https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102473
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Target Milestone|--- |12.0
Last reconfirmed| |2021-09-24
Summary|521.wrf_r 5% slower at |[12 Regression] 521.wrf_r
|-Ofast and generic x86_64 |5% slower at -Ofast and
|tuning after |generic x86_64 tuning after
|r12-3426-g8f323c712ea76c |r12-3426-g8f323c712ea76c
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Looks like at least on Zen movs[hl]dup is on the integer domain so we'l see a
domain crossing penalty here(?). But since this is a generic arch/tuning
regression the SSE2 code path should be what matters - on the committed
testcase I see
foo:
.LFB572:
.cfi_startproc
pxor %xmm0, %xmm0
addss (%rdi), %xmm0
addss 4(%rdi), %xmm0
addss 8(%rdi), %xmm0
addss 12(%rdi), %xmm0
ret
where it seems that the vectorizer doesn't pick up the reduction pattern.
/home/rguenther/src/gcc2/gcc/testsuite/gcc.target/i386/sse2-pr101059.c:20:21:
note: vect_is_simple_use: vectype vector(4) float
/home/rguenther/src/gcc2/gcc/testsuite/gcc.target/i386/sse2-pr101059.c:20:21:
missed: reduc op not supported by target.