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 v1] x86: Move wcslen SSE2 implementation to multiarch/wcslen-sse2.S
Date: Tue, 12 Jul 2022 13:26:17 -0700 [thread overview]
Message-ID: <CAMe9rOruB6-xdFZwGgrERTApnKou20=r1s4szGRG9hVJc4H+2A@mail.gmail.com> (raw)
In-Reply-To: <20220712192910.351121-8-goldstein.w.n@gmail.com>
On Tue, Jul 12, 2022 at 12:29 PM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> This commit doesn't affect libc.so.6, its just housekeeping to prepare
> for adding explicit ISA level support.
>
> Tested build on x86_64 and x86_32 with/without multiarch.
> ---
> sysdeps/x86_64/multiarch/wcslen-sse2.S | 221 ++++++++++++++++++++++++-
> sysdeps/x86_64/wcslen.S | 216 +-----------------------
> 2 files changed, 218 insertions(+), 219 deletions(-)
>
> diff --git a/sysdeps/x86_64/multiarch/wcslen-sse2.S b/sysdeps/x86_64/multiarch/wcslen-sse2.S
> index 2b3a9efd64..944c3bd9c6 100644
> --- a/sysdeps/x86_64/multiarch/wcslen-sse2.S
> +++ b/sysdeps/x86_64/multiarch/wcslen-sse2.S
> @@ -17,10 +17,221 @@
> <https://www.gnu.org/licenses/>. */
>
> #if IS_IN (libc)
> -# define __wcslen __wcslen_sse2
> -
> -# undef weak_alias
> -# define weak_alias(__wcslen, wcslen)
> +# ifndef WCSLEN
> +# define WCSLEN __wcslen_sse2
> +# endif
> #endif
>
> -#include "../wcslen.S"
> +#include <sysdep.h>
> +
> + .text
> +ENTRY (WCSLEN)
> + cmpl $0, (%rdi)
> + jz L(exit_tail0)
> + cmpl $0, 4(%rdi)
> + jz L(exit_tail1)
> + cmpl $0, 8(%rdi)
> + jz L(exit_tail2)
> + cmpl $0, 12(%rdi)
> + jz L(exit_tail3)
> + cmpl $0, 16(%rdi)
> + jz L(exit_tail4)
> + cmpl $0, 20(%rdi)
> + jz L(exit_tail5)
> + cmpl $0, 24(%rdi)
> + jz L(exit_tail6)
> + cmpl $0, 28(%rdi)
> + jz L(exit_tail7)
> +
> + pxor %xmm0, %xmm0
> +
> + lea 32(%rdi), %rax
> + addq $16, %rdi
> + and $-16, %rax
> +
> + pcmpeqd (%rax), %xmm0
> + pmovmskb %xmm0, %edx
> + pxor %xmm1, %xmm1
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm1
> + pmovmskb %xmm1, %edx
> + pxor %xmm2, %xmm2
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm2
> + pmovmskb %xmm2, %edx
> + pxor %xmm3, %xmm3
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm3
> + pmovmskb %xmm3, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm0
> + pmovmskb %xmm0, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm1
> + pmovmskb %xmm1, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm2
> + pmovmskb %xmm2, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm3
> + pmovmskb %xmm3, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm0
> + pmovmskb %xmm0, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm1
> + pmovmskb %xmm1, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm2
> + pmovmskb %xmm2, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd (%rax), %xmm3
> + pmovmskb %xmm3, %edx
> + addq $16, %rax
> + test %edx, %edx
> + jnz L(exit)
> +
> + and $-0x40, %rax
> +
> + .p2align 4
> +L(aligned_64_loop):
> + movaps (%rax), %xmm0
> + movaps 16(%rax), %xmm1
> + movaps 32(%rax), %xmm2
> + movaps 48(%rax), %xmm6
> +
> + pminub %xmm1, %xmm0
> + pminub %xmm6, %xmm2
> + pminub %xmm0, %xmm2
> + pcmpeqd %xmm3, %xmm2
> + pmovmskb %xmm2, %edx
> + addq $64, %rax
> + test %edx, %edx
> + jz L(aligned_64_loop)
> +
> + pcmpeqd -64(%rax), %xmm3
> + pmovmskb %xmm3, %edx
> + addq $48, %rdi
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd %xmm1, %xmm3
> + pmovmskb %xmm3, %edx
> + addq $-16, %rdi
