This is the mail archive of the
glibc-bugs@sourceware.org
mailing list for the glibc project.
Re: [Bug nptl/16630] Use SYSENTER for pthread_cond_broadcast/signal() (i.e. fix "FIXME: Ingo" issue)
- From: OndÅej BÃlka <neleai at seznam dot cz>
- To: bugdal at aerifal dot cx <sourceware-bugzilla at sourceware dot org>
- Cc: glibc-bugs at sourceware dot org
- Date: Tue, 25 Feb 2014 17:11:12 +0100
- Subject: Re: [Bug nptl/16630] Use SYSENTER for pthread_cond_broadcast/signal() (i.e. fix "FIXME: Ingo" issue)
- Authentication-results: sourceware.org; auth=none
- References: <bug-16630-131 at http dot sourceware dot org/bugzilla/> <bug-16630-131-tWjO7oaks6 at http dot sourceware dot org/bugzilla/>
On Tue, Feb 25, 2014 at 05:47:16AM +0000, bugdal at aerifal dot cx wrote:
> http://sourceware.org/bugzilla/show_bug.cgi?id=16630
>
> --- Comment #3 from Rich Felker <bugdal at aerifal dot cx> ---
> On Tue, Feb 25, 2014 at 04:51:18AM +0000, cjones.bugs at gmail dot com wrote:
> > Do you happen to know if that's filed already? I'm curious what it would be
> > replaced with.
>
> There is portable C code for all of these functions already and a
> strong sentiment from some users and developers that the asm is
> unnecessary, error-prone, and lags behind the C in getting fixes and
> improvements. See the related thread on the libc-alpha mailing list:
>
> [RFC][BZ #16549, #16410] Remove pthread_(cond)wait assembly implementations?
>
> Basically I think if it could be demonstrated that the C performs just
> as well (or within a margin of difference that's not significant), the
> asm could be removed.
>
Actually c is around 5000 cycles faster. My guess is that its because
assembly does extra syscall which has bigger impact than
microoptimizations, I did not trace that yet.