From: Szabolcs Nagy <Szabolcs.Nagy@arm.com>
To: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
Cc: 'GNU C Library' <libc-alpha@sourceware.org>
Subject: Re: [PATCH] AArch64: Improve strlen_asimd
Date: Fri, 13 Jan 2023 12:25:35 +0000 [thread overview]
Message-ID: <Y8FNvxktEcu8WVwc@arm.com> (raw)
In-Reply-To: <PAWPR08MB8982BE5A8C1635DBCB89700D83FD9@PAWPR08MB8982.eurprd08.prod.outlook.com>
The 01/12/2023 15:51, Wilco Dijkstra wrote:
> Use shrn for the mask, merge tst+bne into cbnz, and tweak code alignment.
> Performance improves slightly as a result. Passes regress.
>
I prefer to commit this and the other string function optimization
patches and not delay to next release so start using and widely
testing them sooner (we can fix and backport perf regressions).
please commit it, thanks.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
> ---
>
> diff --git a/sysdeps/aarch64/multiarch/strlen_asimd.S b/sysdeps/aarch64/multiarch/strlen_asimd.S
> index ca6ab96ecf2de45def79539facd8e0b86f4edc95..490439491d19c3f14b0228f42248bc8aa6e9e8bd 100644
> --- a/sysdeps/aarch64/multiarch/strlen_asimd.S
> +++ b/sysdeps/aarch64/multiarch/strlen_asimd.S
> @@ -48,6 +48,7 @@
> #define tmp x2
> #define tmpw w2
> #define synd x3
> +#define syndw w3
> #define shift x4
>
> /* For the first 32 bytes, NUL detection works on the principle that
> @@ -87,7 +88,6 @@
>
> ENTRY (__strlen_asimd)
> PTR_ARG (0)
> -
> and tmp1, srcin, MIN_PAGE_SIZE - 1
> cmp tmp1, MIN_PAGE_SIZE - 32
> b.hi L(page_cross)
> @@ -123,7 +123,6 @@ ENTRY (__strlen_asimd)
> add len, len, tmp1, lsr 3
> ret
>
> - .p2align 3
> /* Look for a NUL byte at offset 16..31 in the string. */
> L(bytes16_31):
> ldp data1, data2, [srcin, 16]
> @@ -151,6 +150,7 @@ L(bytes16_31):
> add len, len, tmp1, lsr 3
> ret
>
> + nop
> L(loop_entry):
> bic src, srcin, 31
>
> @@ -166,18 +166,12 @@ L(loop):
> /* Low 32 bits of synd are non-zero if a NUL was found in datav1. */
> cmeq maskv.16b, datav1.16b, 0
> sub len, src, srcin
> - tst synd, 0xffffffff
> - b.ne 1f
> + cbnz syndw, 1f
> cmeq maskv.16b, datav2.16b, 0
> add len, len, 16
> 1:
> /* Generate a bitmask and compute correct byte offset. */
> -#ifdef __AARCH64EB__
> - bic maskv.8h, 0xf0
> -#else
> - bic maskv.8h, 0x0f, lsl 8
> -#endif
> - umaxp maskv.16b, maskv.16b, maskv.16b
> + shrn maskv.8b, maskv.8h, 4
> fmov synd, maskd
> #ifndef __AARCH64EB__
> rbit synd, synd
> @@ -186,8 +180,6 @@ L(loop):
> add len, len, tmp, lsr 2
> ret
>
> - .p2align 4
> -
> L(page_cross):
> bic src, srcin, 31
> mov tmpw, 0x0c03
next prev parent reply other threads:[~2023-01-13 12:26 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-01-12 15:51 Wilco Dijkstra
2023-01-13 12:25 ` Szabolcs Nagy [this message]
2023-01-16 22:16 ` Carlos O'Donell
2023-01-17 16:37 ` Wilco Dijkstra
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=Y8FNvxktEcu8WVwc@arm.com \
--to=szabolcs.nagy@arm.com \
--cc=Wilco.Dijkstra@arm.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).