Page 1 of 1
parallel version yambo give strange result
Posted: Sat May 03, 2014 2:45 pm
by Sheleon
Hi, all
I compiled yambo code 3.4.1 with intel fortran and openmpi, the sequential version give the same results as in the spin-orbit tutorial. However the parallel version give very strange results, say, for the optical properties, it just print NaN for both Im and Re part of the dielectric tensor.
Any advice will be appreciate.
Sheleon
Re: parallel version yambo give strange result
Posted: Sat May 03, 2014 6:21 pm
by Daniele Varsano
Dear Sheleon,
please post the report file for both serial and parallel run and we will have a look.
Thanks,
Daniele
Re: parallel version yambo give strange result
Posted: Sun May 04, 2014 3:45 am
by Sheleon
Dear Daniele,
Thanks for the reply,
here is the report file:
r-para_optics_bse.log
r-seq_optics_bse.log
I have a quick look, nothing diferent was found.
Thanks.
Sheleon
Re: parallel version yambo give strange result
Posted: Sun May 04, 2014 7:14 am
by Davide Sangalli
Dear Sheleon,
it was a bug in the parallel version of yambo_kerr.
I've just relesed the bug-fix in the svn repository. If you downloaded yambo via svn just run 'svn up' and the code will update.
If you downloaded the tgz from qe-forge you need to open the file
src/pol_function/Dipole_transverse.F
and replace the line:
if(trim(global_gauge)=='velocity') call pp_redux_wait(DIP_P(:,:,:,:,i_spin))
with:
if(trim(global_gauge)=='velocity') call pp_redux_wait(DIP_P(:,:,:,ik,i_spin))
Best regards,
Davide
Re: parallel version yambo give strange result
Posted: Sun May 04, 2014 9:34 am
by Sheleon
Dear Davide,
The new version indeed solve the problem.
Thank you very much.
Best Regards,
Sheleon