projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 75464e1) | patch
[x86] refactor combineAddOrSubToADCOrSBB(); NFCI
authorSanjay Patel <spatel@rotateright.com>
Fri, 3 Mar 2017 22:35:11 +0000 (22:35 +0000)
committerSanjay Patel <spatel@rotateright.com>
Fri, 3 Mar 2017 22:35:11 +0000 (22:35 +0000)
commit7ee83b41e01e3377b6c96a418eea2bd8b8b5b940
tree9cb6cf2dfce77c5ae57ac4e3c8a699bb9ee639d8tree | snapshot
parent75464e174a4be825e0517efc1fc0be140630ec39commit | diff
[x86] refactor combineAddOrSubToADCOrSBB(); NFCI

The comments were wrong, and this is not an obvious transform.
This hopefully makes it clearer that we're missing the commuted
patterns for adds. It's less clear that this is actually a good
transform for all micro-arch.

This is prep work for trying to clean up the current adc/sbb
codegen because it's definitely not happening optimally.

llvm-svn: 296918
llvm/lib/Target/X86/X86ISelLowering.cpp diff | blob | history
Domain: System / Toolchain;