[PATCH] D69654: AMDGPU: Switch backend default max workgroup size to 1024

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added a comment.

In D69654#1729035 <https://reviews.llvm.org/D69654#1729035>, @arsenm wrote:

> In D69654#1729022 <https://reviews.llvm.org/D69654#1729022>, @rampitec wrote:
>
> > I do not think that deliberately introducing performance regression is a good way to force FE to do anything.
>
>
> clang already emits the clamp to 256 if unspecified. The bugs from not being correct by default have come up many times


OK, thanks.



================
Comment at: llvm/test/CodeGen/AMDGPU/hsa-metadata-kernel-code-props-v3.ll:15
 ; CHECK:     .kernarg_segment_size: 24
-; CHECK:     .max_flat_workgroup_size: 256
+; CHECK:     .max_flat_workgroup_size: 1024
 ; CHECK:     .name:           test
----------------
arsenm wrote:
> rampitec wrote:
> > And given that getMaxFlatWorkGroupSize() returns 2048 I do not understand how does it work.
> D66812 changes this to 1024
Ok


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69654/new/

https://reviews.llvm.org/D69654