Removed parameter "Consecutive" from isLegalMaskedLoad() / isLegalMaskedStore().

Originally I planned to use the same interface for masked gather/scatter and set isConsecutive to "false" in this case.

Now I'm implementing masked gather/scatter and see that the interface is inconvenient. I want to add interfaces isLegalMaskedGather() / isLegalMaskedScatter() instead of using the "Consecutive" parameter in the existing interfaces.

Differential Revision: http://reviews.llvm.org/D13850

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@250686 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/Analysis/TargetTransformInfo.h b/include/llvm/Analysis/TargetTransformInfo.h
index da94613e26a9c26d7fd45a382083c133e2775442..084fd45f27fb7ddb7eb7694888e76edf292cce33 100644 (file)
--- a/include/llvm/Analysis/TargetTransformInfo.h
+++ b/include/llvm/Analysis/TargetTransformInfo.h
@@ -310,12 +310,11 @@ public:
                              bool HasBaseReg, int64_t Scale,
                              unsigned AddrSpace = 0) const;
 
-  /// \brief Return true if the target works with masked instruction
-  /// AVX2 allows masks for consecutive load and store for i32 and i64 elements.
-  /// AVX-512 architecture will also allow masks for non-consecutive memory
-  /// accesses.
-  bool isLegalMaskedStore(Type *DataType, int Consecutive) const;
-  bool isLegalMaskedLoad(Type *DataType, int Consecutive) const;
+  /// \brief Return true if the target supports masked load/store
+  /// AVX2 and AVX-512 targets allow masks for consecutive load and store for
+  /// 32 and 64 bit elements.
+  bool isLegalMaskedStore(Type *DataType) const;
+  bool isLegalMaskedLoad(Type *DataType) const;
 
   /// \brief Return the cost of the scaling factor used in the addressing
   /// mode represented by AM for this target, for a load/store
@@ -568,8 +567,8 @@ public:
                                      int64_t BaseOffset, bool HasBaseReg,
                                      int64_t Scale,
                                      unsigned AddrSpace) = 0;
-  virtual bool isLegalMaskedStore(Type *DataType, int Consecutive) = 0;
-  virtual bool isLegalMaskedLoad(Type *DataType, int Consecutive) = 0;
+  virtual bool isLegalMaskedStore(Type *DataType) = 0;
+  virtual bool isLegalMaskedLoad(Type *DataType) = 0;
   virtual int getScalingFactorCost(Type *Ty, GlobalValue *BaseGV,
                                    int64_t BaseOffset, bool HasBaseReg,
                                    int64_t Scale, unsigned AddrSpace) = 0;
@@ -693,11 +692,11 @@ public:
     return Impl.isLegalAddressingMode(Ty, BaseGV, BaseOffset, HasBaseReg,
                                       Scale, AddrSpace);
   }
-  bool isLegalMaskedStore(Type *DataType, int Consecutive) override {
-    return Impl.isLegalMaskedStore(DataType, Consecutive);
+  bool isLegalMaskedStore(Type *DataType) override {
+    return Impl.isLegalMaskedStore(DataType);
   }
-  bool isLegalMaskedLoad(Type *DataType, int Consecutive) override {
-    return Impl.isLegalMaskedLoad(DataType, Consecutive);
+  bool isLegalMaskedLoad(Type *DataType) override {
+    return Impl.isLegalMaskedLoad(DataType);
   }
   int getScalingFactorCost(Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset,
                            bool HasBaseReg, int64_t Scale,

diff --git a/include/llvm/Analysis/TargetTransformInfoImpl.h b/include/llvm/Analysis/TargetTransformInfoImpl.h
index 506c7f24c1ddc4f750da10e13dc85059e99e06a9..47609ff52902ccb513686877183c42917e58c019 100644 (file)
--- a/include/llvm/Analysis/TargetTransformInfoImpl.h
+++ b/include/llvm/Analysis/TargetTransformInfoImpl.h
@@ -209,9 +209,9 @@ public:
     return !BaseGV && BaseOffset == 0 && (Scale == 0 || Scale == 1);
   }
 
