[InstCombine] Restrict foldBitCeil to power-of-two integer widths#173849
Merged
[InstCombine] Restrict foldBitCeil to power-of-two integer widths#173849
foldBitCeil to power-of-two integer widths#173849Conversation
Member
|
@llvm/pr-subscribers-llvm-transforms Author: Yunbo Ni (cardigan1008) ChangesThe masking rewrite in For non-power-of-two integer types, This patch restricts the transform to power-of-two bitwidths. Alive2 proof: https://alive2.llvm.org/ce/z/i2E6zT Fixes #173787 Full diff: https://github.com/llvm/llvm-project/pull/173849.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 67d1845832725..b7ab642cde6ea 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -3871,6 +3871,9 @@ static Instruction *foldBitCeil(SelectInst &SI, IRBuilderBase &Builder,
ShouldDropNoWrap))
return nullptr;
+ if (!isPowerOf2_32(BitWidth))
+ return nullptr;
+
if (ShouldDropNoWrap) {
cast<Instruction>(CtlzOp)->setHasNoUnsignedWrap(false);
cast<Instruction>(CtlzOp)->setHasNoSignedWrap(false);
diff --git a/llvm/test/Transforms/InstCombine/bit_ceil.ll b/llvm/test/Transforms/InstCombine/bit_ceil.ll
index 09f90ee05735d..edf1c176ff12c 100644
--- a/llvm/test/Transforms/InstCombine/bit_ceil.ll
+++ b/llvm/test/Transforms/InstCombine/bit_ceil.ll
@@ -337,6 +337,25 @@ define i32 @test_drop_range_attr(i32 %x) {
ret i32 %sel
}
+define i33 @test_bit_ceil_i33_non_pow2(i33 %x) {
+; CHECK-LABEL: @test_bit_ceil_i33_non_pow2(
+; CHECK-NEXT: [[CTLZ:%.*]] = call range(i33 0, 34) i33 @llvm.ctlz.i33(i33 [[X:%.*]], i1 false)
+; CHECK-NEXT: [[TMP2:%.*]] = sub nuw nsw i33 33, [[CTLZ]]
+; CHECK-NEXT: [[SEL:%.*]] = shl nuw i33 1, [[TMP2]]
+; CHECK-NEXT: [[DEC:%.*]] = add i33 [[X]], -1
+; CHECK-NEXT: [[ULT:%.*]] = icmp ult i33 [[DEC]], -2
+; CHECK-NEXT: [[SEL1:%.*]] = select i1 [[ULT]], i33 [[SEL]], i33 1
+; CHECK-NEXT: ret i33 [[SEL1]]
+;
+ %ctlz = call i33 @llvm.ctlz.i33(i33 %x, i1 false)
+ %sub = sub i33 33, %ctlz
+ %shl = shl i33 1, %sub
+ %dec = add i33 %x, -1
+ %ult = icmp ult i33 %dec, -2
+ %sel = select i1 %ult, i33 %shl, i33 1
+ ret i33 %sel
+}
+
define i32 @bit_ceil_plus_nsw(i32 %x) {
; CHECK-LABEL: @bit_ceil_plus_nsw(
; CHECK-NEXT: entry:
@@ -378,5 +397,6 @@ entry:
}
declare i32 @llvm.ctlz.i32(i32, i1 immarg)
+declare i33 @llvm.ctlz.i33(i33, i1 immarg)
declare i64 @llvm.ctlz.i64(i64, i1 immarg)
declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32>, i1)
|
mahesh-attarde
pushed a commit
to mahesh-attarde/llvm-project
that referenced
this pull request
Jan 6, 2026
…lvm#173849) The masking rewrite in `foldBitCeil` assumes a power-of-two bitwidth. For non-power-of-two integer types, `(-ctlz) & (BitWidth - 1)` is not equivalent to `BitWidth - ctlz` and can miscompile. This patch restricts the transform to power-of-two bitwidths. Alive2 proof: https://alive2.llvm.org/ce/z/i2E6zT Fixes llvm#173787
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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 masking rewrite in
foldBitCeilassumes a power-of-two bitwidth.For non-power-of-two integer types,
(-ctlz) & (BitWidth - 1)is not equivalent toBitWidth - ctlzand can miscompile.This patch restricts the transform to power-of-two bitwidths.
Alive2 proof: https://alive2.llvm.org/ce/z/i2E6zT
Fixes #173787