28

I have a huge collection which I would like to process each of the object in a parallel fashion. Doing the same synchronously typically involves a simple "foreach".

Just wondering, what to use in my scenario i.e., either Parallel.ForEach or MyCollection.AsParallel().ForAll().

Can you shed me some light on benefits of using the one vs the other.

svick
  • 236,525
  • 50
  • 385
  • 514
user203687
  • 6,875
  • 12
  • 53
  • 85
  • 1
    This seems related; http://stackoverflow.com/questions/2270097/does-parallel-foreach-require-asparallel?rq=1 – Patrick Jun 04 '13 at 16:01
  • 1
    possible duplicate of [Parallelizing a task using .AsParallel().ForAll or Parallel.ForEach performance issue](http://stackoverflow.com/questions/8415998/parallelizing-a-task-using-asparallel-forall-or-parallel-foreach-performance) – Patrick Jun 04 '13 at 16:02

3 Answers3

31

The way I see it, if all you want to do is to parallelize a foreach, you should use its parallel version, which is Parallel.ForEach().

ParallelEnumerable.ForAll() is for the specific case where you have a LINQ query followed by a foreach and you want to parallelize both of them. So, what you do is that you parallelize the LINQ query by adding AsParallel() and then you parallelize the foreach by changing it into ForAll().

svick
  • 236,525
  • 50
  • 385
  • 514
10

Time both and pick the winner.

I have found Parallel.ForEach to perform better in most delightfully parallel scenarios. It would seem that it uses a more advanced work balancing model. PLINQ is primarily aimed at parallelising queries.

Kirill Shlenskiy
  • 9,367
  • 27
  • 39
4

My 2 cents:

With ForAll you can specify the desired level of parallelism using the WithDegreeOfParallelism extension method.

With ForEach you can only limit the maximum level.

Basically, you can force ForAll to use more threads.

Serge Shultz
  • 5,888
  • 3
  • 27
  • 17