28#include "llvm/IR/IntrinsicsLoongArch.h"
37#define DEBUG_TYPE "loongarch-isel-lowering"
42 cl::desc(
"Trap on integer division by zero."),
54 if (Subtarget.hasBasicF())
56 if (Subtarget.hasBasicD())
60 MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64, MVT::v4f32, MVT::v2f64};
62 MVT::v32i8, MVT::v16i16, MVT::v8i32, MVT::v4i64, MVT::v8f32, MVT::v4f64};
64 if (Subtarget.hasExtLSX())
68 if (Subtarget.hasExtLASX())
69 for (
MVT VT : LASXVTs)
167 if (Subtarget.hasBasicF()) {
191 if (!Subtarget.hasBasicD()) {
202 if (Subtarget.hasBasicD()) {
231 if (Subtarget.hasExtLSX()) {
246 for (
MVT VT : LSXVTs) {
259 for (
MVT VT : {MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64}) {
273 for (
MVT VT : {MVT::v16i8, MVT::v8i16, MVT::v4i32})
275 for (
MVT VT : {MVT::v8i16, MVT::v4i32, MVT::v2i64})
277 for (
MVT VT : {MVT::v4i32, MVT::v2i64}) {
281 for (
MVT VT : {MVT::v4f32, MVT::v2f64}) {
300 if (Subtarget.hasExtLASX()) {
301 for (
MVT VT : LASXVTs) {
315 for (
MVT VT : {MVT::v4i64, MVT::v8i32, MVT::v16i16, MVT::v32i8}) {
329 for (
MVT VT : {MVT::v32i8, MVT::v16i16, MVT::v8i32})
331 for (
MVT VT : {MVT::v16i16, MVT::v8i32, MVT::v4i64})
333 for (
MVT VT : {MVT::v8i32, MVT::v4i32, MVT::v4i64}) {
337 for (
MVT VT : {MVT::v8f32, MVT::v4f64}) {
358 if (Subtarget.hasExtLSX())
381 if (Subtarget.hasLAMCAS())
396 switch (
Op.getOpcode()) {
398 return lowerATOMIC_FENCE(
Op, DAG);
400 return lowerEH_DWARF_CFA(
Op, DAG);
402 return lowerGlobalAddress(
Op, DAG);
404 return lowerGlobalTLSAddress(
Op, DAG);
406 return lowerINTRINSIC_WO_CHAIN(
Op, DAG);
408 return lowerINTRINSIC_W_CHAIN(
Op, DAG);
410 return lowerINTRINSIC_VOID(
Op, DAG);
412 return lowerBlockAddress(
Op, DAG);
414 return lowerJumpTable(
Op, DAG);
416 return lowerShiftLeftParts(
Op, DAG);
418 return lowerShiftRightParts(
Op, DAG,
true);
420 return lowerShiftRightParts(
Op, DAG,
false);
422 return lowerConstantPool(
Op, DAG);
424 return lowerFP_TO_SINT(
Op, DAG);
426 return lowerBITCAST(
Op, DAG);
428 return lowerUINT_TO_FP(
Op, DAG);
430 return lowerSINT_TO_FP(
Op, DAG);
432 return lowerVASTART(
Op, DAG);
434 return lowerFRAMEADDR(
Op, DAG);
436 return lowerRETURNADDR(
Op, DAG);
438 return lowerWRITE_REGISTER(
Op, DAG);
440 return lowerINSERT_VECTOR_ELT(
Op, DAG);
442 return lowerEXTRACT_VECTOR_ELT(
Op, DAG);
444 return lowerBUILD_VECTOR(
Op, DAG);
446 return lowerVECTOR_SHUFFLE(
Op, DAG);
448 return lowerBITREVERSE(
Op, DAG);
455 EVT ResTy =
Op->getValueType(0);
466 for (
unsigned int i = 0; i < NewEltNum; i++) {
469 SDValue RevOp = DAG.
getNode((ResTy == MVT::v16i8 || ResTy == MVT::v32i8)
489 for (
unsigned int i = 0; i < NewEltNum; i++)
490 for (
int j = OrigEltNum / NewEltNum - 1; j >= 0; j--)
491 Mask.push_back(j + (OrigEltNum / NewEltNum) * i);
499template <
typename ValType>
502 unsigned CheckStride,
504 ValType ExpectedIndex,
unsigned ExpectedIndexStride) {
508 if (*
I != -1 && *
I != ExpectedIndex)
510 ExpectedIndex += ExpectedIndexStride;
514 for (
unsigned n = 0; n < CheckStride &&
I !=
End; ++n, ++
I)
533 for (
const auto &M : Mask) {
540 if (SplatIndex == -1)
543 assert(SplatIndex < (
int)Mask.size() &&
"Out of bounds mask index");
544 if (fitsRegularPattern<int>(Mask.begin(), 1, Mask.end(), SplatIndex, 0)) {
545 APInt Imm(64, SplatIndex);
579 int SubMask[4] = {-1, -1, -1, -1};
580 for (
unsigned i = 0; i < 4; ++i) {
581 for (
unsigned j = i; j < Mask.size(); j += 4) {
588 if (Idx < 0 || Idx >= 4)
594 if (SubMask[i] == -1)
598 else if (
Idx != -1 &&
Idx != SubMask[i])
605 for (
int i = 3; i >= 0; --i) {
606 int Idx = SubMask[i];
638 const auto &Begin = Mask.begin();
639 const auto &
End = Mask.end();
640 SDValue OriV1 = V1, OriV2 = V2;
642 if (fitsRegularPattern<int>(Begin, 2,
End, 0, 2))
644 else if (fitsRegularPattern<int>(Begin, 2,
End, Mask.size(), 2))
649 if (fitsRegularPattern<int>(Begin + 1, 2,
End, 0, 2))
651 else if (fitsRegularPattern<int>(Begin + 1, 2,
End, Mask.size(), 2))
678 const auto &Begin = Mask.begin();
679 const auto &
End = Mask.end();
680 SDValue OriV1 = V1, OriV2 = V2;
682 if (fitsRegularPattern<int>(Begin, 2,
End, 1, 2))
684 else if (fitsRegularPattern<int>(Begin, 2,
End, Mask.size() + 1, 2))
689 if (fitsRegularPattern<int>(Begin + 1, 2,
End, 1, 2))
691 else if (fitsRegularPattern<int>(Begin + 1, 2,
End, Mask.size() + 1, 2))
719 const auto &Begin = Mask.begin();
720 const auto &
End = Mask.end();
721 unsigned HalfSize = Mask.size() / 2;
722 SDValue OriV1 = V1, OriV2 = V2;
724 if (fitsRegularPattern<int>(Begin, 2,
End, HalfSize, 1))
726 else if (fitsRegularPattern<int>(Begin, 2,
End, Mask.size() + HalfSize, 1))
731 if (fitsRegularPattern<int>(Begin + 1, 2,
End, HalfSize, 1))
733 else if (fitsRegularPattern<int>(Begin + 1, 2,
End, Mask.size() + HalfSize,
762 const auto &Begin = Mask.begin();
763 const auto &
End = Mask.end();
764 SDValue OriV1 = V1, OriV2 = V2;
766 if (fitsRegularPattern<int>(Begin, 2,
End, 0, 1))
768 else if (fitsRegularPattern<int>(Begin, 2,
End, Mask.size(), 1))
773 if (fitsRegularPattern<int>(Begin + 1, 2,
End, 0, 1))
775 else if (fitsRegularPattern<int>(Begin + 1, 2,
End, Mask.size(), 1))
802 const auto &Begin = Mask.begin();
803 const auto &Mid = Mask.begin() + Mask.size() / 2;
804 const auto &
End = Mask.end();
805 SDValue OriV1 = V1, OriV2 = V2;
807 if (fitsRegularPattern<int>(Begin, 1, Mid, 0, 2))
809 else if (fitsRegularPattern<int>(Begin, 1, Mid, Mask.size(), 2))
814 if (fitsRegularPattern<int>(Mid, 1,
End, 0, 2))
816 else if (fitsRegularPattern<int>(Mid, 1,
End, Mask.size(), 2))
844 const auto &Begin = Mask.begin();
845 const auto &Mid = Mask.begin() + Mask.size() / 2;
846 const auto &
End = Mask.end();
847 SDValue OriV1 = V1, OriV2 = V2;
849 if (fitsRegularPattern<int>(Begin, 1, Mid, 1, 2))
851 else if (fitsRegularPattern<int>(Begin, 1, Mid, Mask.size() + 1, 2))
856 if (fitsRegularPattern<int>(Mid, 1,
End, 1, 2))
858 else if (fitsRegularPattern<int>(Mid, 1,
End, Mask.size() + 1, 2))
900 "Vector type is unsupported for lsx!");
902 "Two operands have different types!");
904 "Unexpected mask size for shuffle!");
905 assert(Mask.size() % 2 == 0 &&
"Expected even mask size.");
953 for (
const auto &M : Mask) {
960 if (SplatIndex == -1)
963 const auto &Begin = Mask.begin();
964 const auto &
End = Mask.end();
965 unsigned HalfSize = Mask.size() / 2;
967 assert(SplatIndex < (
int)Mask.size() &&
"Out of bounds mask index");
968 if (fitsRegularPattern<int>(Begin, 1,
End - HalfSize, SplatIndex, 0) &&
969 fitsRegularPattern<int>(Begin + HalfSize, 1,
End, SplatIndex + HalfSize,
971 APInt Imm(64, SplatIndex);
985 if (Mask.size() <= 4)
1009 const auto &Begin = Mask.begin();
1010 const auto &
End = Mask.end();
1011 unsigned HalfSize = Mask.size() / 2;
1012 unsigned LeftSize = HalfSize / 2;
1013 SDValue OriV1 = V1, OriV2 = V2;
1015 if (fitsRegularPattern<int>(Begin, 2,
End - HalfSize, HalfSize - LeftSize,
1017 fitsRegularPattern<int>(Begin + HalfSize, 2,
End, HalfSize + LeftSize, 1))
1019 else if (fitsRegularPattern<int>(Begin, 2,
End - HalfSize,
1020 Mask.size() + HalfSize - LeftSize, 1) &&
1021 fitsRegularPattern<int>(Begin + HalfSize, 2,
End,
1022 Mask.size() + HalfSize + LeftSize, 1))
1027 if (fitsRegularPattern<int>(Begin + 1, 2,
End - HalfSize, HalfSize - LeftSize,
1029 fitsRegularPattern<int>(Begin + 1 + HalfSize, 2,
End, HalfSize + LeftSize,
1032 else if (fitsRegularPattern<int>(Begin + 1, 2,
End - HalfSize,
1033 Mask.size() + HalfSize - LeftSize, 1) &&
1034 fitsRegularPattern<int>(Begin + 1 + HalfSize, 2,
End,
1035 Mask.size() + HalfSize + LeftSize, 1))
1048 const auto &Begin = Mask.begin();
1049 const auto &
End = Mask.end();
1050 unsigned HalfSize = Mask.size() / 2;
1051 SDValue OriV1 = V1, OriV2 = V2;
1053 if (fitsRegularPattern<int>(Begin, 2,
End - HalfSize, 0, 1) &&
1054 fitsRegularPattern<int>(Begin + HalfSize, 2,
End, HalfSize, 1))
1056 else if (fitsRegularPattern<int>(Begin, 2,
End - HalfSize, Mask.size(), 1) &&
1057 fitsRegularPattern<int>(Begin + HalfSize, 2,
End,
1058 Mask.size() + HalfSize, 1))
1063 if (fitsRegularPattern<int>(Begin + 1, 2,
End - HalfSize, 0, 1) &&
1064 fitsRegularPattern<int>(Begin + 1 + HalfSize, 2,
End, HalfSize, 1))
1066 else if (fitsRegularPattern<int>(Begin + 1, 2,
End - HalfSize, Mask.size(),
1068 fitsRegularPattern<int>(Begin + 1 + HalfSize, 2,
End,
1069 Mask.size() + HalfSize, 1))
1082 const auto &Begin = Mask.begin();
1083 const auto &LeftMid = Mask.begin() + Mask.size() / 4;
1084 const auto &Mid = Mask.begin() + Mask.size() / 2;
1085 const auto &RightMid = Mask.end() - Mask.size() / 4;
1086 const auto &
End = Mask.end();
1087 unsigned HalfSize = Mask.size() / 2;
1088 SDValue OriV1 = V1, OriV2 = V2;
1090 if (fitsRegularPattern<int>(Begin, 1, LeftMid, 0, 2) &&
1091 fitsRegularPattern<int>(Mid, 1, RightMid, HalfSize, 2))
1093 else if (fitsRegularPattern<int>(Begin, 1, LeftMid, Mask.size(), 2) &&
1094 fitsRegularPattern<int>(Mid, 1, RightMid, Mask.size() + HalfSize, 2))
1099 if (fitsRegularPattern<int>(LeftMid, 1, Mid, 0, 2) &&
1100 fitsRegularPattern<int>(RightMid, 1,
End, HalfSize, 2))
1102 else if (fitsRegularPattern<int>(LeftMid, 1, Mid, Mask.size(), 2) &&
1103 fitsRegularPattern<int>(RightMid, 1,
End, Mask.size() + HalfSize, 2))
1117 const auto &Begin = Mask.begin();
1118 const auto &LeftMid = Mask.begin() + Mask.size() / 4;
1119 const auto &Mid = Mask.begin() + Mask.size() / 2;
1120 const auto &RightMid = Mask.end() - Mask.size() / 4;
1121 const auto &
End = Mask.end();
1122 unsigned HalfSize = Mask.size() / 2;
1123 SDValue OriV1 = V1, OriV2 = V2;
1125 if (fitsRegularPattern<int>(Begin, 1, LeftMid, 1, 2) &&
1126 fitsRegularPattern<int>(Mid, 1, RightMid, HalfSize + 1, 2))
1128 else if (fitsRegularPattern<int>(Begin, 1, LeftMid, Mask.size() + 1, 2) &&
1129 fitsRegularPattern<int>(Mid, 1, RightMid, Mask.size() + HalfSize + 1,
1135 if (fitsRegularPattern<int>(LeftMid, 1, Mid, 1, 2) &&
1136 fitsRegularPattern<int>(RightMid, 1,
End, HalfSize + 1, 2))
1138 else if (fitsRegularPattern<int>(LeftMid, 1, Mid, Mask.size() + 1, 2) &&
1139 fitsRegularPattern<int>(RightMid, 1,
End, Mask.size() + HalfSize + 1,
1153 int MaskSize = Mask.size();
1154 int HalfSize = Mask.size() / 2;
1155 const auto &Begin = Mask.begin();
1156 const auto &Mid = Mask.begin() + HalfSize;
1157 const auto &
End = Mask.end();
1169 for (
auto it = Begin; it < Mid; it++) {
1172 else if ((*it >= 0 && *it < HalfSize) ||
1173 (*it >= MaskSize && *it <= MaskSize + HalfSize)) {
1174 int M = *it < HalfSize ? *it : *it - HalfSize;
1179 assert((
int)MaskAlloc.
size() == HalfSize &&
"xvshuf convert failed!");
1181 for (
auto it = Mid; it <
End; it++) {
1184 else if ((*it >= HalfSize && *it < MaskSize) ||
1185 (*it >= MaskSize + HalfSize && *it < MaskSize * 2)) {
1186 int M = *it < MaskSize ? *it - HalfSize : *it - MaskSize;
1191 assert((
int)MaskAlloc.
size() == MaskSize &&
"xvshuf convert failed!");
1222 enum HalfMaskType { HighLaneTy, LowLaneTy,
None };
1224 int MaskSize = Mask.size();
1225 int HalfSize = Mask.size() / 2;
1227 HalfMaskType preMask =
None, postMask =
None;
1229 if (std::all_of(Mask.begin(), Mask.begin() + HalfSize, [&](
int M) {
1230 return M < 0 || (M >= 0 && M < HalfSize) ||
1231 (M >= MaskSize && M < MaskSize + HalfSize);
1233 preMask = HighLaneTy;
1234 else if (std::all_of(Mask.begin(), Mask.begin() + HalfSize, [&](
int M) {
1235 return M < 0 || (M >= HalfSize && M < MaskSize) ||
1236 (M >= MaskSize + HalfSize && M < MaskSize * 2);
1238 preMask = LowLaneTy;
1240 if (std::all_of(Mask.begin() + HalfSize, Mask.end(), [&](
int M) {
1241 return M < 0 || (M >= 0 && M < HalfSize) ||
1242 (M >= MaskSize && M < MaskSize + HalfSize);
1244 postMask = HighLaneTy;
1245 else if (std::all_of(Mask.begin() + HalfSize, Mask.end(), [&](
int M) {
1246 return M < 0 || (M >= HalfSize && M < MaskSize) ||
1247 (M >= MaskSize + HalfSize && M < MaskSize * 2);
1249 postMask = LowLaneTy;
1257 if (preMask == HighLaneTy && postMask == LowLaneTy) {
1260 if (preMask == LowLaneTy && postMask == HighLaneTy) {
1266 if (!V2.isUndef()) {
1273 for (
auto it = Mask.begin(); it < Mask.begin() + HalfSize; it++) {
1274 *it = *it < 0 ? *it : *it - HalfSize;
1276 for (
auto it = Mask.begin() + HalfSize; it < Mask.end(); it++) {
1277 *it = *it < 0 ? *it : *it + HalfSize;
1279 }
else if (preMask == LowLaneTy && postMask == LowLaneTy) {
1285 if (!V2.isUndef()) {
1292 for (
auto it = Mask.begin(); it < Mask.begin() + HalfSize; it++) {
1293 *it = *it < 0 ? *it : *it - HalfSize;
1295 }
else if (preMask == HighLaneTy && postMask == HighLaneTy) {
1301 if (!V2.isUndef()) {
1308 for (
auto it = Mask.begin() + HalfSize; it < Mask.end(); it++) {
1309 *it = *it < 0 ? *it : *it + HalfSize;
1325 "Vector type is unsupported for lasx!");
1327 "Two operands have different types!");
1329 "Unexpected mask size for shuffle!");
1330 assert(Mask.size() % 2 == 0 &&
"Expected even mask size.");
1331 assert(Mask.size() >= 4 &&
"Mask size is less than 4.");
1376 MVT VT =
Op.getSimpleValueType();
1380 bool V1IsUndef = V1.
isUndef();
1381 bool V2IsUndef =
V2.isUndef();
1382 if (V1IsUndef && V2IsUndef)
1395 any_of(OrigMask, [NumElements](
int M) {
return M >= NumElements; })) {
1397 for (
int &M : NewMask)
1398 if (M >= NumElements)
1404 int MaskUpperLimit = OrigMask.
size() * (V2IsUndef ? 1 : 2);
1405 (void)MaskUpperLimit;
1407 [&](
int M) {
return -1 <=
M &&
M < MaskUpperLimit; }) &&
1408 "Out of bounds shuffle index");
1423 if (isa<ConstantSDNode>(
Op))
1425 if (isa<ConstantFPSDNode>(
Op))
1440 EVT ResTy =
Op->getValueType(0);
1442 APInt SplatValue, SplatUndef;
1443 unsigned SplatBitSize;
1448 if ((!Subtarget.hasExtLSX() || !Is128Vec) &&
1449 (!Subtarget.hasExtLASX() || !Is256Vec))
1452 if (
Node->isConstantSplat(SplatValue, SplatUndef, SplatBitSize, HasAnyUndefs,
1454 SplatBitSize <= 64) {
1456 if (SplatBitSize != 8 && SplatBitSize != 16 && SplatBitSize != 32 &&
1462 switch (SplatBitSize) {
1466 ViaVecTy = Is128Vec ? MVT::v16i8 : MVT::v32i8;
1469 ViaVecTy = Is128Vec ? MVT::v8i16 : MVT::v16i16;
1472 ViaVecTy = Is128Vec ? MVT::v4i32 : MVT::v8i32;
1475 ViaVecTy = Is128Vec ? MVT::v2i64 : MVT::v4i64;
1483 if (ViaVecTy != ResTy)
1496 EVT ResTy =
Node->getValueType(0);
1502 for (
unsigned i = 0; i < NumElts; ++i) {
1504 Node->getOperand(i),
1514LoongArchTargetLowering::lowerEXTRACT_VECTOR_ELT(
SDValue Op,
1516 EVT VecTy =
Op->getOperand(0)->getValueType(0);
1521 if (isa<ConstantSDNode>(
Idx) &&
1522 (EltTy == MVT::i32 || EltTy == MVT::i64 || EltTy == MVT::f32 ||
1523 EltTy == MVT::f64 ||
Idx->getAsZExtVal() < NumElts / 2))
1530LoongArchTargetLowering::lowerINSERT_VECTOR_ELT(
SDValue Op,
1532 if (isa<ConstantSDNode>(
Op->getOperand(2)))
1556 if (Subtarget.
is64Bit() &&
Op.getOperand(2).getValueType() == MVT::i32) {
1558 "On LA64, only 64-bit registers can be written.");
1559 return Op.getOperand(0);
1562 if (!Subtarget.
is64Bit() &&
Op.getOperand(2).getValueType() == MVT::i64) {
1564 "On LA32, only 32-bit registers can be written.");
1565 return Op.getOperand(0);
1573 if (!isa<ConstantSDNode>(
Op.getOperand(0))) {
1575 "be a constant integer");
1582 EVT VT =
Op.getValueType();
1585 unsigned Depth =
Op.getConstantOperandVal(0);
1586 int GRLenInBytes = Subtarget.
getGRLen() / 8;
1589 int Offset = -(GRLenInBytes * 2);
1604 if (
Op.getConstantOperandVal(0) != 0) {
1606 "return address can only be determined for the current frame");
1640 const Value *SV = cast<SrcValueSDNode>(
Op.getOperand(2))->getValue();
1648 !Subtarget.hasBasicD() &&
"unexpected target features");
1653 auto *
C = dyn_cast<ConstantSDNode>(Op0.
getOperand(1));
1654 if (
C &&
C->getZExtValue() < UINT64_C(0xFFFFFFFF))
1664 dyn_cast<VTSDNode>(Op0.
getOperand(1))->getVT().bitsLT(MVT::i32))
1668 EVT RetVT =
Op.getValueType();
1670 MakeLibCallOptions CallOptions;
1671 CallOptions.setTypeListBeforeSoften(OpVT, RetVT,
true);
1674 std::tie(Result, Chain) =
1682 !Subtarget.hasBasicD() &&
"unexpected target features");
1689 dyn_cast<VTSDNode>(Op0.
