You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Brian Akins <ba...@web.turner.com> on 2003/12/02 14:40:05 UTC
[PATCH 25137] atomics in worker mpm
Backported from 2.1. Stable for me in various loads.
--
Brian Akins
Senior Systems Engineer
CNN Internet Technologies
Re: [PATCH 25137] atomics in worker mpm
Posted by Brian Akins <ba...@web.turner.com>.
Aaron Bannert wrote:
> Cool! What OS/arch are you using? Also, any idea how well it performs
> compared to before the patch?
Currently Linux on i386 (with non-portable-atomics). I have not noticed
any increase in performance, but I can, now, easily have 4096 active
clients, whereas things got "squirrelly" before. YMMV.
--
Brian Akins
Senior Systems Engineer
CNN Internet Technologies
Re: [PATCH 25137] atomics in worker mpm
Posted by Aaron Bannert <aa...@clove.org>.
On Tue, Dec 02, 2003 at 08:40:05AM -0500, Brian Akins wrote:
> Backported from 2.1. Stable for me in various loads.
Cool! What OS/arch are you using? Also, any idea how well it performs
compared to before the patch?
-aaron
Re: [PATCH 25137] atomics in worker mpm
Posted by Joe Orton <jo...@redhat.com>.
On Thu, Dec 11, 2003 at 12:47:37PM -0800, Aaron Bannert wrote:
> On Thu, Dec 11, 2003 at 08:39:27AM -0500, Brian Akins wrote:
> > I wonder if this
> > >binary would run on an older processor (running a modern version of linux).
> >
> > AFAIK, yes. It's standard x86 assembly.
> >
> >
> > All: Please correct me if I am wrong. I'm sure you will ;)
>
> I'm no x86 asm expert, so maybe someone else can comment on the
> portability of this code.
Me neither, but AFAIK the x86 atomics are supposed to work on i486 and
above: the i386 doesn't have cmpxchgl.
Re: [PATCH 25137] atomics in worker mpm
Posted by Aaron Bannert <aa...@clove.org>.
On Thu, Dec 11, 2003 at 08:39:27AM -0500, Brian Akins wrote:
> I wonder if this
> >binary would run on an older processor (running a modern version of linux).
>
> AFAIK, yes. It's standard x86 assembly.
>
>
> All: Please correct me if I am wrong. I'm sure you will ;)
I'm no x86 asm expert, so maybe someone else can comment on the
portability of this code.
-aaron
Re: [PATCH 25137] atomics in worker mpm
Posted by Brian Akins <ba...@web.turner.com>.
Aaron Bannert wrote:
> Does it use this atomics implementation by default?
AFAIK, no. By default, (ie, without non-portable-atomics), apr_atomic
uses mutexes.
I wonder if this
> binary would run on an older processor (running a modern version of linux).
AFAIK, yes. It's standard x86 assembly.
All: Please correct me if I am wrong. I'm sure you will ;)
--
Brian Akins
Senior Systems Engineer
CNN Internet Technologies
Re: [PATCH 25137] atomics in worker mpm
Posted by Aaron Bannert <aa...@clove.org>.
On Tue, Dec 09, 2003 at 03:24:15PM -0500, Brian Akins wrote:
> I was testing on x86 Linux which appears to do the apr_atomics in assembly.
Does it use this atomics implementation by default? I wonder if this
binary would run on an older processor (running a modern version of linux).
-aaron
Re: [PATCH 25137] atomics in worker mpm
Posted by Brian Akins <ba...@web.turner.com>.
Jeff Trawick wrote:
> Brian Akins wrote:
>
>> Backported from 2.1. Stable for me in various loads.
>
>
> without the non-portable atomic code available and enabled in APR, this
> is going to hurt performance, right? (more mutex operations performed
> in mainline path for the unlucky who use the non-portable atomics???)
> is it noticeable?
I was testing on x86 Linux which appears to do the apr_atomics in assembly.
--
Brian Akins
Senior Systems Engineer
CNN Internet Technologies
Re: [PATCH 25137] atomics in worker mpm
Posted by Jeff Trawick <tr...@attglobal.net>.
Brian Akins wrote:
> Backported from 2.1. Stable for me in various loads.
without the non-portable atomic code available and enabled in APR, this is
going to hurt performance, right? (more mutex operations performed in mainline
path for the unlucky who use the non-portable atomics???) is it noticeable?
the atomics in the pending apr-1.0-dev still leave much to be desired :( no
compare-and-swap interface even on common platforms such as Solaris/Sparc and
AIX... though Greg got c-a-s working on PPC yesterday; I'd expect AIX to be
supported soon...