Hi,
I noticed that inserting particles via insert/pack with particletemplate/sphere and particledistribution/discrete on 8 processors will generate the same distribution of particle sizes on each processor. If I have a dilute species with only 32 particles in the system, each processor will pick the same four sizes for the dilute species. This problem gets progressively worse as the number of processors increases. Now I'm in a situation where I need a larger system size to get an accurate PSD, but I can't increase the number of processors to do it. I was going to just modify the code to give a unique random number seed to each processor, but I wanted to point out this issue and see if the particle generation can be improved in the next release.
David
ckloss | Thu, 09/27/2012 - 12:15
Hi David, >>I noticed that
Hi David,
>>I noticed that inserting particles via insert/pack with particletemplate/sphere and particledistribution/discrete on 8 processors
>>will generate the same distribution of particle sizes on each processor
>> I was going to just modify the code to give a unique random number seed to each processor,
>>but I wanted to point out this issue and see if the particle generation can be improved in the next release
typically this is not a problem, But yes, if you have a very dilute species with O(nprocs) = O(nparticles) this can have a considerable effect.
I changed that for the next release.
Cheers, Christoph