How to collect results from compiled function in ParallelEvaluate?

Question

I am using Mathematica 12.x and have a compiled function, returning many things. I would like to run several instances of this compiled function, one for each core of my CPU, and then collect all of the results. After looking through SE page here I am not sure how to accomplish this.

My function is something like this:

{out1, out2, out3, out4, out5, out6} = compiledFunction[<integer args>]

How should I go about coding the ParallelEvaluate[] to enable the behavior I want?

I would assume something like:

DistributeDefinitions[compiledFunction];
ParallelEvaluate[ SetSystemOptions[ "ParallelOptions" -> "ParallelThreadNumber" -> $ProcessorCount]];
ParallelEvaluate[{out1, out2, out3, out4, out5, out6} = compiledFunction[<integer args>]];

But this does not put all of the thread's instances of functions into the output variables. If I could, I could create out1a, out1b, ..., etc and force them together as out1 = {out1a, out1b, ...} but I am not sure how to do that, and if there's a better way?

Answer 1

I think what you want would be something like:

{out1, out2, out3, out4, out5, out6} = Join @@@ Transpose[
   ParallelEvaluate[compiledFunction[<integer args>]]
];

ParallelEvaluate returns a list of length equal to the number of kernels and each kernel returns a list of length 6, so all you need to do is transpose the dimensions of ParallelEvaluate and join the results together.

Note that if you do:

ParallelEvaluate[{out1, out2, out3, out4, out5, out6} = compiledFunction[<integer args>]];

then all the assignments to out1 etc. will be done in the parallel kernels. In other words: those variables will not get a value in your main session (the master kernel).