aboutsummaryrefslogtreecommitdiff
path: root/llvm/lib/CodeGen
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib/CodeGen')
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp1
-rw-r--r--llvm/lib/CodeGen/AtomicExpandPass.cpp13
-rw-r--r--llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp2
-rw-r--r--llvm/lib/CodeGen/RegisterPressure.cpp4
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp45
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp18
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp7
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h1
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp18
-rw-r--r--llvm/lib/CodeGen/TargetLoweringBase.cpp1
10 files changed, 70 insertions, 40 deletions
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
index c72b6e8..23a3543 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
@@ -3657,6 +3657,7 @@ const MCExpr *AsmPrinter::lowerConstant(const Constant *CV,
break; // Error
}
+ case Instruction::PtrToAddr:
case Instruction::PtrToInt: {
const DataLayout &DL = getDataLayout();
diff --git a/llvm/lib/CodeGen/AtomicExpandPass.cpp b/llvm/lib/CodeGen/AtomicExpandPass.cpp
index 3f3d5dc9..278dd65 100644
--- a/llvm/lib/CodeGen/AtomicExpandPass.cpp
+++ b/llvm/lib/CodeGen/AtomicExpandPass.cpp
@@ -1915,7 +1915,6 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
// TODO: the "order" argument type is "int", not int32. So
// getInt32Ty may be wrong if the arch uses e.g. 16-bit ints.
- ConstantInt *SizeVal64 = ConstantInt::get(Type::getInt64Ty(Ctx), Size);
assert(Ordering != AtomicOrdering::NotAtomic && "expect atomic MO");
Constant *OrderingVal =
ConstantInt::get(Type::getInt32Ty(Ctx), (int)toCABI(Ordering));
@@ -2012,7 +2011,7 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
if (CASExpected) {
AllocaCASExpected = AllocaBuilder.CreateAlloca(CASExpected->getType());
AllocaCASExpected->setAlignment(AllocaAlignment);
- Builder.CreateLifetimeStart(AllocaCASExpected, SizeVal64);
+ Builder.CreateLifetimeStart(AllocaCASExpected);
Builder.CreateAlignedStore(CASExpected, AllocaCASExpected, AllocaAlignment);
Args.push_back(AllocaCASExpected);
}
@@ -2026,7 +2025,7 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
} else {
AllocaValue = AllocaBuilder.CreateAlloca(ValueOperand->getType());
AllocaValue->setAlignment(AllocaAlignment);
- Builder.CreateLifetimeStart(AllocaValue, SizeVal64);
+ Builder.CreateLifetimeStart(AllocaValue);
Builder.CreateAlignedStore(ValueOperand, AllocaValue, AllocaAlignment);
Args.push_back(AllocaValue);
}
@@ -2036,7 +2035,7 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
if (!CASExpected && HasResult && !UseSizedLibcall) {
AllocaResult = AllocaBuilder.CreateAlloca(I->getType());
AllocaResult->setAlignment(AllocaAlignment);
- Builder.CreateLifetimeStart(AllocaResult, SizeVal64);
+ Builder.CreateLifetimeStart(AllocaResult);
Args.push_back(AllocaResult);
}
@@ -2069,7 +2068,7 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
// And then, extract the results...
