[PATCH] D152408: AMDGPU: Avoid endpgm in middle of block for fallback trap lowering.

[Ruiling, Song via Phabricator via llvm-commits]

ruiling accepted this revision.
ruiling added a comment.
This revision is now accepted and ready to land.

I think we definitely should do this! just one inline comment.



================
Comment at: llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp:5669
+  if (std::next(MI.getIterator()) == BB.end()) {
+    BuildMI(BB, BB.end(), DL, B.getTII().get(AMDGPU::S_ENDPGM))
+      .addImm(0);
----------------
I am not sure whether BB might have a fall-through successor here, an assertion saying that won't happen might be helpful. We not only want s_endpgm be terminator, but also the end of function.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152408/new/

https://reviews.llvm.org/D152408