[RISCV][TTI] Add vp.cmp intrinsic cost with functionalOPC. #107504
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This patch make the instruction cost of type-based cmp VP intrinsics will be same as their non-VP counterpart.
llvmbot
added
backend:RISC-V
llvm:analysis
ElvisWang123
requested review from
preames,
lukel97 and
arcbbb
and removed request for
lukel97 and
preames
|
@llvm/pr-subscribers-backend-risc-v @llvm/pr-subscribers-llvm-analysis Author: Elvis Wang (ElvisWang123) ChangesThis patch make the instruction cost of Patch is 83.91 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/107504.diff 2 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index e655200e7a8959..012dcff021291f 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -1015,6 +1015,23 @@ RISCVTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
return getArithmeticInstrCost(*FOp, ICA.getReturnType(), CostKind);
break;
}
+
+ // vp compare
+ case Intrinsic::vp_icmp:
+ case Intrinsic::vp_fcmp: {
+ Intrinsic::ID IID = ICA.getID();
+ std::optional<unsigned> FOp = VPIntrinsic::getFunctionalOpcodeForVP(IID);
+ CmpInst::Predicate Pred;
+
+ if (IID == Intrinsic::vp_icmp)
+ Pred = CmpInst::ICMP_SLT;
+ else
+ Pred = CmpInst::FCMP_OLT;
+
+ assert(FOp && !ICA.getArgTypes().empty());
+ return getCmpSelInstrCost(*FOp, ICA.getArgTypes()[0], ICA.getReturnType(),
+ Pred, CostKind);
+ }
}
if (ST->hasVInstructions() && RetTy->isVectorTy()) {
diff --git a/llvm/test/Analysis/CostModel/RISCV/cmp.ll b/llvm/test/Analysis/CostModel/RISCV/cmp.ll
new file mode 100644
index 00000000000000..40938e000b64ec
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/RISCV/cmp.ll
@@ -0,0 +1,660 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 5
+; RUN: opt < %s -mtriple=riscv32 -mattr=+v,+f -passes="print<cost-model>" -cost-kind=throughput --type-based-intrinsic-cost=true 2>&1 -disable-output | FileCheck %s --check-prefixes=CHECK,RV32
+; RUN: opt < %s -mtriple=riscv64 -mattr=+v,+f -passes="print<cost-model>" -cost-kind=throughput --type-based-intrinsic-cost=true 2>&1 -disable-output | FileCheck %s --check-prefixes=CHECK,RV64
+
+define void @icmp() {
+; RV32-LABEL: 'icmp'
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %1 = icmp slt <2 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %2 = icmp slt <2 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %3 = icmp slt <2 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %4 = icmp slt <2 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %5 = icmp slt <2 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %6 = call <2 x i1> @llvm.vp.icmp.v2i1(<2 x i1> undef, <2 x i1> undef, metadata !"slt", <2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %7 = call <2 x i1> @llvm.vp.icmp.v2i8(<2 x i8> undef, <2 x i8> undef, metadata !"slt", <2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %8 = call <2 x i1> @llvm.vp.icmp.v2i16(<2 x i16> undef, <2 x i16> undef, metadata !"slt", <2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %9 = call <2 x i1> @llvm.vp.icmp.v2i32(<2 x i32> undef, <2 x i32> undef, metadata !"slt", <2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %10 = call <2 x i1> @llvm.vp.icmp.v2i64(<2 x i64> undef, <2 x i64> undef, metadata !"slt", <2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %11 = icmp slt <4 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %12 = icmp slt <4 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %13 = icmp slt <4 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %14 = icmp slt <4 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %15 = icmp slt <4 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %16 = call <4 x i1> @llvm.vp.icmp.v4i1(<4 x i1> undef, <4 x i1> undef, metadata !"slt", <4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %17 = call <4 x i1> @llvm.vp.icmp.v4i8(<4 x i8> undef, <4 x i8> undef, metadata !"slt", <4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %18 = call <4 x i1> @llvm.vp.icmp.v4i16(<4 x i16> undef, <4 x i16> undef, metadata !"slt", <4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %19 = call <4 x i1> @llvm.vp.icmp.v4i32(<4 x i32> undef, <4 x i32> undef, metadata !"slt", <4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %20 = call <4 x i1> @llvm.vp.icmp.v4i64(<4 x i64> undef, <4 x i64> undef, metadata !"slt", <4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = icmp slt <8 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = icmp slt <8 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = icmp slt <8 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %24 = icmp slt <8 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %25 = icmp slt <8 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <8 x i1> @llvm.vp.icmp.v8i1(<8 x i1> undef, <8 x i1> undef, metadata !"slt", <8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <8 x i1> @llvm.vp.icmp.v8i8(<8 x i8> undef, <8 x i8> undef, metadata !"slt", <8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %28 = call <8 x i1> @llvm.vp.icmp.v8i16(<8 x i16> undef, <8 x i16> undef, metadata !"slt", <8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %29 = call <8 x i1> @llvm.vp.icmp.v8i32(<8 x i32> undef, <8 x i32> undef, metadata !"slt", <8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %30 = call <8 x i1> @llvm.vp.icmp.v8i64(<8 x i64> undef, <8 x i64> undef, metadata !"slt", <8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %31 = icmp slt <16 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %32 = icmp slt <16 