Returning Multiple RemoteRefs Error #9457
Closed
JaredCrean2 opened this Issue on Dec 24, 2014 · 3 comments
Projects
None yet
Labels
doc
error handling
parallel
Milestone
No milestone
Assignees
No one assigned
4 participants
@JaredCrean2
@vtjnash
@malmaud
@jakebolewski
Notifications

You’re not receiving notifications from this thread.
@JaredCrean2
JaredCrean2 commented on Dec 24, 2014

I found that returning multiple values using remotecall produces a (rather strange) error. Does anyone know why this would fail in parallel? Although there are a few workaround (using arrays and cell arrays), it would be very useful to be able to directly reuse functions that work in serial for parallel codes.

Here is the script file Script1.jl that triggers the error:

# script to be executed on process 1
# run with julia -p 1 ./Script1.jl

@everywhere include("funcs.jl")

a = 1
b = 2

ret1 = remotecall(2, func1, a, b)
wait(ret1)

ret2, ret3 = remotecall(2, func2, a, b)         # error
wait(ret2)

and here is the function file funcs.jl:

function func1(c, d)

println("c = ", c)
println("d = ", d)

return c

end


function func2(c, d)

println("c = ", c)
println("d = ", d)

return c, d

end 

Here is the error info:

ERROR: start has no method matching start(::RemoteRef)
in include at ./boot.jl:245
in include_from_node1 at loading.jl:128
in process_options at ./client.jl:285
in _start at ./client.jl:354
in _start_3B_1720 at /usr/bin/../lib/x86_64-linux-gnu/julia/sys.so
while loading /home/dev1/julia_tests/ret_values/Script1.jl, in expression starting on line 12

Julia version info
Julia Version 0.3.3
Commit 21d5433* (2014-10-21 20:18 UTC)
Platform Info:
System: Linux (x86_64-linux-gnu)
CPU: Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
WORD_SIZE: 64
BLAS: libopenblas (NO_LAPACK NO_LAPACKE DYNAMIC_ARCH NO_AFFINITY Nehalem)
LAPACK: liblapack.so.3
LIBM: libopenlibm
LLVM: libLLVM-3.3
@vtjnash
The Julia Language member
vtjnash commented on Dec 25, 2014

remotecall always returns a single RemoteRef

help?> remotecall
Base.remotecall(id, func, args...)

   Call a function asynchronously on the given arguments on the
   specified process. Returns a "RemoteRef".

the error message could perhaps be improved. it happens because of:
a, b = RemoteRef()
gets translated into an iterator over the right hand side
@vtjnash vtjnash added error handling doc labels on Dec 25, 2014
@jakebolewski jakebolewski added the parallel label on Jun 2, 2015
@malmaud
malmaud commented on Oct 14, 2015

This is kinda tricky. The docs for remotecall seems pretty clear and the very first section in the parallel computing manual chapter explains that remotecall returns a RemoteReference. But you do have to understand that returning multiple values just means returning a single tuple whose elements are the individual return values.

@JaredCrean2, I'd be interesting in hearing your thoughts on what could make the documentation clearer.
@JaredCrean2
JaredCrean2 commented on Oct 14, 2015

I didn't realize this issue was still open. I just tested with a more recent version of Julia and found if I change the script to:

# script to be executed on process 1
# run with julia -p 1 ./Script1.jl

@everywhere include("funcs.jl")

a = 1
b = 2

ret1 = remotecall(2, func1, a, b)
wait(ret1)

ret2 = remotecall(2, func2, a, b) 
wait(ret2)
a1,b2 = fetch(ret2)  # unpack the tuple here
println("a1, b1 = ", a1, ", ", b2)

then everything works as expected. Returning a single RemoteRef pointing to a tuple seems like a reasonable way to of doing things combined with the documentation, so I'll close.
@JaredCrean2 JaredCrean2 closed this on Oct 14, 2015