Parallel example fails on julia 1.1

[davidanthoff]

This is what I get with master:

~\.julia\dev\BlackBoxOptim\examples [master ≡]> julia .\rosenbrock_parallel.jl
Starting optimization with optimizer XNESOpt{Float64,RandomBound{ContinuousRectSearchSpace}}
0.00 secs, 0 evals, 0 steps
sigma=1.0 |trace(ln_B)|=0.0
0.54 secs, 104 evals, 8 steps, fitness=122353.746434742
sigma=0.9982051576798082 |trace(ln_B)|=5.637851296924623e-18
1.07 secs, 247 evals, 19 steps, fitness=32111.868584104
sigma=0.9931794473616752 |trace(ln_B)|=5.204170427930421e-17
1.60 secs, 416 evals, 32 steps, fitness=15074.967363068
sigma=0.9852835141237257 |trace(ln_B)|=6.591949208711867e-17
2.11 secs, 572 evals, 44 steps, fitness=15074.967363068
sigma=0.9784491185053312 |trace(ln_B)|=5.898059818321144e-17
2.64 secs, 728 evals, 56 steps, fitness=12181.224815595
sigma=0.9693998255007409 |trace(ln_B)|=4.5102810375396984e-17
3.18 secs, 897 evals, 69 steps, fitness=11200.377032287
sigma=0.9611088605732498 |trace(ln_B)|=5.551115123125783e-17
3.70 secs, 1053 evals, 81 steps, fitness=11200.377032287
sigma=0.9541631679900614 |trace(ln_B)|=-8.326672684688674e-17
4.21 secs, 1209 evals, 93 steps, fitness=11200.377032287
sigma=0.947030998226623 |trace(ln_B)|=-2.0816681711721685e-17
4.72 secs, 1365 evals, 105 steps, fitness=10060.451657270
sigma=0.9395602410561272 |trace(ln_B)|=-4.163336342344337e-17
5.24 secs, 1521 evals, 117 steps, fitness=10060.451657270
sigma=0.9332730422696129 |trace(ln_B)|=-9.71445146547012e-17
5.77 secs, 1677 evals, 129 steps, fitness=10060.451657270
sigma=0.9258708463911819 |trace(ln_B)|=-9.71445146547012e-17
6.29 secs, 1833 evals, 141 steps, fitness=10060.451657270
sigma=0.9196027642968452 |trace(ln_B)|=-2.0816681711721685e-16
6.79 secs, 1989 evals, 153 steps, fitness=10060.451657270
sigma=0.9133177227106547 |trace(ln_B)|=-1.942890293094024e-16
7.33 secs, 2158 evals, 166 steps, fitness=9373.475368805
sigma=0.9069943095862962 |trace(ln_B)|=-4.163336342344337e-17
7.87 secs, 2327 evals, 179 steps, fitness=9373.475368805
sigma=0.9023805920988747 |trace(ln_B)|=-8.326672684688674e-17
8.38 secs, 2483 evals, 191 steps, fitness=9373.475368805
sigma=0.8954902893404368 |trace(ln_B)|=-1.3877787807814457e-16
8.91 secs, 2652 evals, 204 steps, fitness=8583.999338502
sigma=0.8892181396948056 |trace(ln_B)|=-2.498001805406602e-16
9.45 secs, 2808 evals, 216 steps, fitness=8583.999338502
sigma=0.8820132750980565 |trace(ln_B)|=-1.3877787807814457e-16
9.98 secs, 2977 evals, 229 steps, fitness=8583.999338502
sigma=0.8731894893256401 |trace(ln_B)|=-5.551115123125783e-17
10.50 secs, 3133 evals, 241 steps, fitness=7489.584123589
sigma=0.8679233801696926 |trace(ln_B)|=-2.498001805406602e-16
11.00 secs, 3289 evals, 253 steps, fitness=7489.584123589
sigma=0.8608442649100014 |trace(ln_B)|=-2.498001805406602e-16
11.52 secs, 3445 evals, 265 steps, fitness=7122.457056131
sigma=0.855225980069676 |trace(ln_B)|=-1.3877787807814457e-16
12.02 secs, 3601 evals, 277 steps, fitness=7122.457056131
sigma=0.8480777088289297 |trace(ln_B)|=-1.942890293094024e-16
12.54 secs, 3757 evals, 289 steps, fitness=6015.103831160
sigma=0.8411899889586818 |trace(ln_B)|=-1.942890293094024e-16
13.07 secs, 3913 evals, 301 steps, fitness=6015.103831160
sigma=0.8342904768757995 |trace(ln_B)|=-2.498001805406602e-16
13.61 secs, 4082 evals, 314 steps, fitness=4780.282264771
sigma=0.8269620765403849 |trace(ln_B)|=-2.498001805406602e-16
14.12 secs, 4238 evals, 326 steps, fitness=4780.282264771
sigma=0.8205543861803173 |trace(ln_B)|=-1.3877787807814457e-16
14.63 secs, 4394 evals, 338 steps, fitness=4780.282264771
sigma=0.8130142586751085 |trace(ln_B)|=-2.7755575615628914e-16
15.13 secs, 4550 evals, 350 steps, fitness=4780.282264771
sigma=0.8085303307890455 |trace(ln_B)|=-3.608224830031759e-16
15.64 secs, 4706 evals, 362 steps, fitness=4780.282264771
sigma=0.8027150651112601 |trace(ln_B)|=-1.3877787807814457e-16
16.15 secs, 4862 evals, 374 steps, fitness=4780.282264771
sigma=0.7948282044027354 |trace(ln_B)|=-3.0531133177191805e-16

