# Nested parallel computation (pmap + @distributed)

**URL:** <https://discourse.julialang.org/t/nested-parallel-computation-pmap-distributed/109075>\
**Category:** General Usage\
**Created:** [January 21, 2024, 9:28pm UTC](https://discourse.julialang.org/t/nested-parallel-computation-pmap-distributed/109075 "2024-01-21T21:28:46Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![structural](https://avatars.discourse-cdn.com/v4/letter/s/34f0e0/32.png) [@structural](https://discourse.julialang.org/u/structural)\
**Post date:** [January 21, 2024, 9:28pm UTC](https://discourse.julialang.org/t/nested-parallel-computation-pmap-distributed/109075/1 "2024-01-21T21:28:46Z")

</div>

I have N datasets (\text{data}^{}\_1,..., \text{data}^{}\_{10}). For each dataset I need to solve \hat{\theta}^{}\_{d}=\arg\min\limits\_\theta h(\theta;\text{data}^{}\_{d}). I am using `pmap` to run the optimizations in parallel for each dataset in `datas` which is a struct which stores the ten datasets. The code looks something like this:

```julia
function compute_h(d,θ)
     Threads.@threads for i in eachindex(d)
           #do something to construct h(d,θ)
     end
     #return h(d,θ)
end

function find_argmin(d)
     # argmin = optimize(compute_h(d,θ), θ_guess)
     # return argmin
end

pool1 = [i for i=1:10]
pmap(d -> find_argmin(h(d)), WorkerPool(pool1), datas)

```

The above code runs without any problems. However, I get the error `ERROR: No active worker available in pool` when I try to parallelize `compute_h` over multiple processors instead of threading it (see below).

```julia
function compute_h(d,θ)
     @distributed for i in eachindex(d)
           #do something to construct h(d,θ)
     end
     #return h(d,θ)
end

```

**My question:** Is there anyway to nest a `@distributed for` loop within a `pmap`? My guess is that at a minimum I will need to specify that the `@distributed for` loop uses workers not already used for the `pmap`. However I have no idea how to specify a `WorkerPool` for the `@distributed for` loop.