From: Sunil Pandey <skpgkp2@gmail.com>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: Noah Goldstein <goldstein.w.n@gmail.com>,
GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH v3 2/3] x86: Move and slightly improve memset_erms
Date: Wed, 13 Jul 2022 20:04:36 -0700 [thread overview]
Message-ID: <CAMAf5_fG22FMc-QRV77gqjD_ctOgTAsbg5HVHwQ=Lce==VKy_A@mail.gmail.com> (raw)
In-Reply-To: <CAMe9rOq-URvjhNzo+VdMQnHSUTr-f5d6GBmWDuBgp1zDcFxyHg@mail.gmail.com>
On Wed, Jun 29, 2022 at 4:30 PM H.J. Lu via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> On Wed, Jun 29, 2022 at 4:07 PM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> >
> > Implementation wise:
> > 1. Remove the VZEROUPPER as memset_{impl}_unaligned_erms does not
> > use the L(stosb) label that was previously defined.
> >
> > 2. Don't give the hotpath (fallthrough) to zero size.
> >
> > Code positioning wise:
> >
> > Move memset_{chk}_erms to its own file. Leaving it in between the
> > memset_{impl}_unaligned both adds unnecessary complexity to the
> > file and wastes space in a relatively hot cache section.
> > ---
> > sysdeps/x86_64/multiarch/Makefile | 1 +
> > sysdeps/x86_64/multiarch/memset-erms.S | 44 +++++++++++++++++++
> > .../multiarch/memset-vec-unaligned-erms.S | 31 -------------
> > 3 files changed, 45 insertions(+), 31 deletions(-)
> > create mode 100644 sysdeps/x86_64/multiarch/memset-erms.S
> >
> > diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
> > index 62a4d96fb8..18cea04423 100644
> > --- a/sysdeps/x86_64/multiarch/Makefile
> > +++ b/sysdeps/x86_64/multiarch/Makefile
> > @@ -30,6 +30,7 @@ sysdep_routines += \
> > memset-avx2-unaligned-erms-rtm \
> > memset-avx512-no-vzeroupper \
> > memset-avx512-unaligned-erms \
> > + memset-erms \
> > memset-evex-unaligned-erms \
> > memset-sse2-unaligned-erms \
> > rawmemchr-avx2 \
> > diff --git a/sysdeps/x86_64/multiarch/memset-erms.S b/sysdeps/x86_64/multiarch/memset-erms.S
> > new file mode 100644
> > index 0000000000..e83cccc731
> > --- /dev/null
> > +++ b/sysdeps/x86_64/multiarch/memset-erms.S
> > @@ -0,0 +1,44 @@
> > +/* memset implement with rep stosb
> > + Copyright (C) 2022 Free Software Foundation, Inc.
> > + This file is part of the GNU C Library.
> > +
> > + The GNU C Library is free software; you can redistribute it and/or
> > + modify it under the terms of the GNU Lesser General Public
> > + License as published by the Free Software Foundation; either
> > + version 2.1 of the License, or (at your option) any later version.
> > +
> > + The GNU C Library is distributed in the hope that it will be useful,
> > + but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + Lesser General Public License for more details.
