aboutsummaryrefslogtreecommitdiff
path: root/llvm/lib/CodeGen
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib/CodeGen')
-rw-r--r--llvm/lib/CodeGen/BranchFolding.cpp48
-rw-r--r--llvm/lib/CodeGen/CommandFlags.cpp7
-rw-r--r--llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp24
-rw-r--r--llvm/lib/CodeGen/GlobalISel/CallLowering.cpp6
-rw-r--r--llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp6
-rw-r--r--llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp4
-rw-r--r--llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp20
-rw-r--r--llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp22
-rw-r--r--llvm/lib/CodeGen/MIRParser/MILexer.cpp2
-rw-r--r--llvm/lib/CodeGen/MIRParser/MILexer.h2
-rw-r--r--llvm/lib/CodeGen/MIRParser/MIParser.cpp12
-rw-r--r--llvm/lib/CodeGen/MIRParser/MIRParser.cpp14
-rw-r--r--llvm/lib/CodeGen/MIRPrinter.cpp14
-rw-r--r--llvm/lib/CodeGen/MachineFunction.cpp5
-rw-r--r--llvm/lib/CodeGen/MachineInstr.cpp6
-rw-r--r--llvm/lib/CodeGen/MachineOperand.cpp4
-rw-r--r--llvm/lib/CodeGen/ModuloSchedule.cpp2
-rw-r--r--llvm/lib/CodeGen/RegisterCoalescer.cpp179
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp138
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp2
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp93
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp10
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp2
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp2
-rw-r--r--llvm/lib/CodeGen/TargetLoweringBase.cpp36
-rw-r--r--llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp17
-rw-r--r--llvm/lib/CodeGen/WindowsSecureHotPatching.cpp13
27 files changed, 536 insertions, 154 deletions
diff --git a/llvm/lib/CodeGen/BranchFolding.cpp b/llvm/lib/CodeGen/BranchFolding.cpp
index 3b3e7a4..dcfd9aa 100644
--- a/llvm/lib/CodeGen/BranchFolding.cpp
+++ b/llvm/lib/CodeGen/BranchFolding.cpp
@@ -2083,22 +2083,55 @@ bool BranchFolder::HoistCommonCodeInSuccs(MachineBasicBlock *MBB) {
if (TBB == FBB) {
MBB->splice(Loc, TBB, TBB->begin(), TIB);
} else {
+ // Merge the debug locations, and hoist and kill the debug instructions from
+ // both branches. FIXME: We could probably try harder to preserve some debug
+ // instructions (but at least this isn't producing wrong locations).
+ MachineInstrBuilder MIRBuilder(*MBB->getParent(), Loc);
+ auto HoistAndKillDbgInstr = [MBB, Loc](MachineBasicBlock::iterator DI) {
+ assert(DI->isDebugInstr() && "Expected a debug instruction");
+ if (DI->isDebugRef()) {
+ const TargetInstrInfo *TII =
+ MBB->getParent()->getSubtarget().getInstrInfo();
+ const MCInstrDesc &DBGV = TII->get(TargetOpcode::DBG_VALUE);
+ DI = BuildMI(*MBB->getParent(), DI->getDebugLoc(), DBGV, false, 0,
+ DI->getDebugVariable(), DI->getDebugExpression());
+ MBB->insert(Loc, &*DI);
+ return;
+ }
+ // Deleting a DBG_PHI results in an undef at the referenced DBG_INSTR_REF.
+ if (DI->isDebugPHI()) {
+ DI->eraseFromParent();
+ return;
+ }
+ // Move DBG_LABELs without modifying them. Set DBG_VALUEs undef.
+ if (!DI->isDebugLabel())
+ DI->setDebugValueUndef();
+ DI->moveBefore(&*Loc);
+ };
+
// TIB and FIB point to the end of the regions to hoist/merge in TBB and
// FBB.
MachineBasicBlock::iterator FE = FIB;
MachineBasicBlock::iterator FI = FBB->begin();
for (MachineBasicBlock::iterator TI :
make_early_inc_range(make_range(TBB->begin(), TIB))) {
- // Move debug instructions and pseudo probes without modifying them.
- // FIXME: This is the wrong thing to do for debug locations, which
- // should at least be killed (and hoisted from BOTH blocks).
- if (TI->isDebugOrPseudoInstr()) {
- TI->moveBefore(&*Loc);
+ // Hoist and kill debug instructions from FBB. After this loop FI points
+ // to the next non-debug instruction to hoist (checked in assert after the
+ // TBB debug instruction handling code).
+ while (FI != FE && FI->isDebugInstr())
+ HoistAndKillDbgInstr(FI++);
+
+ // Kill debug instructions before moving.
+ if (TI->isDebugInstr()) {
+ HoistAndKillDbgInstr(TI);
continue;
}
- // Get the next non-meta instruction in FBB.
- FI = skipDebugInstructionsForward(FI, FE, false);
+ // FI and TI now point to identical non-debug instructions.
+ assert(FI != FE && "Unexpected end of FBB range");
+ // Pseudo probes are excluded from the range when identifying foldable
+ // instructions, so we don't expect to see one now.
+ assert(!TI->isPseudoProbe() && "Unexpected pseudo probe in range");
// NOTE: The loop above checks CheckKillDead but we can't do that here as
// it modifies some kill markers after the check.
assert(TI->isIdenticalTo(*FI, MachineInstr::CheckDefs) &&
@@ -2111,6 +2144,7 @@ bool BranchFolder::HoistCommonCodeInSuccs(MachineBasicBlock *MBB) {
++FI;
}
}
+
FBB->erase(FBB->begin(), FIB);
if (UpdateLiveIns)
diff --git a/llvm/lib/CodeGen/CommandFlags.cpp b/llvm/lib/CodeGen/CommandFlags.cpp
index 9512f79..810dc29 100644
--- a/llvm/lib/CodeGen/CommandFlags.cpp
+++ b/llvm/lib/CodeGen/CommandFlags.cpp
@@ -101,6 +101,7 @@ CGOPT(EABI, EABIVersion)
CGOPT(DebuggerKind, DebuggerTuningOpt)
CGOPT(bool, EnableStackSizeSection)
CGOPT(bool, EnableAddrsig)
+CGOPT(bool, EnableCallGraphSection)
CGOPT(bool, EmitCallSiteInfo)
CGOPT(bool, EnableMachineFunctionSplitter)
CGOPT(bool, EnableStaticDataPartitioning)
@@ -461,6 +462,11 @@ codegen::RegisterCodeGenFlags::RegisterCodeGenFlags() {
cl::init(false));
CGBINDOPT(EnableAddrsig);
+ static cl::opt<bool> EnableCallGraphSection(
+ "call-graph-section", cl::desc("Emit a call graph section"),
+ cl::init(false));
+ CGBINDOPT(EnableCallGraphSection);
+
static cl::opt<bool> EmitCallSiteInfo(
"emit-call-site-info",
cl::desc(
@@ -595,6 +601,7 @@ codegen::InitTargetOptionsFromCodeGenFlags(const Triple &TheTriple) {
Options.EnableMachineFunctionSplitter = getEnableMachineFunctionSplitter();
Options.EnableStaticDataPartitioning = getEnableStaticDataPartitioning();
Options.EmitAddrsig = getEnableAddrsig();
+ Options.EmitCallGraphSection = getEnableCallGraphSection();
Options.EmitCallSiteInfo = getEmitCallSiteInfo();
Options.EnableDebugEntryValues = getEnableDebugEntryValues();
Options.ForceDwarfFrameSection = getForceDwarfFrameSection();
diff --git a/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp b/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp
index 8855740f..9b2851e 100644
--- a/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp
+++ b/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp
@@ -2186,19 +2186,16 @@ Value *ComplexDeinterleavingGraph::replaceNode(IRBuilderBase &Builder,
llvm_unreachable("Deinterleave node should already have ReplacementNode");
break;
case ComplexDeinterleavingOperation::Splat: {
- auto *NewTy = VectorType::getDoubleElementsVectorType(
- cast<VectorType>(Node->Real->getType()));
auto *R = dyn_cast<Instruction>(Node->Real);
auto *I = dyn_cast<Instruction>(Node->Imag);
if (R && I) {
// Splats that are not constant are interleaved where they are located
Instruction *InsertPoint = (I->comesBefore(R) ? R : I)->getNextNode();
IRBuilder<> IRB(InsertPoint);
