Groups


75 of 99+  



julia-users ›
Passing types in @parallel
2 posts by 1 author  



Jason Castiglione 	

11/19/15


I constructed my own type to store LDPC codes.
The module is at
https://github.com/jasontuna/LDPC

It is stored in a Julia module and I load it with
@everywhere using LDPC

I would like to run several simulations in parallel calculating bit error rates for passing codewords through a particular channel. I am getting errors where functions defined in my notebook won't "notice" the LDPC type.

I cut and paste the code for the type in  a notebook, and then load the type with @everywhere and it works then.
What should I do to get the type to load correctly by just using @everywhere for the module?

Here is cut and paste of two simplified examples that have the same error:
1) I write the following in the IJulia notebook

addprocs(6)
@everywhere using Distributions
@everywhere using LDPC

LH = ldpcH(ones(UInt8,1000)*3,ones(UInt8,500)*6,true);

@everywhere function llh(LH::ldpcH)
    return LH.n
end

@parallel (+) for j=1:100
    llh(LH)
end

###############################
and the error I get for each worker looks like 
MethodError: `llh` has no method matching llh(::LDPC.ldpcH)
Closest candidates are:
  llh(::LDPC.ldpcH)
 [inlined code] from In[10]:3
 in anonymous at no file:1500
 in anonymous at multi.jl:892
 in run_work_thunk at multi.jl:645
 [inlined code] from multi.jl:892
 in anonymous at task.jl:63
while loading In[10], in expression starting on line 2



2) on the other hand if I write

addprocs(6)
@everywhere using Distributions

@everywhere type ldpcH
	#for readability I did not cut and paste the contents 
end

@everywhere function llh(LH::ldpcH)
    return LH.n
end
LH = ldpcH(ones(UInt8,1000)*3,ones(UInt8,500)*6,true);
@parallel (+) for j=1:100
    llh(LH)
end

everything works.

 


Jason Castiglione 	

11/20/15


Figured it out. For some reason doing the following works,


using LDPC
@everywhere using LDPC 
- show quoted text -