Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 1 Oct 2000 20:36:43 +0200 (CEST)
From:      "O. Hartmann" <ohartman@ipamzlx.physik.uni-mainz.de>
To:        freebsd-stable@freebsd.org
Cc:        freebsd-emulation@freebsd.org
Subject:   Fortran 90/HPF on FreeBSD 4.x
Message-ID:  <Pine.BSF.4.21.0010011928240.22112-100000@ipamzlx.physik.uni-mainz.de>

next in thread | raw e-mail | index | archive | help
Dear Sirs.

I need help!

I'm looking for a Fortran 90 compiler with SMP/parallel 
capabilities, running under FreeBSD 4.1.

I looked around for some compilers and I found these:
FreeBSD nativ:
 - NAG Fortran F90 (www.nag.co.uk)
 - g77 Fotran 77 (GNU Fortran)

Linux nativ:
 - PGI Compiler Suite (includes F77/F90/HPF/C/C++), MP capable over
   OpenMP (www.pgroup.com)
 - Lahey/Fujitsu F95 for Linux Express V5.5 (www.lahey.com)
 -VAST/F90 (www.psrv.com).

NAG offers also compilers for Linux and you can get a free download 
of theire compiler suits, but for FreeBSD, I would have to obtain a
20 British Pound fee, Linux is completely free.

So, first I tried to install Lahey's F95. I ran into trouble because
of Lahey's glibc-detection routine, it showed me that my FBSD-Linux
emulation uses glibc, not glibc2.1 so the compiler was installed
with the wrong libraries and this happened while their were some
updates in the FBSD code on which I accused the compiler malfunctions.
Well, to come alongside the problem, I tried now to compile some
software and Lahey/Fujitsu works well on F77 code (but I only have
one testprogram at the moment). Another project compiled well, but
the nativ f95 code has some syntax errors, which are NOT shown by the 
PC edition (division by zero, WinNT version does not stop there :-)).

Well, because one of our new oncoming scientists is highly involved 
by meterological, environmental research, radiation-research, he
uses a lot of code which has been developed on PGI's compiler
suite and he told me, that a lot of the stuff needs to run on PGI
compilers. Well, for that, I got a copy of their Linux distribution
on September, 30th and installed it. Version 3.1-1 compiled well and
our scatter function test suite (F77) was compiled well and I got
an executable binary. The other software with the obscure division
by zero error (mentioned above) was compiled well also, but the 
executable drops a core after starting. Yesterday evening and today
I visited Portland again and got a copy of their new compiler R 3.2-2.
I made the same installation, took some modifications in linux86rc file
to make it compiling - and it compiles. But the result is always 
dumping a core, also the F77-compatible source which has been success=
fuly compiled with the prior version. I tried to downgrade again, but
Portland Group cleaned up their downloads and I can only obtain
R 3.1-2 and -3 of the compiler suite.

NAG compiler, Linux version, shows a lot of error messages on the 
f77 standard code and I'm not the professional to clean it up or
apply the right tricky compilation option. Some collegues tried 
it, too, with no success. i do not want me wasting money on a
"free" copy of their FreeBSD nativ compiler because it seems not
realitic that their suite is to be used here at our department.
So I have a focus on Lahey/Fujitsu F95 and especially PGI's
compiler suite, especially last one is prefereable because of the
MP capabilities. Unfortunately the newest version is not willing
to run under FreeBSD 4.1.1-STABLE.

So far on my experiences.

Waht I need, really need, are some little performance hungry f90
native applications, coded as nativ f90 applications without any tricks 
or special library needs (vendor specific) to test and stress Lahey
or PGI compilers and to proof, whether they produce real code or
a heap of bit-trash (sorry for that). 


--> 	Is there anybody who's testing Portland Group's compiler on 
	FreeBSD's Linux emulation? 

-->	Who knows some places where I can download some testsuits
	of simple f90/hpf/f95 code to stress the compiler?
	Please not any kind special applications with special
	needs of applied libraries ...

-->	For the lucky case, Portland Group's compiler is willing
	to work under some circumstances under FreeBSD, where to
	find some parallelizing applications for testing purposes?

Another, maybe crucial question is: how reliable, fast and scalable
is Linux in high performance computing using OpenMP or PVM (clustering)
in comparison to FreeBSD?

FreeBSD seems not to be very wide spread in numerical application's
research using Fortran, I guess (unlike the ancestors using BSD 4.2
or similar, like Ultrix machine or SunOS).

If anybody has any hint, tip or trick in solving my problems to get the PGI
running under FreeBSd, please email me. And: I need essential help in
saving my FreeBSD infrastructure, it should not be replaced by Linux
only due the lack of suitable high performance Fotran compilers.


And, last but not least:
Are there any benchmark comparisons between Linux and FreeBSD? I need some
recent tests, especially how fast is Linux in doing SMP and FreeBSD/SMP.

Last week I had a discussion with some scientists of our computing center
and they told me something about Solaris vs. Linux regarding SMP. I was
told that Linux was faster if the number of CPUs is not higher than 4 pieces,
means: Linux must be really fast because Solaris is said to be a very fast
and well scaling SMP OS?! And FreeBSD? I have no datasheets, so I can not 
decide anyway ...
-
MfG 
O. Hartmann
-------------------------------------------------------------------
ohartman@ipamzlx.physik.uni-mainz.de

Klimadatenserver-Abteilung des IPA 
IT Netz- und Systembetreuung
Johannes Gutenberg-Universitaet Mainz
Becherweg 21
D-55099 Mainz
BRD/Germany

Tel: +496131/3924662
FAX: +496131/3923235



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0010011928240.22112-100000>