[llvm] [CostModel/RISCV] Fix costs of [l]lrint (PR #146058)
Ramkumar Ramachandra via llvm-commits
llvm-commits at lists.llvm.org
Fri Jun 27 04:12:24 PDT 2025
https://github.com/artagnon created https://github.com/llvm/llvm-project/pull/146058
Take the actual instruction cost into account, and don't fallthrough to code that doesn't apply to [l]lrint.
>From 3592a97915ebdf9c6c911ec511be05714760a9dd Mon Sep 17 00:00:00 2001
From: Ramkumar Ramachandra <ramkumar.ramachandra at codasip.com>
Date: Fri, 27 Jun 2025 12:07:43 +0100
Subject: [PATCH] [CostModel/RISCV] Fix costs of [l]lrint
Take the actual instruction cost into account, and don't fallthrough to
code that doesn't apply to [l]lrint.
---
.../Target/RISCV/RISCVTargetTransformInfo.cpp | 12 +++--
llvm/test/Analysis/CostModel/RISCV/fround.ll | 52 +++++++++----------
2 files changed, 33 insertions(+), 31 deletions(-)
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 67a51c12b508e..56773372da905 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -1191,9 +1191,6 @@ static const CostTblEntry VectorIntrinsicCostTable[]{
{Intrinsic::roundeven, MVT::f64, 9},
{Intrinsic::rint, MVT::f32, 7},
{Intrinsic::rint, MVT::f64, 7},
- {Intrinsic::lrint, MVT::i32, 1},
- {Intrinsic::lrint, MVT::i64, 1},
- {Intrinsic::llrint, MVT::i64, 1},
{Intrinsic::nearbyint, MVT::f32, 9},
{Intrinsic::nearbyint, MVT::f64, 9},
{Intrinsic::bswap, MVT::i16, 3},
@@ -1261,12 +1258,17 @@ RISCVTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
auto *RetTy = ICA.getReturnType();
switch (ICA.getID()) {
case Intrinsic::lrint:
- case Intrinsic::llrint:
+ case Intrinsic::llrint: {
// We can't currently lower half or bfloat vector lrint/llrint.
if (auto *VecTy = dyn_cast<VectorType>(ICA.getArgTypes()[0]);
VecTy && VecTy->getElementType()->is16bitFPTy())
return InstructionCost::getInvalid();
- [[fallthrough]];
+ auto LT = getTypeLegalizationCost(RetTy);
+ if (ST->hasVInstructions() && LT.second.isVector())
+ return LT.first *
+ getRISCVInstructionCost(RISCV::VFCVT_X_F_V, LT.second, CostKind);
+ break;
+ }
case Intrinsic::ceil:
case Intrinsic::floor:
case Intrinsic::trunc:
diff --git a/llvm/test/Analysis/CostModel/RISCV/fround.ll b/llvm/test/Analysis/CostModel/RISCV/fround.ll
index a0818d487d151..57b40201662e5 100644
--- a/llvm/test/Analysis/CostModel/RISCV/fround.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/fround.ll
@@ -436,23 +436,23 @@ define void @lrint() {
; CHECK-NEXT: Cost Model: Invalid cost for instruction: %10 = call <vscale x 16 x i64> @llvm.lrint.nxv16i64.nxv16bf16(<vscale x 16 x bfloat> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %11 = call i64 @llvm.lrint.i64.f32(float undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %12 = call <2 x i64> @llvm.lrint.v2i64.v2f32(<2 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %13 = call <4 x i64> @llvm.lrint.v4i64.v4f32(<4 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %14 = call <8 x i64> @llvm.lrint.v8i64.v8f32(<8 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %15 = call <16 x i64> @llvm.lrint.v16i64.v16f32(<16 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %13 = call <4 x i64> @llvm.lrint.v4i64.v4f32(<4 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %14 = call <8 x i64> @llvm.lrint.v8i64.v8f32(<8 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %15 = call <16 x i64> @llvm.lrint.v16i64.v16f32(<16 x float> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %16 = call <vscale x 1 x i64> @llvm.lrint.nxv1i64.nxv1f32(<vscale x 1 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %17 = call <vscale x 2 x i64> @llvm.lrint.nxv2i64.nxv2f32(<vscale x 2 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %18 = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f32(<vscale x 4 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %19 = call <vscale x 8 x i64> @llvm.lrint.nxv8i64.nxv8f32(<vscale x 8 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %20 = call <vscale x 16 x i64> @llvm.lrint.nxv16i64.nxv16f32(<vscale x 16 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %17 = call <vscale x 2 x i64> @llvm.lrint.nxv2i64.nxv2f32(<vscale x 2 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %18 = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f32(<vscale x 4 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %19 = call <vscale x 8 x i64> @llvm.lrint.nxv8i64.nxv8f32(<vscale x 8 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %20 = call <vscale x 16 x i64> @llvm.lrint.nxv16i64.nxv16f32(<vscale x 16 x float> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call i64 @llvm.lrint.i64.f64(double undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <2 x i64> @llvm.lrint.v2i64.v2f64(<2 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <4 x i64> @llvm.lrint.v4i64.v4f64(<4 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %24 = call <8 x i64> @llvm.lrint.v8i64.v8f64(<8 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %25 = call <16 x