[Beowulf] Shared memory

Greg Lindahl lindahl at pathscale.com
Thu Jun 23 15:10:07 PDT 2005


On Thu, Jun 23, 2005 at 09:10:21AM -0700, Michael Will wrote:

> I was just yesterday benchmarking our A3400 quad-opteron with dual cores
> using UnixBench 4.1

Not only is this a terrible collection of microbenchmarks, it doesn't
sound like your test setup was manipulating process affinities. So it
was realistic only for random interactive usage and not OpenMP or MPI,
where the runtime libraries manipulate the process affinities.

Please, please, shoot this benchmark in the head and bury it very
deep. So to speak.

-- greg




More information about the Beowulf mailing list