if (ValueOperand && !UseSizedLibcall)
- Builder.CreateLifetimeEnd(AllocaValue, SizeVal64);
+ Builder.CreateLifetimeEnd(AllocaValue);
if (CASExpected) {
// The final result from the CAS is {load of 'expected' alloca, bool result
@@ -2078,7 +2077,7 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
Value *V = PoisonValue::get(FinalResultTy);
Value *ExpectedOut = Builder.CreateAlignedLoad(
CASExpected->getType(), AllocaCASExpected, AllocaAlignment);
- Builder.CreateLifetimeEnd(AllocaCASExpected, SizeVal64);
+ Builder.CreateLifetimeEnd(AllocaCASExpected);
V = Builder.CreateInsertValue(V, ExpectedOut, 0);
V = Builder.CreateInsertValue(V, Result, 1);
I->replaceAllUsesWith(V);
@@ -2089,7 +2088,7 @@ bool AtomicExpandImpl::expandAtomicOpToLibcall(
else {
V = Builder.CreateAlignedLoad(I->getType(), AllocaResult,
AllocaAlignment);
- Builder.CreateLifetimeEnd(AllocaResult, SizeVal64);
+ Builder.CreateLifetimeEnd(AllocaResult);
}
I->replaceAllUsesWith(V);
}
diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index bbfae57..d30dfa7 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -2209,7 +2209,7 @@ bool IRTranslator::translateKnownIntrinsic(const CallInst &CI, Intrinsic::ID ID,
unsigned Op = ID == Intrinsic::lifetime_start ? TargetOpcode::LIFETIME_START
: TargetOpcode::LIFETIME_END;
- const AllocaInst *AI = dyn_cast<AllocaInst>(CI.getArgOperand(1));
+ const AllocaInst *AI = dyn_cast<AllocaInst>(CI.getArgOperand(0));
if (!AI || !AI->isStaticAlloca())
return true;
diff --git a/llvm/lib/CodeGen/RegisterPressure.cpp b/llvm/lib/CodeGen/RegisterPressure.cpp
index ca51b67..5f37890 100644
--- a/llvm/lib/CodeGen/RegisterPressure.cpp
+++ b/llvm/lib/CodeGen/RegisterPressure.cpp
@@ -1001,7 +1001,7 @@ static void computeMaxPressureDelta(ArrayRef<unsigned> OldMaxPressureVec,
++CritIdx;
if (CritIdx != CritEnd && CriticalPSets[CritIdx].getPSet() == i) {
- int PDiff = (int)PNew - (int)CriticalPSets[CritIdx].getUnitInc();
+ int PDiff = (int)PNew - CriticalPSets[CritIdx].getUnitInc();
if (PDiff > 0) {
Delta.CriticalMax = PressureChange(i);
Delta.CriticalMax.setUnitInc(PDiff);
@@ -1191,7 +1191,7 @@ getUpwardPressureDelta(const MachineInstr *MI, /*const*/ PressureDiff &PDiff,
++CritIdx;
if (CritIdx != CritEnd && CriticalPSets[CritIdx].getPSet() == PSetID) {
- int CritInc = (int)MNew - (int)CriticalPSets[CritIdx].getUnitInc();
+ int CritInc = (int)MNew - CriticalPSets[CritIdx].getUnitInc();
if (CritInc > 0 && CritInc <= std::numeric_limits<int16_t>::max()) {
Delta.CriticalMax = PressureChange(PSetID);
Delta.CriticalMax.setUnitInc(CritInc);
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 7341914..17703f5 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -12843,22 +12843,21 @@ SDValue DAGCombiner::visitMHISTOGRAM(SDNode *N) {
SDLoc DL(HG);
EVT MemVT = HG->getMemoryVT();
+ EVT DataVT = Index.getValueType();
MachineMemOperand *MMO = HG->getMemOperand();
ISD::MemIndexType IndexType = HG->getIndexType();
if (ISD::isConstantSplatVectorAllZeros(Mask.getNode()))
return Chain;
- SDValue Ops[] = {Chain, Inc, Mask, BasePtr, Index,
- HG->getScale(), HG->getIntID()};
- if (refineUniformBase(BasePtr, Index, HG->isIndexScaled(), DAG, DL))
+ if (refineUniformBase(BasePtr, Index, HG->isIndexScaled(), DAG, DL) ||
+ refineIndexType(Index, IndexType, DataVT, DAG)) {
+ SDValue Ops[] = {Chain, Inc, Mask, BasePtr, Index,
+ HG->getScale(), HG->getIntID()};
return DAG.getMaskedHistogram(DAG.getVTList(MVT::Other), MemVT, DL, Ops,
MMO, IndexType);
+ }
- EVT DataVT = Index.getValueType();
- if (refineIndexType(Index, IndexType, DataVT, DAG))
- return DAG.getMaskedHistogram(DAG.getVTList(MVT::Other), MemVT, DL, Ops,
- MMO, IndexType);
return SDValue();
}
@@ -16343,6 +16342,38 @@ SDValue DAGCombiner::visitTRUNCATE(SDNode *N) {
DAG, DL);
}
break;
+ case ISD::ABDU:
+ case ISD::ABDS:
+ // (trunc (abdu/abds a, b)) → (abdu/abds (trunc a), (trunc b))
+ if (!LegalOperations || N0.hasOneUse()) {
+ EVT SrcVT = N0.getValueType();
+ EVT TruncVT = VT;
+ unsigned SrcBits = SrcVT.getScalarSizeInBits();
