Example in Multiple GPUs doc fails

[jeffcbecker]

Describe the bug

A clear and concise description of what the bug is.

Tried to run first example in https://juliagpu.gitlab.io/CUDA.jl/usage/multigpu/ and it fails
To reproduce

julia> using Distributed, CUDA

julia> addprocs(length(devices()))
4-element Array{Int64,1}:
 2
 3
 4
 5

julia> asyncmap((zip(workers(), devices()))) do (p, d)
           remotecall_wait(p) do
               @info "Worker $p uses $d"
               device!(d)
           end
       end
ERROR: On worker 2:
KeyError: key CUDA [052768ef-5323-5732-b1bb-66c8b64840ba] not found
getindex at ./dict.jl:477 [inlined]
root_module at ./loading.jl:967 [inlined]
deserialize_module at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:912
handle_deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:812
deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:735
deserialize_datatype at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:1210
handle_deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:788
deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:735
deserialize_datatype at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:1235
handle_deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:790
deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:735
handle_deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:795
deserialize at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Serialization/src/Serialization.jl:735 [inlined]
deserialize_msg at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Distributed/src/messages.jl:99
#invokelatest#1 at ./essentials.jl:712 [inlined]
invokelatest at ./essentials.jl:711 [inlined]
message_handler_loop at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Distributed/src/process_messages.jl:185
process_tcp_streams at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.4/Distributed/src/process_messages.jl:142
#97 at ./task.jl:358
Stacktrace:
 [1] (::Base.var"#726#728")(::Task) at ./asyncmap.jl:178
 [2] foreach(::Base.var"#726#728", ::Array{Any,1}) at ./abstractarray.jl:1919
 [3] maptwice(::Function, ::Channel{Any}, ::Array{Any,1}, ::Base.Iterators.Zip{Tuple{Array{Int64,1},CUDA.DeviceSet}}) at ./asyncmap.jl:178
 [4] wrap_n_exec_twice at ./asyncmap.jl:154 [inlined]
 [5] #async_usemap#711 at ./asyncmap.jl:103 [inlined]
 [6] #asyncmap#710 at ./asyncmap.jl:81 [inlined]
 [7] asyncmap(::Function, ::Base.Iterators.Zip{Tuple{Array{Int64,1},CUDA.DeviceSet}}) at ./asyncmap.jl:81
 [8] top-level scope at REPL[17]:1

julia> 

The Minimal Working Example (MWE) for this bug:

<details><summary>Manifest.toml</summary>
<p>

Paste your Manifest.toml here, or accurately describe which version of CUDA.jl and its dependencies (GPUArrays.jl, GPUCompiler.jl, LLVM.jl) you are using.

(@v1.4) pkg> status
Status `~/.julia/environments/v1.4/Project.toml`
  [6e4b80f9] BenchmarkTools v0.5.0
  [052768ef] CUDA v0.1.0
  [c5f51814] CUDAdrv v6.3.0
  [be33ccc6] CUDAnative v3.0.4
  [5ae59095] Colors v0.12.3
  [3a865a2d] CuArrays v2.2.2
  [5789e2e9] FileIO v1.3.0
  [0c68f7d7] GPUArrays v3.4.1
  [4e3cecfd] ImageShow v0.2.3
  [d0351b0e] InspectDR v0.3.7
  [da04e1cc] MPI v0.15.0
  [3b7a836e] PGFPlots v3.2.1
  [8314cec4] PGFPlotsX v1.2.8
  [91a5bcdd] Plots v1.5.3
  [d330b81b] PyPlot v2.9.0
  [b8865327] UnicodePlots v1.2.0

</p>
</details>

Expected behavior

A clear and concise description of what you expected to happen.

example should run, not give stack trace

Version info

Details on Julia:

please post the output of:

versioninfo()

Julia Version 1.4.2
Commit 44fa15b150* (2020-05-23 18:35 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-8.0.1 (ORCJIT, skylake)

Details on CUDA:

please post the output of:

CUDA.versioninfo()

v"10.2.0"

Additional context

Add any other context about the problem here.

[kshyatt]

Fixed the backticks so code should be a little more readable

[maleadt]

Should include @everywhere using CUDA