- ReplacementNode = IRB.CreateIntrinsic(Intrinsic::vector_interleave2,
- NewTy, {Node->Real, Node->Imag});
+ ReplacementNode = IRB.CreateVectorInterleave({Node->Real, Node->Imag});
} else {
- ReplacementNode = Builder.CreateIntrinsic(
- Intrinsic::vector_interleave2, NewTy, {Node->Real, Node->Imag});
+ ReplacementNode =
+ Builder.CreateVectorInterleave({Node->Real, Node->Imag});
}
break;
}
@@ -2226,10 +2223,7 @@ Value *ComplexDeinterleavingGraph::replaceNode(IRBuilderBase &Builder,
auto *MaskImag = cast<Instruction>(Node->Imag)->getOperand(0);
auto *A = replaceNode(Builder, Node->Operands[0]);
auto *B = replaceNode(Builder, Node->Operands[1]);
- auto *NewMaskTy = VectorType::getDoubleElementsVectorType(
- cast<VectorType>(MaskReal->getType()));
- auto *NewMask = Builder.CreateIntrinsic(Intrinsic::vector_interleave2,
- NewMaskTy, {MaskReal, MaskImag});
+ auto *NewMask = Builder.CreateVectorInterleave({MaskReal, MaskImag});
ReplacementNode = Builder.CreateSelect(NewMask, A, B);
break;
}
@@ -2260,8 +2254,8 @@ void ComplexDeinterleavingGraph::processReductionSingle(
}
if (!NewInit)
- NewInit = Builder.CreateIntrinsic(Intrinsic::vector_interleave2, NewVTy,
- {Init, Constant::getNullValue(VTy)});
+ NewInit =
+ Builder.CreateVectorInterleave({Init, Constant::getNullValue(VTy)});
NewPHI->addIncoming(NewInit, Incoming);
NewPHI->addIncoming(OperationReplacement, BackEdge);
@@ -2281,16 +2275,12 @@ void ComplexDeinterleavingGraph::processReductionOperation(
auto *OldPHIImag = ReductionInfo[Imag].first;
auto *NewPHI = OldToNewPHI[OldPHIReal];
- auto *VTy = cast<VectorType>(Real->getType());
- auto *NewVTy = VectorType::getDoubleElementsVectorType(VTy);
-
// We have to interleave initial origin values coming from IncomingBlock
Value *InitReal = OldPHIReal->getIncomingValueForBlock(Incoming);
Value *InitImag = OldPHIImag->getIncomingValueForBlock(Incoming);
IRBuilder<> Builder(Incoming->getTerminator());
- auto *NewInit = Builder.CreateIntrinsic(Intrinsic::vector_interleave2, NewVTy,
- {InitReal, InitImag});
+ auto *NewInit = Builder.CreateVectorInterleave({InitReal, InitImag});
NewPHI->addIncoming(NewInit, Incoming);
NewPHI->addIncoming(OperationReplacement, BackEdge);
diff --git a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
index 012d873..9ba1782 100644
--- a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
@@ -1009,7 +1009,8 @@ void CallLowering::insertSRetLoads(MachineIRBuilder &MIRBuilder, Type *RetTy,
for (unsigned I = 0; I < NumValues; ++I) {
Register Addr;
- MIRBuilder.materializePtrAdd(Addr, DemoteReg, OffsetLLTy, Offsets[I]);
+ MIRBuilder.materializeObjectPtrOffset(Addr, DemoteReg, OffsetLLTy,
+ Offsets[I]);
auto *MMO = MF.getMachineMemOperand(PtrInfo, MachineMemOperand::MOLoad,
MRI.getType(VRegs[I]),
commonAlignment(BaseAlign, Offsets[I]));
@@ -1039,7 +1040,8 @@ void CallLowering::insertSRetStores(MachineIRBuilder &MIRBuilder, Type *RetTy,
for (unsigned I = 0; I < NumValues; ++I) {
Register Addr;
- MIRBuilder.materializePtrAdd(Addr, DemoteReg, OffsetLLTy, Offsets[I]);
+ MIRBuilder.materializeObjectPtrOffset(Addr, DemoteReg, OffsetLLTy,
+ Offsets[I]);
auto *MMO = MF.getMachineMemOperand(PtrInfo, MachineMemOperand::MOStore,
MRI.getType(VRegs[I]),
commonAlignment(BaseAlign, Offsets[I]));
diff --git a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
index e8f513a..e84ba91 100644
--- a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
@@ -5949,8 +5949,7 @@ bool CombinerHelper::canCombineFMadOrFMA(MachineInstr &MI,
const TargetOptions &Options = MF->getTarget().Options;
LLT DstType = MRI.getType(MI.getOperand(0).getReg());
- if (CanReassociate &&
- !(Options.UnsafeFPMath || MI.getFlag(MachineInstr::MIFlag::FmReassoc)))
+ if (CanReassociate && !MI.getFlag(MachineInstr::MIFlag::FmReassoc))
return false;
// Floating-point multiply-add with intermediate rounding.
@@ -5962,8 +5961,7 @@ bool CombinerHelper::canCombineFMadOrFMA(MachineInstr &MI,
if (!HasFMAD && !HasFMA)
return false;
- AllowFusionGlobally = Options.AllowFPOpFusion == FPOpFusion::Fast ||
- Options.UnsafeFPMath || HasFMAD;
+ AllowFusionGlobally = Options.AllowFPOpFusion == FPOpFusion::Fast || HasFMAD;
// If the addition is not contractable, do not combine.
if (!AllowFusionGlobally && !MI.getFlag(MachineInstr::MIFlag::FmContract))
return false;
diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index dc5dfab..fd38c30 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -1409,7 +1409,7 @@ bool IRTranslator::translateLoad(const User &U, MachineIRBuilder &MIRBuilder) {
Regs.size() == 1 ? LI.getMetadata(LLVMContext::MD_range) : nullptr;
for (unsigned i = 0; i < Regs.size(); ++i) {
Register Addr;
- MIRBuilder.materializePtrAdd(Addr, Base, OffsetTy, Offsets[i] / 8);
+ MIRBuilder.materializeObjectPtrOffset(Addr, Base, OffsetTy, Offsets[i] / 8);
MachinePointerInfo Ptr(LI.getPointerOperand(), Offsets[i] / 8);
Align BaseAlign = getMemOpAlign(LI);
@@ -1448,7 +1448,7 @@ bool IRTranslator::translateStore(const User &U, MachineIRBuilder &MIRBuilder) {
for (unsigned i = 0; i < Vals.size(); ++i) {
Register Addr;
- MIRBuilder.materializePtrAdd(Addr, Base, OffsetTy, Offsets[i] / 8);
+ MIRBuilder.materializeObjectPtrOffset(Addr, Base, OffsetTy, Offsets[i] / 8);
MachinePointerInfo Ptr(SI.getPointerOperand(), Offsets[i] / 8);
Align BaseAlign = getMemOpAlign(SI);
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index ed7b07f..d9d3569 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -4170,7 +4170,7 @@ LegalizerHelper::LegalizeResult LegalizerHelper::lowerLoad(GAnyLoad &LoadMI) {
auto OffsetCst = MIRBuilder.buildConstant(LLT::scalar(PtrTy.getSizeInBits()),
LargeSplitSize / 8);
Register PtrAddReg = MRI.createGenericVirtualRegister(PtrTy);
- auto SmallPtr = MIRBuilder.buildPtrAdd(PtrAddReg, PtrReg, OffsetCst);
+ auto SmallPtr = MIRBuilder.buildObjectPtrOffset(PtrAddReg, PtrReg, OffsetCst);
auto SmallLoad = MIRBuilder.buildLoadInstr(LoadMI.getOpcode(), AnyExtTy,
SmallPtr, *SmallMMO);
@@ -4277,8 +4277,7 @@ LegalizerHelper::LegalizeResult LegalizerHelper::lowerStore(GStore &StoreMI) {
LLT PtrTy = MRI.getType(PtrReg);
auto OffsetCst = MIRBuilder.buildConstant(
LLT::scalar(PtrTy.getSizeInBits()), LargeSplitSize / 8);
- auto SmallPtr =
- MIRBuilder.buildPtrAdd(PtrTy, PtrReg, OffsetCst);
+ auto SmallPtr = MIRBuilder.buildObjectPtrOffset(PtrTy, PtrReg, OffsetCst);
MachineMemOperand *LargeMMO =
MF.getMachineMemOperand(&MMO, 0, LargeSplitSize / 8);
@@ -5349,7 +5348,8 @@ LegalizerHelper::reduceLoadStoreWidth(GLoadStore &LdStMI, unsigned TypeIdx,
unsigned ByteOffset = Offset / 8;
Register NewAddrReg;
- MIRBuilder.materializePtrAdd(NewAddrReg, AddrReg, OffsetTy, ByteOffset);
+ MIRBuilder.materializeObjectPtrOffset(NewAddrReg, AddrReg, OffsetTy,
+ ByteOffset);
MachineMemOperand *NewMMO =
MF.getMachineMemOperand(&MMO, ByteOffset, PartTy);
@@ -8004,7 +8004,7 @@ LegalizerHelper::lowerFPTRUNC_F64_TO_F16(MachineInstr &MI) {
if (MRI.getType(Src).isVector()) // TODO: Handle vectors directly.
