LLVM  6.0.0svn
Analysis.cpp
Go to the documentation of this file.
1 //===-- Analysis.cpp - CodeGen LLVM IR Analysis Utilities -----------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file defines several CodeGen-specific LLVM IR analysis utilities.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "llvm/CodeGen/Analysis.h"
18 #include "llvm/IR/DataLayout.h"
19 #include "llvm/IR/DerivedTypes.h"
20 #include "llvm/IR/Function.h"
21 #include "llvm/IR/Instructions.h"
22 #include "llvm/IR/IntrinsicInst.h"
23 #include "llvm/IR/LLVMContext.h"
24 #include "llvm/IR/Module.h"
31 
32 using namespace llvm;
33 
34 /// Compute the linearized index of a member in a nested aggregate/struct/array
35 /// by recursing and accumulating CurIndex as long as there are indices in the
36 /// index list.
38  const unsigned *Indices,
39  const unsigned *IndicesEnd,
40  unsigned CurIndex) {
41  // Base case: We're done.
42  if (Indices && Indices == IndicesEnd)
43  return CurIndex;
44 
45  // Given a struct type, recursively traverse the elements.
46  if (StructType *STy = dyn_cast<StructType>(Ty)) {
47  for (StructType::element_iterator EB = STy->element_begin(),
48  EI = EB,
49  EE = STy->element_end();
50  EI != EE; ++EI) {
51  if (Indices && *Indices == unsigned(EI - EB))
52  return ComputeLinearIndex(*EI, Indices+1, IndicesEnd, CurIndex);
53  CurIndex = ComputeLinearIndex(*EI, nullptr, nullptr, CurIndex);
54  }
55  assert(!Indices && "Unexpected out of bound");
56  return CurIndex;
57  }
58  // Given an array type, recursively traverse the elements.
59  else if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
60  Type *EltTy = ATy->getElementType();
61  unsigned NumElts = ATy->getNumElements();
62  // Compute the Linear offset when jumping one element of the array
63  unsigned EltLinearOffset = ComputeLinearIndex(EltTy, nullptr, nullptr, 0);
64  if (Indices) {
65  assert(*Indices < NumElts && "Unexpected out of bound");
66  // If the indice is inside the array, compute the index to the requested
67  // elt and recurse inside the element with the end of the indices list
68  CurIndex += EltLinearOffset* *Indices;
69  return ComputeLinearIndex(EltTy, Indices+1, IndicesEnd, CurIndex);
70  }
71  CurIndex += EltLinearOffset*NumElts;
72  return CurIndex;
73  }
74  // We haven't found the type we're looking for, so keep searching.
75  return CurIndex + 1;
76 }
77 
78 /// ComputeValueVTs - Given an LLVM IR type, compute a sequence of
79 /// EVTs that represent all the individual underlying
80 /// non-aggregate types that comprise it.
81 ///
82 /// If Offsets is non-null, it points to a vector to be filled in
83 /// with the in-memory offsets of each of the individual values.
84 ///
85 void llvm::ComputeValueVTs(const TargetLowering &TLI, const DataLayout &DL,
86  Type *Ty, SmallVectorImpl<EVT> &ValueVTs,
88  uint64_t StartingOffset) {
89  // Given a struct type, recursively traverse the elements.
90  if (StructType *STy = dyn_cast<StructType>(Ty)) {
91  const StructLayout *SL = DL.getStructLayout(STy);
92  for (StructType::element_iterator EB = STy->element_begin(),
93  EI = EB,
94  EE = STy->element_end();
95  EI != EE; ++EI)
96  ComputeValueVTs(TLI, DL, *EI, ValueVTs, Offsets,
97  StartingOffset + SL->getElementOffset(EI - EB));
98  return;
99  }
100  // Given an array type, recursively traverse the elements.
101  if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
102  Type *EltTy = ATy->getElementType();
103  uint64_t EltSize = DL.getTypeAllocSize(EltTy);
104  for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i)
105  ComputeValueVTs(TLI, DL, EltTy, ValueVTs, Offsets,
106  StartingOffset + i * EltSize);
107  return;
108  }
109  // Interpret void as zero return values.
110  if (Ty->isVoidTy())
111  return;
112  // Base case: we can get an EVT for this LLVM IR type.
113  ValueVTs.push_back(TLI.getValueType(DL, Ty));
114  if (Offsets)
115  Offsets->push_back(StartingOffset);
116 }
117 
118 /// ExtractTypeInfo - Returns the type info, possibly bitcast, encoded in V.
120  V = V->stripPointerCasts();
121  GlobalValue *GV = dyn_cast<GlobalValue>(V);
123 
124  if (Var && Var->getName() == "llvm.eh.catch.all.value") {
125  assert(Var->hasInitializer() &&
126  "The EH catch-all value must have an initializer");
127  Value *Init = Var->getInitializer();
128  GV = dyn_cast<GlobalValue>(Init);
129  if (!GV) V = cast<ConstantPointerNull>(Init);
130  }
131 
132  assert((GV || isa<ConstantPointerNull>(V)) &&
133  "TypeInfo must be a global variable or NULL");
134  return GV;
135 }
136 
137 /// hasInlineAsmMemConstraint - Return true if the inline asm instruction being
138 /// processed uses a memory 'm' constraint.
139 bool
141  const TargetLowering &TLI) {
142  for (unsigned i = 0, e = CInfos.size(); i != e; ++i) {
143  InlineAsm::ConstraintInfo &CI = CInfos[i];
144  for (unsigned j = 0, ee = CI.Codes.size(); j != ee; ++j) {
146  if (CType == TargetLowering::C_Memory)
147  return true;
148  }
149 
150  // Indirect operand accesses access memory.
151  if (CI.isIndirect)
152  return true;
153  }
154 
155  return false;
156 }
157 
158 /// getFCmpCondCode - Return the ISD condition code corresponding to
159 /// the given LLVM IR floating-point condition code. This includes
160 /// consideration of global floating-point math flags.
