const SDNode *N, | |||
103 | bool BuildVectorOnly = false); | |||
104 | ||||
105 | /// Return true if the specified node is a BUILD_VECTOR where all of the | |||
106 | /// elements are ~0 or undef. | |||
107 | bool isBuildVectorAllOnes(const SDNode *N); | |||
108 | ||||
109 | /// Return true if the specified node is a BUILD_VECTOR where all of the | |||
110 | /// elements are 0 or undef. | |||
111 | bool isBuildVectorAllZeros(const SDNode *N); | |||
112 | ||||
113 | /// Return true if the specified node is a BUILD_VECTOR node of all | |||
114 | /// ConstantSDNode or undef. | |||
115 | bool isBuildVectorOfConstantSDNodes(const SDNode *N); | |||
116 | ||||
117 | /// Return true if the specified node is a BUILD_VECTOR node of all | |||
118 | /// ConstantFPSDNode or undef. | |||
119 | bool isBuildVectorOfConstantFPSDNodes(const SDNode *N); | |||
120 | ||||
121 | /// Return true if the node has at least one operand and all operands of the | |||
122 | /// specified node are ISD::UNDEF. | |||
123 | bool allOperandsUndef(const SDNode *N); | |||
124 | ||||
125 | } // end namespace ISD | |||
126 | ||||
127 | //===----------------------------------------------------------------------===// | |||
128 | /// Unlike LLVM values, Selection DAG nodes may return multiple | |||
129 | /// values as the result of a computation. Many nodes return multiple values, | |||
130 | /// from loads (which define a token and a return value) to ADDC (which returns | |||
131 | /// a result and a carry value), to calls (which may return an arbitrary number | |||
132 | /// of values). | |||
133 | /// | |||
134 | /// As such, each use of a SelectionDAG computation must indicate the node that | |||
135 | /// computes it as well as which return value to use from that node. This pair | |||
136 | /// of information is represented with the SDValue value type. | |||
137 | /// | |||
138 | class SDValue { | |||
139 | friend struct DenseMapInfo<SDValue>; | |||
140 | ||||
141 | SDNode *Node = nullptr; // The node defining the value we are using. | |||
142 | unsigned ResNo = 0; // Which return value of the node we are using. | |||
143 | ||||
144 | public: | |||
145 | SDValue() = default; | |||
146 | SDValue(SDNode *node, unsigned resno); | |||
147 | ||||
148 | /// get the index which selects a specific result in the SDNode | |||
149 | unsigned getResNo() const { return ResNo; } | |||
150 | ||||
151 | /// get the SDNode which holds the desired result | |||
152 | SDNode *getNode() const { return Node; } | |||
153 | ||||
154 | /// set the SDNode | |||
155 | void setNode(SDNode *N) { Node = N; } | |||
156 | ||||
157 | inline SDNode *operator->() const { return Node; } | |||
158 | ||||
159 | bool operator==(const SDValue &O) const { | |||
160 | return Node == O.Node && ResNo == O.ResNo; | |||
161 | } | |||
162 | bool operator!=(const SDValue &O) const { | |||
163 | return !operator==(O); | |||
164 | } | |||
165 | bool operator<(const SDValue &O) const { | |||
166 | return std::tie(Node, ResNo) < std::tie(O.Node, O.ResNo); | |||
167 | } | |||
168 | explicit operator bool() const { | |||
169 | return Node != nullptr; | |||
170 | } | |||
171 | ||||
172 | SDValue getValue(unsigned R) const { | |||
173 | return SDValue(Node, R); | |||
174 | } | |||
175 | ||||
176 | /// Return true if this node is an operand of N. | |||
177 | bool isOperandOf(const SDNode *N) const; | |||
178 | ||||
179 | /// Return the ValueType of the referenced return value. | |||
180 | inline EVT getValueType() const; | |||
181 | ||||
182 | /// Return the simple ValueType of the referenced return value. | |||
183 | MVT getSimpleValueType() const { | |||
184 | return getValueType().getSimpleVT(); | |||
185 | } | |||
186 | ||||
187 | /// Returns the size of the value in bits. | |||
188 | /// | |||
189 | /// If the value type is a scalable vector type, the scalable property will | |||
190 | /// be set and the runtime size will be a positive integer multiple of the | |||
191 | /// base size. | |||
192 | TypeSize getValueSizeInBits() const { | |||
193 | return getValueType().getSizeInBits(); | |||
194 | } | |||
195 | ||||
196 | uint64_t getScalarValueSizeInBits() const { | |||
197 | return getValueType().getScalarType().getFixedSizeInBits(); | |||
198 | } | |||
199 | ||||
200 | // Forwarding methods - These forward to the corresponding methods in SDNode. | |||
201 | inline unsigned getOpcode() const; | |||
202 | inline unsigned getNumOperands() const; | |||
203 | inline const SDValue &getOperand(unsigned i) const; | |||
204 | inline uint64_t getConstantOperandVal(unsigned i) const; | |||
205 | inline const APInt &getConstantOperandAPInt(unsigned i) const; | |||
206 | inline bool isTargetMemoryOpcode() const; | |||
207 | inline bool isTargetOpcode() const; | |||
208 | inline bool isMachineOpcode() const; | |||
209 | inline bool isUndef() const; | |||
210 | inline unsigned getMachineOpcode() const; | |||
211 | inline const DebugLoc &getDebugLoc() const; | |||
212 | inline void dump() const; | |||
213 | inline void dump(const SelectionDAG *G) const; | |||
214 | inline void dumpr() const; | |||
215 | inline void dumpr(const SelectionDAG *G) const; | |||
216 | ||||
217 | /// Return true if this operand (which must be a chain) reaches the | |||
218 | /// specified operand without crossing any side-effecting instructions. | |||
219 | /// In practice, this looks through token factors and non-volatile loads. | |||
220 | /// In order to remain efficient, this only | |||
221 | /// looks a couple of nodes in, it does not do an exhaustive search. | |||
222 | bool reachesChainWithoutSideEffects(SDValue Dest, | |||
223 | unsigned Depth = 2) const; | |||
224 | ||||
225 | /// Return true if there are no nodes using value ResNo of Node. | |||
226 | inline bool use_empty() const; | |||
227 | ||||
228 | /// Return true if there is exactly one node using value ResNo of Node. | |||
229 | inline bool hasOneUse() const; | |||
230 | }; | |||
231 | ||||
232 | template<> struct DenseMapInfo<SDValue> { | |||
233 | static inline SDValue getEmptyKey() { | |||
234 | SDValue V; | |||
235 | V.ResNo = -1U; | |||
236 | return V; | |||
237 | } | |||
238 | ||||
239 | static inline SDValue getTombstoneKey() { | |||
240 | SDValue V; | |||
241 | V.ResNo = -2U; | |||
242 | return V; | |||
243 | } | |||
244 | ||||
245 | static unsigned getHashValue(const SDValue &Val) { | |||
246 | return ((unsigned)((uintptr_t)Val.getNode() >> 4) ^ | |||
247 | (unsigned)((uintptr_t)Val.getNode() >> 9)) + Val.getResNo(); | |||
248 | } | |||
249 | ||||
250 | static bool isEqual(const SDValue &LHS, const SDValue &RHS) { | |||
251 | return LHS == RHS; | |||
252 | } | |||
253 | }; | |||
254 | ||||
255 | /// Allow casting operators to work directly on | |||
256 | /// SDValues as if they were SDNode*'s. | |||
257 | template<> struct simplify_type<SDValue> { | |||
258 | using SimpleType = SDNode *; | |||
259 | ||||
260 | static SimpleType getSimplifiedValue(SDValue &Val) { | |||
261 | return Val.getNode(); | |||
262 | } | |||
263 | }; | |||
264 | template<> struct simplify_type<const SDValue> { | |||
265 | using SimpleType = /*const*/ SDNode *; | |||
266 | ||||
267 | static SimpleType getSimplifiedValue(const SDValue &Val) { | |||
268 | return Val.getNode(); | |||
269 | } | |||
270 | }; | |||
271 | ||||
272 | /// Represents a use of a SDNode. This class holds an SDValue, | |||
273 | /// which records the SDNode being used and the result number, a | |||
274 | /// pointer to the SDNode using the value, and Next and Prev pointers, | |||
275 | /// which link together all the uses of an SDNode. | |||
276 | /// | |||
277 | class SDUse { | |||
278 | /// Val - The value being used. | |||
279 | SDValue Val; | |||
280 | /// User - The user of this value. | |||
281 | SDNode *User = nullptr; | |||
282 | /// Prev, Next - Pointers to the uses list of the SDNode referred by | |||
283 | /// this operand. | |||
284 | SDUse **Prev = nullptr; | |||
285 | SDUse *Next = nullptr; | |||
286 | ||||
287 | public: | |||
288 | SDUse() = default; | |||
289 | SDUse(const SDUse &U) = delete; | |||
290 | SDUse &operator=(const SDUse &) = delete; | |||
291 | ||||
292 | /// Normally SDUse will just implicitly convert to an SDValue that it holds. | |||
293 | operator const SDValue&() const { return Val; } | |||
294 | ||||
295 | /// If implicit conversion to SDValue doesn't work, the get() method returns | |||
296 | /// the SDValue. | |||
297 | const SDValue &get() const { return Val; } | |||
298 | ||||
299 | /// This returns the SDNode that contains this Use. | |||
300 | SDNode *getUser() { return User; } | |||
301 | ||||
302 | /// Get the next SDUse in the use list. | |||
303 | SDUse *getNext() const { return Next; } | |||
304 | ||||
305 | /// Convenience function for get().getNode(). | |||
306 | SDNode *getNode() const { return Val.getNode(); } | |||
307 | /// Convenience function for get().getResNo(). | |||
308 | unsigned getResNo() const { return Val.getResNo(); } | |||
309 | /// Convenience function for get().getValueType(). | |||
310 | EVT getValueType() const { return Val.getValueType(); } | |||
311 | ||||
312 | /// Convenience function for get().operator== | |||
313 | bool operator==(const SDValue &V) const { | |||
314 | return Val == V; | |||
315 | } | |||
316 | ||||
317 | /// Convenience function for get().operator!= | |||
318 | bool operator!=(const SDValue &V) const { | |||
319 | return Val != V; | |||
320 | } | |||
321 | ||||
322 | /// Convenience function for get().operator< | |||
323 | bool operator<(const SDValue &V) const { | |||
324 | return Val < V; | |||
325 | } | |||
326 | ||||
327 | private: | |||
328 | friend class SelectionDAG; | |||
329 | friend class SDNode; | |||
330 | // TODO: unfriend HandleSDNode once we fix its operand handling. | |||
331 | friend class HandleSDNode; | |||
332 | ||||
333 | void setUser(SDNode *p) { User = p; } | |||
334 | ||||
335 | /// Remove this use from its existing use list, assign it the | |||
336 | /// given value, and add it to the new value's node's use list. | |||
337 | inline void set(const SDValue &V); | |||
338 | /// Like set, but only supports initializing a newly-allocated | |||
339 | /// SDUse with a non-null value. | |||
340 | inline void setInitial(const SDValue &V); | |||
341 | /// Like set, but only sets the Node portion of the value, | |||
342 | /// leaving the ResNo portion unmodified. | |||
343 | inline void setNode(SDNode *N); | |||
344 | ||||
345 | void addToList(SDUse **List) { | |||
346 | Next = *List; | |||
347 | if (Next) Next->Prev = &Next; | |||
348 | Prev = List; | |||
349 | *List = this; | |||
350 | } | |||
351 | ||||
352 | void removeFromList() { | |||
353 | *Prev = Next; | |||
354 | if (Next) Next->Prev = Prev; | |||
355 | } | |||
356 | }; | |||
357 | ||||
358 | /// simplify_type specializations - Allow casting operators to work directly on | |||
359 | /// SDValues as if they were SDNode*'s. | |||
360 | template<> struct simplify_type<SDUse> { | |||
361 | using SimpleType = SDNode *; | |||
362 | ||||
363 | static SimpleType getSimplifiedValue(SDUse &Val) { | |||
364 | return Val.getNode(); | |||
365 | } | |||
366 | }; | |||
367 | ||||
368 | /// These are IR-level optimization flags that may be propagated to SDNodes. | |||
369 | /// TODO: This data structure should be shared by the IR optimizer and the | |||
370 | /// the backend. | |||
371 | struct SDNodeFlags { | |||
372 | private: | |||
373 | bool NoUnsignedWrap : 1; | |||
374 | bool NoSignedWrap : 1; | |||
375 | bool Exact : 1; | |||
376 | bool NoNaNs : 1; | |||
377 | bool NoInfs : 1; | |||
378 | bool NoSignedZeros : 1; | |||
379 | bool AllowReciprocal : 1; | |||
380 | bool AllowContract : 1; | |||
381 | bool ApproximateFuncs : 1; | |||
382 | bool AllowReassociation : 1; | |||
383 | ||||
384 | // We assume instructions do not raise floating-point exceptions by default, | |||
385 | // and only those marked explicitly may do so. We could choose to represent | |||
386 | // this via a positive "FPExcept" flags like on the MI level, but having a | |||
387 | // negative "NoFPExcept" flag here (that defaults to true) makes the flag | |||
388 | // intersection logic more straightforward. | |||
389 | bool NoFPExcept : 1; | |||
390 | ||||
391 | public: | |||
392 | /// Default constructor turns off all optimization flags. | |||
393 | SDNodeFlags() | |||
394 | : NoUnsignedWrap(false), NoSignedWrap(false), Exact(false), NoNaNs(false), | |||
395 | NoInfs(false), NoSignedZeros(false), AllowReciprocal(false), | |||
396 | AllowContract(false), ApproximateFuncs(false), | |||
397 | AllowReassociation(false), NoFPExcept(false) {} | |||
398 | ||||
399 | /// Propagate the fast-math-flags from an IR FPMathOperator. | |||
400 | void copyFMF(const FPMathOperator &FPMO) { | |||
401 | setNoNaNs(FPMO.hasNoNaNs()); | |||
402 | setNoInfs(FPMO.hasNoInfs()); | |||
403 | setNoSignedZeros(FPMO.hasNoSignedZeros()); | |||
404 | setAllowReciprocal(FPMO.hasAllowReciprocal()); | |||
405 | setAllowContract(FPMO.hasAllowContract()); | |||
406 | setApproximateFuncs(FPMO.hasApproxFunc()); | |||
407 | setAllowReassociation(FPMO.hasAllowReassoc()); | |||
408 | } | |||
409 | ||||
410 | // These are mutators for each flag. | |||
411 | void setNoUnsignedWrap(bool b) { NoUnsignedWrap = b; } | |||
412 | void setNoSignedWrap(bool b) { NoSignedWrap = b; } | |||
413 | void setExact(bool b) { Exact = b; } | |||
414 | void setNoNaNs(bool b) { NoNaNs = b; } | |||
415 | void setNoInfs(bool b) { NoInfs = b; } | |||
416 | void setNoSignedZeros(bool b) { NoSignedZeros = b; } | |||
417 | void setAllowReciprocal(bool b) { AllowReciprocal = b; } | |||
418 | void setAllowContract(bool b) { AllowContract = b; } | |||
419 | void setApproximateFuncs(bool b) { ApproximateFuncs = b; } | |||
420 | void setAllowReassociation(bool b) { AllowReassociation = b; } | |||
421 | void setNoFPExcept(bool b) { NoFPExcept = b; } | |||
422 | ||||
423 | // These are accessors for each flag. | |||
424 | bool hasNoUnsignedWrap() const { return NoUnsignedWrap; } | |||
425 | bool hasNoSignedWrap() const { return NoSignedWrap; } | |||
426 | bool hasExact() const { return Exact; } | |||
427 | bool hasNoNaNs() const { return NoNaNs; } | |||
428 | bool hasNoInfs() const { return NoInfs; } | |||
429 | bool hasNoSignedZeros() const { return NoSignedZeros; } | |||
430 | bool hasAllowReciprocal() const { return AllowReciprocal; } | |||
431 | bool hasAllowContract() const { return AllowContract; } | |||
432 | bool hasApproximateFuncs() const { return ApproximateFuncs; } | |||
433 | bool hasAllowReassociation() const { return AllowReassociation; } | |||
434 | bool hasNoFPExcept() const { return NoFPExcept; } | |||
435 | ||||
436 | /// Clear any flags in this flag set that aren't also set in Flags. All | |||
437 | /// flags will be cleared if Flags are undefined. | |||
438 | void intersectWith(const SDNodeFlags Flags) { | |||
439 | NoUnsignedWrap &= Flags.NoUnsignedWrap; | |||
440 | NoSignedWrap &= Flags.NoSignedWrap; | |||
441 | Exact &= Flags.Exact; | |||
442 | NoNaNs &= Flags.NoNaNs; | |||
443 | NoInfs &= Flags.NoInfs; | |||
444 | NoSignedZeros &= Flags.NoSignedZeros; | |||
445 | AllowReciprocal &= Flags.AllowReciprocal; | |||
446 | AllowContract &= Flags.AllowContract; | |||
447 | ApproximateFuncs &= Flags.ApproximateFuncs; | |||
448 | AllowReassociation &= Flags.AllowReassociation; | |||
449 | NoFPExcept &= Flags.NoFPExcept; | |||
450 | } | |||
451 | }; | |||
452 | ||||
453 | /// Represents one node in the SelectionDAG. | |||
454 | /// | |||
455 | class SDNode : public FoldingSetNode, public ilist_node<SDNode> { | |||
456 | private: | |||
457 | /// The operation that this node performs. | |||
458 | int16_t NodeType; | |||
459 | ||||
460 | protected: | |||
461 | // We define a set of mini-helper classes to help us interpret the bits in our | |||
462 | // SubclassData. These are designed to fit within a uint16_t so they pack | |||
463 | // with NodeType. | |||
464 | ||||
