From chemistry-request@server.ccl.net Sat Jun 15 20:10:06 2002
Received: from mail.staff.site.ntu.edu.au ([138.80.69.54])
	by server.ccl.net (8.11.6/8.11.0) with ESMTP id g5G0A3515162
	for <CHEMISTRY@ccl.net>; Sat, 15 Jun 2002 20:10:04 -0400
Received: from mail.staff.site.ntu.edu.au ([138.80.49.23]) by mail.staff.site.ntu.edu.au with Microsoft SMTPSVC(5.0.2195.4905);
	 Sun, 16 Jun 2002 09:40:01 +0930
Received: by mail.staff.site.ntu.edu.au (sSMTP sendmail emulation); Sun, 16 Jun 2002 09:42:13 +0930
Date: Sun, 16 Jun 2002 09:42:13 +0930
From: Brian Salter-Duke <b_duke@octa4.net.au>
To: Comp Chem List <CHEMISTRY@ccl.net>
Subject: GAMESS and Intel Fortran Compiler
Message-ID: <20020616001213.GA916@BSALTERDUKE1>
Mail-Followup-To: Comp Chem List <CHEMISTRY@ccl.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.3.99i
X-OriginalArrivalTime: 16 Jun 2002 00:10:02.0198 (UTC) FILETIME=[28E71F60:01C214CA]

Has any one tried to compile GAMESS with the Intel Fortran Compiler.

I am still with the 26 October 2000 GAMESS version. Everything compiled
fine with:-

ifc -O2 -w $MODULE.f

The -O2 is in fact not required. The -w suppresses warning, mainly about
features being dropped in F95. 

ddisoc.c needs editing to call the C routines with one underscore not
two in the "$ifdef LINUX" block as suggested for Portland Compiler. lked
needs "set LDR='ifc' and 'set LDOPTS=-Vaxlib'" and everything links. 

Everything runs except exam19 which hangs quite late in the run. It is
difficult to judge on the test runs, but it appears to be faster than a
g77 compiler.

The only problem is that "FLUSH--FAILED:: Bad address" appears several
times in the output particularly at each step of the SCF, but the
results are fine and there is no garbage in the PUNCH file as suggested
in unport.src for an AIX bad flush. Should we just take flush out for
IFC and Linux? 

Any thoughts on using other parameters on the compile to improve
performance?

Incidentally on this pentium IV machine running Redhat 7.2, several test
jobs with the g77 compile fail with:-

ddikick: Trapped SIGCHLD: Unexpected termination of a child process.
etc

These are exam07, exam11, exam12 and exam30. exam19 works fine. Does anyone
have an explanation of this g77 behaviour?

Cheers, Brian.

-- 
            Brian Salter-Duke (Brian Duke) b_duke@octa4.net.au  
  Honorary Fellow in Chemistry, NT University, Darwin, NT 0909, Australia.
             Post: Box 1028, Humpty Doo, NT 0836, Australia.
Phone 08-89881600.  Fax 08-89881302.  http://www.octa4.net.au/linden/brian/