return UnableToLegalize;
- if (MIRBuilder.getMF().getTarget().Options.UnsafeFPMath) {
+ if (MI.getFlag(MachineInstr::FmAfn)) {
unsigned Flags = MI.getFlags();
auto Src32 = MIRBuilder.buildFPTrunc(S32, Src, Flags);
MIRBuilder.buildFPTrunc(Dst, Src32, Flags);
@@ -9822,7 +9822,7 @@ LegalizerHelper::lowerMemset(MachineInstr &MI, Register Dst, Register Val,
if (DstOff != 0) {
auto Offset =
MIB.buildConstant(LLT::scalar(PtrTy.getSizeInBits()), DstOff);
- Ptr = MIB.buildPtrAdd(PtrTy, Dst, Offset).getReg(0);
+ Ptr = MIB.buildObjectPtrOffset(PtrTy, Dst, Offset).getReg(0);
}
MIB.buildStore(Value, Ptr, *StoreMMO);
@@ -9962,7 +9962,7 @@ LegalizerHelper::lowerMemcpy(MachineInstr &MI, Register Dst, Register Src,
LLT SrcTy = MRI.getType(Src);
Offset = MIB.buildConstant(LLT::scalar(SrcTy.getSizeInBits()), CurrOffset)
.getReg(0);
- LoadPtr = MIB.buildPtrAdd(SrcTy, Src, Offset).getReg(0);
+ LoadPtr = MIB.buildObjectPtrOffset(SrcTy, Src, Offset).getReg(0);
}
auto LdVal = MIB.buildLoad(CopyTy, LoadPtr, *LoadMMO);
@@ -9970,7 +9970,7 @@ LegalizerHelper::lowerMemcpy(MachineInstr &MI, Register Dst, Register Src,
Register StorePtr = Dst;
if (CurrOffset != 0) {
LLT DstTy = MRI.getType(Dst);
- StorePtr = MIB.buildPtrAdd(DstTy, Dst, Offset).getReg(0);
+ StorePtr = MIB.buildObjectPtrOffset(DstTy, Dst, Offset).getReg(0);
}
MIB.buildStore(LdVal, StorePtr, *StoreMMO);
CurrOffset += CopyTy.getSizeInBytes();
@@ -10060,7 +10060,7 @@ LegalizerHelper::lowerMemmove(MachineInstr &MI, Register Dst, Register Src,
LLT SrcTy = MRI.getType(Src);
auto Offset =
MIB.buildConstant(LLT::scalar(SrcTy.getSizeInBits()), CurrOffset);
- LoadPtr = MIB.buildPtrAdd(SrcTy, Src, Offset).getReg(0);
+ LoadPtr = MIB.buildObjectPtrOffset(SrcTy, Src, Offset).getReg(0);
}
LoadVals.push_back(MIB.buildLoad(CopyTy, LoadPtr, *LoadMMO).getReg(0));
CurrOffset += CopyTy.getSizeInBytes();
@@ -10078,7 +10078,7 @@ LegalizerHelper::lowerMemmove(MachineInstr &MI, Register Dst, Register Src,
LLT DstTy = MRI.getType(Dst);
auto Offset =
MIB.buildConstant(LLT::scalar(DstTy.getSizeInBits()), CurrOffset);
- StorePtr = MIB.buildPtrAdd(DstTy, Dst, Offset).getReg(0);
+ StorePtr = MIB.buildObjectPtrOffset(DstTy, Dst, Offset).getReg(0);
}
MIB.buildStore(LoadVals[I], StorePtr, *StoreMMO);
CurrOffset += CopyTy.getSizeInBytes();
diff --git a/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp b/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
index 121d7e8..27df7e3 100644
--- a/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
@@ -208,11 +208,20 @@ MachineIRBuilder::buildPtrAdd(const DstOp &Res, const SrcOp &Op0,
return buildInstr(TargetOpcode::G_PTR_ADD, {Res}, {Op0, Op1}, Flags);
}
+MachineInstrBuilder MachineIRBuilder::buildObjectPtrOffset(const DstOp &Res,
+ const SrcOp &Op0,
+ const SrcOp &Op1) {
+ return buildPtrAdd(Res, Op0, Op1,
+ MachineInstr::MIFlag::NoUWrap |
+ MachineInstr::MIFlag::InBounds);
+}
+
std::optional<MachineInstrBuilder>
MachineIRBuilder::materializePtrAdd(Register &Res, Register Op0,
- const LLT ValueTy, uint64_t Value) {
+ const LLT ValueTy, uint64_t Value,
+ std::optional<unsigned> Flags) {
assert(Res == 0 && "Res is a result argument");
- assert(ValueTy.isScalar() && "invalid offset type");
+ assert(ValueTy.isScalar() && "invalid offset type");
if (Value == 0) {
Res = Op0;
@@ -221,7 +230,14 @@ MachineIRBuilder::materializePtrAdd(Register &Res, Register Op0,
Res = getMRI()->createGenericVirtualRegister(getMRI()->getType(Op0));
auto Cst = buildConstant(ValueTy, Value);
- return buildPtrAdd(Res, Op0, Cst.getReg(0));
+ return buildPtrAdd(Res, Op0, Cst.getReg(0), Flags);
+}
+
+std::optional<MachineInstrBuilder> MachineIRBuilder::materializeObjectPtrOffset(
+ Register &Res, Register Op0, const LLT ValueTy, uint64_t Value) {
+ return materializePtrAdd(Res, Op0, ValueTy, Value,
+ MachineInstr::MIFlag::NoUWrap |
+ MachineInstr::MIFlag::InBounds);
}
MachineInstrBuilder MachineIRBuilder::buildMaskLowPtrBits(const DstOp &Res,
diff --git a/llvm/lib/CodeGen/MIRParser/MILexer.cpp b/llvm/lib/CodeGen/MIRParser/MILexer.cpp
index 7153902..8b72c29 100644
--- a/llvm/lib/CodeGen/MIRParser/MILexer.cpp
+++ b/llvm/lib/CodeGen/MIRParser/MILexer.cpp
@@ -217,6 +217,7 @@ static MIToken::TokenKind getIdentifierKind(StringRef Identifier) {
.Case("nneg", MIToken::kw_nneg)
.Case("disjoint", MIToken::kw_disjoint)
.Case("samesign", MIToken::kw_samesign)
+ .Case("inbounds", MIToken::kw_inbounds)
.Case("nofpexcept", MIToken::kw_nofpexcept)
.Case("unpredictable", MIToken::kw_unpredictable)
.Case("debug-location", MIToken::kw_debug_location)
@@ -616,6 +617,7 @@ static MIToken::TokenKind getMetadataKeywordKind(StringRef Identifier) {
.Case("!range", MIToken::md_range)
.Case("!DIExpression", MIToken::md_diexpr)
.Case("!DILocation", MIToken::md_dilocation)
+ .Case("!noalias.addrspace", MIToken::md_noalias_addrspace)
.Default(MIToken::Error);
}
diff --git a/llvm/lib/CodeGen/MIRParser/MILexer.h b/llvm/lib/CodeGen/MIRParser/MILexer.h
index d7cd067..0627f17 100644
--- a/llvm/lib/CodeGen/MIRParser/MILexer.h
+++ b/llvm/lib/CodeGen/MIRParser/MILexer.h
@@ -78,6 +78,7 @@ struct MIToken {
kw_nneg,
kw_disjoint,
kw_samesign,
+ kw_inbounds,
kw_debug_location,
kw_debug_instr_number,
kw_dbg_instr_ref,
@@ -151,6 +152,7 @@ struct MIToken {
md_tbaa,
md_alias_scope,
md_noalias,
+ md_noalias_addrspace,
md_range,
md_diexpr,
md_dilocation,
diff --git a/llvm/lib/CodeGen/MIRParser/MIParser.cpp b/llvm/lib/CodeGen/MIRParser/MIParser.cpp
index 3a364d5..6a464d9 100644
--- a/llvm/lib/CodeGen/MIRParser/MIParser.cpp
+++ b/llvm/lib/CodeGen/MIRParser/MIParser.cpp
@@ -1477,7 +1477,8 @@ bool MIParser::parseInstruction(unsigned &OpCode, unsigned &Flags) {
Token.is(MIToken::kw_nneg) ||
Token.is(MIToken::kw_disjoint) ||
Token.is(MIToken::kw_nusw) ||
- Token.is(MIToken::kw_samesign)) {
+ Token.is(MIToken::kw_samesign) ||
+ Token.is(MIToken::kw_inbounds)) {
// clang-format on
// Mine frame and fast math flags
if (Token.is(MIToken::kw_frame_setup))
@@ -1518,6 +1519,8 @@ bool MIParser::parseInstruction(unsigned &OpCode, unsigned &Flags) {
Flags |= MachineInstr::NoUSWrap;
if (Token.is(MIToken::kw_samesign))
Flags |= MachineInstr::SameSign;
+ if (Token.is(MIToken::kw_inbounds))
+ Flags |= MachineInstr::InBounds;
lex();
}
@@ -3482,6 +3485,11 @@ bool MIParser::parseMachineMemoryOperand(MachineMemOperand *&Dest) {
if (parseMDNode(AAInfo.NoAlias))
return true;
break;
+ case MIToken::md_noalias_addrspace:
+ lex();
+ if (parseMDNode(AAInfo.NoAliasAddrSpace))
+ return true;
+ break;
case MIToken::md_range:
lex();
if (parseMDNode(Range))
@@ -3490,7 +3498,7 @@ bool MIParser::parseMachineMemoryOperand(MachineMemOperand *&Dest) {
// TODO: Report an error on duplicate metadata nodes.
default:
return error("expected 'align' or '!tbaa' or '!alias.scope' or "
- "'!noalias' or '!range'");
+ "'!noalias' or '!range' or '!noalias.addrspace'");
}
}
if (expectAndConsume(MIToken::rparen))
diff --git a/llvm/lib/CodeGen/MIRParser/MIRParser.cpp b/llvm/lib/CodeGen/MIRParser/MIRParser.cpp
index 1e9fcf3..3e99e57 100644
--- a/llvm/lib/CodeGen/MIRParser/MIRParser.cpp
+++ b/llvm/lib/CodeGen/MIRParser/MIRParser.cpp
@@ -504,13 +504,21 @@ bool MIRParserImpl::initializeCallSiteInfo(
return error(Error, ArgRegPair.Reg.SourceRange);
CSInfo.ArgRegPairs.emplace_back(Reg, ArgRegPair.ArgNo);
}
+ if (!YamlCSInfo.CalleeTypeIds.empty()) {
+ for (auto CalleeTypeId : YamlCSInfo.CalleeTypeIds) {
+ IntegerType *Int64Ty = Type::getInt64Ty(Context);
+ CSInfo.CalleeTypeIds.push_back(ConstantInt::get(Int64Ty, CalleeTypeId,
+ /*isSigned=*/false));
+ }
+ }
- if (TM.Options.EmitCallSiteInfo)
+ if (TM.Options.EmitCallSiteInfo || TM.Options.EmitCallGraphSection)
MF.addCallSiteInfo(&*CallI, std::move(CSInfo));
}
- if (YamlMF.CallSitesInfo.size() && !TM.Options.EmitCallSiteInfo)
- return error(Twine("Call site info provided but not used"));
+ if (!YamlMF.CallSitesInfo.empty() &&
+ !(TM.Options.EmitCallSiteInfo || TM.Options.EmitCallGraphSection))
+ return error("call site info provided but not used");
return false;
}
diff --git a/llvm/lib/CodeGen/MIRPrinter.cpp b/llvm/lib/CodeGen/MIRPrinter.cpp
index bc4e299..ce1834a 100644
--- a/llvm/lib/CodeGen/MIRPrinter.cpp
+++ b/llvm/lib/CodeGen/MIRPrinter.cpp
@@ -525,24 +525,30 @@ static void convertCallSiteObjects(yaml::MachineFunction &YMF,
const MachineFunction &MF,
ModuleSlotTracker &MST) {
const auto *TRI = MF.getSubtarget().getRegisterInfo();
- for (auto CSInfo : MF.getCallSitesInfo()) {
+ for (auto [MI, CallSiteInfo] : MF.getCallSitesInfo()) {
yaml::CallSiteInfo YmlCS;
yaml::MachineInstrLoc CallLocation;
// Prepare instruction position.