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd -32(%rax), %xmm3
> + pmovmskb %xmm3, %edx
> + addq $-16, %rdi
> + test %edx, %edx
> + jnz L(exit)
> +
> + pcmpeqd %xmm6, %xmm3
> + pmovmskb %xmm3, %edx
> + addq $-16, %rdi
> + test %edx, %edx
> + jz L(aligned_64_loop)
> +
> + .p2align 4
> +L(exit):
> + sub %rdi, %rax
> + shr $2, %rax
> + test %dl, %dl
> + jz L(exit_high)
> +
> + andl $15, %edx
> + jz L(exit_1)
> + ret
> +
> + /* No align here. Naturally aligned % 16 == 1. */
> +L(exit_high):
> + andl $(15 << 8), %edx
> + jz L(exit_3)
> + add $2, %rax
> + ret
> +
> + .p2align 3
> +L(exit_1):
> + add $1, %rax
> + ret
> +
> + .p2align 3
> +L(exit_3):
> + add $3, %rax
> + ret
> +
> + .p2align 3
> +L(exit_tail0):
> + xorl %eax, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail1):
> + movl $1, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail2):
> + movl $2, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail3):
> + movl $3, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail4):
> + movl $4, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail5):
> + movl $5, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail6):
> + movl $6, %eax
> + ret
> +
> + .p2align 3
> +L(exit_tail7):
> + movl $7, %eax
> + ret
> +
> +END (WCSLEN)
> diff --git a/sysdeps/x86_64/wcslen.S b/sysdeps/x86_64/wcslen.S
> index d641141d75..588a0fbe01 100644
> --- a/sysdeps/x86_64/wcslen.S
> +++ b/sysdeps/x86_64/wcslen.S
> @@ -16,218 +16,6 @@
> License along with the GNU C Library; if not, see
> <https://www.gnu.org/licenses/>. */
>
> -#include <sysdep.h>
> -
> - .text
> -ENTRY (__wcslen)
> - cmpl $0, (%rdi)
> - jz L(exit_tail0)
> - cmpl $0, 4(%rdi)
> - jz L(exit_tail1)
> - cmpl $0, 8(%rdi)
> - jz L(exit_tail2)
> - cmpl $0, 12(%rdi)
> - jz L(exit_tail3)
> - cmpl $0, 16(%rdi)
> - jz L(exit_tail4)
> - cmpl $0, 20(%rdi)
> - jz L(exit_tail5)
> - cmpl $0, 24(%rdi)
> - jz L(exit_tail6)
> - cmpl $0, 28(%rdi)
> - jz L(exit_tail7)
> -
> - pxor %xmm0, %xmm0
> -
> - lea 32(%rdi), %rax
> - addq $16, %rdi
> - and $-16, %rax
> -
> - pcmpeqd (%rax), %xmm0
> - pmovmskb %xmm0, %edx
> - pxor %xmm1, %xmm1
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm1
> - pmovmskb %xmm1, %edx
> - pxor %xmm2, %xmm2
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm2
> - pmovmskb %xmm2, %edx
> - pxor %xmm3, %xmm3
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm3
> - pmovmskb %xmm3, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm0
> - pmovmskb %xmm0, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm1
> - pmovmskb %xmm1, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm2
> - pmovmskb %xmm2, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm3
> - pmovmskb %xmm3, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm0
> - pmovmskb %xmm0, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm1
> - pmovmskb %xmm1, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm2
> - pmovmskb %xmm2, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd (%rax), %xmm3
> - pmovmskb %xmm3, %edx
> - addq $16, %rax
> - test %edx, %edx
> - jnz L(exit)
> -
> - and $-0x40, %rax
> -
> - .p2align 4
> -L(aligned_64_loop):
> - movaps (%rax), %xmm0
> - movaps 16(%rax), %xmm1
> - movaps 32(%rax), %xmm2
> - movaps 48(%rax), %xmm6
> -
> - pminub %xmm1, %xmm0
> - pminub %xmm6, %xmm2
> - pminub %xmm0, %xmm2
> - pcmpeqd %xmm3, %xmm2
> - pmovmskb %xmm2, %edx
> - addq $64, %rax
