From: Raphael M Zinsly <rzinsly@linux.ibm.com>
To: Matheus Castanho <msc@linux.ibm.com>, libc-alpha@sourceware.org
Subject: Re: [PATCH v2] powerpc: Add optimized strlen for POWER10
Date: Thu, 22 Apr 2021 12:22:47 -0300 [thread overview]
Message-ID: <3a3e48a5-1083-dcf3-bcc2-32350fa4d20a@linux.ibm.com> (raw)
In-Reply-To: <20210422122911.27758-1-msc@linux.ibm.com>
Hi Matheus, the patch LGTM with some trivial changes.
On 22/04/2021 09:29, Matheus Castanho via Libc-alpha wrote:
> diff --git a/sysdeps/powerpc/powerpc64/le/power10/strlen.S b/sysdeps/powerpc/powerpc64/le/power10/strlen.S
> new file mode 100644
> index 0000000000..7eb37a8f54
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/le/power10/strlen.S
> @@ -0,0 +1,221 @@
> +/* Optimized strlen implementation for POWER10 LE.
> + Copyright (C) 2021 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>
> +
> +#ifndef STRLEN
> +# define STRLEN __strlen
> +# define DEFINE_STRLEN_HIDDEN_DEF 1
> +#endif
> +
> +/* TODO: Replace macros by the actual instructions when minimum binutils becomes
> + >= 2.35. This is used to keep compatibility with older versions. */
> +#define VEXTRACTBM(rt,vrb) \
> + .long(((4)<<(32-6)) \
> + | ((rt)<<(32-11)) \
> + | ((8)<<(32-16)) \
> + | ((vrb)<<(32-21)) \
> + | 1602)
> +
> +#define LXVP(xtp,dq,ra) \
> + .long(((6)<<(32-6)) \
> + | ((((xtp)-32)>>1)<<(32-10)) \
> + | ((1)<<(32-11)) \
> + | ((ra)<<(32-16)) \
> + | dq)
> +
> +#define CHECK16(vreg,offset,addr,label) \
> + lxv vreg+32,offset(addr); \
> + vcmpequb. vreg,vreg,v18; \
> + bne cr6,L(label);
> +
> +/* Load 4 quadwords, merge into one VR for speed and check for NULLs. r6 has #
> + of bytes already checked. */
> +#define CHECK64(offset,addr,label) \
> + li r6,offset; \
> + LXVP(v4+32,offset,addr); \
> + LXVP(v6+32,offset+32,addr); \
> + vminub v14,v4,v5; \
> + vminub v15,v6,v7; \
> + vminub v16,v14,v15; \
> + vcmpequb. v0,v16,v18; \
> + bne cr6,L(label)
> +
> +# define TAIL(vreg,increment) \
nit: the space before define is not needed.
> + vctzlsbb r4,vreg; \
> + subf r3,r3,r5; \
> + addi r4,r4,increment; \
> + add r3,r3,r4; \
> + blr
> +
> +/* Implements the function
> +
> + int [r3] strlen (const void *s [r3])
> +
> + The implementation can load bytes past a matching byte, but only
> + up to the next 64B boundary, so it never crosses a page. */
> +
> +.machine power9
> +
> +ENTRY_TOCLESS (STRLEN, 4)
> + CALL_MCOUNT 1
> +
> + vspltisb v18,0
> + vspltisb v19,-1
> +
> + /* Next 16B-aligned address. Prepare address for L(aligned). */
> + addi r5,r3,16
> + clrrdi r5,r5,4
> +
> + /* Align data and fill bytes not loaded with non matching char. */
> + lvx v0,0,r3
> + lvsr v1,0,r3
> + vperm v0,v19,v0,v1
> +
> + vcmpequb. v6,v0,v18
> + beq cr6,L(aligned)
> +
> + vctzlsbb r3,v6
> + blr
> +
> + /* Test more 112B, 16B at a time. The main loop is optimized for longer
s/112B/176B/
> + strings, so checking the first bytes in 16B chunks benefits a lot
> + small strings. */
> + .p2align 5
> +L(aligned):
--
Raphael Moreira Zinsly
next prev parent reply other threads:[~2021-04-22 15:22 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-04-22 12:29 Matheus Castanho
2021-04-22 15:22 ` Raphael M Zinsly [this message]
2021-04-22 18:20 ` Lucas A. M. Magalhaes
2021-04-22 19:50 ` Matheus Castanho
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=3a3e48a5-1083-dcf3-bcc2-32350fa4d20a@linux.ibm.com \
--to=rzinsly@linux.ibm.com \
--cc=libc-alpha@sourceware.org \
--cc=msc@linux.ibm.com \
/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).