"Neil R. Smith" wrote:
Neil,
I don't know about the Portland compilers but we built GEMPAK on our Sun
E6500
(12 CPUs) with the Sun FORTE -xparallel option and it ran slower than a
version
compiled with standard optimization. Which leads to be believe GEMPAK
is not
parallelized. Can someone confirm or deny this? Has anyone benefitted
on GEMPAK with multi-CPUs?
> Has anyone had success building the gempak package using the
> Portland group compilers? And on a dual cpu box using the SMP
> options? If so, what were your experiences?
> Thanks, -Neil
>
> ---
> Neil R. Smith, Comp. Sys. Mngr. neils@xxxxxxxxxxxxxxxxxx
> Dept. Atmospheric Sci., Texas A&M Univ. 979/845-6272
>From owner-ldm-users@xxxxxxxxxxxxxxxx 25 2001 Aug -0700 19:36:56
Message-ID: <20010826023656.1560.cpmta@xxxxxxxxxxxxxxx>
Date: 25 Aug 2001 19:36:56 -0700
From: stonie.cooper@xxxxxxxxxxxxxxxxx
To: rmullenax@xxxxxxxxxxxx
Subject: Re: Gempak build with PGI compilers
Received: (from majordo@localhost)
by unidata.ucar.edu (UCAR/Unidata) id f7Q2b5P21482
for ldm-users-out; Sat, 25 Aug 2001 20:37:05 -0600 (MDT)
Received: from c007.snv.cp.net (c007-h003.c007.snv.cp.net [209.228.33.209])
by unidata.ucar.edu (UCAR/Unidata) with SMTP id f7Q2b2121467
for <ldm-users@xxxxxxxxxxxxxxxx>; Sat, 25 Aug 2001 20:37:02 -0600 (MDT)
Organization: UCAR/Unidata
Keywords: 200108260237.f7Q2b2121467
Received: (cpmta 1561 invoked from network); 25 Aug 2001 19:36:56 -0700
X-Sent: 26 Aug 2001 02:36:56 GMT
Received: from [216.126.169.251] by mail.planetarydata.com with HTTP;
25 Aug 2001 19:36:56 PDT
Content-Type: text/plain
Content-Disposition: inline
Mime-Version: 1.0
Cc: neils@xxxxxxxxxxxxxxxxxx, gembud@xxxxxxxxxxxxxxxx,
ldm-users@xxxxxxxxxxxxxxxx
X-Mailer: Web Mail 3.9.3.5
X-Sent-From: stonie@xxxxxxxxxxxxxxxxx
Sender: owner-ldm-users@xxxxxxxxxxxxxxxx
Precedence: bulk
Robert,
You are correct - gempak is not parallelized (as in HP FORTRAN) or threaded
(POSIX or otherwise). Depending on the OS, however, gempak can benefit from
SMP boxes; in Linux, with fully symetrical mainboards and a kernel compiled for
SMP, Linux will do _some_ load balancing - even across ipc. Depending on other
load variables, your decoders spawned from LDM may be running on one CPU, while
display programs are running on the other (in a dual cpu situation).
On Sparcs, my experience is that by compiling non-parallelized (non-threaded)
programs without SMP flags - they run a lot faster. As Solaris (or HP-UX)
lends less ability to look "under the hood", I cannot explain why this is the
case.
Stonie
On Sat, 25 August 2001, Robert Mullenax wrote:
>
> "Neil R. Smith" wrote:
>
> Neil,
>
> I don't know about the Portland compilers but we built GEMPAK on our Sun
> E6500
> (12 CPUs) with the Sun FORTE -xparallel option and it ran slower than a
> version
> compiled with standard optimization. Which leads to be believe GEMPAK
> is not
> parallelized. Can someone confirm or deny this? Has anyone benefitted
> on GEMPAK with multi-CPUs?
Stonie R. Cooper
Science Officer
Planetary Data, Incorporated
3495 Liberty Road
Villa Rica, Georgia 30180
ph. (770) 456-0700
