[PATCH] D64512: [InstCombine] Dropping redundant masking before left-shift [0/5] (PR42563)

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added a comment.

> With InstSimplify, we're using a potentially expensive/recursive analysis (similar to ValueTracking), and we know that's where InstCombine spends a lot of time.



> cc'ing @efriedma for further thoughts.



> There's 1 more possibility: move this to AggressiveInstCombine (which is currently only enabled at -O3...).

Note that before we get to the `SimplifyBinOp()` we first need to match 3 or 4 instructions,
which is more strict than the previously-added `reassociateShiftAmtsOfTwoSameDirectionShifts()` fold.
@efriedma  If retargeting *this* fold into AggressiveInstCombine is the only acceptable way forward
please let me know to avoid lack of forward progress/stalling..


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D64512/new/

https://reviews.llvm.org/D64512