Differences
This shows you the differences between two versions of the page.
Next revision | Previous revision | ||
doku:gpaw [2014/03/21 09:31] – external edit 127.0.0.1 | doku:gpaw [2014/08/27 11:27] (current) – jz | ||
---|---|---|---|
Line 2: | Line 2: | ||
=== Settings Scientific linux 6.5 (VSC 1+2) === | === Settings Scientific linux 6.5 (VSC 1+2) === | ||
- | Following | + | Following |
- | * gpaw: svn checkout https:// | + | |
- | * ase: svn checkout https://svn.fysik.dtu.dk/ | + | * gpaw (?): 11253 ; ase (?): 3547 |
+ | * gpaw 0.10.0: 11364 ; ase 3.8.1: 3440 | ||
* gpaw-setups-0.9.9672 | * gpaw-setups-0.9.9672 | ||
* numpy-1.6.2 | * numpy-1.6.2 | ||
- | * MPI Version: impi-4.1.0.024 | + | * MPI Version: impi-4.1.x |
* FFTW from INTEL MKL | * FFTW from INTEL MKL | ||
* files: {{: | * files: {{: | ||
Line 84: | Line 85: | ||
=== running gpaw jobs === | === running gpaw jobs === | ||
- | Job submission on VSC-1 (on VSC-2 also mpich8, mpich4 can be used instead) | + | == Job submission |
< | < | ||
#!/bin/sh | #!/bin/sh | ||
Line 90: | Line 91: | ||
#$ -pe mpich 256 | #$ -pe mpich 256 | ||
#$ -V | #$ -V | ||
- | export OMP_NUM_THREADS=1 | ||
- | NSLOTS_PER_NODE_AVAILABLE=8 | ||
- | NSLOTS_PER_NODE_USED=4 | ||
- | NSLOTS_REDUCED=`echo " | ||
- | echo " | + | mpirun -machinefile |
- | for i in `seq 1 $NSLOTS_PER_NODE_USED` | + | </code> |
- | do | + | |
- | uniq $TMPDIR/machines | + | |
- | done | + | |
- | sort $TMPDIR/ | + | == Job submission using half of the cores on the compute nodes on VSC-2. == |
- | cat $TMPDIR/ | + | If each of your processes require more than 2GB (and less than 4GB) of memory, you can use the parallel environment '' |
- | + | <code> | |
- | mpirun -machinefile | + | #!/bin/sh |
+ | #$ -N Cl5_4x4x1 | ||
+ | #$ -pe mpich8 256 | ||
+ | #$ -V | ||
+ | mpirun -machinefile $TMPDIR/ | ||
</ | </ | ||
+ | If even more memory per process is required the environments '' | ||
+ | Alternatively, | ||
- | Significant speed up is seen when --sl_default is set. First two parameters for the BLACS grid should be similar in size to get an optimal memory distribution on the nodes. | + | Significant speed up is seen **in our test case** |