465 | #if defined(_AIX) && (!defined(__GNUC__4) || defined(__clang__1)) | |||
466 | // Except for GCC; by default, AIX compilers store bit-fields in 4-byte words | |||
467 | // and give the `pack` pragma push semantics. | |||
468 | #define BEGIN_TWO_BYTE_PACK() _Pragma("pack(2)")pack(2) | |||
469 | #define END_TWO_BYTE_PACK() _Pragma("pack(pop)")pack(pop) | |||
470 | #else | |||
471 | #define BEGIN_TWO_BYTE_PACK() | |||
472 | #define END_TWO_BYTE_PACK() | |||
473 | #endif | |||
474 | ||||
475 | BEGIN_TWO_BYTE_PACK() | |||
476 | class SDNodeBitfields { | |||
477 | friend class SDNode; | |||
478 | friend class MemIntrinsicSDNode; | |||
479 | friend class MemSDNode; | |||
480 | friend class SelectionDAG; | |||
481 | ||||
482 | uint16_t HasDebugValue : 1; | |||
483 | uint16_t IsMemIntrinsic : 1; | |||
484 | uint16_t IsDivergent : 1; | |||
485 | }; | |||
486 | enum { NumSDNodeBits = 3 }; | |||
487 | ||||
488 | class ConstantSDNodeBitfields { | |||
489 | friend class ConstantSDNode; | |||
490 | ||||
491 | uint16_t : NumSDNodeBits; | |||
492 | ||||
493 | uint16_t IsOpaque : 1; | |||
494 | }; | |||
495 | ||||
496 | class MemSDNodeBitfields { | |||
497 | friend class MemSDNode; | |||
498 | friend class MemIntrinsicSDNode; | |||
499 | friend class AtomicSDNode; | |||
500 | ||||
501 | uint16_t : NumSDNodeBits; | |||
502 | ||||
503 | uint16_t IsVolatile : 1; | |||
504 | uint16_t IsNonTemporal : 1; | |||
505 | uint16_t IsDereferenceable : 1; | |||
506 | uint16_t IsInvariant : 1; | |||
507 | }; | |||
508 | enum { NumMemSDNodeBits = NumSDNodeBits + 4 }; | |||
509 | ||||
510 | class LSBaseSDNodeBitfields { | |||
511 | friend class LSBaseSDNode; | |||
512 | friend class MaskedLoadStoreSDNode; | |||
513 | friend class MaskedGatherScatterSDNode; | |||
514 | ||||
515 | uint16_t : NumMemSDNodeBits; | |||
516 | ||||
517 | // This storage is shared between disparate class hierarchies to hold an | |||
518 | // enumeration specific to the class hierarchy in use. | |||
519 | // LSBaseSDNode => enum ISD::MemIndexedMode | |||
520 | // MaskedLoadStoreBaseSDNode => enum ISD::MemIndexedMode | |||
521 | // MaskedGatherScatterSDNode => enum ISD::MemIndexType | |||
522 | uint16_t AddressingMode : 3; | |||
523 | }; | |||
524 | enum { NumLSBaseSDNodeBits = NumMemSDNodeBits + 3 }; | |||
525 | ||||
526 | class LoadSDNodeBitfields { | |||
527 | friend class LoadSDNode; | |||
528 | friend class MaskedLoadSDNode; | |||
529 | friend class MaskedGatherSDNode; | |||
530 | ||||
531 | uint16_t : NumLSBaseSDNodeBits; | |||
532 | ||||
533 | uint16_t ExtTy : 2; // enum ISD::LoadExtType | |||
534 | uint16_t IsExpanding : 1; | |||
535 | }; | |||
536 | ||||
537 | class StoreSDNodeBitfields { | |||
538 | friend class StoreSDNode; | |||
539 | friend class MaskedStoreSDNode; | |||
540 | friend class MaskedScatterSDNode; | |||
541 | ||||
542 | uint16_t : NumLSBaseSDNodeBits; | |||
543 | ||||
544 | uint16_t IsTruncating : 1; | |||
545 | uint16_t IsCompressing : 1; | |||
546 | }; | |||
547 | ||||
548 | union { | |||
549 | char RawSDNodeBits[sizeof(uint16_t)]; | |||
550 | SDNodeBitfields SDNodeBits; | |||
551 | ConstantSDNodeBitfields ConstantSDNodeBits; | |||
552 | MemSDNodeBitfields MemSDNodeBits; | |||
553 | LSBaseSDNodeBitfields LSBaseSDNodeBits; | |||
554 | LoadSDNodeBitfields LoadSDNodeBits; | |||
555 | StoreSDNodeBitfields StoreSDNodeBits; | |||
556 | }; | |||
557 | END_TWO_BYTE_PACK() | |||
558 | #undef BEGIN_TWO_BYTE_PACK | |||
559 | #undef END_TWO_BYTE_PACK | |||
560 | ||||
561 | // RawSDNodeBits must cover the entirety of the union. This means that all of | |||
562 | // the union's members must have size <= RawSDNodeBits. We write the RHS as | |||
563 | // "2" instead of sizeof(RawSDNodeBits) because MSVC can't handle the latter. | |||
564 | static_assert(sizeof(SDNodeBitfields) <= 2, "field too wide"); | |||
565 | static_assert(sizeof(ConstantSDNodeBitfields) <= 2, "field too wide"); | |||
566 | static_assert(sizeof(MemSDNodeBitfields) <= 2, "field too wide"); | |||
567 | static_assert(sizeof(LSBaseSDNodeBitfields) <= 2, "field too wide"); | |||
568 | static_assert(sizeof(LoadSDNodeBitfields) <= 2, "field too wide"); | |||
569 | static_assert(sizeof(StoreSDNodeBitfields) <= 2, "field too wide"); | |||
570 | ||||
571 | private: | |||
572 | friend class SelectionDAG; | |||
573 | // TODO: unfriend HandleSDNode once we fix its operand handling. | |||
574 | friend class HandleSDNode; | |||
575 | ||||
576 | /// Unique id per SDNode in the DAG. | |||
577 | int NodeId = -1; | |||
578 | ||||
579 | /// The values that are used by this operation. | |||
580 | SDUse *OperandList = nullptr; | |||
581 | ||||
582 | /// The types of the values this node defines. SDNode's may | |||
583 | /// define multiple values simultaneously. | |||
584 | const EVT *ValueList; | |||
585 | ||||
586 | /// List of uses for this SDNode. | |||
587 | SDUse *UseList = nullptr; | |||
588 | ||||
589 | /// The number of entries in the Operand/Value list. | |||
590 | unsigned short NumOperands = 0; | |||
591 | unsigned short NumValues; | |||
592 | ||||
593 | // The ordering of the SDNodes. It roughly corresponds to the ordering of the | |||
594 | // original LLVM instructions. | |||
595 | // This is used for turning off scheduling, because we'll forgo | |||
596 | // the normal scheduling algorithms and output the instructions according to | |||
597 | // this ordering. | |||
598 | unsigned IROrder; | |||
599 | ||||
600 | /// Source line information. | |||
601 | DebugLoc debugLoc; | |||
602 | ||||
603 | /// Return a pointer to the specified value type. | |||
604 | static const EVT *getValueTypeList(EVT VT); | |||
605 | ||||
606 | SDNodeFlags Flags; | |||
607 | ||||
608 | public: | |||
609 | /// Unique and persistent id per SDNode in the DAG. | |||
610 | /// Used for debug printing. | |||
611 | uint16_t PersistentId; | |||
612 | ||||
613 | //===--------------------------------------------------------------------===// | |||
614 | // Accessors | |||
615 | // | |||
616 | ||||
617 | /// Return the SelectionDAG opcode value for this node. For | |||
618 | /// pre-isel nodes (those for which isMachineOpcode returns false), these | |||
619 | /// are the opcode values in the ISD and <target>ISD namespaces. For | |||
620 | /// post-isel opcodes, see getMachineOpcode. | |||
621 | unsigned getOpcode() const { return (unsigned short)NodeType; } | |||
622 | ||||
623 | /// Test if this node has a target-specific opcode (in the | |||
624 | /// \<target\>ISD namespace). | |||
625 | bool isTargetOpcode() const { return NodeType >= ISD::BUILTIN_OP_END; } | |||
626 | ||||
627 | /// Test if this node has a target-specific opcode that may raise | |||
628 | /// FP exceptions (in the \<target\>ISD namespace and greater than | |||
629 | /// FIRST_TARGET_STRICTFP_OPCODE). Note that all target memory | |||
630 | /// opcode are currently automatically considered to possibly raise | |||
631 | /// FP exceptions as well. | |||
632 | bool isTargetStrictFPOpcode() const { | |||
633 | return NodeType >= ISD::FIRST_TARGET_STRICTFP_OPCODE; | |||
634 | } | |||
635 | ||||
636 | /// Test if this node has a target-specific | |||
637 | /// memory-referencing opcode (in the \<target\>ISD namespace and | |||
638 | /// greater than FIRST_TARGET_MEMORY_OPCODE). | |||
639 | bool isTargetMemoryOpcode() const { | |||
640 | return NodeType >= ISD::FIRST_TARGET_MEMORY_OPCODE; | |||
641 | } | |||
642 | ||||
643 | /// Return true if the type of the node type undefined. | |||
644 | bool isUndef() const { return NodeType == ISD::UNDEF; } | |||
645 | ||||
646 | /// Test if this node is a memory intrinsic (with valid pointer information). | |||
647 | /// INTRINSIC_W_CHAIN and INTRINSIC_VOID nodes are sometimes created for | |||
648 | /// non-memory intrinsics (with chains) that are not really instances of | |||
649 | /// MemSDNode. For such nodes, we need some extra state to determine the | |||
650 | /// proper classof relationship. | |||
651 | bool isMemIntrinsic() const { | |||
652 | return (NodeType == ISD::INTRINSIC_W_CHAIN || | |||
653 | NodeType == ISD::INTRINSIC_VOID) && | |||
654 | SDNodeBits.IsMemIntrinsic; | |||
655 | } | |||
656 | ||||
657 | /// Test if this node is a strict floating point pseudo-op. | |||
658 | bool isStrictFPOpcode() { | |||
659 | switch (NodeType) { | |||
660 | default: | |||
661 | return false; | |||
662 | case ISD::STRICT_FP16_TO_FP: | |||
663 | case ISD::STRICT_FP_TO_FP16: | |||
664 | #define DAG_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN) \ | |||
665 | case ISD::STRICT_##DAGN: | |||
666 | #include "llvm/IR/ConstrainedOps.def" | |||
667 | return true; | |||
668 | } | |||
669 | } | |||
670 | ||||
671 | /// Test if this node has a post-isel opcode, directly | |||
672 | /// corresponding to a MachineInstr opcode. | |||
673 | bool isMachineOpcode() const { return NodeType < 0; } | |||
674 | ||||
675 | /// This may only be called if isMachineOpcode returns | |||
676 | /// true. It returns the MachineInstr opcode value that the node's opcode | |||
677 | /// corresponds to. | |||
678 | unsigned getMachineOpcode() const { | |||
679 | assert(isMachineOpcode() && "Not a MachineInstr opcode!")(static_cast <bool> (isMachineOpcode() && "Not a MachineInstr opcode!" ) ? void (0) : __assert_fail ("isMachineOpcode() && \"Not a MachineInstr opcode!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 679, __extension__ __PRETTY_FUNCTION__)); | |||
680 | return ~NodeType; | |||
681 | } | |||
682 | ||||
683 | bool getHasDebugValue() const { return SDNodeBits.HasDebugValue; } | |||
684 | void setHasDebugValue(bool b) { SDNodeBits.HasDebugValue = b; } | |||
685 | ||||
686 | bool isDivergent() const { return SDNodeBits.IsDivergent; } | |||
687 | ||||
688 | /// Return true if there are no uses of this node. | |||
689 | bool use_empty() const { return UseList == nullptr; } | |||
690 | ||||
691 | /// Return true if there is exactly one use of this node. | |||
692 | bool hasOneUse() const { return hasSingleElement(uses()); } | |||
693 | ||||
694 | /// Return the number of uses of this node. This method takes | |||
695 | /// time proportional to the number of uses. | |||
696 | size_t use_size() const { return std::distance(use_begin(), use_end()); } | |||
697 | ||||
698 | /// Return the unique node id. | |||
699 | int getNodeId() const { return NodeId; } | |||
700 | ||||
701 | /// Set unique node id. | |||
702 | void setNodeId(int Id) { NodeId = Id; } | |||
703 | ||||
704 | /// Return the node ordering. | |||
705 | unsigned getIROrder() const { return IROrder; } | |||
706 | ||||
707 | /// Set the node ordering. | |||
708 | void setIROrder(unsigned Order) { IROrder = Order; } | |||
709 | ||||
710 | /// Return the source location info. | |||
711 | const DebugLoc &getDebugLoc() const { return debugLoc; } | |||
712 | ||||
713 | /// Set source location info. Try to avoid this, putting | |||
714 | /// it in the constructor is preferable. | |||
715 | void setDebugLoc(DebugLoc dl) { debugLoc = std::move(dl); } | |||
716 | ||||
717 | /// This class provides iterator support for SDUse | |||
718 | /// operands that use a specific SDNode. | |||
719 | class use_iterator { | |||
720 | friend class SDNode; | |||
721 | ||||
722 | SDUse *Op = nullptr; | |||
723 | ||||
724 | explicit use_iterator(SDUse *op) : Op(op) {} | |||
725 | ||||
726 | public: | |||
727 | using iterator_category = std::forward_iterator_tag; | |||
728 | using value_type = SDUse; | |||
729 | using difference_type = std::ptrdiff_t; | |||
730 | using pointer = value_type *; | |||
731 | using reference = value_type &; | |||
732 | ||||
733 | use_iterator() = default; | |||
734 | use_iterator(const use_iterator &I) : Op(I.Op) {} | |||
735 | ||||
736 | bool operator==(const use_iterator &x) const { | |||
737 | return Op == x.Op; | |||
738 | } | |||
739 | bool operator!=(const use_iterator &x) const { | |||
740 | return !operator==(x); | |||
741 | } | |||
742 | ||||
743 | /// Return true if this iterator is at the end of uses list. | |||
744 | bool atEnd() const { return Op == nullptr; } | |||
745 | ||||
746 | // Iterator traversal: forward iteration only. | |||
747 | use_iterator &operator++() { // Preincrement | |||
748 | assert(Op && "Cannot increment end iterator!")(static_cast <bool> (Op && "Cannot increment end iterator!" ) ? void (0) : __assert_fail ("Op && \"Cannot increment end iterator!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 748, __extension__ __PRETTY_FUNCTION__)); | |||
749 | Op = Op->getNext(); | |||
750 | return *this; | |||
751 | } | |||
752 | ||||
753 | use_iterator operator++(int) { // Postincrement | |||
754 | use_iterator tmp = *this; ++*this; return tmp; | |||
755 | } | |||
756 | ||||
757 | /// Retrieve a pointer to the current user node. | |||
758 | SDNode *operator*() const { | |||
759 | assert(Op && "Cannot dereference end iterator!")(static_cast <bool> (Op && "Cannot dereference end iterator!" ) ? void (0) : __assert_fail ("Op && \"Cannot dereference end iterator!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 759, __extension__ __PRETTY_FUNCTION__)); | |||
760 | return Op->getUser(); | |||
761 | } | |||
762 | ||||
763 | SDNode *operator->() const { return operator*(); } | |||
764 | ||||
765 | SDUse &getUse() const { return *Op; } | |||
766 | ||||
767 | /// Retrieve the operand # of this use in its user. | |||
768 | unsigned getOperandNo() const { | |||
769 | assert(Op && "Cannot dereference end iterator!")(static_cast <bool> (Op && "Cannot dereference end iterator!" ) ? void (0) : __assert_fail ("Op && \"Cannot dereference end iterator!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 769, __extension__ __PRETTY_FUNCTION__)); | |||
770 | return (unsigned)(Op - Op->getUser()->OperandList); | |||
771 | } | |||
772 | }; | |||
773 | ||||
774 | /// Provide iteration support to walk over all uses of an SDNode. | |||
775 | use_iterator use_begin() const { | |||
776 | return use_iterator(UseList); | |||
777 | } | |||
778 | ||||
779 | static use_iterator use_end() { return use_iterator(nullptr); } | |||
780 | ||||
781 | inline iterator_range<use_iterator> uses() { | |||
782 | return make_range(use_begin(), use_end()); | |||
783 | } | |||
784 | inline iterator_range<use_iterator> uses() const { | |||
785 | return make_range(use_begin(), use_end()); | |||
786 | } | |||
787 | ||||
788 | /// Return true if there are exactly NUSES uses of the indicated value. | |||
789 | /// This method ignores uses of other values defined by this operation. | |||
790 | bool hasNUsesOfValue(unsigned NUses, unsigned Value) const; | |||
791 | ||||
792 | /// Return true if there are any use of the indicated value. | |||
793 | /// This method ignores uses of other values defined by this operation. | |||
794 | bool hasAnyUseOfValue(unsigned Value) const; | |||
795 | ||||
796 | /// Return true if this node is the only use of N. | |||
797 | bool isOnlyUserOf(const SDNode *N) const; | |||
798 | ||||
799 | /// Return true if this node is an operand of N. | |||
800 | bool isOperandOf(const SDNode *N) const; | |||
801 | ||||
802 | /// Return true if this node is a predecessor of N. | |||
803 | /// NOTE: Implemented on top of hasPredecessor and every bit as | |||
804 | /// expensive. Use carefully. | |||
805 | bool isPredecessorOf(const SDNode *N) const { | |||
806 | return N->hasPredecessor(this); | |||
807 | } | |||
808 | ||||
809 | /// Return true if N is a predecessor of this node. | |||
810 | /// N is either an operand of this node, or can be reached by recursively | |||
811 | /// traversing up the operands. | |||
812 | /// NOTE: This is an expensive method. Use it carefully. | |||
813 | bool hasPredecessor(const SDNode *N) const; | |||
814 | ||||
815 | /// Returns true if N is a predecessor of any node in Worklist. This | |||
816 | /// helper keeps Visited and Worklist sets externally to allow unions | |||
817 | /// searches to be performed in parallel, caching of results across | |||