Optimization stopped after 385 steps and 16.61 seconds
Termination reason: Max number of function evaluations (5000) reached
Steps per second = 23.18
Function evals per second = 301.32
Improvements/step = NaN
Total function evaluations = 5005


Best candidate found: [-0.117691, -0.549153, 1.62825, 0.128071, -0.49382, -0.298501, -0.650996, 0.40877, -0.107746, 0.896346, -0.535518, 0.178234, -1.1461, 1.14015, -0.72947, -0.404426, 0.506873, -0.00404523, 0.141733, -0.0141665, 0.475099, 0.851817, -0.153351, 0.702384, 0.593715, -0.610469, 1.86743, 1.31347, -0.0329012, 0.936947, -0.180541, 0.624921, -0.337358, 0.142574, -0.162826, -0.544692, -0.946747, -0.284673, 0.656164, -0.553401, -0.461865, 0.00229038, 0.0829926, 0.780216, 0.999123, -0.903184, -0.0683573, -0.167501, 0.660591, 1.78841]

Fitness: 4780.282264771

ERROR: LoadError: StackOverflowError:
deserialize(::Distributed.ClusterSerializer{Sockets.TCPSocket}, ::Type{RemoteChannel{Channel{BlackBoxOptim.ParallelEvaluatorWorker{FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}}}}}) at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.1\Distributed\src\remotecall.jl:310 (repeats 100 times)
Stacktrace:
 [1] #remotecall_fetch#149(::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}}, ::Function, ::Function, ::Distributed.Worker, ::Function, ::Vararg{Any,N} where N) at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.1\Distributed\src\remotecall.jl:379
 [2] remotecall_fetch(::Function, ::Distributed.Worker, ::Function, ::Vararg{Any,N} where N) at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.1\Distributed\src\remotecall.jl:371
 [3] #remotecall_fetch#152 at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.1\Distributed\src\remotecall.jl:406 [inlined]
 [4] remotecall_fetch at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.1\Distributed\src\remotecall.jl:406 [inlined]
 [5] Type at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.1\Distributed\src\remotecall.jl:108 [inlined]
 [6] #48 at .\none:0 [inlined]
 [7] iterate at .\generator.jl:47 [inlined]
 [8] collect(::Base.Generator{Array{Int64,1},getfield(BlackBoxOptim, Symbol("##48#50")){FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing},FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}}}) at .\array.jl:606
 [9] #ParallelEvaluator#47(::Array{Int64,1}, ::Type, ::FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}, ::TopListArchive{Float64,ScalarFitnessScheme{true}}) at C:\Users\david\.julia\dev\BlackBoxOptim\src\parallel_evaluator.jl:126
 [10] (::getfield(Core, Symbol("#kw#Type")))(::NamedTuple{(:pids,),Tuple{Array{Int64,1}}}, ::Type{BlackBoxOptim.ParallelEvaluator}, ::FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}, ::TopListArchive{Float64,ScalarFitnessScheme{true}}) at .\none:0
