[PATCH] D140493: [SROA] Support promotion in presence of variably-indexed loads

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added a comment.

D140638 <https://reviews.llvm.org/D140638> (`[Codegen][LegalizeIntegerTypes] New legalization strategy for scalar shifts: shift through stack.`)
has taught LLVM how to undo this kind transformation during codegen,
in fact, using effectively the very same profitability heuristic that i use here,
so we don't seem to need a TLI hook here.

I'm still looking into a few codegen improvements,
but otherwise i'm not tracking any other issues with this,
so once i'm done with codegen, this will proceed.
(SROA still needs a "are we done with inlining?" flag, yes.)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D140493/new/

https://reviews.llvm.org/D140493