[RISCV] Fix bug in [l](lrint|lround) vector-cost #151298
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
Follow up on a review of bd66fd0 ([CostModel/RISCV] Fix costs of vector [l](lrint|lround)) post-landing to fix a subtle problem with the cost of vector [l](lrint|lround). We should use source LMUL in the case of a narrowing op. Co-authored-by: Luke Lau <[email protected]>
llvmbot
added
backend:RISC-V
llvm:analysis
Member
|
@llvm/pr-subscribers-llvm-analysis Author: Ramkumar Ramachandra (artagnon) ChangesFollow up on a review of bd66fd0 ( Full diff: https://github.com/llvm/llvm-project/pull/151298.diff 2 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 4eef8de06b278..61dbd066d3e22 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -1280,8 +1280,13 @@ RISCVTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
} else {
Ops = {RISCV::VFCVT_X_F_V};
}
- return std::max(SrcLT.first, LT.first) *
- getRISCVInstructionCost(Ops, LT.second, CostKind);
+
+ // We need to use the source LMUL in the case of a narrowing op, and the
+ // destination LMUL otherwise.
+ if (SrcEltSz > DstEltSz)
+ return SrcLT.first *
+ getRISCVInstructionCost(Ops, SrcLT.second, CostKind);
+ return LT.first * getRISCVInstructionCost(Ops, LT.second, CostKind);
}
break;
}
diff --git a/llvm/test/Analysis/CostModel/RISCV/fround.ll b/llvm/test/Analysis/CostModel/RISCV/fround.ll
index 189e57e5f62ee..235728980cd46 100644
--- a/llvm/test/Analysis/CostModel/RISCV/fround.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/fround.ll
@@ -446,13 +446,13 @@ define void @lrint() {
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %20 = call <vscale x 16 x i32> @llvm.lrint.nxv16i32.nxv16f32(<vscale x 16 x float> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call i32 @llvm.lrint.i32.f64(double poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <2 x i32> @llvm.lrint.v2i32.v2f64(<2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <4 x i32> @llvm.lrint.v4i32.v4f64(<4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %24 = call <8 x i32> @llvm.lrint.v8i32.v8f64(<8 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %25 = call <16 x i32> @llvm.lrint.v16i32.v16f64(<16 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <4 x i32> @llvm.lrint.v4i32.v4f64(<4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <8 x i32> @llvm.lrint.v8i32.v8f64(<8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <16 x i32> @llvm.lrint.v16i32.v16f64(<16 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x i32> @llvm.lrint.nxv1i32.nxv1f64(<vscale x 1 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x i32> @llvm.lrint.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %28 = call <vscale x 4 x i32> @llvm.lrint.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %29 = call <vscale x 8 x i32> @llvm.lrint.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x i32> @llvm.lrint.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x i32> @llvm.lrint.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x i32> @llvm.lrint.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %30 = call i32 @llvm.lrint.i32.bf16(bfloat poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %31 = call <2 x i64> @llvm.lrint.v2i64.v2bf16(<2 x bfloat> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %32 = call <4 x i64> @llvm.lrint.v4i64.v4bf16(<4 x bfloat> poison)
@@ -708,13 +708,13 @@ define void @lround() {
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %20 = call <vscale x 16 x i32> @llvm.lround.nxv16i32.nxv16f32(<vscale x 16 x float> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call i32 @llvm.lround.i32.f64(double poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <2 x i32> @llvm.lround.v2i32.v2f64(<2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <4 x i32> @llvm.lround.v4i32.v4f64(<4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %24 = call <8 x i32> @llvm.lround.v8i32.v8f64(<8 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %25 = call <16 x i32> @llvm.lround.v16i32.v16f64(<16 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <4 x i32> @llvm.lround.v4i32.v4f64(<4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <8 x i32> @llvm.lround.v8i32.v8f64(<8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <16 x i32> @llvm.lround.v16i32.v16f64(<16 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x i32> @llvm.lround.nxv1i32.nxv1f64(<vscale x 1 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x i32> @llvm.lround.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %28 = call <vscale x 4 x i32> @llvm.lround.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %29 = call <vscale x 8 x i32> @llvm.lround.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x i32> @llvm.lround.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x i32> @llvm.lround.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x i32> @llvm.lround.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %30 = call <vscale x 16 x i32> @llvm.lround.nxv16i32.nxv16f64(<vscale x 16 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %31 = call i64 @llvm.lround.i64.bf16(bfloat poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %32 = call <2 x i64> @llvm.lround.v2i64.v2bf16(<2 x bfloat> poison)
|
Member
|
@llvm/pr-subscribers-backend-risc-v Author: Ramkumar Ramachandra (artagnon) ChangesFollow up on a review of bd66fd0 ( Full diff: https://github.com/llvm/llvm-project/pull/151298.diff 2 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 4eef8de06b278..61dbd066d3e22 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -1280,8 +1280,13 @@ RISCVTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
} else {
Ops = {RISCV::VFCVT_X_F_V};
}
- return std::max(SrcLT.first, LT.first) *
- getRISCVInstructionCost(Ops, LT.second, CostKind);
+
+ // We need to use the source LMUL in the case of a narrowing op, and the
+ // destination LMUL otherwise.
