Julia High Performance - Second Edition

Distributed Monte Carlo

In earlier chapters on Threads and GPUs, we have encountered code to calculate the value of Pi using the Monte Carlo simulation. As an example of code that can be easily parallelized, let's try to run this over a distributed Julia cluster. First, we define our function to count the number of hits inside the unit circle: 

@everywhere function darts_in_circle(N)  
    n = 0                      
    for i in 1:N                       
        if rand()^2 + rand()^2 < 1     
            n += 1                     
        end                             
    end                                 
    return n                                  
end

This is then used via pmap to run in parallel on multiple nodes:

function pi_distributed(N, loops)         
    n = sum(pmap((x)->darts_in_circle(N), 1:loops))   
    4 * n / (loops * N)                
end

We also define a serial version to compare timings:

function pi_serial(n)   return 4 * darts_in_circle(n) / nend

Timing ...