[PATCH] D86578: [TargetLowering] Combine known bits for icmp in SimplifySetCC (PR41182)

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/test/CodeGen/X86/fold-rmw-ops.ll:1830
 ; CHECK-NEXT:    # fixup A - offset: 2, value: g8-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
----------------
xbolva00 wrote:
> Regression?
I'm really at a loss as to what is best to do for these cases where we've created a constant conditional branch - the middle-end would have nuked most of these cases already (which probably means doing something in CGP is a waste of time as well) and I'm not certain how often they will be generated in backend.

We can tweak the tests in some cases I suppose, which is a cheat but probably the most realistic approach.

Opinions welcome!


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D86578/new/

https://reviews.llvm.org/D86578