I would like to parallelize the primecount among the cluster.

Primecount does not support distributing computations onto cluster nodes anymore. Up until primecount-6.4 I had added support for MPI to primecount which allowed computations to be automatically distributed onto cluster nodes (see here: https://github.com/kimwalisch/primecount/blob/v6.4/doc/primecount-MPI.md). However there were no primecount MPI users that I was aware of for the few years where this feature was supported and it was a lot of work to maintain this MPI code. Therefore I removed MPI a few years ago (and I am not interested in bringing back MPI support).

But if you need to compute many different primecount results. You could manually distribute these computations onto the different cluster nodes so that each cluster node computes a different primecount result.

primecount will be running to get the nth primes up to specific large number primes (10^20 - 10^30) in large ranges.

Please note that primecount's nth prime functionality only works up to 2^64. This is because primecount's nth prime functionality relies on primesieve's sieve of Eratosthens implementation which is limited to 2^64 due to its O(sqrt(n)) memory usage.

Hence to compute the nth prime > 2^64 using primecount you cannot use primecount alone. You can use primecount to compute the initial nth prime guess, but from there on you will need to iterate over the primes until you find the nth prime using your own implementation (you basically need to reimplement my nth_prime.cpp with support for nth prime > 2^64). Note that this is how David Baugh computed the current nth prime records, I think he used Mathematica to iterate over primes > 2^64.

Uh oh!

Documentation for running primecount in an HPC environment. #77

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions