My guess about the recent sander parallel problems is that the culprit
is likely to be the changes in parms.f ( and perhaps rdparm.f ) that Mike
put in Oct. 3 or so.
The change in parms.f eliminates the "rparms" common block, but this is
used in parallel.f to broadcast the parameters to the non-master nodes.
I haven't tested this yet, but it certainly looks and smells like a bug.
...dac
Received on Sun Oct 15 2006 - 06:07:03 PDT