public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug target/96836] New: [MVE] Consider using define_subst for predicatable operations
@ 2020-08-28 12:06 rsandifo at gcc dot gnu.org
  0 siblings, 0 replies; only message in thread
From: rsandifo at gcc dot gnu.org @ 2020-08-28 12:06 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96836

            Bug ID: 96836
           Summary: [MVE] Consider using define_subst for predicatable
                    operations
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rsandifo at gcc dot gnu.org
                CC: ktkachov at gcc dot gnu.org, ramana at gcc dot gnu.org,
                    rearnsha at gcc dot gnu.org
  Target Milestone: ---
            Target: arm*-*-*

Created attachment 49147
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49147&action=edit
Proof of concept patch.

[This is really just an idea rather than a PR.]

At the moment we use separate patterns for the predicated and
unpredicated forms of MVE instructions.  The predicated ones
have a VPST instruction built in.

This works well, but we might want to consider using define_subst
to define both the predicated and unpredicated forms at the same
time.  As well as shortening the definition, it could help when
making future optimisations, such as:

- putting more than one instruction in the same VPST block
- using VPT instead of a combination of a comparison and a VPST
- using loop predication

I've attached a proof of concept that does this for one arbitrarily-
chosen intrinsic (vorrq_n).  It's not worth submitting on its own,
it would only be useful as part of a wider move to this style.
(Equally, I don't think it would be necessary to convert everything
at once.  We'd just need some indication that the approach is going
somewhere and won't stop at this one intrinsic.)

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2020-08-28 12:06 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-08-28 12:06 [Bug target/96836] New: [MVE] Consider using define_subst for predicatable operations rsandifo at gcc dot gnu.org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).