We had seen that LAM works much better than MPICH. Since I myself was not
involved in making it work on LAM and we compared it with default MPICH
file, probably this is information is not to be taken at face value.
Still, if one desires so, it seems, a fine-tuned LAM performs doubly
faster than the default MPICH under certain circumstances and
configurations, at least for AMBER6 benchmark.
Please treat this info off-the record, I just told what we faced here and
I do admit I may be partially/totally wrong! :-)
Received on Wed Aug 22 2001 - 22:31:58 PDT