- MachineBasicBlock::const_instr_iterator CallI = CSInfo.first->getIterator();
+ MachineBasicBlock::const_instr_iterator CallI = MI->getIterator();
CallLocation.BlockNum = CallI->getParent()->getNumber();
// Get call instruction offset from the beginning of block.
CallLocation.Offset =
std::distance(CallI->getParent()->instr_begin(), CallI);
YmlCS.CallLocation = CallLocation;
+
+ auto [ArgRegPairs, CalleeTypeIds] = CallSiteInfo;
// Construct call arguments and theirs forwarding register info.
- for (auto ArgReg : CSInfo.second.ArgRegPairs) {
+ for (auto ArgReg : ArgRegPairs) {
yaml::CallSiteInfo::ArgRegPair YmlArgReg;
YmlArgReg.ArgNo = ArgReg.ArgNo;
printRegMIR(ArgReg.Reg, YmlArgReg.Reg, TRI);
YmlCS.ArgForwardingRegs.emplace_back(YmlArgReg);
}
+ // Get type ids.
+ for (auto *CalleeTypeId : CalleeTypeIds) {
+ YmlCS.CalleeTypeIds.push_back(CalleeTypeId->getZExtValue());
+ }
YMF.CallSitesInfo.push_back(std::move(YmlCS));
}
@@ -814,6 +820,8 @@ static void printMI(raw_ostream &OS, MFPrintState &State,
OS << "nusw ";
if (MI.getFlag(MachineInstr::SameSign))
OS << "samesign ";
+ if (MI.getFlag(MachineInstr::InBounds))
+ OS << "inbounds ";
// NOTE: Please add new MIFlags also to the MI_FLAGS_STR in
// llvm/utils/update_mir_test_checks.py.
diff --git a/llvm/lib/CodeGen/MachineFunction.cpp b/llvm/lib/CodeGen/MachineFunction.cpp
index 429a17a..60d42e0 100644
--- a/llvm/lib/CodeGen/MachineFunction.cpp
+++ b/llvm/lib/CodeGen/MachineFunction.cpp
@@ -211,8 +211,7 @@ void MachineFunction::init() {
ConstantPool = new (Allocator) MachineConstantPool(getDataLayout());
Alignment = STI->getTargetLowering()->getMinFunctionAlignment();
- // FIXME: Use Function::hasOptSize().
- if (!F.getAlign() && !F.hasFnAttribute(Attribute::OptimizeForSize))
+ if (!F.getAlign() && !F.hasOptSize())
Alignment = std::max(Alignment,
STI->getTargetLowering()->getPrefFunctionAlignment());
@@ -920,7 +919,7 @@ MachineFunction::getCallSiteInfo(const MachineInstr *MI) {
assert(MI->isCandidateForAdditionalCallInfo() &&
"Call site info refers only to call (MI) candidates");
- if (!Target.Options.EmitCallSiteInfo)
+ if (!Target.Options.EmitCallSiteInfo && !Target.Options.EmitCallGraphSection)
return CallSitesInfo.end();
return CallSitesInfo.find(MI);
}
diff --git a/llvm/lib/CodeGen/MachineInstr.cpp b/llvm/lib/CodeGen/MachineInstr.cpp
index da3665b..79047f7 100644
--- a/llvm/lib/CodeGen/MachineInstr.cpp
+++ b/llvm/lib/CodeGen/MachineInstr.cpp
@@ -585,6 +585,8 @@ uint32_t MachineInstr::copyFlagsFromInstruction(const Instruction &I) {
MIFlags |= MachineInstr::MIFlag::NoUSWrap;
if (GEP->hasNoUnsignedWrap())
MIFlags |= MachineInstr::MIFlag::NoUWrap;
+ if (GEP->isInBounds())
+ MIFlags |= MachineInstr::MIFlag::InBounds;
}
// Copy the nonneg flag.
@@ -1860,8 +1862,12 @@ void MachineInstr::print(raw_ostream &OS, ModuleSlotTracker &MST,
OS << "nneg ";
if (getFlag(MachineInstr::Disjoint))
OS << "disjoint ";
+ if (getFlag(MachineInstr::NoUSWrap))
+ OS << "nusw ";
if (getFlag(MachineInstr::SameSign))
OS << "samesign ";
+ if (getFlag(MachineInstr::InBounds))
+ OS << "inbounds ";
// Print the opcode name.
if (TII)
diff --git a/llvm/lib/CodeGen/MachineOperand.cpp b/llvm/lib/CodeGen/MachineOperand.cpp
index 0d25169..c612f8de 100644
--- a/llvm/lib/CodeGen/MachineOperand.cpp
+++ b/llvm/lib/CodeGen/MachineOperand.cpp
@@ -1273,6 +1273,10 @@ void MachineMemOperand::print(raw_ostream &OS, ModuleSlotTracker &MST,
OS << ", !noalias ";
AAInfo.NoAlias->printAsOperand(OS, MST);
}
+ if (AAInfo.NoAliasAddrSpace) {
+ OS << ", !noalias.addrspace ";
+ AAInfo.NoAliasAddrSpace->printAsOperand(OS, MST);
+ }
if (getRanges()) {
OS << ", !range ";
getRanges()->printAsOperand(OS, MST);
diff --git a/llvm/lib/CodeGen/ModuloSchedule.cpp b/llvm/lib/CodeGen/ModuloSchedule.cpp
index 0f742c4..21bf052 100644
--- a/llvm/lib/CodeGen/ModuloSchedule.cpp
+++ b/llvm/lib/CodeGen/ModuloSchedule.cpp
@@ -423,7 +423,7 @@ void ModuloScheduleExpander::generateExistingPhis(
// potentially define two values.
unsigned MaxPhis = PrologStage + 2;
if (!InKernel && (int)PrologStage <= LoopValStage)
- MaxPhis = std::max((int)MaxPhis - (int)LoopValStage, 1);
+ MaxPhis = std::max((int)MaxPhis - LoopValStage, 1);
unsigned NumPhis = std::min(NumStages, MaxPhis);
Register NewReg;
diff --git a/llvm/lib/CodeGen/RegisterCoalescer.cpp b/llvm/lib/CodeGen/RegisterCoalescer.cpp
index 2d7987a..7ede564 100644
--- a/llvm/lib/CodeGen/RegisterCoalescer.cpp
+++ b/llvm/lib/CodeGen/RegisterCoalescer.cpp
@@ -306,7 +306,12 @@ class RegisterCoalescer : private LiveRangeEdit::Delegate {
/// number if it is not zero. If DstReg is a physical register and the
/// existing subregister number of the def / use being updated is not zero,
/// make sure to set it to the correct physical subregister.
- void updateRegDefsUses(Register SrcReg, Register DstReg, unsigned SubIdx);
+ ///
+ /// If \p SubregToRegSrcInst is not empty, we are coalescing a
+ /// `DstReg = SUBREG_TO_REG SrcReg`, which should introduce an
+ /// implicit-def of DstReg on instructions that define SrcReg.
+ void updateRegDefsUses(Register SrcReg, Register DstReg, unsigned SubIdx,
+ ArrayRef<MachineInstr *> SubregToRegSrcInst = {});
/// If the given machine operand reads only undefined lanes add an undef
/// flag.
@@ -1443,6 +1448,7 @@ bool RegisterCoalescer::reMaterializeTrivialDef(const CoalescerPair &CP,
// CopyMI may have implicit operands, save them so that we can transfer them
// over to the newly materialized instruction after CopyMI is removed.
+ LaneBitmask NewMIImplicitOpsMask;
SmallVector<MachineOperand, 4> ImplicitOps;
ImplicitOps.reserve(CopyMI->getNumOperands() -
CopyMI->getDesc().getNumOperands());
@@ -1457,6 +1463,9 @@ bool RegisterCoalescer::reMaterializeTrivialDef(const CoalescerPair &CP,
(MO.getSubReg() == 0 && MO.getReg() == DstOperand.getReg())) &&
"unexpected implicit virtual register def");
ImplicitOps.push_back(MO);
+ if (MO.isDef() && MO.getReg().isVirtual() &&
+ MRI->shouldTrackSubRegLiveness(DstReg))
+ NewMIImplicitOpsMask |= MRI->getMaxLaneMaskForVReg(MO.getReg());
}
}
@@ -1499,14 +1508,11 @@ bool RegisterCoalescer::reMaterializeTrivialDef(const CoalescerPair &CP,
} else {
assert(MO.getReg() == NewMI.getOperand(0).getReg());
- // We're only expecting another def of the main output, so the range
- // should get updated with the regular output range.
- //
- // FIXME: The range updating below probably needs updating to look at
- // the super register if subranges are tracked.
- assert(!MRI->shouldTrackSubRegLiveness(DstReg) &&
- "subrange update for implicit-def of super register may not be "
- "properly handled");
+ // If lanemasks need to be tracked, compile the lanemask of the NewMI
+ // implicit def operands to avoid subranges for the super-regs from
+ // being removed by code later on in this function.
