[PATCH] D156087: [MLIR] Add stage to side effect

donald chen via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Aug 1 15:28:42 PDT 2023


cxy-1993 added a comment.

In D156087#4551469 <https://reviews.llvm.org/D156087#4551469>, @mehdi_amini wrote:

>> I believe that when we say “write full” to a particular resource, this write is certain to occur. Therefore, any writes to this resource before that point can be killed by this operation. If the write is only possible, and we shouldn't explicitly mark it as writing all resources, do you think this approach is feasible?
>
> It might be ok, I am not sure it is conservative in every situation? Assuming it is, it should likely be the default, right now your MemAlloc default to full.

I agree that treating side effects as default on partially shaped memrefs is a cautious approach. Currently, almost all visible ops have side effects on the entire memref, so we can consider this common but less safe behavior as the default and require the definers of specific ops to understand the scope of side effects and make modifications accordingly.

Otherwise, I would have to modify almost all existing side effects to become FullEffect.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D156087/new/

https://reviews.llvm.org/D156087



More information about the llvm-commits mailing list