818 | /// queries and incremental addition to Worklist. Stops early if N is | |||
819 | /// found but will resume. Remember to clear Visited and Worklists | |||
820 | /// if DAG changes. MaxSteps gives a maximum number of nodes to visit before | |||
821 | /// giving up. The TopologicalPrune flag signals that positive NodeIds are | |||
822 | /// topologically ordered (Operands have strictly smaller node id) and search | |||
823 | /// can be pruned leveraging this. | |||
824 | static bool hasPredecessorHelper(const SDNode *N, | |||
825 | SmallPtrSetImpl<const SDNode *> &Visited, | |||
826 | SmallVectorImpl<const SDNode *> &Worklist, | |||
827 | unsigned int MaxSteps = 0, | |||
828 | bool TopologicalPrune = false) { | |||
829 | SmallVector<const SDNode *, 8> DeferredNodes; | |||
830 | if (Visited.count(N)) | |||
831 | return true; | |||
832 | ||||
833 | // Node Id's are assigned in three places: As a topological | |||
834 | // ordering (> 0), during legalization (results in values set to | |||
835 | // 0), new nodes (set to -1). If N has a topolgical id then we | |||
836 | // know that all nodes with ids smaller than it cannot be | |||
837 | // successors and we need not check them. Filter out all node | |||
838 | // that can't be matches. We add them to the worklist before exit | |||
839 | // in case of multiple calls. Note that during selection the topological id | |||
840 | // may be violated if a node's predecessor is selected before it. We mark | |||
841 | // this at selection negating the id of unselected successors and | |||
842 | // restricting topological pruning to positive ids. | |||
843 | ||||
844 | int NId = N->getNodeId(); | |||
845 | // If we Invalidated the Id, reconstruct original NId. | |||
846 | if (NId < -1) | |||
847 | NId = -(NId + 1); | |||
848 | ||||
849 | bool Found = false; | |||
850 | while (!Worklist.empty()) { | |||
851 | const SDNode *M = Worklist.pop_back_val(); | |||
852 | int MId = M->getNodeId(); | |||
853 | if (TopologicalPrune && M->getOpcode() != ISD::TokenFactor && (NId > 0) && | |||
854 | (MId > 0) && (MId < NId)) { | |||
855 | DeferredNodes.push_back(M); | |||
856 | continue; | |||
857 | } | |||
858 | for (const SDValue &OpV : M->op_values()) { | |||
859 | SDNode *Op = OpV.getNode(); | |||
860 | if (Visited.insert(Op).second) | |||
861 | Worklist.push_back(Op); | |||
862 | if (Op == N) | |||
863 | Found = true; | |||
864 | } | |||
865 | if (Found) | |||
866 | break; | |||
867 | if (MaxSteps != 0 && Visited.size() >= MaxSteps) | |||
868 | break; | |||
869 | } | |||
870 | // Push deferred nodes back on worklist. | |||
871 | Worklist.append(DeferredNodes.begin(), DeferredNodes.end()); | |||
872 | // If we bailed early, conservatively return found. | |||
873 | if (MaxSteps != 0 && Visited.size() >= MaxSteps) | |||
874 | return true; | |||
875 | return Found; | |||
876 | } | |||
877 | ||||
878 | /// Return true if all the users of N are contained in Nodes. | |||
879 | /// NOTE: Requires at least one match, but doesn't require them all. | |||
880 | static bool areOnlyUsersOf(ArrayRef<const SDNode *> Nodes, const SDNode *N); | |||
881 | ||||
882 | /// Return the number of values used by this operation. | |||
883 | unsigned getNumOperands() const { return NumOperands; } | |||
884 | ||||
885 | /// Return the maximum number of operands that a SDNode can hold. | |||
886 | static constexpr size_t getMaxNumOperands() { | |||
887 | return std::numeric_limits<decltype(SDNode::NumOperands)>::max(); | |||
888 | } | |||
889 | ||||
890 | /// Helper method returns the integer value of a ConstantSDNode operand. | |||
891 | inline uint64_t getConstantOperandVal(unsigned Num) const; | |||
892 | ||||
893 | /// Helper method returns the APInt of a ConstantSDNode operand. | |||
894 | inline const APInt &getConstantOperandAPInt(unsigned Num) const; | |||
895 | ||||
896 | const SDValue &getOperand(unsigned Num) const { | |||
897 | assert(Num < NumOperands && "Invalid child # of SDNode!")(static_cast <bool> (Num < NumOperands && "Invalid child # of SDNode!" ) ? void (0) : __assert_fail ("Num < NumOperands && \"Invalid child # of SDNode!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 897, __extension__ __PRETTY_FUNCTION__)); | |||
898 | return OperandList[Num]; | |||
899 | } | |||
900 | ||||
901 | using op_iterator = SDUse *; | |||
902 | ||||
903 | op_iterator op_begin() const { return OperandList; } | |||
904 | op_iterator op_end() const { return OperandList+NumOperands; } | |||
905 | ArrayRef<SDUse> ops() const { return makeArrayRef(op_begin(), op_end()); } | |||
906 | ||||
907 | /// Iterator for directly iterating over the operand SDValue's. | |||
908 | struct value_op_iterator | |||
909 | : iterator_adaptor_base<value_op_iterator, op_iterator, | |||
910 | std::random_access_iterator_tag, SDValue, | |||
911 | ptrdiff_t, value_op_iterator *, | |||
912 | value_op_iterator *> { | |||
913 | explicit value_op_iterator(SDUse *U = nullptr) | |||
914 | : iterator_adaptor_base(U) {} | |||
915 | ||||
916 | const SDValue &operator*() const { return I->get(); } | |||
917 | }; | |||
918 | ||||
919 | iterator_range<value_op_iterator> op_values() const { | |||
920 | return make_range(value_op_iterator(op_begin()), | |||
921 | value_op_iterator(op_end())); | |||
922 | } | |||
923 | ||||
924 | SDVTList getVTList() const { | |||
925 | SDVTList X = { ValueList, NumValues }; | |||
926 | return X; | |||
927 | } | |||
928 | ||||
929 | /// If this node has a glue operand, return the node | |||
930 | /// to which the glue operand points. Otherwise return NULL. | |||
931 | SDNode *getGluedNode() const { | |||
932 | if (getNumOperands() != 0 && | |||
933 | getOperand(getNumOperands()-1).getValueType() == MVT::Glue) | |||
934 | return getOperand(getNumOperands()-1).getNode(); | |||
935 | return nullptr; | |||
936 | } | |||
937 | ||||
938 | /// If this node has a glue value with a user, return | |||
939 | /// the user (there is at most one). Otherwise return NULL. | |||
940 | SDNode *getGluedUser() const { | |||
941 | for (use_iterator UI = use_begin(), UE = use_end(); UI != UE; ++UI) | |||
942 | if (UI.getUse().get().getValueType() == MVT::Glue) | |||
943 | return *UI; | |||
944 | return nullptr; | |||
945 | } | |||
946 | ||||
947 | SDNodeFlags getFlags() const { return Flags; } | |||
948 | void setFlags(SDNodeFlags NewFlags) { Flags = NewFlags; } | |||
949 | ||||
950 | /// Clear any flags in this node that aren't also set in Flags. | |||
951 | /// If Flags is not in a defined state then this has no effect. | |||
952 | void intersectFlagsWith(const SDNodeFlags Flags); | |||
953 | ||||
954 | /// Return the number of values defined/returned by this operator. | |||
955 | unsigned getNumValues() const { return NumValues; } | |||
956 | ||||
957 | /// Return the type of a specified result. | |||
958 | EVT getValueType(unsigned ResNo) const { | |||
959 | assert(ResNo < NumValues && "Illegal result number!")(static_cast <bool> (ResNo < NumValues && "Illegal result number!" ) ? void (0) : __assert_fail ("ResNo < NumValues && \"Illegal result number!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 959, __extension__ __PRETTY_FUNCTION__)); | |||
960 | return ValueList[ResNo]; | |||
961 | } | |||
962 | ||||
963 | /// Return the type of a specified result as a simple type. | |||
964 | MVT getSimpleValueType(unsigned ResNo) const { | |||
965 | return getValueType(ResNo).getSimpleVT(); | |||
966 | } | |||
967 | ||||
968 | /// Returns MVT::getSizeInBits(getValueType(ResNo)). | |||
969 | /// | |||
970 | /// If the value type is a scalable vector type, the scalable property will | |||
971 | /// be set and the runtime size will be a positive integer multiple of the | |||
972 | /// base size. | |||
973 | TypeSize getValueSizeInBits(unsigned ResNo) const { | |||
974 | return getValueType(ResNo).getSizeInBits(); | |||
975 | } | |||
976 | ||||
977 | using value_iterator = const EVT *; | |||
978 | ||||
979 | value_iterator value_begin() const { return ValueList; } | |||
980 | value_iterator value_end() const { return ValueList+NumValues; } | |||
981 | iterator_range<value_iterator> values() const { | |||
982 | return llvm::make_range(value_begin(), value_end()); | |||
983 | } | |||
984 | ||||
985 | /// Return the opcode of this operation for printing. | |||
986 | std::string getOperationName(const SelectionDAG *G = nullptr) const; | |||
987 | static const char* getIndexedModeName(ISD::MemIndexedMode AM); | |||
988 | void print_types(raw_ostream &OS, const SelectionDAG *G) const; | |||
989 | void print_details(raw_ostream &OS, const SelectionDAG *G) const; | |||
990 | void print(raw_ostream &OS, const SelectionDAG *G = nullptr) const; | |||
991 | void printr(raw_ostream &OS, const SelectionDAG *G = nullptr) const; | |||
992 | ||||
993 | /// Print a SelectionDAG node and all children down to | |||
994 | /// the leaves. The given SelectionDAG allows target-specific nodes | |||
995 | /// to be printed in human-readable form. Unlike printr, this will | |||
996 | /// print the whole DAG, including children that appear multiple | |||
997 | /// times. | |||
998 | /// | |||
999 | void printrFull(raw_ostream &O, const SelectionDAG *G = nullptr) const; | |||
1000 | ||||
1001 | /// Print a SelectionDAG node and children up to | |||
1002 | /// depth "depth." The given SelectionDAG allows target-specific | |||
1003 | /// nodes to be printed in human-readable form. Unlike printr, this | |||
1004 | /// will print children that appear multiple times wherever they are | |||
1005 | /// used. | |||
1006 | /// | |||
1007 | void printrWithDepth(raw_ostream &O, const SelectionDAG *G = nullptr, | |||
1008 | unsigned depth = 100) const; | |||
1009 | ||||
1010 | /// Dump this node, for debugging. | |||
1011 | void dump() const; | |||
1012 | ||||
1013 | /// Dump (recursively) this node and its use-def subgraph. | |||
1014 | void dumpr() const; | |||
1015 | ||||
1016 | /// Dump this node, for debugging. | |||
1017 | /// The given SelectionDAG allows target-specific nodes to be printed | |||
1018 | /// in human-readable form. | |||
1019 | void dump(const SelectionDAG *G) const; | |||
1020 | ||||
1021 | /// Dump (recursively) this node and its use-def subgraph. | |||
1022 | /// The given SelectionDAG allows target-specific nodes to be printed | |||
1023 | /// in human-readable form. | |||
1024 | void dumpr(const SelectionDAG *G) const; | |||
1025 | ||||
1026 | /// printrFull to dbgs(). The given SelectionDAG allows | |||
1027 | /// target-specific nodes to be printed in human-readable form. | |||
1028 | /// Unlike dumpr, this will print the whole DAG, including children | |||
1029 | /// that appear multiple times. | |||
1030 | void dumprFull(const SelectionDAG *G = nullptr) const; | |||
1031 | ||||
1032 | /// printrWithDepth to dbgs(). The given | |||
1033 | /// SelectionDAG allows target-specific nodes to be printed in | |||
1034 | /// human-readable form. Unlike dumpr, this will print children | |||
1035 | /// that appear multiple times wherever they are used. | |||
1036 | /// | |||
1037 | void dumprWithDepth(const SelectionDAG *G = nullptr, | |||
1038 | unsigned depth = 100) const; | |||
1039 | ||||
1040 | /// Gather unique data for the node. | |||
1041 | void Profile(FoldingSetNodeID &ID) const; | |||
1042 | ||||
1043 | /// This method should only be used by the SDUse class. | |||
1044 | void addUse(SDUse &U) { U.addToList(&UseList); } | |||
1045 | ||||
1046 | protected: | |||
1047 | static SDVTList getSDVTList(EVT VT) { | |||
1048 | SDVTList Ret = { getValueTypeList(VT), 1 }; | |||
1049 | return Ret; | |||
1050 | } | |||
1051 | ||||
1052 | /// Create an SDNode. | |||
1053 | /// | |||
1054 | /// SDNodes are created without any operands, and never own the operand | |||
1055 | /// storage. To add operands, see SelectionDAG::createOperands. | |||
1056 | SDNode(unsigned Opc, unsigned Order, DebugLoc dl, SDVTList VTs) | |||
1057 | : NodeType(Opc), ValueList(VTs.VTs), NumValues(VTs.NumVTs), | |||
1058 | IROrder(Order), debugLoc(std::move(dl)) { | |||
1059 | memset(&RawSDNodeBits, 0, sizeof(RawSDNodeBits)); | |||
1060 | assert(debugLoc.hasTrivialDestructor() && "Expected trivial destructor")(static_cast <bool> (debugLoc.hasTrivialDestructor() && "Expected trivial destructor") ? void (0) : __assert_fail ("debugLoc.hasTrivialDestructor() && \"Expected trivial destructor\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1060, __extension__ __PRETTY_FUNCTION__)); | |||
1061 | assert(NumValues == VTs.NumVTs &&(static_cast <bool> (NumValues == VTs.NumVTs && "NumValues wasn't wide enough for its operands!") ? void (0) : __assert_fail ("NumValues == VTs.NumVTs && \"NumValues wasn't wide enough for its operands!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1062, __extension__ __PRETTY_FUNCTION__)) | |||
1062 | "NumValues wasn't wide enough for its operands!")(static_cast <bool> (NumValues == VTs.NumVTs && "NumValues wasn't wide enough for its operands!") ? void (0) : __assert_fail ("NumValues == VTs.NumVTs && \"NumValues wasn't wide enough for its operands!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1062, __extension__ __PRETTY_FUNCTION__)); | |||
1063 | } | |||
1064 | ||||
1065 | /// Release the operands and set this node to have zero operands. | |||
1066 | void DropOperands(); | |||
1067 | }; | |||
1068 | ||||
1069 | /// Wrapper class for IR location info (IR ordering and DebugLoc) to be passed | |||
1070 | /// into SDNode creation functions. | |||
1071 | /// When an SDNode is created from the DAGBuilder, the DebugLoc is extracted | |||
1072 | /// from the original Instruction, and IROrder is the ordinal position of | |||
1073 | /// the instruction. | |||
1074 | /// When an SDNode is created after the DAG is being built, both DebugLoc and | |||
1075 | /// the IROrder are propagated from the original SDNode. | |||
1076 | /// So SDLoc class provides two constructors besides the default one, one to | |||
1077 | /// be used by the DAGBuilder, the other to be used by others. | |||
1078 | class SDLoc { | |||
1079 | private: | |||
1080 | DebugLoc DL; | |||
1081 | int IROrder = 0; | |||
1082 | ||||
1083 | public: | |||
1084 | SDLoc() = default; | |||
1085 | SDLoc(const SDNode *N) : DL(N->getDebugLoc()), IROrder(N->getIROrder()) {} | |||
1086 | SDLoc(const SDValue V) : SDLoc(V.getNode()) {} | |||
1087 | SDLoc(const Instruction *I, int Order) : IROrder(Order) { | |||
1088 | assert(Order >= 0 && "bad IROrder")(static_cast <bool> (Order >= 0 && "bad IROrder" ) ? void (0) : __assert_fail ("Order >= 0 && \"bad IROrder\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1088, __extension__ __PRETTY_FUNCTION__)); | |||
1089 | if (I) | |||
1090 | DL = I->getDebugLoc(); | |||
1091 | } | |||
1092 | ||||
1093 | unsigned getIROrder() const { return IROrder; } | |||
1094 | const DebugLoc &getDebugLoc() const { return DL; } | |||
1095 | }; | |||
1096 | ||||
1097 | // Define inline functions from the SDValue class. | |||
1098 | ||||
1099 | inline SDValue::SDValue(SDNode *node, unsigned resno) | |||
1100 | : Node(node), ResNo(resno) { | |||
1101 | // Explicitly check for !ResNo to avoid use-after-free, because there are | |||
1102 | // callers that use SDValue(N, 0) with a deleted N to indicate successful | |||
1103 | // combines. | |||
1104 | assert((!Node || !ResNo || ResNo < Node->getNumValues()) &&(static_cast <bool> ((!Node || !ResNo || ResNo < Node ->getNumValues()) && "Invalid result number for the given node!" ) ? void (0) : __assert_fail ("(!Node || !ResNo || ResNo < Node->getNumValues()) && \"Invalid result number for the given node!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1105, __extension__ __PRETTY_FUNCTION__)) | |||
1105 | "Invalid result number for the given node!")(static_cast <bool> ((!Node || !ResNo || ResNo < Node ->getNumValues()) && "Invalid result number for the given node!" ) ? void (0) : __assert_fail ("(!Node || !ResNo || ResNo < Node->getNumValues()) && \"Invalid result number for the given node!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1105, __extension__ __PRETTY_FUNCTION__)); | |||
