This is the mail archive of the
libc-ports@sources.redhat.com
mailing list for the libc-ports project.
Re: [PATCH] ARM: Add Cortex-A15 optimized NEON and VFP memcpy routines, with IFUNC.
- From: Will Newton <will dot newton at linaro dot org>
- To: Måns Rullgård <mans at mansr dot com>
- Cc: libc-ports at sourceware dot org, Patch Tracking <patches at linaro dot org>
- Date: Mon, 15 Apr 2013 11:49:36 +0100
- Subject: Re: [PATCH] ARM: Add Cortex-A15 optimized NEON and VFP memcpy routines, with IFUNC.
- References: <516BCEE5 dot 9070809 at linaro dot org> <yw1x8v4k6rcc dot fsf at unicorn dot mansr dot com> <CANu=DmjJUZ319+7_M8cyxMga_rYxbGb_QSs87Q29JBdkKX_97g at mail dot gmail dot com> <yw1x4nf86pi8 dot fsf at unicorn dot mansr dot com>
On 15 April 2013 11:45, Måns Rullgård <mans@mansr.com> wrote:
> Will Newton <will.newton@linaro.org> writes:
>
>> On 15 April 2013 11:06, Måns Rullgård <mans@mansr.com> wrote:
>>
>> Hi Måns,
>>
>>>> Add a high performance memcpy routine optimized for Cortex-A15 with
>>>> variants for use in the presence of NEON and VFP hardware, selected
>>>> at runtime using indirect function support.
>>>
>>> How does this perform on Cortex-A9?
>>
>> The code is also faster on A9 although the gains are not quite as
>> pronounced.
>
> That should be expected. As long as it's not slower, it's all well.
>
> Out of interest, did you compare this memcpy to any other NEON versions
> floating about, such as the one in bionic?
It's been tested against all the versions of memcpy in the
cortex-strings package on launchpad, which includes a version from
bionic, although I don't believe it uses NEON. AFAIK the latest code
in bionic for A15 is an ARM Ltd. copyright version similar to the one
in newlib, which this code is faster than in almost all cases.
--
Will Newton
Toolchain Working Group, Linaro