Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-developers-subscribe AT sources DOT redhat DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin-developers/>
List-Post: <mailto:cygwin-developers AT sources DOT redhat DOT com>
List-Help: <mailto:cygwin-developers-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-developers-owner AT sources DOT redhat DOT com
Delivered-To: mailing list cygwin-developers AT sources DOT redhat DOT com
Subject: Re: Quick testfeedback...
From: Robert Collins <robert DOT collins AT itdomain DOT com DOT au>
To: Jason Tishler <jason AT tishler DOT net>
Cc: cygwin-developers AT cygwin DOT com
In-Reply-To: <20010911125735.D1752@dothill.com>
References: <20010911125735 DOT D1752 AT dothill DOT com>
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
X-Mailer: Evolution/0.13 (Preview Release)
Date: 12 Sep 2001 09:20:11 +1000
Message-Id: <1000250412.30377.1.camel@lifelesswks>
Mime-Version: 1.0
X-OriginalArrivalTime: 11 Sep 2001 23:07:02.0355 (UTC) FILETIME=[77841A30:01C13B16]

On Wed, 2001-09-12 at 02:57, Jason Tishler wrote:
> Rob,
> 
> On Tue, Sep 11, 2001 at 11:10:45AM -0400, Jason Tishler wrote:
> > On Tue, Sep 11, 2001 at 10:00:11PM +1000, Robert Collins wrote:
> > > I have tested this out on win95 for regressions, but not on NT
> > > unfortunately... If some kind NT/2k user could test this I would be very
> > > appreciative.
> > 
> > I ran your test suite with the latest CVS and then again after applying
> > your patch.
> 
> I also ran the Python regression test for both cases:
> 
>     threads (CVS):   4:55  # latest CVS
>     threads (patch): 3:22  # latest CVS + patch
>     no threads:      2:26  # latest CVS + patch (but shouldn't matter)
>     extra (patch):   0:47  # four extra tests only run in the threaded case
> 
> The above indicates that the use of critical sections instead of mutuxes
> is a big win -- at least for Python.  Additionally, the threaded case
> using critical sections performs almost identically to the non threaded
> one, if one accounts for the extra tests:
> 
>     threads without extra: 2:35  # 3:22 - 0:47
>     no threads:            2:26
> 
> Hence, the threaded case only incurs about a 6% performance penalty.
> Not bad!
> 
> Would you be willing to check in your patch?

Yes, once I get feedback on the earlier reported broadcast fault. A
broken broadcast() will break nearly every pthread app... a bad thing.

Rob