[PATCH] D102830: [AMDGPU] Avoid null export insertion when unifying exit blocks

[Carl Ritson via Phabricator via llvm-commits]

critson created this revision.
critson added reviewers: foad, piotr, ruiling.
Herald added subscribers: kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, nhaehnle, jvesely, kzhuravl, arsenm.
critson requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.

Avoid adding a null export by unifying existing "done" exports
in the unified exit block of a shader with divergent exits.

If multiple incompatible "done" exports exist then they cannot be
unified and a null export is inserted, and "done" bits cleared.
Note: all existing front-ends emit a single export done so this
should not happen in practice.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D102830

Files:
  llvm/lib/Target/AMDGPU/AMDGPUUnifyDivergentExitNodes.cpp
  llvm/test/CodeGen/AMDGPU/kill-infinite-loop.ll
  llvm/test/CodeGen/AMDGPU/multi-divergent-exit-region.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D102830.346653.patch
Type: text/x-patch
Size: 20295 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210520/24e621a4/attachment.bin>