getOperand(1))->getVT().bitsLE(MVT::i32))
1693 EVT RetVT =
Op.getValueType();
1695 MakeLibCallOptions CallOptions;
1696 CallOptions.setTypeListBeforeSoften(OpVT, RetVT,
true);
1699 std::tie(Result, Chain) =
1710 if (
Op.getValueType() == MVT::f32 && Op0.
getValueType() == MVT::i32 &&
1711 Subtarget.
is64Bit() && Subtarget.hasBasicF()) {
1727 if (
Op.getValueSizeInBits() > 32 && Subtarget.hasBasicF() &&
1728 !Subtarget.hasBasicD()) {
1752 N->getOffset(), Flags);
1760template <
class NodeTy>
1763 bool IsLocal)
const {
1774 assert(Subtarget.
is64Bit() &&
"Large code model requires LA64");
1826 return getAddr(cast<BlockAddressSDNode>(
Op), DAG,
1832 return getAddr(cast<JumpTableSDNode>(
Op), DAG,
1838 return getAddr(cast<ConstantPoolSDNode>(
Op), DAG,
1845 assert(
N->getOffset() == 0 &&
"unexpected offset in global node");
1849 if (GV->
isDSOLocal() && isa<GlobalVariable>(GV)) {
1850 if (
auto GCM = dyn_cast<GlobalVariable>(GV)->
getCodeModel())
1859 unsigned Opc,
bool UseGOT,
1877 if (Opc == LoongArch::PseudoLA_TLS_LE && !Large)
1918 Args.push_back(Entry);
1950LoongArchTargetLowering::lowerGlobalTLSAddress(
SDValue Op,
1957 assert((!Large || Subtarget.
is64Bit()) &&
"Large code model requires LA64");
1960 assert(
N->getOffset() == 0 &&
"unexpected offset in global node");
1974 return getDynamicTLSAddr(
N, DAG,
1975 Large ? LoongArch::PseudoLA_TLS_GD_LARGE
1976 : LoongArch::PseudoLA_TLS_GD,
1983 return getDynamicTLSAddr(
N, DAG,
1984 Large ? LoongArch::PseudoLA_TLS_LD_LARGE
1985 : LoongArch::PseudoLA_TLS_LD,
1990 return getStaticTLSAddr(
N, DAG,
1991 Large ? LoongArch::PseudoLA_TLS_IE_LARGE
1992 : LoongArch::PseudoLA_TLS_IE,
1999 return getStaticTLSAddr(
N, DAG, LoongArch::PseudoLA_TLS_LE,
2003 return getTLSDescAddr(
N, DAG,
2004 Large ? LoongArch::PseudoLA_TLS_DESC_LARGE
2005 : LoongArch::PseudoLA_TLS_DESC,
2009template <
unsigned N>
2012 auto *CImm = cast<ConstantSDNode>(
Op->getOperand(ImmOp));
2014 if ((IsSigned && !isInt<N>(CImm->getSExtValue())) ||
2015 (!IsSigned && !isUInt<N>(CImm->getZExtValue()))) {
2017 ": argument out of range.");
2024LoongArchTargetLowering::lowerINTRINSIC_WO_CHAIN(
SDValue Op,
2027 switch (
Op.getConstantOperandVal(0)) {
2030 case Intrinsic::thread_pointer: {
2034 case Intrinsic::loongarch_lsx_vpickve2gr_d:
2035 case Intrinsic::loongarch_lsx_vpickve2gr_du:
2036 case Intrinsic::loongarch_lsx_vreplvei_d:
2037 case Intrinsic::loongarch_lasx_xvrepl128vei_d:
2038 return checkIntrinsicImmArg<1>(
Op, 2, DAG);
2039 case Intrinsic::loongarch_lsx_vreplvei_w:
2040 case Intrinsic::loongarch_lasx_xvrepl128vei_w:
2041 case Intrinsic::loongarch_lasx_xvpickve2gr_d:
2042 case Intrinsic::loongarch_lasx_xvpickve2gr_du:
2043 case Intrinsic::loongarch_lasx_xvpickve_d:
2044 case Intrinsic::loongarch_lasx_xvpickve_d_f:
2045 return checkIntrinsicImmArg<2>(
Op, 2, DAG);
2046 case Intrinsic::loongarch_lasx_xvinsve0_d:
2047 return checkIntrinsicImmArg<2>(
Op, 3, DAG);
2048 case Intrinsic::loongarch_lsx_vsat_b:
2049 case Intrinsic::loongarch_lsx_vsat_bu:
2050 case Intrinsic::loongarch_lsx_vrotri_b:
2051 case Intrinsic::loongarch_lsx_vsllwil_h_b:
2052 case Intrinsic::loongarch_lsx_vsllwil_hu_bu:
2053 case Intrinsic::loongarch_lsx_vsrlri_b:
2054 case Intrinsic::loongarch_lsx_vsrari_b:
2055 case Intrinsic::loongarch_lsx_vreplvei_h:
2056 case Intrinsic::loongarch_lasx_xvsat_b:
2057 case Intrinsic::loongarch_lasx_xvsat_bu:
2058 case Intrinsic::loongarch_lasx_xvrotri_b:
2059 case Intrinsic::loongarch_lasx_xvsllwil_h_b:
2060 case Intrinsic::loongarch_lasx_xvsllwil_hu_bu:
2061 case Intrinsic::loongarch_lasx_xvsrlri_b:
2062 case Intrinsic::loongarch_lasx_xvsrari_b:
2063 case Intrinsic::loongarch_lasx_xvrepl128vei_h:
2064 case Intrinsic::loongarch_lasx_xvpickve_w:
2065 case Intrinsic::loongarch_lasx_xvpickve_w_f:
2066 return checkIntrinsicImmArg<3>(
Op, 2, DAG);
2067 case Intrinsic::loongarch_lasx_xvinsve0_w:
2068 return checkIntrinsicImmArg<3>(
Op, 3, DAG);
2069 case Intrinsic::loongarch_lsx_vsat_h:
2070 case Intrinsic::loongarch_lsx_vsat_hu:
2071 case Intrinsic::loongarch_lsx_vrotri_h:
2072 case Intrinsic::loongarch_lsx_vsllwil_w_h:
2073 case Intrinsic::loongarch_lsx_vsllwil_wu_hu:
2074 case Intrinsic::loongarch_lsx_vsrlri_h:
2075 case Intrinsic::loongarch_lsx_vsrari_h:
2076 case Intrinsic::loongarch_lsx_vreplvei_b:
2077 case Intrinsic::loongarch_lasx_xvsat_h:
2078 case Intrinsic::loongarch_lasx_xvsat_hu:
2079 case Intrinsic::loongarch_lasx_xvrotri_h:
2080 case Intrinsic::loongarch_lasx_xvsllwil_w_h:
2081 case Intrinsic::loongarch_lasx_xvsllwil_wu_hu:
2082 case Intrinsic::loongarch_lasx_xvsrlri_h:
2083 case Intrinsic::loongarch_lasx_xvsrari_h:
2084 case Intrinsic::loongarch_lasx_xvrepl128vei_b:
2085 return checkIntrinsicImmArg<4>(
Op, 2, DAG);
2086 case Intrinsic::loongarch_lsx_vsrlni_b_h:
2087 case Intrinsic::loongarch_lsx_vsrani_b_h:
2088 case Intrinsic::loongarch_lsx_vsrlrni_b_h:
2089 case Intrinsic::loongarch_lsx_vsrarni_b_h:
2090 case Intrinsic::loongarch_lsx_vssrlni_b_h:
2091 case Intrinsic::loongarch_lsx_vssrani_b_h:
2092 case Intrinsic::loongarch_lsx_vssrlni_bu_h:
2093 case Intrinsic::loongarch_lsx_vssrani_bu_h:
2094 case Intrinsic::loongarch_lsx_vssrlrni_b_h:
2095 case Intrinsic::loongarch_lsx_vssrarni_b_h:
2096 case Intrinsic::loongarch_lsx_vssrlrni_bu_h:
2097 case Intrinsic::loongarch_lsx_vssrarni_bu_h:
2098 case Intrinsic::loongarch_lasx_xvsrlni_b_h:
2099 case Intrinsic::loongarch_lasx_xvsrani_b_h:
2100 case Intrinsic::loongarch_lasx_xvsrlrni_b_h:
2101 case Intrinsic::loongarch_lasx_xvsrarni_b_h:
2102 case Intrinsic::loongarch_lasx_xvssrlni_b_h:
2103 case Intrinsic::loongarch_lasx_xvssrani_b_h:
2104 case Intrinsic::loongarch_lasx_xvssrlni_bu_h:
2105 case Intrinsic::loongarch_lasx_xvssrani_bu_h:
2106 case Intrinsic::loongarch_lasx_xvssrlrni_b_h:
2107 case Intrinsic::loongarch_lasx_xvssrarni_b_h:
2108 case Intrinsic::loongarch_lasx_xvssrlrni_bu_h:
2109 case Intrinsic::loongarch_lasx_xvssrarni_bu_h:
2110 return checkIntrinsicImmArg<4>(
Op, 3, DAG);
2111 case Intrinsic::loongarch_lsx_vsat_w:
2112 case Intrinsic::loongarch_lsx_vsat_wu:
2113 case Intrinsic::loongarch_lsx_vrotri_w:
2114 case Intrinsic::loongarch_lsx_vsllwil_d_w:
2115 case Intrinsic::loongarch_lsx_vsllwil_du_wu:
2116 case Intrinsic::loongarch_lsx_vsrlri_w:
2117 case Intrinsic::loongarch_lsx_vsrari_w:
2118 case Intrinsic::loongarch_lsx_vslei_bu:
2119 case Intrinsic::loongarch_lsx_vslei_hu:
2120 case Intrinsic::loongarch_lsx_vslei_wu:
2121 case Intrinsic::loongarch_lsx_vslei_du:
2122 case Intrinsic::loongarch_lsx_vslti_bu:
2123 case Intrinsic::loongarch_lsx_vslti_hu:
2124 case Intrinsic::loongarch_lsx_vslti_wu:
2125 case Intrinsic::loongarch_lsx_vslti_du:
2126 case Intrinsic::loongarch_lsx_vbsll_v:
2127 case Intrinsic::loongarch_lsx_vbsrl_v:
2128 case Intrinsic::loongarch_lasx_xvsat_w:
2129 case Intrinsic::loongarch_lasx_xvsat_wu:
2130 case Intrinsic::loongarch_lasx_xvrotri_w:
2131 case Intrinsic::loongarch_lasx_xvsllwil_d_w:
2132 case Intrinsic::loongarch_lasx_xvsllwil_du_wu:
2133 case Intrinsic::loongarch_lasx_xvsrlri_w:
2134 case Intrinsic::loongarch_lasx_xvsrari_w:
2135 case Intrinsic::loongarch_lasx_xvslei_bu:
2136 case Intrinsic::loongarch_lasx_xvslei_hu:
2137 case Intrinsic::loongarch_lasx_xvslei_wu:
2138 case Intrinsic::loongarch_lasx_xvslei_du:
2139 case Intrinsic::loongarch_lasx_xvslti_bu:
2140 case Intrinsic::loongarch_lasx_xvslti_hu:
2141 case Intrinsic::loongarch_lasx_xvslti_wu:
2142 case Intrinsic::loongarch_lasx_xvslti_du:
2143 case Intrinsic::loongarch_lasx_xvbsll_v:
2144 case Intrinsic::loongarch_lasx_xvbsrl_v:
2145 return checkIntrinsicImmArg<5>(
Op, 2, DAG);
2146 case Intrinsic::loongarch_lsx_vseqi_b:
2147 case Intrinsic::loongarch_lsx_vseqi_h:
2148 case Intrinsic::loongarch_lsx_vseqi_w:
2149 case Intrinsic::loongarch_lsx_vseqi_d:
2150 case Intrinsic::loongarch_lsx_vslei_b:
2151 case Intrinsic::loongarch_lsx_vslei_h:
2152 case Intrinsic::loongarch_lsx_vslei_w:
2153 case Intrinsic::loongarch_lsx_vslei_d:
2154 case Intrinsic::loongarch_lsx_vslti_b:
2155 case Intrinsic::loongarch_lsx_vslti_h:
2156 case Intrinsic::loongarch_lsx_vslti_w:
2157 case Intrinsic::loongarch_lsx_vslti_d:
2158 case Intrinsic::loongarch_lasx_xvseqi_b:
2159 case Intrinsic::loongarch_lasx_xvseqi_h:
2160 case Intrinsic::loongarch_lasx_xvseqi_w:
2161 case Intrinsic::loongarch_lasx_xvseqi_d:
2162 case Intrinsic::loongarch_lasx_xvslei_b:
2163 case Intrinsic::loongarch_lasx_xvslei_h:
2164 case Intrinsic::loongarch_lasx_xvslei_w:
2165 case Intrinsic::loongarch_lasx_xvslei_d:
2166 case Intrinsic::loongarch_lasx_xvslti_b:
2167 case Intrinsic::loongarch_lasx_xvslti_h:
2168 case Intrinsic::loongarch_lasx_xvslti_w:
2169 case Intrinsic::loongarch_lasx_xvslti_d:
2170 return checkIntrinsicImmArg<5>(
Op, 2, DAG,
true);
2171 case Intrinsic::loongarch_lsx_vsrlni_h_w:
2172 case Intrinsic::loongarch_lsx_vsrani_h_w:
2173 case Intrinsic::loongarch_lsx_vsrlrni_h_w:
2174 case Intrinsic::loongarch_lsx_vsrarni_h_w:
2175 case Intrinsic::loongarch_lsx_vssrlni_h_w:
2176 case Intrinsic::loongarch_lsx_vssrani_h_w:
2177 case Intrinsic::loongarch_lsx_vssrlni_hu_w:
2178 case Intrinsic::loongarch_lsx_vssrani_hu_w:
2179 case Intrinsic::loongarch_lsx_vssrlrni_h_w:
2180 case Intrinsic::loongarch_lsx_vssrarni_h_w:
2181 case Intrinsic::loongarch_lsx_vssrlrni_hu_w:
2182 case Intrinsic::loongarch_lsx_vssrarni_hu_w:
2183 case Intrinsic::loongarch_lsx_vfrstpi_b:
2184 case Intrinsic::loongarch_lsx_vfrstpi_h:
2185 case Intrinsic::loongarch_lasx_xvsrlni_h_w:
2186 case Intrinsic::loongarch_lasx_xvsrani_h_w:
2187 case Intrinsic::loongarch_lasx_xvsrlrni_h_w:
2188 case Intrinsic::loongarch_lasx_xvsrarni_h_w:
2189 case Intrinsic::loongarch_lasx_xvssrlni_h_w:
2190 case Intrinsic::loongarch_lasx_xvssrani_h_w:
2191 case Intrinsic::loongarch_lasx_xvssrlni_hu_w:
2192 case Intrinsic::loongarch_lasx_xvssrani_hu_w:
2193 case Intrinsic::loongarch_lasx_xvssrlrni_h_w:
2194 case Intrinsic::loongarch_lasx_xvssrarni_h_w:
2195 case Intrinsic::loongarch_lasx_xvssrlrni_hu_w:
2196 case Intrinsic::loongarch_lasx_xvssrarni_hu_w:
2197 case Intrinsic::loongarch_lasx_xvfrstpi_b:
2198 case Intrinsic::loongarch_lasx_xvfrstpi_h:
2199 return checkIntrinsicImmArg<5>(
Op, 3, DAG);
2200 case Intrinsic::loongarch_lsx_vsat_d:
2201 case Intrinsic::loongarch_lsx_vsat_du:
2202 case Intrinsic::loongarch_lsx_vrotri_d:
2203 case Intrinsic::loongarch_lsx_vsrlri_d:
2204 case Intrinsic::loongarch_lsx_vsrari_d:
2205 case Intrinsic::loongarch_lasx_xvsat_d:
2206 case Intrinsic::loongarch_lasx_xvsat_du:
2207 case Intrinsic::loongarch_lasx_xvrotri_d:
2208 case Intrinsic::loongarch_lasx_xvsrlri_d:
2209 case Intrinsic::loongarch_lasx_xvsrari_d:
2210 return checkIntrinsicImmArg<6>(
Op, 2, DAG);
2211 case Intrinsic::loongarch_lsx_vsrlni_w_d:
2212 case Intrinsic::loongarch_lsx_vsrani_w_d:
2213 case Intrinsic::loongarch_lsx_vsrlrni_w_d:
2214 case Intrinsic::loongarch_lsx_vsrarni_w_d:
2215 case Intrinsic::loongarch_lsx_vssrlni_w_d:
2216 case Intrinsic::loongarch_lsx_vssrani_w_d:
2217 case Intrinsic::loongarch_lsx_vssrlni_wu_d:
2218 case Intrinsic::loongarch_lsx_vssrani_wu_d:
2219 case Intrinsic::loongarch_lsx_vssrlrni_w_d:
2220 case Intrinsic::loongarch_lsx_vssrarni_w_d:
2221 case Intrinsic::loongarch_lsx_vssrlrni_wu_d:
2222 case Intrinsic::loongarch_lsx_vssrarni_wu_d:
2223 case Intrinsic::loongarch_lasx_xvsrlni_w_d:
2224 case Intrinsic::loongarch_lasx_xvsrani_w_d:
2225 case Intrinsic::loongarch_lasx_xvsrlrni_w_d:
2226 case Intrinsic::loongarch_lasx_xvsrarni_w_d:
2227 case Intrinsic::loongarch_lasx_xvssrlni_w_d:
2228 case Intrinsic::loongarch_lasx_xvssrani_w_d:
2229 case Intrinsic::loongarch_lasx_xvssrlni_wu_d:
2230 case Intrinsic::loongarch_lasx_xvssrani_wu_d:
2231 case Intrinsic::loongarch_lasx_xvssrlrni_w_d:
2232 case Intrinsic::loongarch_lasx_xvssrarni_w_d:
2233 case Intrinsic::loongarch_lasx_xvssrlrni_wu_d:
2234 case Intrinsic::loongarch_lasx_xvssrarni_wu_d:
2235 return checkIntrinsicImmArg<6>(
Op, 3, DAG);
2236 case Intrinsic::loongarch_lsx_vsrlni_d_q:
2237 case Intrinsic::loongarch_lsx_vsrani_d_q:
2238 case Intrinsic::loongarch_lsx_vsrlrni_d_q:
2239 case Intrinsic::loongarch_lsx_vsrarni_d_q:
2240 case Intrinsic::loongarch_lsx_vssrlni_d_q:
2241 case Intrinsic::loongarch_lsx_vssrani_d_q:
2242 case Intrinsic::loongarch_lsx_vssrlni_du_q:
2243 case Intrinsic::loongarch_lsx_vssrani_du_q:
2244 case Intrinsic::loongarch_lsx_vssrlrni_d_q:
2245 case Intrinsic::loongarch_lsx_vssrarni_d_q:
2246 case Intrinsic::loongarch_lsx_vssrlrni_du_q:
2247 case Intrinsic::loongarch_lsx_vssrarni_du_q:
2248 case Intrinsic::loongarch_lasx_xvsrlni_d_q:
2249 case Intrinsic::loongarch_lasx_xvsrani_d_q:
2250 case Intrinsic::loongarch_lasx_xvsrlrni_d_q:
2251 case Intrinsic::loongarch_lasx_xvsrarni_d_q:
2252 case Intrinsic::loongarch_lasx_xvssrlni_d_q:
2253 case Intrinsic::loongarch_lasx_xvssrani_d_q:
2254 case Intrinsic::loongarch_lasx_xvssrlni_du_q:
2255 case Intrinsic::loongarch_lasx_xvssrani_du_q:
2256 case Intrinsic::loongarch_lasx_xvssrlrni_d_q:
2257 case Intrinsic::loongarch_lasx_xvssrarni_d_q:
2258 case Intrinsic::loongarch_lasx_xvssrlrni_du_q:
2259 case Intrinsic::loongarch_lasx_xvssrarni_du_q:
2260 return checkIntrinsicImmArg<7>(
Op, 3, DAG);
2261 case Intrinsic::loongarch_lsx_vnori_b:
2262 case Intrinsic::loongarch_lsx_vshuf4i_b:
2263 case Intrinsic::loongarch_lsx_vshuf4i_h:
2264 case Intrinsic::loongarch_lsx_vshuf4i_w:
2265 case Intrinsic::loongarch_lasx_xvnori_b:
2266 case Intrinsic::loongarch_lasx_xvshuf4i_b:
2267 case Intrinsic::loongarch_lasx_xvshuf4i_h:
2268 case Intrinsic::loongarch_lasx_xvshuf4i_w:
2269 case Intrinsic::loongarch_lasx_xvpermi_d:
2270 return checkIntrinsicImmArg<8>(
Op, 2, DAG);
2271 case Intrinsic::loongarch_lsx_vshuf4i_d:
2272 case Intrinsic::loongarch_lsx_vpermi_w:
2273 case Intrinsic::loongarch_lsx_vbitseli_b:
2274 case Intrinsic::loongarch_lsx_vextrins_b:
2275 case Intrinsic::loongarch_lsx_vextrins_h:
2276 case Intrinsic::loongarch_lsx_vextrins_w:
2277 case Intrinsic::loongarch_lsx_vextrins_d:
2278 case Intrinsic::loongarch_lasx_xvshuf4i_d:
2279 case Intrinsic::loongarch_lasx_xvpermi_w:
2280 case Intrinsic::loongarch_lasx_xvpermi_q:
2281 case Intrinsic::loongarch_lasx_xvbitseli_b:
2282 case Intrinsic::loongarch_lasx_xvextrins_b:
2283 case Intrinsic::loongarch_lasx_xvextrins_h:
2284 case Intrinsic::loongarch_lasx_xvextrins_w:
2285 case Intrinsic::loongarch_lasx_xvextrins_d:
2286 return checkIntrinsicImmArg<8>(
Op, 3, DAG);
2287 case Intrinsic::loongarch_lsx_vrepli_b:
2288 case Intrinsic::loongarch_lsx_vrepli_h:
2289 case Intrinsic::loongarch_lsx_vrepli_w:
2290 case Intrinsic::loongarch_lsx_vrepli_d:
2291 case Intrinsic::loongarch_lasx_xvrepli_b:
2292 case Intrinsic::loongarch_lasx_xvrepli_h:
2293 case Intrinsic::loongarch_lasx_xvrepli_w:
2294 case Intrinsic::loongarch_lasx_xvrepli_d:
2295 return checkIntrinsicImmArg<10>(
Op, 1, DAG,
true);
2296 case Intrinsic::loongarch_lsx_vldi:
2297 case Intrinsic::loongarch_lasx_xvldi:
2298 return checkIntrinsicImmArg<13>(
Op, 1, DAG,
true);
2313LoongArchTargetLowering::lowerINTRINSIC_W_CHAIN(
SDValue Op,
2317 EVT VT =
Op.getValueType();
2319 const StringRef ErrorMsgOOR =
"argument out of range";
2320 const StringRef ErrorMsgReqLA64 =
"requires loongarch64";
2321 const StringRef ErrorMsgReqF =
"requires basic 'f' target feature";
2323 switch (
Op.getConstantOperandVal(1)) {
2326 case Intrinsic::loongarch_crc_w_b_w:
2327 case Intrinsic::loongarch_crc_w_h_w:
2328 case Intrinsic::loongarch_crc_w_w_w:
2329 case Intrinsic::loongarch_crc_w_d_w:
2330 case Intrinsic::loongarch_crcc_w_b_w:
2331 case Intrinsic::loongarch_crcc_w_h_w:
2332 case Intrinsic::loongarch_crcc_w_w_w:
2333 case Intrinsic::loongarch_crcc_w_d_w:
2335 case Intrinsic::loongarch_csrrd_w:
2336 case Intrinsic::loongarch_csrrd_d: {
2337 unsigned Imm =
Op.getConstantOperandVal(2);
2338 return !isUInt<14>(Imm)
2343 case Intrinsic::loongarch_csrwr_w:
2344 case Intrinsic::loongarch_csrwr_d: {
2345 unsigned Imm =
Op.getConstantOperandVal(3);
2346 return !isUInt<14>(Imm)
2349 {Chain,
Op.getOperand(2),
2352 case Intrinsic::loongarch_csrxchg_w:
2353 case Intrinsic::loongarch_csrxchg_d: {
2354 unsigned Imm =
Op.getConstantOperandVal(4);
2355 return !isUInt<14>(Imm)
2358 {Chain,
Op.getOperand(2),
Op.getOperand(3),
2361 case Intrinsic::loongarch_iocsrrd_d: {
2366#define IOCSRRD_CASE(NAME, NODE) \
2367 case Intrinsic::loongarch_##NAME: { \
2368 return DAG.getNode(LoongArchISD::NODE, DL, {GRLenVT, MVT::Other}, \
2369 {Chain, Op.getOperand(2)}); \
2375 case Intrinsic::loongarch_cpucfg: {
2377 {Chain,
Op.getOperand(2)});
2379 case Intrinsic::loongarch_lddir_d: {
2380 unsigned Imm =
Op.getConstantOperandVal(3);
2381 return !isUInt<8>(Imm)
2385 case Intrinsic::loongarch_movfcsr2gr: {
2386 if (!Subtarget.hasBasicF())
2388 unsigned Imm =
Op.getConstantOperandVal(2);
2389 return !isUInt<2>(Imm)
2394 case Intrinsic::loongarch_lsx_vld:
2395 case Intrinsic::loongarch_lsx_vldrepl_b:
2396 case Intrinsic::loongarch_lasx_xvld:
2397 case Intrinsic::loongarch_lasx_xvldrepl_b:
2398 return !isInt<12>(cast<ConstantSDNode>(
Op.getOperand(3))->getSExtValue())
2401 case Intrinsic::loongarch_lsx_vldrepl_h:
2402 case Intrinsic::loongarch_lasx_xvldrepl_h:
2403 return !isShiftedInt<11, 1>(
2404 cast<ConstantSDNode>(
Op.getOperand(3))->getSExtValue())
2406 Op,
"argument out of range or not a multiple of 2", DAG)
2408 case Intrinsic::loongarch_lsx_vldrepl_w:
2409 case Intrinsic::loongarch_lasx_xvldrepl_w:
2410 return !isShiftedInt<10, 2>(
2411 cast<ConstantSDNode>(
Op.getOperand(3))->getSExtValue())
2413 Op,
"argument out of range or not a multiple of 4", DAG)
2415 case Intrinsic::loongarch_lsx_vldrepl_d:
2416 case Intrinsic::loongarch_lasx_xvldrepl_d:
2417 return !isShiftedInt<9, 3>(
2418 cast<ConstantSDNode>(
Op.getOperand(3))->getSExtValue())
2420 Op,
"argument out of range or not a multiple of 8", DAG)
2431 return Op.getOperand(0);
2439 uint64_t IntrinsicEnum =
Op.getConstantOperandVal(1);
2441 const StringRef ErrorMsgOOR =
"argument out of range";
2442 const StringRef ErrorMsgReqLA64 =
"requires loongarch64";
2443 const StringRef ErrorMsgReqLA32 =
"requires loongarch32";
2444 const StringRef ErrorMsgReqF =
"requires basic 'f' target feature";
2446 switch (IntrinsicEnum) {
2450 case Intrinsic::loongarch_cacop_d:
2451 case Intrinsic::loongarch_cacop_w: {
2452 if (IntrinsicEnum == Intrinsic::loongarch_cacop_d && !Subtarget.
