[llvm-branch-commits] [llvm] [ConstantFolding] Set signed/implicitTrunc when handling GEP offsets (PR #116864)

[via llvm-branch-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-analysis

Author: Rose (AreaZR)

<details>
<summary>Changes</summary>

GEP offsets have sext_or_trunc semantics. We were already doing this for the outer-most GEP, but not for the inner ones.

I believe one of the sanitizer buildbot failures was due to this, but I did not manage to reproduce the issue or come up with a test case. Usually the problematic case will already be folded away due to index type canonicalization.

(cherry picked from commit a18dd29077c84fc076a4ed431d9e815a3d0b6f24)

---
Full diff: https://github.com/llvm/llvm-project/pull/116864.diff


1 Files Affected:

- (modified) llvm/lib/Analysis/ConstantFolding.cpp (+2-1) 


``````````diff
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index ff30fece5fce93..a73f3c7ded78be 100644
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -924,7 +924,8 @@ Constant *SymbolicallyEvaluateGEP(const GEPOperator *GEP,
     Ptr = cast<Constant>(GEP->getOperand(0));
     SrcElemTy = GEP->getSourceElementType();
     Offset = Offset.sadd_ov(
-        APInt(BitWidth, DL.getIndexedOffsetInType(SrcElemTy, NestedOps)),
+        APInt(BitWidth, DL.getIndexedOffsetInType(SrcElemTy, NestedOps),
+              /*isSigned=*/true, /*implicitTrunc=*/true),
         Overflow);
   }
 

``````````

</details>


https://github.com/llvm/llvm-project/pull/116864