Mailing-List: contact cygwin-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-subscribe AT sources DOT redhat DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin/>
List-Post: <mailto:cygwin AT sources DOT redhat DOT com>
List-Help: <mailto:cygwin-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-owner AT sources DOT redhat DOT com
Delivered-To: mailing list cygwin AT sources DOT redhat DOT com
Message-ID: <003b01c1158f$f8281e90$9865fea9@timayum4srqln4>
From: "Tim Prince" <tprince AT computer DOT org>
To: "Francois Colbert" <francois DOT colbert AT videotron DOT ca>, <cygwin AT cygwin DOT com>
References: <000701c1158b$4ab07400$59cdfdcf AT lambda>
Subject: Re: gprof prints incorrect results
Date: Wed, 25 Jul 2001 22:01:01 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4522.1200
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4522.1200


----- Original Message -----
From: "Francois Colbert" <francois DOT colbert AT videotron DOT ca>
To: <cygwin AT cygwin DOT com>
Sent: Wednesday, July 25, 2001 9:27 PM
Subject: gprof prints incorrect results


> Hi.
>
> I tried to use the gprof utility, but I was unable to get useful
> results with the Cygwin version.
>
> I made a simple test program (see below: prof.cpp) and tried to
> profile it on a Linux system (with pre-installed utilities) and on a
> Windows system (with Cygwin utilities: gprof version 2.11.90, gcc
> version 2.95.2-6.).  On each platform, I compiled the program using
> the command line
>
>    gcc -pg prof.cpp -o prof.exe
>
> I executed prof.exe to generate a gmon.out file, and then I ran gprof
> using the command line
>
>    gprof --brief prof.exe gmon.out > prof-win32.txt
>
> under Windows, and
>
>    gprof --brief prof.exe gmon.out > prof-linux.txt
>
> under Linux.
>
>
> The two resulting text files, prof-win32.txt and prof-linux.txt, are
> listed below.  I didn't expect the timing data to be similar, because
> they are the result of the execution of two different executables in
> two different contexts.
>
> Nonetheless, I expected the statistics related to the call graph to be
> exactly the same.  They aren't.  I identified the following anomalies
> in the prof-win32.txt file:
>
> 1- In the flat profile section, number-of-calls stats are missing.
>
> 2- In the call graph section, the last column doesn't show any useful
>    data, since every function either is "spontaneous" or appears to be
>    called only by itself.
>
> 3- In the functions index at the end of the file, the DoSomething
>    function is missing.  It wasn't missing for every execution of the
>    test program, though.
I expect this would be improved if you built the entire run-time library
with -pg.  The cygwin setup doesn't provide for automatically linking a
profiling version of the run-time.  A possible work-around is to build a
separate copy of the run-time libraries and link specifically against
those libraries.  Recursive calling will be "spontaneous" no matter what
you do.



--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/