161 ///
163  switch (Pred) {
164  case FCmpInst::FCMP_FALSE: return ISD::SETFALSE;
165  case FCmpInst::FCMP_OEQ: return ISD::SETOEQ;
166  case FCmpInst::FCMP_OGT: return ISD::SETOGT;
167  case FCmpInst::FCMP_OGE: return ISD::SETOGE;
168  case FCmpInst::FCMP_OLT: return ISD::SETOLT;
169  case FCmpInst::FCMP_OLE: return ISD::SETOLE;
170  case FCmpInst::FCMP_ONE: return ISD::SETONE;
171  case FCmpInst::FCMP_ORD: return ISD::SETO;
172  case FCmpInst::FCMP_UNO: return ISD::SETUO;
173  case FCmpInst::FCMP_UEQ: return ISD::SETUEQ;
174  case FCmpInst::FCMP_UGT: return ISD::SETUGT;
175  case FCmpInst::FCMP_UGE: return ISD::SETUGE;
176  case FCmpInst::FCMP_ULT: return ISD::SETULT;
177  case FCmpInst::FCMP_ULE: return ISD::SETULE;
178  case FCmpInst::FCMP_UNE: return ISD::SETUNE;
179  case FCmpInst::FCMP_TRUE: return ISD::SETTRUE;
180  default: llvm_unreachable("Invalid FCmp predicate opcode!");
181  }
182 }
183 
185  switch (CC) {
186  case ISD::SETOEQ: case ISD::SETUEQ: return ISD::SETEQ;
187  case ISD::SETONE: case ISD::SETUNE: return ISD::SETNE;
188  case ISD::SETOLT: case ISD::SETULT: return ISD::SETLT;
189  case ISD::SETOLE: case ISD::SETULE: return ISD::SETLE;
190  case ISD::SETOGT: case ISD::SETUGT: return ISD::SETGT;
191  case ISD::SETOGE: case ISD::SETUGE: return ISD::SETGE;
192  default: return CC;
193  }
194 }
195 
196 /// getICmpCondCode - Return the ISD condition code corresponding to
197 /// the given LLVM IR integer condition code.
198 ///
200  switch (Pred) {
201  case ICmpInst::ICMP_EQ: return ISD::SETEQ;
202  case ICmpInst::ICMP_NE: return ISD::SETNE;
203  case ICmpInst::ICMP_SLE: return ISD::SETLE;
204  case ICmpInst::ICMP_ULE: return ISD::SETULE;
205  case ICmpInst::ICMP_SGE: return ISD::SETGE;
206  case ICmpInst::ICMP_UGE: return ISD::SETUGE;
207  case ICmpInst::ICMP_SLT: return ISD::SETLT;
208  case ICmpInst::ICMP_ULT: return ISD::SETULT;
209  case ICmpInst::ICMP_SGT: return ISD::SETGT;
210  case ICmpInst::ICMP_UGT: return ISD::SETUGT;
211  default:
212  llvm_unreachable("Invalid ICmp predicate opcode!");
213  }
214 }
215 
216 static bool isNoopBitcast(Type *T1, Type *T2,
217  const TargetLoweringBase& TLI) {
218  return T1 == T2 || (T1->isPointerTy() && T2->isPointerTy()) ||
219  (isa<VectorType>(T1) && isa<VectorType>(T2) &&
220  TLI.isTypeLegal(EVT::getEVT(T1)) && TLI.isTypeLegal(EVT::getEVT(T2)));
221 }
222 
223 /// Look through operations that will be free to find the earliest source of
224 /// this value.
225 ///
226 /// @param ValLoc If V has aggegate type, we will be interested in a particular
227 /// scalar component. This records its address; the reverse of this list gives a
228 /// sequence of indices appropriate for an extractvalue to locate the important
229 /// value. This value is updated during the function and on exit will indicate
230 /// similar information for the Value returned.
231 ///
232 /// @param DataBits If this function looks through truncate instructions, this
233 /// will record the smallest size attained.
234 static const Value *getNoopInput(const Value *V,
236  unsigned &DataBits,
237  const TargetLoweringBase &TLI,
238  const DataLayout &DL) {
239  while (true) {
240  // Try to look through V1; if V1 is not an instruction, it can't be looked
241  // through.
242  const Instruction *I = dyn_cast<Instruction>(V);
243  if (!I || I->getNumOperands() == 0) return V;
244  const Value *NoopInput = nullptr;
245 
246  Value *Op = I->getOperand(0);
247  if (isa<BitCastInst>(I)) {
248  // Look through truly no-op bitcasts.
249  if (isNoopBitcast(Op->getType(), I->getType(), TLI))
250  NoopInput = Op;
251  } else if (isa<GetElementPtrInst>(I)) {
252  // Look through getelementptr
253  if (cast<GetElementPtrInst>(I)->hasAllZeroIndices())
254  NoopInput = Op;
255  } else if (isa<IntToPtrInst>(I)) {
256  // Look through inttoptr.
257  // Make sure this isn't a truncating or extending cast. We could
258  // support this eventually, but don't bother for now.
259  if (!isa<VectorType>(I->getType()) &&
260  DL.getPointerSizeInBits() ==
261  cast<IntegerType>(Op->getType())->getBitWidth())
262  NoopInput = Op;
263  } else if (isa<PtrToIntInst>(I)) {
264  // Look through ptrtoint.
265  // Make sure this isn't a truncating or extending cast. We could
266  // support this eventually, but don't bother for now.
