You should also check out the LAM FAQ in the Heterogeneity section
(hmm... it seems that we're missing a 32/64 question, but all the other
concepts apply -- I'll add a 32/64 question soon), and the man page for
lamboot(1) and mpirun(1) for an explanation of the "cpu=2" and "mpirun
C ..." semantics.
On Feb 23, 2005, at 9:46 PM, Tim Prince wrote:
> At 09:59 AM 2/23/2005, you wrote:
>
>> Hi all,
>>
>> I have a question which puzzles me, about running LAM.
>> I have a 32-bit Dual Intel Xeon machine, and an AMD 64-bit Dual
>> Opteron machine.
>> Both of them are running Solaris 10, Build 72.
>>
>> I want to know whether there is a difference in performance between
>> running LAM across two 32-bit machines, or across one 32, and one
>> 64-bit machine.
>> I use HPL 1.0a to try to investigate this issue, but the performance
>> ratings are the same, for both conditions.
>
> I think your combination of questions exceeds anyone's desire to
> tackle it. Since your 64-bit machines may be capable of running
> 32-bit software with better performance than your 32-bit machines, you
> should consider running 32-bit lam-mpi and the same application build
> throughout. If your work is anywhere close to balanced, your
> performance will be limited by the slower machines.
>
>
> Tim Prince
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
