[llvm] 1a75d88 - [X86] Move getGatherOverhead/getScatterOverhead into X86TargetTransformInfo.

[Craig Topper via llvm-commits]

Author: Craig Topper
Date: 2020-07-26T10:38:42-07:00
New Revision: 1a75d88b3eb4e0cc5494281f24f362ece3d4aa4a

URL: https://github.com/llvm/llvm-project/commit/1a75d88b3eb4e0cc5494281f24f362ece3d4aa4a
DIFF: https://github.com/llvm/llvm-project/commit/1a75d88b3eb4e0cc5494281f24f362ece3d4aa4a.diff

LOG: [X86] Move getGatherOverhead/getScatterOverhead into X86TargetTransformInfo.

These cost methods don't make much sense in X86Subtarget. Make
them methods in X86's TTI and move the feature checks from the
X86Subtarget constructor into these methods.

Reviewed By: RKSimon

Differential Revision: https://reviews.llvm.org/D84594

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86Subtarget.cpp
    llvm/lib/Target/X86/X86Subtarget.h
    llvm/lib/Target/X86/X86TargetTransformInfo.cpp
    llvm/lib/Target/X86/X86TargetTransformInfo.h

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86Subtarget.cpp b/llvm/lib/Target/X86/X86Subtarget.cpp
index 07e913e13911..095aea3a1063 100644
--- a/llvm/lib/Target/X86/X86Subtarget.cpp
+++ b/llvm/lib/Target/X86/X86Subtarget.cpp
@@ -262,17 +262,6 @@ void X86Subtarget::initSubtargetFeatures(StringRef CPU, StringRef FS) {
            isTargetKFreeBSD() || In64BitMode)
     stackAlignment = Align(16);
 
-  // Some CPUs have more overhead for gather. The specified overhead is relative
-  // to the Load operation. "2" is the number provided by Intel architects. This
-  // parameter is used for cost estimation of Gather Op and comparison with
-  // other alternatives.
-  // TODO: Remove the explicit hasAVX512()?, That would mean we would only
-  // enable gather with a -march.
-  if (hasAVX512() || (hasAVX2() && hasFastGather()))
-    GatherOverhead = 2;
-  if (hasAVX512())
-    ScatterOverhead = 2;
-
   // Consume the vector width attribute or apply any target specific limit.
   if (PreferVectorWidthOverride)
     PreferVectorWidth = PreferVectorWidthOverride;

diff  --git a/llvm/lib/Target/X86/X86Subtarget.h b/llvm/lib/Target/X86/X86Subtarget.h
index 5b5ab4b969aa..e555dfdd638a 100644
--- a/llvm/lib/Target/X86/X86Subtarget.h
+++ b/llvm/lib/Target/X86/X86Subtarget.h
@@ -506,10 +506,6 @@ class X86Subtarget final : public X86GenSubtargetInfo {
   /// True if compiling for 16-bit, false for 32-bit or 64-bit.
   bool In16BitMode = false;
 
-  /// Contains the Overhead of gather\scatter instructions
-  int GatherOverhead = 1024;
-  int ScatterOverhead = 1024;
-
   X86SelectionDAGInfo TSInfo;
   // Ordering here is important. X86InstrInfo initializes X86RegisterInfo which
   // X86TargetLowering needs.
@@ -678,8 +674,6 @@ class X86Subtarget final : public X86GenSubtargetInfo {
   bool isPMADDWDSlow() const { return IsPMADDWDSlow; }
   bool isUnalignedMem16Slow() const { return IsUAMem16Slow; }
   bool isUnalignedMem32Slow() const { return IsUAMem32Slow; }
-  int getGatherOverhead() const { return GatherOverhead; }
-  int getScatterOverhead() const { return ScatterOverhead; }
   bool hasSSEUnalignedMem() const { return HasSSEUnalignedMem; }
   bool hasCmpxchg16b() const { return HasCmpxchg16b && is64Bit(); }
   bool useLeaForSP() const { return UseLeaForSP; }

diff  --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index cc18e55656ef..491078fd0542 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -3848,6 +3848,26 @@ X86TTIImpl::getCFInstrCost(unsigned Opcode, TTI::TargetCostKind CostKind) {
   return CostKind == TTI::TCK_RecipThroughput ? 0 : 1;
 }
 
+int X86TTIImpl::getGatherOverhead() const {
+  // Some CPUs have more overhead for gather. The specified overhead is relative
+  // to the Load operation. "2" is the number provided by Intel architects. This
+  // parameter is used for cost estimation of Gather Op and comparison with
+  // other alternatives.
+  // TODO: Remove the explicit hasAVX512()?, That would mean we would only
+  // enable gather with a -march.
+  if (ST->hasAVX512() || (ST->hasAVX2() && ST->hasFastGather()))
+    return 2;
+
+  return 1024;
+}
+
+int X86TTIImpl::getScatterOverhead() const {
+  if (ST->hasAVX512())
+    return 2;
+
+  return 1024;
+}
+
 // Return an average cost of Gather / Scatter instruction, maybe improved later
 int X86TTIImpl::getGSVectorCost(unsigned Opcode, Type *SrcVTy, const Value *Ptr,
                                 Align Alignment, unsigned AddressSpace) {
@@ -3906,8 +3926,8 @@ int X86TTIImpl::getGSVectorCost(unsigned Opcode, Type *SrcVTy, const Value *Ptr,
   // The gather / scatter cost is given by Intel architects. It is a rough
   // number since we are looking at one instruction in a time.
   const int GSOverhead = (Opcode == Instruction::Load)
-                             ? ST->getGatherOverhead()
-                             : ST->getScatterOverhead();
+                             ? getGatherOverhead()
+                             : getScatterOverhead();
   return GSOverhead + VF * getMemoryOpCost(Opcode, SrcVTy->getScalarType(),
                                            MaybeAlign(Alignment), AddressSpace,
                                            TTI::TCK_RecipThroughput);

diff  --git a/llvm/lib/Target/X86/X86TargetTransformInfo.h b/llvm/lib/Target/X86/X86TargetTransformInfo.h
index ca875fa68523..5ccaf409790d 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.h
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.h
@@ -244,6 +244,9 @@ class X86TTIImpl : public BasicTTIImplBase<X86TTIImpl> {
   int getGSVectorCost(unsigned Opcode, Type *DataTy, const Value *Ptr,
                       Align Alignment, unsigned AddressSpace);
 
+  int getGatherOverhead() const;
+  int getScatterOverhead() const;
+
   /// @}
 };