On 4/13/06, Brady Catherman <[EMAIL PROTECTED]> wrote:
For our Mac OS run I was able to work with Apple to get performance
numbers up. Mac OS started much, much slower than Linux and failed to
scale at all. One of the Apple engineers emailed with several
recommendations to improve performance. The biggest was to not use
Mpich 1. Once I compiled against Mpich 2 everything went much faster
on the Apple side of things.

There's one thing I had better try out then, we were just talking about
potentially upgrading recently and decided against  it as the Myrinet-GM
implementation of Mpich was built against 1.2.6.

Anyways, the Apple run of HPL was compiled against the Accelerate
framework while the Gentoo side was compiled against the most recent
Goto libraries (in January 2006). The Apple side was compiled with a
custom version of GCC 4.0.2 (XCode doesn't come with fortran) and the
Linux side was compiled with GCC 4.0.2 from Gentoo's ebuild.

I may have made it harder for myself in that I compiled a 64 bit
kernel and 64 bit platform.

Our setup is XServe G5's 2.3GHz with 4GB of RAM, running form a local
80GB SATA hard drive. At the time we just had a simple GigE
connection between nodes. Not we have a GigE based FNN.

Same basic setup here except the majority of our G5's are 2Ghz with 2G of RAM.
We used the IBM xlf compiler and Goto's libraries in OS X and development
ATLAS 3.7.11 (development for G5 support), gcc is version 3.4.4 in Gentoo
and Debian.

The Gentoo image is true 64-bit whereas the Debian is 64-bit kernel and 32-bit
userland.  Gentoo ends up with a tiny advantage here so far, but we've only
tested across 8 nodes, so I wouldn't really count the results yet.

Hopefully we'll get our new netbooter up and working soon and I'll run some
more comparisons across a much larger set of nodes.

Thanks,

Justin Bronder
University of Maine
ACRL

with gcc
What environment was your partner using?


Reply via email to