From chemistry-request %-% at %-% ccl.net Mon Jun 9 13:35:34 2003 Received: from kff1.uchicago.edu (kff1.uchicago.edu [128.135.160.87]) by server.ccl.net (8.12.8/8.12.8) with ESMTP id h59HZYkr001168 for ; Mon, 9 Jun 2003 13:35:34 -0400 Received: from kff1.uchicago.edu (localhost [127.0.0.1]) by kff1.uchicago.edu (8.12.3/8.12.3/Debian-6.4) with ESMTP id h59HZYfA003482 for ; Mon, 9 Jun 2003 12:35:34 -0500 Received: from localhost (cmtaylor <-at-> localhost) by kff1.uchicago.edu (8.12.3/8.12.3/Debian-6.4) with ESMTP id h59HZWbS003478 for ; Mon, 9 Jun 2003 12:35:34 -0500 Date: Mon, 9 Jun 2003 12:35:32 -0500 (CDT) From: caroline taylor To: CCL list Subject: CCL: ifc -O0 failure Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CCLers, We have run into an anomoly using the ifc compiler with optimization. When builing the code we've been using to benchmark, we've used the various optimization flags, with the following results (numerical result and runtime): -O0: 6.710886E+07 (21.960u 0.020s 0:22.19 99.0%) -O1: 1.718282E+08 (17.560u 0.010s 0:17.76 98.9%) -O2: 1.718282E+08 (17.550u 0.010s 0:17.74 98.9%) -O3: 1.718282E+08 (17.580u 0.000s 0:17.75 99.0%) -O4: 1.718282E+08 (17.570u 0.000s 0:17.75 98.9%) The 0 level optimization is clearly inconsistent, and far larger than round off. This has been checked on multiple machines (all linux). Has anyone run into this before? Is there a known problem with this level, or is something else going on? The very simple code we've used for checking the compiler is below: PROGRAM tyme IMPLICIT NONE REAL :: sum INTEGER :: i INTEGER, PARAMETER :: imax = 100000000 sum = 0.0 DO i = 1, imax sum = sum + DEXP( DBLE(i)/DBLE(imax) ) END DO write(*,*) sum, i END PROGRAM tyme Thanks for your time! Caroline --------------------------------------------- Dr. Caroline M. Taylor Department of Chemistry University of Chicago, James Franck Institute 5640 S. Ellis Avenue Chicago, Illinois 60637 (773) 702-7223 ---------------------------------------------