Re: [PATCH] AArch64: Update A64FX memset not to degrade at 16KB

[naohirot--- via Libc-alpha <libc-alpha@sourceware.org>]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="gb2312", Size: 2235 bytes --]

Hi Szabolcs,

I noticed that my email address is not right in the commit log,
would you please fix it to <naohirot@fujitsu.com>?

commit 23777232c23f80809613bdfa329f63aadf992922
Author: Naohiro Tamura via Libc-alpha <libc-alpha@sourceware.org>
Date:   Fri Aug 27 05:03:04 2021 +0000

Thanks.
Naohiro

________________________________________
From: Tamura, Naohiro/Ìï´å Ö±ŽÚ <naohirot@fujitsu.com>
Sent: Monday, 6 September 2021 08:51
To: Szabolcs Nagy
Cc: Wilco Dijkstra; libc-alpha@sourceware.org
Subject: Re: [PATCH] AArch64: Update A64FX memset not to degrade at 16KB

Hi Szabolcs,

Thanks!

Naohiro
________________________________________
From: Szabolcs Nagy <szabolcs.nagy@arm.com>
Sent: Saturday, 4 September 2021 00:01
To: Tamura, Naohiro/Ìï´å Ö±ŽÚ
Cc: Wilco Dijkstra; libc-alpha@sourceware.org
Subject: Re: [PATCH] AArch64: Update A64FX memset not to degrade at 16KB

The 08/27/2021 05:03, Naohiro Tamura via Libc-alpha wrote:
> This patch updates unroll8 code so as not to degrade at the peak
> performance 16KB for both FX1000 and FX700.
>
> Inserted 2 instructions at the beginning of the unroll8 loop,
> cmp and branch, are a workaround that is found heuristically.
>
> Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>

thanks, i committed this now.


> ---
>  sysdeps/aarch64/multiarch/memset_a64fx.S | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/sysdeps/aarch64/multiarch/memset_a64fx.S b/sysdeps/aarch64/multiarch/memset_a64fx.S
> index 7bf759b6a753..f7dfdaace7cf 100644
> --- a/sysdeps/aarch64/multiarch/memset_a64fx.S
> +++ b/sysdeps/aarch64/multiarch/memset_a64fx.S
> @@ -96,7 +96,14 @@ L(vl_agnostic): // VL Agnostic
>  L(unroll8):
>       sub     count, count, tmp1
>       .p2align 4
> -1:   st1b_unroll 0, 7
> +     // The 2 instructions at the beginning of the following loop,
> +     // cmp and branch, are a workaround so as not to degrade at
> +     // the peak performance 16KB.
> +     // It is found heuristically and the branch condition, b.ne,
> +     // is chosen intentionally never to jump.
> +1:   cmp     xzr, xzr
> +     b.ne    1b
> +     st1b_unroll 0, 7
>       add     dst, dst, tmp1
>       subs    count, count, tmp1
>       b.hi    1b
> --
> 2.17.1
>