From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <hjl.tools@gmail.com>
Received: from mail-oi1-x235.google.com (mail-oi1-x235.google.com
 [IPv6:2607:f8b0:4864:20::235])
 by sourceware.org (Postfix) with ESMTPS id A17F83856254
 for <libc-alpha@sourceware.org>; Tue, 12 Jul 2022 20:26:54 +0000 (GMT)
DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org A17F83856254
Received: by mail-oi1-x235.google.com with SMTP id n206so3022592oia.6
 for <libc-alpha@sourceware.org>; Tue, 12 Jul 2022 13:26:54 -0700 (PDT)
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;
 d=1e100.net; s=20210112;
 h=x-gm-message-state:mime-version:references:in-reply-to:from:date
 :message-id:subject:to:cc;
 bh=KRxuEjF2qz2XaoGZioNrSbWh5eJCuNe0kXFdVZpNb28=;
 b=AISYA78Yx0l6IB4xUh3Cf68IPihh5CIDcT4JwCv3o8eM66BctvMO1o7YZq4W4asx8x
 jzeTSEuVjODVDxojPAvY60XT6prvQ9zjRzIPnoPvSQa4QSbmY6JGcDkxu1k3Mpf9NkRX
 /8k3XTc0aFRZS0nRi1UnPIsSd+9xJqw1qy7ghF+22UanT6m5nO9FnoIDEFuId/IKXTUD
 GWFyU6uDF7ApJ0X09PPBB47agHLJeUl2KI79KkbZxmCURPZrE+v9/UtwegljHi3aqi2Y
 ol6GepGSNHag18/vXKJXXnSiuajpWTPrChoRrHSAf7aWEbDNPiBDX1KPvUsc2kb8YpDu
 NBRg==
X-Gm-Message-State: AJIora8+Rk0Ul6b3s1aDdWmgxZ7f0k3+q2TjTG6vMWH8uTMRkys8RM+3
 l8ohqC6wqMp4wSgFtjx3JmQJ8G9ye1oxSpWQsoI=
X-Google-Smtp-Source: AGRyM1ujff1bedlGG6V6IAf2nVRrA69VsNr6F6L3i2W9rtTPhTO+DpykvH3BzuLUw5v0xyUnPN7ntxndbST/Psjd0xs=
X-Received: by 2002:a05:6808:14c3:b0:337:a1dc:89d5 with SMTP id
 f3-20020a05680814c300b00337a1dc89d5mr2887061oiw.201.1657657613876; Tue, 12
 Jul 2022 13:26:53 -0700 (PDT)
MIME-Version: 1.0
References: <20220712192910.351121-1-goldstein.w.n@gmail.com>
 <20220712192910.351121-8-goldstein.w.n@gmail.com>
In-Reply-To: <20220712192910.351121-8-goldstein.w.n@gmail.com>
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Tue, 12 Jul 2022 13:26:17 -0700
Message-ID: <CAMe9rOruB6-xdFZwGgrERTApnKou20=r1s4szGRG9hVJc4H+2A@mail.gmail.com>
Subject: Re: [PATCH v1] x86: Move wcslen SSE2 implementation to
 multiarch/wcslen-sse2.S
To: Noah Goldstein <goldstein.w.n@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>,
 "Carlos O'Donell" <carlos@systemhalted.org>
Content-Type: text/plain; charset="UTF-8"
X-Spam-Status: No, score=-3024.8 required=5.0 tests=BAYES_00, DKIM_SIGNED,
 DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, FREEMAIL_FROM, GIT_PATCH_0,
 KAM_SHORT, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP,
 T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.6
X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on
 server2.sourceware.org
X-BeenThere: libc-alpha@sourceware.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Libc-alpha mailing list <libc-alpha.sourceware.org>
List-Unsubscribe: <https://sourceware.org/mailman/options/libc-alpha>,
 <mailto:libc-alpha-request@sourceware.org?subject=unsubscribe>
List-Archive: <https://sourceware.org/pipermail/libc-alpha/>
List-Post: <mailto:libc-alpha@sourceware.org>
List-Help: <mailto:libc-alpha-request@sourceware.org?subject=help>
List-Subscribe: <https://sourceware.org/mailman/listinfo/libc-alpha>,
 <mailto:libc-alpha-request@sourceware.org?subject=subscribe>
X-List-Received-Date: Tue, 12 Jul 2022 20:26:56 -0000

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.