+ if (MRI->shouldTrackSubRegLiveness(MO.getReg()))
+ NewMIImplicitOpsMask |= MRI->getMaxLaneMaskForVReg(MO.getReg());
}
}
}
@@ -1606,7 +1612,8 @@ bool RegisterCoalescer::reMaterializeTrivialDef(const CoalescerPair &CP,
CurrIdx.getRegSlot(NewMI.getOperand(0).isEarlyClobber());
VNInfo::Allocator &Alloc = LIS->getVNInfoAllocator();
for (LiveInterval::SubRange &SR : DstInt.subranges()) {
- if ((SR.LaneMask & DstMask).none()) {
+ if ((SR.LaneMask & DstMask).none() &&
+ (SR.LaneMask & NewMIImplicitOpsMask).none()) {
LLVM_DEBUG(dbgs()
<< "Removing undefined SubRange "
<< PrintLaneMask(SR.LaneMask) << " : " << SR << "\n");
@@ -1870,11 +1877,14 @@ void RegisterCoalescer::addUndefFlag(const LiveInterval &Int, SlotIndex UseIdx,
}
}
-void RegisterCoalescer::updateRegDefsUses(Register SrcReg, Register DstReg,
- unsigned SubIdx) {
+void RegisterCoalescer::updateRegDefsUses(
+ Register SrcReg, Register DstReg, unsigned SubIdx,
+ ArrayRef<MachineInstr *> SubregToRegSrcInsts) {
bool DstIsPhys = DstReg.isPhysical();
LiveInterval *DstInt = DstIsPhys ? nullptr : &LIS->getInterval(DstReg);
+ // Coalescing a COPY may expose reads of 'undef' subregisters.
+ // If so, then explicitly propagate 'undef' to those operands.
if (DstInt && DstInt->hasSubRanges() && DstReg != SrcReg) {
for (MachineOperand &MO : MRI->reg_operands(DstReg)) {
if (MO.isUndef())
@@ -1891,6 +1901,15 @@ void RegisterCoalescer::updateRegDefsUses(Register SrcReg, Register DstReg,
}
}
+ // If DstInt already has a subrange for the unused lanes, then we shouldn't
+ // create duplicate subranges when we update the interval for unused lanes.
+ LaneBitmask DstIntLaneMask;
+ if (DstInt && MRI->shouldTrackSubRegLiveness(DstReg)) {
+ for (LiveInterval::SubRange &SR : DstInt->subranges())
+ DstIntLaneMask |= SR.LaneMask;
+ }
+
+ // Go through all instructions to replace uses of 'SrcReg' by 'DstReg'.
SmallPtrSet<MachineInstr *, 8> Visited;
for (MachineRegisterInfo::reg_instr_iterator I = MRI->reg_instr_begin(SrcReg),
E = MRI->reg_instr_end();
@@ -1914,6 +1933,80 @@ void RegisterCoalescer::updateRegDefsUses(Register SrcReg, Register DstReg,
if (DstInt && !Reads && SubIdx && !UseMI->isDebugInstr())
Reads = DstInt->liveAt(LIS->getInstructionIndex(*UseMI));
+ bool RequiresImplicitRedef = false;
+ if (!SubregToRegSrcInsts.empty()) {
+ // We can only add an implicit-def and undef if the sub registers match,
+ // e.g.
+ // %0:gr32 = INSTX
+ // %0.sub8:gr32 = INSTY // top 24 bits of %0 still defined
+ // %1:gr64 = SUBREG_TO_REG 0, %0, %subreg.sub32
+ //
+ // This cannot be transformed into:
+ // %1.sub32:gr64 = INSTX
+ // undef %1.sub8:gr64 = INSTY , implicit-def %1
+ //
+ // Because that would thrash the top 24 bits of %1.sub32.
+ if (is_contained(SubregToRegSrcInsts, UseMI) &&
+ all_of(UseMI->defs(),
+ [&SubIdx, &SrcReg](const MachineOperand &MO) -> bool {
+ if (MO.getReg() != SrcReg || !MO.getSubReg() || MO.isUndef())
+ return true;
+ return SubIdx == MO.getSubReg();
+ })) {
+ // Add implicit-def of super-register to express that the whole
+ // register is defined by the instruction.
+ MachineInstrBuilder MIB(*MF, UseMI);
+ MIB.addReg(DstReg, RegState::ImplicitDefine);
+ RequiresImplicitRedef = true;
+ }
+
+ // If the coalesed instruction doesn't fully define the register, we need
+ // to preserve the original super register liveness for SUBREG_TO_REG.
+ //
+ // We pretended SUBREG_TO_REG was a regular copy for coalescing purposes,
+ // but it introduces liveness for other subregisters. Downstream users may
+ // have been relying on those bits, so we need to ensure their liveness is
+ // captured with a def of other lanes.
+ if (DstInt && MRI->shouldTrackSubRegLiveness(DstReg)) {
+ // First check if there is sufficient granularity in terms of subranges.
+ LaneBitmask DstMask = MRI->getMaxLaneMaskForVReg(DstInt->reg());
+ LaneBitmask UsedLanes = TRI->getSubRegIndexLaneMask(SubIdx);
+ LaneBitmask UnusedLanes = DstMask & ~UsedLanes;
+ if ((UnusedLanes & ~DstIntLaneMask).any()) {
+ BumpPtrAllocator &Allocator = LIS->getVNInfoAllocator();
+ DstInt->createSubRangeFrom(Allocator, UnusedLanes, *DstInt);
+ DstIntLaneMask |= UnusedLanes;
+ }
+
+ // After duplicating the live ranges for the low/hi bits, we
+ // need to update the subranges of the DstReg interval such that
+ // for a case like this:
+ //
+ // entry:
+ // 16B %1:gpr32 = INSTRUCTION (<=> UseMI)
+ // :
+ // if.then:
+ // 32B %1:gpr32 = MOVIMM32 ..
+ // 48B %0:gpr64 = SUBREG_TO_REG 0, %1, sub32
+ //
+ // Only the MOVIMM32 require a def of the top lanes and any intervals
+ // for the top 32-bits of the def at 16B should be removed.
+ for (LiveInterval::SubRange &SR : DstInt->subranges()) {
+ if (!Writes || RequiresImplicitRedef ||
+ (SR.LaneMask & UnusedLanes).none())
+ continue;
+
+ assert((SR.LaneMask & UnusedLanes) == SR.LaneMask &&
+ "Unexpected lanemask. Subrange needs finer granularity");
+
+ SlotIndex UseIdx = LIS->getInstructionIndex(*UseMI).getRegSlot(false);
+ auto SegmentI = SR.find(UseIdx);
+ if (SegmentI != SR.end())
+ SR.removeSegment(SegmentI, true);
+ }
+ }
+ }
+
// Replace SrcReg with DstReg in all UseMI operands.
for (unsigned Op : Ops) {
MachineOperand &MO = UseMI->getOperand(Op);
@@ -1922,7 +2015,7 @@ void RegisterCoalescer::updateRegDefsUses(Register SrcReg, Register DstReg,
// turn a full def into a read-modify-write sub-register def and vice
// versa.
if (SubIdx && MO.isDef())
- MO.setIsUndef(!Reads);
+ MO.setIsUndef(!Reads || RequiresImplicitRedef);
// A subreg use of a partially undef (super) register may be a complete
// undef use now and then has to be marked that way.
@@ -2025,6 +2118,30 @@ void RegisterCoalescer::setUndefOnPrunedSubRegUses(LiveInterval &LI,
LIS->shrinkToUses(&LI);
}
+/// For a given use of value \p Idx, it returns the def in the current block,
+/// or otherwise all possible defs in preceding blocks.
+static bool FindDefInBlock(SmallPtrSetImpl<MachineBasicBlock *> &VisitedBlocks,
+ SmallVector<MachineInstr *> &Instrs,
+ LiveIntervals *LIS, LiveInterval &SrcInt,
+ MachineBasicBlock *MBB, VNInfo *Idx) {
+ if (!Idx->isPHIDef()) {
+ MachineInstr *Def = LIS->getInstructionFromIndex(Idx->def);
+ assert(Def && "Unable to find a def for SUBREG_TO_REG source operand");
+ Instrs.push_back(Def);
+ return true;
+ }
+
+ bool Any = false;
+ if (VisitedBlocks.count(MBB))
+ return false;
+ VisitedBlocks.insert(MBB);
+ for (MachineBasicBlock *Pred : MBB->predecessors()) {
+ Any |= FindDefInBlock(VisitedBlocks, Instrs, LIS, SrcInt, Pred,
+ SrcInt.getVNInfoBefore(LIS->getMBBEndIdx(Pred)));
+ }
+ return Any;
+}
+
bool RegisterCoalescer::joinCopy(
MachineInstr *CopyMI, bool &Again,
SmallPtrSetImpl<MachineInstr *> &CurrentErasedInstrs) {
@@ -2156,6 +2273,35 @@ bool RegisterCoalescer::joinCopy(
});
}
+ SmallVector<MachineInstr *> SubregToRegSrcInsts;
+ if (CopyMI->isSubregToReg()) {
+ // For the case where the copy instruction is a SUBREG_TO_REG, e.g.
+ //
+ // %0:gpr32 = movimm32 ..
+ // %1:gpr64 = SUBREG_TO_REG 0, %0, sub32
+ // ...
+ // %0:gpr32 = COPY <something>
+ //
+ // After joining liveranges, the original `movimm32` will need an
+ // implicit-def to make it explicit that the entire register is written,
+ // i.e.
+ //
+ // undef %0.sub32:gpr64 = movimm32 ..., implicit-def %0
+ // ...
+ // undef %0.sub32:gpr64 = COPY <something> // Note that this does not
+ // // require an implicit-def,
+ // // because it has nothing to
+ // // do with the SUBREG_TO_REG.