+ unsigned TruncBits = TruncVT.getScalarSizeInBits();
+ unsigned NeededBits = SrcBits - TruncBits;
+
+ SDValue A = N0.getOperand(0);
+ SDValue B = N0.getOperand(1);
+ bool CanFold = false;
+
+ if (N0.getOpcode() == ISD::ABDU) {
+ KnownBits KnownA = DAG.computeKnownBits(A);
+ KnownBits KnownB = DAG.computeKnownBits(B);
+ CanFold = KnownA.countMinLeadingZeros() >= NeededBits &&
+ KnownB.countMinLeadingZeros() >= NeededBits;
+ } else {
+ unsigned SignBitsA = DAG.ComputeNumSignBits(A);
+ unsigned SignBitsB = DAG.ComputeNumSignBits(B);
+ CanFold = SignBitsA > NeededBits && SignBitsB > NeededBits;
+ }
+
+ if (CanFold && TLI.isOperationLegal(N0.getOpcode(), VT)) {
+ SDValue NewA = DAG.getNode(ISD::TRUNCATE, DL, TruncVT, A);
+ SDValue NewB = DAG.getNode(ISD::TRUNCATE, DL, TruncVT, B);
+ return DAG.getNode(N0.getOpcode(), DL, TruncVT, NewA, NewB);
+ }
+ }
+ break;
}
return SDValue();
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index b9e72c9..5ef1746 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -1371,7 +1371,7 @@ void SelectionDAG::init(MachineFunction &NewMF,
const TargetLibraryInfo *LibraryInfo,
UniformityInfo *NewUA, ProfileSummaryInfo *PSIin,
BlockFrequencyInfo *BFIin, MachineModuleInfo &MMIin,
- FunctionVarLocs const *VarLocs, bool HasDivergency) {
+ FunctionVarLocs const *VarLocs) {
MF = &NewMF;
SDAGISelPass = PassPtr;
ORE = &NewORE;
@@ -1384,7 +1384,6 @@ void SelectionDAG::init(MachineFunction &NewMF,
BFI = BFIin;
MMI = &MMIin;
FnVarLocs = VarLocs;
- DivergentTarget = HasDivergency;
}
SelectionDAG::~SelectionDAG() {
@@ -2331,8 +2330,7 @@ SDValue SelectionDAG::getRegister(Register Reg, EVT VT) {
return SDValue(E, 0);
auto *N = newSDNode<RegisterSDNode>(Reg, VTs);
- N->SDNodeBits.IsDivergent =
- DivergentTarget && TLI->isSDNodeSourceOfDivergence(N, FLI, UA);
+ N->SDNodeBits.IsDivergent = TLI->isSDNodeSourceOfDivergence(N, FLI, UA);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
@@ -12264,8 +12262,6 @@ static bool gluePropagatesDivergence(const SDNode *Node) {
}
bool SelectionDAG::calculateDivergence(SDNode *N) {
- if (!DivergentTarget)
- return false;
if (TLI->isSDNodeAlwaysUniform(N)) {
assert(!TLI->isSDNodeSourceOfDivergence(N, FLI, UA) &&
"Conflicting divergence information!");
@@ -12285,8 +12281,6 @@ bool SelectionDAG::calculateDivergence(SDNode *N) {
}
void SelectionDAG::updateDivergence(SDNode *N) {
- if (!DivergentTarget)
- return;
SmallVector<SDNode *, 16> Worklist(1, N);
do {
N = Worklist.pop_back_val();
@@ -13847,20 +13841,16 @@ void SelectionDAG::createOperands(SDNode *Node, ArrayRef<SDValue> Vals) {
Ops[I].setInitial(Vals[I]);
EVT VT = Ops[I].getValueType();
- // Take care of the Node's operands iff target has divergence
// Skip Chain. It does not carry divergence.
- if (DivergentTarget && VT != MVT::Other &&
+ if (VT != MVT::Other &&
(VT != MVT::Glue || gluePropagatesDivergence(Ops[I].getNode())) &&
Ops[I].getNode()->isDivergent()) {
- // Node is going to be divergent if at least one of its operand is
- // divergent, unless it belongs to the "AlwaysUniform" exemptions.
IsDivergent = true;
}
}
Node->NumOperands = Vals.size();
Node->OperandList = Ops;
- // Check the divergence of the Node itself.
- if (DivergentTarget && !TLI->isSDNodeAlwaysUniform(Node)) {
+ if (!TLI->isSDNodeAlwaysUniform(Node)) {
IsDivergent |= TLI->isSDNodeSourceOfDivergence(Node, FLI, UA);
Node->SDNodeBits.IsDivergent = IsDivergent;
}
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index f5f5c14..48ab797 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -3977,6 +3977,11 @@ void SelectionDAGBuilder::visitSIToFP(const User &I) {
setValue(&I, DAG.getNode(ISD::SINT_TO_FP, getCurSDLoc(), DestVT, N));
}
+void SelectionDAGBuilder::visitPtrToAddr(const User &I) {
+ // FIXME: this is not correct for pointers with addr width != pointer width
+ visitPtrToInt(I);
+}
+
void SelectionDAGBuilder::visitPtrToInt(const User &I) {
// What to do depends on the size of the integer and the size of the pointer.