is64Bit())
2454 if (IntrinsicEnum == Intrinsic::loongarch_cacop_w && Subtarget.
is64Bit())
2458 int Imm2 = cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue();
2459 if (!isUInt<5>(Imm1) || !isInt<12>(Imm2))
2463 case Intrinsic::loongarch_dbar: {
2465 return !isUInt<15>(Imm)
2470 case Intrinsic::loongarch_ibar: {
2472 return !isUInt<15>(Imm)
2477 case Intrinsic::loongarch_break: {
2479 return !isUInt<15>(Imm)
2484 case Intrinsic::loongarch_movgr2fcsr: {
2485 if (!Subtarget.hasBasicF())
2488 return !isUInt<2>(Imm)
2495 case Intrinsic::loongarch_syscall: {
2497 return !isUInt<15>(Imm)
2502#define IOCSRWR_CASE(NAME, NODE) \
2503 case Intrinsic::loongarch_##NAME: { \
2504 SDValue Op3 = Op.getOperand(3); \
2505 return Subtarget.is64Bit() \
2506 ? DAG.getNode(LoongArchISD::NODE, DL, MVT::Other, Chain, \
2507 DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op2), \
2508 DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op3)) \
2509 : DAG.getNode(LoongArchISD::NODE, DL, MVT::Other, Chain, Op2, \
2516 case Intrinsic::loongarch_iocsrwr_d: {
2524#define ASRT_LE_GT_CASE(NAME) \
2525 case Intrinsic::loongarch_##NAME: { \
2526 return !Subtarget.is64Bit() \
2527 ? emitIntrinsicErrorMessage(Op, ErrorMsgReqLA64, DAG) \
2532#undef ASRT_LE_GT_CASE
2533 case Intrinsic::loongarch_ldpte_d: {
2534 unsigned Imm =
Op.getConstantOperandVal(3);
2540 case Intrinsic::loongarch_lsx_vst:
2541 case Intrinsic::loongarch_lasx_xvst:
2542 return !isInt<12>(cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue())
2545 case Intrinsic::loongarch_lasx_xvstelm_b:
2546 return (!isInt<8>(cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2547 !isUInt<5>(
Op.getConstantOperandVal(5)))
2550 case Intrinsic::loongarch_lsx_vstelm_b:
2551 return (!isInt<8>(cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2552 !isUInt<4>(
Op.getConstantOperandVal(5)))
2555 case Intrinsic::loongarch_lasx_xvstelm_h:
2556 return (!isShiftedInt<8, 1>(
2557 cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2558 !isUInt<4>(
Op.getConstantOperandVal(5)))
2560 Op,
"argument out of range or not a multiple of 2", DAG)
2562 case Intrinsic::loongarch_lsx_vstelm_h:
2563 return (!isShiftedInt<8, 1>(
2564 cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2565 !isUInt<3>(
Op.getConstantOperandVal(5)))
2567 Op,
"argument out of range or not a multiple of 2", DAG)
2569 case Intrinsic::loongarch_lasx_xvstelm_w:
2570 return (!isShiftedInt<8, 2>(
2571 cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2572 !isUInt<3>(
Op.getConstantOperandVal(5)))
2574 Op,
"argument out of range or not a multiple of 4", DAG)
2576 case Intrinsic::loongarch_lsx_vstelm_w:
2577 return (!isShiftedInt<8, 2>(
2578 cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2579 !isUInt<2>(
Op.getConstantOperandVal(5)))
2581 Op,
"argument out of range or not a multiple of 4", DAG)
2583 case Intrinsic::loongarch_lasx_xvstelm_d:
2584 return (!isShiftedInt<8, 3>(
2585 cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2586 !isUInt<2>(
Op.getConstantOperandVal(5)))
2588 Op,
"argument out of range or not a multiple of 8", DAG)
2590 case Intrinsic::loongarch_lsx_vstelm_d:
2591 return (!isShiftedInt<8, 3>(
2592 cast<ConstantSDNode>(
Op.getOperand(4))->getSExtValue()) ||
2593 !isUInt<1>(
Op.getConstantOperandVal(5)))
2595 Op,
"argument out of range or not a multiple of 8", DAG)
2606 EVT VT =
Lo.getValueType();
2647 EVT VT =
Lo.getValueType();
2739 NewOp0 = DAG.
getNode(ExtOpc,
DL, MVT::i64,
N->getOperand(0));
2740 NewRes = DAG.
getNode(WOpcode,
DL, MVT::i64, NewOp0);
2744 NewOp0 = DAG.
getNode(ExtOpc,
DL, MVT::i64,
N->getOperand(0));
2750 NewRes = DAG.
getNode(WOpcode,
DL, MVT::i64, NewOp0, NewOp1);
2777 StringRef ErrorMsg,
bool WithChain =
true) {
2782 Results.push_back(
N->getOperand(0));
2785template <
unsigned N>
2790 const StringRef ErrorMsgOOR =
"argument out of range";
2791 unsigned Imm =
Node->getConstantOperandVal(2);
2792 if (!isUInt<N>(Imm)) {
2825 switch (
N->getConstantOperandVal(0)) {
2828 case Intrinsic::loongarch_lsx_vpickve2gr_b:
2829 replaceVPICKVE2GRResults<4>(
N,
Results, DAG, Subtarget,
2832 case Intrinsic::loongarch_lsx_vpickve2gr_h:
2833 case Intrinsic::loongarch_lasx_xvpickve2gr_w:
2834 replaceVPICKVE2GRResults<3>(
N,
Results, DAG, Subtarget,
2837 case Intrinsic::loongarch_lsx_vpickve2gr_w:
2838 replaceVPICKVE2GRResults<2>(
N,
Results, DAG, Subtarget,
2841 case Intrinsic::loongarch_lsx_vpickve2gr_bu:
2842 replaceVPICKVE2GRResults<4>(
N,
Results, DAG, Subtarget,
2845 case Intrinsic::loongarch_lsx_vpickve2gr_hu:
2846 case Intrinsic::loongarch_lasx_xvpickve2gr_wu:
2847 replaceVPICKVE2GRResults<3>(
N,
Results, DAG, Subtarget,
2850 case Intrinsic::loongarch_lsx_vpickve2gr_wu:
2851 replaceVPICKVE2GRResults<2>(
N,
Results, DAG, Subtarget,
2854 case Intrinsic::loongarch_lsx_bz_b:
2855 case Intrinsic::loongarch_lsx_bz_h:
2856 case Intrinsic::loongarch_lsx_bz_w:
2857 case Intrinsic::loongarch_lsx_bz_d:
2858 case Intrinsic::loongarch_lasx_xbz_b:
2859 case Intrinsic::loongarch_lasx_xbz_h:
2860 case Intrinsic::loongarch_lasx_xbz_w:
2861 case Intrinsic::loongarch_lasx_xbz_d:
2865 case Intrinsic::loongarch_lsx_bz_v:
2866 case Intrinsic::loongarch_lasx_xbz_v:
2870 case Intrinsic::loongarch_lsx_bnz_b:
2871 case Intrinsic::loongarch_lsx_bnz_h:
2872 case Intrinsic::loongarch_lsx_bnz_w:
2873 case Intrinsic::loongarch_lsx_bnz_d:
2874 case Intrinsic::loongarch_lasx_xbnz_b:
2875 case Intrinsic::loongarch_lasx_xbnz_h:
2876 case Intrinsic::loongarch_lasx_xbnz_w:
2877 case Intrinsic::loongarch_lasx_xbnz_d:
2881 case Intrinsic::loongarch_lsx_bnz_v:
2882 case Intrinsic::loongarch_lasx_xbnz_v:
2892 EVT VT =
N->getValueType(0);
2893 switch (
N->getOpcode()) {
2899 "Unexpected custom legalisation");
2907 "Unexpected custom legalisation");
2909 Subtarget.hasDiv32() && VT == MVT::i32
2917 "Unexpected custom legalisation");
2926 "Unexpected custom legalisation");
2931 "Unexpected custom legalisation");
2936 if (Src.getValueType() == MVT::f16)
2947 EVT OpVT = Src.getValueType();
2951 std::tie(Result, Chain) =
2958 EVT SrcVT = Src.getValueType();
2959 if (VT == MVT::i32 && SrcVT == MVT::f32 && Subtarget.
is64Bit() &&
2960 Subtarget.hasBasicF()) {
2969 "Unexpected custom legalisation");
2972 TLI.expandFP_TO_UINT(
N, Tmp1, Tmp2, DAG);
2978 assert((VT == MVT::i16 || VT == MVT::i32) &&
2979 "Unexpected custom legalization");
3000 assert((VT == MVT::i8 || (VT == MVT::i32 && Subtarget.
is64Bit())) &&
3001 "Unexpected custom legalization");
3021 "Unexpected custom legalisation");
3029 const StringRef ErrorMsgOOR =
"argument out of range";
3030 const StringRef ErrorMsgReqLA64 =
"requires loongarch64";
3031 const StringRef ErrorMsgReqF =
"requires basic 'f' target feature";
3033 switch (
N->getConstantOperandVal(1)) {
3036 case Intrinsic::loongarch_movfcsr2gr: {
3037 if (!Subtarget.hasBasicF()) {
3042 if (!isUInt<2>(Imm)) {
3054#define CRC_CASE_EXT_BINARYOP(NAME, NODE) \
3055 case Intrinsic::loongarch_##NAME: { \
3056 SDValue NODE = DAG.getNode( \
3057 LoongArchISD::NODE, DL, {MVT::i64, MVT::Other}, \
3058 {Chain, DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op2), \
3059 DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, N->getOperand(3))}); \
3060 Results.push_back(DAG.getNode(ISD::TRUNCATE, DL, VT, NODE.getValue(0))); \
3061 Results.push_back(NODE.getValue(1)); \
3070#undef CRC_CASE_EXT_BINARYOP
3072#define CRC_CASE_EXT_UNARYOP(NAME, NODE) \
3073 case Intrinsic::loongarch_##NAME: { \
3074 SDValue NODE = DAG.getNode( \
3075 LoongArchISD::NODE, DL, {MVT::i64, MVT::Other}, \
3077 DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, N->getOperand(3))}); \
3078 Results.push_back(DAG.getNode(ISD::TRUNCATE, DL, VT, NODE.getValue(0))); \
3079 Results.push_back(NODE.getValue(1)); \
3084#undef CRC_CASE_EXT_UNARYOP
3085#define CSR_CASE(ID) \
3086 case Intrinsic::loongarch_##ID: { \
3087 if (!Subtarget.is64Bit()) \
3088 emitErrorAndReplaceIntrinsicResults(N, Results, DAG, ErrorMsgReqLA64); \
3096 case Intrinsic::loongarch_csrrd_w: {
3098 if (!isUInt<14>(Imm)) {
3110 case Intrinsic::loongarch_csrwr_w: {
3111 unsigned Imm =
N->getConstantOperandVal(3);
3112 if (!isUInt<14>(Imm)) {
3125 case Intrinsic::loongarch_csrxchg_w: {
3126 unsigned Imm =
N->getConstantOperandVal(4);
3127 if (!isUInt<14>(Imm)) {
3141#define IOCSRRD_CASE(NAME, NODE) \
3142 case Intrinsic::loongarch_##NAME: { \
3143 SDValue IOCSRRDResults = \
3144 DAG.getNode(LoongArchISD::NODE, DL, {MVT::i64, MVT::Other}, \
3145 {Chain, DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op2)}); \
3146 Results.push_back( \
3147 DAG.getNode(ISD::TRUNCATE, DL, VT, IOCSRRDResults.getValue(0))); \
3148 Results.push_back(IOCSRRDResults.getValue(1)); \
3155 case Intrinsic::loongarch_cpucfg: {
3164 case Intrinsic::loongarch_lddir_d: {
3177 "On LA64, only 64-bit registers can be read.");
3180 "On LA32, only 32-bit registers can be read.");
3182 Results.push_back(
N->getOperand(0));
3193 OpVT == MVT::f64 ? RTLIB::LROUND_F64 : RTLIB::LROUND_F32;
3210 SDValue FirstOperand =
N->getOperand(0);
3211 SDValue SecondOperand =
N->getOperand(1);
3212 unsigned FirstOperandOpc = FirstOperand.
getOpcode();
3213 EVT ValTy =
N->getValueType(0);
3216 unsigned SMIdx, SMLen;
3222 if (!(CN = dyn_cast<ConstantSDNode>(SecondOperand)) ||
3233 if (!(CN = dyn_cast<ConstantSDNode>(FirstOperand.
getOperand(1))))
3274 NewOperand = FirstOperand;
3277 msb = lsb + SMLen - 1;
3281 if (FirstOperandOpc ==
ISD::SRA || FirstOperandOpc ==
ISD::SRL || lsb == 0)
3302 SDValue FirstOperand =
N->getOperand(0);
3304 EVT ValTy =
N->getValueType(0);
3307 unsigned MaskIdx, MaskLen;
3313 !(CN = dyn_cast<ConstantSDNode>(FirstOperand.
getOperand(1))) ||
3318 if (!(CN = dyn_cast<ConstantSDNode>(
N->getOperand(1))))
3322 if (MaskIdx <= Shamt && Shamt <= MaskIdx + MaskLen - 1)
3335 EVT ValTy =
N->getValueType(0);
3336 SDValue N0 =
N->getOperand(0), N1 =
N->getOperand(1);
3340 unsigned MaskIdx0, MaskLen0, MaskIdx1, MaskLen1;
3342 bool SwapAndRetried =
false;
3347 if (ValBits != 32 && ValBits != 64)
3357 (CN0 = dyn_cast<ConstantSDNode>(N0.
getOperand(1))) &&
3360 (CN1 = dyn_cast<ConstantSDNode>(N1.getOperand(1))) &&
3362 MaskIdx0 == MaskIdx1 && MaskLen0 == MaskLen1 &&
3363 (CN1 = dyn_cast<ConstantSDNode>(N1.getOperand(0).getOperand(1))) &&
3365 (MaskIdx0 + MaskLen0 <= ValBits)) {
3379 (CN0 = dyn_cast<ConstantSDNode>(N0.
getOperand(1))) &&
3382 (CN1 = dyn_cast<ConstantSDNode>(N1.getOperand(1))) &&
3384 (CN1 = dyn_cast<ConstantSDNode>(N1.getOperand(0).getOperand(1))) &&
3386 MaskLen0 == MaskLen1 && MaskIdx1 == 0 &&
3387 (MaskIdx0 + MaskLen0 <= ValBits)) {
3402 (CN0 = dyn_cast<ConstantSDNode>(N0.
getOperand(1))) &&
3404 (MaskIdx0 + MaskLen0 <= 64) &&
3405 (CN1 = dyn_cast<ConstantSDNode>(N1->getOperand(1))) &&
3412 ? (MaskIdx0 + (MaskLen0 & 31) - 1)
3413 : (MaskIdx0 + MaskLen0 - 1),
3425 (CN0 = dyn_cast<ConstantSDNode>(N0.
getOperand(1))) &&
3427 MaskIdx0 == 0 && (CN1 = dyn_cast<ConstantSDNode>(N1.getOperand(1))) &&
3429 (MaskIdx0 + MaskLen0 <= ValBits)) {
3444 (CN0 = dyn_cast<ConstantSDNode>(N0.
