diff options
Diffstat (limited to 'llvm/lib/Target')
-rw-r--r-- | llvm/lib/Target/AArch64/AArch64PromoteConstant.cpp | 3 | ||||
-rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp | 2 | ||||
-rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp | 4 | ||||
-rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp | 15 | ||||
-rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp | 8 | ||||
-rw-r--r-- | llvm/lib/Target/ARM/ARMParallelDSP.cpp | 6 | ||||
-rw-r--r-- | llvm/lib/Target/NVPTX/NVPTXLowerArgs.cpp | 3 | ||||
-rw-r--r-- | llvm/lib/Target/WebAssembly/WebAssemblyLowerEmscriptenEHSjLj.cpp | 11 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86ISelLowering.cpp | 6 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86InterleavedAccess.cpp | 15 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86WinEHState.cpp | 8 |
11 files changed, 44 insertions, 37 deletions
diff --git a/llvm/lib/Target/AArch64/AArch64PromoteConstant.cpp b/llvm/lib/Target/AArch64/AArch64PromoteConstant.cpp index 1da95bfb9a8..a594ecb71fc 100644 --- a/llvm/lib/Target/AArch64/AArch64PromoteConstant.cpp +++ b/llvm/lib/Target/AArch64/AArch64PromoteConstant.cpp @@ -493,7 +493,8 @@ void AArch64PromoteConstant::insertDefinitions(Function &F, for (const auto &IPI : InsertPts) { // Create the load of the global variable. IRBuilder<> Builder(IPI.first); - LoadInst *LoadedCst = Builder.CreateLoad(&PromotedGV); + LoadInst *LoadedCst = + Builder.CreateLoad(PromotedGV.getValueType(), &PromotedGV); LLVM_DEBUG(dbgs() << "**********\n"); LLVM_DEBUG(dbgs() << "New def: "); LLVM_DEBUG(LoadedCst->print(dbgs())); diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp index 38804723c92..75982075325 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp @@ -806,7 +806,7 @@ bool AMDGPUCodeGenPrepare::visitLoadInst(LoadInst &I) { Type *I32Ty = Builder.getInt32Ty(); Type *PT = PointerType::get(I32Ty, I.getPointerAddressSpace()); Value *BitCast= Builder.CreateBitCast(I.getPointerOperand(), PT); - LoadInst *WidenLoad = Builder.CreateLoad(BitCast); + LoadInst *WidenLoad = Builder.CreateLoad(I32Ty, BitCast); WidenLoad->copyMetadata(I); // If we have range metadata, we need to convert the type, and not make diff --git a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp index f259f8311ec..581e229b4a0 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp @@ -1357,12 +1357,12 @@ bool AMDGPULibCalls::fold_sincos(CallInst *CI, IRBuilder<> &B, if (!isSin) { // CI->cos, UI->sin B.SetInsertPoint(&*ItOld); UI->replaceAllUsesWith(&*Call); - Instruction *Reload = B.CreateLoad(Alloc); + Instruction *Reload = B.CreateLoad(Alloc->getAllocatedType(), Alloc); CI->replaceAllUsesWith(Reload); UI->eraseFromParent(); CI->eraseFromParent(); } else { // CI->sin, UI->cos - Instruction *Reload = B.CreateLoad(Alloc); + Instruction *Reload = B.CreateLoad(Alloc->getAllocatedType(), Alloc); UI->replaceAllUsesWith(Reload); CI->replaceAllUsesWith(Call); UI->eraseFromParent(); diff --git a/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp b/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp index aaad8b6e48a..b2637418ebb 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp @@ -132,6 +132,7 @@ bool AMDGPULowerKernelArguments::runOnFunction(Function &F) { KernArgBaseAlign); Value *ArgPtr; + Type *AdjustedArgTy; if (DoShiftOpt) { // FIXME: Handle aggregate types // Since we don't have sub-dword scalar loads, avoid doing an extload by // loading earlier than the argument address, and extracting the relevant @@ -144,25 +145,25 @@ bool AMDGPULowerKernelArguments::runOnFunction(Function &F) { KernArgSegment, AlignDownOffset, Arg.getName() + ".kernarg.offset.align.down"); - ArgPtr = Builder.CreateBitCast(ArgPtr, - Builder.getInt32Ty()->getPointerTo(AS), - ArgPtr->getName() + ".cast"); + AdjustedArgTy = Builder.getInt32Ty(); } else { ArgPtr = Builder.CreateConstInBoundsGEP1_64( KernArgSegment, EltOffset, Arg.getName() + ".kernarg.offset"); - ArgPtr = Builder.CreateBitCast(ArgPtr, ArgTy->getPointerTo(AS), - ArgPtr->getName() + ".cast"); + AdjustedArgTy = ArgTy; } if (IsV3 && Size >= 32) { V4Ty = VectorType::get(VT->getVectorElementType(), 4); // Use the hack that clang uses to avoid SelectionDAG ruining v3 loads - ArgPtr = Builder.CreateBitCast(ArgPtr, V4Ty->getPointerTo(AS)); + AdjustedArgTy = V4Ty; } - LoadInst *Load = Builder.CreateAlignedLoad(ArgPtr, AdjustedAlign); + ArgPtr = Builder.CreateBitCast(ArgPtr, AdjustedArgTy->getPointerTo(AS), + ArgPtr->getName() + ".cast"); + LoadInst *Load = + Builder.CreateAlignedLoad(AdjustedArgTy, ArgPtr, AdjustedAlign); Load->setMetadata(LLVMContext::MD_invariant_load, MDNode::get(Ctx, {})); MDBuilder MDB(Ctx); diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp index 8bf2b13c654..5f05ce7d2a2 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp @@ -245,10 +245,10 @@ AMDGPUPromoteAlloca::getLocalSizeYZ(IRBuilder<> &Builder) { // 32-bit and extract sequence is already present, and it is probably easier // to CSE this. The loads should be mergable later anyway. Value *GEPXY = Builder.CreateConstInBoundsGEP1_64(CastDispatchPtr, 1); - LoadInst *LoadXY = Builder.CreateAlignedLoad(GEPXY, 4); + LoadInst *LoadXY = Builder.CreateAlignedLoad(I32Ty, GEPXY, 4); Value *GEPZU = Builder.CreateConstInBoundsGEP1_64(CastDispatchPtr, 2); - LoadInst *LoadZU = Builder.CreateAlignedLoad(GEPZU, 4); + LoadInst *LoadZU = Builder.CreateAlignedLoad(I32Ty, GEPZU, 4); MDNode *MD = MDNode::get(Mod->getContext(), None); LoadXY->setMetadata(LLVMContext::MD_invariant_load, MD); @@ -426,7 +426,7 @@ static bool tryPromoteAllocaToVector(AllocaInst *Alloca) { Value *Index = calculateVectorIndex(Ptr, GEPVectorIdx); Value *BitCast = Builder.CreateBitCast(Alloca, VecPtrTy); - Value *VecValue = Builder.CreateLoad(BitCast); + Value *VecValue = Builder.CreateLoad(VectorTy, BitCast); Value *ExtractElement = Builder.CreateExtractElement(VecValue, Index); Inst->replaceAllUsesWith(ExtractElement); Inst->eraseFromParent(); @@ -441,7 +441,7 @@ static bool tryPromoteAllocaToVector(AllocaInst *Alloca) { Value *Ptr = SI->getPointerOperand(); Value *Index = calculateVectorIndex(Ptr, GEPVectorIdx); Value *BitCast = Builder.CreateBitCast(Alloca, VecPtrTy); - Value *VecValue = Builder.CreateLoad(BitCast); + Value *VecValue = Builder.CreateLoad(VectorTy, BitCast); Value *NewVecValue = Builder.CreateInsertElement(VecValue, SI->getValueOperand(), Index); diff --git a/llvm/lib/Target/ARM/ARMParallelDSP.cpp b/llvm/lib/Target/ARM/ARMParallelDSP.cpp index fc846c15834..9730c32577e 100644 --- a/llvm/lib/Target/ARM/ARMParallelDSP.cpp +++ b/llvm/lib/Target/ARM/ARMParallelDSP.cpp @@ -688,12 +688,12 @@ bool ARMParallelDSP::MatchSMLAD(Function &F) { } static LoadInst *CreateLoadIns(IRBuilder<NoFolder> &IRB, LoadInst &BaseLoad, - const Type *LoadTy) { + Type *LoadTy) { const unsigned AddrSpace = BaseLoad.getPointerAddressSpace(); Value *VecPtr = IRB.CreateBitCast(BaseLoad.getPointerOperand(), LoadTy->getPointerTo(AddrSpace)); - return IRB.CreateAlignedLoad(VecPtr, BaseLoad.getAlignment()); + return IRB.CreateAlignedLoad(LoadTy, VecPtr, BaseLoad.getAlignment()); } Instruction *ARMParallelDSP::CreateSMLADCall(LoadInst *VecLd0, LoadInst *VecLd1, @@ -709,7 +709,7 @@ Instruction *ARMParallelDSP::CreateSMLADCall(LoadInst *VecLd0, LoadInst *VecLd1, ++BasicBlock::iterator(InsertAfter)); // Replace the reduction chain with an intrinsic call - const Type *Ty = IntegerType::get(M->getContext(), 32); + Type *Ty = IntegerType::get(M->getContext(), 32); LoadInst *NewLd0 = CreateLoadIns(Builder, VecLd0[0], Ty); LoadInst *NewLd1 = CreateLoadIns(Builder, VecLd1[0], Ty); Value* Args[] = { NewLd0, NewLd1, Acc }; diff --git a/llvm/lib/Target/NVPTX/NVPTXLowerArgs.cpp b/llvm/lib/Target/NVPTX/NVPTXLowerArgs.cpp index 35fa61a1c1d..8c7f6e63e1f 100644 --- a/llvm/lib/Target/NVPTX/NVPTXLowerArgs.cpp +++ b/llvm/lib/Target/NVPTX/NVPTXLowerArgs.cpp @@ -169,7 +169,8 @@ void NVPTXLowerArgs::handleByValParam(Argument *Arg) { Value *ArgInParam = new AddrSpaceCastInst( Arg, PointerType::get(StructType, ADDRESS_SPACE_PARAM), Arg->getName(), FirstInst); - LoadInst *LI = new LoadInst(ArgInParam, Arg->getName(), FirstInst); + LoadInst *LI = + new LoadInst(StructType, ArgInParam, Arg->getName(), FirstInst); new StoreInst(LI, AllocA, FirstInst); } diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyLowerEmscriptenEHSjLj.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyLowerEmscriptenEHSjLj.cpp index 16f6af3ef64..98d519e1fa2 100644 --- a/llvm/lib/Target/WebAssembly/WebAssemblyLowerEmscriptenEHSjLj.cpp +++ b/llvm/lib/Target/WebAssembly/WebAssemblyLowerEmscriptenEHSjLj.cpp @@ -445,7 +445,8 @@ Value *WebAssemblyLowerEmscriptenEHSjLj::wrapInvoke(CallOrInvoke *CI) { // Post-invoke // %__THREW__.val = __THREW__; __THREW__ = 0; - Value *Threw = IRB.CreateLoad(ThrewGV, ThrewGV->getName() + ".val"); + Value *Threw = + IRB.CreateLoad(IRB.getInt32Ty(), ThrewGV, ThrewGV->getName() + ".val"); IRB.CreateStore(IRB.getInt32(0), ThrewGV); return Threw; } @@ -548,8 +549,8 @@ void WebAssemblyLowerEmscriptenEHSjLj::wrapTestSetjmp( BasicBlock *ElseBB1 = BasicBlock::Create(C, "if.else1", F); BasicBlock *EndBB1 = BasicBlock::Create(C, "if.end", F); Value *ThrewCmp = IRB.CreateICmpNE(Threw, IRB.getInt32(0)); - Value *ThrewValue = - IRB.CreateLoad(ThrewValueGV, ThrewValueGV->getName() + ".val"); + Value *ThrewValue = IRB.CreateLoad(IRB.getInt32Ty(), ThrewValueGV, + ThrewValueGV->getName() + ".val"); Value *ThrewValueCmp = IRB.CreateICmpNE(ThrewValue, IRB.getInt32(0)); Value *Cmp1 = IRB.CreateAnd(ThrewCmp, ThrewValueCmp, "cmp1"); IRB.CreateCondBr(Cmp1, ThenBB1, ElseBB1); @@ -561,8 +562,8 @@ void WebAssemblyLowerEmscriptenEHSjLj::wrapTestSetjmp( BasicBlock *EndBB2 = BasicBlock::Create(C, "if.end2", F); Value *ThrewInt = IRB.CreateIntToPtr(Threw, Type::getInt32PtrTy(C), Threw->getName() + ".i32p"); - Value *LoadedThrew = - IRB.CreateLoad(ThrewInt, ThrewInt->getName() + ".loaded"); + Value *LoadedThrew = IRB.CreateLoad(IRB.getInt32Ty(), ThrewInt, + ThrewInt->getName() + ".loaded"); Value *ThenLabel = IRB.CreateCall( TestSetjmpF, {LoadedThrew, SetjmpTable, SetjmpTableSize}, "label"); Value *Cmp2 = IRB.CreateICmpEQ(ThenLabel, IRB.getInt32(0)); diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 1e32eaf0654..00a9d7cbf1c 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -25362,7 +25362,6 @@ X86TargetLowering::lowerIdempotentRMWIntoFencedLoad(AtomicRMWInst *AI) const { // We must restrict the ordering to avoid generating loads with Release or // ReleaseAcquire orderings. auto Order = AtomicCmpXchgInst::getStrongestFailureOrdering(AI->getOrdering()); - auto Ptr = AI->getPointerOperand(); // Before the load we need a fence. Here is an example lifted from // http://www.hpl.hp.com/techreports/2012/HPL-2012-68.pdf showing why a fence @@ -25397,8 +25396,9 @@ X86TargetLowering::lowerIdempotentRMWIntoFencedLoad(AtomicRMWInst *AI) const { Builder.CreateCall(MFence, {}); // Finally we can emit the atomic load. - LoadInst *Loaded = Builder.CreateAlignedLoad(Ptr, - AI->getType()->getPrimitiveSizeInBits()); + LoadInst *Loaded = + Builder.CreateAlignedLoad(AI->getType(), AI->getPointerOperand(), + AI->getType()->getPrimitiveSizeInBits()); Loaded->setAtomic(Order, SSID); AI->replaceAllUsesWith(Loaded); AI->eraseFromParent(); diff --git a/llvm/lib/Target/X86/X86InterleavedAccess.cpp b/llvm/lib/Target/X86/X86InterleavedAccess.cpp index 0922883cd17..717ed2b0631 100644 --- a/llvm/lib/Target/X86/X86InterleavedAccess.cpp +++ b/llvm/lib/Target/X86/X86InterleavedAccess.cpp @@ -193,7 +193,7 @@ void X86InterleavedAccessGroup::decompose( // Decompose the load instruction. LoadInst *LI = cast<LoadInst>(VecInst); - Type *VecBasePtrTy = SubVecTy->getPointerTo(LI->getPointerAddressSpace()); + Type *VecBaseTy, *VecBasePtrTy; Value *VecBasePtr; unsigned int NumLoads = NumSubVectors; // In the case of stride 3 with a vector of 32 elements load the information @@ -201,18 +201,21 @@ void X86InterleavedAccessGroup::decompose( // [0,1...,VF/2-1,VF/2+VF,VF/2+VF+1,...,2VF-1] unsigned VecLength = DL.getTypeSizeInBits(VecWidth); if (VecLength == 768 || VecLength == 1536) { - Type *VecTran = - VectorType::get(Type::getInt8Ty(LI->getContext()), 16)->getPointerTo(); - VecBasePtr = Builder.CreateBitCast(LI->getPointerOperand(), VecTran); + VecBaseTy = VectorType::get(Type::getInt8Ty(LI->getContext()), 16); + VecBasePtrTy = VecBaseTy->getPointerTo(LI->getPointerAddressSpace()); + VecBasePtr = Builder.CreateBitCast(LI->getPointerOperand(), VecBasePtrTy); NumLoads = NumSubVectors * (VecLength / 384); - } else + } else { + VecBaseTy = SubVecTy; + VecBasePtrTy = VecBaseTy->getPointerTo(LI->getPointerAddressSpace()); VecBasePtr = Builder.CreateBitCast(LI->getPointerOperand(), VecBasePtrTy); + } // Generate N loads of T type. for (unsigned i = 0; i < NumLoads; i++) { // TODO: Support inbounds GEP. Value *NewBasePtr = Builder.CreateGEP(VecBasePtr, Builder.getInt32(i)); Instruction *NewLoad = - Builder.CreateAlignedLoad(NewBasePtr, LI->getAlignment()); + Builder.CreateAlignedLoad(VecBaseTy, NewBasePtr, LI->getAlignment()); DecomposedVectors.push_back(NewLoad); } } diff --git a/llvm/lib/Target/X86/X86WinEHState.cpp b/llvm/lib/Target/X86/X86WinEHState.cpp index 18d3e019e0b..3e31d99f656 100644 --- a/llvm/lib/Target/X86/X86WinEHState.cpp +++ b/llvm/lib/Target/X86/X86WinEHState.cpp @@ -432,7 +432,7 @@ void WinEHStatePass::linkExceptionRegistration(IRBuilder<> &Builder, // Next = [fs:00] Constant *FSZero = Constant::getNullValue(LinkTy->getPointerTo()->getPointerTo(257)); - Value *Next = Builder.CreateLoad(FSZero); + Value *Next = Builder.CreateLoad(LinkTy->getPointerTo(), FSZero); Builder.CreateStore(Next, Builder.CreateStructGEP(LinkTy, Link, 0)); // [fs:00] = Link Builder.CreateStore(Link, FSZero); @@ -447,8 +447,8 @@ void WinEHStatePass::unlinkExceptionRegistration(IRBuilder<> &Builder) { } Type *LinkTy = getEHLinkRegistrationType(); // [fs:00] = Link->Next - Value *Next = - Builder.CreateLoad(Builder.CreateStructGEP(LinkTy, Link, 0)); + Value *Next = Builder.CreateLoad(LinkTy->getPointerTo(), + Builder.CreateStructGEP(LinkTy, Link, 0)); Constant *FSZero = Constant::getNullValue(LinkTy->getPointerTo()->getPointerTo(257)); Builder.CreateStore(Next, FSZero); @@ -783,7 +783,7 @@ void WinEHStatePass::addStateStores(Function &F, WinEHFuncInfo &FuncInfo) { if (InCleanup) { Value *StateField = Builder.CreateStructGEP(nullptr, RegNode, StateFieldIndex); - State = Builder.CreateLoad(StateField); + State = Builder.CreateLoad(Builder.getInt32Ty(), StateField); } else { State = Builder.getInt32(getStateForCallSite(BlockColors, FuncInfo, CS)); } |