[PATCH] D113748: [X86] combineX86ShufflesConstants - constant fold from target shuffles unless optsize = true

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

In D113748#3126969 <https://reviews.llvm.org/D113748#3126969>, @lebedev.ri wrote:

> I think this makes sense, the more computations we not-repeatedly-do at run-time, the better.
> Is there any understanding of the bloat it was guarding against?
> Any clues as to how much does this grow the constant pool?

shufflevector already constant folds (all the time - it has no optsize limit which is a shame), so I expect most of the constant pool size to come from there already - the reason we see so much pmuludq code changes is because we create the target shuffles directly during lowering


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113748/new/

https://reviews.llvm.org/D113748