1106 | assert(ResNo < -2U && "Cannot use result numbers reserved for DenseMaps.")(static_cast <bool> (ResNo < -2U && "Cannot use result numbers reserved for DenseMaps." ) ? void (0) : __assert_fail ("ResNo < -2U && \"Cannot use result numbers reserved for DenseMaps.\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1106, __extension__ __PRETTY_FUNCTION__)); | |||
1107 | } | |||
1108 | ||||
1109 | inline unsigned SDValue::getOpcode() const { | |||
1110 | return Node->getOpcode(); | |||
1111 | } | |||
1112 | ||||
1113 | inline EVT SDValue::getValueType() const { | |||
1114 | return Node->getValueType(ResNo); | |||
1115 | } | |||
1116 | ||||
1117 | inline unsigned SDValue::getNumOperands() const { | |||
1118 | return Node->getNumOperands(); | |||
1119 | } | |||
1120 | ||||
1121 | inline const SDValue &SDValue::getOperand(unsigned i) const { | |||
1122 | return Node->getOperand(i); | |||
| ||||
1123 | } | |||
1124 | ||||
1125 | inline uint64_t SDValue::getConstantOperandVal(unsigned i) const { | |||
1126 | return Node->getConstantOperandVal(i); | |||
1127 | } | |||
1128 | ||||
1129 | inline const APInt &SDValue::getConstantOperandAPInt(unsigned i) const { | |||
1130 | return Node->getConstantOperandAPInt(i); | |||
1131 | } | |||
1132 | ||||
1133 | inline bool SDValue::isTargetOpcode() const { | |||
1134 | return Node->isTargetOpcode(); | |||
1135 | } | |||
1136 | ||||
1137 | inline bool SDValue::isTargetMemoryOpcode() const { | |||
1138 | return Node->isTargetMemoryOpcode(); | |||
1139 | } | |||
1140 | ||||
1141 | inline bool SDValue::isMachineOpcode() const { | |||
1142 | return Node->isMachineOpcode(); | |||
1143 | } | |||
1144 | ||||
1145 | inline unsigned SDValue::getMachineOpcode() const { | |||
1146 | return Node->getMachineOpcode(); | |||
1147 | } | |||
1148 | ||||
1149 | inline bool SDValue::isUndef() const { | |||
1150 | return Node->isUndef(); | |||
1151 | } | |||
1152 | ||||
1153 | inline bool SDValue::use_empty() const { | |||
1154 | return !Node->hasAnyUseOfValue(ResNo); | |||
1155 | } | |||
1156 | ||||
1157 | inline bool SDValue::hasOneUse() const { | |||
1158 | return Node->hasNUsesOfValue(1, ResNo); | |||
1159 | } | |||
1160 | ||||
1161 | inline const DebugLoc &SDValue::getDebugLoc() const { | |||
1162 | return Node->getDebugLoc(); | |||
1163 | } | |||
1164 | ||||
1165 | inline void SDValue::dump() const { | |||
1166 | return Node->dump(); | |||
1167 | } | |||
1168 | ||||
1169 | inline void SDValue::dump(const SelectionDAG *G) const { | |||
1170 | return Node->dump(G); | |||
1171 | } | |||
1172 | ||||
1173 | inline void SDValue::dumpr() const { | |||
1174 | return Node->dumpr(); | |||
1175 | } | |||
1176 | ||||
1177 | inline void SDValue::dumpr(const SelectionDAG *G) const { | |||
1178 | return Node->dumpr(G); | |||
1179 | } | |||
1180 | ||||
1181 | // Define inline functions from the SDUse class. | |||
1182 | ||||
1183 | inline void SDUse::set(const SDValue &V) { | |||
1184 | if (Val.getNode()) removeFromList(); | |||
1185 | Val = V; | |||
1186 | if (V.getNode()) V.getNode()->addUse(*this); | |||
1187 | } | |||
1188 | ||||
1189 | inline void SDUse::setInitial(const SDValue &V) { | |||
1190 | Val = V; | |||
1191 | V.getNode()->addUse(*this); | |||
1192 | } | |||
1193 | ||||
1194 | inline void SDUse::setNode(SDNode *N) { | |||
1195 | if (Val.getNode()) removeFromList(); | |||
1196 | Val.setNode(N); | |||
1197 | if (N) N->addUse(*this); | |||
1198 | } | |||
1199 | ||||
1200 | /// This class is used to form a handle around another node that | |||
1201 | /// is persistent and is updated across invocations of replaceAllUsesWith on its | |||
1202 | /// operand. This node should be directly created by end-users and not added to | |||
1203 | /// the AllNodes list. | |||
1204 | class HandleSDNode : public SDNode { | |||
1205 | SDUse Op; | |||
1206 | ||||
1207 | public: | |||
1208 | explicit HandleSDNode(SDValue X) | |||
1209 | : SDNode(ISD::HANDLENODE, 0, DebugLoc(), getSDVTList(MVT::Other)) { | |||
1210 | // HandleSDNodes are never inserted into the DAG, so they won't be | |||
1211 | // auto-numbered. Use ID 65535 as a sentinel. | |||
1212 | PersistentId = 0xffff; | |||
1213 | ||||
1214 | // Manually set up the operand list. This node type is special in that it's | |||
1215 | // always stack allocated and SelectionDAG does not manage its operands. | |||
1216 | // TODO: This should either (a) not be in the SDNode hierarchy, or (b) not | |||
1217 | // be so special. | |||
1218 | Op.setUser(this); | |||
1219 | Op.setInitial(X); | |||
1220 | NumOperands = 1; | |||
1221 | OperandList = &Op; | |||
1222 | } | |||
1223 | ~HandleSDNode(); | |||
1224 | ||||
1225 | const SDValue &getValue() const { return Op; } | |||
1226 | }; | |||
1227 | ||||
1228 | class AddrSpaceCastSDNode : public SDNode { | |||
1229 | private: | |||
1230 | unsigned SrcAddrSpace; | |||
1231 | unsigned DestAddrSpace; | |||
1232 | ||||
1233 | public: | |||
1234 | AddrSpaceCastSDNode(unsigned Order, const DebugLoc &dl, EVT VT, | |||
1235 | unsigned SrcAS, unsigned DestAS); | |||
1236 | ||||
1237 | unsigned getSrcAddressSpace() const { return SrcAddrSpace; } | |||
1238 | unsigned getDestAddressSpace() const { return DestAddrSpace; } | |||
1239 | ||||
1240 | static bool classof(const SDNode *N) { | |||
1241 | return N->getOpcode() == ISD::ADDRSPACECAST; | |||
1242 | } | |||
1243 | }; | |||
1244 | ||||
1245 | /// This is an abstract virtual class for memory operations. | |||
1246 | class MemSDNode : public SDNode { | |||
1247 | private: | |||
1248 | // VT of in-memory value. | |||
1249 | EVT MemoryVT; | |||
1250 | ||||
1251 | protected: | |||
1252 | /// Memory reference information. | |||
1253 | MachineMemOperand *MMO; | |||
1254 | ||||
1255 | public: | |||
1256 | MemSDNode(unsigned Opc, unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
1257 | EVT memvt, MachineMemOperand *MMO); | |||
1258 | ||||
1259 | bool readMem() const { return MMO->isLoad(); } | |||
1260 | bool writeMem() const { return MMO->isStore(); } | |||
1261 | ||||
1262 | /// Returns alignment and volatility of the memory access | |||
1263 | Align getOriginalAlign() const { return MMO->getBaseAlign(); } | |||
1264 | Align getAlign() const { return MMO->getAlign(); } | |||
1265 | // FIXME: Remove once transition to getAlign is over. | |||
1266 | unsigned getAlignment() const { return MMO->getAlign().value(); } | |||
1267 | ||||
1268 | /// Return the SubclassData value, without HasDebugValue. This contains an | |||
1269 | /// encoding of the volatile flag, as well as bits used by subclasses. This | |||
1270 | /// function should only be used to compute a FoldingSetNodeID value. | |||
1271 | /// The HasDebugValue bit is masked out because CSE map needs to match | |||
1272 | /// nodes with debug info with nodes without debug info. Same is about | |||
1273 | /// isDivergent bit. | |||
1274 | unsigned getRawSubclassData() const { | |||
1275 | uint16_t Data; | |||
1276 | union { | |||
1277 | char RawSDNodeBits[sizeof(uint16_t)]; | |||
1278 | SDNodeBitfields SDNodeBits; | |||
1279 | }; | |||
1280 | memcpy(&RawSDNodeBits, &this->RawSDNodeBits, sizeof(this->RawSDNodeBits)); | |||
1281 | SDNodeBits.HasDebugValue = 0; | |||
1282 | SDNodeBits.IsDivergent = false; | |||
1283 | memcpy(&Data, &RawSDNodeBits, sizeof(RawSDNodeBits)); | |||
1284 | return Data; | |||
1285 | } | |||
1286 | ||||
1287 | bool isVolatile() const { return MemSDNodeBits.IsVolatile; } | |||
1288 | bool isNonTemporal() const { return MemSDNodeBits.IsNonTemporal; } | |||
1289 | bool isDereferenceable() const { return MemSDNodeBits.IsDereferenceable; } | |||
1290 | bool isInvariant() const { return MemSDNodeBits.IsInvariant; } | |||
1291 | ||||
1292 | // Returns the offset from the location of the access. | |||
1293 | int64_t getSrcValueOffset() const { return MMO->getOffset(); } | |||
1294 | ||||
1295 | /// Returns the AA info that describes the dereference. | |||
1296 | AAMDNodes getAAInfo() const { return MMO->getAAInfo(); } | |||
1297 | ||||
1298 | /// Returns the Ranges that describes the dereference. | |||
1299 | const MDNode *getRanges() const { return MMO->getRanges(); } | |||
1300 | ||||
1301 | /// Returns the synchronization scope ID for this memory operation. | |||
1302 | SyncScope::ID getSyncScopeID() const { return MMO->getSyncScopeID(); } | |||
1303 | ||||
1304 | /// Return the atomic ordering requirements for this memory operation. For | |||
1305 | /// cmpxchg atomic operations, return the atomic ordering requirements when | |||
1306 | /// store occurs. | |||
1307 | AtomicOrdering getSuccessOrdering() const { | |||
1308 | return MMO->getSuccessOrdering(); | |||
1309 | } | |||
1310 | ||||
1311 | /// Return a single atomic ordering that is at least as strong as both the | |||
1312 | /// success and failure orderings for an atomic operation. (For operations | |||
1313 | /// other than cmpxchg, this is equivalent to getSuccessOrdering().) | |||
1314 | AtomicOrdering getMergedOrdering() const { return MMO->getMergedOrdering(); } | |||
1315 | ||||
1316 | /// Return true if the memory operation ordering is Unordered or higher. | |||
1317 | bool isAtomic() const { return MMO->isAtomic(); } | |||
1318 | ||||
1319 | /// Returns true if the memory operation doesn't imply any ordering | |||
1320 | /// constraints on surrounding memory operations beyond the normal memory | |||
1321 | /// aliasing rules. | |||
1322 | bool isUnordered() const { return MMO->isUnordered(); } | |||
1323 | ||||
1324 | /// Returns true if the memory operation is neither atomic or volatile. | |||
1325 | bool isSimple() const { return !isAtomic() && !isVolatile(); } | |||
1326 | ||||
1327 | /// Return the type of the in-memory value. | |||
1328 | EVT getMemoryVT() const { return MemoryVT; } | |||
1329 | ||||
1330 | /// Return a MachineMemOperand object describing the memory | |||
1331 | /// reference performed by operation. | |||
1332 | MachineMemOperand *getMemOperand() const { return MMO; } | |||
1333 | ||||
1334 | const MachinePointerInfo &getPointerInfo() const { | |||
1335 | return MMO->getPointerInfo(); | |||
1336 | } | |||
1337 | ||||
1338 | /// Return the address space for the associated pointer | |||
1339 | unsigned getAddressSpace() const { | |||
1340 | return getPointerInfo().getAddrSpace(); | |||
1341 | } | |||
1342 | ||||
1343 | /// Update this MemSDNode's MachineMemOperand information | |||
1344 | /// to reflect the alignment of NewMMO, if it has a greater alignment. | |||
1345 | /// This must only be used when the new alignment applies to all users of | |||
1346 | /// this MachineMemOperand. | |||
1347 | void refineAlignment(const MachineMemOperand *NewMMO) { | |||
1348 | MMO->refineAlignment(NewMMO); | |||
1349 | } | |||
1350 | ||||
1351 | const SDValue &getChain() const { return getOperand(0); } | |||
1352 | ||||
1353 | const SDValue &getBasePtr() const { | |||
1354 | switch (getOpcode()) { | |||
1355 | case ISD::STORE: | |||
1356 | case ISD::MSTORE: | |||
1357 | return getOperand(2); | |||
1358 | case ISD::MGATHER: | |||
1359 | case ISD::MSCATTER: | |||
1360 | return getOperand(3); | |||
1361 | default: | |||
1362 | return getOperand(1); | |||
1363 | } | |||
1364 | } | |||
1365 | ||||
1366 | // Methods to support isa and dyn_cast | |||
1367 | static bool classof(const SDNode *N) { | |||
1368 | // For some targets, we lower some target intrinsics to a MemIntrinsicNode | |||
1369 | // with either an intrinsic or a target opcode. | |||
1370 | switch (N->getOpcode()) { | |||
1371 | case ISD::LOAD: | |||
1372 | case ISD::STORE: | |||
1373 | case ISD::PREFETCH: | |||
1374 | case ISD::ATOMIC_CMP_SWAP: | |||
1375 | case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS: | |||
1376 | case ISD::ATOMIC_SWAP: | |||
1377 | case ISD::ATOMIC_LOAD_ADD: | |||
1378 | case ISD::ATOMIC_LOAD_SUB: | |||
1379 | case ISD::ATOMIC_LOAD_AND: | |||
1380 | case ISD::ATOMIC_LOAD_CLR: | |||
1381 | case ISD::ATOMIC_LOAD_OR: | |||
1382 | case ISD::ATOMIC_LOAD_XOR: | |||
1383 | case ISD::ATOMIC_LOAD_NAND: | |||
1384 | case ISD::ATOMIC_LOAD_MIN: | |||
1385 | case ISD::ATOMIC_LOAD_MAX: | |||
1386 | case ISD::ATOMIC_LOAD_UMIN: | |||
1387 | case ISD::ATOMIC_LOAD_UMAX: | |||
1388 | case ISD::ATOMIC_LOAD_FADD: | |||
1389 | case ISD::ATOMIC_LOAD_FSUB: | |||
1390 | case ISD::ATOMIC_LOAD: | |||
1391 | case ISD::ATOMIC_STORE: | |||
1392 | case ISD::MLOAD: | |||
1393 | case ISD::MSTORE: | |||
1394 | case ISD::MGATHER: | |||
1395 | case ISD::MSCATTER: | |||
1396 | return true; | |||
1397 | default: | |||
1398 | return N->isMemIntrinsic() || N->isTargetMemoryOpcode(); | |||
1399 | } | |||
1400 | } | |||
1401 | }; | |||
1402 | ||||
1403 | /// This is an SDNode representing atomic operations. | |||
1404 | class AtomicSDNode : public MemSDNode { | |||
1405 | public: | |||
1406 | AtomicSDNode(unsigned Opc, unsigned Order, const DebugLoc &dl, SDVTList VTL, | |||
1407 | EVT MemVT, MachineMemOperand *MMO) | |||
1408 | : MemSDNode(Opc, Order, dl, VTL, MemVT, MMO) { | |||
1409 | assert(((Opc != ISD::ATOMIC_LOAD && Opc != ISD::ATOMIC_STORE) ||(static_cast <bool> (((Opc != ISD::ATOMIC_LOAD && Opc != ISD::ATOMIC_STORE) || MMO->isAtomic()) && "then why are we using an AtomicSDNode?" ) ? void (0) : __assert_fail ("((Opc != ISD::ATOMIC_LOAD && Opc != ISD::ATOMIC_STORE) || MMO->isAtomic()) && \"then why are we using an AtomicSDNode?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1410, __extension__ __PRETTY_FUNCTION__)) | |||
1410 | MMO->isAtomic()) && "then why are we using an AtomicSDNode?")(static_cast <bool> (((Opc != ISD::ATOMIC_LOAD && Opc != ISD::ATOMIC_STORE) || MMO->isAtomic()) && "then why are we using an AtomicSDNode?" ) ? void (0) : __assert_fail ("((Opc != ISD::ATOMIC_LOAD && Opc != ISD::ATOMIC_STORE) || MMO->isAtomic()) && \"then why are we using an AtomicSDNode?\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1410, __extension__ __PRETTY_FUNCTION__)); | |||
1411 | } | |||
1412 | ||||
1413 | const SDValue &getBasePtr() const { return getOperand(1); } | |||
1414 | const SDValue &getVal() const { return getOperand(2); } | |||
1415 | ||||
1416 | /// Returns true if this SDNode represents cmpxchg atomic operation, false | |||
1417 | /// otherwise. | |||
1418 | bool isCompareAndSwap() const { | |||
1419 | unsigned Op = getOpcode(); | |||
1420 | return Op == ISD::ATOMIC_CMP_SWAP || | |||
1421 | Op == ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS; | |||
1422 | } | |||
1423 | ||||
1424 | /// For cmpxchg atomic operations, return the atomic ordering requirements | |||
1425 | /// when store does not occur. | |||
1426 | AtomicOrdering getFailureOrdering() const { | |||
1427 | assert(isCompareAndSwap() && "Must be cmpxchg operation")(static_cast <bool> (isCompareAndSwap() && "Must be cmpxchg operation" ) ? void (0) : __assert_fail ("isCompareAndSwap() && \"Must be cmpxchg operation\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1427, __extension__ __PRETTY_FUNCTION__)); | |||
1428 | return MMO->getFailureOrdering(); | |||
1429 | } | |||
1430 | ||||
1431 | // Methods to support isa and dyn_cast | |||
1432 | static bool classof(const SDNode *N) { | |||
1433 | return N->getOpcode() == ISD::ATOMIC_CMP_SWAP || | |||
1434 | N->getOpcode() == ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS || | |||
1435 | N->getOpcode() == ISD::ATOMIC_SWAP || | |||
1436 | N->getOpcode() == ISD::ATOMIC_LOAD_ADD || | |||
1437 | N->getOpcode() == ISD::ATOMIC_LOAD_SUB || | |||
1438 | N->getOpcode() == ISD::ATOMIC_LOAD_AND || | |||
1439 | N->getOpcode() == ISD::ATOMIC_LOAD_CLR || | |||
1440 | N->getOpcode() == ISD::ATOMIC_LOAD_OR || | |||
1441 | N->getOpcode() == ISD::ATOMIC_LOAD_XOR || | |||
1442 | N->getOpcode() == ISD::ATOMIC_LOAD_NAND || | |||
1443 | N->getOpcode() == ISD::ATOMIC_LOAD_MIN || | |||
1444 | N->getOpcode() == ISD::ATOMIC_LOAD_MAX || | |||
