How to run two tasks on parallel?

General Usage Performance
21

As an example, showing the difference between the thread/non-threaded version:

tmin=1;
tmax=10;
D1 = [1.0 2 3 4 5 6 7 8 9 10];
D2 = [10.0 9 8 7 6 5 4 3 2 1];
X = zeros(10);
Y = zeros(10);

function calculate!(X, Y, D1, D2, tmin,tmax)
    for n in tmin+1:tmax
        X[n] = D1[n]*D2[n] + Y[n-1];
        Y[n] = D1[n-1]*D2[n-1] + X[n-1];
    end
    nothing
end

function calculate_threaded!(X, Y, D1, D2, tmin, tmax)
    for n in tmin+1:tmax
        Xn = Threads.@spawn D1[n]*D2[n] + Y[n-1]; 
        Yn = Threads.@spawn D1[n-1]*D2[n-1] + X[n-1];
        X[n] = fetch(Xn)
        Y[n] = fetch(Yn)
    end
    nothing
end

We benchmark to see:

using BenchmarkTools
@btime calculate!(X, Y, D1, D2, tmin, tmax)
# 31.539 ns (0 allocations: 0 bytes)
@btime calculate_threaded!(X, Y, D1, D2, tmin, tmax)
# 41.249 μs (125 allocations: 9.53 KiB)

Using locks will likely only make this worse. You will only see a performance benefit if the calculations in “Threads.@spawn” is significant (like a big matrix calculation).

1 Like