This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 2/2][BZ #16640] Remove strtok assembly implementation.
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>, Mike Frysinger <vapier at gentoo dot org>
- Cc: OndÅej BÃlka <neleai at seznam dot cz>, libc-alpha at sourceware dot org
- Date: Thu, 13 Mar 2014 10:39:07 -0400
- Subject: Re: [PATCH 2/2][BZ #16640] Remove strtok assembly implementation.
- Authentication-results: sourceware.org; auth=none
- References: <20140227123238 dot GA26291 at domone dot podge> <20140227124206 dot GA26474 at domone dot podge> <5318A03D dot 3000705 at redhat dot com> <20140306163241 dot GA11843 at domone dot podge> <5318B58B dot 5040704 at redhat dot com> <20140306205212 dot GB11843 at domone dot podge> <53192422 dot 2050101 at redhat dot com> <20140310212923 dot GG6407 at domone dot podge> <53214E1B dot 6040601 at redhat dot com> <Pine dot LNX dot 4 dot 64 dot 1403131251210 dot 4888 at digraph dot polyomino dot org dot uk>
On 03/13/2014 08:53 AM, Joseph S. Myers wrote:
> On Thu, 13 Mar 2014, Carlos O'Donell wrote:
>
>> Are you saying then that the resulting code generated with gcc 4.4
>> is approximately the same performance as that generated with a
>> newer gcc, both of which are 2-4x faster than the assembly
>> implementation?
>>
>> If that's the case then please commit your patch.
>
> Carlos, you appear to have omitted an important qualification here. As
> Mike noted in
> <https://sourceware.org/ml/libc-alpha/2014-03/msg00221.html>, there needs
> to be a proper justification for *machines without SSE4.2*. Without such
> a justification showing this change does not regress performance on such
> hardware, we don't have consensus for this patch.
Joseph,
Thanks, I'd missed Mike's comments, it looks like threading didn't work
correctly for that comment.
Ondrej,
Are you able to test on a machine without SSE4.2 and describe the
performance differences?
If not, are you able to wire this up as an IFUNC for selection based
on SSSE4.2 availability?
Cheers,
Carlos.