[LLVM][LangRef] Redefine out-of-range stepvector values as being truncated. #173494
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
…cated. The LangRef current defines out-of-range stepvector values as poison. This property is at odds with both the expansion used for fixed-length vectors and the equivalent ISD node, both of which implicitly truncate out-of-range values. NOTE: In order to keep the PR mostly NFC I would like to defer the follow extensions to seperate PRs. 1) The new definition means the "8-bit" restriction can be lifted because that only existed due to problematic cases like `<vscale x n x i1> stepvector()`, which by definition is mostly poison. Defering because I'm unsure of the code generation support for smaller types, as a minimum we're missing test coverage. 2) The instcombine can fire in many more case, and the current constant handling can be a simplification rather than a combine.
llvmbot
added
llvm:instcombine
Member
|
@llvm/pr-subscribers-llvm-ir @llvm/pr-subscribers-llvm-transforms Author: Paul Walker (paulwalker-arm) ChangesThe LangRef current defines out-of-range stepvector values as poison. This property is at odds with both the expansion used for fixed-length vectors and the equivalent ISD node, both of which implicitly truncate out-of-range values. NOTE: In order to keep the PR mostly NFC I would like to defer the follow extensions to seperate PRs.
Full diff: https://github.com/llvm/llvm-project/pull/173494.diff 3 Files Affected:
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index d99280f05e73f..5b462b87acb0f 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -20828,7 +20828,7 @@ of integers whose elements contain a linear sequence of values starting from 0
with a step of 1. This intrinsic can only be used for vectors with integer
elements that are at least 8 bits in size. If the sequence value exceeds
the allowed limit for the element type then the result for that lane is
-a poison value.
+truncated.
These intrinsics work for both fixed and scalable vectors. While this intrinsic
supports all vector types, the recommended way to express this operation for
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
index 98e2d9ebe4fc2..f5e8d341e3493 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -448,7 +448,7 @@ Instruction *InstCombinerImpl::visitExtractElementInst(ExtractElementInst &EI) {
if (IndexC->getValue().getActiveBits() <= BitWidth)
Idx = ConstantInt::get(Ty, IndexC->getValue().zextOrTrunc(BitWidth));
else
- Idx = PoisonValue::get(Ty);
+ return nullptr;
return replaceInstUsesWith(EI, Idx);
}
}
diff --git a/llvm/test/Transforms/InstCombine/vscale_extractelement.ll b/llvm/test/Transforms/InstCombine/vscale_extractelement.ll
index 9ac8a92abb689..ec58307a253d1 100644
--- a/llvm/test/Transforms/InstCombine/vscale_extractelement.ll
+++ b/llvm/test/Transforms/InstCombine/vscale_extractelement.ll
@@ -214,12 +214,15 @@ entry:
ret i64 %1
}
-; Check that poison is returned when the extracted element has wrapped.
+; TODO: stepvector now wraps rather than poisons elements when the value does
+; not fit, so this should return 0.
define i8 @ext_lane256_from_stepvec() {
; CHECK-LABEL: @ext_lane256_from_stepvec(
; CHECK-NEXT: entry:
-; CHECK-NEXT: ret i8 poison
+; CHECK-NEXT: [[TMP0:%.*]] = call <vscale x 512 x i8> @llvm.stepvector.nxv512i8()
+; CHECK-NEXT: [[TMP1:%.*]] = extractelement <vscale x 512 x i8> [[TMP0]], i64 256
+; CHECK-NEXT: ret i8 [[TMP1]]
;
entry:
%0 = call <vscale x 512 x i8> @llvm.stepvector.nxv512i8()
|
nikic
reviewed
Dec 24, 2025
| unsigned BitWidth = Ty->getIntegerBitWidth(); | ||
| Value *Idx; | ||
| // Return index when its value does not exceed the allowed limit | ||
| // for the element type of the vector, otherwise return undefined. |
Contributor
There was a problem hiding this comment.
This comment is outdated now.
Why not drop the if entirely here instead of keeping the nullptr return?
Collaborator
Author
There was a problem hiding this comment.
Why not drop the if entirely here instead of keeping the nullptr return?
That's point two in the message. This whole functionality belongs in simplifyExtractElementInst, so I figured it's better to do the work once in a separate PR?
Contributor
There was a problem hiding this comment.
I'm fine with that. Maybe worth noting that InstCombine could handle the generalized case of a variable index extract (where a zext/trunc instruction may have to be generated).
topperc
approved these changes
Dec 24, 2025
Collaborator
topperc
left a comment
topperc
left a comment
There was a problem hiding this comment.
LGTM. The RISC-V instruction vid.v we use for stepvector has the truncate behavior.
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/95/builds/19539 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/157/builds/43267 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/129/builds/35525 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/186/builds/14985 Here is the relevant piece of the build log for the reference |
nikic
added a commit
that referenced
this pull request
Dec 25, 2025
LV can create step vectors that wrap around, e.g. `step-vector i1` with VF>2. Allow truncation when creating the vector constant to avoid an assertion failure with #171456. After #173494 the definition of the llvm.stepvector intrinsic has been changed to make it have wrapping semantics, so the semantics for the fixed and scalable case match now.
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Dec 25, 2025
…#173229) LV can create step vectors that wrap around, e.g. `step-vector i1` with VF>2. Allow truncation when creating the vector constant to avoid an assertion failure with llvm/llvm-project#171456. After llvm/llvm-project#173494 the definition of the llvm.stepvector intrinsic has been changed to make it have wrapping semantics, so the semantics for the fixed and scalable case match now.
mahesh-attarde
pushed a commit
to mahesh-attarde/llvm-project
that referenced
this pull request
Jan 6, 2026
…cated. (llvm#173494) The LangRef current defines out-of-range stepvector values as poison. This property is at odds with both the expansion used for fixed-length vectors and the equivalent ISD node, both of which implicitly truncate out-of-range values.
mahesh-attarde
pushed a commit
to mahesh-attarde/llvm-project
that referenced
this pull request
Jan 6, 2026
LV can create step vectors that wrap around, e.g. `step-vector i1` with VF>2. Allow truncation when creating the vector constant to avoid an assertion failure with llvm#171456. After llvm#173494 the definition of the llvm.stepvector intrinsic has been changed to make it have wrapping semantics, so the semantics for the fixed and scalable case match now.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 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.
The LangRef current defines out-of-range stepvector values as poison. This property is at odds with both the expansion used for fixed-length vectors and the equivalent ISD node, both of which implicitly truncate out-of-range values.
NOTE: In order to keep the PR mostly NFC I would like to defer the follow extensions to seperate PRs.
The new definition means the "8-bit" restriction can be lifted
because that only existed due to problematic cases like
<vscale x n x i1> stepvector(), which by definition is mostly poison.Defering because I'm unsure of the code generation support for
smaller types, as a minimum we're missing test coverage.
The instcombine can fire in many more case, and the current constant
handling can be a simplification rather than a combine.