+ if (SrcEltSz > DstEltSz)
+ return SrcLT.first *
+ getRISCVInstructionCost(Ops, SrcLT.second, CostKind);
+ return LT.first * getRISCVInstructionCost(Ops, LT.second, CostKind);
}
break;
}
diff --git a/llvm/test/Analysis/CostModel/RISCV/fround.ll b/llvm/test/Analysis/CostModel/RISCV/fround.ll
index 189e57e5f62ee..235728980cd46 100644
--- a/llvm/test/Analysis/CostModel/RISCV/fround.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/fround.ll
@@ -446,13 +446,13 @@ define void @lrint() {
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %20 = call <vscale x 16 x i32> @llvm.lrint.nxv16i32.nxv16f32(<vscale x 16 x float> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call i32 @llvm.lrint.i32.f64(double poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <2 x i32> @llvm.lrint.v2i32.v2f64(<2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <4 x i32> @llvm.lrint.v4i32.v4f64(<4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %24 = call <8 x i32> @llvm.lrint.v8i32.v8f64(<8 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %25 = call <16 x i32> @llvm.lrint.v16i32.v16f64(<16 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <4 x i32> @llvm.lrint.v4i32.v4f64(<4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <8 x i32> @llvm.lrint.v8i32.v8f64(<8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <16 x i32> @llvm.lrint.v16i32.v16f64(<16 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x i32> @llvm.lrint.nxv1i32.nxv1f64(<vscale x 1 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x i32> @llvm.lrint.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %28 = call <vscale x 4 x i32> @llvm.lrint.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %29 = call <vscale x 8 x i32> @llvm.lrint.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x i32> @llvm.lrint.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x i32> @llvm.lrint.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x i32> @llvm.lrint.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %30 = call i32 @llvm.lrint.i32.bf16(bfloat poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %31 = call <2 x i64> @llvm.lrint.v2i64.v2bf16(<2 x bfloat> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %32 = call <4 x i64> @llvm.lrint.v4i64.v4bf16(<4 x bfloat> poison)
@@ -708,13 +708,13 @@ define void @lround() {
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %20 = call <vscale x 16 x i32> @llvm.lround.nxv16i32.nxv16f32(<vscale x 16 x float> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %21 = call i32 @llvm.lround.i32.f64(double poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %22 = call <2 x i32> @llvm.lround.v2i32.v2f64(<2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %23 = call <4 x i32> @llvm.lround.v4i32.v4f64(<4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %24 = call <8 x i32> @llvm.lround.v8i32.v8f64(<8 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %25 = call <16 x i32> @llvm.lround.v16i32.v16f64(<16 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %23 = call <4 x i32> @llvm.lround.v4i32.v4f64(<4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %24 = call <8 x i32> @llvm.lround.v8i32.v8f64(<8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %25 = call <16 x i32> @llvm.lround.v16i32.v16f64(<16 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x i32> @llvm.lround.nxv1i32.nxv1f64(<vscale x 1 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %27 = call <vscale x 2 x i32> @llvm.lround.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %28 = call <vscale x 4 x i32> @llvm.lround.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %29 = call <vscale x 8 x i32> @llvm.lround.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x i32> @llvm.lround.nxv2i32.nxv2f64(<vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x i32> @llvm.lround.nxv4i32.nxv4f64(<vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x i32> @llvm.lround.nxv8i32.nxv8f64(<vscale x 8 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %30 = call <vscale x 16 x i32> @llvm.lround.nxv16i32.nxv16f64(<vscale x 16 x double> poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %31 = call i64 @llvm.lround.i64.bf16(bfloat poison)
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %32 = call <2 x i64> @llvm.lround.v2i64.v2bf16(<2 x bfloat> poison)
|
lukel97
approved these changes
Jul 30, 2025
Contributor
Author
|
@topperc I would appreciate it if you could also have a look at this. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Follow up on a review of bd66fd0 (
[CostModel/RISCV] Fix costs of vector [l](lrint|lround)) post-landing to fix a subtle problem with the cost of vector[l](lrint|lround). We should use source LMUL in the case of a narrowing op.