Large simulations

From: Sanjay H A <sanju_smg_at_lycos.com>
Date: Thu, 06 Mar 2008 05:04:46 -0500 (EST)

Hi All

I am running Gadget2 with lcdm_gas.parm .
I am using this application thinking that It is long running simulation (It should take hours on 32+ processors). Correct me if I am wrong.

To begin with I am taken lcdm_gas.parm example. Currecntly I executed on single proccessor.
For each step it is taking 1-2 secs. But I want to increase the cpu time taken by these steps should be in minutes ( on higher processor configuration).

Another thing I noticed in the screen output, there is differences in activities in each step ( as shown below step: 597, 598). Why is that

Begin Step 597, Time: 0.98836, Redshift: 0.0117773, Systemstep: 0.00231173, Dloga: 0.00234169
Start force computation...
Begin tree force.
tree is done.
Start density computation...
ngb iteration 1: need to repeat for 0000000003 particles.
Start hydro-force computation...
force computation done.


Begin Step 598, Time: 0.990677, Redshift: 0.00941078, Systemstep: 0.00231715, Dloga: 0.00234169
domain decomposition...
NTopleaves= 57
work-load balance=1 memory-balance=1
domain decomposition done.
begin Peano-Hilbert order...
Peano-Hilbert done.
Start force computation...
Tree construction.
Tree construction done.
Begin tree force.
tree is done.
Start density computation...
ngb iteration 1: need to repeat for 0000000467 particles.
Start hydro-force computation...
force computation done.


My requirement is, I have to know 1 importatnt parameter by changing that, I can get larger CPU time for each step. ( In Molecular Dynamics Simulation, Number of paricals is the Major parameter, by changing that we can vary CPU time)

Thanx in advance
Regards
Sanjay

---------[ Received Mail Content ]----------

Subject : Re: [gadget-list] Varying the Execution time (problem size)

Date : Wed, 05 Mar 2008 14:58:24 +0100

From : Revaz Yves <yves.revaz_at_epfl.ch>

To : Gadget General Discussion <gadget-list_at_MPA-Garching.MPG.DE>







Sanjay,



Well, I'm not sure to understand clearly what you want to do.

If you want to increase the CPU time needed to run the example

lcdm_gas.param,

you can of course increase the number of particles, i.e, creating a new

gassphere_littleendian.dat

file, which is not so easy at the beginning.

Easier to do is to vary the accuracy of time integration (see section

Accuracy of time integration

in gassphere.param). For example, decreasing ErrTolIntAccuracy will

increase the CPU time.

Or decreasing ErrTolTheta will also strongly increase the CPU time.



Cheers,



yves



Sanjay H A wrote:

> Hi All

> I am new to Gadget2.

> I am able to install Gadget2 and able to run one of the example

> program (lcdm_gas.param) using

>

> /usr/local/mpich-1.2.6/ch_p4/bin/mpirun -np 4 -machinefile hosts

> Gadget2 lcdm_gas.param

>

> Now I want to know, how to vary the execution time of the same example

> (on same number of processor) by varying any one of the parameter (

> which affects more on execution time, like in Molecular Dynamics

> Simulation, we can observe variation in Execution time by changing the

> number of particals).

>

> Can I know the parameter name for each example and also procedure to

> vary that parameter.

>

> Thanx in Advance

> Regards

> Sanjay

>

>

>

>

>









-----------------------------------------------------------



If you wish to unsubscribe from this mailing, send mail to

minimalist_at_MPA-Garching.MPG.de with a subject of: unsubscribe gadget-list

A web-archive of this mailing list is available here:

http://www.mpa-garching.mpg.de/gadget/gadget-list

Received on 2008-03-06 11:05:12