Julia 0.6 pmap function

Question

I want to paralelize a decomposition algorithm for optimization named Progressive Hedging. This optimization is stored on a function named PH which recieve arguments for the model, some arguments are matrix but PH needs just a vector from that matrix in this way.

for s = 1:nS
    res = PH(k,s,data,Lines,Ag,Gx,Pmax[:,s],Prmax[:,s],COpe[:,s])
    push!(data,res)
end

So PH needs just a vector from Pmax, Prmax and COpe.

To paralelize I try to do this.

 pmap(s -> PH(k,s,data,Lines,Ag,Gx,Pmax[:,s],Prmax[:,s],COpe[:,]),1:nS)

But I get this:

The applicable method may be too new: running in world age 21846, while current world is 21965.

I'm using Julia 0.6 maybe the way i'm programming is from a older version.

Any idea?

Try to look at: docs.julialang.org/en/stable/manual/methods/… — Liso
You did not give sufficient code to reproduce the issue. World age issues are due to defining functions with eval. I assume you must be doing this in PH but we have no idea how you've defined PH. — Chris Rackauckas

PRobbe PRobbe · Accepted Answer · 2017-11-15T20:27:31

I recently had a similar issue with pmap() in 0.6. Try to assign the argument f in pmap(f,c...) to a concrete function, i.e.,

createPH(s) = PH(k,s,data,Lines,Ag,Gx,Pmax[:,s],Prmax[:,s],COpe[:,])
pmap(createPH,1:nS)

This fixed the issue for me. (Also note that in 0.6.0 a warning is generated, and not a world age error)

Julia 0.6 pmap function

1 Answers