Iced Tea in Crete 2012:Parallel programming

From WikiEducator
Jump to: navigation, search
Parallel programming
Convenor: Michael Hunger
Participants:
  • John Kostaras
  • Ignasi Marimon-Clos
  • Giannis Skitsas
  • Heinz Kabuts
  • James Scobbie
Summary:

What do you use to parallelise your application?

  • File:Akka.pdf / Scala
  • Erlang
  • Fork/join / ParallelArray
  • GPars
  • Disruptor
  • Executors/Futures
  • Completion Service
  • @Asynch

In the previous listing Akka is partially an outsider since it a library but also part of a product (see Typesafe Stack and Typesafe Console).

Recommendations:

Recommendations go here

  • small task may cause too much overhead
  • big tasks may reduce your parallelizaion
  • Queues present contention issues in either head, tail or in the counter when very balanced
  • try to get prefetch to act. Using Objects massively in your design destroys any prefetch hope
  • ...