problem in parallel calculation

Questions regarding the compilation of VASP on various platforms: hardware, compilers and libraries, etc.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
zhuanghl
Newbie
Newbie
Posts: 31
Joined: Thu Dec 08, 2005 2:15 pm
License Nr.: 471

problem in parallel calculation

#1 Post by zhuanghl » Sat Mar 04, 2006 9:18 am

I use completely the same input files for both the parallel and serial version of the code.There is not any problem in the serial calculation,while some warning like:

BRMIX: very serious problems
the old and the new charge density differ
old charge density: 180.09651 new 180.26392
0.571E+01

appear at some step of the parallel calculation.I don't know how so and whether it affect the final result.Thanks!
Last edited by zhuanghl on Sat Mar 04, 2006 9:18 am, edited 1 time in total.

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

problem in parallel calculation

#2 Post by admin » Tue Mar 28, 2006 8:23 am

1) please check if parallelisation affects other parameters of the run, like NBANDS: NBANDS is set to an integer multiple of the number of nodes. If the number of empty bands is too low, this may affect the results (especially if you use the RMM-DIIS algorithm for electronic convergence).
Set NBANDS to a well-defined value in INCAR and test again.
2) check whether the calculations have converged to the same final results though: Mixing does not affect the converged results themselves, however it may change the speed of convergence.
3) if NBANDS seems reasonable in the parallel run, and the results differ for exactly the same input, I think this is an indication that something may be wrong with your parallel environment. Try to decrease the compiler optimisation flags re-compile VASP and do another test run. If this does not work as well, ask the people who installed the computers to make checks of the hardware / compiler installation.
Last edited by admin on Tue Mar 28, 2006 8:23 am, edited 1 time in total.

mindflying

problem in parallel calculation

#3 Post by mindflying » Thu Jul 12, 2007 5:46 am

Hi, zhuanghl,

How did you get the problem solved?
I got the same problem when compiling parallel vasp, and serial vasp has no problem.

The machine is opteron. I can compile the parallel vasp smoothly, though with several annoying messages as follows:
ifort: command line warning #10121: overriding '-xO' with '-xW'
ifort: command line remark #10148: option '-tp' not supported
fortcom: Warning: elpol.f90, line 693: The statement following this DEC loop optimization directive must be an iterative do-stmt, a vector assignment, an OMP pdo-directive, or an OMP parallel-do-directive.
!DIR$ IVDEP
------^

In fact I got the "BRMIX" problem with serial version when using setting as follows:
FFT3D = fftw3d.o fft3dlib.o /usr/local/lib64/libfftw3.a
, but it's ok when I swithed to the other setting:
FFT3D = fft3dfurth.o fft3dlib.o


Thanks.
Last edited by mindflying on Thu Jul 12, 2007 5:46 am, edited 1 time in total.

Post Reply