getOperand(1))) &&
3446 (CN1 = dyn_cast<ConstantSDNode>(N1)) &&
3452 DAG.
getConstant(ValBits == 32 ? (MaskIdx0 + (MaskLen0 & 31) - 1)
3453 : (MaskIdx0 + MaskLen0 - 1),
3468 unsigned MaskIdx, MaskLen;
3469 if (N1.getOpcode() ==
ISD::SHL && N1.getOperand(0).getOpcode() ==
ISD::AND &&
3470 (CNMask = dyn_cast<ConstantSDNode>(N1.getOperand(0).getOperand(1))) &&
3472 MaskIdx == 0 && (CNShamt = dyn_cast<ConstantSDNode>(N1.getOperand(1))) &&
3494 (CNMask = dyn_cast<ConstantSDNode>(N1.getOperand(1))) &&
3496 N1.getOperand(0).getOpcode() ==
ISD::SHL &&
3497 (CNShamt = dyn_cast<ConstantSDNode>(N1.getOperand(0).getOperand(1))) &&
3510 if (!SwapAndRetried) {
3512 SwapAndRetried =
true;
3516 SwapAndRetried =
false;
3528 (CNMask = dyn_cast<ConstantSDNode>(N1.getOperand(1))) &&
3542 if (!SwapAndRetried) {
3544 SwapAndRetried =
true;
3554 switch (V.getNode()->getOpcode()) {
3556 LoadSDNode *LoadNode = cast<LoadSDNode>(V.getNode());
3565 VTSDNode *TypeNode = cast<VTSDNode>(V.getNode()->getOperand(1));
3566 if ((TypeNode->
getVT() == MVT::i8) || (TypeNode->
getVT() == MVT::i16)) {
3573 VTSDNode *TypeNode = cast<VTSDNode>(V.getNode()->getOperand(1));
3574 if ((TypeNode->
getVT() == MVT::i8) || (TypeNode->
getVT() == MVT::i16)) {
3651 SDNode *AndNode =
N->getOperand(0).getNode();
3659 SDValue CmpInputValue =
N->getOperand(1);
3667 CN = dyn_cast<ConstantSDNode>(CmpInputValue);
3670 AndInputValue1 = AndInputValue1.
getOperand(0);
3674 if (AndInputValue2 != CmpInputValue)
3707 TruncInputValue1, TruncInputValue2);
3729template <
unsigned N>
3733 bool IsSigned =
false) {
3735 auto *CImm = cast<ConstantSDNode>(
Node->getOperand(ImmOp));
3737 if ((IsSigned && !isInt<N>(CImm->getSExtValue())) ||
3738 (!IsSigned && !isUInt<N>(CImm->getZExtValue()))) {
3740 ": argument out of range.");
3746template <
unsigned N>
3750 EVT ResTy =
Node->getValueType(0);
3751 auto *CImm = cast<ConstantSDNode>(
Node->getOperand(ImmOp));
3754 if ((IsSigned && !isInt<N>(CImm->getSExtValue())) ||
3755 (!IsSigned && !isUInt<N>(CImm->getZExtValue()))) {
3757 ": argument out of range.");
3762 IsSigned ? CImm->getSExtValue() : CImm->getZExtValue(), IsSigned),
3768 EVT ResTy =
Node->getValueType(0);
3776 EVT ResTy =
Node->getValueType(0);
3785template <
unsigned N>
3788 EVT ResTy =
Node->getValueType(0);
3789 auto *CImm = cast<ConstantSDNode>(
Node->getOperand(2));
3791 if (!isUInt<N>(CImm->getZExtValue())) {
3793 ": argument out of range.");
3803template <
unsigned N>
3806 EVT ResTy =
Node->getValueType(0);
3807 auto *CImm = cast<ConstantSDNode>(
Node->getOperand(2));
3809 if (!isUInt<N>(CImm->getZExtValue())) {
3811 ": argument out of range.");
3820template <
unsigned N>
3823 EVT ResTy =
Node->getValueType(0);
3824 auto *CImm = cast<ConstantSDNode>(
Node->getOperand(2));
3826 if (!isUInt<N>(CImm->getZExtValue())) {
3828 ": argument out of range.");
3842 switch (
N->getConstantOperandVal(0)) {
3845 case Intrinsic::loongarch_lsx_vadd_b:
3846 case Intrinsic::loongarch_lsx_vadd_h:
3847 case Intrinsic::loongarch_lsx_vadd_w:
3848 case Intrinsic::loongarch_lsx_vadd_d:
3849 case Intrinsic::loongarch_lasx_xvadd_b:
3850 case Intrinsic::loongarch_lasx_xvadd_h:
3851 case Intrinsic::loongarch_lasx_xvadd_w:
3852 case Intrinsic::loongarch_lasx_xvadd_d:
3855 case Intrinsic::loongarch_lsx_vaddi_bu:
3856 case Intrinsic::loongarch_lsx_vaddi_hu:
3857 case Intrinsic::loongarch_lsx_vaddi_wu:
3858 case Intrinsic::loongarch_lsx_vaddi_du:
3859 case Intrinsic::loongarch_lasx_xvaddi_bu:
3860 case Intrinsic::loongarch_lasx_xvaddi_hu:
3861 case Intrinsic::loongarch_lasx_xvaddi_wu:
3862 case Intrinsic::loongarch_lasx_xvaddi_du:
3864 lowerVectorSplatImm<5>(
N, 2, DAG));
3865 case Intrinsic::loongarch_lsx_vsub_b:
3866 case Intrinsic::loongarch_lsx_vsub_h:
3867 case Intrinsic::loongarch_lsx_vsub_w:
3868 case Intrinsic::loongarch_lsx_vsub_d:
3869 case Intrinsic::loongarch_lasx_xvsub_b:
3870 case Intrinsic::loongarch_lasx_xvsub_h:
3871 case Intrinsic::loongarch_lasx_xvsub_w:
3872 case Intrinsic::loongarch_lasx_xvsub_d:
3875 case Intrinsic::loongarch_lsx_vsubi_bu:
3876 case Intrinsic::loongarch_lsx_vsubi_hu:
3877 case Intrinsic::loongarch_lsx_vsubi_wu:
3878 case Intrinsic::loongarch_lsx_vsubi_du:
3879 case Intrinsic::loongarch_lasx_xvsubi_bu:
3880 case Intrinsic::loongarch_lasx_xvsubi_hu:
3881 case Intrinsic::loongarch_lasx_xvsubi_wu:
3882 case Intrinsic::loongarch_lasx_xvsubi_du:
3884 lowerVectorSplatImm<5>(
N, 2, DAG));
3885 case Intrinsic::loongarch_lsx_vneg_b:
3886 case Intrinsic::loongarch_lsx_vneg_h:
3887 case Intrinsic::loongarch_lsx_vneg_w:
3888 case Intrinsic::loongarch_lsx_vneg_d:
3889 case Intrinsic::loongarch_lasx_xvneg_b:
3890 case Intrinsic::loongarch_lasx_xvneg_h:
3891 case Intrinsic::loongarch_lasx_xvneg_w:
3892 case Intrinsic::loongarch_lasx_xvneg_d:
3896 APInt(
N->getValueType(0).getScalarType().getSizeInBits(), 0,
3898 SDLoc(
N),
N->getValueType(0)),
3900 case Intrinsic::loongarch_lsx_vmax_b:
3901 case Intrinsic::loongarch_lsx_vmax_h:
3902 case Intrinsic::loongarch_lsx_vmax_w:
3903 case Intrinsic::loongarch_lsx_vmax_d:
3904 case Intrinsic::loongarch_lasx_xvmax_b:
3905 case Intrinsic::loongarch_lasx_xvmax_h:
3906 case Intrinsic::loongarch_lasx_xvmax_w:
3907 case Intrinsic::loongarch_lasx_xvmax_d:
3910 case Intrinsic::loongarch_lsx_vmax_bu:
3911 case Intrinsic::loongarch_lsx_vmax_hu:
3912 case Intrinsic::loongarch_lsx_vmax_wu:
3913 case Intrinsic::loongarch_lsx_vmax_du:
3914 case Intrinsic::loongarch_lasx_xvmax_bu:
3915 case Intrinsic::loongarch_lasx_xvmax_hu:
3916 case Intrinsic::loongarch_lasx_xvmax_wu:
3917 case Intrinsic::loongarch_lasx_xvmax_du:
3920 case Intrinsic::loongarch_lsx_vmaxi_b:
3921 case Intrinsic::loongarch_lsx_vmaxi_h:
3922 case Intrinsic::loongarch_lsx_vmaxi_w:
3923 case Intrinsic::loongarch_lsx_vmaxi_d:
3924 case Intrinsic::loongarch_lasx_xvmaxi_b:
3925 case Intrinsic::loongarch_lasx_xvmaxi_h:
3926 case Intrinsic::loongarch_lasx_xvmaxi_w:
3927 case Intrinsic::loongarch_lasx_xvmaxi_d:
3929 lowerVectorSplatImm<5>(
N, 2, DAG,
true));
3930 case Intrinsic::loongarch_lsx_vmaxi_bu:
3931 case Intrinsic::loongarch_lsx_vmaxi_hu:
3932 case Intrinsic::loongarch_lsx_vmaxi_wu:
3933 case Intrinsic::loongarch_lsx_vmaxi_du:
3934 case Intrinsic::loongarch_lasx_xvmaxi_bu:
3935 case Intrinsic::loongarch_lasx_xvmaxi_hu:
3936 case Intrinsic::loongarch_lasx_xvmaxi_wu:
3937 case Intrinsic::loongarch_lasx_xvmaxi_du:
3939 lowerVectorSplatImm<5>(
N, 2, DAG));
3940 case Intrinsic::loongarch_lsx_vmin_b:
3941 case Intrinsic::loongarch_lsx_vmin_h:
3942 case Intrinsic::loongarch_lsx_vmin_w:
3943 case Intrinsic::loongarch_lsx_vmin_d:
3944 case Intrinsic::loongarch_lasx_xvmin_b:
3945 case Intrinsic::loongarch_lasx_xvmin_h:
3946 case Intrinsic::loongarch_lasx_xvmin_w:
3947 case Intrinsic::loongarch_lasx_xvmin_d:
3950 case Intrinsic::loongarch_lsx_vmin_bu:
3951 case Intrinsic::loongarch_lsx_vmin_hu:
3952 case Intrinsic::loongarch_lsx_vmin_wu:
3953 case Intrinsic::loongarch_lsx_vmin_du:
3954 case Intrinsic::loongarch_lasx_xvmin_bu:
3955 case Intrinsic::loongarch_lasx_xvmin_hu:
3956 case Intrinsic::loongarch_lasx_xvmin_wu:
3957 case Intrinsic::loongarch_lasx_xvmin_du:
3960 case Intrinsic::loongarch_lsx_vmini_b:
3961 case Intrinsic::loongarch_lsx_vmini_h:
3962 case Intrinsic::loongarch_lsx_vmini_w:
3963 case Intrinsic::loongarch_lsx_vmini_d:
3964 case Intrinsic::loongarch_lasx_xvmini_b:
3965 case Intrinsic::loongarch_lasx_xvmini_h:
3966 case Intrinsic::loongarch_lasx_xvmini_w:
3967 case Intrinsic::loongarch_lasx_xvmini_d:
3969 lowerVectorSplatImm<5>(
N, 2, DAG,
true));
3970 case Intrinsic::loongarch_lsx_vmini_bu:
3971 case Intrinsic::loongarch_lsx_vmini_hu:
3972 case Intrinsic::loongarch_lsx_vmini_wu:
3973 case Intrinsic::loongarch_lsx_vmini_du:
3974 case Intrinsic::loongarch_lasx_xvmini_bu:
3975 case Intrinsic::loongarch_lasx_xvmini_hu:
3976 case Intrinsic::loongarch_lasx_xvmini_wu:
3977 case Intrinsic::loongarch_lasx_xvmini_du:
3979 lowerVectorSplatImm<5>(
N, 2, DAG));
3980 case Intrinsic::loongarch_lsx_vmul_b:
3981 case Intrinsic::loongarch_lsx_vmul_h:
3982 case Intrinsic::loongarch_lsx_vmul_w:
3983 case Intrinsic::loongarch_lsx_vmul_d:
3984 case Intrinsic::loongarch_lasx_xvmul_b:
3985 case Intrinsic::loongarch_lasx_xvmul_h:
3986 case Intrinsic::loongarch_lasx_xvmul_w:
3987 case Intrinsic::loongarch_lasx_xvmul_d:
3990 case Intrinsic::loongarch_lsx_vmadd_b:
3991 case Intrinsic::loongarch_lsx_vmadd_h:
3992 case Intrinsic::loongarch_lsx_vmadd_w:
3993 case Intrinsic::loongarch_lsx_vmadd_d:
3994 case Intrinsic::loongarch_lasx_xvmadd_b:
3995 case Intrinsic::loongarch_lasx_xvmadd_h:
3996 case Intrinsic::loongarch_lasx_xvmadd_w:
3997 case Intrinsic::loongarch_lasx_xvmadd_d: {
3998 EVT ResTy =
N->getValueType(0);
4003 case Intrinsic::loongarch_lsx_vmsub_b:
4004 case Intrinsic::loongarch_lsx_vmsub_h:
4005 case Intrinsic::loongarch_lsx_vmsub_w:
4006 case Intrinsic::loongarch_lsx_vmsub_d:
4007 case Intrinsic::loongarch_lasx_xvmsub_b:
4008 case Intrinsic::loongarch_lasx_xvmsub_h:
4009 case Intrinsic::loongarch_lasx_xvmsub_w:
4010 case Intrinsic::loongarch_lasx_xvmsub_d: {
4011 EVT ResTy =
N->getValueType(0);
4016 case Intrinsic::loongarch_lsx_vdiv_b:
4017 case Intrinsic::loongarch_lsx_vdiv_h:
4018 case Intrinsic::loongarch_lsx_vdiv_w:
4019 case Intrinsic::loongarch_lsx_vdiv_d:
4020 case Intrinsic::loongarch_lasx_xvdiv_b:
4021 case Intrinsic::loongarch_lasx_xvdiv_h:
4022 case Intrinsic::loongarch_lasx_xvdiv_w:
4023 case Intrinsic::loongarch_lasx_xvdiv_d:
4026 case Intrinsic::loongarch_lsx_vdiv_bu:
4027 case Intrinsic::loongarch_lsx_vdiv_hu:
4028 case Intrinsic::loongarch_lsx_vdiv_wu:
4029 case Intrinsic::loongarch_lsx_vdiv_du:
4030 case Intrinsic::loongarch_lasx_xvdiv_bu:
4031 case Intrinsic::loongarch_lasx_xvdiv_hu:
4032 case Intrinsic::loongarch_lasx_xvdiv_wu:
4033 case Intrinsic::loongarch_lasx_xvdiv_du:
4036 case Intrinsic::loongarch_lsx_vmod_b:
4037 case Intrinsic::loongarch_lsx_vmod_h:
4038 case Intrinsic::loongarch_lsx_vmod_w:
4039 case Intrinsic::loongarch_lsx_vmod_d:
4040 case Intrinsic::loongarch_lasx_xvmod_b:
4041 case Intrinsic::loongarch_lasx_xvmod_h:
4042 case Intrinsic::loongarch_lasx_xvmod_w:
4043 case Intrinsic::loongarch_lasx_xvmod_d:
4046 case Intrinsic::loongarch_lsx_vmod_bu:
4047 case Intrinsic::loongarch_lsx_vmod_hu:
4048 case Intrinsic::loongarch_lsx_vmod_wu:
4049 case Intrinsic::loongarch_lsx_vmod_du:
4050 case Intrinsic::loongarch_lasx_xvmod_bu:
4051 case Intrinsic::loongarch_lasx_xvmod_hu:
4052 case Intrinsic::loongarch_lasx_xvmod_wu:
4053 case Intrinsic::loongarch_lasx_xvmod_du:
4056 case Intrinsic::loongarch_lsx_vand_v:
4057 case Intrinsic::loongarch_lasx_xvand_v:
4060 case Intrinsic::loongarch_lsx_vor_v:
4061 case Intrinsic::loongarch_lasx_xvor_v:
4064 case Intrinsic::loongarch_lsx_vxor_v:
4065 case Intrinsic::loongarch_lasx_xvxor_v:
4068 case Intrinsic::loongarch_lsx_vnor_v:
4069 case Intrinsic::loongarch_lasx_xvnor_v: {
4074 case Intrinsic::loongarch_lsx_vandi_b:
4075 case Intrinsic::loongarch_lasx_xvandi_b:
4077 lowerVectorSplatImm<8>(
N, 2, DAG));
4078 case Intrinsic::loongarch_lsx_vori_b:
4079 case Intrinsic::loongarch_lasx_xvori_b:
4081 lowerVectorSplatImm<8>(
N, 2, DAG));
4082 case Intrinsic::loongarch_lsx_vxori_b:
4083 case Intrinsic::loongarch_lasx_xvxori_b:
4085 lowerVectorSplatImm<8>(
N, 2, DAG));
4086 case Intrinsic::loongarch_lsx_vsll_b:
4087 case Intrinsic::loongarch_lsx_vsll_h:
4088 case Intrinsic::loongarch_lsx_vsll_w:
4089 case Intrinsic::loongarch_lsx_vsll_d:
4090 case Intrinsic::loongarch_lasx_xvsll_b:
4091 case Intrinsic::loongarch_lasx_xvsll_h:
4092 case Intrinsic::loongarch_lasx_xvsll_w:
4093 case Intrinsic::loongarch_lasx_xvsll_d:
4096 case Intrinsic::loongarch_lsx_vslli_b:
4097 case Intrinsic::loongarch_lasx_xvslli_b:
4099 lowerVectorSplatImm<3>(
N, 2, DAG));
4100 case Intrinsic::loongarch_lsx_vslli_h:
4101 case Intrinsic::loongarch_lasx_xvslli_h:
4103 lowerVectorSplatImm<4>(
N, 2, DAG));
4104 case Intrinsic::loongarch_lsx_vslli_w:
4105 case Intrinsic::loongarch_lasx_xvslli_w:
4107 lowerVectorSplatImm<5>(
N, 2, DAG));
4108 case Intrinsic::loongarch_lsx_vslli_d:
4109 case Intrinsic::loongarch_lasx_xvslli_d:
4111 lowerVectorSplatImm<6>(
N, 2, DAG));
4112 case Intrinsic::loongarch_lsx_vsrl_b:
4113 case Intrinsic::loongarch_lsx_vsrl_h:
4114 case Intrinsic::loongarch_lsx_vsrl_w:
4115 case Intrinsic::loongarch_lsx_vsrl_d:
4116 case Intrinsic::loongarch_lasx_xvsrl_b:
4117 case Intrinsic::loongarch_lasx_xvsrl_h:
4118 case Intrinsic::loongarch_lasx_xvsrl_w:
4119 case Intrinsic::loongarch_lasx_xvsrl_d:
4122 case Intrinsic::loongarch_lsx_vsrli_b:
4123 case Intrinsic::loongarch_lasx_xvsrli_b:
4125 lowerVectorSplatImm<3>(
N, 2, DAG));
4126 case Intrinsic::loongarch_lsx_vsrli_h:
4127 case Intrinsic::loongarch_lasx_xvsrli_h:
4129 lowerVectorSplatImm<4>(
N, 2, DAG));
4130 case Intrinsic::loongarch_lsx_vsrli_w:
4131 case Intrinsic::loongarch_lasx_xvsrli_w:
4133 lowerVectorSplatImm<5>(
N, 2, DAG));
4134 case Intrinsic::loongarch_lsx_vsrli_d:
4135 case Intrinsic::loongarch_lasx_xvsrli_d:
4137 lowerVectorSplatImm<6>(
N, 2, DAG));
4138 case Intrinsic::loongarch_lsx_vsra_b:
4139 case Intrinsic::loongarch_lsx_vsra_h:
4140 case Intrinsic::loongarch_lsx_vsra_w:
4141 case Intrinsic::loongarch_lsx_vsra_d:
4142 case Intrinsic::loongarch_lasx_xvsra_b:
4143 case Intrinsic::loongarch_lasx_xvsra_h:
4144 case Intrinsic::loongarch_lasx_xvsra_w:
4145 case Intrinsic::loongarch_lasx_xvsra_d:
4148 case Intrinsic::loongarch_lsx_vsrai_b:
4149 case Intrinsic::loongarch_lasx_xvsrai_b:
4151 lowerVectorSplatImm<3>(
N, 2, DAG));
4152 case Intrinsic::loongarch_lsx_vsrai_h:
4153 case Intrinsic::loongarch_lasx_xvsrai_h:
4155 lowerVectorSplatImm<4>(
N, 2, DAG));
4156 case Intrinsic::loongarch_lsx_vsrai_w:
4157 case Intrinsic::loongarch_lasx_xvsrai_w:
4159 lowerVectorSplatImm<5>(
N, 2, DAG));
4160 case Intrinsic::loongarch_lsx_vsrai_d:
4161 case Intrinsic::loongarch_lasx_xvsrai_d:
4163 lowerVectorSplatImm<6>(
N, 2, DAG));
4164 case Intrinsic::loongarch_lsx_vclz_b:
4165 case Intrinsic::loongarch_lsx_vclz_h:
4166 case Intrinsic::loongarch_lsx_vclz_w:
4167 case Intrinsic::loongarch_lsx_vclz_d:
4168 case Intrinsic::loongarch_lasx_xvclz_b:
4169 case Intrinsic::loongarch_lasx_xvclz_h:
4170 case Intrinsic::loongarch_lasx_xvclz_w:
4171 case Intrinsic::loongarch_lasx_xvclz_d:
4173 case Intrinsic::loongarch_lsx_vpcnt_b:
4174 case Intrinsic::loongarch_lsx_vpcnt_h:
4175 case Intrinsic::loongarch_lsx_vpcnt_w:
4176 case Intrinsic::loongarch_lsx_vpcnt_d:
4177 case Intrinsic::loongarch_lasx_xvpcnt_b:
4178 case Intrinsic::loongarch_lasx_xvpcnt_h:
4179 case Intrinsic::loongarch_lasx_xvpcnt_w:
4180 case Intrinsic::loongarch_lasx_xvpcnt_d:
4182 case Intrinsic::loongarch_lsx_vbitclr_b:
4183 case Intrinsic::loongarch_lsx_vbitclr_h:
4184 case Intrinsic::loongarch_lsx_vbitclr_w:
4185 case Intrinsic::loongarch_lsx_vbitclr_d:
4186 case Intrinsic::loongarch_lasx_xvbitclr_b:
4187 case Intrinsic::loongarch_lasx_xvbitclr_h:
4188 case Intrinsic::loongarch_lasx_xvbitclr_w:
4189 case Intrinsic::loongarch_lasx_xvbitclr_d:
4191 case Intrinsic::loongarch_lsx_vbitclri_b:
4192 case Intrinsic::loongarch_lasx_xvbitclri_b:
4193 return lowerVectorBitClearImm<3>(
N, DAG);
4194 case Intrinsic::loongarch_lsx_vbitclri_h:
4195 case Intrinsic::loongarch_lasx_xvbitclri_h:
4196 return lowerVectorBitClearImm<4>(
N, DAG);
4197 case Intrinsic::loongarch_lsx_vbitclri_w:
