[PATCH] D101924: [X86] Improve costmodel for scalar byte swaps
Roman Lebedev via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu May 6 07:03:10 PDT 2021
lebedev.ri marked 2 inline comments as done.
lebedev.ri added inline comments.
================
Comment at: llvm/lib/Target/X86/X86.td:519
+ : SubtargetFeature<"fast-movbe", "HasFastMOVBE", "true",
+ "Prefer a movbe over a single-use load + bswap / single-use bswap + store">;
+
----------------
RKSimon wrote:
> lebedev.ri wrote:
> > RKSimon wrote:
> > > You say 'prefer' - is a future intention to alter isel depending on this? I'm not sure that's actually useful.
> > > You say 'prefer' - is a future intention to alter isel depending on this?
> >
> > Not really. I was only interested in doing what is contained in this patch,
> > i.e. modelling bswap as free in costmodel, if it will be free.
> I guess technically different codegen might still occur if the costs come out slightly different and the vectorizer 'prefers' scalar code - sorry for the noise.
That wasn't noise. "Do you intend to prohibit formation of movbe if not HasFastMOVBE" seems like a good question.
(i do not.)
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D101924/new/
https://reviews.llvm.org/D101924
More information about the llvm-commits
mailing list