1445 | N->getOpcode() == ISD::ATOMIC_LOAD_UMIN || | |||
1446 | N->getOpcode() == ISD::ATOMIC_LOAD_UMAX || | |||
1447 | N->getOpcode() == ISD::ATOMIC_LOAD_FADD || | |||
1448 | N->getOpcode() == ISD::ATOMIC_LOAD_FSUB || | |||
1449 | N->getOpcode() == ISD::ATOMIC_LOAD || | |||
1450 | N->getOpcode() == ISD::ATOMIC_STORE; | |||
1451 | } | |||
1452 | }; | |||
1453 | ||||
1454 | /// This SDNode is used for target intrinsics that touch | |||
1455 | /// memory and need an associated MachineMemOperand. Its opcode may be | |||
1456 | /// INTRINSIC_VOID, INTRINSIC_W_CHAIN, PREFETCH, or a target-specific opcode | |||
1457 | /// with a value not less than FIRST_TARGET_MEMORY_OPCODE. | |||
1458 | class MemIntrinsicSDNode : public MemSDNode { | |||
1459 | public: | |||
1460 | MemIntrinsicSDNode(unsigned Opc, unsigned Order, const DebugLoc &dl, | |||
1461 | SDVTList VTs, EVT MemoryVT, MachineMemOperand *MMO) | |||
1462 | : MemSDNode(Opc, Order, dl, VTs, MemoryVT, MMO) { | |||
1463 | SDNodeBits.IsMemIntrinsic = true; | |||
1464 | } | |||
1465 | ||||
1466 | // Methods to support isa and dyn_cast | |||
1467 | static bool classof(const SDNode *N) { | |||
1468 | // We lower some target intrinsics to their target opcode | |||
1469 | // early a node with a target opcode can be of this class | |||
1470 | return N->isMemIntrinsic() || | |||
1471 | N->getOpcode() == ISD::PREFETCH || | |||
1472 | N->isTargetMemoryOpcode(); | |||
1473 | } | |||
1474 | }; | |||
1475 | ||||
1476 | /// This SDNode is used to implement the code generator | |||
1477 | /// support for the llvm IR shufflevector instruction. It combines elements | |||
1478 | /// from two input vectors into a new input vector, with the selection and | |||
1479 | /// ordering of elements determined by an array of integers, referred to as | |||
1480 | /// the shuffle mask. For input vectors of width N, mask indices of 0..N-1 | |||
1481 | /// refer to elements from the LHS input, and indices from N to 2N-1 the RHS. | |||
1482 | /// An index of -1 is treated as undef, such that the code generator may put | |||
1483 | /// any value in the corresponding element of the result. | |||
1484 | class ShuffleVectorSDNode : public SDNode { | |||
1485 | // The memory for Mask is owned by the SelectionDAG's OperandAllocator, and | |||
1486 | // is freed when the SelectionDAG object is destroyed. | |||
1487 | const int *Mask; | |||
1488 | ||||
1489 | protected: | |||
1490 | friend class SelectionDAG; | |||
1491 | ||||
1492 | ShuffleVectorSDNode(EVT VT, unsigned Order, const DebugLoc &dl, const int *M) | |||
1493 | : SDNode(ISD::VECTOR_SHUFFLE, Order, dl, getSDVTList(VT)), Mask(M) {} | |||
1494 | ||||
1495 | public: | |||
1496 | ArrayRef<int> getMask() const { | |||
1497 | EVT VT = getValueType(0); | |||
1498 | return makeArrayRef(Mask, VT.getVectorNumElements()); | |||
1499 | } | |||
1500 | ||||
1501 | int getMaskElt(unsigned Idx) const { | |||
1502 | assert(Idx < getValueType(0).getVectorNumElements() && "Idx out of range!")(static_cast <bool> (Idx < getValueType(0).getVectorNumElements () && "Idx out of range!") ? void (0) : __assert_fail ("Idx < getValueType(0).getVectorNumElements() && \"Idx out of range!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1502, __extension__ __PRETTY_FUNCTION__)); | |||
1503 | return Mask[Idx]; | |||
1504 | } | |||
1505 | ||||
1506 | bool isSplat() const { return isSplatMask(Mask, getValueType(0)); } | |||
1507 | ||||
1508 | int getSplatIndex() const { | |||
1509 | assert(isSplat() && "Cannot get splat index for non-splat!")(static_cast <bool> (isSplat() && "Cannot get splat index for non-splat!" ) ? void (0) : __assert_fail ("isSplat() && \"Cannot get splat index for non-splat!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1509, __extension__ __PRETTY_FUNCTION__)); | |||
1510 | EVT VT = getValueType(0); | |||
1511 | for (unsigned i = 0, e = VT.getVectorNumElements(); i != e; ++i) | |||
1512 | if (Mask[i] >= 0) | |||
1513 | return Mask[i]; | |||
1514 | ||||
1515 | // We can choose any index value here and be correct because all elements | |||
1516 | // are undefined. Return 0 for better potential for callers to simplify. | |||
1517 | return 0; | |||
1518 | } | |||
1519 | ||||
1520 | static bool isSplatMask(const int *Mask, EVT VT); | |||
1521 | ||||
1522 | /// Change values in a shuffle permute mask assuming | |||
1523 | /// the two vector operands have swapped position. | |||
1524 | static void commuteMask(MutableArrayRef<int> Mask) { | |||
1525 | unsigned NumElems = Mask.size(); | |||
1526 | for (unsigned i = 0; i != NumElems; ++i) { | |||
1527 | int idx = Mask[i]; | |||
1528 | if (idx < 0) | |||
1529 | continue; | |||
1530 | else if (idx < (int)NumElems) | |||
1531 | Mask[i] = idx + NumElems; | |||
1532 | else | |||
1533 | Mask[i] = idx - NumElems; | |||
1534 | } | |||
1535 | } | |||
1536 | ||||
1537 | static bool classof(const SDNode *N) { | |||
1538 | return N->getOpcode() == ISD::VECTOR_SHUFFLE; | |||
1539 | } | |||
1540 | }; | |||
1541 | ||||
1542 | class ConstantSDNode : public SDNode { | |||
1543 | friend class SelectionDAG; | |||
1544 | ||||
1545 | const ConstantInt *Value; | |||
1546 | ||||
1547 | ConstantSDNode(bool isTarget, bool isOpaque, const ConstantInt *val, EVT VT) | |||
1548 | : SDNode(isTarget ? ISD::TargetConstant : ISD::Constant, 0, DebugLoc(), | |||
1549 | getSDVTList(VT)), | |||
1550 | Value(val) { | |||
1551 | ConstantSDNodeBits.IsOpaque = isOpaque; | |||
1552 | } | |||
1553 | ||||
1554 | public: | |||
1555 | const ConstantInt *getConstantIntValue() const { return Value; } | |||
1556 | const APInt &getAPIntValue() const { return Value->getValue(); } | |||
1557 | uint64_t getZExtValue() const { return Value->getZExtValue(); } | |||
1558 | int64_t getSExtValue() const { return Value->getSExtValue(); } | |||
1559 | uint64_t getLimitedValue(uint64_t Limit = UINT64_MAX(18446744073709551615UL)) { | |||
1560 | return Value->getLimitedValue(Limit); | |||
1561 | } | |||
1562 | MaybeAlign getMaybeAlignValue() const { return Value->getMaybeAlignValue(); } | |||
1563 | Align getAlignValue() const { return Value->getAlignValue(); } | |||
1564 | ||||
1565 | bool isOne() const { return Value->isOne(); } | |||
1566 | bool isNullValue() const { return Value->isZero(); } | |||
1567 | bool isAllOnesValue() const { return Value->isMinusOne(); } | |||
1568 | bool isMaxSignedValue() const { return Value->isMaxValue(true); } | |||
1569 | bool isMinSignedValue() const { return Value->isMinValue(true); } | |||
1570 | ||||
1571 | bool isOpaque() const { return ConstantSDNodeBits.IsOpaque; } | |||
1572 | ||||
1573 | static bool classof(const SDNode *N) { | |||
1574 | return N->getOpcode() == ISD::Constant || | |||
1575 | N->getOpcode() == ISD::TargetConstant; | |||
1576 | } | |||
1577 | }; | |||
1578 | ||||
1579 | uint64_t SDNode::getConstantOperandVal(unsigned Num) const { | |||
1580 | return cast<ConstantSDNode>(getOperand(Num))->getZExtValue(); | |||
1581 | } | |||
1582 | ||||
1583 | const APInt &SDNode::getConstantOperandAPInt(unsigned Num) const { | |||
1584 | return cast<ConstantSDNode>(getOperand(Num))->getAPIntValue(); | |||
1585 | } | |||
1586 | ||||
1587 | class ConstantFPSDNode : public SDNode { | |||
1588 | friend class SelectionDAG; | |||
1589 | ||||
1590 | const ConstantFP *Value; | |||
1591 | ||||
1592 | ConstantFPSDNode(bool isTarget, const ConstantFP *val, EVT VT) | |||
1593 | : SDNode(isTarget ? ISD::TargetConstantFP : ISD::ConstantFP, 0, | |||
1594 | DebugLoc(), getSDVTList(VT)), | |||
1595 | Value(val) {} | |||
1596 | ||||
1597 | public: | |||
1598 | const APFloat& getValueAPF() const { return Value->getValueAPF(); } | |||
1599 | const ConstantFP *getConstantFPValue() const { return Value; } | |||
1600 | ||||
1601 | /// Return true if the value is positive or negative zero. | |||
1602 | bool isZero() const { return Value->isZero(); } | |||
1603 | ||||
1604 | /// Return true if the value is a NaN. | |||
1605 | bool isNaN() const { return Value->isNaN(); } | |||
1606 | ||||
1607 | /// Return true if the value is an infinity | |||
1608 | bool isInfinity() const { return Value->isInfinity(); } | |||
1609 | ||||
1610 | /// Return true if the value is negative. | |||
1611 | bool isNegative() const { return Value->isNegative(); } | |||
1612 | ||||
1613 | /// We don't rely on operator== working on double values, as | |||
1614 | /// it returns true for things that are clearly not equal, like -0.0 and 0.0. | |||
1615 | /// As such, this method can be used to do an exact bit-for-bit comparison of | |||
1616 | /// two floating point values. | |||
1617 | ||||
1618 | /// We leave the version with the double argument here because it's just so | |||
1619 | /// convenient to write "2.0" and the like. Without this function we'd | |||
1620 | /// have to duplicate its logic everywhere it's called. | |||
1621 | bool isExactlyValue(double V) const { | |||
1622 | return Value->getValueAPF().isExactlyValue(V); | |||
1623 | } | |||
1624 | bool isExactlyValue(const APFloat& V) const; | |||
1625 | ||||
1626 | static bool isValueValidForType(EVT VT, const APFloat& Val); | |||
1627 | ||||
1628 | static bool classof(const SDNode *N) { | |||
1629 | return N->getOpcode() == ISD::ConstantFP || | |||
1630 | N->getOpcode() == ISD::TargetConstantFP; | |||
1631 | } | |||
1632 | }; | |||
1633 | ||||
1634 | /// Returns true if \p V is a constant integer zero. | |||
1635 | bool isNullConstant(SDValue V); | |||
1636 | ||||
1637 | /// Returns true if \p V is an FP constant with a value of positive zero. | |||
1638 | bool isNullFPConstant(SDValue V); | |||
1639 | ||||
1640 | /// Returns true if \p V is an integer constant with all bits set. | |||
1641 | bool isAllOnesConstant(SDValue V); | |||
1642 | ||||
1643 | /// Returns true if \p V is a constant integer one. | |||
1644 | bool isOneConstant(SDValue V); | |||
1645 | ||||
1646 | /// Return the non-bitcasted source operand of \p V if it exists. | |||
1647 | /// If \p V is not a bitcasted value, it is returned as-is. | |||
1648 | SDValue peekThroughBitcasts(SDValue V); | |||
1649 | ||||
1650 | /// Return the non-bitcasted and one-use source operand of \p V if it exists. | |||
1651 | /// If \p V is not a bitcasted one-use value, it is returned as-is. | |||
1652 | SDValue peekThroughOneUseBitcasts(SDValue V); | |||
1653 | ||||
1654 | /// Return the non-extracted vector source operand of \p V if it exists. | |||
1655 | /// If \p V is not an extracted subvector, it is returned as-is. | |||
1656 | SDValue peekThroughExtractSubvectors(SDValue V); | |||
1657 | ||||
1658 | /// Returns true if \p V is a bitwise not operation. Assumes that an all ones | |||
1659 | /// constant is canonicalized to be operand 1. | |||
1660 | bool isBitwiseNot(SDValue V, bool AllowUndefs = false); | |||
1661 | ||||
1662 | /// Returns the SDNode if it is a constant splat BuildVector or constant int. | |||
1663 | ConstantSDNode *isConstOrConstSplat(SDValue N, bool AllowUndefs = false, | |||
1664 | bool AllowTruncation = false); | |||
1665 | ||||
1666 | /// Returns the SDNode if it is a demanded constant splat BuildVector or | |||
1667 | /// constant int. | |||
1668 | ConstantSDNode *isConstOrConstSplat(SDValue N, const APInt &DemandedElts, | |||
1669 | bool AllowUndefs = false, | |||
1670 | bool AllowTruncation = false); | |||
1671 | ||||
1672 | /// Returns the SDNode if it is a constant splat BuildVector or constant float. | |||
1673 | ConstantFPSDNode *isConstOrConstSplatFP(SDValue N, bool AllowUndefs = false); | |||
1674 | ||||
1675 | /// Returns the SDNode if it is a demanded constant splat BuildVector or | |||
1676 | /// constant float. | |||
1677 | ConstantFPSDNode *isConstOrConstSplatFP(SDValue N, const APInt &DemandedElts, | |||
1678 | bool AllowUndefs = false); | |||
1679 | ||||
1680 | /// Return true if the value is a constant 0 integer or a splatted vector of | |||
1681 | /// a constant 0 integer (with no undefs by default). | |||
1682 | /// Build vector implicit truncation is not an issue for null values. | |||
1683 | bool isNullOrNullSplat(SDValue V, bool AllowUndefs = false); | |||
1684 | ||||
1685 | /// Return true if the value is a constant 1 integer or a splatted vector of a | |||
1686 | /// constant 1 integer (with no undefs). | |||
1687 | /// Does not permit build vector implicit truncation. | |||
1688 | bool isOneOrOneSplat(SDValue V, bool AllowUndefs = false); | |||
1689 | ||||
1690 | /// Return true if the value is a constant -1 integer or a splatted vector of a | |||
1691 | /// constant -1 integer (with no undefs). | |||
1692 | /// Does not permit build vector implicit truncation. | |||
1693 | bool isAllOnesOrAllOnesSplat(SDValue V, bool AllowUndefs = false); | |||
1694 | ||||
1695 | /// Return true if \p V is either a integer or FP constant. | |||
1696 | inline bool isIntOrFPConstant(SDValue V) { | |||
1697 | return isa<ConstantSDNode>(V) || isa<ConstantFPSDNode>(V); | |||
1698 | } | |||
1699 | ||||
1700 | class GlobalAddressSDNode : public SDNode { | |||
1701 | friend class SelectionDAG; | |||
1702 | ||||
1703 | const GlobalValue *TheGlobal; | |||
1704 | int64_t Offset; | |||
1705 | unsigned TargetFlags; | |||
1706 | ||||
1707 | GlobalAddressSDNode(unsigned Opc, unsigned Order, const DebugLoc &DL, | |||
1708 | const GlobalValue *GA, EVT VT, int64_t o, | |||
1709 | unsigned TF); | |||
1710 | ||||
1711 | public: | |||
1712 | const GlobalValue *getGlobal() const { return TheGlobal; } | |||
1713 | int64_t getOffset() const { return Offset; } | |||
1714 | unsigned getTargetFlags() const { return TargetFlags; } | |||
1715 | // Return the address space this GlobalAddress belongs to. | |||
1716 | unsigned getAddressSpace() const; | |||
1717 | ||||
1718 | static bool classof(const SDNode *N) { | |||
1719 | return N->getOpcode() == ISD::GlobalAddress || | |||
1720 | N->getOpcode() == ISD::TargetGlobalAddress || | |||
1721 | N->getOpcode() == ISD::GlobalTLSAddress || | |||
1722 | N->getOpcode() == ISD::TargetGlobalTLSAddress; | |||
1723 | } | |||
1724 | }; | |||
1725 | ||||
1726 | class FrameIndexSDNode : public SDNode { | |||
1727 | friend class SelectionDAG; | |||
1728 | ||||
1729 | int FI; | |||
1730 | ||||
1731 | FrameIndexSDNode(int fi, EVT VT, bool isTarg) | |||
1732 | : SDNode(isTarg ? ISD::TargetFrameIndex : ISD::FrameIndex, | |||
1733 | 0, DebugLoc(), getSDVTList(VT)), FI(fi) { | |||
1734 | } | |||
1735 | ||||
1736 | public: | |||
1737 | int getIndex() const { return FI; } | |||
1738 | ||||
1739 | static bool classof(const SDNode *N) { | |||
1740 | return N->getOpcode() == ISD::FrameIndex || | |||
1741 | N->getOpcode() == ISD::TargetFrameIndex; | |||
1742 | } | |||
1743 | }; | |||
1744 | ||||
1745 | /// This SDNode is used for LIFETIME_START/LIFETIME_END values, which indicate | |||
1746 | /// the offet and size that are started/ended in the underlying FrameIndex. | |||
1747 | class LifetimeSDNode : public SDNode { | |||
1748 | friend class SelectionDAG; | |||
1749 | int64_t Size; | |||
1750 | int64_t Offset; // -1 if offset is unknown. | |||
1751 | ||||
1752 | LifetimeSDNode(unsigned Opcode, unsigned Order, const DebugLoc &dl, | |||
1753 | SDVTList VTs, int64_t Size, int64_t Offset) | |||
1754 | : SDNode(Opcode, Order, dl, VTs), Size(Size), Offset(Offset) {} | |||
1755 | public: | |||
1756 | int64_t getFrameIndex() const { | |||
1757 | return cast<FrameIndexSDNode>(getOperand(1))->getIndex(); | |||
1758 | } | |||
1759 | ||||
1760 | bool hasOffset() const { return Offset >= 0; } | |||
1761 | int64_t getOffset() const { | |||
1762 | assert(hasOffset() && "offset is unknown")(static_cast <bool> (hasOffset() && "offset is unknown" ) ? void (0) : __assert_fail ("hasOffset() && \"offset is unknown\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1762, __extension__ __PRETTY_FUNCTION__)); | |||
1763 | return Offset; | |||
1764 | } | |||
1765 | int64_t getSize() const { | |||
