Threads are not speeding up computation a lot

General Usage Performance
8

here is an attempt :


using Base.Threads
using Base.Iterators
using BenchmarkTools
# include("./secp256k1.jl")

function gen_keys_range(k_range, out_matrix=nothing)
    if isnothing(out_matrix)
        data_size = length(k_range)
        keys = zeros(BigInt, data_size, 2)
    else
        keys = out_matrix
    end
    for (i, k) in enumerate(k_range)
        x,y = big(1), big(1)
        keys[i,1] = x
        keys[i,2] = y
    end
    return keys, k_range
end

function gen_keys_1(k_start, k_num)
    nth = Threads.nthreads()
    keys = zeros(BigInt, k_num, 2)
    k_end = k_start + k_num - 1
    part_size = div(k_num, nth)
    tasks = [Threads.@spawn(gen_keys_range(k_r)) for k_r in partition(k_start:k_end, part_size)]
    results = fetch.(tasks)
    for res in results
        k, k_r = res
        v_start = k_r.start - k_start + 1
        v_stop = k_r.stop - k_start + 1
        @views keys[v_start:v_stop,:] = k
    end

    return keys

end

nth = Threads.nthreads()
println("Threads num $nth")

function test_gen_keys_range()
    gen_keys_1(123,10000)
end


@btime test_gen_keys_range()
""

giving

> julia -t 1 --project main.jl
Threads num 1
  655.200 μs (40045 allocations: 1.07 MiB)
> julia -t auto --project main.jl
Threads num 20
  256.800 μs (40649 allocations: 1.09 MiB)

I don’t have your file so I just made two bigint for the test, results are stable after 4 threads because of memory