267  if (!isa<VectorType>(I->getType()) &&
268  DL.getPointerSizeInBits() ==
269  cast<IntegerType>(I->getType())->getBitWidth())
270  NoopInput = Op;
271  } else if (isa<TruncInst>(I) &&
272  TLI.allowTruncateForTailCall(Op->getType(), I->getType())) {
273  DataBits = std::min(DataBits, I->getType()->getPrimitiveSizeInBits());
274  NoopInput = Op;
275  } else if (auto CS = ImmutableCallSite(I)) {
276  const Value *ReturnedOp = CS.getReturnedArgOperand();
277  if (ReturnedOp && isNoopBitcast(ReturnedOp->getType(), I->getType(), TLI))
278  NoopInput = ReturnedOp;
279  } else if (const InsertValueInst *IVI = dyn_cast<InsertValueInst>(V)) {
280  // Value may come from either the aggregate or the scalar
281  ArrayRef<unsigned> InsertLoc = IVI->getIndices();
282  if (ValLoc.size() >= InsertLoc.size() &&
283  std::equal(InsertLoc.begin(), InsertLoc.end(), ValLoc.rbegin())) {
284  // The type being inserted is a nested sub-type of the aggregate; we
285  // have to remove those initial indices to get the location we're
286  // interested in for the operand.
287  ValLoc.resize(ValLoc.size() - InsertLoc.size());
288  NoopInput = IVI->getInsertedValueOperand();
289  } else {
290  // The struct we're inserting into has the value we're interested in, no
291  // change of address.
292  NoopInput = Op;
293  }
294  } else if (const ExtractValueInst *EVI = dyn_cast<ExtractValueInst>(V)) {
295  // The part we're interested in will inevitably be some sub-section of the
296  // previous aggregate. Combine the two paths to obtain the true address of
297  // our element.
298  ArrayRef<unsigned> ExtractLoc = EVI->getIndices();
299  ValLoc.append(ExtractLoc.rbegin(), ExtractLoc.rend());
300  NoopInput = Op;
301  }
302  // Terminate if we couldn't find anything to look through.
303  if (!NoopInput)
304  return V;
305 
306  V = NoopInput;
307  }
308 }
309 
310 /// Return true if this scalar return value only has bits discarded on its path
311 /// from the "tail call" to the "ret". This includes the obvious noop
312 /// instructions handled by getNoopInput above as well as free truncations (or
313 /// extensions prior to the call).
314 static bool slotOnlyDiscardsData(const Value *RetVal, const Value *CallVal,
315  SmallVectorImpl<unsigned> &RetIndices,
316  SmallVectorImpl<unsigned> &CallIndices,
317  bool AllowDifferingSizes,
318  const TargetLoweringBase &TLI,
319  const DataLayout &DL) {
320 
321  // Trace the sub-value needed by the return value as far back up the graph as
322  // possible, in the hope that it will intersect with the value produced by the
323  // call. In the simple case with no "returned" attribute, the hope is actually
324  // that we end up back at the tail call instruction itself.
325  unsigned BitsRequired = UINT_MAX;
326  RetVal = getNoopInput(RetVal, RetIndices, BitsRequired, TLI, DL);
327 
328  // If this slot in the value returned is undef, it doesn't matter what the
329  // call puts there, it'll be fine.
330  if (isa<UndefValue>(RetVal))
331  return true;
332 
333  // Now do a similar search up through the graph to find where the value
334  // actually returned by the "tail call" comes from. In the simple case without
335  // a "returned" attribute, the search will be blocked immediately and the loop
336  // a Noop.
337  unsigned BitsProvided = UINT_MAX;
338  CallVal = getNoopInput(CallVal, CallIndices, BitsProvided, TLI, DL);
339 
340  // There's no hope if we can't actually trace them to (the same part of!) the
341  // same value.
342  if (CallVal != RetVal || CallIndices != RetIndices)
343  return false;
344 
345  // However, intervening truncates may have made the call non-tail. Make sure
346  // all the bits that are needed by the "ret" have been provided by the "tail
347  // call". FIXME: with sufficiently cunning bit-tracking, we could look through
348  // extensions too.
349  if (BitsProvided < BitsRequired ||
350  (!AllowDifferingSizes && BitsProvided != BitsRequired))
351  return false;
352 
353  return true;
354 }
355 
356 /// For an aggregate type, determine whether a given index is within bounds or
357 /// not.
358 static bool indexReallyValid(CompositeType *T, unsigned Idx) {
359  if (ArrayType *AT = dyn_cast<ArrayType>(T))
360  return Idx < AT->getNumElements();
361 
362  return Idx < cast<StructType>(T)->getNumElements();
363 }
364 
365 /// Move the given iterators to the next leaf type in depth first traversal.
366 ///
367 /// Performs a depth-first traversal of the type as specified by its arguments,
368 /// stopping at the next leaf node (which may be a legitimate scalar type or an
369 /// empty struct or array).
370 ///
371 /// @param SubTypes List of the partial components making up the type from
372 /// outermost to innermost non-empty aggregate. The element currently
373 /// represented is SubTypes.back()->getTypeAtIndex(Path.back() - 1).
374 ///
375 /// @param Path Set of extractvalue indices leading from the outermost type
376 /// (SubTypes[0]) to the leaf node currently represented.
377 ///
378 /// @returns true if a new type was found, false otherwise. Calling this
379 /// function again on a finished iterator will repeatedly return
380 /// false. SubTypes.back()->getTypeAtIndex(Path.back()) is either an empty
381 /// aggregate or a non-aggregate
384  // First march back up the tree until we can successfully increment one of the
385  // coordinates in Path.
386  while (!Path.empty() && !indexReallyValid(SubTypes.back(), Path.back() + 1)) {
387  Path.pop_back();
388  SubTypes.pop_back();
389  }
390 
391  // If we reached the top, then the iterator is done.
392  if (Path.empty())
393  return false;
394 
395  // We know there's *some* valid leaf now, so march back down the tree picking
396  // out the left-most element at each node.
