[PATCH] D72737: [AMDGPU] Bundle loads before post-RA scheduler

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added a comment.

In D72737#1820974 <https://reviews.llvm.org/D72737#1820974>, @kerbowa wrote:

> In D72737#1820946 <https://reviews.llvm.org/D72737#1820946>, @rampitec wrote:
>
> > In D72737#1820865 <https://reviews.llvm.org/D72737#1820865>, @kerbowa wrote:
> >
> > > By the way I think we actually do try to insert waitcnt into bundles. I wonder if it would be worth it to just update the memory legalizer to work with bundles as well.
> >
> >
> > I saw waitcount inserted after the bundle if I do not unpack them, so probably we do not do it always if we really do.
>
>
> Yeah maybe. I remember it was added here: c04aab9c0646461bc187808920b3d5ee7f5cc5ab
>  Was the waitcnt after the bundle that you saw in the correct place?


Yes, it was correct. It was just vmcnt(0) because we could not tear apart load bundle and need all the loaded values in a store bundle. When I unpack it each store was waiting for vmcnt(3) which is better.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72737/new/

https://reviews.llvm.org/D72737