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %33 = icmp slt <16 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %34 = icmp slt <16 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %35 = icmp slt <16 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %36 = call <16 x i1> @llvm.vp.icmp.v16i1(<16 x i1> undef, <16 x i1> undef, metadata !"slt", <16 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %37 = call <16 x i1> @llvm.vp.icmp.v16i8(<16 x i8> undef, <16 x i8> undef, metadata !"slt", <16 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %38 = call <16 x i1> @llvm.vp.icmp.v16i16(<16 x i16> undef, <16 x i16> undef, metadata !"slt", <16 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %39 = call <16 x i1> @llvm.vp.icmp.v16i32(<16 x i32> undef, <16 x i32> undef, metadata !"slt", <16 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %40 = call <16 x i1> @llvm.vp.icmp.v16i64(<16 x i64> undef, <16 x i64> undef, metadata !"slt", <16 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %41 = icmp slt <32 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %42 = icmp slt <32 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %43 = icmp slt <32 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %44 = icmp slt <32 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %45 = icmp slt <32 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %46 = call <32 x i1> @llvm.vp.icmp.v32i1(<32 x i1> undef, <32 x i1> undef, metadata !"slt", <32 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %47 = call <32 x i1> @llvm.vp.icmp.v32i8(<32 x i8> undef, <32 x i8> undef, metadata !"slt", <32 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %48 = call <32 x i1> @llvm.vp.icmp.v32i16(<32 x i16> undef, <32 x i16> undef, metadata !"slt", <32 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %49 = call <32 x i1> @llvm.vp.icmp.v32i32(<32 x i32> undef, <32 x i32> undef, metadata !"slt", <32 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %50 = call <32 x i1> @llvm.vp.icmp.v32i64(<32 x i64> undef, <32 x i64> undef, metadata !"slt", <32 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %51 = icmp slt <64 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %52 = icmp slt <64 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %53 = icmp slt <64 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %54 = icmp slt <64 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %55 = icmp slt <64 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %56 = call <64 x i1> @llvm.vp.icmp.v64i1(<64 x i1> undef, <64 x i1> undef, metadata !"slt", <64 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %57 = call <64 x i1> @llvm.vp.icmp.v64i8(<64 x i8> undef, <64 x i8> undef, metadata !"slt", <64 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %58 = call <64 x i1> @llvm.vp.icmp.v64i16(<64 x i16> undef, <64 x i16> undef, metadata !"slt", <64 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %59 = call <64 x i1> @llvm.vp.icmp.v64i32(<64 x i32> undef, <64 x i32> undef, metadata !"slt", <64 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %60 = call <64 x i1> @llvm.vp.icmp.v64i64(<64 x i64> undef, <64 x i64> undef, metadata !"slt", <64 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %61 = icmp slt <128 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %62 = icmp slt <128 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %63 = icmp slt <128 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %64 = icmp slt <128 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %65 = icmp slt <128 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %66 = call <128 x i1> @llvm.vp.icmp.v128i1(<128 x i1> undef, <128 x i1> undef, metadata !"slt", <128 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %67 = call <128 x i1> @llvm.vp.icmp.v128i8(<128 x i8> undef, <128 x i8> undef, metadata !"slt", <128 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %68 = call <128 x i1> @llvm.vp.icmp.v128i16(<128 x i16> undef, <128 x i16> undef, metadata !"slt", <128 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %69 = call <128 x i1> @llvm.vp.icmp.v128i32(<128 x i32> undef, <128 x i32> undef, metadata !"slt", <128 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %70 = call <128 x i1> @llvm.vp.icmp.v128i64(<128 x i64> undef, <128 x i64> undef, metadata !"slt", <128 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %71 = icmp slt <256 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %72 = icmp slt <256 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %73 = icmp slt <256 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %74 = icmp slt <256 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %75 = icmp slt <256 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %76 = call <256 x i1> @llvm.vp.icmp.v256i1(<256 x i1> undef, <256 x i1> undef, metadata !"slt", <256 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %77 = call <256 x i1> @llvm.vp.icmp.v256i8(<256 x i8> undef, <256 x i8> undef, metadata !"slt", <256 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %78 = call <256 x i1> @llvm.vp.icmp.v256i16(<256 x i16> undef, <256 x i16> undef, metadata !"slt", <256 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %79 = call <256 x i1> @llvm.vp.icmp.v256i32(<256 x i32> undef, <256 x i32> undef, metadata !"slt", <256 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %80 = call <256 x i1> @llvm.vp.icmp.v256i64(<256 x i64> undef, <256 x i64> undef, metadata !"slt", <256 