397  ++Path.back();
398  Type *DeeperType = SubTypes.back()->getTypeAtIndex(Path.back());
399  while (DeeperType->isAggregateType()) {
400  CompositeType *CT = cast<CompositeType>(DeeperType);
401  if (!indexReallyValid(CT, 0))
402  return true;
403 
404  SubTypes.push_back(CT);
405  Path.push_back(0);
406 
407  DeeperType = CT->getTypeAtIndex(0U);
408  }
409 
410  return true;
411 }
412 
413 /// Find the first non-empty, scalar-like type in Next and setup the iterator
414 /// components.
415 ///
416 /// Assuming Next is an aggregate of some kind, this function will traverse the
417 /// tree from left to right (i.e. depth-first) looking for the first
418 /// non-aggregate type which will play a role in function return.
419 ///
420 /// For example, if Next was {[0 x i64], {{}, i32, {}}, i32} then we would setup
421 /// Path as [1, 1] and SubTypes as [Next, {{}, i32, {}}] to represent the first
422 /// i32 in that type.
423 static bool firstRealType(Type *Next,
426  // First initialise the iterator components to the first "leaf" node
427  // (i.e. node with no valid sub-type at any index, so {} does count as a leaf
428  // despite nominally being an aggregate).
429  while (Next->isAggregateType() &&
430  indexReallyValid(cast<CompositeType>(Next), 0)) {
431  SubTypes.push_back(cast<CompositeType>(Next));
432  Path.push_back(0);
433  Next = cast<CompositeType>(Next)->getTypeAtIndex(0U);
434  }
435 
436  // If there's no Path now, Next was originally scalar already (or empty
437  // leaf). We're done.
438  if (Path.empty())
439  return true;
440 
441  // Otherwise, use normal iteration to keep looking through the tree until we
442  // find a non-aggregate type.
443  while (SubTypes.back()->getTypeAtIndex(Path.back())->isAggregateType()) {
444  if (!advanceToNextLeafType(SubTypes, Path))
445  return false;
446  }
447 
448  return true;
449 }
450 
451 /// Set the iterator data-structures to the next non-empty, non-aggregate
452 /// subtype.
455  do {
456  if (!advanceToNextLeafType(SubTypes, Path))
457  return false;
458 
459  assert(!Path.empty() && "found a leaf but didn't set the path?");
460  } while (SubTypes.back()->getTypeAtIndex(Path.back())->isAggregateType());
461 
462  return true;
463 }
464 
465 
466 /// Test if the given instruction is in a position to be optimized
467 /// with a tail-call. This roughly means that it's in a block with
468 /// a return and there's nothing that needs to be scheduled
469 /// between it and the return.
470 ///
471 /// This function only tests target-independent requirements.
473  const Instruction *I = CS.getInstruction();
474  const BasicBlock *ExitBB = I->getParent();
475  const TerminatorInst *Term = ExitBB->getTerminator();
476  const ReturnInst *Ret = dyn_cast<ReturnInst>(Term);
477 
478  // The block must end in a return statement or unreachable.
479  //
480  // FIXME: Decline tailcall if it's not guaranteed and if the block ends in
481  // an unreachable, for now. The way tailcall optimization is currently
482  // implemented means it will add an epilogue followed by a jump. That is
483  // not profitable. Also, if the callee is a special function (e.g.
484  // longjmp on x86), it can end up causing miscompilation that has not
485  // been fully understood.
486  if (!Ret &&
487  (!TM.Options.GuaranteedTailCallOpt || !isa<UnreachableInst>(Term)))
488  return false;
489 
490  // If I will have a chain, make sure no other instruction that will have a
491  // chain interposes between I and the return.
492  if (I->mayHaveSideEffects() || I->mayReadFromMemory() ||
494  for (BasicBlock::const_iterator BBI = std::prev(ExitBB->end(), 2);; --BBI) {
495  if (&*BBI == I)
496  break;
497  // Debug info intrinsics do not get in the way of tail call optimization.
498  if (isa<DbgInfoIntrinsic>(BBI))
499  continue;
500  if (BBI->mayHaveSideEffects() || BBI->mayReadFromMemory() ||
502  return false;
503  }
504 
505  const Function *F = ExitBB->getParent();
507  F, I, Ret, *TM.getSubtargetImpl(*F)->getTargetLowering());
508 }
509 
511  const ReturnInst *Ret,
512  const TargetLoweringBase &TLI,
513  bool *AllowDifferingSizes) {
514  // ADS may be null, so don't write to it directly.
515  bool DummyADS;
516  bool &ADS = AllowDifferingSizes ? *AllowDifferingSizes : DummyADS;
517  ADS = true;
518 
520  AttrBuilder CalleeAttrs(cast<CallInst>(I)->getAttributes(),
522 
523  // Noalias is completely benign as far as calling convention goes, it
524  // shouldn't affect whether the call is a tail call.
525  CallerAttrs.removeAttribute(Attribute::NoAlias);
526  CalleeAttrs.removeAttribute(Attribute::NoAlias);
527 
528  if (CallerAttrs.contains(Attribute::ZExt)) {
529  if (!CalleeAttrs.contains(Attribute::ZExt))
530  return false;
531 
532  ADS = false;
533  CallerAttrs.removeAttribute(Attribute::ZExt);
534  CalleeAttrs.removeAttribute(Attribute::ZExt);
535  } else if (CallerAttrs.contains(Attribute::SExt)) {
536  if (!CalleeAttrs.contains(Attribute::SExt))
537  return false;
538 
539  ADS = false;
540  CallerAttrs.removeAttribute(Attribute::SExt);
541  CalleeAttrs.removeAttribute(Attribute::SExt);
542  }
543 
544  // If they're still different, there's some facet we don't understand
545  // (currently only "inreg", but in future who knows). It may be OK but the
546  // only safe option is to reject the tail call.
547  return CallerAttrs == CalleeAttrs;
548 }
549 
551  const Instruction *I,
552  const ReturnInst *Ret,
553  const TargetLoweringBase &TLI) {
554  // If the block ends with a void return or unreachable, it doesn't matter
555  // what the call's return type is.