4198 case Intrinsic::loongarch_lasx_xvbitclri_w:
4199 return lowerVectorBitClearImm<5>(
N, DAG);
4200 case Intrinsic::loongarch_lsx_vbitclri_d:
4201 case Intrinsic::loongarch_lasx_xvbitclri_d:
4202 return lowerVectorBitClearImm<6>(
N, DAG);
4203 case Intrinsic::loongarch_lsx_vbitset_b:
4204 case Intrinsic::loongarch_lsx_vbitset_h:
4205 case Intrinsic::loongarch_lsx_vbitset_w:
4206 case Intrinsic::loongarch_lsx_vbitset_d:
4207 case Intrinsic::loongarch_lasx_xvbitset_b:
4208 case Intrinsic::loongarch_lasx_xvbitset_h:
4209 case Intrinsic::loongarch_lasx_xvbitset_w:
4210 case Intrinsic::loongarch_lasx_xvbitset_d: {
4211 EVT VecTy =
N->getValueType(0);
4217 case Intrinsic::loongarch_lsx_vbitseti_b:
4218 case Intrinsic::loongarch_lasx_xvbitseti_b:
4219 return lowerVectorBitSetImm<3>(
N, DAG);
4220 case Intrinsic::loongarch_lsx_vbitseti_h:
4221 case Intrinsic::loongarch_lasx_xvbitseti_h:
4222 return lowerVectorBitSetImm<4>(
N, DAG);
4223 case Intrinsic::loongarch_lsx_vbitseti_w:
4224 case Intrinsic::loongarch_lasx_xvbitseti_w:
4225 return lowerVectorBitSetImm<5>(
N, DAG);
4226 case Intrinsic::loongarch_lsx_vbitseti_d:
4227 case Intrinsic::loongarch_lasx_xvbitseti_d:
4228 return lowerVectorBitSetImm<6>(
N, DAG);
4229 case Intrinsic::loongarch_lsx_vbitrev_b:
4230 case Intrinsic::loongarch_lsx_vbitrev_h:
4231 case Intrinsic::loongarch_lsx_vbitrev_w:
4232 case Intrinsic::loongarch_lsx_vbitrev_d:
4233 case Intrinsic::loongarch_lasx_xvbitrev_b:
4234 case Intrinsic::loongarch_lasx_xvbitrev_h:
4235 case Intrinsic::loongarch_lasx_xvbitrev_w:
4236 case Intrinsic::loongarch_lasx_xvbitrev_d: {
4237 EVT VecTy =
N->getValueType(0);
4243 case Intrinsic::loongarch_lsx_vbitrevi_b:
4244 case Intrinsic::loongarch_lasx_xvbitrevi_b:
4245 return lowerVectorBitRevImm<3>(
N, DAG);
4246 case Intrinsic::loongarch_lsx_vbitrevi_h:
4247 case Intrinsic::loongarch_lasx_xvbitrevi_h:
4248 return lowerVectorBitRevImm<4>(
N, DAG);
4249 case Intrinsic::loongarch_lsx_vbitrevi_w:
4250 case Intrinsic::loongarch_lasx_xvbitrevi_w:
4251 return lowerVectorBitRevImm<5>(
N, DAG);
4252 case Intrinsic::loongarch_lsx_vbitrevi_d:
4253 case Intrinsic::loongarch_lasx_xvbitrevi_d:
4254 return lowerVectorBitRevImm<6>(
N, DAG);
4255 case Intrinsic::loongarch_lsx_vfadd_s:
4256 case Intrinsic::loongarch_lsx_vfadd_d:
4257 case Intrinsic::loongarch_lasx_xvfadd_s:
4258 case Intrinsic::loongarch_lasx_xvfadd_d:
4261 case Intrinsic::loongarch_lsx_vfsub_s:
4262 case Intrinsic::loongarch_lsx_vfsub_d:
4263 case Intrinsic::loongarch_lasx_xvfsub_s:
4264 case Intrinsic::loongarch_lasx_xvfsub_d:
4267 case Intrinsic::loongarch_lsx_vfmul_s:
4268 case Intrinsic::loongarch_lsx_vfmul_d:
4269 case Intrinsic::loongarch_lasx_xvfmul_s:
4270 case Intrinsic::loongarch_lasx_xvfmul_d:
4273 case Intrinsic::loongarch_lsx_vfdiv_s:
4274 case Intrinsic::loongarch_lsx_vfdiv_d:
4275 case Intrinsic::loongarch_lasx_xvfdiv_s:
4276 case Intrinsic::loongarch_lasx_xvfdiv_d:
4279 case Intrinsic::loongarch_lsx_vfmadd_s:
4280 case Intrinsic::loongarch_lsx_vfmadd_d:
4281 case Intrinsic::loongarch_lasx_xvfmadd_s:
4282 case Intrinsic::loongarch_lasx_xvfmadd_d:
4284 N->getOperand(2),
N->getOperand(3));
4285 case Intrinsic::loongarch_lsx_vinsgr2vr_b:
4287 N->getOperand(1),
N->getOperand(2),
4288 legalizeIntrinsicImmArg<4>(
N, 3, DAG, Subtarget));
4289 case Intrinsic::loongarch_lsx_vinsgr2vr_h:
4290 case Intrinsic::loongarch_lasx_xvinsgr2vr_w:
4292 N->getOperand(1),
N->getOperand(2),
4293 legalizeIntrinsicImmArg<3>(
N, 3, DAG, Subtarget));
4294 case Intrinsic::loongarch_lsx_vinsgr2vr_w:
4295 case Intrinsic::loongarch_lasx_xvinsgr2vr_d:
4297 N->getOperand(1),
N->getOperand(2),
4298 legalizeIntrinsicImmArg<2>(
N, 3, DAG, Subtarget));
4299 case Intrinsic::loongarch_lsx_vinsgr2vr_d:
4301 N->getOperand(1),
N->getOperand(2),
4302 legalizeIntrinsicImmArg<1>(
N, 3, DAG, Subtarget));
4303 case Intrinsic::loongarch_lsx_vreplgr2vr_b:
4304 case Intrinsic::loongarch_lsx_vreplgr2vr_h:
4305 case Intrinsic::loongarch_lsx_vreplgr2vr_w:
4306 case Intrinsic::loongarch_lsx_vreplgr2vr_d:
4307 case Intrinsic::loongarch_lasx_xvreplgr2vr_b:
4308 case Intrinsic::loongarch_lasx_xvreplgr2vr_h:
4309 case Intrinsic::loongarch_lasx_xvreplgr2vr_w:
4310 case Intrinsic::loongarch_lasx_xvreplgr2vr_d:
4314 case Intrinsic::loongarch_lsx_vreplve_b:
4315 case Intrinsic::loongarch_lsx_vreplve_h:
4316 case Intrinsic::loongarch_lsx_vreplve_w:
4317 case Intrinsic::loongarch_lsx_vreplve_d:
4318 case Intrinsic::loongarch_lasx_xvreplve_b:
4319 case Intrinsic::loongarch_lasx_xvreplve_h:
4320 case Intrinsic::loongarch_lasx_xvreplve_w:
4321 case Intrinsic::loongarch_lasx_xvreplve_d:
4333 switch (
N->getOpcode()) {
4370 MF->
insert(It, BreakMBB);
4374 SinkMBB->splice(SinkMBB->end(),
MBB, std::next(
MI.getIterator()),
MBB->
end());
4375 SinkMBB->transferSuccessorsAndUpdatePHIs(
MBB);
4393 BreakMBB->addSuccessor(SinkMBB);
4405 switch (
MI.getOpcode()) {
4408 case LoongArch::PseudoVBZ:
4409 CondOpc = LoongArch::VSETEQZ_V;
4411 case LoongArch::PseudoVBZ_B:
4412 CondOpc = LoongArch::VSETANYEQZ_B;
4414 case LoongArch::PseudoVBZ_H:
4415 CondOpc = LoongArch::VSETANYEQZ_H;
4417 case LoongArch::PseudoVBZ_W:
4418 CondOpc = LoongArch::VSETANYEQZ_W;
4420 case LoongArch::PseudoVBZ_D:
4421 CondOpc = LoongArch::VSETANYEQZ_D;
4423 case LoongArch::PseudoVBNZ:
4424 CondOpc = LoongArch::VSETNEZ_V;
4426 case LoongArch::PseudoVBNZ_B:
4427 CondOpc = LoongArch::VSETALLNEZ_B;
4429 case LoongArch::PseudoVBNZ_H:
4430 CondOpc = LoongArch::VSETALLNEZ_H;
4432 case LoongArch::PseudoVBNZ_W:
4433 CondOpc = LoongArch::VSETALLNEZ_W;
4435 case LoongArch::PseudoVBNZ_D:
4436 CondOpc = LoongArch::VSETALLNEZ_D;
4438 case LoongArch::PseudoXVBZ:
4439 CondOpc = LoongArch::XVSETEQZ_V;
4441 case LoongArch::PseudoXVBZ_B:
4442 CondOpc = LoongArch::XVSETANYEQZ_B;
4444 case LoongArch::PseudoXVBZ_H:
4445 CondOpc = LoongArch::XVSETANYEQZ_H;
4447 case LoongArch::PseudoXVBZ_W:
4448 CondOpc = LoongArch::XVSETANYEQZ_W;
4450 case LoongArch::PseudoXVBZ_D:
4451 CondOpc = LoongArch::XVSETANYEQZ_D;
4453 case LoongArch::PseudoXVBNZ:
4454 CondOpc = LoongArch::XVSETNEZ_V;
4456 case LoongArch::PseudoXVBNZ_B:
4457 CondOpc = LoongArch::XVSETALLNEZ_B;
4459 case LoongArch::PseudoXVBNZ_H:
4460 CondOpc = LoongArch::XVSETALLNEZ_H;
4462 case LoongArch::PseudoXVBNZ_W:
4463 CondOpc = LoongArch::XVSETALLNEZ_W;
4465 case LoongArch::PseudoXVBNZ_D:
4466 CondOpc = LoongArch::XVSETALLNEZ_D;
4481 F->insert(It, FalseBB);
4482 F->insert(It, TrueBB);
4483 F->insert(It, SinkBB);
4486 SinkBB->
splice(SinkBB->
end(), BB, std::next(
MI.getIterator()), BB->
end());
4490 Register FCC =
MRI.createVirtualRegister(&LoongArch::CFRRegClass);
4499 Register RD1 =
MRI.createVirtualRegister(&LoongArch::GPRRegClass);
4507 Register RD2 =
MRI.createVirtualRegister(&LoongArch::GPRRegClass);
4515 MI.getOperand(0).getReg())
4522 MI.eraseFromParent();
4531 switch (
MI.getOpcode()) {
4534 case LoongArch::PseudoXVINSGR2VR_B:
4536 InsOp = LoongArch::VINSGR2VR_B;
4538 case LoongArch::PseudoXVINSGR2VR_H:
4540 InsOp = LoongArch::VINSGR2VR_H;
4552 unsigned Idx =
MI.getOperand(3).getImm();
4555 if (
Idx >= HalfSize) {
4556 ScratchReg1 =
MRI.createVirtualRegister(RC);
4557 BuildMI(*BB,
MI,
DL,
TII->get(LoongArch::XVPERMI_Q), ScratchReg1)
4563 Register ScratchSubReg1 =
MRI.createVirtualRegister(SubRC);
4564 Register ScratchSubReg2 =
MRI.createVirtualRegister(SubRC);
4566 .
addReg(ScratchReg1, 0, LoongArch::sub_128);
4573 if (
Idx >= HalfSize)
4574 ScratchReg2 =
MRI.createVirtualRegister(RC);
4576 BuildMI(*BB,
MI,
DL,
TII->get(LoongArch::SUBREG_TO_REG), ScratchReg2)
4579 .
addImm(LoongArch::sub_128);
4581 if (
Idx >= HalfSize)
4587 MI.eraseFromParent();
4594 assert(Subtarget.hasExtLSX());
4601 Register ScratchReg1 =
MRI.createVirtualRegister(RC);
4602 Register ScratchReg2 =
MRI.createVirtualRegister(RC);
4603 Register ScratchReg3 =
MRI.createVirtualRegister(RC);
4607 TII->get(Subtarget.
is64Bit() ? LoongArch::VINSGR2VR_D
4608 : LoongArch::VINSGR2VR_W),
4615 TII->get(Subtarget.
is64Bit() ? LoongArch::VPCNT_D : LoongArch::VPCNT_W),
4619 TII->get(Subtarget.
is64Bit() ? LoongArch::VPICKVE2GR_D
4620 : LoongArch::VPICKVE2GR_W),
4625 MI.eraseFromParent();
4634 switch (
MI.getOpcode()) {
4637 case LoongArch::DIV_W:
4638 case LoongArch::DIV_WU:
4639 case LoongArch::MOD_W:
4640 case LoongArch::MOD_WU:
4641 case LoongArch::DIV_D:
4642 case LoongArch::DIV_DU:
4643 case LoongArch::MOD_D:
4644 case LoongArch::MOD_DU:
4647 case LoongArch::WRFCSR: {
4649 LoongArch::FCSR0 +
MI.getOperand(0).getImm())
4650 .
addReg(
MI.getOperand(1).getReg());
4651 MI.eraseFromParent();
4654 case LoongArch::RDFCSR: {
4657 MI.getOperand(0).getReg())
4658 .
addReg(LoongArch::FCSR0 +
MI.getOperand(1).getImm());
4660 MI.eraseFromParent();
4663 case LoongArch::PseudoVBZ:
4664 case LoongArch::PseudoVBZ_B:
4665 case LoongArch::PseudoVBZ_H:
4666 case LoongArch::PseudoVBZ_W:
4667 case LoongArch::PseudoVBZ_D:
4668 case LoongArch::PseudoVBNZ:
4669 case LoongArch::PseudoVBNZ_B:
4670 case LoongArch::PseudoVBNZ_H:
4671 case LoongArch::PseudoVBNZ_W:
4672 case LoongArch::PseudoVBNZ_D:
4673 case LoongArch::PseudoXVBZ:
4674 case LoongArch::PseudoXVBZ_B:
4675 case LoongArch::PseudoXVBZ_H:
4676 case LoongArch::PseudoXVBZ_W:
4677 case LoongArch::PseudoXVBZ_D:
4678 case LoongArch::PseudoXVBNZ:
4679 case LoongArch::PseudoXVBNZ_B:
4680 case LoongArch::PseudoXVBNZ_H:
4681 case LoongArch::PseudoXVBNZ_W:
4682 case LoongArch::PseudoXVBNZ_D:
4684 case LoongArch::PseudoXVINSGR2VR_B:
4685 case LoongArch::PseudoXVINSGR2VR_H:
4687 case LoongArch::PseudoCTPOP:
4689 case TargetOpcode::STATEPOINT:
4695 MI.addOperand(*
MI.getMF(),
4697 LoongArch::R1,
true,
4708 unsigned *
Fast)
const {
4709 if (!Subtarget.hasUAL())
4723#define NODE_NAME_CASE(node) \
4724 case LoongArchISD::node: \
4725 return "LoongArchISD::" #node;
4805#undef NODE_NAME_CASE
4818 LoongArch::R7, LoongArch::R8, LoongArch::R9,
4819 LoongArch::R10, LoongArch::R11};
4823 LoongArch::F3, LoongArch::F4, LoongArch::F5,
4824 LoongArch::F6, LoongArch::F7};
4827 LoongArch::F0_64, LoongArch::F1_64, LoongArch::F2_64, LoongArch::F3_64,
4828 LoongArch::F4_64, LoongArch::F5_64, LoongArch::F6_64, LoongArch::F7_64};
4831 LoongArch::VR3, LoongArch::VR4, LoongArch::VR5,
4832 LoongArch::VR6, LoongArch::VR7};
4835 LoongArch::XR3, LoongArch::XR4, LoongArch::XR5,
4836 LoongArch::XR6, LoongArch::XR7};
4842 unsigned ValNo2,
MVT ValVT2,
MVT LocVT2,
4844 unsigned GRLenInBytes = GRLen / 8;
4877 unsigned ValNo,
MVT ValVT,
4879 CCState &State,
bool IsFixed,
bool IsRet,
4881 unsigned GRLen =
DL.getLargestLegalIntTypeSizeInBits();
4882 assert((GRLen == 32 || GRLen == 64) &&
"Unspport GRLen");
4883 MVT GRLenVT = GRLen == 32 ? MVT::i32 : MVT::i64;
4888 if (IsRet && ValNo > 1)
4892 bool UseGPRForFloat =
true;
4902 UseGPRForFloat = !IsFixed;
4911 UseGPRForFloat =
true;
4913 if (UseGPRForFloat && ValVT == MVT::f32) {
4916 }
else if (UseGPRForFloat && GRLen == 64 && ValVT == MVT::f64) {
4919 }
else if (UseGPRForFloat && GRLen == 32 && ValVT == MVT::f64) {
4930 unsigned TwoGRLenInBytes = (2 * GRLen) / 8;
4932 DL.getTypeAllocSize(OrigTy) == TwoGRLenInBytes) {
4935 if (RegIdx != std::size(
ArgGPRs) && RegIdx % 2 == 1)
4944 "PendingLocs and PendingArgFlags out of sync");
4962 PendingLocs.
size() <= 2) {
4963 assert(PendingLocs.
size() == 2 &&
"Unexpected PendingLocs.size()");
4968 PendingLocs.
clear();
4969 PendingArgFlags.
clear();
4976 unsigned StoreSizeBytes = GRLen / 8;
4979 if (ValVT == MVT::f32 && !UseGPRForFloat)
4981 else if (ValVT == MVT::f64 && !UseGPRForFloat)
4995 if (!PendingLocs.
empty()) {
4997 assert(PendingLocs.
size() > 2 &&
"Unexpected PendingLocs.size()");
4998 for (
auto &It : PendingLocs) {
5000 It.convertToReg(Reg);
5005 PendingLocs.clear();
5006 PendingArgFlags.
clear();
5009 assert((!UseGPRForFloat || LocVT == GRLenVT) &&
5010 "Expected an GRLenVT at this stage");
5027void LoongArchTargetLowering::analyzeInputArgs(
5030 LoongArchCCAssignFn Fn)
const {
5032 for (
unsigned i = 0, e =
Ins.size(); i != e; ++i) {
5034 Type *ArgTy =
nullptr;
5036 ArgTy = FType->getReturnType();
5037 else if (Ins[i].isOrigArg())
5038 ArgTy = FType->getParamType(Ins[i].getOrigArgIndex());
5042 CCInfo,
true, IsRet, ArgTy)) {
5043 LLVM_DEBUG(
dbgs() <<
"InputArg #" << i <<
" has unhandled type " << ArgVT
5050void LoongArchTargetLowering::analyzeOutputArgs(
5053 CallLoweringInfo *CLI, LoongArchCCAssignFn Fn)
const {
5054 for (
unsigned i = 0, e = Outs.
size(); i != e; ++i) {
5055 MVT ArgVT = Outs[i].VT;
5056 Type *OrigTy = CLI ? CLI->getArgs()[Outs[i].OrigArgIndex].Ty :
nullptr;
5060 CCInfo, Outs[i].IsFixed, IsRet, OrigTy)) {
5061 LLVM_DEBUG(
dbgs() <<
"OutputArg #" << i <<
" has unhandled type " << ArgVT
5102 if (In.isOrigArg()) {
5107 if ((
BitWidth <= 32 && In.Flags.isSExt()) ||
5108 (
BitWidth < 32 && In.Flags.isZExt())) {
5168 if (LocVT == MVT::i32 || LocVT == MVT::i64) {
5172 LoongArch::R23, LoongArch::R24, LoongArch::R25,
5173 LoongArch::R26, LoongArch::R27, LoongArch::R28,
5174 LoongArch::R29, LoongArch::R30, LoongArch::R31};
5181 if (LocVT == MVT::f32) {
5184 static const MCPhysReg FPR32List[] = {LoongArch::F24, LoongArch::F25,
5185 LoongArch::F26, LoongArch::F27};
5192 if (LocVT == MVT::f64) {
5195 static const MCPhysReg FPR64List[] = {LoongArch::F28_64, LoongArch::F29_64,
5196 LoongArch::F30_64, LoongArch::F31_64};
5225 "GHC calling convention requires the F and D extensions");
5230 unsigned GRLenInBytes = Subtarget.
getGRLen() / 8;
5232 std::vector<SDValue> OutChains;
5241 analyzeInputArgs(MF, CCInfo, Ins,
false,
CC_LoongArch);
5243 for (
unsigned i = 0, e = ArgLocs.
size(); i != e; ++i) {
5255 unsigned ArgIndex = Ins[i].OrigArgIndex;
5256 unsigned ArgPartOffset = Ins[i].PartOffset;
5257 assert(ArgPartOffset == 0);
5258 while (i + 1 != e && Ins[i + 1].OrigArgIndex == ArgIndex) {
5260 unsigned PartOffset = Ins[i + 1].PartOffset - ArgPartOffset;
5283 int VaArgOffset, VarArgsSaveSize;
5289 VarArgsSaveSize = 0;
5291 VarArgsSaveSize = GRLenInBytes * (ArgRegs.
size() -
Idx);
5292 VaArgOffset = -VarArgsSaveSize;
5298 LoongArchFI->setVarArgsFrameIndex(FI);
5306 VarArgsSaveSize += GRLenInBytes;
5311 for (
unsigned I =
Idx;
I < ArgRegs.
size();
5312 ++
I, VaArgOffset += GRLenInBytes) {
5320 cast<StoreSDNode>(Store.getNode())
5322 ->setValue((
Value *)
nullptr);
5323 OutChains.push_back(Store);
5325 LoongArchFI->setVarArgsSaveSize(VarArgsSaveSize);
5330 if (!OutChains.empty()) {
5331 OutChains.push_back(Chain);
5346 if (
N->getNumValues() != 1)
5348 if (!
N->hasNUsesOfValue(1, 0))
5351 SDNode *Copy = *
N->user_begin();
5357 if (Copy->getGluedNode())
5361 bool HasRet =
false;
5362 for (
SDNode *Node : Copy->users()) {
5371 Chain = Copy->getOperand(0);
5376bool LoongArchTargetLowering::isEligibleForTailCallOptimization(
5380 auto CalleeCC = CLI.CallConv;
5381 auto &Outs = CLI.Outs;
5383 auto CallerCC = Caller.getCallingConv();
5390 for (
auto &VA : ArgLocs)
5396 auto IsCallerStructRet = Caller.hasStructRetAttr();
5397 auto IsCalleeStructRet = Outs.
empty() ?
false : Outs[0].Flags.isSRet();
5398 if (IsCallerStructRet || IsCalleeStructRet)
5402 for (
auto &Arg : Outs)
5403 if (Arg.Flags.isByVal())
5408 const uint32_t *CallerPreserved =
TRI->getCallPreservedMask(MF, CallerCC);
5409 if (CalleeCC != CallerCC) {
5410 const uint32_t *CalleePreserved =
TRI->getCallPreservedMask(MF, CalleeCC);
5411 if (!
