[LoopVectorizer] Add a -force-vscale-for-tuning override option. #156916
+13
−0
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
|
@llvm/pr-subscribers-vectorizers @llvm/pr-subscribers-llvm-transforms Author: David Green (davemgreen) ChangesIt can be useful for debugging and tuning to be able to alter the VScaleForTuning. This adds a quick option to the vectorizer for it. It overrides the VScaleForTuning in the vectorizer even when the vscale is known, as the options is a "force". Full diff: https://github.com/llvm/llvm-project/pull/156916.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 3fbeef1211954..bf84a571e679e 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -303,6 +303,10 @@ static cl::opt<bool> ForceTargetSupportsScalableVectors(
"Pretend that scalable vectors are supported, even if the target does "
"not support them. This flag should only be used for testing."));
+static cl::opt<unsigned>
+ VScaleForTuningOpt("force-vscale-for-tuning", cl::Hidden,
+ cl::desc("Force a vscale for tuning factor in the loop vectorizer"));
+
static cl::opt<unsigned> SmallLoopCost(
"small-loop-cost", cl::init(20), cl::Hidden,
cl::desc(
@@ -1473,6 +1477,11 @@ class LoopVectorizationCostModel {
/// vscale_range.min == vscale_range.max then return vscale_range.max, else
/// return the value returned by the corresponding TTI method.
void initializeVScaleForTuning() {
+ if (VScaleForTuningOpt.getNumOccurrences()) {
+ VScaleForTuning = VScaleForTuningOpt;
+ return;
+ }
+
const Function *Fn = TheLoop->getHeader()->getParent();
if (Fn->hasFnAttribute(Attribute::VScaleRange)) {
auto Attr = Fn->getFnAttribute(Attribute::VScaleRange);
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-vectorization-cost-tuning.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-vectorization-cost-tuning.ll
index c4aee69db70b3..16d3786681ffa 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-vectorization-cost-tuning.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-vectorization-cost-tuning.ll
@@ -7,6 +7,10 @@
; RUN: -force-target-instruction-cost=1 -passes=loop-vectorize -S -debug-only=loop-vectorize --disable-output < %s 2>&1 \
; RUN: | FileCheck %s --check-prefixes=VSCALEFORTUNING1
+; RUN: opt -mtriple=aarch64 -mattr=+sve -mcpu=generic -force-vscale-for-tuning=2 \
+; RUN: -force-target-instruction-cost=1 -passes=loop-vectorize -S -debug-only=loop-vectorize --disable-output < %s 2>&1 \
+; RUN: | FileCheck %s --check-prefixes=VSCALEFORTUNING2
+
; RUN: opt -mtriple=aarch64 -mcpu=neoverse-v1 \
; RUN: -force-target-instruction-cost=1 -passes=loop-vectorize -S -debug-only=loop-vectorize --disable-output < %s 2>&1 \
; RUN: | FileCheck %s --check-prefixes=VSCALEFORTUNING2
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
It can be useful for debugging and tuning to be able to alter the VScaleForTuning. This adds a quick option to the vectorizer for it
bc701eb to
23ced2e
Compare
david-arm
reviewed
Sep 4, 2025
| @@ -1473,6 +1477,11 @@ class LoopVectorizationCostModel { | |||
| /// vscale_range.min == vscale_range.max then return vscale_range.max, else | |||
| /// return the value returned by the corresponding TTI method. | |||
| void initializeVScaleForTuning() { | |||
| if (VScaleForTuningOpt.getNumOccurrences()) { | |||
There was a problem hiding this comment.
Thanks for this! I like the idea of having a flag to force the choice of vscale, but should this be sanitised according to vscale_range on the function or at least emit a warning that it's architecturally unsupported? For example, if the function has the vscale_range(1, 16) attribute then only power-of-2 vscale values are permitted leading to choices of 1,2,4,8 or 16.
fhahn
reviewed
Sep 4, 2025
Comment on lines
+1480
to
+1483
| if (VScaleForTuningOpt.getNumOccurrences()) { | ||
| VScaleForTuning = VScaleForTuningOpt; | ||
| return; | ||
| } |
There was a problem hiding this comment.
Is the purpose of the option to deliberately override the function attribute?
Does this have to be LV specific? Could this be handled in getVScaleForTuning?
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.
It can be useful for debugging and tuning to be able to alter the VScaleForTuning. This adds a quick option to the vectorizer for it. It overrides the VScaleForTuning in the vectorizer even when the vscale is known, as the options is a "force".