+ LiveInterval &SrcInt =
+ LIS->getInterval(CP.isFlipped() ? CP.getDstReg() : CP.getSrcReg());
+ SlotIndex SubregToRegSlotIdx = LIS->getInstructionIndex(*CopyMI);
+ SmallPtrSet<MachineBasicBlock *, 8> VisitedBlocks;
+ if (!FindDefInBlock(VisitedBlocks, SubregToRegSrcInsts, LIS, SrcInt,
+ CopyMI->getParent(),
+ SrcInt.Query(SubregToRegSlotIdx).valueIn()))
+ llvm_unreachable("SUBREG_TO_REG src requires a def");
+ }
+
ShrinkMask = LaneBitmask::getNone();
ShrinkMainRange = false;
@@ -2225,9 +2371,12 @@ bool RegisterCoalescer::joinCopy(
// Rewrite all SrcReg operands to DstReg.
// Also update DstReg operands to include DstIdx if it is set.
- if (CP.getDstIdx())
+ if (CP.getDstIdx()) {
+ assert(SubregToRegSrcInsts.empty() && "can this happen?");
updateRegDefsUses(CP.getDstReg(), CP.getDstReg(), CP.getDstIdx());
- updateRegDefsUses(CP.getSrcReg(), CP.getDstReg(), CP.getSrcIdx());
+ }
+ updateRegDefsUses(CP.getSrcReg(), CP.getDstReg(), CP.getSrcIdx(),
+ SubregToRegSrcInsts);
// Shrink subregister ranges if necessary.
if (ShrinkMask.any()) {
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index d3df434..a43020e 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -35,6 +35,7 @@
#include "llvm/CodeGen/ByteProvider.h"
#include "llvm/CodeGen/DAGCombine.h"
#include "llvm/CodeGen/ISDOpcodes.h"
+#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/SDPatternMatch.h"
@@ -15262,23 +15263,31 @@ SDValue DAGCombiner::visitAssertExt(SDNode *N) {
}
}
- // If we have (AssertZext (and (AssertSext X, iX), M), iY) and Y is smaller
- // than X, and the And doesn't change the lower iX bits, we can move the
- // AssertZext in front of the And and drop the AssertSext.
if (Opcode == ISD::AssertZext && N0.getOpcode() == ISD::AND &&
- N0.hasOneUse() && N0.getOperand(0).getOpcode() == ISD::AssertSext &&
isa<ConstantSDNode>(N0.getOperand(1))) {
- SDValue BigA = N0.getOperand(0);
- EVT BigA_AssertVT = cast<VTSDNode>(BigA.getOperand(1))->getVT();
const APInt &Mask = N0.getConstantOperandAPInt(1);
- if (AssertVT.bitsLT(BigA_AssertVT) &&
- Mask.countr_one() >= BigA_AssertVT.getScalarSizeInBits()) {
- SDLoc DL(N);
- SDValue NewAssert =
- DAG.getNode(Opcode, DL, N->getValueType(0), BigA.getOperand(0), N1);
- return DAG.getNode(ISD::AND, DL, N->getValueType(0), NewAssert,
- N0.getOperand(1));
+
+ // If we have (AssertZext (and (AssertSext X, iX), M), iY) and Y is smaller
+ // than X, and the And doesn't change the lower iX bits, we can move the
+ // AssertZext in front of the And and drop the AssertSext.
+ if (N0.getOperand(0).getOpcode() == ISD::AssertSext && N0.hasOneUse()) {
+ SDValue BigA = N0.getOperand(0);
+ EVT BigA_AssertVT = cast<VTSDNode>(BigA.getOperand(1))->getVT();
+ if (AssertVT.bitsLT(BigA_AssertVT) &&
+ Mask.countr_one() >= BigA_AssertVT.getScalarSizeInBits()) {
+ SDLoc DL(N);
+ SDValue NewAssert =
+ DAG.getNode(Opcode, DL, N->getValueType(0), BigA.getOperand(0), N1);
+ return DAG.getNode(ISD::AND, DL, N->getValueType(0), NewAssert,
+ N0.getOperand(1));
+ }
}
+
+ // Remove AssertZext entirely if the mask guarantees the assertion cannot
+ // fail.
+ // TODO: Use KB countMinLeadingZeros to handle non-constant masks?
+ if (Mask.isIntN(AssertVT.getScalarSizeInBits()))
+ return N0;
}
return SDValue();
@@ -22778,8 +22787,10 @@ SDValue DAGCombiner::visitLIFETIME_END(SDNode *N) {
const BaseIndexOffset StoreBase = BaseIndexOffset::match(ST, DAG);
// If we store purely within object bounds just before its lifetime ends,
// we can remove the store.
- if (LifetimeEndBase.contains(DAG, LifetimeEnd->getSize() * 8, StoreBase,
- StoreSize.getFixedValue() * 8)) {
+ MachineFrameInfo &MFI = DAG.getMachineFunction().getFrameInfo();
+ if (LifetimeEndBase.contains(
+ DAG, MFI.getObjectSize(LifetimeEnd->getFrameIndex()) * 8,
+ StoreBase, StoreSize.getFixedValue() * 8)) {
LLVM_DEBUG(dbgs() << "\nRemoving store:"; StoreBase.dump();
dbgs() << "\nwithin LIFETIME_END of : ";
LifetimeEndBase.dump(); dbgs() << "\n");
@@ -28971,13 +28982,100 @@ SDValue DAGCombiner::SimplifySelectCC(const SDLoc &DL, SDValue N0, SDValue N1,
return SDValue();
}
+static SDValue matchMergedBFX(SDValue Root, SelectionDAG &DAG,
+ const TargetLowering &TLI) {
+ // Match a pattern such as:
+ // (X | (X >> C0) | (X >> C1) | ...) & Mask
+ // This extracts contiguous parts of X and ORs them together before comparing.
+ // We can optimize this so that we directly check (X & SomeMask) instead,
+ // eliminating the shifts.
+
+ EVT VT = Root.getValueType();
+
+ // TODO: Support vectors?
+ if (!VT.isScalarInteger() || Root.getOpcode() != ISD::AND)
+ return SDValue();
+
+ SDValue N0 = Root.getOperand(0);
+ SDValue N1 = Root.getOperand(1);
+
+ if (N0.getOpcode() != ISD::OR || !isa<ConstantSDNode>(N1))
+ return SDValue();
+
+ APInt RootMask = cast<ConstantSDNode>(N1)->getAsAPIntVal();
+
+ SDValue Src;
+ const auto IsSrc = [&](SDValue V) {
+ if (!Src) {
+ Src = V;
+ return true;
+ }
+
+ return Src == V;
+ };
+
+ SmallVector<SDValue> Worklist = {N0};
+ APInt PartsMask(VT.getSizeInBits(), 0);
+ while (!Worklist.empty()) {
+ SDValue V = Worklist.pop_back_val();
+ if (!V.hasOneUse() && (Src && Src != V))
+ return SDValue();
+
+ if (V.getOpcode() == ISD::OR) {
+ Worklist.push_back(V.getOperand(0));
+ Worklist.push_back(V.getOperand(1));
+ continue;
+ }
+
+ if (V.getOpcode() == ISD::SRL) {
+ SDValue ShiftSrc = V.getOperand(0);
+ SDValue ShiftAmt = V.getOperand(1);
+
+ if (!IsSrc(ShiftSrc) || !isa<ConstantSDNode>(ShiftAmt))
+ return SDValue();
+
+ auto ShiftAmtVal = cast<ConstantSDNode>(ShiftAmt)->getAsZExtVal();
+ if (ShiftAmtVal > RootMask.getBitWidth())
+ return SDValue();
+
+ PartsMask |= (RootMask << ShiftAmtVal);
+ continue;
+ }
+
+ if (IsSrc(V)) {
+ PartsMask |= RootMask;
+ continue;
+ }
+
+ return SDValue();
+ }
+
+ if (!Src)
+ return SDValue();
+
+ SDLoc DL(Root);
+ return DAG.getNode(ISD::AND, DL, VT,
+ {Src, DAG.getConstant(PartsMask, DL, VT)});
+}
+
/// This is a stub for TargetLowering::SimplifySetCC.
SDValue DAGCombiner::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
ISD::CondCode Cond, const SDLoc &DL,
bool foldBooleans) {
TargetLowering::DAGCombinerInfo
DagCombineInfo(DAG, Level, false, this);
- return TLI.SimplifySetCC(VT, N0, N1, Cond, foldBooleans, DagCombineInfo, DL);
+ if (SDValue C =
+ TLI.SimplifySetCC(VT, N0, N1, Cond, foldBooleans, DagCombineInfo, DL))
+ return C;
+
+ if (ISD::isIntEqualitySetCC(Cond) && N0.getOpcode() == ISD::AND &&
+ isNullConstant(N1)) {
+
+ if (SDValue Res = matchMergedBFX(N0, DAG, TLI))
+ return DAG.getSetCC(DL, VT, Res, N1, Cond);
+ }
+
+ return SDValue();
}
/// Given an ISD::SDIV node expressing a divide by constant, return
@@ -29415,7 +29513,7 @@ bool DAGCombiner::mayAlias(SDNode *Op0, SDNode *Op1) const {
MachineMemOperand *MMO;
};
- auto getCharacteristics = [](SDNode *N) -> MemUseCharacteristics {
+ auto getCharacteristics = [this](SDNode *N) -> MemUseCharacteristics {
if (const auto *LSN = dyn_cast<LSBaseSDNode>(N)) {
int64_t Offset = 0;
if (auto *C = dyn_cast<ConstantSDNode>(LSN->getOffset()))
@@ -29428,13 +29526,15 @@ bool DAGCombiner::mayAlias(SDNode *Op0, SDNode *Op1) const {
LSN->getBasePtr(), Offset /*base offset*/,
LocationSize::precise(Size), LSN->getMemOperand()};
}
- if (const auto *LN = cast<LifetimeSDNode>(N))
+ if (const auto *LN = cast<LifetimeSDNode>(N)) {
+ MachineFrameInfo &MFI = DAG.getMachineFunction().getFrameInfo();
return {false /*isVolatile*/,
/*isAtomic*/ false,
LN->getOperand(1),
0,
- LocationSize::precise(LN->getSize()),
+ LocationSize::precise(MFI.getObjectSize(LN->getFrameIndex())),
(MachineMemOperand *)nullptr};
+ }
// Default.
