[SVE] Remove usages of VectorType::getNumElements() from AMDGPU

Reviewers: efriedma, arsenm, david-arm, fpetrogalli

Reviewed By: efriedma

Subscribers: dmgreen, arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, tschuett, hiraditya, rkruppe, psnobl, kerbowa, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D79807

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
index 1acc7b0..fb5b425 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
@@ -255,7 +255,7 @@ Type *AMDGPUCodeGenPrepare::getI32Ty(IRBuilder<> &B, const Type *T) const {
 
   if (T->isIntegerTy())
     return B.getInt32Ty();
-  return VectorType::get(B.getInt32Ty(), cast<VectorType>(T)->getNumElements());
+  return FixedVectorType::get(B.getInt32Ty(), cast<FixedVectorType>(T));
 }
 
 bool AMDGPUCodeGenPrepare::isSigned(const BinaryOperator &I) const {
@@ -477,7 +477,7 @@ bool AMDGPUCodeGenPrepare::isU24(Value *V, unsigned ScalarSize) const {
 
 static void extractValues(IRBuilder<> &Builder,
                           SmallVectorImpl<Value *> &Values, Value *V) {
-  VectorType *VT = dyn_cast<VectorType>(V->getType());
+  auto *VT = dyn_cast<FixedVectorType>(V->getType());
   if (!VT) {
     Values.push_back(V);
     return;
@@ -777,7 +777,7 @@ bool AMDGPUCodeGenPrepare::visitFDiv(BinaryOperator &FDiv) {
   Value *Den = FDiv.getOperand(1);
 
   Value *NewFDiv = nullptr;
-  if (VectorType *VT = dyn_cast<VectorType>(FDiv.getType())) {
+  if (auto *VT = dyn_cast<FixedVectorType>(FDiv.getType())) {
     NewFDiv = UndefValue::get(VT);
 
     // FIXME: Doesn't do the right thing for cases where the vector is partially
@@ -1233,7 +1233,7 @@ bool AMDGPUCodeGenPrepare::visitBinaryOperator(BinaryOperator &I) {
     IRBuilder<> Builder(&I);
     Builder.SetCurrentDebugLocation(I.getDebugLoc());
 
-    if (VectorType *VT = dyn_cast<VectorType>(Ty)) {
+    if (auto *VT = dyn_cast<FixedVectorType>(Ty)) {
       NewDiv = UndefValue::get(VT);
 
       for (unsigned N = 0, E = VT->getNumElements(); N != E; ++N) {

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
index 4e726c3..a077364 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
@@ -186,7 +186,7 @@ std::string MetadataStreamerV2::getTypeName(Type *Ty, bool Signed) const {
   case Type::DoubleTyID:
     return "double";
   case Type::FixedVectorTyID: {
-    auto VecTy = cast<VectorType>(Ty);
+    auto VecTy = cast<FixedVectorType>(Ty);
     auto ElTy = VecTy->getElementType();
     auto NumElements = VecTy->getNumElements();
     return (Twine(getTypeName(ElTy, Signed)) + Twine(NumElements)).str();
@@ -633,7 +633,7 @@ std::string MetadataStreamerV3::getTypeName(Type *Ty, bool Signed) const {
   case Type::DoubleTyID:
     return "double";
   case Type::FixedVectorTyID: {
-    auto VecTy = cast<VectorType>(Ty);
+    auto VecTy = cast<FixedVectorType>(Ty);
     auto ElTy = VecTy->getElementType();
     auto NumElements = VecTy->getNumElements();
     return (Twine(getTypeName(ElTy, Signed)) + Twine(NumElements)).str();

diff --git a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
index 4fb8583..86e437e 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
@@ -1126,8 +1126,8 @@ bool AMDGPULibCalls::fold_pow(CallInst *CI, IRBuilder<> &B,
     Type* rTy = opr0->getType();
     Type* nTyS = eltType->isDoubleTy() ? B.getInt64Ty() : B.getInt32Ty();
     Type *nTy = nTyS;
-    if (const VectorType *vTy = dyn_cast<VectorType>(rTy))
-      nTy = VectorType::get(nTyS, vTy->getNumElements());
+    if (const auto *vTy = dyn_cast<FixedVectorType>(rTy))
+      nTy = FixedVectorType::get(nTyS, vTy);
     unsigned size = nTy->getScalarSizeInBits();
     opr_n = CI->getArgOperand(1);
     if (opr_n->getType()->isIntegerTy())

diff --git a/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp b/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp
index b6de5cb..52e192e 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULowerKernelArguments.cpp
@@ -135,7 +135,7 @@ bool AMDGPULowerKernelArguments::runOnFunction(Function &F) {
         continue;
     }
 
-    VectorType *VT = dyn_cast<VectorType>(ArgTy);
+    auto *VT = dyn_cast<FixedVectorType>(ArgTy);
     bool IsV3 = VT && VT->getNumElements() == 3;
     bool DoShiftOpt = Size < 32 && !ArgTy->isAggregateType();
 