556  if (!Ret || Ret->getNumOperands() == 0) return true;
557 
558  // If the return value is undef, it doesn't matter what the call's
559  // return type is.
560  if (isa<UndefValue>(Ret->getOperand(0))) return true;
561 
562  // Make sure the attributes attached to each return are compatible.
563  bool AllowDifferingSizes;
564  if (!attributesPermitTailCall(F, I, Ret, TLI, &AllowDifferingSizes))
565  return false;
566 
567  const Value *RetVal = Ret->getOperand(0), *CallVal = I;
568  // Intrinsic like llvm.memcpy has no return value, but the expanded
569  // libcall may or may not have return value. On most platforms, it
570  // will be expanded as memcpy in libc, which returns the first
571  // argument. On other platforms like arm-none-eabi, memcpy may be
572  // expanded as library call without return value, like __aeabi_memcpy.
573  const CallInst *Call = cast<CallInst>(I);
574  if (Function *F = Call->getCalledFunction()) {
575  Intrinsic::ID IID = F->getIntrinsicID();
576  if (((IID == Intrinsic::memcpy &&
577  TLI.getLibcallName(RTLIB::MEMCPY) == StringRef("memcpy")) ||
578  (IID == Intrinsic::memmove &&
579  TLI.getLibcallName(RTLIB::MEMMOVE) == StringRef("memmove")) ||
580  (IID == Intrinsic::memset &&
581  TLI.getLibcallName(RTLIB::MEMSET) == StringRef("memset"))) &&
582  RetVal == Call->getArgOperand(0))
583  return true;
584  }
585 
586  SmallVector<unsigned, 4> RetPath, CallPath;
587  SmallVector<CompositeType *, 4> RetSubTypes, CallSubTypes;
588 
589  bool RetEmpty = !firstRealType(RetVal->getType(), RetSubTypes, RetPath);
590  bool CallEmpty = !firstRealType(CallVal->getType(), CallSubTypes, CallPath);
591 
592  // Nothing's actually returned, it doesn't matter what the callee put there
593  // it's a valid tail call.
594  if (RetEmpty)
595  return true;
596 
597  // Iterate pairwise through each of the value types making up the tail call
598  // and the corresponding return. For each one we want to know whether it's
599  // essentially going directly from the tail call to the ret, via operations
600  // that end up not generating any code.
601  //
602  // We allow a certain amount of covariance here. For example it's permitted
603  // for the tail call to define more bits than the ret actually cares about
604  // (e.g. via a truncate).
605  do {
606  if (CallEmpty) {
607  // We've exhausted the values produced by the tail call instruction, the
608  // rest are essentially undef. The type doesn't really matter, but we need
609  // *something*.
610  Type *SlotType = RetSubTypes.back()->getTypeAtIndex(RetPath.back());
611  CallVal = UndefValue::get(SlotType);
612  }
613 
614  // The manipulations performed when we're looking through an insertvalue or
615  // an extractvalue would happen at the front of the RetPath list, so since
616  // we have to copy it anyway it's more efficient to create a reversed copy.
617  SmallVector<unsigned, 4> TmpRetPath(RetPath.rbegin(), RetPath.rend());
618  SmallVector<unsigned, 4> TmpCallPath(CallPath.rbegin(), CallPath.rend());
619 
620  // Finally, we can check whether the value produced by the tail call at this
621  // index is compatible with the value we return.
622  if (!slotOnlyDiscardsData(RetVal, CallVal, TmpRetPath, TmpCallPath,
623  AllowDifferingSizes, TLI,
624  F->getParent()->getDataLayout()))
625  return false;
626 
627  CallEmpty = !nextRealType(CallSubTypes, CallPath);
628  } while(nextRealType(RetSubTypes, RetPath));
629 
630  return true;
631 }
632 
634  DenseMap<const MachineBasicBlock *, int> &FuncletMembership, int Funclet,
635  const MachineBasicBlock *MBB) {
637  while (!Worklist.empty()) {
638  const MachineBasicBlock *Visiting = Worklist.pop_back_val();
639  // Don't follow blocks which start new funclets.
640  if (Visiting->isEHPad() && Visiting != MBB)
641  continue;
642 
643  // Add this MBB to our funclet.
644  auto P = FuncletMembership.insert(std::make_pair(Visiting, Funclet));
645 
646  // Don't revisit blocks.
647  if (!P.second) {
648  assert(P.first->second == Funclet && "MBB is part of two funclets!");
649  continue;
650  }
651 
652  // Returns are boundaries where funclet transfer can occur, don't follow
653  // successors.
654  if (Visiting->isReturnBlock())
655  continue;
656 
657  for (const MachineBasicBlock *Succ : Visiting->successors())
658  Worklist.push_back(Succ);
659  }
660 }
661 
665 
666  // We don't have anything to do if there aren't any EH pads.
667  if (!MF.hasEHFunclets())
668  return FuncletMembership;
669 
670  int EntryBBNumber = MF.front().getNumber();
671  bool IsSEH = isAsynchronousEHPersonality(
673 
679  for (const MachineBasicBlock &MBB : MF) {
680  if (MBB.isEHFuncletEntry()) {
681  FuncletBlocks.push_back(&MBB);
682  } else if (IsSEH && MBB.isEHPad()) {
683  SEHCatchPads.push_back(&MBB);
684  } else if (MBB.pred_empty()) {
685  UnreachableBlocks.push_back(&MBB);
686  }
687 
688  MachineBasicBlock::const_iterator MBBI = MBB.getFirstTerminator();
689 
690  // CatchPads are not funclets for SEH so do not consider CatchRet to
691  // transfer control to another funclet.
692  if (MBBI == MBB.end() || MBBI->getOpcode() != TII->getCatchReturnOpcode())
693  continue;
694 
695  // FIXME: SEH CatchPads are not necessarily in the parent function:
696  // they could be inside a finally block.
