[CPMD-list] TCPU

[Axel Kohlmeyer]

On Tue, 01 Oct 2002 18:51:00 -0400  HW Sheng wrote:

> Hi, Axel and Daniel,
> 
> Many thanks! Now I have a better picture of the network computing. Still, I
> am concerned about the latency. I reckon, there might be a way to reduce the
> overhead (in my case, it's an outrageous 50%) spent on the inter-connetion.

hi howard.

that is because you are using _dual_ machines, 
so you only have half of each ethernet for each cpu and 
a single cpu is already saturating the network bandwith 
(latencies usually don't matter so much for large cpmd jobs).

try running the exact same job on 4 machines while using a
single cpu and on 2 machines with both cpus and compare the timings.

also the network timing results at the end of the job are 
not reliable on linux machines when using lam, since they don't 
use wall time but cpu+system times as base for the timings.

best regards,
	axel.

> But, anyways, I appriciate your help.
> 
> Howard
> 
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://www.cpmd.org/mailman/listinfo/cpmd-list
> 

--

=======================================================================
Axel Kohlmeyer      e-mail:  axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum                   http://www.theochem.ruhr-uni-bochum.de
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.