# Pmap cannot write into SharedArray

**URL:** <https://discourse.julialang.org/t/pmap-cannot-write-into-sharedarray/3177>\
**Category:** General Usage\
**Tags:** parallel\
**Created:** [April 11, 2017, 11:36pm UTC](https://discourse.julialang.org/t/pmap-cannot-write-into-sharedarray/3177 "2017-04-11T23:36:15Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Arrigo\_Benedetti](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/arrigo_benedetti/32/25545_2.png) [@Arrigo\_Benedetti](https://discourse.julialang.org/u/Arrigo_Benedetti)\
**Post date:** [April 11, 2017, 11:36pm UTC](https://discourse.julialang.org/t/pmap-cannot-write-into-sharedarray/3177/1 "2017-04-11T23:36:15Z")

</div>

Consider the following code:

```julia
nw = nworkers()
@show nw

N=20
@eval @everywhere N = $N

spw = ceil(Int, N / nw)
@show spw

@everywhere yr = SharedArray(Float64, N)
@everywhere yr = 1:N

@everywhere sigma = SharedArray(Float64, N)

@everywhere function dummy_batch(arange, yr, sigma)
    @show arange
    for i in arange
      sigma[i] = yr[i] * yr[i]
      @show sigma[i]
    end
end

my_ranges = [(i:min(i+spw-1,N)) for i = 1:spw:N]
@show my_ranges

@time pmap((arg)->dummy_batch(arg,yr,sigma), my_ranges)

@show sigma
@show sdata(sigma)

```

As you can see from the following log the SharedArray is assigned the correct values by the the workers however after pmap sigma is all zeros. I have used SharedArrays in the past with 0.4.7, I am running this code with 0.5.1. Did anything change in the new version?

-Arrigo

nw = 5  
spw = 4  
my\_ranges = UnitRange{Int64}[1:4,5:8,9:12,13:16,17:20]  
From worker 2: arange = 1:4  
From worker 2: sigma[i] = 1.0  
From worker 5: arange = 9:12  
From worker 4: arange = 5:8  
From worker 2: sigma[i] = 4.0  
From worker 5: sigma[i] = 81.0  
From worker 4: sigma[i] = 25.0  
From worker 2: sigma[i] = 9.0  
From worker 5: sigma[i] = 100.0  
From worker 4: sigma[i] = 36.0  
From worker 2: sigma[i] = 16.0  
From worker 5: sigma[i] = 121.0  
From worker 4: sigma[i] = 49.0  
From worker 5: sigma[i] = 144.0  
From worker 4: sigma[i] = 64.0  
From worker 3: arange = 17:20  
From worker 3: sigma[i] = 289.0  
From worker 3: sigma[i] = 324.0  
From worker 3: sigma[i] = 361.0  
From worker 3: sigma[i] = 400.0  
From worker 6: arange = 13:16  
From worker 6: sigma[i] = 169.0  
From worker 6: sigma[i] = 196.0  
From worker 6: sigma[i] = 225.0  
. From worker 6: sigma[i] = 256.0  
041796 seconds (5.67 M allocations: 240.650 MB, 2.37% gc time)  
sigma = [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0]  
sdata(sigma) = [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0]

---

<div class="post-metadata">

**Author:** ![amit.murthy](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/amit.murthy/32/47_2.png) [@amit.murthy](https://discourse.julialang.org/u/amit.murthy)\
**Post date:** [April 12, 2017, 4:18am UTC](https://discourse.julialang.org/t/pmap-cannot-write-into-sharedarray/3177/2 "2017-04-12T04:18:13Z")

</div>

> [@Arrigo\_Benedetti](#):
>
> @everywhere sigma = SharedArray(Float64, N)

The `@everywhere` will define different instances of the SharedArray on each worker.

Also in 0.5, globals referenced in a closure are not automatically serialized to remote nodes and the pmap will throw UndefVar errors if the globals are not localized.

The following should work:

```julia
nw = nworkers()
@show nw

N=20
@eval @everywhere N = $N

spw = ceil(Int, N / nw)
@show spw

yr = 1:N
sigma = SharedArray(Float64, N)

@everywhere function dummy_batch(arange, yr, sigma)
    @show arange
    for i in arange
      sigma[i] = yr[i] * yr[i]
      @show sigma[i]
    end
end

my_ranges = [(i:min(i+spw-1,N)) for i = 1:spw:N]
@show my_ranges

let yr=yr, sigma=sigma
    @time pmap((arg)->dummy_batch(arg,yr,sigma), my_ranges)
end

@show sigma
@show sdata(sigma)

```