0

I wrote an application that uses scala parallel collections and akka actors, and I would now like to study its "strong scaling" properties, i.e. how the running time for a given problem instance changes, as a function of the number of cores/threads available.

What would be a proper way of going about this? How can I tell the application to use only up to n cores/threads?

mitchus
  • 4,677
  • 3
  • 35
  • 70
  • possible duplicate of [limiting number of threads used by the JVM](http://stackoverflow.com/questions/17341883/limiting-number-of-threads-used-by-the-jvm) – om-nom-nom Oct 08 '13 at 16:13
  • In AKKA there are various thread pools that you can change in application.conf. For instance IO workers can be configured. Also look at `thread-pool-executor` setting. Basically, if you use remote AKKA many other factors come into play, so I guess you want to use local actors for this kind of benchmarking first. – yǝsʞǝla Oct 08 '13 at 16:33
  • @om-nom-nom : my question is not about the JVM in general, but about scala+akka. So configurations of akka and parallel collections would be perfectly good answers, which is clearly not the case for the linked question. (Also, that question has no accepted answer.) – mitchus Oct 08 '13 at 19:24
  • 2
    If you want to control the parallelism for the thread pool behind parallel collections, check this link out: http://docs.scala-lang.org/overviews/parallel-collections/configuration.html. For akka, check out any documentation related to dispatchers like this: http://doc.akka.io/docs/akka/2.2.1/scala/dispatchers.html. – cmbaxter Oct 09 '13 at 00:15

0 Answers0