TRI->regmaskSubsetEqual(CallerPreserved, CalleePreserved))
5449 analyzeOutputArgs(MF, ArgCCInfo, Outs,
false, &CLI,
CC_LoongArch);
5453 IsTailCall = isEligibleForTailCallOptimization(ArgCCInfo, CLI, MF, ArgLocs);
5459 "site marked musttail");
5466 for (
unsigned i = 0, e = Outs.
size(); i != e; ++i) {
5468 if (!Flags.isByVal())
5472 unsigned Size = Flags.getByValSize();
5473 Align Alignment = Flags.getNonZeroByValAlign();
5480 Chain = DAG.
getMemcpy(Chain,
DL, FIPtr, Arg, SizeNode, Alignment,
5482 false,
nullptr, std::nullopt,
5494 for (
unsigned i = 0, j = 0, e = ArgLocs.
size(); i != e; ++i) {
5496 SDValue ArgValue = OutVals[i];
5509 unsigned ArgIndex = Outs[i].OrigArgIndex;
5510 unsigned ArgPartOffset = Outs[i].PartOffset;
5511 assert(ArgPartOffset == 0);
5516 while (i + 1 != e && Outs[i + 1].OrigArgIndex == ArgIndex) {
5517 SDValue PartValue = OutVals[i + 1];
5518 unsigned PartOffset = Outs[i + 1].PartOffset - ArgPartOffset;
5528 int FI = cast<FrameIndexSDNode>(SpillSlot)->getIndex();
5532 for (
const auto &Part : Parts) {
5533 SDValue PartValue = Part.first;
5534 SDValue PartOffset = Part.second;
5541 ArgValue = SpillSlot;
5547 if (Flags.isByVal())
5548 ArgValue = ByValArgs[j++];
5555 assert(!IsTailCall &&
"Tail call not allowed if stack is used "
5556 "for passing parameters");
5559 if (!StackPtr.getNode())
5572 if (!MemOpChains.
empty())
5578 for (
auto &Reg : RegsToPass) {
5579 Chain = DAG.
getCopyToReg(Chain,
DL, Reg.first, Reg.second, Glue);
5606 for (
auto &Reg : RegsToPass)
5612 const uint32_t *Mask =
TRI->getCallPreservedMask(MF, CallConv);
5613 assert(Mask &&
"Missing call preserved mask for calling convention");
5631 assert(Subtarget.
is64Bit() &&
"Medium code model requires LA64");
5635 assert(Subtarget.
is64Bit() &&
"Large code model requires LA64");
5658 analyzeInputArgs(MF, RetCCInfo, Ins,
true,
CC_LoongArch);
5661 for (
auto &VA : RVLocs) {
5681 CCState CCInfo(CallConv, IsVarArg, MF, RVLocs, Context);
5683 for (
unsigned i = 0, e = Outs.
size(); i != e; ++i) {
5687 Outs[i].Flags, CCInfo,
true,
true,
5714 for (
unsigned i = 0, e = RVLocs.
size(); i < e; ++i) {
5738 if (!Subtarget.hasExtLSX())
5741 if (VT == MVT::f32) {
5742 uint64_t masked = Imm.bitcastToAPInt().getZExtValue() & 0x7e07ffff;
5743 return (masked == 0x3e000000 || masked == 0x40000000);
5746 if (VT == MVT::f64) {
5747 uint64_t masked = Imm.bitcastToAPInt().getZExtValue() & 0x7fc0ffffffffffff;
5748 return (masked == 0x3fc0000000000000 || masked == 0x4000000000000000);
5754bool LoongArchTargetLowering::isFPImmLegal(
const APFloat &Imm,
EVT VT,
5755 bool ForCodeSize)
const {
5757 if (VT == MVT::f32 && !Subtarget.hasBasicF())
5759 if (VT == MVT::f64 && !Subtarget.hasBasicD())
5761 return (Imm.isZero() || Imm.isExactlyValue(1.0) ||
isFPImmVLDILegal(Imm, VT));
5772bool LoongArchTargetLowering::shouldInsertFencesForAtomic(
5775 return isa<LoadInst>(
I) || isa<StoreInst>(
I);
5777 if (isa<LoadInst>(
I))
5782 Type *Ty =
I->getOperand(0)->getType();
5801 return Y.getValueType().isScalarInteger() && !isa<ConstantSDNode>(
Y);
5807 unsigned Intrinsic)
const {
5808 switch (Intrinsic) {
5811 case Intrinsic::loongarch_masked_atomicrmw_xchg_i32:
5812 case Intrinsic::loongarch_masked_atomicrmw_add_i32:
5813 case Intrinsic::loongarch_masked_atomicrmw_sub_i32:
5814 case Intrinsic::loongarch_masked_atomicrmw_nand_i32:
5816 Info.memVT = MVT::i32;
5817 Info.ptrVal =
I.getArgOperand(0);
5836 "Unable to expand");
5837 unsigned MinWordSize = 4;
5850 Intrinsic::ptrmask, {PtrTy, IntTy},
5851 {
Addr, ConstantInt::get(IntTy, ~(
uint64_t)(MinWordSize - 1))},
nullptr,
5855 Value *PtrLSB = Builder.
CreateAnd(AddrInt, MinWordSize - 1,
"PtrLSB");
5857 ShiftAmt = Builder.
CreateTrunc(ShiftAmt, WordType,
"ShiftAmt");
5859 ConstantInt::get(WordType,
5863 Value *ValOperand_Shifted =
5865 ShiftAmt,
"ValOperand_Shifted");
5868 NewOperand = Builder.
CreateOr(ValOperand_Shifted, Inv_Mask,
"AndOperand");
5870 NewOperand = ValOperand_Shifted;
5896 if (Subtarget.hasLAM_BH() && Subtarget.
is64Bit() &&
5904 if (Subtarget.hasLAMCAS()) {
5926 return Intrinsic::loongarch_masked_atomicrmw_xchg_i64;
5928 return Intrinsic::loongarch_masked_atomicrmw_add_i64;
5930 return Intrinsic::loongarch_masked_atomicrmw_sub_i64;
5932 return Intrinsic::loongarch_masked_atomicrmw_nand_i64;
5934 return Intrinsic::loongarch_masked_atomicrmw_umax_i64;
5936 return Intrinsic::loongarch_masked_atomicrmw_umin_i64;
5938 return Intrinsic::loongarch_masked_atomicrmw_max_i64;
5940 return Intrinsic::loongarch_masked_atomicrmw_min_i64;
5950 return Intrinsic::loongarch_masked_atomicrmw_xchg_i32;
5952 return Intrinsic::loongarch_masked_atomicrmw_add_i32;
5954 return Intrinsic::loongarch_masked_atomicrmw_sub_i32;
5956 return Intrinsic::loongarch_masked_atomicrmw_nand_i32;
5968 if (Subtarget.hasLAMCAS())
5981 Value *FailureOrdering =
5985 Intrinsic::ID CmpXchgIntrID = Intrinsic::loongarch_masked_cmpxchg_i64;
5991 CmpXchgIntrID, Tys, {AlignedAddr, CmpVal, NewVal, Mask, FailureOrdering});
6015 unsigned GRLen = Subtarget.
getGRLen();
6044 {AlignedAddr, Incr, Mask, SextShamt, Ordering});
6047 Builder.
CreateCall(LlwOpScwLoop, {AlignedAddr, Incr, Mask, Ordering});
6074 const Constant *PersonalityFn)
const {
6075 return LoongArch::R4;
6079 const Constant *PersonalityFn)
const {
6080 return LoongArch::R5;
6091 int RefinementSteps = VT.
getScalarType() == MVT::f64 ? 2 : 1;
6092 return RefinementSteps;
6097 int &RefinementSteps,
6098 bool &UseOneConstNR,
6099 bool Reciprocal)
const {
6100 if (Subtarget.hasFrecipe()) {
6104 if (VT == MVT::f32 || (VT == MVT::f64 && Subtarget.hasBasicD()) ||
6105 (VT == MVT::v4f32 && Subtarget.hasExtLSX()) ||
6106 (VT == MVT::v2f64 && Subtarget.hasExtLSX()) ||
6107 (VT == MVT::v8f32 && Subtarget.hasExtLASX()) ||
6108 (VT == MVT::v4f64 && Subtarget.hasExtLASX())) {
6127 int &RefinementSteps)
const {
6128 if (Subtarget.hasFrecipe()) {
6132 if (VT == MVT::f32 || (VT == MVT::f64 && Subtarget.hasBasicD()) ||
6133 (VT == MVT::v4f32 && Subtarget.hasExtLSX()) ||
6134 (VT == MVT::v2f64 && Subtarget.hasExtLSX()) ||
6135 (VT == MVT::v8f32 && Subtarget.hasExtLASX()) ||
6136 (VT == MVT::v4f64 && Subtarget.hasExtLASX())) {
6153LoongArchTargetLowering::getConstraintType(
StringRef Constraint)
const {
6171 if (Constraint.
size() == 1) {
6172 switch (Constraint[0]) {
6187 if (Constraint ==
"ZC" || Constraint ==
"ZB")
6203std::pair<unsigned, const TargetRegisterClass *>
6204LoongArchTargetLowering::getRegForInlineAsmConstraint(
6208 if (Constraint.
size() == 1) {
6209 switch (Constraint[0]) {
6214 return std::make_pair(0U, &LoongArch::GPRRegClass);
6216 if (Subtarget.hasBasicF() && VT == MVT::f32)
6217 return std::make_pair(0U, &LoongArch::FPR32RegClass);
6218 if (Subtarget.hasBasicD() && VT == MVT::f64)
6219 return std::make_pair(0U, &LoongArch::FPR64RegClass);
6220 if (Subtarget.hasExtLSX() &&
6221 TRI->isTypeLegalForClass(LoongArch::LSX128RegClass, VT))
6222 return std::make_pair(0U, &LoongArch::LSX128RegClass);
6223 if (Subtarget.hasExtLASX() &&
6224 TRI->isTypeLegalForClass(LoongArch::LASX256RegClass, VT))
6225 return std::make_pair(0U, &LoongArch::LASX256RegClass);
6245 bool IsFP = Constraint[2] ==
'f';
6246 std::pair<StringRef, StringRef> Temp = Constraint.
split(
'$');
6247 std::pair<unsigned, const TargetRegisterClass *>
R;
6249 TRI, join_items(
"", Temp.first, Temp.second), VT);
6252 unsigned RegNo =
R.first;
6253 if (LoongArch::F0 <= RegNo && RegNo <= LoongArch::F31) {
6254 if (Subtarget.hasBasicD() && (VT == MVT::f64 || VT == MVT::Other)) {
6255 unsigned DReg = RegNo - LoongArch::F0 + LoongArch::F0_64;
6256 return std::make_pair(DReg, &LoongArch::FPR64RegClass);
6266void LoongArchTargetLowering::LowerAsmOperandForConstraint(
6270 if (Constraint.
size() == 1) {
6271 switch (Constraint[0]) {
6274 if (
auto *
C = dyn_cast<ConstantSDNode>(
Op)) {
6276 if (isInt<16>(CVal))
6283 if (
auto *
C = dyn_cast<ConstantSDNode>(
Op)) {
6285 if (isInt<12>(CVal))
6292 if (
auto *
C = dyn_cast<ConstantSDNode>(
Op))
6293 if (
C->getZExtValue() == 0)
6299 if (
auto *
C = dyn_cast<ConstantSDNode>(
Op)) {
6301 if (isUInt<12>(CVal))
6313#define GET_REGISTER_MATCHER
6314#include "LoongArchGenAsmMatcher.inc"
6320 std::string NewRegName =
Name.second.str();
6322 if (Reg == LoongArch::NoRegister)
6324 if (Reg == LoongArch::NoRegister)
6328 if (!ReservedRegs.
test(Reg))
6344 if (
auto *ConstNode = dyn_cast<ConstantSDNode>(
C.getNode())) {
6345 const APInt &Imm = ConstNode->getAPIntValue();
6347 if ((Imm + 1).isPowerOf2() || (Imm - 1).isPowerOf2() ||
6348 (1 - Imm).isPowerOf2() || (-1 - Imm).isPowerOf2())
6351 if (ConstNode->hasOneUse() &&
6352 ((Imm - 2).isPowerOf2() || (Imm - 4).isPowerOf2() ||
6353 (Imm - 8).isPowerOf2() || (Imm - 16).isPowerOf2()))
6359 if (ConstNode->hasOneUse() && !(Imm.sge(-2048) && Imm.sle(4095))) {
6360 unsigned Shifts = Imm.countr_zero();
6366 APInt ImmPop = Imm.ashr(Shifts);
6367 if (ImmPop == 3 || ImmPop == 5 || ImmPop == 9 || ImmPop == 17)
6371 APInt ImmSmall =
APInt(Imm.getBitWidth(), 1ULL << Shifts,
true);
6372 if ((Imm - ImmSmall).isPowerOf2() || (Imm + ImmSmall).isPowerOf2() ||
6373 (ImmSmall - Imm).isPowerOf2())
6383 Type *Ty,
unsigned AS,
6399 !(isShiftedInt<14, 2>(AM.
BaseOffs) && Subtarget.hasUAL()))
6426 return isInt<12>(Imm);
6430 return isInt<12>(Imm);
6437 if (
auto *LD = dyn_cast<LoadSDNode>(Val)) {
6438 EVT MemVT = LD->getMemoryVT();
6439 if ((MemVT == MVT::i8 || MemVT == MVT::i16) &&
6450 return Subtarget.
is64Bit() && SrcVT == MVT::i32 && DstVT == MVT::i64;
6459 if (
Y.getValueType().isVector())
6462 return !isa<ConstantSDNode>(
Y);
6471 Type *Ty,
bool IsSigned)
const {
6492 Align &PrefAlign)
const {
6493 if (!isa<MemIntrinsic>(CI))
6498 PrefAlign =
Align(8);
6501 PrefAlign =
Align(4);
unsigned const MachineRegisterInfo * MRI
static MCRegister MatchRegisterName(StringRef Name)
static bool checkValueWidth(SDValue V, unsigned width, ISD::LoadExtType &ExtType)
static SDValue performORCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI, const AArch64Subtarget *Subtarget, const AArch64TargetLowering &TLI)
static SDValue performANDCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI)
static SDValue performSETCCCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI, SelectionDAG &DAG)
static msgpack::DocNode getNode(msgpack::DocNode DN, msgpack::Type Type, MCValue Val)
#define NODE_NAME_CASE(node)
MachineBasicBlock MachineBasicBlock::iterator DebugLoc DL
static MCRegister MatchRegisterAltName(StringRef Name)
Maps from the set of all alternative registernames to a register number.
Function Alias Analysis Results
static uint64_t getConstant(const Value *IndexValue)
static SDValue getTargetNode(GlobalAddressSDNode *N, const SDLoc &DL, EVT Ty, SelectionDAG &DAG, unsigned Flags)
Analysis containing CSE Info
static SDValue convertValVTToLocVT(SelectionDAG &DAG, SDValue Val, const CCValAssign &VA, const SDLoc &DL)
static SDValue unpackFromMemLoc(SelectionDAG &DAG, SDValue Chain, const CCValAssign &VA, const SDLoc &DL)
static SDValue convertLocVTToValVT(SelectionDAG &DAG, SDValue Val, const CCValAssign &VA, const SDLoc &DL)
static SDValue unpackFromRegLoc(const CSKYSubtarget &Subtarget, SelectionDAG &DAG, SDValue Chain, const CCValAssign &VA, const SDLoc &DL)
Returns the sub type a function will return at a given Idx Should correspond to the result type of an ExtractValue instruction executed with just that one unsigned Idx
const HexagonInstrInfo * TII
static SDValue performINTRINSIC_WO_CHAINCombine(SDNode *N, SelectionDAG &DAG, TargetLowering::DAGCombinerInfo &DCI, const LoongArchSubtarget &Subtarget)
static SDValue lowerVECTOR_SHUFFLE_VREPLVEI(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VREPLVEI (if possible).
const MCPhysReg ArgFPR32s[]
static SDValue lowerVECTOR_SHUFFLE_VSHUF4I(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VSHUF4I (if possible).
static SDValue lowerVECTOR_SHUFFLE_VPICKEV(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VPICKEV (if possible).
static SDValue lowerVECTOR_SHUFFLE_XVPICKOD(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVPICKOD (if possible).
static bool fitsRegularPattern(typename SmallVectorImpl< ValType >::const_iterator Begin, unsigned CheckStride, typename SmallVectorImpl< ValType >::const_iterator End, ValType ExpectedIndex, unsigned ExpectedIndexStride)
Determine whether a range fits a regular pattern of values.
static void canonicalizeShuffleVectorByLane(const SDLoc &DL, MutableArrayRef< int > Mask, MVT VT, SDValue &V1, SDValue &V2, SelectionDAG &DAG)
Shuffle vectors by lane to generate more optimized instructions.
static SDValue emitIntrinsicErrorMessage(SDValue Op, StringRef ErrorMsg, SelectionDAG &DAG)
static cl::opt< bool > ZeroDivCheck("loongarch-check-zero-division", cl::Hidden, cl::desc("Trap on integer division by zero."), cl::init(false))
static SDValue lower256BitShuffle(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Dispatching routine to lower various 256-bit LoongArch vector shuffles.
static int getEstimateRefinementSteps(EVT VT, const LoongArchSubtarget &Subtarget)
static void emitErrorAndReplaceIntrinsicResults(SDNode *N, SmallVectorImpl< SDValue > &Results, SelectionDAG &DAG, StringRef ErrorMsg, bool WithChain=true)
static SDValue checkIntrinsicImmArg(SDValue Op, unsigned ImmOp, SelectionDAG &DAG, bool IsSigned=false)
static SDValue lowerVECTOR_SHUFFLE_XVSHUF4I(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVSHUF4I (if possible).
static SDValue lowerVECTOR_SHUFFLE_VILVH(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VILVH (if possible).
static Align getPrefTypeAlign(EVT VT, SelectionDAG &DAG)
static bool CC_LoongArch(const DataLayout &DL, LoongArchABI::ABI ABI, unsigned ValNo, MVT ValVT, CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed, bool IsRet, Type *OrigTy)
static SDValue performSRLCombine(SDNode *N, SelectionDAG &DAG, TargetLowering::DAGCombinerInfo &DCI, const LoongArchSubtarget &Subtarget)
static SDValue lowerVectorBitSetImm(SDNode *Node, SelectionDAG &DAG)
static SDValue lowerVECTOR_SHUFFLE_XVPACKOD(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVPACKOD (if possible).
#define CRC_CASE_EXT_BINARYOP(NAME, NODE)
static SDValue lowerVectorBitRevImm(SDNode *Node, SelectionDAG &DAG)
static SDValue truncateVecElts(SDNode *Node, SelectionDAG &DAG)
static MachineBasicBlock * insertDivByZeroTrap(MachineInstr &MI, MachineBasicBlock *MBB)
static SDValue customLegalizeToWOpWithSExt(SDNode *N, SelectionDAG &DAG)
static SDValue lowerVectorBitClear(SDNode *Node, SelectionDAG &DAG)
static bool CC_LoongArch_GHC(unsigned ValNo, MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags, CCState &State)
static SDValue lowerVECTOR_SHUFFLE_VPACKEV(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VPACKEV (if possible).
static void replaceVPICKVE2GRResults(SDNode *Node, SmallVectorImpl< SDValue > &Results, SelectionDAG &DAG, const LoongArchSubtarget &Subtarget, unsigned ResOp)
static SDValue legalizeIntrinsicImmArg(SDNode *Node, unsigned ImmOp, SelectionDAG &DAG, const LoongArchSubtarget &Subtarget, bool IsSigned=false)
static SDValue emitIntrinsicWithChainErrorMessage(SDValue Op, StringRef ErrorMsg, SelectionDAG &DAG)
static bool CC_LoongArchAssign2GRLen(unsigned GRLen, CCState &State, CCValAssign VA1, ISD::ArgFlagsTy ArgFlags1, unsigned ValNo2, MVT ValVT2, MVT LocVT2, ISD::ArgFlagsTy ArgFlags2)
const MCPhysReg ArgFPR64s[]
static MachineBasicBlock * emitPseudoCTPOP(MachineInstr &MI, MachineBasicBlock *BB, const LoongArchSubtarget &Subtarget)
#define IOCSRWR_CASE(NAME, NODE)
#define CRC_CASE_EXT_UNARYOP(NAME, NODE)
static SDValue lowerVECTOR_SHUFFLE_VPACKOD(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VPACKOD (if possible).
static MachineBasicBlock * emitPseudoXVINSGR2VR(MachineInstr &MI, MachineBasicBlock *BB, const LoongArchSubtarget &Subtarget)
static SDValue lowerVectorSplatImm(SDNode *Node, unsigned ImmOp, SelectionDAG &DAG, bool IsSigned=false)
const MCPhysReg ArgGPRs[]
static SDValue lowerVECTOR_SHUFFLE_XVILVL(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVILVL (if possible).
static SDValue customLegalizeToWOp(SDNode *N, SelectionDAG &DAG, int NumOp, unsigned ExtOpc=ISD::ANY_EXTEND)
static void replaceVecCondBranchResults(SDNode *N, SmallVectorImpl< SDValue > &Results, SelectionDAG &DAG, const LoongArchSubtarget &Subtarget, unsigned ResOp)
#define ASRT_LE_GT_CASE(NAME)
static bool isConstantOrUndef(const SDValue Op)
static SDValue lowerVECTOR_SHUFFLE_XVPACKEV(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVPACKEV (if possible).
static MachineBasicBlock * emitVecCondBranchPseudo(MachineInstr &MI, MachineBasicBlock *BB, const LoongArchSubtarget &Subtarget)
static SDValue lowerVECTOR_SHUFFLE_XVILVH(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVILVH (if possible).
static SDValue lowerVECTOR_SHUFFLE_XVSHUF(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVSHUF (if possible).
static SDValue performBITREV_WCombine(SDNode *N, SelectionDAG &DAG, TargetLowering::DAGCombinerInfo &DCI, const LoongArchSubtarget &Subtarget)
#define IOCSRRD_CASE(NAME, NODE)
static SDValue lower128BitShuffle(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Dispatching routine to lower various 128-bit LoongArch vector shuffles.
static SDValue lowerVECTOR_SHUFFLE_XVPICKEV(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVPICKEV (if possible).
static SDValue lowerVECTOR_SHUFFLE_VILVL(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VILVL (if possible).
static SDValue lowerVectorBitClearImm(SDNode *Node, SelectionDAG &DAG)
static bool isConstantOrUndefBUILD_VECTOR(const BuildVectorSDNode *Op)
static void replaceINTRINSIC_WO_CHAINResults(SDNode *N, SmallVectorImpl< SDValue > &Results, SelectionDAG &DAG, const LoongArchSubtarget &Subtarget)
static SDValue lowerVECTOR_SHUFFLE_XVREPLVEI(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into XVREPLVEI (if possible).
static SDValue lowerVECTOR_SHUFFLE_VPICKOD(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VPICKOD (if possible).
static Intrinsic::ID getIntrinsicForMaskedAtomicRMWBinOp(unsigned GRLen, AtomicRMWInst::BinOp BinOp)
static SDValue lowerVECTOR_SHUFFLE_VSHUF(const SDLoc &DL, ArrayRef< int > Mask, MVT VT, SDValue V1, SDValue V2, SelectionDAG &DAG)
Lower VECTOR_SHUFFLE into VSHUF.
static LoongArchISD::NodeType getLoongArchWOpcode(unsigned Opcode)
unsigned const TargetRegisterInfo * TRI
static GCMetadataPrinterRegistry::Add< OcamlGCMetadataPrinter > Y("ocaml", "ocaml 3.10-compatible collector")
static CodeModel::Model getCodeModel(const PPCSubtarget &S, const TargetMachine &TM, const MachineOperand &MO)
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
This file defines the 'Statistic' class, which is designed to be an easy way to expose various metric...
#define STATISTIC(VARNAME, DESC)
Class for arbitrary precision integers.
bool isSubsetOf(const APInt &RHS) const
This operation checks that all bits set in this APInt are also set in RHS.
This class represents an incoming formal argument to a Function.
ArrayRef - Represent a constant reference to an array (0 or more elements consecutively in memory),...
size_t size() const
size - Get the array size.
An instruction that atomically checks whether a specified value is in a memory location,...
Value * getCompareOperand()
AtomicOrdering getFailureOrdering() const
Returns the failure ordering constraint of this cmpxchg instruction.
an instruction that atomically reads a memory location, combines it with another value,...
