From d7c219a506ec9aabe7c5d36c0da55656af487b73 Mon Sep 17 00:00:00 2001 From: Rosie Sumpter Date: Wed, 9 Jun 2021 10:00:16 +0100 Subject: [PATCH] [CostModel][AArch64] Improve the cost estimate of CTPOP intrinsic Added a case for CTPOP to AArch64TTIImpl::getIntrinsicInstrCost so that the cost estimate matches the codegen in test/CodeGen/AArch64/arm64-vpopcnt.ll Differential Revision: https://reviews.llvm.org/D103952 --- .../Target/AArch64/AArch64TargetTransformInfo.cpp | 25 +++++++++ llvm/test/Analysis/CostModel/AArch64/ctpop.ll | 61 +++++----------------- 2 files changed, 38 insertions(+), 48 deletions(-) diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp index 4390951..37e9852 100644 --- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp +++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp @@ -300,6 +300,31 @@ AArch64TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA, return LegalisationCost.first * Entry->Cost; break; } + case Intrinsic::ctpop: { + static const CostTblEntry CtpopCostTbl[] = { + {ISD::CTPOP, MVT::v2i64, 4}, + {ISD::CTPOP, MVT::v4i32, 3}, + {ISD::CTPOP, MVT::v8i16, 2}, + {ISD::CTPOP, MVT::v16i8, 1}, + {ISD::CTPOP, MVT::i64, 4}, + {ISD::CTPOP, MVT::v2i32, 3}, + {ISD::CTPOP, MVT::v4i16, 2}, + {ISD::CTPOP, MVT::v8i8, 1}, + {ISD::CTPOP, MVT::i32, 5}, + }; + auto LT = TLI->getTypeLegalizationCost(DL, RetTy); + MVT MTy = LT.second; + if (const auto *Entry = CostTableLookup(CtpopCostTbl, ISD::CTPOP, MTy)) { + // Extra cost of +1 when illegal vector types are legalized by promoting + // the integer type. + int ExtraCost = MTy.isVector() && MTy.getScalarSizeInBits() != + RetTy->getScalarSizeInBits() + ? 1 + : 0; + return LT.first * Entry->Cost + ExtraCost; + } + break; + } default: break; } diff --git a/llvm/test/Analysis/CostModel/AArch64/ctpop.ll b/llvm/test/Analysis/CostModel/AArch64/ctpop.ll index f276f04..bf6cb04 100644 --- a/llvm/test/Analysis/CostModel/AArch64/ctpop.ll +++ b/llvm/test/Analysis/CostModel/AArch64/ctpop.ll @@ -4,41 +4,33 @@ ; Verify the cost of scalar ctpop instructions. define i64 @test_ctpop_i64(i64 %a) { -; ; CHECK-LABEL: 'test_ctpop_i64' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i64 @llvm.ctpop.i64(i64 %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call i64 @llvm.ctpop.i64(i64 %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i64 %ctpop -; %ctpop = call i64 @llvm.ctpop.i64(i64 %a) ret i64 %ctpop } define i32 @test_ctpop_i32(i32 %a) { -; ; CHECK-LABEL: 'test_ctpop_i32' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i32 @llvm.ctpop.i32(i32 %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %ctpop = call i32 @llvm.ctpop.i32(i32 %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 %ctpop -; %ctpop = call i32 @llvm.ctpop.i32(i32 %a) ret i32 %ctpop } define i16 @test_ctpop_i16(i16 %a) { -; ; CHECK-LABEL: 'test_ctpop_i16' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i16 @llvm.ctpop.i16(i16 %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %ctpop = call i16 @llvm.ctpop.i16(i16 %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i16 %ctpop -; %ctpop = call i16 @llvm.ctpop.i16(i16 %a) ret i16 %ctpop } define i8 @test_ctpop_i8(i8 %a) { -; ; CHECK-LABEL: 'test_ctpop_i8' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i8 @llvm.ctpop.i8(i8 %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %ctpop = call i8 @llvm.ctpop.i8(i8 %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i8 %ctpop -; %ctpop = call i8 @llvm.ctpop.i8(i8 %a) ret i8 %ctpop } @@ -51,140 +43,113 @@ declare i8 @llvm.ctpop.i8(i8) ; Verify the cost of vector ctpop instructions. define <2 x i64> @test_ctpop_v2i64(<2 x i64> %a) { -; ; CHECK-LABEL: 'test_ctpop_v2i64' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %ctpop -; %ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %a) ret <2 x i64> %ctpop } define <2 x i32> @test_ctpop_v2i32(<2 x i32> %a) { -; ; CHECK-LABEL: 'test_ctpop_v2i32' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %ctpop -; %ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %a) ret <2 x i32> %ctpop } define <4 x i32> @test_ctpop_v4i32(<4 x i32> %a) { -; ; CHECK-LABEL: 'test_ctpop_v4i32' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %ctpop -; %ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %a) ret <4 x i32> %ctpop } define <2 x i16> @test_ctpop_v2i16(<2 x i16> %a) { -; ; CHECK-LABEL: 'test_ctpop_v2i16' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i16> @llvm.ctpop.v2i16(<2 x i16> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <2 x i16> @llvm.ctpop.v2i16(<2 x i16> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i16> %ctpop -; %ctpop = call <2 x i16> @llvm.ctpop.v2i16(<2 x i16> %a) ret <2 x i16> %ctpop } define <4 x i16> @test_ctpop_v4i16(<4 x i16> %a) { -; ; CHECK-LABEL: 'test_ctpop_v4i16' ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <4 x i16> @llvm.ctpop.v4i16(<4 x i16> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i16> %ctpop -; %ctpop = call <4 x i16> @llvm.ctpop.v4i16(<4 x i16> %a) ret <4 x i16> %ctpop } define <8 x i16> @test_ctpop_v8i16(<8 x i16> %a) { -; ; CHECK-LABEL: 'test_ctpop_v8i16' ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %ctpop -; %ctpop = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %a) ret <8 x i16> %ctpop } define <2 x i8> @test_ctpop_v2i8(<2 x i8> %a) { -; ; CHECK-LABEL: 'test_ctpop_v2i8' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i8> %ctpop -; %ctpop = call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %a) ret <2 x i8> %ctpop } define <4 x i8> @test_ctpop_v4i8(<4 x i8> %a) { -; ; CHECK-LABEL: 'test_ctpop_v4i8' -; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <4 x i8> @llvm.ctpop.v4i8(<4 x i8> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %ctpop = call <4 x i8> @llvm.ctpop.v4i8(<4 x i8> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i8> %ctpop -; %ctpop = call <4 x i8> @llvm.ctpop.v4i8(<4 x i8> %a) ret <4 x i8> %ctpop } define <8 x i8> @test_ctpop_v8i8(<8 x i8> %a) { -; ; CHECK-LABEL: 'test_ctpop_v8i8' ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %ctpop = call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i8> %ctpop -; %ctpop = call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> %a) ret <8 x i8> %ctpop } define <16 x i8> @test_ctpop_v16i8(<16 x i8> %a) { -; ; CHECK-LABEL: 'test_ctpop_v16i8' ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %ctpop = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %ctpop -; %ctpop = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %a) ret <16 x i8> %ctpop } define <4 x i64> @test_ctpop_v4i64(<4 x i64> %a) { -; ; CHECK-LABEL: 'test_ctpop_v4i64' -; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %ctpop -; %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %a) ret <4 x i64> %ctpop } define <8 x i32> @test_ctpop_v8i32(<8 x i32> %a) { -; ; CHECK-LABEL: 'test_ctpop_v8i32' -; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %ctpop -; %ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %a) ret <8 x i32> %ctpop } define <16 x i16> @test_ctpop_v16i16(<16 x i16> %a) { -; ; CHECK-LABEL: 'test_ctpop_v16i16' ; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %ctpop -; %ctpop = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %a) ret <16 x i16> %ctpop } define <32 x i8> @test_ctpop_v32i8(<32 x i8> %a) { ; CHECK-LABEL: 'test_ctpop_v32i8' -; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %a) +; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %a) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %ctpop -; %ctpop = call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %a) ret <32 x i8> %ctpop } -- 2.7.4