[PATCH] D105072: [GISel] Support llvm.memcpy.inline

[Amara Emerson via Phabricator via llvm-commits]

aemerson added a comment.

Why change the existing memcpy test? I would have thought the logical way to test this was to use a G_MEMCPY_INLINE with a length much larger than the heuristics would normally allow for G_MEMCPY expansion, rather than change existing test coverage.



================
Comment at: llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp:1347
+  if (!LenVRegAndVal)
+    return false; // FIXME: support dynamically sized G_MEMCPY_INLINE
+
----------------
Dynamically sized memcpy.inline isn't legal according to the langref, so we should probably assert here.


================
Comment at: llvm/test/CodeGen/AArch64/GlobalISel/inline-memcpy-opt.mir:51
+  attributes #1 = { argmemonly nounwind }
+  attributes #2 = { optsize }
+  attributes #3 = { minsize }
----------------
This IR section is probably unnecessary.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D105072/new/

https://reviews.llvm.org/D105072