jit doesn't like threading?!?

You're probably right, and there are probably other
Python parts of FEniCS that are not threadsafe.

The thing is, Python threads are not really concurrent,
so you may not get the speedup you're looking for
anyway without significant work... Therefore, thread
safety has not really been an issue.

However, the jit is designed to work with multiple
processes. Maybe you'll have better luck with
the multiprocessing module, or just use the existing
MPI support.

Martin

On 30 August 2011 10:35, Thomas Fraunholz
<email address hidden> wrote:
> New question #169569 on DOLFIN:
> https://answers.launchpad.net/dolfin/+question/169569
>
> Hi,
> I tried to distribute my computations on several threads using the python threading module. Everything works fine using a single thread. But it seems to me that jit is somehow confused by adding an additional thread. As I am not quite well informed about the jit module I would like to know if somebody has already seen this before or knows about this constraint. It would help a lot if somebody knows for sure that this should work fine (or perhaps not).
> Thanks a lot!
> Thomas
>
>
> --------
>> python main.py
> Thread-1 working!
> Thread-2 working!
> Solving linear variational problem.
> Solving linear variational problem.
> 3 cells out of 6 marked for refinement (50.0%).
> Number of cells increased from 6 to 10 (66.7% increase).
> Exception in thread Thread-1:
> Traceback (most recent call last):
>  File "/usr/lib/python2.7/threading.py", line 552, in __bootstrap_inner
>    self.run()
>  File ".../methods.py", line 74, in run
>    eta, eta_list = afem_estimate(param, mesh, u_h, f)
>  File ".../methods.py", line 206, in afem_estimate
>    eta_E = assemble(eta_e, mesh=space.mesh())
>  File "/usr/lib/python2.7/dist-packages/dolfin/fem/assembling.py", line 157, in assemble
>    common_cell=common_cell)
>  File "/usr/lib/python2.7/dist-packages/dolfin/fem/form.py", line 54, in __init__
>    common_cell)
>  File "/usr/lib/python2.7/dist-packages/dolfin/compilemodules/jit.py", line 62, in mpi_jit
>    return local_jit(*args, **kwargs)
>  File "/usr/lib/python2.7/dist-packages/dolfin/compilemodules/jit.py", line 139, in jit
>    return jit_compile(form, parameters=p, common_cell=common_cell)
>  File "/usr/lib/python2.7/dist-packages/ffc/jitcompiler.py", line 73, in jit
>    return jit_form(object, parameters, common_cell)
>  File "/usr/lib/python2.7/dist-packages/ffc/jitcompiler.py", line 102, in jit_form
>    module = instant.import_module(jit_object, cache_dir=cache_dir)
>  File "/usr/lib/python2.7/dist-packages/instant/cache.py", line 156, in import_module
>    return check_disk_cache(modulename, cache_dir, moduleids)
>  File "/usr/lib/python2.7/dist-packages/instant/cache.py", line 126, in check_disk_cache
>    release_lock(lock)
>  File "/usr/lib/python2.7/dist-packages/instant/locking.py", line 46, in release_lock
>    lockname = _lock_names[lock.fileno()]
> ValueError: I/O operation on closed file
>
> Solving linear variational problem.
> 5 cells out of 10 marked for refinement (50.0%).
> Number of cells increased from 10 to 18 (80.0% increase).
> Solving linear variational problem.
> 8 cells out of 18 marked for refinement (44.4%).
> Number of cells increased from 18 to 30 (66.7% increase).
> Solving linear variational problem.
> 12 cells out of 30 marked for refinement (40.0%).
> Number of cells increased from 30 to 57 (90.0% increase).
>
> --
> You received this question notification because you are a member of
> DOLFIN Team, which is an answer contact for DOLFIN.
>

Thanks for the helpfull advice! Multiprocessing works fine.