 [11] make_evaluator(::FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}, ::Nothing, ::DictChain{Symbol,Any}) at C:\Users\david\.julia\dev\BlackBoxOptim\src\opt_controller.jl:12
 [12] BlackBoxOptim.OptRunController(::XNESOpt{Float64,RandomBound{ContinuousRectSearchSpace}}, ::FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}, ::DictChain{Symbol,Any}) at C:\Users\david\.julia\dev\BlackBoxOptim\src\opt_controller.jl:97
 [13] run!(::BlackBoxOptim.OptController{XNESOpt{Float64,RandomBound{ContinuousRectSearchSpace}},FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}}) at C:\Users\david\.julia\dev\BlackBoxOptim\src\opt_controller.jl:435
 [14] #bboptimize#84 at C:\Users\david\.julia\dev\BlackBoxOptim\src\bboptimize.jl:66 [inlined]
 [15] bboptimize(::BlackBoxOptim.OptController{XNESOpt{Float64,RandomBound{ContinuousRectSearchSpace}},FunctionBasedProblem{ScalarFitnessScheme{true},ContinuousRectSearchSpace,Nothing}}) at C:\Users\david\.julia\dev\BlackBoxOptim\src\bboptimize.jl:63
 [16] top-level scope at util.jl:213
 [17] include at .\boot.jl:326 [inlined]
 [18] include_relative(::Module, ::String) at .\loading.jl:1038
 [19] include(::Module, ::String) at .\sysimg.jl:29
 [20] exec_options(::Base.JLOptions) at .\client.jl:267
 [21] _start() at .\client.jl:436
in expression starting at C:\Users\david\.julia\dev\BlackBoxOptim\examples\rosenbrock_parallel.jl:28

[isentropic]

@robertfeldt, I was wondering if there is anyone working on this, should we expect a fix anytime soon? One of my projects would be vastly sped up if this works.

[robertfeldt]

Hmm, parallel/distributed use of Julia is not my main competence so it would probably be more efficient if someone with more experience in this looked at it. Maybe @alyst has some input? If not I'll try to get to it soon and just investigate from scratch. Doesn't look too hard to debug.

[wallacmj]

Seems to be related to this issues JuliaLang/julia#30679

You can get around this by adding the following line before calling the optimizer.

for i in procs()
    @everywhere struct A end
    @fetchfrom i A()
end

[robertfeldt]

Hmm, I tried both with and without this supposed fix but Julia 1.1 dies with Bus error: 10 on my MB Pro.

[wallacmj]

Might not work for 1.1. It works for me on the platform below.

Julia Version 1.0.3
Commit 099e826241 (2018-12-18 01:34 UTC)
Platform Info:
OS: Linux (x86_64-pc-linux-gnu)
CPU: Intel(R) Core(TM) i7-4720HQ CPU @ 2.60GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-6.0.0 (ORCJIT, haswell)

[ayadlin]

for i in procs()
    @everywhere struct A end
    @fetchfrom i A()
end

Tried this and it worked - Would anyone care to explain what this spinet is doing?
Is it just fetching whatever is in process i and integrating with the rest?

Thanks,
A