Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

release/18.x: [InstCombine] Drop UB-implying attrs/metadata after speculating an instruction (#85542) #85562

Merged
merged 1 commit into from
Mar 18, 2024

Conversation

llvmbot
Copy link
Collaborator

@llvmbot llvmbot commented Mar 17, 2024

Backport 252d019

Requested by: @dtcxzyw

…struction (llvm#85542)

When speculating an instruction in `InstCombinerImpl::FoldOpIntoSelect`,
the call may result in undefined behavior. This patch drops all
UB-implying attrs/metadata to fix this.

Fixes llvm#85536.

(cherry picked from commit 252d019)
@llvmbot llvmbot requested a review from nikic as a code owner March 17, 2024 06:21
@llvmbot llvmbot added this to the LLVM 18.X Release milestone Mar 17, 2024
@llvmbot
Copy link
Collaborator Author

llvmbot commented Mar 17, 2024

@nikic What do you think about merging this PR to the release branch?

@llvmbot
Copy link
Collaborator Author

llvmbot commented Mar 17, 2024

@llvm/pr-subscribers-llvm-transforms

Author: None (llvmbot)

Changes

Backport 252d019

Requested by: @dtcxzyw


Full diff: https://github.com/llvm/llvm-project/pull/85562.diff

2 Files Affected:

  • (modified) llvm/lib/Transforms/InstCombine/InstructionCombining.cpp (+1)
  • (modified) llvm/test/Transforms/InstCombine/intrinsic-select.ll (+40)
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 5d207dcfd18dd4..6f0cf9d9c8f187 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1455,6 +1455,7 @@ static Value *foldOperationIntoSelectOperand(Instruction &I, SelectInst *SI,
                                              Value *NewOp, InstCombiner &IC) {
   Instruction *Clone = I.clone();
   Clone->replaceUsesOfWith(SI, NewOp);
+  Clone->dropUBImplyingAttrsAndMetadata();
   IC.InsertNewInstBefore(Clone, SI->getIterator());
   return Clone;
 }
diff --git a/llvm/test/Transforms/InstCombine/intrinsic-select.ll b/llvm/test/Transforms/InstCombine/intrinsic-select.ll
index a203b28bcb82a8..f37226bbd5b09c 100644
--- a/llvm/test/Transforms/InstCombine/intrinsic-select.ll
+++ b/llvm/test/Transforms/InstCombine/intrinsic-select.ll
@@ -240,3 +240,43 @@ define i32 @vec_to_scalar_select_vector(<2 x i1> %b) {
   %c = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %s)
   ret i32 %c
 }
+
+define i8 @test_drop_noundef(i1 %cond, i8 %val) {
+; CHECK-LABEL: @test_drop_noundef(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = call i8 @llvm.smin.i8(i8 [[VAL:%.*]], i8 0)
+; CHECK-NEXT:    [[RET:%.*]] = select i1 [[COND:%.*]], i8 -1, i8 [[TMP0]]
+; CHECK-NEXT:    ret i8 [[RET]]
+;
+entry:
+  %sel = select i1 %cond, i8 -1, i8 %val
+  %ret = call noundef i8 @llvm.smin.i8(i8 %sel, i8 0)
+  ret i8 %ret
+}
+
+define i1 @pr85536(i32 %a) {
+; CHECK-LABEL: @pr85536(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp ult i32 [[A:%.*]], 31
+; CHECK-NEXT:    [[SHL1:%.*]] = shl nsw i32 -1, [[A]]
+; CHECK-NEXT:    [[ZEXT:%.*]] = zext i32 [[SHL1]] to i64
+; CHECK-NEXT:    [[SHL2:%.*]] = shl i64 [[ZEXT]], 48
+; CHECK-NEXT:    [[SHR:%.*]] = ashr exact i64 [[SHL2]], 48
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.smin.i64(i64 [[SHR]], i64 0)
+; CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 65535
+; CHECK-NEXT:    [[RET1:%.*]] = icmp eq i64 [[TMP1]], 0
+; CHECK-NEXT:    [[RET:%.*]] = select i1 [[CMP1]], i1 [[RET1]], i1 false
+; CHECK-NEXT:    ret i1 [[RET]]
+;
+entry:
+  %cmp1 = icmp ugt i32 %a, 30
+  %shl1 = shl nsw i32 -1, %a
+  %zext = zext i32 %shl1 to i64
+  %shl2 = shl i64 %zext, 48
+  %shr = ashr exact i64 %shl2, 48
+  %sel = select i1 %cmp1, i64 -1, i64 %shr
+  %smin = call noundef i64 @llvm.smin.i64(i64 %sel, i64 0)
+  %masked = and i64 %smin, 65535
+  %ret = icmp eq i64 %masked, 0
+  ret i1 %ret
+}

Copy link
Contributor

@nikic nikic left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@tstellar tstellar merged commit 7fd9979 into llvm:release/18.x Mar 18, 2024
12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Development

Successfully merging this pull request may close these issues.

None yet

4 participants