return {false /*isvolatile*/,
/*isAtomic*/ false,
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index 74172b2..ba0ab23 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -3853,7 +3853,7 @@ bool SelectionDAGLegalize::ExpandNode(SDNode *Node) {
break;
case ISD::FP_TO_FP16:
LLVM_DEBUG(dbgs() << "Legalizing FP_TO_FP16\n");
- if (!TLI.useSoftFloat() && TM.Options.UnsafeFPMath) {
+ if (Node->getFlags().hasApproximateFuncs() && !TLI.useSoftFloat()) {
SDValue Op = Node->getOperand(0);
MVT SVT = Op.getSimpleValueType();
if ((SVT == MVT::f64 || SVT == MVT::f80) &&
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 773ff48..02d1100 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -784,10 +784,6 @@ static void AddNodeIDCustom(FoldingSetNodeID &ID, const SDNode *N) {
case ISD::TargetFrameIndex:
ID.AddInteger(cast<FrameIndexSDNode>(N)->getIndex());
break;
- case ISD::LIFETIME_START:
- case ISD::LIFETIME_END:
- ID.AddInteger(cast<LifetimeSDNode>(N)->getSize());
- break;
case ISD::PSEUDO_PROBE:
ID.AddInteger(cast<PseudoProbeSDNode>(N)->getGuid());
ID.AddInteger(cast<PseudoProbeSDNode>(N)->getIndex());
@@ -7847,20 +7843,43 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
}
}
- // Perform trivial constant folding.
- if (SDValue SV = FoldConstantArithmetic(Opcode, DL, VT, {N1, N2}, Flags))
- return SV;
+ if (N1.getOpcode() == ISD::POISON || N2.getOpcode() == ISD::POISON) {
+ switch (Opcode) {
+ case ISD::XOR:
+ case ISD::ADD:
+ case ISD::PTRADD:
+ case ISD::SUB:
+ case ISD::SIGN_EXTEND_INREG:
+ case ISD::UDIV:
+ case ISD::SDIV:
+ case ISD::UREM:
+ case ISD::SREM:
+ case ISD::MUL:
+ case ISD::AND:
+ case ISD::SSUBSAT:
+ case ISD::USUBSAT:
+ case ISD::UMIN:
+ case ISD::OR:
+ case ISD::SADDSAT:
+ case ISD::UADDSAT:
+ case ISD::UMAX:
+ case ISD::SMAX:
+ case ISD::SMIN:
+ // fold op(arg1, poison) -> poison, fold op(poison, arg2) -> poison.
+ return N2.getOpcode() == ISD::POISON ? N2 : N1;
+ }
+ }
// Canonicalize an UNDEF to the RHS, even over a constant.
- if (N1.isUndef()) {
+ if (N1.getOpcode() == ISD::UNDEF && N2.getOpcode() != ISD::UNDEF) {
if (TLI->isCommutativeBinOp(Opcode)) {
std::swap(N1, N2);
} else {
switch (Opcode) {
case ISD::PTRADD:
case ISD::SUB:
- // fold op(undef, arg2) -> undef, fold op(poison, arg2) ->poison.
- return N1.getOpcode() == ISD::POISON ? getPOISON(VT) : getUNDEF(VT);
+ // fold op(undef, non_undef_arg2) -> undef.
+ return N1;
case ISD::SIGN_EXTEND_INREG:
case ISD::UDIV:
case ISD::SDIV:
@@ -7868,18 +7887,17 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
case ISD::SREM:
case ISD::SSUBSAT:
case ISD::USUBSAT:
- // fold op(undef, arg2) -> 0, fold op(poison, arg2) -> poison.
- return N1.getOpcode() == ISD::POISON ? getPOISON(VT)
- : getConstant(0, DL, VT);
+ // fold op(undef, non_undef_arg2) -> 0.
+ return getConstant(0, DL, VT);
}
}
}
// Fold a bunch of operators when the RHS is undef.
- if (N2.isUndef()) {
+ if (N2.getOpcode() == ISD::UNDEF) {
switch (Opcode) {
case ISD::XOR:
- if (N1.isUndef())
+ if (N1.getOpcode() == ISD::UNDEF)
// Handle undef ^ undef -> 0 special case. This is a common
// idiom (misuse).
return getConstant(0, DL, VT);
@@ -7887,29 +7905,48 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
case ISD::ADD:
case ISD::PTRADD:
case ISD::SUB:
+ // fold op(arg1, undef) -> undef.
+ return N2;
case ISD::UDIV:
case ISD::SDIV:
case ISD::UREM:
case ISD::SREM:
- // fold op(arg1, undef) -> undef, fold op(arg1, poison) -> poison.
- return N2.getOpcode() == ISD::POISON ? getPOISON(VT) : getUNDEF(VT);
+ // fold op(arg1, undef) -> poison.
+ return getPOISON(VT);
case ISD::MUL:
case ISD::AND:
case ISD::SSUBSAT:
case ISD::USUBSAT:
- // fold op(arg1, undef) -> 0, fold op(arg1, poison) -> poison.
- return N2.getOpcode() == ISD::POISON ? getPOISON(VT)
- : getConstant(0, DL, VT);
+ case ISD::UMIN:
+ // fold op(undef, undef) -> undef, fold op(arg1, undef) -> 0.
+ return N1.getOpcode() == ISD::UNDEF ? N2 : getConstant(0, DL, VT);
case ISD::OR:
case ISD::SADDSAT:
case ISD::UADDSAT:
- // fold op(arg1, undef) -> an all-ones constant, fold op(arg1, poison) ->
- // poison.
- return N2.getOpcode() == ISD::POISON ? getPOISON(VT)
- : getAllOnesConstant(DL, VT);
+ case ISD::UMAX:
+ // fold op(undef, undef) -> undef, fold op(arg1, undef) -> -1.
+ return N1.getOpcode() == ISD::UNDEF ? N2 : getAllOnesConstant(DL, VT);
+ case ISD::SMAX:
+ // fold op(undef, undef) -> undef, fold op(arg1, undef) -> MAX_INT.
+ return N1.getOpcode() == ISD::UNDEF
+ ? N2
+ : getConstant(
+ APInt::getSignedMaxValue(VT.getScalarSizeInBits()), DL,
+ VT);
+ case ISD::SMIN:
+ // fold op(undef, undef) -> undef, fold op(arg1, undef) -> MIN_INT.
+ return N1.getOpcode() == ISD::UNDEF
+ ? N2
+ : getConstant(
+ APInt::getSignedMinValue(VT.getScalarSizeInBits()), DL,
+ VT);
}
}
+ // Perform trivial constant folding.
+ if (SDValue SV = FoldConstantArithmetic(Opcode, DL, VT, {N1, N2}, Flags))
+ return SV;
+
// Memoize this node if possible.