1766 | assert(hasOffset() && "offset is unknown")(static_cast <bool> (hasOffset() && "offset is unknown" ) ? void (0) : __assert_fail ("hasOffset() && \"offset is unknown\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1766, __extension__ __PRETTY_FUNCTION__)); | |||
1767 | return Size; | |||
1768 | } | |||
1769 | ||||
1770 | // Methods to support isa and dyn_cast | |||
1771 | static bool classof(const SDNode *N) { | |||
1772 | return N->getOpcode() == ISD::LIFETIME_START || | |||
1773 | N->getOpcode() == ISD::LIFETIME_END; | |||
1774 | } | |||
1775 | }; | |||
1776 | ||||
1777 | /// This SDNode is used for PSEUDO_PROBE values, which are the function guid and | |||
1778 | /// the index of the basic block being probed. A pseudo probe serves as a place | |||
1779 | /// holder and will be removed at the end of compilation. It does not have any | |||
1780 | /// operand because we do not want the instruction selection to deal with any. | |||
1781 | class PseudoProbeSDNode : public SDNode { | |||
1782 | friend class SelectionDAG; | |||
1783 | uint64_t Guid; | |||
1784 | uint64_t Index; | |||
1785 | uint32_t Attributes; | |||
1786 | ||||
1787 | PseudoProbeSDNode(unsigned Opcode, unsigned Order, const DebugLoc &Dl, | |||
1788 | SDVTList VTs, uint64_t Guid, uint64_t Index, uint32_t Attr) | |||
1789 | : SDNode(Opcode, Order, Dl, VTs), Guid(Guid), Index(Index), | |||
1790 | Attributes(Attr) {} | |||
1791 | ||||
1792 | public: | |||
1793 | uint64_t getGuid() const { return Guid; } | |||
1794 | uint64_t getIndex() const { return Index; } | |||
1795 | uint32_t getAttributes() const { return Attributes; } | |||
1796 | ||||
1797 | // Methods to support isa and dyn_cast | |||
1798 | static bool classof(const SDNode *N) { | |||
1799 | return N->getOpcode() == ISD::PSEUDO_PROBE; | |||
1800 | } | |||
1801 | }; | |||
1802 | ||||
1803 | class JumpTableSDNode : public SDNode { | |||
1804 | friend class SelectionDAG; | |||
1805 | ||||
1806 | int JTI; | |||
1807 | unsigned TargetFlags; | |||
1808 | ||||
1809 | JumpTableSDNode(int jti, EVT VT, bool isTarg, unsigned TF) | |||
1810 | : SDNode(isTarg ? ISD::TargetJumpTable : ISD::JumpTable, | |||
1811 | 0, DebugLoc(), getSDVTList(VT)), JTI(jti), TargetFlags(TF) { | |||
1812 | } | |||
1813 | ||||
1814 | public: | |||
1815 | int getIndex() const { return JTI; } | |||
1816 | unsigned getTargetFlags() const { return TargetFlags; } | |||
1817 | ||||
1818 | static bool classof(const SDNode *N) { | |||
1819 | return N->getOpcode() == ISD::JumpTable || | |||
1820 | N->getOpcode() == ISD::TargetJumpTable; | |||
1821 | } | |||
1822 | }; | |||
1823 | ||||
1824 | class ConstantPoolSDNode : public SDNode { | |||
1825 | friend class SelectionDAG; | |||
1826 | ||||
1827 | union { | |||
1828 | const Constant *ConstVal; | |||
1829 | MachineConstantPoolValue *MachineCPVal; | |||
1830 | } Val; | |||
1831 | int Offset; // It's a MachineConstantPoolValue if top bit is set. | |||
1832 | Align Alignment; // Minimum alignment requirement of CP. | |||
1833 | unsigned TargetFlags; | |||
1834 | ||||
1835 | ConstantPoolSDNode(bool isTarget, const Constant *c, EVT VT, int o, | |||
1836 | Align Alignment, unsigned TF) | |||
1837 | : SDNode(isTarget ? ISD::TargetConstantPool : ISD::ConstantPool, 0, | |||
1838 | DebugLoc(), getSDVTList(VT)), | |||
1839 | Offset(o), Alignment(Alignment), TargetFlags(TF) { | |||
1840 | assert(Offset >= 0 && "Offset is too large")(static_cast <bool> (Offset >= 0 && "Offset is too large" ) ? void (0) : __assert_fail ("Offset >= 0 && \"Offset is too large\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1840, __extension__ __PRETTY_FUNCTION__)); | |||
1841 | Val.ConstVal = c; | |||
1842 | } | |||
1843 | ||||
1844 | ConstantPoolSDNode(bool isTarget, MachineConstantPoolValue *v, EVT VT, int o, | |||
1845 | Align Alignment, unsigned TF) | |||
1846 | : SDNode(isTarget ? ISD::TargetConstantPool : ISD::ConstantPool, 0, | |||
1847 | DebugLoc(), getSDVTList(VT)), | |||
1848 | Offset(o), Alignment(Alignment), TargetFlags(TF) { | |||
1849 | assert(Offset >= 0 && "Offset is too large")(static_cast <bool> (Offset >= 0 && "Offset is too large" ) ? void (0) : __assert_fail ("Offset >= 0 && \"Offset is too large\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1849, __extension__ __PRETTY_FUNCTION__)); | |||
1850 | Val.MachineCPVal = v; | |||
1851 | Offset |= 1 << (sizeof(unsigned)*CHAR_BIT8-1); | |||
1852 | } | |||
1853 | ||||
1854 | public: | |||
1855 | bool isMachineConstantPoolEntry() const { | |||
1856 | return Offset < 0; | |||
1857 | } | |||
1858 | ||||
1859 | const Constant *getConstVal() const { | |||
1860 | assert(!isMachineConstantPoolEntry() && "Wrong constantpool type")(static_cast <bool> (!isMachineConstantPoolEntry() && "Wrong constantpool type") ? void (0) : __assert_fail ("!isMachineConstantPoolEntry() && \"Wrong constantpool type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1860, __extension__ __PRETTY_FUNCTION__)); | |||
1861 | return Val.ConstVal; | |||
1862 | } | |||
1863 | ||||
1864 | MachineConstantPoolValue *getMachineCPVal() const { | |||
1865 | assert(isMachineConstantPoolEntry() && "Wrong constantpool type")(static_cast <bool> (isMachineConstantPoolEntry() && "Wrong constantpool type") ? void (0) : __assert_fail ("isMachineConstantPoolEntry() && \"Wrong constantpool type\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 1865, __extension__ __PRETTY_FUNCTION__)); | |||
1866 | return Val.MachineCPVal; | |||
1867 | } | |||
1868 | ||||
1869 | int getOffset() const { | |||
1870 | return Offset & ~(1 << (sizeof(unsigned)*CHAR_BIT8-1)); | |||
1871 | } | |||
1872 | ||||
1873 | // Return the alignment of this constant pool object, which is either 0 (for | |||
1874 | // default alignment) or the desired value. | |||
1875 | Align getAlign() const { return Alignment; } | |||
1876 | unsigned getTargetFlags() const { return TargetFlags; } | |||
1877 | ||||
1878 | Type *getType() const; | |||
1879 | ||||
1880 | static bool classof(const SDNode *N) { | |||
1881 | return N->getOpcode() == ISD::ConstantPool || | |||
1882 | N->getOpcode() == ISD::TargetConstantPool; | |||
1883 | } | |||
1884 | }; | |||
1885 | ||||
1886 | /// Completely target-dependent object reference. | |||
1887 | class TargetIndexSDNode : public SDNode { | |||
1888 | friend class SelectionDAG; | |||
1889 | ||||
1890 | unsigned TargetFlags; | |||
1891 | int Index; | |||
1892 | int64_t Offset; | |||
1893 | ||||
1894 | public: | |||
1895 | TargetIndexSDNode(int Idx, EVT VT, int64_t Ofs, unsigned TF) | |||
1896 | : SDNode(ISD::TargetIndex, 0, DebugLoc(), getSDVTList(VT)), | |||
1897 | TargetFlags(TF), Index(Idx), Offset(Ofs) {} | |||
1898 | ||||
1899 | unsigned getTargetFlags() const { return TargetFlags; } | |||
1900 | int getIndex() const { return Index; } | |||
1901 | int64_t getOffset() const { return Offset; } | |||
1902 | ||||
1903 | static bool classof(const SDNode *N) { | |||
1904 | return N->getOpcode() == ISD::TargetIndex; | |||
1905 | } | |||
1906 | }; | |||
1907 | ||||
1908 | class BasicBlockSDNode : public SDNode { | |||
1909 | friend class SelectionDAG; | |||
1910 | ||||
1911 | MachineBasicBlock *MBB; | |||
1912 | ||||
1913 | /// Debug info is meaningful and potentially useful here, but we create | |||
1914 | /// blocks out of order when they're jumped to, which makes it a bit | |||
1915 | /// harder. Let's see if we need it first. | |||
1916 | explicit BasicBlockSDNode(MachineBasicBlock *mbb) | |||
1917 | : SDNode(ISD::BasicBlock, 0, DebugLoc(), getSDVTList(MVT::Other)), MBB(mbb) | |||
1918 | {} | |||
1919 | ||||
1920 | public: | |||
1921 | MachineBasicBlock *getBasicBlock() const { return MBB; } | |||
1922 | ||||
1923 | static bool classof(const SDNode *N) { | |||
1924 | return N->getOpcode() == ISD::BasicBlock; | |||
1925 | } | |||
1926 | }; | |||
1927 | ||||
1928 | /// A "pseudo-class" with methods for operating on BUILD_VECTORs. | |||
1929 | class BuildVectorSDNode : public SDNode { | |||
1930 | public: | |||
1931 | // These are constructed as SDNodes and then cast to BuildVectorSDNodes. | |||
1932 | explicit BuildVectorSDNode() = delete; | |||
1933 | ||||
1934 | /// Check if this is a constant splat, and if so, find the | |||
1935 | /// smallest element size that splats the vector. If MinSplatBits is | |||
1936 | /// nonzero, the element size must be at least that large. Note that the | |||
1937 | /// splat element may be the entire vector (i.e., a one element vector). | |||
1938 | /// Returns the splat element value in SplatValue. Any undefined bits in | |||
1939 | /// that value are zero, and the corresponding bits in the SplatUndef mask | |||
1940 | /// are set. The SplatBitSize value is set to the splat element size in | |||
1941 | /// bits. HasAnyUndefs is set to true if any bits in the vector are | |||
1942 | /// undefined. isBigEndian describes the endianness of the target. | |||
1943 | bool isConstantSplat(APInt &SplatValue, APInt &SplatUndef, | |||
1944 | unsigned &SplatBitSize, bool &HasAnyUndefs, | |||
1945 | unsigned MinSplatBits = 0, | |||
1946 | bool isBigEndian = false) const; | |||
1947 | ||||
1948 | /// Returns the demanded splatted value or a null value if this is not a | |||
1949 | /// splat. | |||
1950 | /// | |||
1951 | /// The DemandedElts mask indicates the elements that must be in the splat. | |||
1952 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
1953 | /// the vector width and set the bits where elements are undef. | |||
1954 | SDValue getSplatValue(const APInt &DemandedElts, | |||
1955 | BitVector *UndefElements = nullptr) const; | |||
1956 | ||||
1957 | /// Returns the splatted value or a null value if this is not a splat. | |||
1958 | /// | |||
1959 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
1960 | /// the vector width and set the bits where elements are undef. | |||
1961 | SDValue getSplatValue(BitVector *UndefElements = nullptr) const; | |||
1962 | ||||
1963 | /// Find the shortest repeating sequence of values in the build vector. | |||
1964 | /// | |||
1965 | /// e.g. { u, X, u, X, u, u, X, u } -> { X } | |||
1966 | /// { X, Y, u, Y, u, u, X, u } -> { X, Y } | |||
1967 | /// | |||
1968 | /// Currently this must be a power-of-2 build vector. | |||
1969 | /// The DemandedElts mask indicates the elements that must be present, | |||
1970 | /// undemanded elements in Sequence may be null (SDValue()). If passed a | |||
1971 | /// non-null UndefElements bitvector, it will resize it to match the original | |||
1972 | /// vector width and set the bits where elements are undef. If result is | |||
1973 | /// false, Sequence will be empty. | |||
1974 | bool getRepeatedSequence(const APInt &DemandedElts, | |||
1975 | SmallVectorImpl<SDValue> &Sequence, | |||
1976 | BitVector *UndefElements = nullptr) const; | |||
1977 | ||||
1978 | /// Find the shortest repeating sequence of values in the build vector. | |||
1979 | /// | |||
1980 | /// e.g. { u, X, u, X, u, u, X, u } -> { X } | |||
1981 | /// { X, Y, u, Y, u, u, X, u } -> { X, Y } | |||
1982 | /// | |||
1983 | /// Currently this must be a power-of-2 build vector. | |||
1984 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
1985 | /// the original vector width and set the bits where elements are undef. | |||
1986 | /// If result is false, Sequence will be empty. | |||
1987 | bool getRepeatedSequence(SmallVectorImpl<SDValue> &Sequence, | |||
1988 | BitVector *UndefElements = nullptr) const; | |||
1989 | ||||
1990 | /// Returns the demanded splatted constant or null if this is not a constant | |||
1991 | /// splat. | |||
1992 | /// | |||
1993 | /// The DemandedElts mask indicates the elements that must be in the splat. | |||
1994 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
1995 | /// the vector width and set the bits where elements are undef. | |||
1996 | ConstantSDNode * | |||
1997 | getConstantSplatNode(const APInt &DemandedElts, | |||
1998 | BitVector *UndefElements = nullptr) const; | |||
1999 | ||||
2000 | /// Returns the splatted constant or null if this is not a constant | |||
2001 | /// splat. | |||
2002 | /// | |||
2003 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
2004 | /// the vector width and set the bits where elements are undef. | |||
2005 | ConstantSDNode * | |||
2006 | getConstantSplatNode(BitVector *UndefElements = nullptr) const; | |||
2007 | ||||
2008 | /// Returns the demanded splatted constant FP or null if this is not a | |||
2009 | /// constant FP splat. | |||
2010 | /// | |||
2011 | /// The DemandedElts mask indicates the elements that must be in the splat. | |||
2012 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
2013 | /// the vector width and set the bits where elements are undef. | |||
2014 | ConstantFPSDNode * | |||
2015 | getConstantFPSplatNode(const APInt &DemandedElts, | |||
2016 | BitVector *UndefElements = nullptr) const; | |||
2017 | ||||
2018 | /// Returns the splatted constant FP or null if this is not a constant | |||
2019 | /// FP splat. | |||
2020 | /// | |||
2021 | /// If passed a non-null UndefElements bitvector, it will resize it to match | |||
2022 | /// the vector width and set the bits where elements are undef. | |||
2023 | ConstantFPSDNode * | |||
2024 | getConstantFPSplatNode(BitVector *UndefElements = nullptr) const; | |||
2025 | ||||
2026 | /// If this is a constant FP splat and the splatted constant FP is an | |||
2027 | /// exact power or 2, return the log base 2 integer value. Otherwise, | |||
2028 | /// return -1. | |||
2029 | /// | |||
2030 | /// The BitWidth specifies the necessary bit precision. | |||
2031 | int32_t getConstantFPSplatPow2ToLog2Int(BitVector *UndefElements, | |||
2032 | uint32_t BitWidth) const; | |||
2033 | ||||
2034 | bool isConstant() const; | |||
2035 | ||||
2036 | static bool classof(const SDNode *N) { | |||
2037 | return N->getOpcode() == ISD::BUILD_VECTOR; | |||
2038 | } | |||
2039 | }; | |||
2040 | ||||
2041 | /// An SDNode that holds an arbitrary LLVM IR Value. This is | |||
2042 | /// used when the SelectionDAG needs to make a simple reference to something | |||
2043 | /// in the LLVM IR representation. | |||
2044 | /// | |||
2045 | class SrcValueSDNode : public SDNode { | |||
2046 | friend class SelectionDAG; | |||
2047 | ||||
2048 | const Value *V; | |||
2049 | ||||
2050 | /// Create a SrcValue for a general value. | |||
2051 | explicit SrcValueSDNode(const Value *v) | |||
2052 | : SDNode(ISD::SRCVALUE, 0, DebugLoc(), getSDVTList(MVT::Other)), V(v) {} | |||
2053 | ||||
2054 | public: | |||
2055 | /// Return the contained Value. | |||
2056 | const Value *getValue() const { return V; } | |||
2057 | ||||
2058 | static bool classof(const SDNode *N) { | |||
2059 | return N->getOpcode() == ISD::SRCVALUE; | |||
2060 | } | |||
2061 | }; | |||
2062 | ||||
2063 | class MDNodeSDNode : public SDNode { | |||
2064 | friend class SelectionDAG; | |||
2065 | ||||
2066 | const MDNode *MD; | |||
2067 | ||||
2068 | explicit MDNodeSDNode(const MDNode *md) | |||
2069 | : SDNode(ISD::MDNODE_SDNODE, 0, DebugLoc(), getSDVTList(MVT::Other)), MD(md) | |||
2070 | {} | |||
2071 | ||||
2072 | public: | |||
2073 | const MDNode *getMD() const { return MD; } | |||
2074 | ||||
2075 | static bool classof(const SDNode *N) { | |||
2076 | return N->getOpcode() == ISD::MDNODE_SDNODE; | |||
2077 | } | |||
2078 | }; | |||
2079 | ||||
2080 | class RegisterSDNode : public SDNode { | |||
2081 | friend class SelectionDAG; | |||
2082 | ||||
2083 | Register Reg; | |||
2084 | ||||
2085 | RegisterSDNode(Register reg, EVT VT) | |||
2086 | : SDNode(ISD::Register, 0, DebugLoc(), getSDVTList(VT)), Reg(reg) {} | |||
2087 | ||||
2088 | public: | |||
2089 | Register getReg() const { return Reg; } | |||
2090 | ||||
2091 | static bool classof(const SDNode *N) { | |||
2092 | return N->getOpcode() == ISD::Register; | |||
2093 | } | |||
2094 | }; | |||
2095 | ||||
2096 | class RegisterMaskSDNode : public SDNode { | |||
2097 | friend class SelectionDAG; | |||
2098 | ||||
2099 | // The memory for RegMask is not owned by the node. | |||
2100 | const uint32_t *RegMask; | |||
