[llvm] [RISCV] Improve getInterleavedMemoryOpCost for interleave groups with tail gaps. (PR #192074)
Mel Chen via llvm-commits
llvm-commits at lists.llvm.org
Mon Apr 27 03:41:13 PDT 2026
Mel-Chen wrote:
> > > Is it possible to add a test under `Analysis/CostModel` for this?
> >
> >
> > Unfortunately not currently AFAIK since the costs that are printed there are from `TargetTransformInfo::getInstructionCost`, which only sees one instruction at a time, e.g. first the load and then the shufflevectors or deinterleave. I agree it would be nice though if there was a way to test it there.
>
> Yes, I think we can not check `getInterleavedMemoryOpCost()` by Analysis/CostModel because it is required the interleaved analysis by vectorizer. If we want to verify the cost for interleaved accesses, a possible way is using `-debug-only=loop-vectorize` and inspecting the `Cost of X for VF Y: ... `. @eas WDYT?
Ping @eas
https://github.com/llvm/llvm-project/pull/192074
More information about the llvm-commits
mailing list