[IR] Fix UB on Op<2> in ShuffleVector predicates

[rnk]

This Op<2> usage was missed in 1ee6ec2, which replaced the third shuffle operand with a vector of integer mask constants.

I noticed this when attempting to make changes to the layout of llvm::Value.

[llvmbot]

@llvm/pr-subscribers-llvm-ir

Author: Reid Kleckner (rnk)

Changes

This Op<2> usage was missed in 1ee6ec2, which replaced the third shuffle operand with a vector of integer mask constants.

I noticed this when attempting to make changes to the layout of llvm::Value.

---

Full diff: https://github.com/llvm/llvm-project/pull/75549.diff

1 Files Affected:

• (modified) llvm/lib/IR/Instructions.cpp (+1-8)

diff --git a/llvm/lib/IR/Instructions.cpp b/llvm/lib/IR/Instructions.cpp
index bc228a577c6a6c..299b4e74677dcc 100644
--- a/llvm/lib/IR/Instructions.cpp
+++ b/llvm/lib/IR/Instructions.cpp
@@ -2412,9 +2412,6 @@ bool ShuffleVectorInst::isInsertSubvectorMask(ArrayRef<int> Mask,
 }
 
 bool ShuffleVectorInst::isIdentityWithPadding() const {
-  if (isa<UndefValue>(Op<2>()))
-    return false;
-
   // FIXME: Not currently possible to express a shuffle mask for a scalable
   // vector for this case.
   if (isa<ScalableVectorType>(getType()))
@@ -2439,9 +2436,6 @@ bool ShuffleVectorInst::isIdentityWithPadding() const {
 }
 
 bool ShuffleVectorInst::isIdentityWithExtract() const {
-  if (isa<UndefValue>(Op<2>()))
-    return false;
-
   // FIXME: Not currently possible to express a shuffle mask for a scalable
   // vector for this case.
   if (isa<ScalableVectorType>(getType()))
@@ -2457,8 +2451,7 @@ bool ShuffleVectorInst::isIdentityWithExtract() const {
 
 bool ShuffleVectorInst::isConcat() const {
   // Vector concatenation is differentiated from identity with padding.
-  if (isa<UndefValue>(Op<0>()) || isa<UndefValue>(Op<1>()) ||
-      isa<UndefValue>(Op<2>()))
+  if (isa<UndefValue>(Op<0>()) || isa<UndefValue>(Op<1>()))
     return false;
 
   // FIXME: Not currently possible to express a shuffle mask for a scalable

[nikic]

LGTM

[rnk]

It occurs to me that there's a missed opportunity to add some kind of bounds check assertion to the static index Op<> accessor helpers.