[llvm] [LV][EVL] Generate negative strided load/store for reversed load/store (PR #123608)

[Elvis Wang via llvm-commits]

ElvisWang123 wrote:

> Thanks! I think I have followed your suggestion and ported the new cost model. But it seems that the legacy cost model and VPlan-based cost model have different decisions of best factor. I debugged it and I don't think there is anything wrong.

I think the legacy model doesn't calculate the correct cost when using strided load/store to replace the reverse load/store so the VF mismatch.
You can implement this calculation in the legacy model. Or use `planContainsAdditionalSimplifications` to bypass the assertion.


https://github.com/llvm/llvm-project/pull/123608