697  const MachineBasicBlock *Successor = MBBI->getOperand(0).getMBB();
698  const MachineBasicBlock *SuccessorColor = MBBI->getOperand(1).getMBB();
699  CatchRetSuccessors.push_back(
700  {Successor, IsSEH ? EntryBBNumber : SuccessorColor->getNumber()});
701  }
702 
703  // We don't have anything to do if there aren't any EH pads.
704  if (FuncletBlocks.empty())
705  return FuncletMembership;
706 
707  // Identify all the basic blocks reachable from the function entry.
708  collectFuncletMembers(FuncletMembership, EntryBBNumber, &MF.front());
709  // All blocks not part of a funclet are in the parent function.
710  for (const MachineBasicBlock *MBB : UnreachableBlocks)
711  collectFuncletMembers(FuncletMembership, EntryBBNumber, MBB);
712  // Next, identify all the blocks inside the funclets.
713  for (const MachineBasicBlock *MBB : FuncletBlocks)
714  collectFuncletMembers(FuncletMembership, MBB->getNumber(), MBB);
715  // SEH CatchPads aren't really funclets, handle them separately.
716  for (const MachineBasicBlock *MBB : SEHCatchPads)
717  collectFuncletMembers(FuncletMembership, EntryBBNumber, MBB);
718  // Finally, identify all the targets of a catchret.
719  for (std::pair<const MachineBasicBlock *, int> CatchRetPair :
720  CatchRetSuccessors)
721  collectFuncletMembers(FuncletMembership, CatchRetPair.second,
722  CatchRetPair.first);
723  return FuncletMembership;
724 }
static unsigned getBitWidth(Type *Ty, const DataLayout &DL)
Returns the bitwidth of the given scalar or pointer type.
Return a value (possibly void), from a function.
A parsed version of the target data layout string in and methods for querying it. ...
Definition: DataLayout.h:109
ISD::CondCode getICmpCondCode(ICmpInst::Predicate Pred)
getICmpCondCode - Return the ISD condition code corresponding to the given LLVM IR integer condition ...
Definition: Analysis.cpp:199
static void collectFuncletMembers(DenseMap< const MachineBasicBlock *, int > &FuncletMembership, int Funclet, const MachineBasicBlock *MBB)
Definition: Analysis.cpp:633
This instruction extracts a struct member or array element value from an aggregate value...
Compute iterated dominance frontiers using a linear time algorithm.
Definition: AllocatorList.h:24
iterator begin() const
Definition: ArrayRef.h:137
LLVM_ATTRIBUTE_ALWAYS_INLINE size_type size() const
Definition: SmallVector.h:136
const StructLayout * getStructLayout(StructType *Ty) const
Returns a StructLayout object, indicating the alignment of the struct, its size, and the offsets of i...
Definition: DataLayout.cpp:562
This class represents a function call, abstracting a target machine&#39;s calling convention.
virtual const TargetLowering * getTargetLowering() const
reverse_iterator rbegin() const
Definition: ArrayRef.h:140
unsigned less or equal
Definition: InstrTypes.h:886
static const Value * getNoopInput(const Value *V, SmallVectorImpl< unsigned > &ValLoc, unsigned &DataBits, const TargetLoweringBase &TLI, const DataLayout &DL)
Look through operations that will be free to find the earliest source of this value.
Definition: Analysis.cpp:234
unsigned less than
Definition: InstrTypes.h:885
0 1 0 0 True if ordered and less than
Definition: InstrTypes.h:866
Offsets
Offsets in bytes from the start of the input buffer.
Definition: SIInstrInfo.h:910
1 1 1 0 True if unordered or not equal
Definition: InstrTypes.h:876
unsigned getPointerSizeInBits(unsigned AS=0) const
Layout pointer size, in bits FIXME: The defaults need to be removed once all of the backends/clients ...
Definition: DataLayout.h:346
The two locations do not alias at all.
Definition: AliasAnalysis.h:85
F(f)
GlobalValue * ExtractTypeInfo(Value *V)
ExtractTypeInfo - Returns the type info, possibly bitcast, encoded in V.
Definition: Analysis.cpp:119
bool returnTypeIsEligibleForTailCall(const Function *F, const Instruction *I, const ReturnInst *Ret, const TargetLoweringBase &TLI)
Test if given that the input instruction is in the tail call position if the return type or any attri...
Definition: Analysis.cpp:550
ConstraintCodeVector Codes
Code - The constraint code, either the register name (in braces) or the constraint letter/number...
Definition: InlineAsm.h:149
iterator_range< succ_iterator > successors()
std::pair< iterator, bool > insert(const std::pair< KeyT, ValueT > &KV)
Definition: DenseMap.h:191
1 0 0 1 True if unordered or equal
Definition: InstrTypes.h:871
Used to lazily calculate structure layout information for a target machine, based on the DataLayout s...
Definition: DataLayout.h:493
1 0 0 0 True if unordered: isnan(X) | isnan(Y)
Definition: InstrTypes.h:870
bool isReturnBlock() const
Convenience function that returns true if the block ends in a return instruction. ...
const DataLayout & getDataLayout() const
Get the data layout for the module&#39;s target platform.
Definition: Module.cpp:361
const HexagonInstrInfo * TII
Class to represent struct types.
Definition: DerivedTypes.h:201
This class consists of common code factored out of the SmallVector class to reduce code duplication b...
Definition: APFloat.h:42
0 1 0 1 True if ordered and less than or equal
Definition: InstrTypes.h:867
InstrTy * getInstruction() const
Definition: CallSite.h:92
unsigned getCatchReturnOpcode() const
This class defines information used to lower LLVM code to legal SelectionDAG operators that the targe...
AttributeList getAttributes(LLVMContext &C, ID id)
Return the attributes for an intrinsic.
Type * getType() const
All values are typed, get the type of this value.
Definition: Value.h:245
#define T
Class to represent array types.
Definition: DerivedTypes.h:369
AttributeList getAttributes() const
Return the attribute list for this Function.