SDNode *N;
SDVTList VTs = getVTList(VT);
@@ -9360,8 +9397,7 @@ SDValue SelectionDAG::getMemIntrinsicNode(unsigned Opcode, const SDLoc &dl,
}
SDValue SelectionDAG::getLifetimeNode(bool IsStart, const SDLoc &dl,
- SDValue Chain, int FrameIndex,
- int64_t Size) {
+ SDValue Chain, int FrameIndex) {
const unsigned Opcode = IsStart ? ISD::LIFETIME_START : ISD::LIFETIME_END;
const auto VTs = getVTList(MVT::Other);
SDValue Ops[2] = {
@@ -9373,13 +9409,12 @@ SDValue SelectionDAG::getLifetimeNode(bool IsStart, const SDLoc &dl,
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
ID.AddInteger(FrameIndex);
- ID.AddInteger(Size);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
- LifetimeSDNode *N = newSDNode<LifetimeSDNode>(Opcode, dl.getIROrder(),
- dl.getDebugLoc(), VTs, Size);
+ LifetimeSDNode *N =
+ newSDNode<LifetimeSDNode>(Opcode, dl.getIROrder(), dl.getDebugLoc(), VTs);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 1636465..306e068 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -3923,11 +3923,15 @@ void SelectionDAGBuilder::visitFPTrunc(const User &I) {
// FPTrunc is never a no-op cast, no need to check
SDValue N = getValue(I.getOperand(0));
SDLoc dl = getCurSDLoc();
+ SDNodeFlags Flags;
+ if (auto *TruncInst = dyn_cast<FPMathOperator>(&I))
+ Flags.copyFMF(*TruncInst);
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
EVT DestVT = TLI.getValueType(DAG.getDataLayout(), I.getType());
setValue(&I, DAG.getNode(ISD::FP_ROUND, dl, DestVT, N,
DAG.getTargetConstant(
- 0, dl, TLI.getPointerTy(DAG.getDataLayout()))));
+ 0, dl, TLI.getPointerTy(DAG.getDataLayout())),
+ Flags));
}
void SelectionDAGBuilder::visitFPExt(const User &I) {
@@ -7594,8 +7598,6 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I,
if (TM.getOptLevel() == CodeGenOptLevel::None)
return;
- const int64_t ObjectSize =
- cast<ConstantInt>(I.getArgOperand(0))->getSExtValue();
const AllocaInst *LifetimeObject = cast<AllocaInst>(I.getArgOperand(1));
// First check that the Alloca is static, otherwise it won't have a
@@ -7605,7 +7607,7 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I,
return;
const int FrameIndex = SI->second;
- Res = DAG.getLifetimeNode(IsStart, sdl, getRoot(), FrameIndex, ObjectSize);
+ Res = DAG.getLifetimeNode(IsStart, sdl, getRoot(), FrameIndex);
DAG.setRoot(Res);
return;
}
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
index 9474587..900da76 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
@@ -946,8 +946,6 @@ void SDNode::print_details(raw_ostream &OS, const SelectionDAG *G) const {
<< " -> "
<< ASC->getDestAddressSpace()
<< ']';
- } else if (const LifetimeSDNode *LN = dyn_cast<LifetimeSDNode>(this)) {
- OS << "<0 to " << LN->getSize() << ">";
} else if (const auto *AA = dyn_cast<AssertAlignSDNode>(this)) {
OS << '<' << AA->getAlign().value() << '>';
}
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 1764910..48d6b99 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -9471,7 +9471,7 @@ SDValue TargetLowering::CTTZTableLookup(SDNode *Node, SelectionDAG &DAG,
ISD::SRL, DL, VT,
DAG.getNode(ISD::MUL, DL, VT, DAG.getNode(ISD::AND, DL, VT, Op, Neg),
DAG.getConstant(DeBruijn, DL, VT)),
- DAG.getConstant(ShiftAmt, DL, VT));
+ DAG.getShiftAmountConstant(ShiftAmt, VT, DL));
Lookup = DAG.getSExtOrTrunc(Lookup, DL, getPointerTy(TD));
SmallVector<uint8_t> Table(BitWidth, 0);
diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index d4a3455..3c91b0e 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -806,7 +806,17 @@ void TargetLoweringBase::initActions() {
ISD::SDIVFIX, ISD::SDIVFIXSAT,
ISD::UDIVFIX, ISD::UDIVFIXSAT,
ISD::FP_TO_SINT_SAT, ISD::FP_TO_UINT_SAT,
- ISD::IS_FPCLASS},
+ ISD::IS_FPCLASS, ISD::FCBRT,
+ ISD::FLOG, ISD::FLOG2,
+ ISD::FLOG10, ISD::FEXP,
+ ISD::FEXP2, ISD::FEXP10,
+ ISD::FFLOOR, ISD::FNEARBYINT,
+ ISD::FCEIL, ISD::FRINT,
+ ISD::FTRUNC, ISD::FROUNDEVEN,
+ ISD::FTAN, ISD::FACOS,
+ ISD::FASIN, ISD::FATAN,
+ ISD::FCOSH, ISD::FSINH,
+ ISD::FTANH, ISD::FATAN2},
VT, Expand);
// Overflow operations default to expand
@@ -852,13 +862,12 @@ void TargetLoweringBase::initActions() {
// These operations default to expand for vector types.
if (VT.isVector())
- setOperationAction(
- {ISD::FCOPYSIGN, ISD::SIGN_EXTEND_INREG, ISD::ANY_EXTEND_VECTOR_INREG,
- ISD::SIGN_EXTEND_VECTOR_INREG, ISD::ZERO_EXTEND_VECTOR_INREG,
- ISD::SPLAT_VECTOR, ISD::LRINT, ISD::LLRINT, ISD::LROUND,
- ISD::LLROUND, ISD::FTAN, ISD::FACOS, ISD::FASIN, ISD::FATAN,
- ISD::FCOSH, ISD::FSINH, ISD::FTANH, ISD::FATAN2},
- VT, Expand);
+ setOperationAction({ISD::FCOPYSIGN, ISD::SIGN_EXTEND_INREG,
+ ISD::ANY_EXTEND_VECTOR_INREG,
+ ISD::SIGN_EXTEND_VECTOR_INREG,
+ ISD::ZERO_EXTEND_VECTOR_INREG, ISD::SPLAT_VECTOR,
+ ISD::LRINT, ISD::LLRINT, ISD::LROUND, ISD::LLROUND},
+ VT, Expand);
// Constrained floating-point operations default to expand.
#define DAG_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN) \
@@ -914,15 +923,6 @@ void TargetLoweringBase::initActions() {
{MVT::bf16, MVT::f16, MVT::f32, MVT::f64, MVT::f80, MVT::f128},
Expand);
- // These library functions default to expand.
- setOperationAction({ISD::FCBRT, ISD::FLOG, ISD::FLOG2, ISD::FLOG10,
- ISD::FEXP, ISD::FEXP2, ISD::FEXP10, ISD::FFLOOR,
- ISD::FNEARBYINT, ISD::FCEIL, ISD::FRINT, ISD::FTRUNC,
- ISD::FROUNDEVEN, ISD::FTAN, ISD::FACOS, ISD::FASIN,
- ISD::FATAN, ISD::FCOSH, ISD::FSINH, ISD::FTANH,
- ISD::FATAN2},
- {MVT::f32, MVT::f64, MVT::f128}, Expand);
-
// Insert custom handling default for llvm.canonicalize.*.
setOperationAction(ISD::FCANONICALIZE,
{MVT::f16, MVT::f32, MVT::f64, MVT::f128}, Expand);
@@ -2062,7 +2062,7 @@ void TargetLoweringBase::insertSSPDeclarations(Module &M) const {
// FreeBSD has "__stack_chk_guard" defined externally on libc.so
if (M.getDirectAccessExternalData() &&
- !TM.getTargetTriple().isWindowsGNUEnvironment() &&
+ !TM.getTargetTriple().isOSCygMing() &&
!(TM.getTargetTriple().isPPC64() &&
TM.getTargetTriple().isOSFreeBSD()) &&
(!TM.getTargetTriple().isOSDarwin() ||
diff --git a/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp b/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
index 408d07b..e9172f4 100644
--- a/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
@@ -1060,27 +1060,27 @@ MCSection *TargetLoweringObjectFileELF::getSectionForConstant(
auto &Context = getContext();
if (Kind.isMergeableConst4() && MergeableConst4Section)
- return Context.getELFSection(".rodata.cst4." + SectionSuffix,
+ return Context.getELFSection(".rodata.cst4." + SectionSuffix + ".",
ELF::SHT_PROGBITS,
ELF::SHF_ALLOC | ELF::SHF_MERGE, 4);
if (Kind.isMergeableConst8() && MergeableConst8Section)
- return Context.getELFSection(".rodata.cst8." + SectionSuffix,
+ return Context.getELFSection(".rodata.cst8." + SectionSuffix + ".",
ELF::SHT_PROGBITS,
ELF::SHF_ALLOC | ELF::SHF_MERGE, 8);
if (Kind.isMergeableConst16() && MergeableConst16Section)
- return Context.getELFSection(".rodata.cst16." + SectionSuffix,
+ return Context.getELFSection(".rodata.cst16." + SectionSuffix + ".",
ELF::SHT_PROGBITS,
ELF::SHF_ALLOC | ELF::SHF_MERGE, 16);
if (Kind.isMergeableConst32() && MergeableConst32Section)
- return Context.getELFSection(".rodata.cst32." + SectionSuffix,
+ return Context.getELFSection(".rodata.cst32." + SectionSuffix + ".",
ELF::SHT_PROGBITS,
ELF::SHF_ALLOC | ELF::SHF_MERGE, 32);
if (Kind.isReadOnly())
- return Context.getELFSection(".rodata." + SectionSuffix, ELF::SHT_PROGBITS,
- ELF::SHF_ALLOC);
+ return Context.getELFSection(".rodata." + SectionSuffix + ".",
+ ELF::SHT_PROGBITS, ELF::SHF_ALLOC);
assert(Kind.isReadOnlyWithRel() && "Unknown section kind");
- return Context.getELFSection(".data.rel.ro." + SectionSuffix,
+ return Context.getELFSection(".data.rel.ro." + SectionSuffix + ".",
ELF::SHT_PROGBITS,
ELF::SHF_ALLOC | ELF::SHF_WRITE);
}
@@ -1734,7 +1734,8 @@ MCSection *TargetLoweringObjectFileCOFF::getExplicitSectionGlobal(
Name == getInstrProfSectionName(IPSK_covdata, Triple::COFF,
/*AddSegmentInfo=*/false) ||
Name == getInstrProfSectionName(IPSK_covname, Triple::COFF,
- /*AddSegmentInfo=*/false))
+ /*AddSegmentInfo=*/false) ||
+ Name == ".llvmbc" || Name == ".llvmcmd")
Kind = SectionKind::getMetadata();
int Selection = 0;
unsigned Characteristics = getCOFFSectionFlags(Kind, TM);
diff --git a/llvm/lib/CodeGen/WindowsSecureHotPatching.cpp b/llvm/lib/CodeGen/WindowsSecureHotPatching.cpp
index 6267207..fd54190 100644
--- a/llvm/lib/CodeGen/WindowsSecureHotPatching.cpp
+++ b/llvm/lib/CodeGen/WindowsSecureHotPatching.cpp
@@ -369,6 +369,19 @@ static GlobalVariable *getOrCreateRefVariable(
AddrOfOldGV, Twine("__ref_").concat(GV->getName()),
nullptr, GlobalVariable::NotThreadLocal);
+ // RefGV is created with isConstant = false, but we want to place RefGV into
+ // .rdata, not .data. It is important that the GlobalVariable be mutable
+ // from the compiler's point of view, so that the optimizer does not remove
+ // the global variable entirely and replace all references to it with its
+ // initial value.
+ //
+ // When the Windows hot-patch loader applies a hot-patch, it maps the
+ // pages of .rdata as read/write so that it can set each __ref_* variable
+ // to point to the original variable in the base image. Afterward, pages in
+ // .rdata are remapped as read-only. This protects the __ref_* variables from
+ // being overwritten during execution.
+ RefGV->setSection(".rdata");
+
// Create debug info for the replacement global variable.
DataLayout Layout = M->getDataLayout();
DIType *DebugType = DebugInfo.createPointerType(