This is the mail archive of the
mailing list for the glibc project.
Re: [PATCH 5/8] Move i486/pthread_spin_trylock.S to pthread_spin_trylock.S
- From: Torvald Riegel <triegel at redhat dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Joseph Myers <joseph at codesourcery dot com>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 26 Aug 2015 22:56:56 +0200
- Subject: Re: [PATCH 5/8] Move i486/pthread_spin_trylock.S to pthread_spin_trylock.S
- Authentication-results: sourceware.org; auth=none
- References: <20150826173721 dot GF21261 at intel dot com> <alpine dot DEB dot 2 dot 10 dot 1508261800370 dot 13146 at digraph dot polyomino dot org dot uk> <CAMe9rOp7zoMCsMLY2WoEucLKkrR6oJZg9XDJa2S-OJmfA08C8A at mail dot gmail dot com>
On Wed, 2015-08-26 at 12:33 -0700, H.J. Lu wrote:
> On Wed, Aug 26, 2015 at 11:02 AM, Joseph Myers <firstname.lastname@example.org> wrote:
> > On Wed, 26 Aug 2015, H.J. Lu wrote:
> >> Since glibc doesn't support i386 any more, we can move
> >> i486/pthread_spin_trylock.S to pthread_spin_trylock.S
> > The i686 version wasn't just a #include - you're losing the HAVE_CMOV
> > define in this patch.
> > This illustrates the use of testing these patches by making sure that
> > stripped installed shared libraries (for each of i486, i586 and i686) are
> > identical before and after the patch series.
> Here is the updated patch. There are no changes in generated
> shared libraries for i486, i586 amd i686 targets.
> OK for master?
Could you instead please try to replace the custom asm implementation
with a C implementation, preferably the generic one in
nptl/pthread_spin_trylock.c? And, if necessary, and improve the latter
if there is a significant performance difference for uncontended locks?