| 7:32:12 PM PST - Sun, Nov 30th 2014 |
|
Hi,
I succesfully compiled nwchem 6.5 code on my high performance computer. When I test NWPW module using the example provided by NWChem named s2.nw, I found I can use at most 50 cpus to compute correctly. If cpus are more than 50, errors will occour. the main error is "TCGMSG DMatrix_dgemm1_rot2 not implemented"
In addition, I test my own nwpw example, and found when the parallel scale is large such as 128 cpus, the program looks like stoping at reading the potential file. But if the parallel scale is small, everthing seems ok.
I compile NWChem 6.5 code by intel compiler 2013 and openmpi 1.6.5 by "-O2" argument (If I use -O3, nwpw module can not run even for searial computation.) .
Does somebody can help me? what's wrong?
|