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
Date: Sun, 10 Jun 2001 23:51:31 -0400
From: Christopher Faylor <cgf AT redhat DOT com>
To: cygwin-developers AT cygwin DOT com
Subject: Re: Recent changes to path_conv::check -- ouch
Message-ID: <20010610235131.A10749@redhat.com>
Reply-To: cygwin-developers AT cygwin DOT com
Mail-Followup-To: cygwin-developers AT cygwin DOT com
References: <20010610201750 DOT A8503 AT redhat DOT com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.3.11i
In-Reply-To: <20010610201750.A8503@redhat.com>; from cgf@redhat.com on Sun, Jun 10, 2001 at 08:17:50PM -0400

[following up on myself]
Kazuhiro's patch has forced me to look into this function a little
more.  It looks like we're just calling rootdir repeatedly rather than
invoking the OS.  I guess that is ok.  I'd rather not have even that
perfomance penalty but I don't know if it can be avoided.

cgf

On Sun, Jun 10, 2001 at 08:17:50PM -0400, Christopher Faylor wrote:
>Egor,
>I guess I missed this before but you added a potentially expensive
>operation to the middle of the path_conv check loop by calling
>update_fs_info repeatedly.  This looks like we will suffer a performance
>hit in the non-ntea case for the privilege of a performance gain for the
>use of ntea for symlinks.  I don't think that's acceptable.
>
>It also looks like even in the ideal case, we're guaranteed to call
>update_fs_info twice on the same path.
>
>Also, you seemed to have changed the name of some recently added
>elements to the path_conv class.  vol_flags became fs_flags and
>vol_serial became fs_serial.
>
>path_conv is probably the most frequently called function in cygwin so
>we've tried very hard to keep the code flow as optimized as possible by
>not calling the OS too often.  Maybe I'm wrong, but I think you've
>probably added a performance hit with your recent changes.
>
>Have you run any before/after performance checks on your code?