// We can either truncate, zero extend, or no-op, accordingly.
@@ -7597,7 +7602,7 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I,
if (TM.getOptLevel() == CodeGenOptLevel::None)
return;
- const AllocaInst *LifetimeObject = dyn_cast<AllocaInst>(I.getArgOperand(1));
+ const AllocaInst *LifetimeObject = dyn_cast<AllocaInst>(I.getArgOperand(0));
if (!LifetimeObject)
return;
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
index c251755..e0835e6 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
@@ -574,6 +574,7 @@ private:
void visitFPToSI(const User &I);
void visitUIToFP(const User &I);
void visitSIToFP(const User &I);
+ void visitPtrToAddr(const User &I);
void visitPtrToInt(const User &I);
void visitIntToPtr(const User &I);
void visitBitCast(const User &I);
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
index 26071ed..ece50ed 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
@@ -480,10 +480,7 @@ void SelectionDAGISel::initializeAnalysisResults(
MachineModuleInfo &MMI =
MAMP.getCachedResult<MachineModuleAnalysis>(*Fn.getParent())->getMMI();
- TTI = &FAM.getResult<TargetIRAnalysis>(Fn);
-
- CurDAG->init(*MF, *ORE, MFAM, LibInfo, UA, PSI, BFI, MMI, FnVarLocs,
- TTI->hasBranchDivergence(&Fn));
+ CurDAG->init(*MF, *ORE, MFAM, LibInfo, UA, PSI, BFI, MMI, FnVarLocs);
// Now get the optional analyzes if we want to.
// This is based on the possibly changed OptLevel (after optnone is taken
@@ -501,6 +498,10 @@ void SelectionDAGISel::initializeAnalysisResults(
BatchAA = std::nullopt;
SP = &FAM.getResult<SSPLayoutAnalysis>(Fn);
+
+#if !defined(NDEBUG) && LLVM_ENABLE_ABI_BREAKING_CHECKS
+ TTI = &FAM.getResult<TargetIRAnalysis>(Fn);
+#endif
}
void SelectionDAGISel::initializeAnalysisResults(MachineFunctionPass &MFP) {
@@ -536,10 +537,7 @@ void SelectionDAGISel::initializeAnalysisResults(MachineFunctionPass &MFP) {
MachineModuleInfo &MMI =
MFP.getAnalysis<MachineModuleInfoWrapperPass>().getMMI();
- TTI = &MFP.getAnalysis<TargetTransformInfoWrapperPass>().getTTI(Fn);
-
- CurDAG->init(*MF, *ORE, &MFP, LibInfo, UA, PSI, BFI, MMI, FnVarLocs,
- TTI->hasBranchDivergence(&Fn));
+ CurDAG->init(*MF, *ORE, &MFP, LibInfo, UA, PSI, BFI, MMI, FnVarLocs);
// Now get the optional analyzes if we want to.
// This is based on the possibly changed OptLevel (after optnone is taken
@@ -558,6 +556,10 @@ void SelectionDAGISel::initializeAnalysisResults(MachineFunctionPass &MFP) {
BatchAA = std::nullopt;
SP = &MFP.getAnalysis<StackProtector>().getLayoutInfo();
+
+#if !defined(NDEBUG) && LLVM_ENABLE_ABI_BREAKING_CHECKS
+ TTI = &MFP.getAnalysis<TargetTransformInfoWrapperPass>().getTTI(Fn);
+#endif
}
bool SelectionDAGISel::runOnMachineFunction(MachineFunction &mf) {
diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index bf4c9f9..d80a229 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -1893,6 +1893,7 @@ int TargetLoweringBase::InstructionOpcodeToISD(unsigned Opcode) const {
case SIToFP: return ISD::SINT_TO_FP;
case FPTrunc: return ISD::FP_ROUND;
case FPExt: return ISD::FP_EXTEND;
+ case PtrToAddr: return ISD::BITCAST;
case PtrToInt: return ISD::BITCAST;
case IntToPtr: return ISD::BITCAST;
case BitCast: return ISD::BITCAST;
generated by cgit v1.1 at 2025-08-09 11:30:04 +0000