Re: Linux Multi-CPU performance

Scott,

It looks like you are using a 2 (dual core) CPUS is this so? Some
things to check.
1. Run your program on 2 CPUS
2. Did you optimize your compiler command for your xeon? I presume you
use the same compilation command in both cases.
3. Did you build MPI implementation yourself, or did it come with the
compiler. Is it MPI, or OpenMP, another?
3a. If MPI are you using ch_p4 or ch_shmem? If ch_p4 what is the
communication layer, ssh, rsh?

I am running a ch_shmem MPI on my dual core dell and all seems well.

If you post your short program I can rerun it to give you an idea.

-Tom

.

Relevant Pages

  • Re: mpi(ch) on dual-core CPUs (X2 - core intel or AMD)
    ... >> efficient as 2 CPUs on the same dualCPU-motherboard setup in MPI jobs? ... > bandwidth then neither a two socket Intel system nor a dual core Intel ... > well if I had chosen to use boards with chipsets from VIA or Nvidia. ... > All in all I would go with Intel based boards and slow clocked Pentiums, ...
    (comp.parallel.mpi)
  • Re: very low performance for very small packets under MPICH
    ... >be true independent of platform, MPI, OS etc. ... * New interconnects like InfiniPath are giving better bandwidth ... at smaller packet sizes ... * InfiniPath also gets a benefit from using all the cpus on the ...
    (comp.parallel.mpi)