public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug target/101464] New: Replace zveroupper with vpxor
@ 2021-07-15 18:26 hjl.tools at gmail dot com
  2021-07-16  1:54 ` [Bug target/101464] " crazylht at gmail dot com
                   ` (3 more replies)
  0 siblings, 4 replies; 5+ messages in thread
From: hjl.tools at gmail dot com @ 2021-07-15 18:26 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101464

            Bug ID: 101464
           Summary: Replace zveroupper with vpxor
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hjl.tools at gmail dot com
  Target Milestone: ---
            Target: i386, x86-64

[hjl@gnu-cfl-2 pr101456]$ cat y.c
#include <x86intrin.h>

extern __m256d x, y;

void
foo (void)
{
  x = y;
}
[hjl@gnu-cfl-2 pr101456]$ gcc -S -O2 -mavx2 y.c
[hjl@gnu-cfl-2 pr101456]$ cat y.s
        .file   "y.c"
        .text
        .p2align 4
        .globl  foo
        .type   foo, @function
foo:
.LFB5667:
        .cfi_startproc
        vmovapd y(%rip), %ymm0
        vmovapd %ymm0, x(%rip)
        vzeroupper
        ret
        .cfi_endproc
.LFE5667:
        .size   foo, .-foo
        .ident  "GCC: (GNU) 11.1.1 20210531 (Red Hat 11.1.1-3)"
        .section        .note.GNU-stack,"",@progbits
[hjl@gnu-cfl-2 pr101456]$ 

vzeroupper can be replaced by a single pxor.

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [Bug target/101464] Replace zveroupper with vpxor
  2021-07-15 18:26 [Bug target/101464] New: Replace zveroupper with vpxor hjl.tools at gmail dot com
@ 2021-07-16  1:54 ` crazylht at gmail dot com
  2021-07-16  2:00 ` hjl.tools at gmail dot com
                   ` (2 subsequent siblings)
  3 siblings, 0 replies; 5+ messages in thread
From: crazylht at gmail dot com @ 2021-07-16  1:54 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101464

Hongtao.liu <crazylht at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |crazylht at gmail dot com

--- Comment #1 from Hongtao.liu <crazylht at gmail dot com> ---
>From the implementation perspective, we need to record all usage of sse
registers which is AVX_U128_DIRTY and clear them all.

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [Bug target/101464] Replace zveroupper with vpxor
  2021-07-15 18:26 [Bug target/101464] New: Replace zveroupper with vpxor hjl.tools at gmail dot com
  2021-07-16  1:54 ` [Bug target/101464] " crazylht at gmail dot com
@ 2021-07-16  2:00 ` hjl.tools at gmail dot com
  2021-07-16  6:26 ` rguenth at gcc dot gnu.org
  2021-07-16 12:07 ` hjl.tools at gmail dot com
  3 siblings, 0 replies; 5+ messages in thread
From: hjl.tools at gmail dot com @ 2021-07-16  2:00 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101464

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> ---
Created attachment 51159
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51159&action=edit
A patch

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [Bug target/101464] Replace zveroupper with vpxor
  2021-07-15 18:26 [Bug target/101464] New: Replace zveroupper with vpxor hjl.tools at gmail dot com
  2021-07-16  1:54 ` [Bug target/101464] " crazylht at gmail dot com
  2021-07-16  2:00 ` hjl.tools at gmail dot com
@ 2021-07-16  6:26 ` rguenth at gcc dot gnu.org
  2021-07-16 12:07 ` hjl.tools at gmail dot com
  3 siblings, 0 replies; 5+ messages in thread
From: rguenth at gcc dot gnu.org @ 2021-07-16  6:26 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101464

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
But in the long discussion involving vzeroupper and HTM aborts in glibc memory
functions it came up that vpxor is _not_ properly clearing the upper state and
thus does _not_ remove the penalty that following non-VEX SSE instructions get?

^ permalink raw reply	[flat|nested] 5+ messages in thread

* [Bug target/101464] Replace zveroupper with vpxor
  2021-07-15 18:26 [Bug target/101464] New: Replace zveroupper with vpxor hjl.tools at gmail dot com
                   ` (2 preceding siblings ...)
  2021-07-16  6:26 ` rguenth at gcc dot gnu.org
@ 2021-07-16 12:07 ` hjl.tools at gmail dot com
  3 siblings, 0 replies; 5+ messages in thread
From: hjl.tools at gmail dot com @ 2021-07-16 12:07 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101464

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #4 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Richard Biener from comment #3)
> But in the long discussion involving vzeroupper and HTM aborts in glibc
> memory functions it came up that vpxor is _not_ properly clearing the upper
> state and thus does _not_ remove the penalty that following non-VEX SSE
> instructions get?

From

https://sourceware.org/bugzilla/show_bug.cgi?id=27457

we can't use VPXOR to replace VZEROUPPER.

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2021-07-16 12:07 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-15 18:26 [Bug target/101464] New: Replace zveroupper with vpxor hjl.tools at gmail dot com
2021-07-16  1:54 ` [Bug target/101464] " crazylht at gmail dot com
2021-07-16  2:00 ` hjl.tools at gmail dot com
2021-07-16  6:26 ` rguenth at gcc dot gnu.org
2021-07-16 12:07 ` hjl.tools at gmail dot com

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).