K-point, G-vectors and Bands parallelization
Posted: Wed Mar 23, 2011 5:43 pm
Dear all
I’m dealing with big gold clusters and to speed-up calculations I need to use the K-point, G-vectors and Bands parallelization. I performed many different tests increasing the energy cutoff. With low energy cutoffs the program completes the jobs without problems. On the contrary when the cutoff is higher (approx. 24) Abinit does’t print the WFK file. I attached an input files for which Abinit doesn’t write the WFK file and also the respective output. The last rows of this file are:
“----iterations are completed or convergence reached----
outwf : write wavefunction to file /storage/…
-P-0000 leave_test : synchronization done...
Subroutine Unknown:0:COMMENT
MPI/IO accessing FORTRAN file header: detected record mark length=4
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 3 in communicator MPI_COMM_WORLD
with errorcode 1.
NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
Did anyone get into the same trouble?
Did he/she fixed it? How he/she fixed?
Thank you all for your help.
Luca Sementa
IPCF-CNR
Via G. Moruzzi 1
56124 Pisa (ITALY)
I’m dealing with big gold clusters and to speed-up calculations I need to use the K-point, G-vectors and Bands parallelization. I performed many different tests increasing the energy cutoff. With low energy cutoffs the program completes the jobs without problems. On the contrary when the cutoff is higher (approx. 24) Abinit does’t print the WFK file. I attached an input files for which Abinit doesn’t write the WFK file and also the respective output. The last rows of this file are:
“----iterations are completed or convergence reached----
outwf : write wavefunction to file /storage/…
-P-0000 leave_test : synchronization done...
Subroutine Unknown:0:COMMENT
MPI/IO accessing FORTRAN file header: detected record mark length=4
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 3 in communicator MPI_COMM_WORLD
with errorcode 1.
NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
Did anyone get into the same trouble?
Did he/she fixed it? How he/she fixed?
Thank you all for your help.
Luca Sementa
IPCF-CNR
Via G. Moruzzi 1
56124 Pisa (ITALY)