-  bool isLegalMaskedStore(Type *DataType, int Consecutive) { return false; }
+  bool isLegalMaskedStore(Type *DataType) { return false; }
 
-  bool isLegalMaskedLoad(Type *DataType, int Consecutive) { return false; }
+  bool isLegalMaskedLoad(Type *DataType) { return false; }
 
   int getScalingFactorCost(Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset,
                            bool HasBaseReg, int64_t Scale, unsigned AddrSpace) {

diff --git a/lib/Analysis/TargetTransformInfo.cpp b/lib/Analysis/TargetTransformInfo.cpp
index 75e940c74dd874e35a742d1c006d46a0072fab0d..427e1ed8ceb6f2feb028eb99560ef8ced44da956 100644 (file)
--- a/lib/Analysis/TargetTransformInfo.cpp
+++ b/lib/Analysis/TargetTransformInfo.cpp
@@ -113,14 +113,12 @@ bool TargetTransformInfo::isLegalAddressingMode(Type *Ty, GlobalValue *BaseGV,
                                         Scale, AddrSpace);
 }
 
-bool TargetTransformInfo::isLegalMaskedStore(Type *DataType,
-                                             int Consecutive) const {
-  return TTIImpl->isLegalMaskedStore(DataType, Consecutive);
+bool TargetTransformInfo::isLegalMaskedStore(Type *DataType) const {
+  return TTIImpl->isLegalMaskedStore(DataType);
 }
 
-bool TargetTransformInfo::isLegalMaskedLoad(Type *DataType,
-                                            int Consecutive) const {
-  return TTIImpl->isLegalMaskedLoad(DataType, Consecutive);
+bool TargetTransformInfo::isLegalMaskedLoad(Type *DataType) const {
+  return TTIImpl->isLegalMaskedLoad(DataType);
 }
 
 int TargetTransformInfo::getScalingFactorCost(Type *Ty, GlobalValue *BaseGV,

diff --git a/lib/CodeGen/CodeGenPrepare.cpp b/lib/CodeGen/CodeGenPrepare.cpp
index 27bf83a3c7b46ef20bfce04d32fcbebc1a0f5c39..ad007a43ecb3c0fe0019d55561bc0d512a9d7dcc 100644 (file)
--- a/lib/CodeGen/CodeGenPrepare.cpp
+++ b/lib/CodeGen/CodeGenPrepare.cpp
@@ -1384,7 +1384,7 @@ bool CodeGenPrepare::optimizeCallInst(CallInst *CI, bool& ModifiedDT) {
     }
     case Intrinsic::masked_load: {
       // Scalarize unsupported vector masked load
-      if (!TTI->isLegalMaskedLoad(CI->getType(), 1)) {
+      if (!TTI->isLegalMaskedLoad(CI->getType())) {
         ScalarizeMaskedLoad(CI);
         ModifiedDT = true;
         return true;
@@ -1392,7 +1392,7 @@ bool CodeGenPrepare::optimizeCallInst(CallInst *CI, bool& ModifiedDT) {
       return false;
     }
     case Intrinsic::masked_store: {
-      if (!TTI->isLegalMaskedStore(CI->getArgOperand(0)->getType(), 1)) {
+      if (!TTI->isLegalMaskedStore(CI->getArgOperand(0)->getType())) {
         ScalarizeMaskedStore(CI);
         ModifiedDT = true;
         return true;

diff --git a/lib/Target/X86/X86TargetTransformInfo.cpp b/lib/Target/X86/X86TargetTransformInfo.cpp
index c0a12a82a32d7bf488b8367dce13ba67bf78a783..e69b713b8af720dbdc67cda6876774d99ece25d5 100644 (file)
--- a/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -899,8 +899,8 @@ int X86TTIImpl::getMaskedMemoryOpCost(unsigned Opcode, Type *SrcTy,
   unsigned NumElem = SrcVTy->getVectorNumElements();
   VectorType *MaskTy =
     VectorType::get(Type::getInt8Ty(getGlobalContext()), NumElem);
-  if ((Opcode == Instruction::Load && !isLegalMaskedLoad(SrcVTy, 1)) ||
-      (Opcode == Instruction::Store && !isLegalMaskedStore(SrcVTy, 1)) ||
+  if ((Opcode == Instruction::Load && !isLegalMaskedLoad(SrcVTy)) ||
+      (Opcode == Instruction::Store && !isLegalMaskedStore(SrcVTy)) ||
       !isPowerOf2_32(NumElem)) {
     // Scalarization
     int MaskSplitCost = getScalarizationOverhead(MaskTy, false, true);
@@ -1189,19 +1189,16 @@ int X86TTIImpl::getIntImmCost(Intrinsic::ID IID, unsigned Idx, const APInt &Imm,
   return X86TTIImpl::getIntImmCost(Imm, Ty);
 }
 
