[PATCH] D62024: [InstCombine] fold shuffles of insert_subvectors

Tue May 21 10:01:13 PDT 2019

spatel added a comment.

In D62024#1509309 <https://reviews.llvm.org/D62024#1509309>, @efriedma wrote:

> > We use "alignTo(7, 3)":
>
> Hmm.  It looks like the idea is that we should perform the shuffle in some type that's a multiple of the source vectors, so we can construct the source operands using CONCAT_VECTORS.  Not sure why we aren't just generating an INSERT_SUBVECTOR into undef, though.

INSERT_SUBVECTOR makes sense to me, but I just tried that experiment and...fallout: x86 has regressions/crashing, AMDGPU is infinite looping.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D62024/new/

https://reviews.llvm.org/D62024