I have rewritten the parallel (PVM) code of
ZZZZZZ so that it should also run now on a parallel machine
with many processors (the number of processors
in the previous version was limited to the number of
legal moves in the starting position).
Before I release version 3.34, I would like to
know if someone can test it for me on a machine
with 64 processors or more (I only have access
to a cluster of 8 workstations) and tell me
(1) if it works (2) if communication with the slave-driver
is a bottleneck (3) how much idle time there is
(4) the speedup and (5) tune some of the #defines
to find out the maximum speedup possible.
(2) and (3) would require knowledge of PVM profiling
(and some of the PVM profiling tools).
Regards
Gijsbert Wiesenekker
ZZZZZZ author