[PATCH] D117494: AMDGPU: Account for usage HIP-style dynamic LDS

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

In D117494#3248955 <https://reviews.llvm.org/D117494#3248955>, @yaxunl wrote:

> This may cause some perf degradation.
>
> Do we know why promote alloca does not work for dynamic LDS? Is there a better way to handle this? Thanks.

We can't take LDS if the user may be using an unknown amount. The only way we could continue to do this is if we had an explicit optimization hint provided telling us either the max dynamic total size or allocation


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D117494/new/

https://reviews.llvm.org/D117494