This is the mail archive of the
mailing list for the glibc project.
Re: [PATCH, AArch64] Add optimized strchr.
- From: Richard Earnshaw <rearnsha at arm dot com>
- To: Will Newton <will dot newton at linaro dot org>
- Cc: libc-alpha <libc-alpha at sourceware dot org>
- Date: Tue, 17 Jun 2014 11:45:40 +0100
- Subject: Re: [PATCH, AArch64] Add optimized strchr.
- Authentication-results: sourceware.org; auth=none
- References: <53983583 dot 3040606 at arm dot com> <CANu=DmhQ8KkFbiKO-2HsUaMbdPjaGJTtHKvtJ5nRPr=uaMDwsQ at mail dot gmail dot com>
On 17/06/14 11:37, Will Newton wrote:
> On 11 June 2014 11:54, Richard Earnshaw <email@example.com> wrote:
> Hi Richard,
>> Implementation of strchr for AArch64. Speedups taken from micro-bench
>> show the improvements relative to the standard C code.
>> The use of LD1 means we have identical code for both big- and
>> little-endian systems.
>> A speedup (which is taken as the time to run the reference version
>> divided by the time to run the new version) >1 implies an improvement
>> (<1 a regression). The table only shows one sample that regresses and
>> that's by marginally less than 1%. The geomean for all the samples is
>> Measurements are on Cortex-A57.
>> <date> Richard Earnshaw <firstname.lastname@example.org>
>> * sysdeps/aarch64/strchr.S: New file.
> Some of the indentation around the defines of tmp1 and vrepchr seems a
> bit off, but other than that this looks OK.
vrepchr is fine (it's the + at the beginning of the line in patch format
that causes the tab to go one column more).
tmp1 has a tab rather than a space between the define and tmp1, that
should preferably be replaced with a normal space but it's not a real