Merged main:a4823377fd2f into amd-gfx:21a2c4086495

Local branch amd-gfx 21a2c40 Merged main:ab97c9bdb747 into amd-gfx:89f031ec9f94 Remote branch main a482337 [ARM] Add basic masked load/store costs
jaebaek · Dec 12, 2020 · dd34bef · dd34bef
2 parents 21a2c40 + a482337
commit dd34bef
Show file tree

Hide file tree

Showing 3 changed files with 418 additions and 396 deletions.
diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -1318,6 +1318,24 @@ int ARMTTIImpl::getMemoryOpCost(unsigned Opcode, Type *Src,
                                            CostKind, I);
 }
 
+unsigned ARMTTIImpl::getMaskedMemoryOpCost(unsigned Opcode, Type *Src,
+                                           Align Alignment,
+                                           unsigned AddressSpace,
+                                           TTI::TargetCostKind CostKind) {
+  if (ST->hasMVEIntegerOps()) {
+    if (Opcode == Instruction::Load && isLegalMaskedLoad(Src, Alignment))
+      return ST->getMVEVectorCostFactor();
+    if (Opcode == Instruction::Store && isLegalMaskedStore(Src, Alignment))
+      return ST->getMVEVectorCostFactor();
+  }
+  if (!isa<FixedVectorType>(Src))
+    return BaseT::getMaskedMemoryOpCost(Opcode, Src, Alignment, AddressSpace,
+                                        CostKind);
+  // Scalar cost, which is currently very high due to the efficiency of the
+  // generated code.
+  return cast<FixedVectorType>(Src)->getNumElements() * 8;
+}
+
 int ARMTTIImpl::getInterleavedMemoryOpCost(
     unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
     Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,

diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.h b/llvm/lib/Target/ARM/ARMTargetTransformInfo.h
@@ -229,6 +229,10 @@ class ARMTTIImpl : public BasicTTIImplBase<ARMTTIImpl> {
                       TTI::TargetCostKind CostKind,
                       const Instruction *I = nullptr);
 
+  unsigned getMaskedMemoryOpCost(unsigned Opcode, Type *Src, Align Alignment,
+                                 unsigned AddressSpace,
+                                 TTI::TargetCostKind CostKind);
+
   int getInterleavedMemoryOpCost(
       unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
       Align Alignment, unsigned AddressSpace,