public inbox for
 help / color / mirror / Atom feed
* [PATCH][0/5][committed] aarch64: Reimplement [US]Q(R)SHR(U)N(2) patterns with standard RTL codes
@ 2023-06-16 13:07 Kyrylo Tkachov
  0 siblings, 0 replies; only message in thread
From: Kyrylo Tkachov @ 2023-06-16 13:07 UTC (permalink / raw)
  To: gcc-patches

Hi all,

This patch series reimplements the MD patterns for the instructions that
perform narrowing right shifts with optional rounding and saturation
using standard RTL codes rather than unspecs.  This includes the scalar
forms and the *2 forms that write to the high half of the result vector.
This allows us to get rid of a number of unspecs and should significantly
improve the simplification capabilities around these instructions.
I attempted to compress as many forms as possible with iterators and the
end result looks reasonably orthogonal with a few small exceptions described
in the individual patches.

The semantics are pretty well exercised by tests in advsimd-intrinsics.exp and
in many of those tests the intrinsics involved are now entirely evaluated at
compile-time and disappear from the output at optimisation levels. The validation
against the reference numbers still passes (though I got many failures during
development as I was getting little things wrong, so the tests are working as

Bootstrapped and tested on aarch64-none-linux-gnu and aarch64_be-none-elf.
Pushing to trunk.

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

only message in thread, other threads:[~2023-06-16 13:07 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-06-16 13:07 [PATCH][0/5][committed] aarch64: Reimplement [US]Q(R)SHR(U)N(2) patterns with standard RTL codes Kyrylo Tkachov

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).