-bool X86TTIImpl::isLegalMaskedLoad(Type *DataTy, int Consecutive) {
-  int DataWidth = DataTy->getPrimitiveSizeInBits();
+bool X86TTIImpl::isLegalMaskedLoad(Type *DataTy) {
+  Type *ScalarTy = DataTy->getScalarType();
+  int DataWidth = ScalarTy->isPointerTy() ? DL.getPointerSizeInBits() :
+    ScalarTy->getPrimitiveSizeInBits();
 
-  // Todo: AVX512 allows gather/scatter, works with strided and random as well
-  if ((DataWidth < 32) || (Consecutive == 0))
-    return false;
-  if (ST->hasAVX512() || ST->hasAVX2())
-    return true;
-  return false;
+  return (DataWidth >= 32 && ST->hasAVX2());
 }
 
-bool X86TTIImpl::isLegalMaskedStore(Type *DataType, int Consecutive) {
-  return isLegalMaskedLoad(DataType, Consecutive);
+bool X86TTIImpl::isLegalMaskedStore(Type *DataType) {
+  return isLegalMaskedLoad(DataType);
 }
 
 bool X86TTIImpl::areInlineCompatible(const Function *Caller,

diff --git a/lib/Target/X86/X86TargetTransformInfo.h b/lib/Target/X86/X86TargetTransformInfo.h
index efcae715c56a132a99e76a9d8233c0b92d261955..5ddde2a45a570961bffdaa9678be1555a2b49711 100644 (file)
--- a/lib/Target/X86/X86TargetTransformInfo.h
+++ b/lib/Target/X86/X86TargetTransformInfo.h
@@ -88,8 +88,8 @@ public:
   int getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm, Type *Ty);
   int getIntImmCost(Intrinsic::ID IID, unsigned Idx, const APInt &Imm,
                     Type *Ty);
-  bool isLegalMaskedLoad(Type *DataType, int Consecutive);
-  bool isLegalMaskedStore(Type *DataType, int Consecutive);
+  bool isLegalMaskedLoad(Type *DataType);
+  bool isLegalMaskedStore(Type *DataType);
   bool areInlineCompatible(const Function *Caller,
                            const Function *Callee) const;
 

diff --git a/lib/Transforms/Vectorize/LoopVectorize.cpp b/lib/Transforms/Vectorize/LoopVectorize.cpp
index ec91e138e8642babd168141cfc8d2f5e5c9d14e0..37430fef30b722689df75450f1b3901ba02f8846 100644 (file)
--- a/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -1226,12 +1226,12 @@ public:
   /// Returns true if the target machine supports masked store operation
   /// for the given \p DataType and kind of access to \p Ptr.
   bool isLegalMaskedStore(Type *DataType, Value *Ptr) {
-    return TTI->isLegalMaskedStore(DataType, isConsecutivePtr(Ptr));
+    return isConsecutivePtr(Ptr) && TTI->isLegalMaskedStore(DataType);
   }
   /// Returns true if the target machine supports masked load operation
   /// for the given \p DataType and kind of access to \p Ptr.
   bool isLegalMaskedLoad(Type *DataType, Value *Ptr) {
-    return TTI->isLegalMaskedLoad(DataType, isConsecutivePtr(Ptr));
+    return isConsecutivePtr(Ptr) && TTI->isLegalMaskedLoad(DataType);
   }
   /// Returns true if vector representation of the instruction \p I
   /// requires mask.