Definition: Function.h:205
CondCode
ISD::CondCode enum - These are ordered carefully to make the bitfields below work out...
Definition: ISDOpcodes.h:916
int getNumber() const
MachineBasicBlocks are uniquely numbered at the function level, unless they&#39;re not in a MachineFuncti...
virtual const TargetInstrInfo * getInstrInfo() const
ISD::CondCode getFCmpCondCode(FCmpInst::Predicate Pred)
getFCmpCondCode - Return the ISD condition code corresponding to the given LLVM IR floating-point con...
Definition: Analysis.cpp:162
Type::subtype_iterator element_iterator
Definition: DerivedTypes.h:301
void ComputeValueVTs(const TargetLowering &TLI, const DataLayout &DL, Type *Ty, SmallVectorImpl< EVT > &ValueVTs, SmallVectorImpl< uint64_t > *Offsets=nullptr, uint64_t StartingOffset=0)
ComputeValueVTs - Given an LLVM IR type, compute a sequence of EVTs that represent all the individual...
Definition: Analysis.cpp:85
Value * getOperand(unsigned i) const
Definition: User.h:154
bool hasInlineAsmMemConstraint(InlineAsm::ConstraintInfoVector &CInfos, const TargetLowering &TLI)
hasInlineAsmMemConstraint - Return true if the inline asm instruction being processed uses a memory &#39;...
Definition: Analysis.cpp:140
TargetInstrInfo - Interface to description of machine instruction set.
bool isVoidTy() const
Return true if this is &#39;void&#39;.
Definition: Type.h:141
static bool indexReallyValid(CompositeType *T, unsigned Idx)
For an aggregate type, determine whether a given index is within bounds or not.
Definition: Analysis.cpp:358
static bool nextRealType(SmallVectorImpl< CompositeType *> &SubTypes, SmallVectorImpl< unsigned > &Path)
Set the iterator data-structures to the next non-empty, non-aggregate subtype.
Definition: Analysis.cpp:453
#define P(N)
Subclasses of this class are all able to terminate a basic block.
Definition: InstrTypes.h:54
LLVM Basic Block Representation.
Definition: BasicBlock.h:59
const TargetSubtargetInfo & getSubtarget() const
getSubtarget - Return the subtarget for which this machine code is being compiled.
The instances of the Type class are immutable: once they are created, they are never changed...
Definition: Type.h:46
virtual ConstraintType getConstraintType(StringRef Constraint) const
Given a constraint, return the type of constraint it is for this target.
bool isIndirect
isIndirect - True if this operand is an indirect operand.
Definition: InlineAsm.h:145
size_t size() const
size - Get the array size.
Definition: ArrayRef.h:149
bool hasEHFunclets() const
bool isPointerTy() const
True if this is an instance of PointerType.
Definition: Type.h:221
bool mayHaveSideEffects() const
Return true if the instruction may have side effects.
Definition: Instruction.h:504
EHPersonality classifyEHPersonality(const Value *Pers)
See if the given exception handling personality function is one that we understand.
Predicate
This enumeration lists the possible predicates for CmpInst subclasses.
Definition: InstrTypes.h:860
unsigned GuaranteedTailCallOpt
GuaranteedTailCallOpt - This flag is enabled when -tailcallopt is specified on the commandline...
static bool firstRealType(Type *Next, SmallVectorImpl< CompositeType *> &SubTypes, SmallVectorImpl< unsigned > &Path)
Find the first non-empty, scalar-like type in Next and setup the iterator components.
Definition: Analysis.cpp:423
0 1 1 1 True if ordered (no nans)
Definition: InstrTypes.h:869
1 1 1 1 Always true (always folded)
Definition: InstrTypes.h:877
virtual bool allowTruncateForTailCall(Type *FromTy, Type *ToTy) const
Return true if a truncation from FromTy to ToTy is permitted when deciding whether a call is in tail ...
static UndefValue * get(Type *T)
Static factory methods - Return an &#39;undef&#39; object of the specified type.
Definition: Constants.cpp:1320
const Value * stripPointerCasts() const
Strip off pointer casts, all-zero GEPs, and aliases.
Definition: Value.cpp:527
static bool slotOnlyDiscardsData(const Value *RetVal, const Value *CallVal, SmallVectorImpl< unsigned > &RetIndices, SmallVectorImpl< unsigned > &CallIndices, bool AllowDifferingSizes, const TargetLoweringBase &TLI, const DataLayout &DL)
Return true if this scalar return value only has bits discarded on its path from the "tail call" to t...
Definition: Analysis.cpp:314
const MachineBasicBlock & front() const
1 1 0 1 True if unordered, less than, or equal
Definition: InstrTypes.h:875
#define llvm_unreachable(msg)
Marks that the current location is not supposed to be reachable.
EVT getValueType(const DataLayout &DL, Type *Ty, bool AllowUnknown=false) const
Return the EVT corresponding to this LLVM type.
bool isInTailCallPosition(ImmutableCallSite CS, const TargetMachine &TM)
Test if the given instruction is in a position to be optimized with a tail-call.
Definition: Analysis.cpp:472
signed greater than
Definition: InstrTypes.h:887
std::vector< ConstraintInfo > ConstraintInfoVector
Definition: InlineAsm.h:116
This base class for TargetLowering contains the SelectionDAG-independent parts that can be used from ...
0 0 1 0 True if ordered and greater than
Definition: InstrTypes.h:864
Iterator for intrusive lists based on ilist_node.
unsigned getNumOperands() const
Definition: User.h:176
static bool isNoopBitcast(Type *T1, Type *T2, const TargetLoweringBase &TLI)
Definition: Analysis.cpp:216
iterator end()
Definition: BasicBlock.h:254
1 1 0 0 True if unordered or less than
Definition: InstrTypes.h:874
Module.h This file contains the declarations for the Module class.
