[Pw_forum] Benchmarking: openmpi-1.3 vs openmpi-1.2.8

[Axel Kohlmeyer]

On Sun, 15 Feb 2009, Mahmoud Payami wrote:

MP> Dear ALL,
MP> 
MP> I performed a benchmarking test using openmpi-1.2.8 and openmpi-1.3 for a 
MP> 2-atom Au cluster.
MP> The out files show that in all parts "1.3" takes shorter time than "1.2.8". 
MP> However, the wall time in "1.3" is slightly increased. I do not know how to 
MP> resolve it. Is the bottleneck the connection cables?
MP> Bests,
MP>           Mahmoud Payami
MP> 
MP> ----------------------------------------
MP> #out_1.2.8
MP> PWSCF : 2m 5.29s CPU time, 4m43.10s wall time

wow. your wall time is over double 
the CPU time. this is _bad_. over half
of your cpu power is wasted.

what kind of setup are you running? what 
interconnect? what types of machines?
without context there is little to learn
from this.

[...]

MP> #out-1.3
MP> PWSCF : 1m55.59s CPU time, 4m44.04s wall time

same here. in practice on the wall time matters,
so there is no significant difference. but it
also is highly likely, that you are not running
in an efficient way. this is not really useful in
telling you anything about the two mpi versions.

cheers,
   axel.

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.