[PATCH] D145329: AMDGPU: Always split blocks for si_end_cf

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Mar 6 05:07:56 PST 2023


arsenm added a comment.

I was assuming WQM needs to split blocks more aggressively itself to avoid the same problems



================
Comment at: llvm/test/CodeGen/AMDGPU/collapse-endcf.ll:264
+; GCN-O0-NEXT:  .LBB1_5: ; %bb.inner.end
+; GCN-O0-NEXT:    v_readlane_b32 s0, v1, 4
+; GCN-O0-NEXT:    v_readlane_b32 s1, v1, 5
----------------
critson wrote:
> Is this reordering fixing the bug mentioned in the description?
> (Exec mask is restored before buffer_load, rather than after.)
Yes, previously we would only correctly handle spills used for the exec source value, not other spills


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D145329/new/

https://reviews.llvm.org/D145329



More information about the llvm-commits mailing list