[llvm] [VPlan] Optz WidenGEP with single-scalar operands (PR #166740)

Luke Lau via llvm-commits llvm-commits at lists.llvm.org
Thu Nov 6 05:30:51 PST 2025


lukel97 wrote:

> This is different to VPWidenSelect I think, which even with scalar condition produces a vector result for vector operands?
> 
> if all operands are single scalar, we should be able to handle this in narrowToSingleScalar, and remove the special handling in the recipe altogether?

I agree it would be nice to handle it there but I don't think we have a recipe to represent an arbitrary scalar GEP yet. I think we only have VPInstruction:PtrAdd. 

https://github.com/llvm/llvm-project/pull/166740


More information about the llvm-commits mailing list