From: "H.J. Lu" <hjl.tools@gmail.com>
To: Noah Goldstein <goldstein.w.n@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>,
"Carlos O'Donell" <carlos@systemhalted.org>
Subject: Re: [PATCH v2 2/3] x86: Move and slightly improve memset_erms
Date: Wed, 29 Jun 2022 15:18:43 -0700 [thread overview]
Message-ID: <CAMe9rOon=yvCpUbBqtttEKE8gWWk_ikMYNiSxQ+kokHSMzLvqQ@mail.gmail.com> (raw)
In-Reply-To: <20220629221204.1242709-2-goldstein.w.n@gmail.com>
On Wed, Jun 29, 2022 at 3:12 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 | 25 +++++++++++++++
> .../multiarch/memset-vec-unaligned-erms.S | 31 -------------------
> 3 files changed, 26 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..1fce0c9fcc
> --- /dev/null
> +++ b/sysdeps/x86_64/multiarch/memset-erms.S
> @@ -0,0 +1,25 @@
Need copyright notice.
> +#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
>
--
H.J.
next prev parent reply other threads:[~2022-06-29 22:19 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 [this message]
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
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='CAMe9rOon=yvCpUbBqtttEKE8gWWk_ikMYNiSxQ+kokHSMzLvqQ@mail.gmail.com' \
--to=hjl.tools@gmail.com \
--cc=carlos@systemhalted.org \
--cc=goldstein.w.n@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).