[PATCH] D104205: [X86] Schedule-model second (mask) output of GATHER instruction
Simon Pilgrim via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 15 01:47:07 PDT 2021
RKSimon accepted this revision.
RKSimon added a comment.
This revision is now accepted and ready to land.
LGTM with one (optional) rename minor - cheers
================
Comment at: llvm/lib/Target/X86/X86Schedule.td:128
def WriteMove : SchedWrite;
+def WriteGatherMaskWriteback : SchedWrite;
def WriteCopy : WriteSequence<[WriteLoad, WriteStore]>; // mem->mem copy
----------------
RKSimon wrote:
> Maybe call this WriteVecMaskedGather to more closely match the other vector load/store classes? Maybe move it to be with the WriteVecMaskedLoad defs as well for clarity.
>
> I'm assuming we don't need separate float/integer classes for this?
OK, very pedantic but maybe call this WriteVecMaskedGatherWriteback so that it matches the naming convention for vector loads and will match if we add masked gather classes in the future (although tbh they are so diverse in characteristic that it might be easier to just use customs).
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D104205/new/
https://reviews.llvm.org/D104205
More information about the llvm-commits
mailing list