RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 12:12:47 +01:00
Code Issues Projects Releases Wiki Activity

[TTI] NFC: Change getScalarizationOverhead and getOperandsScalarizationOverhead to return InstructionCost.

Browse Source 
This patch migrates the TTI cost interfaces to return an InstructionCost.

See this patch for the introduction of the type: https://reviews.llvm.org/D91174
See this thread for context: http://lists.llvm.org/pipermail/llvm-dev/2020-November/146408.html

Reviewed By: sdesmalen

Differential Revision: https://reviews.llvm.org/D101283
This commit is contained in:
dfukalov 2021-04-22 12:41:01 +03:00 committed by Daniil Fukalov
parent 1892e228c0
commit 584872cffa
10 changed files with 59 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
include/llvm/Analysis/TargetTransformInfo.h
View File

@ -728,13 +728,14 @@ public:
/// Estimate the overhead of scalarizing an instruction. Insert and Extract
/// are set if the demanded result elements need to be inserted and/or
/// extracted from vectors.
unsigned getScalarizationOverhead(VectorType *Ty, const APInt &DemandedElts,
InstructionCost getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) const;
/// Estimate the overhead of scalarizing an instructions unique
/// non-constant operands. The (potentially vector) types to use for each of
/// argument are passes via Tys.
unsigned getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
InstructionCost getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) const;
/// If target has efficient vector element load/store instructions, it can
@ -1499,10 +1500,11 @@ public:
virtual bool shouldBuildLookupTablesForConstant(Constant *C) = 0;
virtual bool shouldBuildRelLookupTables() = 0;
virtual bool useColdCCForColdCall(Function &F) = 0;
virtual unsigned getScalarizationOverhead(VectorType *Ty,
virtual InstructionCost getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) = 0;
virtual unsigned
bool Insert,
bool Extract) = 0;
virtual InstructionCost
getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) = 0;
virtual bool supportsEfficientVectorElementLoadStore() = 0;
@ -1900,11 +1902,13 @@ public:
return Impl.useColdCCForColdCall(F);
}
unsigned getScalarizationOverhead(VectorType *Ty, const APInt &DemandedElts,
InstructionCost getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) override {
return Impl.getScalarizationOverhead(Ty, DemandedElts, Insert, Extract);
}
unsigned getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
InstructionCost
getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) override {
return Impl.getOperandsScalarizationOverhead(Args, Tys);
}

5
include/llvm/Analysis/TargetTransformInfoImpl.h
View File

@ -301,12 +301,13 @@ public:
bool useColdCCForColdCall(Function &F) const { return false; }
unsigned getScalarizationOverhead(VectorType *Ty, const APInt &DemandedElts,
InstructionCost getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) const {
return 0;
}
unsigned getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
InstructionCost getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) const {
return 0;
}

23
include/llvm/CodeGen/BasicTTIImpl.h
View File

@ -213,8 +213,8 @@ private:
getMemoryOpCost(Opcode, VT->getElementType(), Alignment, 0, CostKind));
// Next, compute the cost of packing the result in a vector.
int PackingCost = getScalarizationOverhead(VT, Opcode != Instruction::Store,
Opcode == Instruction::Store);
InstructionCost PackingCost = getScalarizationOverhead(
VT, Opcode != Instruction::Store, Opcode == Instruction::Store);
InstructionCost ConditionalCost = 0;
if (VariableMask) {
@ -650,7 +650,8 @@ public:
/// Estimate the overhead of scalarizing an instruction. Insert and Extract
/// are set if the demanded result elements need to be inserted and/or
/// extracted from vectors.
unsigned getScalarizationOverhead(VectorType *InTy, const APInt &DemandedElts,
InstructionCost getScalarizationOverhead(VectorType *InTy,
const APInt &DemandedElts,
bool Insert, bool Extract) {
/// FIXME: a bitfield is not a reasonable abstraction for talking about
/// which elements are needed from a scalable vector
@ -670,11 +671,11 @@ public:
Cost += thisT()->getVectorInstrCost(Instruction::ExtractElement, Ty, i);
}
return *Cost.getValue();
return Cost;
}
/// Helper wrapper for the DemandedElts variant of getScalarizationOverhead.
unsigned getScalarizationOverhead(VectorType *InTy, bool Insert,
InstructionCost getScalarizationOverhead(VectorType *InTy, bool Insert,
bool Extract) {
auto *Ty = cast<FixedVectorType>(InTy);
@ -685,11 +686,11 @@ public:
/// Estimate the overhead of scalarizing an instructions unique
/// non-constant operands. The (potentially vector) types to use for each of
/// argument are passes via Tys.
unsigned getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
InstructionCost getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) {
assert(Args.size() == Tys.size() && "Expected matching Args and Tys");
unsigned Cost = 0;
InstructionCost Cost = 0;
SmallPtrSet<const Value*, 4> UniqueOperands;
for (int I = 0, E = Args.size(); I != E; I++) {
// Disregard things like metadata arguments.
@ -712,12 +713,10 @@ public:
/// instruction, with return type RetTy and arguments Args of type Tys. If
/// Args are unknown (empty), then the cost associated with one argument is
/// added as a heuristic.
unsigned getScalarizationOverhead(VectorType *RetTy,
InstructionCost getScalarizationOverhead(VectorType *RetTy,
ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) {
unsigned Cost = 0;
Cost += getScalarizationOverhead(RetTy, true, false);
InstructionCost Cost = getScalarizationOverhead(RetTy, true, false);
if (!Args.empty())
Cost += getOperandsScalarizationOverhead(Args, Tys);
else
@ -756,7 +755,7 @@ public:
bool IsFloat = Ty->isFPOrFPVectorTy();
// Assume that floating point arithmetic operations cost twice as much as
// integer operations.
unsigned OpCost = (IsFloat ? 2 : 1);
InstructionCost OpCost = (IsFloat ? 2 : 1);
if (TLI->isOperationLegalOrPromote(ISD, LT.second)) {
// The operation is legal. Assume it costs 1.

