[PATCH] D138874: [InstCombine] canonicalize trunc + insert as bitcast + shuffle, part 3

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

So I think the question is - do we treat this as a canonicalization issue (in which case this code can remain in InstCombine) or do we make it a cost driven fold and move it to VectorCombine?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D138874/new/

https://reviews.llvm.org/D138874