i64> @llvm.lrint.v16i64.v16f64(<16 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <4 x i64> @llvm.lrint.v4i64.v4f64(<4 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <8 x i64> @llvm.lrint.v8i64.v8f64(<8 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <16 x i64> @llvm.lrint.v16i64.v16f64(<16 x double> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x i64> @llvm.lrint.nxv1i64.nxv1f64(<vscale x 1 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x i64> @llvm.lrint.nxv2i64.nxv2f64(<vscale x 2 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %28 = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f64(<vscale x 4 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %29 = call <vscale x 8 x i64> @llvm.lrint.nxv8i64.nxv8f64(<vscale x 8 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x i64> @llvm.lrint.nxv2i64.nxv2f64(<vscale x 2 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f64(<vscale x 4 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x i64> @llvm.lrint.nxv8i64.nxv8f64(<vscale x 8 x double> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
call i64 @llvm.lrint.i64.bf16(bfloat undef)
@@ -528,23 +528,23 @@ define void @llrint() {
; CHECK-NEXT: Cost Model: Invalid cost for instruction: %10 = call <vscale x 16 x i64> @llvm.llrint.nxv16i64.nxv16bf16(<vscale x 16 x bfloat> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %11 = call i64 @llvm.llrint.i64.f32(float undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %12 = call <2 x i64> @llvm.llrint.v2i64.v2f32(<2 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %13 = call <4 x i64> @llvm.llrint.v4i64.v4f32(<4 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %14 = call <8 x i64> @llvm.llrint.v8i64.v8f32(<8 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %15 = call <16 x i64> @llvm.llrint.v16i64.v16f32(<16 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %13 = call <4 x i64> @llvm.llrint.v4i64.v4f32(<4 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %14 = call <8 x i64> @llvm.llrint.v8i64.v8f32(<8 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %15 = call <16 x i64> @llvm.llrint.v16i64.v16f32(<16 x float> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %16 = call <vscale x 1 x i64> @llvm.llrint.nxv1i64.nxv1f32(<vscale x 1 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %17 = call <vscale x 2 x i64> @llvm.llrint.nxv2i64.nxv2f32(<vscale x 2 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %18 = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f32(<vscale x 4 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %19 = call <vscale x 8 x i64> @llvm.llrint.nxv8i64.nxv8f32(<vscale x 8 x float> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %20 = call <vscale x 16 x i64> @llvm.llrint.nxv16i64.nxv16f32(<vscale x 16 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %17 = call <vscale x 2 x i64> @llvm.llrint.nxv2i64.nxv2f32(<vscale x 2 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %18 = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f32(<vscale x 4 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %19 = call <vscale x 8 x i64> @llvm.llrint.nxv8i64.nxv8f32(<vscale x 8 x float> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %20 = call <vscale x 16 x i64> @llvm.llrint.nxv16i64.nxv16f32(<vscale x 16 x float> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call i64 @llvm.llrint.i64.f64(double undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <2 x i64> @llvm.llrint.v2i64.v2f64(<2 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <4 x i64> @llvm.llrint.v4i64.v4f64(<4 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %24 = call <8 x i64> @llvm.llrint.v8i64.v8f64(<8 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %25 = call <16 x i64> @llvm.llrint.v16i64.v16f64(<16 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <4 x i64> @llvm.llrint.v4i64.v4f64(<4 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <8 x i64> @llvm.llrint.v8i64.v8f64(<8 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <16 x i64> @llvm.llrint.v16i64.v16f64(<16 x double> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x i64> @llvm.llrint.nxv1i64.nxv1f64(<vscale x 1 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x i64> @llvm.llrint.nxv2i64.nxv2f64(<vscale x 2 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %28 = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f64(<vscale x 4 x double> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %29 = call <vscale x 8 x i64> @llvm.llrint.nxv8i64.nxv8f64(<vscale x 8 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x i64> @llvm.llrint.nxv2i64.nxv2f64(<vscale x 2 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f64(<vscale x 4 x double> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x i64> @llvm.llrint.nxv8i64.nxv8f64(<vscale x 8 x double> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
call i64 @llvm.llrint.i64.bf16(bfloat undef)
More information about the llvm-commits
mailing list