DenseMap< const MachineBasicBlock *, int > getFuncletMembership(const MachineFunction &MF)
Definition: Analysis.cpp:663
iterator end() const
Definition: ArrayRef.h:138
bool isAggregateType() const
Return true if the type is an aggregate type.
Definition: Type.h:255
signed less than
Definition: InstrTypes.h:889
LLVM_NODISCARD T pop_back_val()
Definition: SmallVector.h:385
static bool advanceToNextLeafType(SmallVectorImpl< CompositeType *> &SubTypes, SmallVectorImpl< unsigned > &Path)
Move the given iterators to the next leaf type in depth first traversal.
Definition: Analysis.cpp:382
ISD::CondCode getFCmpCodeWithoutNaN(ISD::CondCode CC)
getFCmpCodeWithoutNaN - Given an ISD condition code comparing floats, return the equivalent code if w...
Definition: Analysis.cpp:184
Intrinsic::ID getIntrinsicID() const LLVM_READONLY
getIntrinsicID - This method returns the ID number of the specified function, or Intrinsic::not_intri...
Definition: Function.h:175
virtual const TargetSubtargetInfo * getSubtargetImpl(const Function &) const
Virtual method implemented by subclasses that returns a reference to that target&#39;s TargetSubtargetInf...
signed less or equal
Definition: InstrTypes.h:890
bool isTypeLegal(EVT VT) const
Return true if the target has native support for the specified value type.
void append(in_iter in_start, in_iter in_end)
Add the specified range to the end of the SmallVector.
Definition: SmallVector.h:398
Common super class of ArrayType, StructType and VectorType.
Definition: DerivedTypes.h:162
uint64_t getTypeAllocSize(Type *Ty) const
Returns the offset in bytes between successive objects of the specified type, including alignment pad...
Definition: DataLayout.h:405
Function * getCalledFunction() const
Return the function called, or null if this is an indirect function invocation.
reverse_iterator rend() const
Definition: ArrayRef.h:141
uint64_t getElementOffset(unsigned Idx) const
Definition: DataLayout.h:515
bool isEHPad() const
Returns true if the block is a landing pad.
LLVM_NODISCARD bool empty() const
Definition: SmallVector.h:61
unsigned greater or equal
Definition: InstrTypes.h:884
Value * getArgOperand(unsigned i) const
getArgOperand/setArgOperand - Return/set the i-th call argument.
TargetOptions Options
Definition: TargetMachine.h:96
Establish a view to a call site for examination.
Definition: CallSite.h:695
const Function * getParent() const
Return the enclosing method, or null if none.
Definition: BasicBlock.h:108
#define I(x, y, z)
Definition: MD5.cpp:58
bool mayReadFromMemory() const
Return true if this instruction may read memory.
0 1 1 0 True if ordered and operands are unequal
Definition: InstrTypes.h:868
LLVM_NODISCARD std::enable_if<!is_simple_type< Y >::value, typename cast_retty< X, const Y >::ret_type >::type dyn_cast(const Y &Val)
Definition: Casting.h:323
1 0 1 0 True if unordered or greater than
Definition: InstrTypes.h:872
static EVT getEVT(Type *Ty, bool HandleUnknown=false)
Return the value type corresponding to the specified type.
Definition: ValueTypes.cpp:311
const Function * getFunction() const
getFunction - Return the LLVM function that this machine code represents
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
bool isAsynchronousEHPersonality(EHPersonality Pers)
Returns true if this personality function catches asynchronous exceptions.
bool isSafeToSpeculativelyExecute(const Value *V, const Instruction *CtxI=nullptr, const DominatorTree *DT=nullptr)
Return true if the instruction does not have any effects besides calculating the result and does not ...
unsigned getPrimitiveSizeInBits() const LLVM_READONLY
Return the basic size of this type if it is a primitive type.
Definition: Type.cpp:115
0 0 0 1 True if ordered and equal
Definition: InstrTypes.h:863
Module * getParent()
Get the module that this global value is contained inside of...
Definition: GlobalValue.h:545
LLVM Value Representation.
Definition: Value.h:73
Constant * getPersonalityFn() const
Get the personality function associated with this function.
Definition: Function.cpp:1255
1 0 1 1 True if unordered, greater than, or equal
Definition: InstrTypes.h:873
Primary interface to the complete machine description for the target machine.
Definition: TargetMachine.h:57
unsigned greater than
Definition: InstrTypes.h:883
StringRef - Represent a constant reference to a string, i.e.
Definition: StringRef.h:49
const TerminatorInst * getTerminator() const LLVM_READONLY
Returns the terminator instruction if the block is well formed or null if the block is not well forme...
Definition: BasicBlock.cpp:120
bool attributesPermitTailCall(const Function *F, const Instruction *I, const ReturnInst *Ret, const TargetLoweringBase &TLI, bool *AllowDifferingSizes=nullptr)
Test if given that the input instruction is in the tail call position, if there is an attribute misma...
Definition: Analysis.cpp:510
0 0 1 1 True if ordered and greater than or equal
Definition: InstrTypes.h:865
unsigned ComputeLinearIndex(Type *Ty, const unsigned *Indices, const unsigned *IndicesEnd, unsigned CurIndex=0)
Compute the linearized index of a member in a nested aggregate/struct/array.
Definition: Analysis.cpp:37
#define T1
0 0 0 0 Always false (always folded)
Definition: InstrTypes.h:862
signed greater or equal
Definition: InstrTypes.h:888
This file describes how to lower LLVM code to machine code.
Type * getTypeAtIndex(const Value *V) const
Given an index value into the type, return the type of the element.
Definition: Type.cpp:517
const BasicBlock * getParent() const
Definition: Instruction.h:66
const char * getLibcallName(RTLIB::Libcall Call) const
Get the libcall routine name for the specified libcall.
This instruction inserts a struct field of array element value into an aggregate value.
void resize(size_type N)
Definition: SmallVector.h:355