[PATCH] D124196: [AMDGPU][SILowerSGPRSpills] Spill SGPRs to virtual VGPRs

[Christudasan Devadasan via Phabricator via llvm-commits]

cdevadas added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/spill-sgpr-to-virtual-vgpr.mir:194
+  ; GCN-NEXT:   [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+  ; GCN-NEXT:   S_CBRANCH_SCC1 %bb.2, implicit killed $scc
+  ; GCN-NEXT: {{  $}}
----------------
cdevadas wrote:
> arsenm wrote:
> > Needs a case where the insert block has no terminators
> I couldn't write one successfully. Will try some unstructured flow to force one.
I don't think such a case exists. A fall-through block will have only one successor and that becomes the nearest dominator for its children.
It would be true even for any unstructured flow.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D124196/new/

https://reviews.llvm.org/D124196