> > +
> > + You should have received a copy of the GNU Lesser General Public
> > + License along with the GNU C Library; if not, see
> > + <https://www.gnu.org/licenses/>. */
> > +
> > +
> > +#include <sysdep.h>
> > +
> > +#if defined USE_MULTIARCH && IS_IN (libc)
> > + .text
> > +ENTRY (__memset_chk_erms)
> > + cmp %RDX_LP, %RCX_LP
> > + jb HIDDEN_JUMPTARGET (__chk_fail)
> > +END (__memset_chk_erms)
> > +
> > +/* Only used to measure performance of REP STOSB. */
> > +ENTRY (__memset_erms)
> > + /* Skip zero length. */
> > + test %RDX_LP, %RDX_LP
> > + jz L(stosb_return_zero)
> > + mov %RDX_LP, %RCX_LP
> > + movzbl %sil, %eax
> > + mov %RDI_LP, %RDX_LP
> > + rep stosb
> > + mov %RDX_LP, %RAX_LP
> > + ret
> > +L(stosb_return_zero):
> > + movq %rdi, %rax
> > + ret
> > +END (__memset_erms)
> > +#endif
> > diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
> > index abc12d9cda..905d0fa464 100644
> > --- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
> > +++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
> > @@ -156,37 +156,6 @@ L(entry_from_wmemset):
> > #if defined USE_MULTIARCH && IS_IN (libc)
> > END (MEMSET_SYMBOL (__memset, unaligned))
> >
> > -# if VEC_SIZE == 16
> > -ENTRY (__memset_chk_erms)
> > - cmp %RDX_LP, %RCX_LP
> > - jb HIDDEN_JUMPTARGET (__chk_fail)
> > -END (__memset_chk_erms)
> > -
> > -/* Only used to measure performance of REP STOSB. */
> > -ENTRY (__memset_erms)
> > - /* Skip zero length. */
> > - test %RDX_LP, %RDX_LP
> > - jnz L(stosb)
> > - movq %rdi, %rax
> > - ret
> > -# else
> > -/* Provide a hidden symbol to debugger. */
> > - .hidden MEMSET_SYMBOL (__memset, erms)
> > -ENTRY (MEMSET_SYMBOL (__memset, erms))
> > -# endif
> > -L(stosb):
> > - mov %RDX_LP, %RCX_LP
> > - movzbl %sil, %eax
> > - mov %RDI_LP, %RDX_LP
> > - rep stosb
> > - mov %RDX_LP, %RAX_LP
> > - VZEROUPPER_RETURN
> > -# if VEC_SIZE == 16
> > -END (__memset_erms)
> > -# else
> > -END (MEMSET_SYMBOL (__memset, erms))
> > -# endif
> > -
> > # if defined SHARED && IS_IN (libc)
> > ENTRY_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned_erms))
> > cmp %RDX_LP, %RCX_LP
> > --
> > 2.34.1
> >
>
> LGTM.
>
> Thanks.
>
> --
> H.J.
I would like to backport this patch to release branches.
Any comments or objections?
--Sunil
next prev parent reply other threads:[~2022-07-14 3:05 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-06-28 15:27 [PATCH v1 1/3] x86: Add definition for __wmemset_chk AVX2 RTM in ifunc impl list Noah Goldstein
2022-06-28 15:27 ` [PATCH v1 2/3] x86: Move and slightly improve memset_erms Noah Goldstein
2022-06-29 19:25 ` H.J. Lu
2022-06-29 19:32 ` Noah Goldstein
2022-06-29 22:11 ` Noah Goldstein
2022-06-28 15:27 ` [PATCH v1 3/3] x86: Add support for building {w}memset{_chk} with explicit ISA level Noah Goldstein
2022-06-29 19:30 ` H.J. Lu
2022-06-29 22:12 ` Noah Goldstein
2022-06-29 19:21 ` [PATCH v1 1/3] x86: Add definition for __wmemset_chk AVX2 RTM in ifunc impl list H.J. Lu
2022-06-29 23:08 ` Noah Goldstein
2022-07-14 3:03 ` Sunil Pandey
2022-06-29 22:12 ` [PATCH v2 " Noah Goldstein
2022-06-29 22:12 ` [PATCH v2 2/3] x86: Move and slightly improve memset_erms Noah Goldstein
2022-06-29 22:18 ` H.J. Lu
2022-06-29 23:09 ` Noah Goldstein
2022-06-29 22:12 ` [PATCH v2 3/3] x86: Add support for building {w}memset{_chk} with explicit ISA level Noah Goldstein
2022-06-29 22:53 ` H.J. Lu
2022-06-29 23:07 ` [PATCH v3 1/3] x86: Add definition for __wmemset_chk AVX2 RTM in ifunc impl list Noah Goldstein
2022-06-29 23:07 ` [PATCH v3 2/3] x86: Move and slightly improve memset_erms Noah Goldstein
2022-06-29 23:30 ` H.J. Lu
2022-07-14 3:04 ` Sunil Pandey [this message]
2022-06-29 23:07 ` [PATCH v3 3/3] x86: Add support for building {w}memset{_chk} with explicit ISA level Noah Goldstein
2022-06-29 23:28 ` H.J. Lu
2022-07-01 22:45 ` H.J. Lu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAMAf5_fG22FMc-QRV77gqjD_ctOgTAsbg5HVHwQ=Lce==VKy_A@mail.gmail.com' \
--to=skpgkp2@gmail.com \
--cc=goldstein.w.n@gmail.com \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).