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPrintfRuntimeBinding.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPrintfRuntimeBinding.cpp
index 543d38d..e6cdbe9 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUPrintfRuntimeBinding.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPrintfRuntimeBinding.cpp
@@ -218,10 +218,10 @@ bool AMDGPUPrintfRuntimeBinding::lowerPrintfForGpu(
         //
         if (ArgSize % DWORD_ALIGN != 0) {
           llvm::Type *ResType = llvm::Type::getInt32Ty(Ctx);
-          VectorType *LLVMVecType = llvm::dyn_cast<llvm::VectorType>(ArgType);
+          auto *LLVMVecType = llvm::dyn_cast<llvm::FixedVectorType>(ArgType);
           int NumElem = LLVMVecType ? LLVMVecType->getNumElements() : 1;
           if (LLVMVecType && NumElem > 1)
-            ResType = llvm::VectorType::get(ResType, NumElem);
+            ResType = llvm::FixedVectorType::get(ResType, NumElem);
           Builder.SetInsertPoint(CI);
           Builder.SetCurrentDebugLocation(CI->getDebugLoc());
           if (OpConvSpecifiers[ArgCount - 1] == 'x' ||
@@ -479,7 +479,7 @@ bool AMDGPUPrintfRuntimeBinding::lowerPrintfForGpu(
           }
         } else if (isa<FixedVectorType>(ArgType)) {
           Type *IType = NULL;
-          uint32_t EleCount = cast<VectorType>(ArgType)->getNumElements();
+          uint32_t EleCount = cast<FixedVectorType>(ArgType)->getNumElements();
           uint32_t EleSize = ArgType->getScalarSizeInBits();
           uint32_t TotalSize = EleCount * EleSize;
           if (EleCount == 3) {

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
index 44fa99f..ae691b7 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -297,9 +297,9 @@ Value *AMDGPUPromoteAlloca::getWorkitemID(IRBuilder<> &Builder, unsigned N) {
   return CI;
 }
 
-static VectorType *arrayTypeToVecType(ArrayType *ArrayTy) {
-  return VectorType::get(ArrayTy->getElementType(),
-                         ArrayTy->getNumElements());
+static FixedVectorType *arrayTypeToVecType(ArrayType *ArrayTy) {
+  return FixedVectorType::get(ArrayTy->getElementType(),
+                              ArrayTy->getNumElements());
 }
 
 static Value *stripBitcasts(Value *V) {
@@ -390,7 +390,7 @@ static bool tryPromoteAllocaToVector(AllocaInst *Alloca, const DataLayout &DL) {
   }
 
   Type *AllocaTy = Alloca->getAllocatedType();
-  VectorType *VectorTy = dyn_cast<VectorType>(AllocaTy);
+  auto *VectorTy = dyn_cast<FixedVectorType>(AllocaTy);
   if (auto *ArrayTy = dyn_cast<ArrayType>(AllocaTy)) {
     if (VectorType::isValidElementType(ArrayTy->getElementType()) &&
         ArrayTy->getNumElements() > 0)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPURewriteOutArguments.cpp b/llvm/lib/Target/AMDGPU/AMDGPURewriteOutArguments.cpp
index 5a2cd48..9c3d96d 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPURewriteOutArguments.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURewriteOutArguments.cpp
@@ -208,8 +208,8 @@ bool AMDGPURewriteOutArguments::doInitialization(Module &M) {
 
 #ifndef NDEBUG
 bool AMDGPURewriteOutArguments::isVec3ToVec4Shuffle(Type *Ty0, Type* Ty1) const {
-  VectorType *VT0 = dyn_cast<VectorType>(Ty0);
-  VectorType *VT1 = dyn_cast<VectorType>(Ty1);
+  auto *VT0 = dyn_cast<FixedVectorType>(Ty0);
+  auto *VT1 = dyn_cast<FixedVectorType>(Ty1);
   if (!VT0 || !VT1)
     return false;
 

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index d4048cf..faa3d85 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -909,7 +909,7 @@ bool GCNTTIImpl::rewriteIntrinsicWithAddressSpace(
 unsigned GCNTTIImpl::getShuffleCost(TTI::ShuffleKind Kind, VectorType *VT,
                                     int Index, VectorType *SubTp) {
   if (ST->hasVOP3PInsts()) {
-    if (VT->getNumElements() == 2 &&
+    if (cast<FixedVectorType>(VT)->getNumElements() == 2 &&
         DL.getTypeSizeInBits(VT->getElementType()) == 16) {
       // With op_sel VOP3P instructions freely can access the low half or high
       // half of a register, so any swizzle is free.

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index d810013..550d4c1 100644 (file)
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -938,9 +938,8 @@ unsigned SITargetLowering::getVectorTypeBreakdownForCallingConv(
 static EVT memVTFromImageData(Type *Ty, unsigned DMaskLanes) {
   assert(DMaskLanes != 0);
 
-  if (auto *VT = dyn_cast<VectorType>(Ty)) {
-    unsigned NumElts = std::min(DMaskLanes,
-                                static_cast<unsigned>(VT->getNumElements()));
+  if (auto *VT = dyn_cast<FixedVectorType>(Ty)) {
+    unsigned NumElts = std::min(DMaskLanes, VT->getNumElements());
     return EVT::getVectorVT(Ty->getContext(),
                             EVT::getEVT(VT->getElementType()),
                             NumElts);