Align getAlign() const
Return the alignment of the memory that is being allocated by the instruction.
BinOp
This enumeration lists the possible modifications atomicrmw can make.
@ USubCond
Subtract only if no unsigned overflow.
@ Min
*p = old <signed v ? old : v
@ USubSat
*p = usub.sat(old, v) usub.sat matches the behavior of llvm.usub.sat.
@ UIncWrap
Increment one up to a maximum value.
@ Max
*p = old >signed v ? old : v
@ UMin
*p = old <unsigned v ? old : v
@ UMax
*p = old >unsigned v ? old : v
@ UDecWrap
Decrement one until a minimum value or zero.
Value * getPointerOperand()
bool isFloatingPointOperation() const
BinOp getOperation() const
SyncScope::ID getSyncScopeID() const
Returns the synchronization scope ID of this rmw instruction.
AtomicOrdering getOrdering() const
Returns the ordering constraint of this rmw instruction.
LLVM Basic Block Representation.
bool test(unsigned Idx) const
A "pseudo-class" with methods for operating on BUILD_VECTORs.
CCState - This class holds information needed while lowering arguments and return values.
unsigned getFirstUnallocated(ArrayRef< MCPhysReg > Regs) const
getFirstUnallocated - Return the index of the first unallocated register in the set,...
SmallVectorImpl< ISD::ArgFlagsTy > & getPendingArgFlags()
MCRegister AllocateReg(MCPhysReg Reg)
AllocateReg - Attempt to allocate one register.
int64_t AllocateStack(unsigned Size, Align Alignment)
AllocateStack - Allocate a chunk of stack space with the specified size and alignment.
void AnalyzeCallOperands(const SmallVectorImpl< ISD::OutputArg > &Outs, CCAssignFn Fn)
AnalyzeCallOperands - Analyze the outgoing arguments to a call, incorporating info about the passed v...
uint64_t getStackSize() const
Returns the size of the currently allocated portion of the stack.
SmallVectorImpl< CCValAssign > & getPendingLocs()
void AnalyzeFormalArguments(const SmallVectorImpl< ISD::InputArg > &Ins, CCAssignFn Fn)
AnalyzeFormalArguments - Analyze an array of argument values, incorporating info about the formals in...
void addLoc(const CCValAssign &V)
CCValAssign - Represent assignment of one arg/retval to a location.
static CCValAssign getPending(unsigned ValNo, MVT ValVT, MVT LocVT, LocInfo HTP, unsigned ExtraInfo=0)
Register getLocReg() const
LocInfo getLocInfo() const
static CCValAssign getReg(unsigned ValNo, MVT ValVT, MCRegister Reg, MVT LocVT, LocInfo HTP, bool IsCustom=false)
static CCValAssign getMem(unsigned ValNo, MVT ValVT, int64_t Offset, MVT LocVT, LocInfo HTP, bool IsCustom=false)
int64_t getLocMemOffset() const
unsigned getValNo() const
bool isMustTailCall() const
Tests if this call site must be tail call optimized.
This class represents a function call, abstracting a target machine's calling convention.
This is the shared class of boolean and integer constants.
bool isMinusOne() const
This function will return true iff every bit in this constant is set to true.
bool isZero() const
This is just a convenience method to make client code smaller for a common code.
uint64_t getZExtValue() const
int64_t getSExtValue() const
This is an important base class in LLVM.
This class represents an Operation in the Expression.
uint64_t getNumOperands() const
A parsed version of the target data layout string in and methods for querying it.
unsigned getPointerSizeInBits(unsigned AS=0) const
Layout pointer size, in bits FIXME: The defaults need to be removed once all of the backends/clients ...
Align getPrefTypeAlign(Type *Ty) const
Returns the preferred stack/global alignment for the specified type.
FunctionType * getFunctionType() const
Returns the FunctionType for me.
CallingConv::ID getCallingConv() const
getCallingConv()/setCallingConv(CC) - These method get and set the calling convention of this functio...
Argument * getArg(unsigned i) const
Common base class shared among various IRBuilders.
Value * CreateSExt(Value *V, Type *DestTy, const Twine &Name="")
Value * CreateLShr(Value *LHS, Value *RHS, const Twine &Name="", bool isExact=false)
IntegerType * getInt32Ty()
Fetch the type representing a 32-bit integer.
IntegerType * getInt64Ty()
Fetch the type representing a 64-bit integer.
CallInst * CreateIntrinsic(Intrinsic::ID ID, ArrayRef< Type * > Types, ArrayRef< Value * > Args, FMFSource FMFSource={}, const Twine &Name="")
Create a call to intrinsic ID with Args, mangled using Types.
Value * CreateNot(Value *V, const Twine &Name="")
Value * CreateSub(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Value * CreateBitCast(Value *V, Type *DestTy, const Twine &Name="")
ConstantInt * getIntN(unsigned N, uint64_t C)
Get a constant N-bit value, zero extended or truncated from a 64-bit value.
Value * CreateShl(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Value * CreateZExt(Value *V, Type *DestTy, const Twine &Name="", bool IsNonNeg=false)
LLVMContext & getContext() const
Value * CreateAnd(Value *LHS, Value *RHS, const Twine &Name="")
Value * CreatePtrToInt(Value *V, Type *DestTy, const Twine &Name="")
CallInst * CreateCall(FunctionType *FTy, Value *Callee, ArrayRef< Value * > Args={}, const Twine &Name="", MDNode *FPMathTag=nullptr)
AtomicRMWInst * CreateAtomicRMW(AtomicRMWInst::BinOp Op, Value *Ptr, Value *Val, MaybeAlign Align, AtomicOrdering Ordering, SyncScope::ID SSID=SyncScope::System)
Value * CreateTrunc(Value *V, Type *DestTy, const Twine &Name="", bool IsNUW=false, bool IsNSW=false)
Value * CreateOr(Value *LHS, Value *RHS, const Twine &Name="")
This provides a uniform API for creating instructions and inserting them into a basic block: either a...
const Module * getModule() const
Return the module owning the function this instruction belongs to or nullptr it the function does not...
InstListType::iterator eraseFromParent()
This method unlinks 'this' from the containing basic block and deletes it.
const DataLayout & getDataLayout() const
Get the data layout of the module this instruction belongs to.
Class to represent integer types.
This is an important class for using LLVM in a threaded context.
void emitError(const Instruction *I, const Twine &ErrorStr)
emitError - Emit an error message to the currently installed error handler with optional location inf...
This class is used to represent ISD::LOAD nodes.
ISD::LoadExtType getExtensionType() const
Return whether this is a plain node, or one of the varieties of value-extending loads.
LoongArchMachineFunctionInfo - This class is derived from MachineFunctionInfo and contains private Lo...
void addSExt32Register(Register Reg)
const LoongArchRegisterInfo * getRegisterInfo() const override
const LoongArchInstrInfo * getInstrInfo() const override
unsigned getMaxBytesForAlignment() const
Align getPrefFunctionAlignment() const
unsigned getGRLen() const
Align getPrefLoopAlignment() const
bool isUsedByReturnOnly(SDNode *N, SDValue &Chain) const override
Return true if result of the specified node is used by a return node only.
SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const override
This method will be invoked for all target nodes and for any target-independent nodes that the target...
SDValue getSqrtEstimate(SDValue Operand, SelectionDAG &DAG, int Enabled, int &RefinementSteps, bool &UseOneConstNR, bool Reciprocal) const override
Hooks for building estimates in place of slower divisions and square roots.
bool isLegalICmpImmediate(int64_t Imm) const override
Return true if the specified immediate is legal icmp immediate, that is the target has icmp instructi...
TargetLowering::AtomicExpansionKind shouldExpandAtomicCmpXchgInIR(AtomicCmpXchgInst *CI) const override
Returns how the given atomic cmpxchg should be expanded by the IR-level AtomicExpand pass.
Value * emitMaskedAtomicCmpXchgIntrinsic(IRBuilderBase &Builder, AtomicCmpXchgInst *CI, Value *AlignedAddr, Value *CmpVal, Value *NewVal, Value *Mask, AtomicOrdering Ord) const override
Perform a masked cmpxchg using a target-specific intrinsic.
EVT getSetCCResultType(const DataLayout &DL, LLVMContext &Context, EVT VT) const override
Return the ValueType of the result of SETCC operations.
bool isFMAFasterThanFMulAndFAdd(const MachineFunction &MF, EVT VT) const override
Return true if an FMA operation is faster than a pair of fmul and fadd instructions.
SDValue LowerCall(TargetLowering::CallLoweringInfo &CLI, SmallVectorImpl< SDValue > &InVals) const override
This hook must be implemented to lower calls into the specified DAG.
bool decomposeMulByConstant(LLVMContext &Context, EVT VT, SDValue C) const override
Return true if it is profitable to transform an integer multiplication-by-constant into simpler opera...
bool isSExtCheaperThanZExt(EVT SrcVT, EVT DstVT) const override
Return true if sign-extension from FromTy to ToTy is cheaper than zero-extension.
TargetLowering::AtomicExpansionKind shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const override
Returns how the IR-level AtomicExpand pass should expand the given AtomicRMW, if at all.
bool allowsMisalignedMemoryAccesses(EVT VT, unsigned AddrSpace=0, Align Alignment=Align(1), MachineMemOperand::Flags Flags=MachineMemOperand::MONone, unsigned *Fast=nullptr) const override
Determine if the target supports unaligned memory accesses.
bool isCheapToSpeculateCtlz(Type *Ty) const override
Return true if it is cheap to speculate a call to intrinsic ctlz.
SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const override
This callback is invoked for operations that are unsupported by the target, which are registered to u...
bool shouldAlignPointerArgs(CallInst *CI, unsigned &MinSize, Align &PrefAlign) const override
Return true if the pointer arguments to CI should be aligned by aligning the object whose address is ...
Value * emitMaskedAtomicRMWIntrinsic(IRBuilderBase &Builder, AtomicRMWInst *AI, Value *AlignedAddr, Value *Incr, Value *Mask, Value *ShiftAmt, AtomicOrdering Ord) const override
Perform a masked atomicrmw using a target-specific intrinsic.
bool isZExtFree(SDValue Val, EVT VT2) const override
Return true if zero-extending the specific node Val to type VT2 is free (either because it's implicit...
Register getExceptionPointerRegister(const Constant *PersonalityFn) const override
If a physical register, this returns the register that receives the exception address on entry to an ...
bool signExtendConstant(const ConstantInt *CI) const override
Return true if this constant should be sign extended when promoting to a larger type.
const char * getTargetNodeName(unsigned Opcode) const override
This method returns the name of a target specific DAG node.
bool getTgtMemIntrinsic(IntrinsicInfo &Info, const CallInst &I, MachineFunction &MF, unsigned Intrinsic) const override
Given an intrinsic, checks if on the target the intrinsic will need to map to a MemIntrinsicNode (tou...
bool isLegalAddImmediate(int64_t Imm) const override
Return true if the specified immediate is legal add immediate, that is the target has add instruction...
bool isCheapToSpeculateCttz(Type *Ty) const override
Return true if it is cheap to speculate a call to intrinsic cttz.
bool isLegalAddressingMode(const DataLayout &DL, const AddrMode &AM, Type *Ty, unsigned AS, Instruction *I=nullptr) const override
Return true if the addressing mode represented by AM is legal for this target, for a load/store of th...
bool shouldSignExtendTypeInLibCall(Type *Ty, bool IsSigned) const override
Returns true if arguments should be sign-extended in lib calls.
bool isFPImmVLDILegal(const APFloat &Imm, EVT VT) const
bool shouldExtendTypeInLibCall(EVT Type) const override
Returns true if arguments should be extended in lib calls.
Register getRegisterByName(const char *RegName, LLT VT, const MachineFunction &MF) const override
Return the register ID of the name passed in.
bool hasAndNot(SDValue Y) const override
Return true if the target has a bitwise and-not operation: X = ~A & B This can be used to simplify se...
bool isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const override
Return true if folding a constant offset with the given GlobalAddress is legal.
bool CanLowerReturn(CallingConv::ID CallConv, MachineFunction &MF, bool IsVarArg, const SmallVectorImpl< ISD::OutputArg > &Outs, LLVMContext &Context) const override
This hook should be implemented to check whether the return values described by the Outs array can fi...
Register getExceptionSelectorRegister(const Constant *PersonalityFn) const override
If a physical register, this returns the register that receives the exception typeid on entry to a la...
void ReplaceNodeResults(SDNode *N, SmallVectorImpl< SDValue > &Results, SelectionDAG &DAG) const override
This callback is invoked when a node result type is illegal for the target, and the operation was reg...
void emitExpandAtomicRMW(AtomicRMWInst *AI) const override
Perform a atomicrmw expansion using a target-specific way.
ISD::NodeType getExtendForAtomicCmpSwapArg() const override
Returns how the platform's atomic compare and swap expects its comparison value to be extended (ZERO_...
LoongArchTargetLowering(const TargetMachine &TM, const LoongArchSubtarget &STI)
SDValue LowerReturn(SDValue Chain, CallingConv::ID CallConv, bool IsVarArg, const SmallVectorImpl< ISD::OutputArg > &Outs, const SmallVectorImpl< SDValue > &OutVals, const SDLoc &DL, SelectionDAG &DAG) const override
This hook must be implemented to lower outgoing return values, described by the Outs array,...
bool hasAndNotCompare(SDValue Y) const override
Return true if the target should transform: (X & Y) == Y —> (~X & Y) == 0 (X & Y) !...
SDValue getRecipEstimate(SDValue Operand, SelectionDAG &DAG, int Enabled, int &RefinementSteps) const override
Return a reciprocal estimate value for the input operand.
SDValue LowerFormalArguments(SDValue Chain, CallingConv::ID CallConv, bool IsVarArg, const SmallVectorImpl< ISD::InputArg > &Ins, const SDLoc &DL, SelectionDAG &DAG, SmallVectorImpl< SDValue > &InVals) const override
This hook must be implemented to lower the incoming (formal) arguments, described by the Ins array,...
bool mayBeEmittedAsTailCall(const CallInst *CI) const override
Return true if the target may be able emit the call instruction as a tail call.
Wrapper class representing physical registers. Should be passed by value.
bool hasFeature(unsigned Feature) const
bool is128BitVector() const
Return true if this is a 128-bit vector type.
unsigned getVectorNumElements() const
bool isVector() const
Return true if this is a vector value type.
static auto fixedlen_vector_valuetypes()
bool is256BitVector() const
Return true if this is a 256-bit vector type.
bool isScalarInteger() const
Return true if this is an integer, not including vectors.
bool isFloatingPoint() const
Return true if this is a FP or a vector FP type.
static MVT getIntegerVT(unsigned BitWidth)
MVT changeVectorElementTypeToInteger() const
Return a vector with the same number of elements as this vector, but with the element type converted ...
void transferSuccessorsAndUpdatePHIs(MachineBasicBlock *FromMBB)
Transfers all the successors, as in transferSuccessors, and update PHI operands in the successor bloc...
const BasicBlock * getBasicBlock() const
Return the LLVM basic block that this instance corresponded to originally.
void addSuccessor(MachineBasicBlock *Succ, BranchProbability Prob=BranchProbability::getUnknown())
Add Succ as a successor of this MachineBasicBlock.
const MachineFunction * getParent() const
Return the MachineFunction containing this basic block.
void splice(iterator Where, MachineBasicBlock *Other, iterator From)
Take an instruction from MBB 'Other' at the position From, and insert it into this MBB right before '...
The MachineFrameInfo class represents an abstract stack frame until prolog/epilog code is inserted.
int CreateFixedObject(uint64_t Size, int64_t SPOffset, bool IsImmutable, bool isAliased=false)
Create a new object at a fixed location on the stack.
int CreateStackObject(uint64_t Size, Align Alignment, bool isSpillSlot, const AllocaInst *Alloca=nullptr, uint8_t ID=0)
Create a new statically sized stack object, returning a nonnegative identifier to represent it.
void setFrameAddressIsTaken(bool T)
void setHasTailCall(bool V=true)
void setReturnAddressIsTaken(bool s)
const TargetSubtargetInfo & getSubtarget() const
getSubtarget - Return the subtarget for which this machine code is being compiled.
MachineMemOperand * getMachineMemOperand(MachinePointerInfo PtrInfo, MachineMemOperand::Flags f, LLT MemTy, Align base_alignment, const AAMDNodes &AAInfo=AAMDNodes(), const MDNode *Ranges=nullptr, SyncScope::ID SSID=SyncScope::System, AtomicOrdering Ordering=AtomicOrdering::NotAtomic, AtomicOrdering FailureOrdering=AtomicOrdering::NotAtomic)
getMachineMemOperand - Allocate a new MachineMemOperand.
MachineFrameInfo & getFrameInfo()
getFrameInfo - Return the frame info object for the current function.
MachineRegisterInfo & getRegInfo()
getRegInfo - Return information about the registers currently in use.
const DataLayout & getDataLayout() const
Return the DataLayout attached to the Module associated to this MF.
Function & getFunction()
Return the LLVM function that this machine code represents.
Ty * getInfo()
getInfo - Keep track of various per-function pieces of information for backends that would like to do...
Register addLiveIn(MCRegister PReg, const TargetRegisterClass *RC)
addLiveIn - Add the specified physical register as a live-in value and create a corresponding virtual...
MachineBasicBlock * CreateMachineBasicBlock(const BasicBlock *BB=nullptr, std::optional< UniqueBBID > BBID=std::nullopt)
CreateMachineBasicBlock - Allocate a new MachineBasicBlock.
void insert(iterator MBBI, MachineBasicBlock *MBB)
const MachineInstrBuilder & addImm(int64_t Val) const
Add a new immediate operand.
const MachineInstrBuilder & addReg(Register RegNo, unsigned flags=0, unsigned SubReg=0) const
Add a new virtual register operand.
const MachineInstrBuilder & addMBB(MachineBasicBlock *MBB, unsigned TargetFlags=0) const
Representation of each machine instruction.
const MachineOperand & getOperand(unsigned i) const
A description of a memory reference used in the backend.
Flags
Flags values. These may be or'd together.
@ MOVolatile
The memory access is volatile.
@ MODereferenceable
The memory access is dereferenceable (i.e., doesn't trap).
@ MOLoad
The memory access reads data.
@ MOInvariant
The memory access always returns the same value (or traps).
@ MOStore
The memory access writes data.
MachineOperand class - Representation of each machine instruction operand.
void setIsKill(bool Val=true)
void setIsUndef(bool Val=true)
Register getReg() const
getReg - Returns the register number.
static MachineOperand CreateReg(Register Reg, bool isDef, bool isImp=false, bool isKill=false, bool isDead=false, bool isUndef=false, bool isEarlyClobber=false, unsigned SubReg=0, bool isDebug=false, bool isInternalRead=false, bool isRenamable=false)
MachineRegisterInfo - Keep track of information for virtual and physical registers,...
Register createVirtualRegister(const TargetRegisterClass *RegClass, StringRef Name="")
createVirtualRegister - Create and return a new virtual register in the function with the specified r...
void addLiveIn(MCRegister Reg, Register vreg=Register())
addLiveIn - Add the specified register as a live-in.
EVT getMemoryVT() const
Return the type of the in-memory value.
MutableArrayRef - Represent a mutable reference to an array (0 or more elements consecutively in memo...
Class to represent pointers.
unsigned getAddressSpace() const
Return the address space of the Pointer type.
Wrapper class representing virtual and physical registers.
Wrapper class for IR location info (IR ordering and DebugLoc) to be passed into SDNode creation funct...
Represents one node in the SelectionDAG.
unsigned getOpcode() const
Return the SelectionDAG opcode value for this node.
size_t use_size() const
Return the number of uses of this node.
uint64_t getAsZExtVal() const
Helper method returns the zero-extended integer value of a ConstantSDNode.
const SDValue & getOperand(unsigned Num) const
EVT getValueType(unsigned ResNo) const
Return the type of a specified result.
Unlike LLVM values, Selection DAG nodes may return multiple values as the result of a computation.
SDNode * getNode() const
get the SDNode which holds the desired result
SDValue getValue(unsigned R) const
EVT getValueType() const
Return the ValueType of the referenced return value.
const SDValue & getOperand(unsigned i) const
uint64_t getScalarValueSizeInBits() const
uint64_t getConstantOperandVal(unsigned i) const
MVT getSimpleValueType() const
Return the simple ValueType of the referenced return value.
unsigned getOpcode() const
This is used to represent a portion of an LLVM function in a low-level Data Dependence DAG representa...
SDValue getExtLoad(ISD::LoadExtType ExtType, const SDLoc &dl, EVT VT, SDValue Chain, SDValue Ptr, MachinePointerInfo PtrInfo, EVT MemVT, MaybeAlign Alignment=MaybeAlign(), MachineMemOperand::Flags MMOFlags=MachineMemOperand::MONone, const AAMDNodes &AAInfo=AAMDNodes())
SDValue getTargetGlobalAddress(const GlobalValue *GV, const SDLoc &DL, EVT VT, int64_t offset=0, unsigned TargetFlags=0)
SDValue getCopyToReg(SDValue Chain, const SDLoc &dl, Register Reg, SDValue N)
SDValue getMergeValues(ArrayRef< SDValue > Ops, const SDLoc &dl)
Create a MERGE_VALUES node from the given operands.
SDVTList getVTList(EVT VT)
Return an SDVTList that represents the list of values specified.
MachineSDNode * getMachineNode(unsigned Opcode, const SDLoc &dl, EVT VT)
These are used for target selectors to create a new node with specified return type(s),...
SDValue getMemcpy(SDValue Chain, const SDLoc &dl, SDValue Dst, SDValue Src, SDValue Size, Align Alignment, bool isVol, bool AlwaysInline, const CallInst *CI, std::optional< bool > OverrideTailCall, MachinePointerInfo DstPtrInfo, MachinePointerInfo SrcPtrInfo, const AAMDNodes &AAInfo=AAMDNodes(), AAResults *AA=nullptr)
SDValue getSetCC(const SDLoc &DL, EVT VT, SDValue LHS, SDValue RHS, ISD::CondCode Cond, SDValue Chain=SDValue(), bool IsSignaling=false)
Helper function to make it easier to build SetCC's if you just have an ISD::CondCode instead of an SD...
SDValue getRegister(Register Reg, EVT VT)
SDValue getLoad(EVT VT, const SDLoc &dl, SDValue Chain, SDValue Ptr, MachinePointerInfo PtrInfo, MaybeAlign Alignment=MaybeAlign(), MachineMemOperand::Flags MMOFlags=MachineMemOperand::MONone, const AAMDNodes &AAInfo=AAMDNodes(), const MDNode *Ranges=nullptr)
Loads are not normal binary operators: their result type is not determined by their operands,...
void addNoMergeSiteInfo(const SDNode *Node, bool NoMerge)
Set NoMergeSiteInfo to be associated with Node if NoMerge is true.