4
lib/Analysis/TargetTransformInfo.cpp
View File

@ -470,14 +470,14 @@ bool TargetTransformInfo::useColdCCForColdCall(Function &F) const {
return TTIImpl->useColdCCForColdCall(F);
}
unsigned
InstructionCost
TargetTransformInfo::getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) const {
return TTIImpl->getScalarizationOverhead(Ty, DemandedElts, Insert, Extract);
}
unsigned TargetTransformInfo::getOperandsScalarizationOverhead(
InstructionCost TargetTransformInfo::getOperandsScalarizationOverhead(
ArrayRef<const Value *> Args, ArrayRef<Type *> Tys) const {
return TTIImpl->getOperandsScalarizationOverhead(Args, Tys);
}

4
lib/Target/ARM/ARMTargetTransformInfo.cpp
View File

@ -1514,8 +1514,8 @@ InstructionCost ARMTTIImpl::getGatherScatterOpCost(
NumElems * LT.first * ST->getMVEVectorCostFactor(CostKind);
// The scalarization cost should be a lot higher. We use the number of vector
// elements plus the scalarization overhead.
unsigned ScalarCost = NumElems * LT.first +
BaseT::getScalarizationOverhead(VTy, true, false) +
InstructionCost ScalarCost =
NumElems * LT.first + BaseT::getScalarizationOverhead(VTy, true, false) +
BaseT::getScalarizationOverhead(VTy, false, true);
if (EltSize < 8 || Alignment < EltSize / 8)

7
lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
View File

@ -122,13 +122,12 @@ ElementCount HexagonTTIImpl::getMinimumVF(unsigned ElemWidth,
return ElementCount::getFixed((8 * ST.getVectorLength()) / ElemWidth);
}
unsigned HexagonTTIImpl::getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) {
InstructionCost HexagonTTIImpl::getScalarizationOverhead(
VectorType *Ty, const APInt &DemandedElts, bool Insert, bool Extract) {
return BaseT::getScalarizationOverhead(Ty, DemandedElts, Insert, Extract);
}
unsigned
InstructionCost
HexagonTTIImpl::getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys) {
return BaseT::getOperandsScalarizationOverhead(Args, Tys);

5
lib/Target/Hexagon/HexagonTargetTransformInfo.h
View File

@ -104,9 +104,10 @@ public:
return true;
}
unsigned getScalarizationOverhead(VectorType *Ty, const APInt &DemandedElts,
InstructionCost getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract);
unsigned getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
InstructionCost getOperandsScalarizationOverhead(ArrayRef<const Value *> Args,
ArrayRef<Type *> Tys);
InstructionCost getCallInstrCost(Function *F, Type *RetTy,
ArrayRef<Type *> Tys,

2
lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
View File

@ -541,7 +541,7 @@ InstructionCost SystemZTTIImpl::getArithmeticInstrCost(
// There is no native support for FRem.
if (Opcode == Instruction::FRem) {
SmallVector<Type *> Tys(Args.size(), Ty);
unsigned Cost =
InstructionCost Cost =
(VF * LIBCALL_COST) + getScalarizationOverhead(VTy, Args, Tys);
// FIXME: VF 2 for float is currently just as expensive as for VF 4.
if (VF == 2 && ScalarBits == 32)

7
lib/Target/X86/X86TargetTransformInfo.cpp
View File

@ -3121,10 +3121,11 @@ InstructionCost X86TTIImpl::getVectorInstrCost(unsigned Opcode, Type *Val,
return BaseT::getVectorInstrCost(Opcode, Val, Index) + RegisterFileMoveCost;
}
unsigned X86TTIImpl::getScalarizationOverhead(VectorType *Ty,
InstructionCost X86TTIImpl::getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract) {
unsigned Cost = 0;
bool Insert,
bool Extract) {
InstructionCost Cost = 0;
// For insertions, a ISD::BUILD_VECTOR style vector initialization can be much
// cheaper than an accumulation of ISD::INSERT_VECTOR_ELT.

3
lib/Target/X86/X86TargetTransformInfo.h
View File

@ -140,7 +140,8 @@ public:
const Instruction *I = nullptr);
InstructionCost getVectorInstrCost(unsigned Opcode, Type *Val,
unsigned Index);
unsigned getScalarizationOverhead(VectorType *Ty, const APInt &DemandedElts,
InstructionCost getScalarizationOverhead(VectorType *Ty,
const APInt &DemandedElts,
bool Insert, bool Extract);
InstructionCost getMemoryOpCost(unsigned Opcode, Type *Src,
MaybeAlign Alignment, unsigned AddressSpace,