2101 | ||||
2102 | RegisterMaskSDNode(const uint32_t *mask) | |||
2103 | : SDNode(ISD::RegisterMask, 0, DebugLoc(), getSDVTList(MVT::Untyped)), | |||
2104 | RegMask(mask) {} | |||
2105 | ||||
2106 | public: | |||
2107 | const uint32_t *getRegMask() const { return RegMask; } | |||
2108 | ||||
2109 | static bool classof(const SDNode *N) { | |||
2110 | return N->getOpcode() == ISD::RegisterMask; | |||
2111 | } | |||
2112 | }; | |||
2113 | ||||
2114 | class BlockAddressSDNode : public SDNode { | |||
2115 | friend class SelectionDAG; | |||
2116 | ||||
2117 | const BlockAddress *BA; | |||
2118 | int64_t Offset; | |||
2119 | unsigned TargetFlags; | |||
2120 | ||||
2121 | BlockAddressSDNode(unsigned NodeTy, EVT VT, const BlockAddress *ba, | |||
2122 | int64_t o, unsigned Flags) | |||
2123 | : SDNode(NodeTy, 0, DebugLoc(), getSDVTList(VT)), | |||
2124 | BA(ba), Offset(o), TargetFlags(Flags) {} | |||
2125 | ||||
2126 | public: | |||
2127 | const BlockAddress *getBlockAddress() const { return BA; } | |||
2128 | int64_t getOffset() const { return Offset; } | |||
2129 | unsigned getTargetFlags() const { return TargetFlags; } | |||
2130 | ||||
2131 | static bool classof(const SDNode *N) { | |||
2132 | return N->getOpcode() == ISD::BlockAddress || | |||
2133 | N->getOpcode() == ISD::TargetBlockAddress; | |||
2134 | } | |||
2135 | }; | |||
2136 | ||||
2137 | class LabelSDNode : public SDNode { | |||
2138 | friend class SelectionDAG; | |||
2139 | ||||
2140 | MCSymbol *Label; | |||
2141 | ||||
2142 | LabelSDNode(unsigned Opcode, unsigned Order, const DebugLoc &dl, MCSymbol *L) | |||
2143 | : SDNode(Opcode, Order, dl, getSDVTList(MVT::Other)), Label(L) { | |||
2144 | assert(LabelSDNode::classof(this) && "not a label opcode")(static_cast <bool> (LabelSDNode::classof(this) && "not a label opcode") ? void (0) : __assert_fail ("LabelSDNode::classof(this) && \"not a label opcode\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2144, __extension__ __PRETTY_FUNCTION__)); | |||
2145 | } | |||
2146 | ||||
2147 | public: | |||
2148 | MCSymbol *getLabel() const { return Label; } | |||
2149 | ||||
2150 | static bool classof(const SDNode *N) { | |||
2151 | return N->getOpcode() == ISD::EH_LABEL || | |||
2152 | N->getOpcode() == ISD::ANNOTATION_LABEL; | |||
2153 | } | |||
2154 | }; | |||
2155 | ||||
2156 | class ExternalSymbolSDNode : public SDNode { | |||
2157 | friend class SelectionDAG; | |||
2158 | ||||
2159 | const char *Symbol; | |||
2160 | unsigned TargetFlags; | |||
2161 | ||||
2162 | ExternalSymbolSDNode(bool isTarget, const char *Sym, unsigned TF, EVT VT) | |||
2163 | : SDNode(isTarget ? ISD::TargetExternalSymbol : ISD::ExternalSymbol, 0, | |||
2164 | DebugLoc(), getSDVTList(VT)), | |||
2165 | Symbol(Sym), TargetFlags(TF) {} | |||
2166 | ||||
2167 | public: | |||
2168 | const char *getSymbol() const { return Symbol; } | |||
2169 | unsigned getTargetFlags() const { return TargetFlags; } | |||
2170 | ||||
2171 | static bool classof(const SDNode *N) { | |||
2172 | return N->getOpcode() == ISD::ExternalSymbol || | |||
2173 | N->getOpcode() == ISD::TargetExternalSymbol; | |||
2174 | } | |||
2175 | }; | |||
2176 | ||||
2177 | class MCSymbolSDNode : public SDNode { | |||
2178 | friend class SelectionDAG; | |||
2179 | ||||
2180 | MCSymbol *Symbol; | |||
2181 | ||||
2182 | MCSymbolSDNode(MCSymbol *Symbol, EVT VT) | |||
2183 | : SDNode(ISD::MCSymbol, 0, DebugLoc(), getSDVTList(VT)), Symbol(Symbol) {} | |||
2184 | ||||
2185 | public: | |||
2186 | MCSymbol *getMCSymbol() const { return Symbol; } | |||
2187 | ||||
2188 | static bool classof(const SDNode *N) { | |||
2189 | return N->getOpcode() == ISD::MCSymbol; | |||
2190 | } | |||
2191 | }; | |||
2192 | ||||
2193 | class CondCodeSDNode : public SDNode { | |||
2194 | friend class SelectionDAG; | |||
2195 | ||||
2196 | ISD::CondCode Condition; | |||
2197 | ||||
2198 | explicit CondCodeSDNode(ISD::CondCode Cond) | |||
2199 | : SDNode(ISD::CONDCODE, 0, DebugLoc(), getSDVTList(MVT::Other)), | |||
2200 | Condition(Cond) {} | |||
2201 | ||||
2202 | public: | |||
2203 | ISD::CondCode get() const { return Condition; } | |||
2204 | ||||
2205 | static bool classof(const SDNode *N) { | |||
2206 | return N->getOpcode() == ISD::CONDCODE; | |||
2207 | } | |||
2208 | }; | |||
2209 | ||||
2210 | /// This class is used to represent EVT's, which are used | |||
2211 | /// to parameterize some operations. | |||
2212 | class VTSDNode : public SDNode { | |||
2213 | friend class SelectionDAG; | |||
2214 | ||||
2215 | EVT ValueType; | |||
2216 | ||||
2217 | explicit VTSDNode(EVT VT) | |||
2218 | : SDNode(ISD::VALUETYPE, 0, DebugLoc(), getSDVTList(MVT::Other)), | |||
2219 | ValueType(VT) {} | |||
2220 | ||||
2221 | public: | |||
2222 | EVT getVT() const { return ValueType; } | |||
2223 | ||||
2224 | static bool classof(const SDNode *N) { | |||
2225 | return N->getOpcode() == ISD::VALUETYPE; | |||
2226 | } | |||
2227 | }; | |||
2228 | ||||
2229 | /// Base class for LoadSDNode and StoreSDNode | |||
2230 | class LSBaseSDNode : public MemSDNode { | |||
2231 | public: | |||
2232 | LSBaseSDNode(ISD::NodeType NodeTy, unsigned Order, const DebugLoc &dl, | |||
2233 | SDVTList VTs, ISD::MemIndexedMode AM, EVT MemVT, | |||
2234 | MachineMemOperand *MMO) | |||
2235 | : MemSDNode(NodeTy, Order, dl, VTs, MemVT, MMO) { | |||
2236 | LSBaseSDNodeBits.AddressingMode = AM; | |||
2237 | assert(getAddressingMode() == AM && "Value truncated")(static_cast <bool> (getAddressingMode() == AM && "Value truncated") ? void (0) : __assert_fail ("getAddressingMode() == AM && \"Value truncated\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2237, __extension__ __PRETTY_FUNCTION__)); | |||
2238 | } | |||
2239 | ||||
2240 | const SDValue &getOffset() const { | |||
2241 | return getOperand(getOpcode() == ISD::LOAD ? 2 : 3); | |||
2242 | } | |||
2243 | ||||
2244 | /// Return the addressing mode for this load or store: | |||
2245 | /// unindexed, pre-inc, pre-dec, post-inc, or post-dec. | |||
2246 | ISD::MemIndexedMode getAddressingMode() const { | |||
2247 | return static_cast<ISD::MemIndexedMode>(LSBaseSDNodeBits.AddressingMode); | |||
2248 | } | |||
2249 | ||||
2250 | /// Return true if this is a pre/post inc/dec load/store. | |||
2251 | bool isIndexed() const { return getAddressingMode() != ISD::UNINDEXED; } | |||
2252 | ||||
2253 | /// Return true if this is NOT a pre/post inc/dec load/store. | |||
2254 | bool isUnindexed() const { return getAddressingMode() == ISD::UNINDEXED; } | |||
2255 | ||||
2256 | static bool classof(const SDNode *N) { | |||
2257 | return N->getOpcode() == ISD::LOAD || | |||
2258 | N->getOpcode() == ISD::STORE; | |||
2259 | } | |||
2260 | }; | |||
2261 | ||||
2262 | /// This class is used to represent ISD::LOAD nodes. | |||
2263 | class LoadSDNode : public LSBaseSDNode { | |||
2264 | friend class SelectionDAG; | |||
2265 | ||||
2266 | LoadSDNode(unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
2267 | ISD::MemIndexedMode AM, ISD::LoadExtType ETy, EVT MemVT, | |||
2268 | MachineMemOperand *MMO) | |||
2269 | : LSBaseSDNode(ISD::LOAD, Order, dl, VTs, AM, MemVT, MMO) { | |||
2270 | LoadSDNodeBits.ExtTy = ETy; | |||
2271 | assert(readMem() && "Load MachineMemOperand is not a load!")(static_cast <bool> (readMem() && "Load MachineMemOperand is not a load!" ) ? void (0) : __assert_fail ("readMem() && \"Load MachineMemOperand is not a load!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2271, __extension__ __PRETTY_FUNCTION__)); | |||
2272 | assert(!writeMem() && "Load MachineMemOperand is a store!")(static_cast <bool> (!writeMem() && "Load MachineMemOperand is a store!" ) ? void (0) : __assert_fail ("!writeMem() && \"Load MachineMemOperand is a store!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2272, __extension__ __PRETTY_FUNCTION__)); | |||
2273 | } | |||
2274 | ||||
2275 | public: | |||
2276 | /// Return whether this is a plain node, | |||
2277 | /// or one of the varieties of value-extending loads. | |||
2278 | ISD::LoadExtType getExtensionType() const { | |||
2279 | return static_cast<ISD::LoadExtType>(LoadSDNodeBits.ExtTy); | |||
2280 | } | |||
2281 | ||||
2282 | const SDValue &getBasePtr() const { return getOperand(1); } | |||
2283 | const SDValue &getOffset() const { return getOperand(2); } | |||
2284 | ||||
2285 | static bool classof(const SDNode *N) { | |||
2286 | return N->getOpcode() == ISD::LOAD; | |||
2287 | } | |||
2288 | }; | |||
2289 | ||||
2290 | /// This class is used to represent ISD::STORE nodes. | |||
2291 | class StoreSDNode : public LSBaseSDNode { | |||
2292 | friend class SelectionDAG; | |||
2293 | ||||
2294 | StoreSDNode(unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
2295 | ISD::MemIndexedMode AM, bool isTrunc, EVT MemVT, | |||
2296 | MachineMemOperand *MMO) | |||
2297 | : LSBaseSDNode(ISD::STORE, Order, dl, VTs, AM, MemVT, MMO) { | |||
2298 | StoreSDNodeBits.IsTruncating = isTrunc; | |||
2299 | assert(!readMem() && "Store MachineMemOperand is a load!")(static_cast <bool> (!readMem() && "Store MachineMemOperand is a load!" ) ? void (0) : __assert_fail ("!readMem() && \"Store MachineMemOperand is a load!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2299, __extension__ __PRETTY_FUNCTION__)); | |||
2300 | assert(writeMem() && "Store MachineMemOperand is not a store!")(static_cast <bool> (writeMem() && "Store MachineMemOperand is not a store!" ) ? void (0) : __assert_fail ("writeMem() && \"Store MachineMemOperand is not a store!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2300, __extension__ __PRETTY_FUNCTION__)); | |||
2301 | } | |||
2302 | ||||
2303 | public: | |||
2304 | /// Return true if the op does a truncation before store. | |||
2305 | /// For integers this is the same as doing a TRUNCATE and storing the result. | |||
2306 | /// For floats, it is the same as doing an FP_ROUND and storing the result. | |||
2307 | bool isTruncatingStore() const { return StoreSDNodeBits.IsTruncating; } | |||
2308 | void setTruncatingStore(bool Truncating) { | |||
2309 | StoreSDNodeBits.IsTruncating = Truncating; | |||
2310 | } | |||
2311 | ||||
2312 | const SDValue &getValue() const { return getOperand(1); } | |||
2313 | const SDValue &getBasePtr() const { return getOperand(2); } | |||
2314 | const SDValue &getOffset() const { return getOperand(3); } | |||
2315 | ||||
2316 | static bool classof(const SDNode *N) { | |||
2317 | return N->getOpcode() == ISD::STORE; | |||
2318 | } | |||
2319 | }; | |||
2320 | ||||
2321 | /// This base class is used to represent MLOAD and MSTORE nodes | |||
2322 | class MaskedLoadStoreSDNode : public MemSDNode { | |||
2323 | public: | |||
2324 | friend class SelectionDAG; | |||
2325 | ||||
2326 | MaskedLoadStoreSDNode(ISD::NodeType NodeTy, unsigned Order, | |||
2327 | const DebugLoc &dl, SDVTList VTs, | |||
2328 | ISD::MemIndexedMode AM, EVT MemVT, | |||
2329 | MachineMemOperand *MMO) | |||
2330 | : MemSDNode(NodeTy, Order, dl, VTs, MemVT, MMO) { | |||
2331 | LSBaseSDNodeBits.AddressingMode = AM; | |||
2332 | assert(getAddressingMode() == AM && "Value truncated")(static_cast <bool> (getAddressingMode() == AM && "Value truncated") ? void (0) : __assert_fail ("getAddressingMode() == AM && \"Value truncated\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2332, __extension__ __PRETTY_FUNCTION__)); | |||
2333 | } | |||
2334 | ||||
2335 | // MaskedLoadSDNode (Chain, ptr, offset, mask, passthru) | |||
2336 | // MaskedStoreSDNode (Chain, data, ptr, offset, mask) | |||
2337 | // Mask is a vector of i1 elements | |||
2338 | const SDValue &getOffset() const { | |||
2339 | return getOperand(getOpcode() == ISD::MLOAD ? 2 : 3); | |||
2340 | } | |||
2341 | const SDValue &getMask() const { | |||
2342 | return getOperand(getOpcode() == ISD::MLOAD ? 3 : 4); | |||
2343 | } | |||
2344 | ||||
2345 | /// Return the addressing mode for this load or store: | |||
2346 | /// unindexed, pre-inc, pre-dec, post-inc, or post-dec. | |||
2347 | ISD::MemIndexedMode getAddressingMode() const { | |||
2348 | return static_cast<ISD::MemIndexedMode>(LSBaseSDNodeBits.AddressingMode); | |||
2349 | } | |||
2350 | ||||
2351 | /// Return true if this is a pre/post inc/dec load/store. | |||
2352 | bool isIndexed() const { return getAddressingMode() != ISD::UNINDEXED; } | |||
2353 | ||||
2354 | /// Return true if this is NOT a pre/post inc/dec load/store. | |||
2355 | bool isUnindexed() const { return getAddressingMode() == ISD::UNINDEXED; } | |||
2356 | ||||
2357 | static bool classof(const SDNode *N) { | |||
2358 | return N->getOpcode() == ISD::MLOAD || | |||
2359 | N->getOpcode() == ISD::MSTORE; | |||
2360 | } | |||
2361 | }; | |||
2362 | ||||
2363 | /// This class is used to represent an MLOAD node | |||
2364 | class MaskedLoadSDNode : public MaskedLoadStoreSDNode { | |||
2365 | public: | |||
2366 | friend class SelectionDAG; | |||
2367 | ||||
2368 | MaskedLoadSDNode(unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
2369 | ISD::MemIndexedMode AM, ISD::LoadExtType ETy, | |||
2370 | bool IsExpanding, EVT MemVT, MachineMemOperand *MMO) | |||
2371 | : MaskedLoadStoreSDNode(ISD::MLOAD, Order, dl, VTs, AM, MemVT, MMO) { | |||
2372 | LoadSDNodeBits.ExtTy = ETy; | |||
2373 | LoadSDNodeBits.IsExpanding = IsExpanding; | |||
2374 | } | |||
2375 | ||||
2376 | ISD::LoadExtType getExtensionType() const { | |||
2377 | return static_cast<ISD::LoadExtType>(LoadSDNodeBits.ExtTy); | |||
2378 | } | |||
2379 | ||||
2380 | const SDValue &getBasePtr() const { return getOperand(1); } | |||
2381 | const SDValue &getOffset() const { return getOperand(2); } | |||
2382 | const SDValue &getMask() const { return getOperand(3); } | |||
2383 | const SDValue &getPassThru() const { return getOperand(4); } | |||
2384 | ||||
2385 | static bool classof(const SDNode *N) { | |||
2386 | return N->getOpcode() == ISD::MLOAD; | |||
2387 | } | |||
2388 | ||||
2389 | bool isExpandingLoad() const { return LoadSDNodeBits.IsExpanding; } | |||
2390 | }; | |||
2391 | ||||
2392 | /// This class is used to represent an MSTORE node | |||
2393 | class MaskedStoreSDNode : public MaskedLoadStoreSDNode { | |||
2394 | public: | |||
2395 | friend class SelectionDAG; | |||
2396 | ||||
2397 | MaskedStoreSDNode(unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
2398 | ISD::MemIndexedMode AM, bool isTrunc, bool isCompressing, | |||
2399 | EVT MemVT, MachineMemOperand *MMO) | |||
2400 | : MaskedLoadStoreSDNode(ISD::MSTORE, Order, dl, VTs, AM, MemVT, MMO) { | |||
2401 | StoreSDNodeBits.IsTruncating = isTrunc; | |||
2402 | StoreSDNodeBits.IsCompressing = isCompressing; | |||
2403 | } | |||
2404 | ||||
2405 | /// Return true if the op does a truncation before store. | |||
2406 | /// For integers this is the same as doing a TRUNCATE and storing the result. | |||
2407 | /// For floats, it is the same as doing an FP_ROUND and storing the result. | |||
2408 | bool isTruncatingStore() const { return StoreSDNodeBits.IsTruncating; } | |||
2409 | ||||
2410 | /// Returns true if the op does a compression to the vector before storing. | |||
2411 | /// The node contiguously stores the active elements (integers or floats) | |||
2412 | /// in src (those with their respective bit set in writemask k) to unaligned | |||
2413 | /// memory at base_addr. | |||
2414 | bool isCompressingStore() const { return StoreSDNodeBits.IsCompressing; } | |||
2415 | ||||
2416 | const SDValue &getValue() const { return getOperand(1); } | |||
2417 | const SDValue &getBasePtr() const { return getOperand(2); } | |||
2418 | const SDValue &getOffset() const { return getOperand(3); } | |||
2419 | const SDValue &getMask() const { return getOperand(4); } | |||
2420 | ||||
2421 | static bool classof(const SDNode *N) { | |||
2422 | return N->getOpcode() == ISD::MSTORE; | |||
2423 | } | |||
2424 | }; | |||
2425 | ||||
2426 | /// This is a base class used to represent | |||
2427 | /// MGATHER and MSCATTER nodes | |||
2428 | /// | |||
2429 | class MaskedGatherScatterSDNode : public MemSDNode { | |||
