Month: May 2018

Julia is particularly attractive for its ability to run in parallel, not only on clusters, but also on the multiple cores of your laptop processor. Here is a minimalistic example using pmap().

# For parallel computing with julia -p NumberOfAdditionalCPU

@everywhere begin # this part will be available on all CPUs

function computation(runid_is_useless)
  result = rand()
  return result
end # function computation()

end # @everywhere

### Main part - runs only on main instance of Julia.

Nprocs = nprocs()
print("Number of processors ",Nprocs,".\n")

## Launching computations on Nprocs parallel processors
results = @time pmap(computation,1:Nprocs)

## Post-processing
for i in 1:Nprocs
  print(@sprintf("Result of CPU %i is %f\n",i,results[i]))
end # for i

On my laptop, my processor has 4 cores, and I run this code with julia -p 3 code.jl which gives

Number of processors 4.
0.548238 seconds (214.75 k allocations: 9.788 MB, 5.76% gc time)
Result of CPU 1 is 0.725449
Result of CPU 2 is 0.377958
Result of CPU 3 is 0.443606
Result of CPU 4 is 0.016641

The same code with -p 39 can also be run on the 40 CPUs cluster of my department! For more material on parallel computing with Julia, you may take a look at Julia documentation.