test for 4096 cpus. was RE: [CPMD-list] Empty ENERGIES file on x86-64 systems

[Alexandr Isayev]

 Dear CPMD community,
 Igor, Axel

KII> what test cases have you run and on how many processors?
KII> It would be interesting if you could post the results
KII> when you are ready.

Sure I will. I am running Si512 right now, and several unit cells
of my system [144 atoms per unit cell]. Unfortunately it is monoclinic unit cell,
so I had a lot of fun before browsed the list archives :-)
Therefore, these are NOT the proper systems to run on zillions of CPUs.

KII> Sometime ago CPMD developers promised to make available
KII> the test cases which could scale to a large number of cpus.
KII> Given that yours has 4096, would not it be nice
KII> to run something suitable on it?
KII> Is there a _relaible_ standard test case which could
KII> fit the bill?

Please look on page 35.
http://researchcomp.stanford.edu/hpc/archives/BlueGene.pdf
If this Pd-H sample system became available for test purposes
it might provide "golden standard" for testers.
However, this question should be addressed to the authors of the
project.

KII> [e.g. it might be tempting to put a quick but not
KII> robust wavefunction optimization method in the test. 
KII> But then you end up having different number of iterations 
KII> on different machines and how do you compare those?]

KII> Best,
KII> Igor

Sincerely,
Alex




-------------------------------------------------------
Alexandr Isayev,
Graduate Research Assistant, and System Administrator
@ Computational Center for Molecular Structure
and Interactions (CCMSI),
Jackson State University,
Jackson, MS USA
   Tel:  +(601) 979-1134
e-mail:  alex(at)ccmsi.us
   Web:  http://www.ccmsi.us
--------------------------------------------------------