[PATCH] D141728: [AMDGPU] Tune scheduler on GFX10 and GFX11

[Austin Kerbow via Phabricator via llvm-commits]

kerbowa added a comment.

I think this makes sense, but it also makes the concept of an occupancy target a misnomer.

Additionally, I worry that we may be over-prioritizing RP reduction when we consistently see cases where higher RP isn't necessarily leading to better performance.

This situation is basically why the HighRPReschedule stage exists - where we target a higher occupancy to see if occupancy was dropped because the heuristics are not focusing RP reduction soon enough. This HighRP stage will only increase the target occupancy by one and decrease the critical register limit by an additional 10. We could probably be more aggressive with these numbers in cases where we have spilling. For example, setting the HighRPErrorMargin to something very high like 200 also eliminates the spilling in the testcase for the relevant ticket (SWDEV-377300).

That may be a more targeted change, but if your feeling is that this patch will be a more general improvement it looks good to me.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D141728/new/

https://reviews.llvm.org/D141728