[PATCH] D69288: [GISel][ArtifactCombiner] Relax the constraint to combine unmerge with concat_vectors

[Quentin Colombet via Phabricator via llvm-commits]

qcolombet added a comment.

> Are you planning on fixing the regression in the near future? If so splitting the patches this way seems fine

I plan to fix the regressions, but I cannot commit on a deadline yet. In particular, I don't expect I will have time the next coming few weeks.

The reason I still believe it makes sense to land the patch as is is because the code without this patch will assert on inputs that looks like `test_unmerge_values_s128_of_zext_of_concat_vectors` (yes the name doesn't make sense, I'll fix that :)). I.e., I would rather generate worse code (like you mentioned) than potentially crashing the compiler.

What do you think?


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69288/new/

https://reviews.llvm.org/D69288