[PATCH] D150246: AMDGPU: remove an illegal transform for shl(or)

[Jay Foad via Phabricator via llvm-commits]

foad added a comment.

> The right way is to transform the pattern to `(a << c2) | (c1 << c2)`
> But the right transformation does not do any help on folding the
> constant offset into the memory instructions.

It should help because `SelectionDAG::isBaseWithConstantOffset` knows how to match OR (if the known bits do not overlap) as well as ADD.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D150246/new/

https://reviews.llvm.org/D150246