2430 | public: | |||
2431 | friend class SelectionDAG; | |||
2432 | ||||
2433 | MaskedGatherScatterSDNode(ISD::NodeType NodeTy, unsigned Order, | |||
2434 | const DebugLoc &dl, SDVTList VTs, EVT MemVT, | |||
2435 | MachineMemOperand *MMO, ISD::MemIndexType IndexType) | |||
2436 | : MemSDNode(NodeTy, Order, dl, VTs, MemVT, MMO) { | |||
2437 | LSBaseSDNodeBits.AddressingMode = IndexType; | |||
2438 | assert(getIndexType() == IndexType && "Value truncated")(static_cast <bool> (getIndexType() == IndexType && "Value truncated") ? void (0) : __assert_fail ("getIndexType() == IndexType && \"Value truncated\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2438, __extension__ __PRETTY_FUNCTION__)); | |||
2439 | } | |||
2440 | ||||
2441 | /// How is Index applied to BasePtr when computing addresses. | |||
2442 | ISD::MemIndexType getIndexType() const { | |||
2443 | return static_cast<ISD::MemIndexType>(LSBaseSDNodeBits.AddressingMode); | |||
2444 | } | |||
2445 | void setIndexType(ISD::MemIndexType IndexType) { | |||
2446 | LSBaseSDNodeBits.AddressingMode = IndexType; | |||
2447 | } | |||
2448 | bool isIndexScaled() const { | |||
2449 | return (getIndexType() == ISD::SIGNED_SCALED) || | |||
2450 | (getIndexType() == ISD::UNSIGNED_SCALED); | |||
2451 | } | |||
2452 | bool isIndexSigned() const { | |||
2453 | return (getIndexType() == ISD::SIGNED_SCALED) || | |||
2454 | (getIndexType() == ISD::SIGNED_UNSCALED); | |||
2455 | } | |||
2456 | ||||
2457 | // In the both nodes address is Op1, mask is Op2: | |||
2458 | // MaskedGatherSDNode (Chain, passthru, mask, base, index, scale) | |||
2459 | // MaskedScatterSDNode (Chain, value, mask, base, index, scale) | |||
2460 | // Mask is a vector of i1 elements | |||
2461 | const SDValue &getBasePtr() const { return getOperand(3); } | |||
2462 | const SDValue &getIndex() const { return getOperand(4); } | |||
2463 | const SDValue &getMask() const { return getOperand(2); } | |||
2464 | const SDValue &getScale() const { return getOperand(5); } | |||
2465 | ||||
2466 | static bool classof(const SDNode *N) { | |||
2467 | return N->getOpcode() == ISD::MGATHER || | |||
2468 | N->getOpcode() == ISD::MSCATTER; | |||
2469 | } | |||
2470 | }; | |||
2471 | ||||
2472 | /// This class is used to represent an MGATHER node | |||
2473 | /// | |||
2474 | class MaskedGatherSDNode : public MaskedGatherScatterSDNode { | |||
2475 | public: | |||
2476 | friend class SelectionDAG; | |||
2477 | ||||
2478 | MaskedGatherSDNode(unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
2479 | EVT MemVT, MachineMemOperand *MMO, | |||
2480 | ISD::MemIndexType IndexType, ISD::LoadExtType ETy) | |||
2481 | : MaskedGatherScatterSDNode(ISD::MGATHER, Order, dl, VTs, MemVT, MMO, | |||
2482 | IndexType) { | |||
2483 | LoadSDNodeBits.ExtTy = ETy; | |||
2484 | } | |||
2485 | ||||
2486 | const SDValue &getPassThru() const { return getOperand(1); } | |||
2487 | ||||
2488 | ISD::LoadExtType getExtensionType() const { | |||
2489 | return ISD::LoadExtType(LoadSDNodeBits.ExtTy); | |||
2490 | } | |||
2491 | ||||
2492 | static bool classof(const SDNode *N) { | |||
2493 | return N->getOpcode() == ISD::MGATHER; | |||
2494 | } | |||
2495 | }; | |||
2496 | ||||
2497 | /// This class is used to represent an MSCATTER node | |||
2498 | /// | |||
2499 | class MaskedScatterSDNode : public MaskedGatherScatterSDNode { | |||
2500 | public: | |||
2501 | friend class SelectionDAG; | |||
2502 | ||||
2503 | MaskedScatterSDNode(unsigned Order, const DebugLoc &dl, SDVTList VTs, | |||
2504 | EVT MemVT, MachineMemOperand *MMO, | |||
2505 | ISD::MemIndexType IndexType, bool IsTrunc) | |||
2506 | : MaskedGatherScatterSDNode(ISD::MSCATTER, Order, dl, VTs, MemVT, MMO, | |||
2507 | IndexType) { | |||
2508 | StoreSDNodeBits.IsTruncating = IsTrunc; | |||
2509 | } | |||
2510 | ||||
2511 | /// Return true if the op does a truncation before store. | |||
2512 | /// For integers this is the same as doing a TRUNCATE and storing the result. | |||
2513 | /// For floats, it is the same as doing an FP_ROUND and storing the result. | |||
2514 | bool isTruncatingStore() const { return StoreSDNodeBits.IsTruncating; } | |||
2515 | ||||
2516 | const SDValue &getValue() const { return getOperand(1); } | |||
2517 | ||||
2518 | static bool classof(const SDNode *N) { | |||
2519 | return N->getOpcode() == ISD::MSCATTER; | |||
2520 | } | |||
2521 | }; | |||
2522 | ||||
2523 | /// An SDNode that represents everything that will be needed | |||
2524 | /// to construct a MachineInstr. These nodes are created during the | |||
2525 | /// instruction selection proper phase. | |||
2526 | /// | |||
2527 | /// Note that the only supported way to set the `memoperands` is by calling the | |||
2528 | /// `SelectionDAG::setNodeMemRefs` function as the memory management happens | |||
2529 | /// inside the DAG rather than in the node. | |||
2530 | class MachineSDNode : public SDNode { | |||
2531 | private: | |||
2532 | friend class SelectionDAG; | |||
2533 | ||||
2534 | MachineSDNode(unsigned Opc, unsigned Order, const DebugLoc &DL, SDVTList VTs) | |||
2535 | : SDNode(Opc, Order, DL, VTs) {} | |||
2536 | ||||
2537 | // We use a pointer union between a single `MachineMemOperand` pointer and | |||
2538 | // a pointer to an array of `MachineMemOperand` pointers. This is null when | |||
2539 | // the number of these is zero, the single pointer variant used when the | |||
2540 | // number is one, and the array is used for larger numbers. | |||
2541 | // | |||
2542 | // The array is allocated via the `SelectionDAG`'s allocator and so will | |||
2543 | // always live until the DAG is cleaned up and doesn't require ownership here. | |||
2544 | // | |||
2545 | // We can't use something simpler like `TinyPtrVector` here because `SDNode` | |||
2546 | // subclasses aren't managed in a conforming C++ manner. See the comments on | |||
2547 | // `SelectionDAG::MorphNodeTo` which details what all goes on, but the | |||
2548 | // constraint here is that these don't manage memory with their constructor or | |||
2549 | // destructor and can be initialized to a good state even if they start off | |||
2550 | // uninitialized. | |||
2551 | PointerUnion<MachineMemOperand *, MachineMemOperand **> MemRefs = {}; | |||
2552 | ||||
2553 | // Note that this could be folded into the above `MemRefs` member if doing so | |||
2554 | // is advantageous at some point. We don't need to store this in most cases. | |||
2555 | // However, at the moment this doesn't appear to make the allocation any | |||
2556 | // smaller and makes the code somewhat simpler to read. | |||
2557 | int NumMemRefs = 0; | |||
2558 | ||||
2559 | public: | |||
2560 | using mmo_iterator = ArrayRef<MachineMemOperand *>::const_iterator; | |||
2561 | ||||
2562 | ArrayRef<MachineMemOperand *> memoperands() const { | |||
2563 | // Special case the common cases. | |||
2564 | if (NumMemRefs == 0) | |||
2565 | return {}; | |||
2566 | if (NumMemRefs == 1) | |||
2567 | return makeArrayRef(MemRefs.getAddrOfPtr1(), 1); | |||
2568 | ||||
2569 | // Otherwise we have an actual array. | |||
2570 | return makeArrayRef(MemRefs.get<MachineMemOperand **>(), NumMemRefs); | |||
2571 | } | |||
2572 | mmo_iterator memoperands_begin() const { return memoperands().begin(); } | |||
2573 | mmo_iterator memoperands_end() const { return memoperands().end(); } | |||
2574 | bool memoperands_empty() const { return memoperands().empty(); } | |||
2575 | ||||
2576 | /// Clear out the memory reference descriptor list. | |||
2577 | void clearMemRefs() { | |||
2578 | MemRefs = nullptr; | |||
2579 | NumMemRefs = 0; | |||
2580 | } | |||
2581 | ||||
2582 | static bool classof(const SDNode *N) { | |||
2583 | return N->isMachineOpcode(); | |||
2584 | } | |||
2585 | }; | |||
2586 | ||||
2587 | /// An SDNode that records if a register contains a value that is guaranteed to | |||
2588 | /// be aligned accordingly. | |||
2589 | class AssertAlignSDNode : public SDNode { | |||
2590 | Align Alignment; | |||
2591 | ||||
2592 | public: | |||
2593 | AssertAlignSDNode(unsigned Order, const DebugLoc &DL, EVT VT, Align A) | |||
2594 | : SDNode(ISD::AssertAlign, Order, DL, getSDVTList(VT)), Alignment(A) {} | |||
2595 | ||||
2596 | Align getAlign() const { return Alignment; } | |||
2597 | ||||
2598 | static bool classof(const SDNode *N) { | |||
2599 | return N->getOpcode() == ISD::AssertAlign; | |||
2600 | } | |||
2601 | }; | |||
2602 | ||||
2603 | class SDNodeIterator { | |||
2604 | const SDNode *Node; | |||
2605 | unsigned Operand; | |||
2606 | ||||
2607 | SDNodeIterator(const SDNode *N, unsigned Op) : Node(N), Operand(Op) {} | |||
2608 | ||||
2609 | public: | |||
2610 | using iterator_category = std::forward_iterator_tag; | |||
2611 | using value_type = SDNode; | |||
2612 | using difference_type = std::ptrdiff_t; | |||
2613 | using pointer = value_type *; | |||
2614 | using reference = value_type &; | |||
2615 | ||||
2616 | bool operator==(const SDNodeIterator& x) const { | |||
2617 | return Operand == x.Operand; | |||
2618 | } | |||
2619 | bool operator!=(const SDNodeIterator& x) const { return !operator==(x); } | |||
2620 | ||||
2621 | pointer operator*() const { | |||
2622 | return Node->getOperand(Operand).getNode(); | |||
2623 | } | |||
2624 | pointer operator->() const { return operator*(); } | |||
2625 | ||||
2626 | SDNodeIterator& operator++() { // Preincrement | |||
2627 | ++Operand; | |||
2628 | return *this; | |||
2629 | } | |||
2630 | SDNodeIterator operator++(int) { // Postincrement | |||
2631 | SDNodeIterator tmp = *this; ++*this; return tmp; | |||
2632 | } | |||
2633 | size_t operator-(SDNodeIterator Other) const { | |||
2634 | assert(Node == Other.Node &&(static_cast <bool> (Node == Other.Node && "Cannot compare iterators of two different nodes!" ) ? void (0) : __assert_fail ("Node == Other.Node && \"Cannot compare iterators of two different nodes!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2635, __extension__ __PRETTY_FUNCTION__)) | |||
2635 | "Cannot compare iterators of two different nodes!")(static_cast <bool> (Node == Other.Node && "Cannot compare iterators of two different nodes!" ) ? void (0) : __assert_fail ("Node == Other.Node && \"Cannot compare iterators of two different nodes!\"" , "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/CodeGen/SelectionDAGNodes.h" , 2635, __extension__ __PRETTY_FUNCTION__)); | |||
2636 | return Operand - Other.Operand; | |||
2637 | } | |||
2638 | ||||
2639 | static SDNodeIterator begin(const SDNode *N) { return SDNodeIterator(N, 0); } | |||
2640 | static SDNodeIterator end (const SDNode *N) { | |||
2641 | return SDNodeIterator(N, N->getNumOperands()); | |||
2642 | } | |||
2643 | ||||
2644 | unsigned getOperand() const { return Operand; } | |||
2645 | const SDNode *getNode() const { return Node; } | |||
2646 | }; | |||
2647 | ||||
2648 | template <> struct GraphTraits<SDNode*> { | |||
2649 | using NodeRef = SDNode *; | |||
2650 | using ChildIteratorType = SDNodeIterator; | |||
2651 | ||||
2652 | static NodeRef getEntryNode(SDNode *N) { return N; } | |||
2653 | ||||
2654 | static ChildIteratorType child_begin(NodeRef N) { | |||
2655 | return SDNodeIterator::begin(N); | |||
2656 | } | |||
2657 | ||||
2658 | static ChildIteratorType child_end(NodeRef N) { | |||
2659 | return SDNodeIterator::end(N); | |||
2660 | } | |||
2661 | }; | |||
2662 | ||||
2663 | /// A representation of the largest SDNode, for use in sizeof(). | |||
2664 | /// | |||
2665 | /// This needs to be a union because the largest node differs on 32 bit systems | |||
2666 | /// with 4 and 8 byte pointer alignment, respectively. | |||
2667 | using LargestSDNode = AlignedCharArrayUnion<AtomicSDNode, TargetIndexSDNode, | |||
2668 | BlockAddressSDNode, | |||
2669 | GlobalAddressSDNode, | |||
2670 | PseudoProbeSDNode>; | |||
2671 | ||||
2672 | /// The SDNode class with the greatest alignment requirement. | |||
2673 | using MostAlignedSDNode = GlobalAddressSDNode; | |||
2674 | ||||
2675 | namespace ISD { | |||
2676 | ||||
2677 | /// Returns true if the specified node is a non-extending and unindexed load. | |||
2678 | inline bool isNormalLoad(const SDNode *N) { | |||
2679 | const LoadSDNode *Ld = dyn_cast<LoadSDNode>(N); | |||
2680 | return Ld && Ld->getExtensionType() == ISD::NON_EXTLOAD && | |||
2681 | Ld->getAddressingMode() == ISD::UNINDEXED; | |||
2682 | } | |||
2683 | ||||
2684 | /// Returns true if the specified node is a non-extending load. | |||
2685 | inline bool isNON_EXTLoad(const SDNode *N) { | |||
2686 | return isa<LoadSDNode>(N) && | |||
2687 | cast<LoadSDNode>(N)->getExtensionType() == ISD::NON_EXTLOAD; | |||
2688 | } | |||
2689 | ||||
2690 | /// Returns true if the specified node is a EXTLOAD. | |||
2691 | inline bool isEXTLoad(const SDNode *N) { | |||
2692 | return isa<LoadSDNode>(N) && | |||
2693 | cast<LoadSDNode>(N)->getExtensionType() == ISD::EXTLOAD; | |||
2694 | } | |||
2695 | ||||
2696 | /// Returns true if the specified node is a SEXTLOAD. | |||
2697 | inline bool isSEXTLoad(const SDNode *N) { | |||
2698 | return isa<LoadSDNode>(N) && | |||
2699 | cast<LoadSDNode>(N)->getExtensionType() == ISD::SEXTLOAD; | |||
2700 | } | |||
2701 | ||||
2702 | /// Returns true if the specified node is a ZEXTLOAD. | |||
2703 | inline bool isZEXTLoad(const SDNode *N) { | |||
2704 | return isa<LoadSDNode>(N) && | |||
2705 | cast<LoadSDNode>(N)->getExtensionType() == ISD::ZEXTLOAD; | |||
2706 | } | |||
2707 | ||||
2708 | /// Returns true if the specified node is an unindexed load. | |||
2709 | inline bool isUNINDEXEDLoad(const SDNode *N) { | |||
2710 | return isa<LoadSDNode>(N) && | |||
2711 | cast<LoadSDNode>(N)->getAddressingMode() == ISD::UNINDEXED; | |||
2712 | } | |||
2713 | ||||
2714 | /// Returns true if the specified node is a non-truncating | |||
2715 | /// and unindexed store. | |||
2716 | inline bool isNormalStore(const SDNode *N) { | |||
2717 | const StoreSDNode *St = dyn_cast<StoreSDNode>(N); | |||
2718 | return St && !St->isTruncatingStore() && | |||
2719 | St->getAddressingMode() == ISD::UNINDEXED; | |||
2720 | } | |||
2721 | ||||
2722 | /// Returns true if the specified node is an unindexed store. | |||
2723 | inline bool isUNINDEXEDStore(const SDNode *N) { | |||
2724 | return isa<StoreSDNode>(N) && | |||
2725 | cast<StoreSDNode>(N)->getAddressingMode() == ISD::UNINDEXED; | |||
2726 | } | |||
2727 | ||||
2728 | /// Attempt to match a unary predicate against a scalar/splat constant or | |||
2729 | /// every element of a constant BUILD_VECTOR. | |||
2730 | /// If AllowUndef is true, then UNDEF elements will pass nullptr to Match. | |||
2731 | bool matchUnaryPredicate(SDValue Op, | |||
2732 | std::function<bool(ConstantSDNode *)> Match, | |||
2733 | bool AllowUndefs = false); | |||
2734 | ||||
2735 | /// Attempt to match a binary predicate against a pair of scalar/splat | |||
2736 | /// constants or every element of a pair of constant BUILD_VECTORs. | |||
2737 | /// If AllowUndef is true, then UNDEF elements will pass nullptr to Match. | |||
2738 | /// If AllowTypeMismatch is true then RetType + ArgTypes don't need to match. | |||
2739 | bool matchBinaryPredicate( | |||
2740 | SDValue LHS, SDValue RHS, | |||
2741 | std::function<bool(ConstantSDNode *, ConstantSDNode *)> Match, | |||
2742 | bool AllowUndefs = false, bool AllowTypeMismatch = false); | |||
2743 | ||||
2744 | /// Returns true if the specified value is the overflow result from one | |||
2745 | /// of the overflow intrinsic nodes. | |||
2746 | inline bool isOverflowIntrOpRes(SDValue Op) { | |||
2747 | unsigned Opc = Op.getOpcode(); | |||
2748 | return (Op.getResNo() == 1 && | |||
2749 | (Opc == ISD::SADDO || Opc == ISD::UADDO || Opc == ISD::SSUBO || | |||
2750 | Opc == ISD::USUBO || Opc == ISD::SMULO || Opc == ISD::UMULO)); | |||
2751 | } | |||
2752 | ||||
2753 | } // end namespace ISD | |||
2754 | ||||
2755 | } // end namespace llvm | |||
2756 | ||||
2757 | #endif // LLVM_CODEGEN_SELECTIONDAGNODES_H |