Skip to content

[AArch64] Add cost model for @experimental.vector.match #118512

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Dec 11, 2024
Merged

[AArch64] Add cost model for @experimental.vector.match #118512

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
955435e
[AArch64] Add cost model for @experimental.vector.match
rj-jesus Dec 3, 2024
b0c35de
Implement cost calculation for generic lowering
rj-jesus Dec 5, 2024
15df9dd
Move base cost calculation to getTypeBasedIntrinsicInstrCost
rj-jesus Dec 10, 2024
3353d5e
Add shortcut to call onto getTypeBasedIntrinsicInstrCost directly
rj-jesus Dec 10, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 31 additions & 0 deletions llvm/include/llvm/CodeGen/BasicTTIImpl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1935,6 +1935,8 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {

return Cost;
}
case Intrinsic::experimental_vector_match:
return thisT()->getTypeBasedIntrinsicInstrCost(ICA, CostKind);
}

// Assume that we need to scalarize this intrinsic.)
Expand Down Expand Up @@ -2190,6 +2192,35 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
case Intrinsic::vector_reduce_fminimum:
return thisT()->getMinMaxReductionCost(getMinMaxReductionIntrinsicOp(IID),
VecOpTy, ICA.getFlags(), CostKind);
case Intrinsic::experimental_vector_match: {
auto *SearchTy = cast<VectorType>(ICA.getArgTypes()[0]);
auto *NeedleTy = cast<FixedVectorType>(ICA.getArgTypes()[1]);
unsigned SearchSize = NeedleTy->getNumElements();

// If we're not expanding the intrinsic then we assume this is cheap to
// implement.
EVT SearchVT = getTLI()->getValueType(DL, SearchTy);
if (!getTLI()->shouldExpandVectorMatch(SearchVT, SearchSize))
return getTypeLegalizationCost(RetTy).first;

// Approximate the cost based on the expansion code in
// SelectionDAGBuilder.
InstructionCost Cost = 0;
Cost += thisT()->getVectorInstrCost(Instruction::ExtractElement, NeedleTy,
CostKind, 1, nullptr, nullptr);
Cost += thisT()->getVectorInstrCost(Instruction::InsertElement, SearchTy,
CostKind, 0, nullptr, nullptr);
Cost += thisT()->getShuffleCost(TTI::SK_Broadcast, SearchTy, std::nullopt,
CostKind, 0, nullptr);
Cost += thisT()->getCmpSelInstrCost(BinaryOperator::ICmp, SearchTy, RetTy,
CmpInst::ICMP_EQ, CostKind);
Cost +=
thisT()->getArithmeticInstrCost(BinaryOperator::Or, RetTy, CostKind);
Cost *= SearchSize;
Cost +=
thisT()->getArithmeticInstrCost(BinaryOperator::And, RetTy, CostKind);
return Cost;
}
case Intrinsic::abs:
ISD = ISD::ABS;
break;
Expand Down
17 changes: 17 additions & 0 deletions llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -905,6 +905,23 @@ AArch64TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
}
break;
}
case Intrinsic::experimental_vector_match: {
auto *NeedleTy = cast<FixedVectorType>(ICA.getArgTypes()[1]);
EVT SearchVT = getTLI()->getValueType(DL, ICA.getArgTypes()[0]);
unsigned SearchSize = NeedleTy->getNumElements();
if (!getTLI()->shouldExpandVectorMatch(SearchVT, SearchSize)) {
// Base cost for MATCH instructions. At least on the Neoverse V2 and
// Neoverse V3, these are cheap operations with the same latency as a
// vector ADD. In most cases, however, we also need to do an extra DUP.
// For fixed-length vectors we currently need an extra five--six
// instructions besides the MATCH.
InstructionCost Cost = 4;
if (isa<FixedVectorType>(RetTy))
Cost += 10;
return Cost;
}
break;
}
default:
break;
}
Expand Down
48 changes: 48 additions & 0 deletions llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1360,6 +1360,54 @@ define void @histogram_nxv4i64(<vscale x 4 x ptr> %buckets, <vscale x 4 x i1> %m
ret void
}

