Is using parallel computing to create JuMP constraints possible and useful?

Specific Domains Optimization (Mathematical)
, ,
1

I was wondering if there is any advantage to generate JuMP constraints with @threads regarding the CPU times.

    Threads.@threads for c in set_curtailments
        @constraint(model, sum(x[:, c]) .<= y[c] * nrow(data_sites))
        @constraint(model, sum(x[s, c] * data_sites[s, 2] for s in set_sites) >= pars.powerContractualized * y[c])
    end

and equivalent formulation with container constraints:

        @constraint(model, [c in set_curtailments], sum(x[:, c]) .<= y[c] * nrow(data_sites))
        @constraint(model, [c in set_curtailments], sum(x[s, c] * data_sites[s, 2] for s in set_sites) >= pars.powerContractualized * y[c])

I would intuitively expect the first to be much faster if the number of CPU cores available is greater than 2.

Also, I could not test it because I get:

[7608] signal (11.1): Erreur de segmentation which is a segmentation fault when using the Threads.

2

You cannot use parallelism when building a JuMP model