VASP Stops after >> parallel 3dFFT charge
Posted: Wed May 27, 2009 8:57 am
Hi,
I'm running VASP 4.6 in a SGI Xeon250 cluster. Executable created with Intel 10 compilers and Intel mkl libraries, as well as /opt/intel/mkl/10.1.2.024/lib/em64t/libfftw3xf_intel.a for FFT.
Any job I run, no matter the number of processors, ends after:
I use Torque/MPI, and this is the output of the job.err:
Apparently it gets stuck when doing FFTw stuff, any ideas?
I'm running VASP 4.6 in a SGI Xeon250 cluster. Executable created with Intel 10 compilers and Intel mkl libraries, as well as /opt/intel/mkl/10.1.2.024/lib/em64t/libfftw3xf_intel.a for FFT.
Any job I run, no matter the number of processors, ends after:
Code: Select all
parallel 3dFFT wavefunction:
    minimum data exchange during FFTs selected (reduces bandwidth)
 parallel 3dFFT charge:
    minimum data exchange during FFTs selected (reduces bandwidth)
Code: Select all
numero de procesadores para ejecutar la tarea:  4
 running on    4 nodes
 distr:  one band on    1 nodes,    4 groups
 vasp.4.6.28 25Jul05 complexÂ
 POSCAR found :  4 types and  152 ions
 LDA part: xc-table for Ceperly-Alder, standard interpolation
 POSCAR, INCAR and KPOINTS ok, starting setup
 FFT: planning ...           1
MPI: On host cl1n002, Program /usr/local/bin/vasp, Rank 0, Process 23930 received signal SIGSEGV(11)
MPI: --------stack traceback-------
sh: idb: command not found
MPI: -----stack traceback ends-----
MPI: On host cl1n002, Program /usr/local/bin/vasp, Rank 0, Process 23930: Dumping core on signal SIGSEGV(11) into directory /data1/al
do-scratch/vasp_4.6/SRC/bench1
MPI: MPI_COMM_WORLD rank 0 has terminated without calling MPI_Finalize()
MPI: aborting job
MPI: Received signal 11
Apparently it gets stuck when doing FFTw stuff, any ideas?