ISYM=0 vs ISYM=2 for CPU efficiency

Queries about input and output files, running specific calculations, etc.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
Eduard
Newbie
Newbie
Posts: 11
Joined: Fri Dec 07, 2007 12:01 pm
License Nr.: 749

ISYM=0 vs ISYM=2 for CPU efficiency

#1 Post by Eduard » Wed Jun 08, 2011 10:05 pm

Dear All,

I am studying a system which has no symmetry. Nevertheless, simply by curiosity I wanted to check the impact of the ISYM tag on the computational time. I have run the same calculation, once with ISYM=0 and once with ISYM=2.

Here is the outcome:

ISYM=0
=======
Total CPU time used (sec): 334510.688
User time (sec): 329592.986
System time (sec): 4917.711
Elapsed time (sec): 388932.160

E0= 840.18495 eV


ISYM = 2
========
Total CPU time used (sec): 389981.406
User time (sec): 384049.926
System time (sec): 5931.483
Elapsed time (sec): 807039.157

E0= -840.18517 eV

By default I have not expected any significant difference between the computational time of the 2 runs. Nevertheless, there is a significant slow down of the computational time when the ISYM=2 for a non-symmetric system. Why is this so ?

Looking into the OSZICAR, the number of the ionic relaxation steps is the same (37 in both cases) and also the total energies (given above). Why does the ISYM=2 presence in the INCAR slow down the electronic relaxation ? Notice also the large elapsed time for the ISYM=2 calculation. What is the explanation ?

Ed
Last edited by Eduard on Wed Jun 08, 2011 10:05 pm, edited 1 time in total.

alex
Hero Member
Hero Member
Posts: 585
Joined: Tue Nov 16, 2004 2:21 pm
License Nr.: 5-67
Location: Germany

ISYM=0 vs ISYM=2 for CPU efficiency

#2 Post by alex » Thu Jun 09, 2011 10:25 am

Hi Ed,

for the huge differences in elapsed time I'd blame your computer. Maybe NFS was down for some time or so. It happens to my system if I do huge file transfers (>~ 100 GB or so).

About user time I have no idea. The manual states it should be more efficient (at least in parts).

Have you also checked timings for a single electronic step?

Cheers,

alex
Last edited by alex on Thu Jun 09, 2011 10:25 am, edited 1 time in total.

admin
Administrator
Administrator
Posts: 2921
Joined: Tue Aug 03, 2004 8:18 am
License Nr.: 458

ISYM=0 vs ISYM=2 for CPU efficiency

#3 Post by admin » Thu Jun 09, 2011 2:13 pm

the speedup of course also depends on the symmetry of your cell (i.e the reduction of the irreducible part of the BZ with respect to the BZ). Please check the number of symmetry operations of your cell
Last edited by admin on Thu Jun 09, 2011 2:13 pm, edited 1 time in total.

prasenjit

ISYM=0 vs ISYM=2 for CPU efficiency

#4 Post by prasenjit » Fri Oct 26, 2012 10:24 pm

[align=center]It happens because, when ISYM = 2 then before entering scf cycle, the system checks the symmetries present in your structure. after doing all possible checks, it reaches to the conclusion that there is actually no symmetry present in your structure. So then it start doing the same number of scf cycle as ISYM = 0. And hence it takes same number of cycles. On the other hand, ISYM = 0 do not do any symmetry checks, so it immediately enter the calculation part.[/align]
Last edited by prasenjit on Fri Oct 26, 2012 10:24 pm, edited 1 time in total.

Post Reply