pmap_memory_usage.jl

addprocs(4)
blas_set_num_threads(1)

function f(K, k, y, masks, ranges)
    function dowork(i)
        z = K[masks[:, i], masks[:, i]] \ y[ranges[i, 1]:ranges[i, 2]]
        z = k[:, ranges[i, 1]:ranges[i, 2]] * z
        return z
    end
    r = pmap(dowork, collect(1:size(ranges, 1)))
    return r
end

n = 4000
m = 100

K = randn(n, n)
K = K' * K
K = K + eye(n)

masks = rand(Bool, n, m)
sizes = vec(sum(masks, 1))

k = randn(m, sum(sizes))
y = randn(sum(sizes))

ranges = hcat(vcat(0, cumsum(sizes)[1:end-1]) .+ 1, cumsum(sizes))

K = convert(SharedArray, K)
masks = convert(SharedArray, masks)
k = convert(SharedArray, k)
y = convert(SharedArray, y)
ranges = convert(SharedArray, ranges)

r = f(K, k, y, masks, ranges)