> - test %edx, %edx
> - jz L(aligned_64_loop)
> -
> - pcmpeqd -64(%rax), %xmm3
> - pmovmskb %xmm3, %edx
> - addq $48, %rdi
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd %xmm1, %xmm3
> - pmovmskb %xmm3, %edx
> - addq $-16, %rdi
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd -32(%rax), %xmm3
> - pmovmskb %xmm3, %edx
> - addq $-16, %rdi
> - test %edx, %edx
> - jnz L(exit)
> -
> - pcmpeqd %xmm6, %xmm3
> - pmovmskb %xmm3, %edx
> - addq $-16, %rdi
> - test %edx, %edx
> - jz L(aligned_64_loop)
> -
> - .p2align 4
> -L(exit):
> - sub %rdi, %rax
> - shr $2, %rax
> - test %dl, %dl
> - jz L(exit_high)
> -
> - andl $15, %edx
> - jz L(exit_1)
> - ret
> -
> - /* No align here. Naturally aligned % 16 == 1. */
> -L(exit_high):
> - andl $(15 << 8), %edx
> - jz L(exit_3)
> - add $2, %rax
> - ret
> -
> - .p2align 3
> -L(exit_1):
> - add $1, %rax
> - ret
> -
> - .p2align 3
> -L(exit_3):
> - add $3, %rax
> - ret
> -
> - .p2align 3
> -L(exit_tail0):
> - xorl %eax, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail1):
> - movl $1, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail2):
> - movl $2, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail3):
> - movl $3, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail4):
> - movl $4, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail5):
> - movl $5, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail6):
> - movl $6, %eax
> - ret
> -
> - .p2align 3
> -L(exit_tail7):
> - movl $7, %eax
> - ret
> -
> -END (__wcslen)
> -
> +#define WCSLEN __wcslen
> +#include "multiarch/wcslen-sse2.S"
> weak_alias(__wcslen, wcslen)
> --
> 2.34.1
>
LGTM.
Thanks.
--
H.J.
next prev parent reply other threads:[~2022-07-12 20:26 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-12 19:29 [PATCH v1] x86: Move strlen SSE2 implementation to multiarch/strlen-sse2.S Noah Goldstein
2022-07-12 19:29 ` [PATCH v1] x86: Move strcpy SSE2 implementation to multiarch/strcpy-sse2.S Noah Goldstein
2022-07-12 23:23 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Move memrchr SSE2 implementation to multiarch/memrchr-sse2.S Noah Goldstein
2022-07-12 22:58 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Move strrchr SSE2 implementation to multiarch/strrchr-sse2.S Noah Goldstein
2022-07-12 22:28 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Move strchr SSE2 implementation to multiarch/strchr-sse2.S Noah Goldstein
2022-07-12 21:27 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Move strcat SSE2 implementation to multiarch/strcat-sse2.S Noah Goldstein
2022-07-12 21:16 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Move wcschr SSE2 implementation to multiarch/wcschr-sse2.S Noah Goldstein
2022-07-12 20:55 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Move wcslen SSE2 implementation to multiarch/wcslen-sse2.S Noah Goldstein
2022-07-12 20:26 ` H.J. Lu [this message]
2022-07-12 19:29 ` [PATCH v1] x86: Remove unneeded rtld-wmemcmp Noah Goldstein
2022-07-12 19:44 ` H.J. Lu
2022-07-12 19:29 ` [PATCH v1] x86: Add missing rtm tests for strcmp family Noah Goldstein
2022-07-12 19:59 ` H.J. Lu
2022-07-12 23:29 ` [PATCH v1] x86: Move strlen SSE2 implementation to multiarch/strlen-sse2.S H.J. Lu
2022-07-13 4:06 ` Noah Goldstein
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='CAMe9rOruB6-xdFZwGgrERTApnKou20=r1s4szGRG9hVJc4H+2A@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).