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %81 = icmp slt <vscale x 1 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %82 = icmp slt <vscale x 1 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %83 = icmp slt <vscale x 1 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %84 = icmp slt <vscale x 1 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %85 = icmp slt <vscale x 1 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %86 = call <vscale x 1 x i1> @llvm.vp.icmp.nxv1i1(<vscale x 1 x i1> undef, <vscale x 1 x i1> undef, metadata !"slt", <vscale x 1 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %87 = call <vscale x 1 x i1> @llvm.vp.icmp.nxv1i8(<vscale x 1 x i8> undef, <vscale x 1 x i8> undef, metadata !"slt", <vscale x 1 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %88 = call <vscale x 1 x i1> @llvm.vp.icmp.nxv1i16(<vscale x 1 x i16> undef, <vscale x 1 x i16> undef, metadata !"slt", <vscale x 1 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %89 = call <vscale x 1 x i1> @llvm.vp.icmp.nxv1i32(<vscale x 1 x i32> undef, <vscale x 1 x i32> undef, metadata !"slt", <vscale x 1 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %90 = call <vscale x 1 x i1> @llvm.vp.icmp.nxv1i64(<vscale x 1 x i64> undef, <vscale x 1 x i64> undef, metadata !"slt", <vscale x 1 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %91 = icmp slt <vscale x 2 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %92 = icmp slt <vscale x 2 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %93 = icmp slt <vscale x 2 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %94 = icmp slt <vscale x 2 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %95 = icmp slt <vscale x 2 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %96 = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i1(<vscale x 2 x i1> undef, <vscale x 2 x i1> undef, metadata !"slt", <vscale x 2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %97 = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i8(<vscale x 2 x i8> undef, <vscale x 2 x i8> undef, metadata !"slt", <vscale x 2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %98 = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i16(<vscale x 2 x i16> undef, <vscale x 2 x i16> undef, metadata !"slt", <vscale x 2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %99 = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> undef, metadata !"slt", <vscale x 2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %100 = call <vscale x 2 x i1> @llvm.vp.icmp.nxv2i64(<vscale x 2 x i64> undef, <vscale x 2 x i64> undef, metadata !"slt", <vscale x 2 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %101 = icmp slt <vscale x 4 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %102 = icmp slt <vscale x 4 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %103 = icmp slt <vscale x 4 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %104 = icmp slt <vscale x 4 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %105 = icmp slt <vscale x 4 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %106 = call <vscale x 4 x i1> @llvm.vp.icmp.nxv4i1(<vscale x 4 x i1> undef, <vscale x 4 x i1> undef, metadata !"slt", <vscale x 4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %107 = call <vscale x 4 x i1> @llvm.vp.icmp.nxv4i8(<vscale x 4 x i8> undef, <vscale x 4 x i8> undef, metadata !"slt", <vscale x 4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %108 = call <vscale x 4 x i1> @llvm.vp.icmp.nxv4i16(<vscale x 4 x i16> undef, <vscale x 4 x i16> undef, metadata !"slt", <vscale x 4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %109 = call <vscale x 4 x i1> @llvm.vp.icmp.nxv4i32(<vscale x 4 x i32> undef, <vscale x 4 x i32> undef, metadata !"slt", <vscale x 4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %110 = call <vscale x 4 x i1> @llvm.vp.icmp.nxv4i64(<vscale x 4 x i64> undef, <vscale x 4 x i64> undef, metadata !"slt", <vscale x 4 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %111 = icmp slt <vscale x 8 x i1> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %112 = icmp slt <vscale x 8 x i8> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %113 = icmp slt <vscale x 8 x i16> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %114 = icmp slt <vscale x 8 x i32> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %115 = icmp slt <vscale x 8 x i64> undef, undef
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %116 = call <vscale x 8 x i1> @llvm.vp.icmp.nxv8i1(<vscale x 8 x i1> undef, <vscale x 8 x i1> undef, metadata !"slt", <vscale x 8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %117 = call <vscale x 8 x i1> @llvm.vp.icmp.nxv8i8(<vscale x 8 x i8> undef, <vscale x 8 x i8> undef, metadata !"slt", <vscale x 8 x i1> undef, i32 undef)
+; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %118 = call <vscale x 8 x i1> @llvm.vp.icmp.nxv8i16(<vscale x 8 x ...
[truncated]
|
preames
reviewed
Sep 6, 2024
| CmpInst::Predicate Pred; | ||
|
|
||
| if (IID == Intrinsic::vp_icmp) | ||
| Pred = CmpInst::ICMP_SLT; |
There was a problem hiding this comment.
You need to get the code from the intrinsic as the cost of the comparison does depend on the predicate.
If you have a type based query (i.e. no args), then you probably shouldn't return a cost here at all.
|
@preames Gentle ping! |
preames
approved these changes
Sep 11, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch make the instruction cost of
cmpVP intrinsics as same as their non-VP counterpart.