[CIG-SEISMO] about computational efficiency
Yingzi Ying
yingzi.ying at me.com
Tue Dec 18 08:50:20 PST 2012
Hi,
I tested a job with 720*720*50 meshes generated by xmeshfem3d and ran
xspecfem3d on a cluster using 27 nodes, each of which has two Intel Xeon
E5650 hex-core processors (12 cores) and 24GB of RAM. After 24 hours'
running, xspecfem3d only solved 5500 time steps.
I am not sure if this efficiency if reasonable, or there might be the
so-called "over-load memory problem" that there is not enough free
memory left for system communication?
Best regards,
Yingzi
More information about the CIG-SEISMO
mailing list