Document Thread Safety

[adomasbaliuka]

In which cases is sampling thread-safe? It seems that (at least) sampling from Hypergeometric apparently isn't.

This might be due to this C-code in Rmath-julia. It should be documented if sampling from distributions is thread-safe.

Below example runs fine without the @threads macro. With the macro, it seemingly runs forever and prints warnings to the console (which seem to come from that C library Rmath-julia).

using Random, Distributions

Random.seed!(42)

function sample_KkC(n; N, Q)
    total_errors = Distributions.Binomial(N, Q)
    K = rand(total_errors)
    d_hyper = Hypergeometric(K, N - K, n)
    k = rand(d_hyper)

    return k
end

function main()
    function f(Q)
        objective(n) = [sample_KkC(n; N = 819_200, Q) for _ = 1:100]
        vals = [10, 100]
        objective.(vals)
    end

    Qs = [0.05, 0.055, 0.1, 0.2, 0.3]

    Threads.@threads for i in eachindex(Qs)
        f(Qs[i])
    end
end

main()

output

rhyper.c: afc(i), i=-15 < 0 -- SHOULD NOT HAPPEN!
rhyper(*, n1=-262145576, n2=100, k=10000): branch III: giving up after 1384304160 rejections
rhyper(*, n1=-287340072, n2=100, k=10000): branch III: giving up after 1384304160 rejections

System Info

julia> versioninfo()
Julia Version 1.10.0
Commit 3120989f39b (2023-12-25 18:01 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 8 × 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, tigerlake)
  Threads: 11 on 8 virtual cores

(jl_Lv9tPw) pkg> st
Status `/tmp/jl_Lv9tPw/Project.toml`
  [31c24e10] Distributions v0.25.107

[adomasbaliuka]

Should be fixed by JuliaStats/Rmath-julia#46 but does not yet work correctly on Distributions v0.25.110. Will check when updated further.