SDValue getNOT(const SDLoc &DL, SDValue Val, EVT VT)
Create a bitwise NOT operation as (XOR Val, -1).
const TargetLowering & getTargetLoweringInfo() const
SDValue getTargetJumpTable(int JTI, EVT VT, unsigned TargetFlags=0)
SDValue getUNDEF(EVT VT)
Return an UNDEF node. UNDEF does not have a useful SDLoc.
SDValue getCALLSEQ_END(SDValue Chain, SDValue Op1, SDValue Op2, SDValue InGlue, const SDLoc &DL)
Return a new CALLSEQ_END node, which always must have a glue result (to ensure it's not CSE'd).
SDValue getBuildVector(EVT VT, const SDLoc &DL, ArrayRef< SDValue > Ops)
Return an ISD::BUILD_VECTOR node.
bool isSplatValue(SDValue V, const APInt &DemandedElts, APInt &UndefElts, unsigned Depth=0) const
Test whether V has a splatted value for all the demanded elements.
SDValue getBitcast(EVT VT, SDValue V)
Return a bitcast using the SDLoc of the value operand, and casting to the provided type.
SDValue getCopyFromReg(SDValue Chain, const SDLoc &dl, Register Reg, EVT VT)
void setNodeMemRefs(MachineSDNode *N, ArrayRef< MachineMemOperand * > NewMemRefs)
Mutate the specified machine node's memory references to the provided list.
const DataLayout & getDataLayout() const
SDValue getConstant(uint64_t Val, const SDLoc &DL, EVT VT, bool isTarget=false, bool isOpaque=false)
Create a ConstantSDNode wrapping a constant value.
SDValue getSignedTargetConstant(int64_t Val, const SDLoc &DL, EVT VT, bool isOpaque=false)
void ReplaceAllUsesWith(SDValue From, SDValue To)
Modify anything using 'From' to use 'To' instead.
SDValue getCommutedVectorShuffle(const ShuffleVectorSDNode &SV)
Returns an ISD::VECTOR_SHUFFLE node semantically equivalent to the shuffle node in input but with swa...
SDValue getStore(SDValue Chain, const SDLoc &dl, SDValue Val, SDValue Ptr, MachinePointerInfo PtrInfo, Align Alignment, MachineMemOperand::Flags MMOFlags=MachineMemOperand::MONone, const AAMDNodes &AAInfo=AAMDNodes())
Helper function to build ISD::STORE nodes.
SDValue getSignedConstant(int64_t Val, const SDLoc &DL, EVT VT, bool isTarget=false, bool isOpaque=false)
SDValue getCALLSEQ_START(SDValue Chain, uint64_t InSize, uint64_t OutSize, const SDLoc &DL)
Return a new CALLSEQ_START node, that starts new call frame, in which InSize bytes are set up inside ...
SDValue getExternalSymbol(const char *Sym, EVT VT)
const TargetMachine & getTarget() const
SDValue getIntPtrConstant(uint64_t Val, const SDLoc &DL, bool isTarget=false)
SDValue getValueType(EVT)
SDValue getNode(unsigned Opcode, const SDLoc &DL, EVT VT, ArrayRef< SDUse > Ops)
Gets or creates the specified node.
SDValue getTargetConstant(uint64_t Val, const SDLoc &DL, EVT VT, bool isOpaque=false)
SDValue getTargetBlockAddress(const BlockAddress *BA, EVT VT, int64_t Offset=0, unsigned TargetFlags=0)
MachineFunction & getMachineFunction() const
SDValue getFrameIndex(int FI, EVT VT, bool isTarget=false)
KnownBits computeKnownBits(SDValue Op, unsigned Depth=0) const
Determine which bits of Op are known to be either zero or one and return them in Known.
SDValue getRegisterMask(const uint32_t *RegMask)
LLVMContext * getContext() const
SDValue getTargetExternalSymbol(const char *Sym, EVT VT, unsigned TargetFlags=0)
SDValue CreateStackTemporary(TypeSize Bytes, Align Alignment)
Create a stack temporary based on the size in bytes and the alignment.
SDValue getTargetConstantPool(const Constant *C, EVT VT, MaybeAlign Align=std::nullopt, int Offset=0, unsigned TargetFlags=0)
SDValue getEntryNode() const
Return the token chain corresponding to the entry of the function.
SDValue getVectorShuffle(EVT VT, const SDLoc &dl, SDValue N1, SDValue N2, ArrayRef< int > Mask)
Return an ISD::VECTOR_SHUFFLE node.
This SDNode is used to implement the code generator support for the llvm IR shufflevector instruction...
ArrayRef< int > getMask() const
This class consists of common code factored out of the SmallVector class to reduce code duplication b...
typename SuperClass::const_iterator const_iterator
void push_back(const T &Elt)
This is a 'vector' (really, a variable-sized array), optimized for the case when the array is small.
StackOffset holds a fixed and a scalable offset in bytes.
StringRef - Represent a constant reference to a string, i.e.
std::pair< StringRef, StringRef > split(char Separator) const
Split into two substrings around the first occurrence of a separator character.
bool starts_with(StringRef Prefix) const
Check if this string starts with the given Prefix.
constexpr size_t size() const
size - Get the string size.
A switch()-like statement whose cases are string literals.
StringSwitch & Case(StringLiteral S, T Value)
TargetInstrInfo - Interface to description of machine instruction set.
void setBooleanVectorContents(BooleanContent Ty)
Specify how the target extends the result of a vector boolean value from a vector of i1 to a wider ty...
void setOperationAction(unsigned Op, MVT VT, LegalizeAction Action)
Indicate that the specified operation does not work with the specified type and indicate what to do a...
MachineBasicBlock * emitPatchPoint(MachineInstr &MI, MachineBasicBlock *MBB) const
Replace/modify any TargetFrameIndex operands with a targte-dependent sequence of memory operands that...
virtual const TargetRegisterClass * getRegClassFor(MVT VT, bool isDivergent=false) const
Return the register class that should be used for the specified value type.
const TargetMachine & getTargetMachine() const
virtual bool isZExtFree(Type *FromTy, Type *ToTy) const
Return true if any actual instruction that defines a value of type FromTy implicitly zero-extends the...
void setMaxBytesForAlignment(unsigned MaxBytes)
void setPrefLoopAlignment(Align Alignment)
Set the target's preferred loop alignment.
void setMaxAtomicSizeInBitsSupported(unsigned SizeInBits)
Set the maximum atomic operation size supported by the backend.
void setMinFunctionAlignment(Align Alignment)
Set the target's minimum function alignment.
void setBooleanContents(BooleanContent Ty)
Specify how the target extends the result of integer and floating point boolean values from i1 to a w...
void computeRegisterProperties(const TargetRegisterInfo *TRI)
Once all of the register classes are added, this allows us to compute derived properties we expose.
void addRegisterClass(MVT VT, const TargetRegisterClass *RC)
Add the specified register class as an available regclass for the specified value type.
virtual MVT getPointerTy(const DataLayout &DL, uint32_t AS=0) const
Return the pointer type for the given address space, defaults to the pointer type from the data layou...
void setPrefFunctionAlignment(Align Alignment)
Set the target's preferred function alignment.
void setTruncStoreAction(MVT ValVT, MVT MemVT, LegalizeAction Action)
Indicate that the specified truncating store does not work with the specified type and indicate what ...
@ ZeroOrOneBooleanContent
@ ZeroOrNegativeOneBooleanContent
void setMinCmpXchgSizeInBits(unsigned SizeInBits)
Sets the minimum cmpxchg or ll/sc size supported by the backend.
void setStackPointerRegisterToSaveRestore(Register R)
If set to a physical register, this specifies the register that llvm.savestack/llvm....
AtomicExpansionKind
Enum that specifies what an atomic load/AtomicRMWInst is expanded to, if at all.
void setCondCodeAction(ArrayRef< ISD::CondCode > CCs, MVT VT, LegalizeAction Action)
Indicate that the specified condition code is or isn't supported on the target and indicate what to d...
void setTargetDAGCombine(ArrayRef< ISD::NodeType > NTs)
Targets should invoke this method for each target independent node that they want to provide a custom...
void setLoadExtAction(unsigned ExtType, MVT ValVT, MVT MemVT, LegalizeAction Action)
Indicate that the specified load with extension does not work with the specified type and indicate wh...
LegalizeTypeAction getTypeAction(LLVMContext &Context, EVT VT) const
Return how we should legalize values of this type, either it is already legal (return 'Legal') or we ...
std::vector< ArgListEntry > ArgListTy
This class defines information used to lower LLVM code to legal SelectionDAG operators that the targe...
std::pair< SDValue, SDValue > makeLibCall(SelectionDAG &DAG, RTLIB::Libcall LC, EVT RetVT, ArrayRef< SDValue > Ops, MakeLibCallOptions CallOptions, const SDLoc &dl, SDValue Chain=SDValue()) const
Returns a pair of (return value, chain).
virtual InlineAsm::ConstraintCode getInlineAsmMemConstraint(StringRef ConstraintCode) const
virtual ConstraintType getConstraintType(StringRef Constraint) const
Given a constraint, return the type of constraint it is for this target.
std::pair< SDValue, SDValue > LowerCallTo(CallLoweringInfo &CLI) const
This function lowers an abstract call to a function into an actual call.
virtual std::pair< unsigned, const TargetRegisterClass * > getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI, StringRef Constraint, MVT VT) const
Given a physical register constraint (e.g.
bool verifyReturnAddressArgumentIsConstant(SDValue Op, SelectionDAG &DAG) const
virtual void LowerAsmOperandForConstraint(SDValue Op, StringRef Constraint, std::vector< SDValue > &Ops, SelectionDAG &DAG) const
Lower the specified operand into the Ops vector.
Primary interface to the complete machine description for the target machine.
bool useTLSDESC() const
Returns true if this target uses TLS Descriptors.
bool useEmulatedTLS() const
Returns true if this target uses emulated TLS.
bool shouldAssumeDSOLocal(const GlobalValue *GV) const
CodeModel::Model getCodeModel() const
Returns the code model.
TargetRegisterInfo base class - We assume that the target defines a static array of TargetRegisterDes...
virtual const TargetInstrInfo * getInstrInfo() const
Twine - A lightweight data structure for efficiently representing the concatenation of temporary valu...
The instances of the Type class are immutable: once they are created, they are never changed.
unsigned getIntegerBitWidth() const
static IntegerType * getIntNTy(LLVMContext &C, unsigned N)
bool isIntegerTy() const
True if this is an instance of IntegerType.
TypeSize getPrimitiveSizeInBits() const LLVM_READONLY
Return the basic size of this type if it is a primitive type.
This class is used to represent EVT's, which are used to parameterize some operations.
LLVM Value Representation.
Type * getType() const
All values are typed, get the type of this value.
void replaceAllUsesWith(Value *V)
Change all uses of this to point to a new Value.
self_iterator getIterator()
#define llvm_unreachable(msg)
Marks that the current location is not supposed to be reachable.
constexpr char Args[]
Key for Kernel::Metadata::mArgs.
@ GHC
Used by the Glasgow Haskell Compiler (GHC).
@ Fast
Attempts to make calls as fast as possible (e.g.
@ C
The default llvm calling convention, compatible with C.
NodeType
ISD::NodeType enum - This enum defines the target-independent operators for a SelectionDAG.
@ SETCC
SetCC operator - This evaluates to a true value iff the condition is true.
@ STACKRESTORE
STACKRESTORE has two operands, an input chain and a pointer to restore to it returns an output chain.
@ STACKSAVE
STACKSAVE - STACKSAVE has one operand, an input chain.
@ STRICT_FSETCC
STRICT_FSETCC/STRICT_FSETCCS - Constrained versions of SETCC, used for floating-point operands only.
@ SMUL_LOHI
SMUL_LOHI/UMUL_LOHI - Multiply two integers of type iN, producing a signed/unsigned value of type i[2...
@ BSWAP
Byte Swap and Counting operators.
@ VAEND
VAEND, VASTART - VAEND and VASTART have three operands: an input chain, pointer, and a SRCVALUE.
@ ADD
Simple integer binary arithmetic operators.
@ LOAD
LOAD and STORE have token chains as their first operand, then the same operands as an LLVM load/store...
@ ANY_EXTEND
ANY_EXTEND - Used for integer types. The high bits are undefined.
@ FMA
FMA - Perform a * b + c with no intermediate rounding step.
@ INTRINSIC_VOID
OUTCHAIN = INTRINSIC_VOID(INCHAIN, INTRINSICID, arg1, arg2, ...) This node represents a target intrin...
@ SINT_TO_FP
[SU]INT_TO_FP - These operators convert integers (whose interpreted sign depends on the first letter)...
@ CONCAT_VECTORS
CONCAT_VECTORS(VECTOR0, VECTOR1, ...) - Given a number of values of vector type with the same length ...
@ FADD
Simple binary floating point operators.
@ MEMBARRIER
MEMBARRIER - Compiler barrier only; generate a no-op.
@ ATOMIC_FENCE
OUTCHAIN = ATOMIC_FENCE(INCHAIN, ordering, scope) This corresponds to the fence instruction.
@ FP16_TO_FP
FP16_TO_FP, FP_TO_FP16 - These operators are used to perform promotions and truncation for half-preci...
@ BITCAST
BITCAST - This operator converts between integer, vector and FP values, as if the value was stored to...
@ BUILTIN_OP_END
BUILTIN_OP_END - This must be the last enum value in this list.
@ SIGN_EXTEND
Conversion operators.
@ FSINCOS
FSINCOS - Compute both fsin and fcos as a single operation.
@ FNEG
Perform various unary floating-point operations inspired by libm.
@ BR_CC
BR_CC - Conditional branch.
@ BR_JT
BR_JT - Jumptable branch.
@ IS_FPCLASS
Performs a check of floating point class property, defined by IEEE-754.
@ SELECT
Select(COND, TRUEVAL, FALSEVAL).
@ UNDEF
UNDEF - An undefined node.
@ VACOPY
VACOPY - VACOPY has 5 operands: an input chain, a destination pointer, a source pointer,...
@ MULHU
MULHU/MULHS - Multiply high - Multiply two integers of type iN, producing an unsigned/signed value of...
@ SHL
Shift and rotation operations.
@ VECTOR_SHUFFLE
VECTOR_SHUFFLE(VEC1, VEC2) - Returns a vector, of the same type as VEC1/VEC2.
@ FMINNUM_IEEE
FMINNUM_IEEE/FMAXNUM_IEEE - Perform floating-point minimumNumber or maximumNumber on two values,...
@ READ_REGISTER
READ_REGISTER, WRITE_REGISTER - This node represents llvm.register on the DAG, which implements the n...
@ EXTRACT_VECTOR_ELT
EXTRACT_VECTOR_ELT(VECTOR, IDX) - Returns a single element from VECTOR identified by the (potentially...
@ CopyToReg
CopyToReg - This node has three operands: a chain, a register number to set to this value,...
@ ZERO_EXTEND
ZERO_EXTEND - Used for integer types, zeroing the new bits.
@ DEBUGTRAP
DEBUGTRAP - Trap intended to get the attention of a debugger.
@ SELECT_CC
Select with condition operator - This selects between a true value and a false value (ops #2 and #3) ...
@ DYNAMIC_STACKALLOC
DYNAMIC_STACKALLOC - Allocate some number of bytes on the stack aligned to a specified boundary.
@ SIGN_EXTEND_INREG
SIGN_EXTEND_INREG - This operator atomically performs a SHL/SRA pair to sign extend a small value in ...
@ SMIN
[US]{MIN/MAX} - Binary minimum or maximum of signed or unsigned integers.
@ FP_EXTEND
X = FP_EXTEND(Y) - Extend a smaller FP type into a larger FP type.
@ VSELECT
Select with a vector condition (op #0) and two vector operands (ops #1 and #2), returning a vector re...
@ EH_DWARF_CFA
EH_DWARF_CFA - This node represents the pointer to the DWARF Canonical Frame Address (CFA),...
@ FRAMEADDR
FRAMEADDR, RETURNADDR - These nodes represent llvm.frameaddress and llvm.returnaddress on the DAG.
@ FP_TO_SINT
FP_TO_[US]INT - Convert a floating point value to a signed or unsigned integer.
@ AND
Bitwise operators - logical and, logical or, logical xor.
@ TRAP
TRAP - Trapping instruction.
@ INTRINSIC_WO_CHAIN
RESULT = INTRINSIC_WO_CHAIN(INTRINSICID, arg1, arg2, ...) This node represents a target intrinsic fun...
@ INSERT_VECTOR_ELT
INSERT_VECTOR_ELT(VECTOR, VAL, IDX) - Returns VECTOR with the element at IDX replaced with VAL.
@ TokenFactor
TokenFactor - This node takes multiple tokens as input and produces a single token result.
@ TRUNCATE
TRUNCATE - Completely drop the high bits.
@ VAARG
VAARG - VAARG has four operands: an input chain, a pointer, a SRCVALUE, and the alignment.
@ SHL_PARTS
SHL_PARTS/SRA_PARTS/SRL_PARTS - These operators are used for expanded integer shift operations.
@ AssertSext
AssertSext, AssertZext - These nodes record if a register contains a value that has already been zero...
@ INTRINSIC_W_CHAIN
RESULT,OUTCHAIN = INTRINSIC_W_CHAIN(INCHAIN, INTRINSICID, arg1, ...) This node represents a target in...
@ BUILD_VECTOR
BUILD_VECTOR(ELT0, ELT1, ELT2, ELT3,...) - Return a fixed-width vector with the specified,...
CondCode
ISD::CondCode enum - These are ordered carefully to make the bitfields below work out,...
LoadExtType
LoadExtType enum - This enum defines the three variants of LOADEXT (load with extension).
Function * getOrInsertDeclaration(Module *M, ID id, ArrayRef< Type * > Tys={})
Look up the Function declaration of the intrinsic id in the Module M.
ABI getTargetABI(StringRef ABIName)
Libcall getSINTTOFP(EVT OpVT, EVT RetVT)
getSINTTOFP - Return the SINTTOFP_*_* value for the given types, or UNKNOWN_LIBCALL if there is none.
Libcall getUINTTOFP(EVT OpVT, EVT RetVT)
getUINTTOFP - Return the UINTTOFP_*_* value for the given types, or UNKNOWN_LIBCALL if there is none.
Libcall
RTLIB::Libcall enum - This enum defines all of the runtime library calls the backend can emit.
Libcall getFPTOSINT(EVT OpVT, EVT RetVT)
getFPTOSINT - Return the FPTOSINT_*_* value for the given types, or UNKNOWN_LIBCALL if there is none.
@ SingleThread
Synchronized with respect to signal handlers executing in the same thread.
Reg
All possible values of the reg field in the ModR/M byte.
initializer< Ty > init(const Ty &Val)
This is an optimization pass for GlobalISel generic memory operations.
bool all_of(R &&range, UnaryPredicate P)
Provide wrappers to std::all_of which take ranges instead of having to pass begin/end explicitly.
MachineInstrBuilder BuildMI(MachineFunction &MF, const MIMetadata &MIMD, const MCInstrDesc &MCID)
Builder interface. Specify how to create the initial instruction itself.
constexpr bool isShiftedMask_64(uint64_t Value)
Return true if the argument contains a non-empty sequence of ones with the remainder zero (64 bit ver...
bool any_of(R &&range, UnaryPredicate P)
Provide wrappers to std::any_of which take ranges instead of having to pass begin/end explicitly.
raw_ostream & dbgs()
dbgs() - This returns a reference to a raw_ostream for debugging messages.
void report_fatal_error(Error Err, bool gen_crash_diag=true)
Report a serious error, calling any installed error handler.
constexpr bool isUInt(uint64_t x)
Checks if an unsigned integer fits into the given bit width.
AtomicOrdering
Atomic ordering for LLVM's memory model.
unsigned getKillRegState(bool B)
DWARFExpression::Operation Op
constexpr unsigned BitWidth
void swap(llvm::BitVector &LHS, llvm::BitVector &RHS)
Implement std::swap in terms of BitVector swap.
This struct is a compact representation of a valid (non-zero power of two) alignment.
EVT changeVectorElementTypeToInteger() const
Return a vector with the same number of elements as this vector, but with the element type converted ...
TypeSize getStoreSize() const
Return the number of bytes overwritten by a store of the specified value type.
bool isSimple() const
Test if the given EVT is simple (as opposed to being extended).
TypeSize getSizeInBits() const
Return the size of the specified value type in bits.
uint64_t getScalarSizeInBits() const
MVT getSimpleVT() const
Return the SimpleValueType held in the specified simple EVT.
bool is128BitVector() const
Return true if this is a 128-bit vector type.
uint64_t getFixedSizeInBits() const
Return the size of the specified fixed width value type in bits.
static EVT getFloatingPointVT(unsigned BitWidth)
Returns the EVT that represents a floating-point type with the given number of bits.
bool isVector() const
Return true if this is a vector value type.
EVT getScalarType() const
If this is a vector type, return the element type, otherwise return this.
bool is256BitVector() const
Return true if this is a 256-bit vector type.
Type * getTypeForEVT(LLVMContext &Context) const
This method returns an LLVM type corresponding to the specified EVT.
EVT getVectorElementType() const
Given a vector type, return the type of each element.
bool isScalarInteger() const
Return true if this is an integer, but not a vector.
unsigned getVectorNumElements() const
Given a vector type, return the number of elements it contains.
Align getNonZeroOrigAlign() const
Register getFrameRegister(const MachineFunction &MF) const override
BitVector getReservedRegs(const MachineFunction &MF) const override
This class contains a discriminated union of information about pointers in memory operands,...
static MachinePointerInfo getGOT(MachineFunction &MF)
Return a MachinePointerInfo record that refers to a GOT entry.
static MachinePointerInfo getFixedStack(MachineFunction &MF, int FI, int64_t Offset=0)
Return a MachinePointerInfo record that refers to the specified FrameIndex.
This represents a list of ValueType's that has been intern'd by a SelectionDAG.
This represents an addressing mode of: BaseGV + BaseOffs + BaseReg + Scale*ScaleReg + ScalableOffset*...
This structure contains all information that is necessary for lowering calls.
SmallVector< ISD::InputArg, 32 > Ins
SmallVector< ISD::OutputArg, 32 > Outs
SmallVector< SDValue, 32 > OutVals
bool isBeforeLegalizeOps() const
This structure is used to pass arguments to makeLibCall function.
MakeLibCallOptions & setTypeListBeforeSoften(ArrayRef< EVT > OpsVT, EVT RetVT, bool Value=true)