True 64 vs. Linux code on an Alpha

[Craig Tierney]

Which version of the Compaq Compilers for Linux are you using?
One of the users here has reported significant speed up using
the v1.1 Beta Fortran compiler.

Craig


On Mon, Apr 09, 2001 at 12:24:08PM -0400, Josip Loncaric wrote:
> I'm posting this on behalf of a friend who does not subscribe.  Can
> anyone help him?  His group is
> 
> > finding that compiling under True64 and executing under Linux is
> > yielding 20-35% faster code than compiling using the DEC Linux FORTRAN
> > compiler and running under Linux. Can this be the page coloring issue
> > raising it's head?
> 
> However, incorrect 'getcwd' results under Linux are causing the
> following problem:
> 
> > We have an alpha running True 64 that we are using to compile code
> > (Fortran, C) using the Compaq compilers. When we move the executable to
> > an alpha running Red Hat Linux 7, any file i/o causes the code to crash.
> > Executing the code with strace displayed
> > 
> > open("getwd: can\'t stat //junk", O_RDWR|O_CREAT, 0666) = -1 ENOENT (No
> > such file or directory)
> > 
> > which corresponds to the statement in the Fortran code
> > 
> > open(10, file='junk', status='unknown', form='formatted')
> > 
> > Because the file name given in the open statement does not give an
> > absolute path, the os is being queried through the getcwd function to
> > determined the current working directory. This directory is prepended to
> > the file name. When the code is run under Linux, the incorrect path of
> > "//" is returned by getcwd. If all of the fortran statements are
> > modified to give absolute paths, the code runs correctly under Linux.
> > Also, if a c subroutine named getcwd is linked in that returns the
> > directory that the code will be run in, it works properly. The problem
> > with these two solutions is that the code will always try to do i/o in
> > the hard-wired directory regardless of where the executable resides.
> > This often creates confusion. We would like to compile these codes under
> > True 64 instead of Linux because we have seen a significantly better run
> > times for computationally intensive codes. Does anyone know of a simple
> > workaround for this problem?
> 
> 
> Thanks in advance,
> Josip
> 
> -- 
> Dr. Josip Loncaric, Research Fellow               mailto:josip at icase.edu
> ICASE, Mail Stop 132C           PGP key at http://www.icase.edu./~josip/
> NASA Langley Research Center             mailto:j.loncaric at larc.nasa.gov
> Hampton, VA 23681-2199, USA    Tel. +1 757 864-2192  Fax +1 757 864-6134
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

-- 
Craig Tierney (ctierney at hpti.com)
phone: 303-497-3112