[llvm] [SLP] Avoid seeding related affine loop address computations (PR #226220)

Alexey Bataev via llvm-commits llvm-commits at lists.llvm.org
Sun Oct 11 09:48:52 PDT 2026


================
@@ -36692,6 +36707,12 @@ bool SLPVectorizerPass::vectorizeGEPIndices(BasicBlock *BB, BoUpSLP &R) {
         Bundle[BundleIndex++] = GEPIdx;
       }
 
+      if (isStrengthReducibleIndexBundle(Bundle, *SE, *LI, SLPReVec)) {
+        LLVM_DEBUG(dbgs() << "SLP: Not vectorizing strength-reducible address "
+                             "computations.\n");
+        continue;
----------------
alexey-bataev wrote:

```suggestion
        // Left to loop strength reduction, keep the once-used seeds away.
        SmallVector<Instruction *> Arithmetic;
        for (Value *Idx : Bundle)
          collectIndexArithmetic(Idx, *LI, Arithmetic);
        for (Instruction *I : Arithmetic)
          SkippedIndexOps[I->getParent()].push_back(I);
        continue;
```

https://github.com/llvm/llvm-project/pull/226220


More information about the llvm-commits mailing list