define void @match() #3 {
; CHECK-VSCALE-1-LABEL: 'match'
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %match_nxv16i8_v16i8 = call <vscale x 16 x i1> @llvm.experimental.vector.match.nxv16i8.v16i8(<vscale x 16 x i8> undef, <16 x i8> undef, <vscale x 16 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %match_nxv8i16_v8i16 = call <vscale x 8 x i1> @llvm.experimental.vector.match.nxv8i16.v8i16(<vscale x 8 x i16> undef, <8 x i16> undef, <vscale x 8 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %match_nxv4i32_v4i32 = call <vscale x 4 x i1> @llvm.experimental.vector.match.nxv4i32.v4i32(<vscale x 4 x i32> undef, <4 x i32> undef, <vscale x 4 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %match_nxv2i64_v2i64 = call <vscale x 2 x i1> @llvm.experimental.vector.match.nxv2i64.v2i64(<vscale x 2 x i64> undef, <2 x i64> undef, <vscale x 2 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %match_v16i8_v16i8 = call <16 x i1> @llvm.experimental.vector.match.v16i8.v16i8(<16 x i8> undef, <16 x i8> undef, <16 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %match_v8i16_v8i16 = call <8 x i1> @llvm.experimental.vector.match.v8i16.v8i16(<8 x i16> undef, <8 x i16> undef, <8 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %match_v4i32_v4i32 = call <4 x i1> @llvm.experimental.vector.match.v4i32.v4i32(<4 x i32> undef, <4 x i32> undef, <4 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %match_v2i64_v2i64 = call <2 x i1> @llvm.experimental.vector.match.v2i64.v2i64(<2 x i64> undef, <2 x i64> undef, <2 x i1> undef)
; CHECK-VSCALE-1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-VSCALE-2-LABEL: 'match'
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %match_nxv16i8_v16i8 = call <vscale x 16 x i1> @llvm.experimental.vector.match.nxv16i8.v16i8(<vscale x 16 x i8> undef, <16 x i8> undef, <vscale x 16 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %match_nxv8i16_v8i16 = call <vscale x 8 x i1> @llvm.experimental.vector.match.nxv8i16.v8i16(<vscale x 8 x i16> undef, <8 x i16> undef, <vscale x 8 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %match_nxv4i32_v4i32 = call <vscale x 4 x i1> @llvm.experimental.vector.match.nxv4i32.v4i32(<vscale x 4 x i32> undef, <4 x i32> undef, <vscale x 4 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %match_nxv2i64_v2i64 = call <vscale x 2 x i1> @llvm.experimental.vector.match.nxv2i64.v2i64(<vscale x 2 x i64> undef, <2 x i64> undef, <vscale x 2 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %match_v16i8_v16i8 = call <16 x i1> @llvm.experimental.vector.match.v16i8.v16i8(<16 x i8> undef, <16 x i8> undef, <16 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %match_v8i16_v8i16 = call <8 x i1> @llvm.experimental.vector.match.v8i16.v8i16(<8 x i16> undef, <8 x i16> undef, <8 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %match_v4i32_v4i32 = call <4 x i1> @llvm.experimental.vector.match.v4i32.v4i32(<4 x i32> undef, <4 x i32> undef, <4 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %match_v2i64_v2i64 = call <2 x i1> @llvm.experimental.vector.match.v2i64.v2i64(<2 x i64> undef, <2 x i64> undef, <2 x i1> undef)
; CHECK-VSCALE-2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; TYPE_BASED_ONLY-LABEL: 'match'
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %match_nxv16i8_v16i8 = call <vscale x 16 x i1> @llvm.experimental.vector.match.nxv16i8.v16i8(<vscale x 16 x i8> undef, <16 x i8> undef, <vscale x 16 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %match_nxv8i16_v8i16 = call <vscale x 8 x i1> @llvm.experimental.vector.match.nxv8i16.v8i16(<vscale x 8 x i16> undef, <8 x i16> undef, <vscale x 8 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %match_nxv4i32_v4i32 = call <vscale x 4 x i1> @llvm.experimental.vector.match.nxv4i32.v4i32(<vscale x 4 x i32> undef, <4 x i32> undef, <vscale x 4 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %match_nxv2i64_v2i64 = call <vscale x 2 x i1> @llvm.experimental.vector.match.nxv2i64.v2i64(<vscale x 2 x i64> undef, <2 x i64> undef, <vscale x 2 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %match_v16i8_v16i8 = call <16 x i1> @llvm.experimental.vector.match.v16i8.v16i8(<16 x i8> undef, <16 x i8> undef, <16 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %match_v8i16_v8i16 = call <8 x i1> @llvm.experimental.vector.match.v8i16.v8i16(<8 x i16> undef, <8 x i16> undef, <8 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %match_v4i32_v4i32 = call <4 x i1> @llvm.experimental.vector.match.v4i32.v4i32(<4 x i32> undef, <4 x i32> undef, <4 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %match_v2i64_v2i64 = call <2 x i1> @llvm.experimental.vector.match.v2i64.v2i64(<2 x i64> undef, <2 x i64> undef, <2 x i1> undef)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;

%match_nxv16i8_v16i8 = call <vscale x 16 x i1> @llvm.experimental.vector.match.nxv16i8.v16i8(<vscale x 16 x i8> undef, <16 x i8> undef, <vscale x 16 x i1> undef)
%match_nxv8i16_v8i16 = call <vscale x 8 x i1> @llvm.experimental.vector.match.nxv8i16.v8i16(<vscale x 8 x i16> undef, <8 x i16> undef, <vscale x 8 x i1> undef)
%match_nxv4i32_v4i32 = call <vscale x 4 x i1> @llvm.experimental.vector.match.nxv4i32.v4i32(<vscale x 4 x i32> undef, <4 x i32> undef, <vscale x 4 x i1> undef)
%match_nxv2i64_v2i64 = call <vscale x 2 x i1> @llvm.experimental.vector.match.nxv2i64.v2i64(<vscale x 2 x i64> undef, <2 x i64> undef, <vscale x 2 x i1> undef)

%match_v16i8_v16i8 = call <16 x i1> @llvm.experimental.vector.match.v16i8.v16i8(<16 x i8> undef, <16 x i8> undef, <16 x i1> undef)
%match_v8i16_v8i16 = call <8 x i1> @llvm.experimental.vector.match.v8i16.v8i16(<8 x i16> undef, <8 x i16> undef, <8 x i1> undef)
%match_v4i32_v4i32 = call <4 x i1> @llvm.experimental.vector.match.v4i32.v4i32(<4 x i32> undef, <4 x i32> undef, <4 x i1> undef)
%match_v2i64_v2i64 = call <2 x i1> @llvm.experimental.vector.match.v2i64.v2i64(<2 x i64> undef, <2 x i64> undef, <2 x i1> undef)

ret void
}

declare <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64, i64)
declare <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64, i64)
declare <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64, i64)
Expand Down
Loading