Seeding the random number generator

[james@let.rug.nl]

Hi,

I recently did some training runs in parallel on a linux cluster only
to find that I got identical results in each run. Could this be down
to the processes being started at the same time?

If the random number is seeded via the system clock, then since the
clock is synchronised across the cluster, there might be a danger of
this happening when the process scheduler starts the parallel tasks
within the same unit of time on the system clock. I had this problem
with some other code that was fixed when I used a different method of
initialising the random number generator. How do I do that in PDP++?

Cheers,

James