dolfin team mailing list archive

[Johannes Ring <johannr@xxxxxxxxx>]

On Tue, Sep 22, 2009 at 9:56 AM, Garth N. Wells <gnw20@xxxxxxxxx> wrote:
>
>
> Anders Logg wrote:
>> On Mon, Sep 21, 2009 at 10:18:21PM +0200, Johan Hake wrote:
>>> On Monday 21 September 2009 22:00:54 Anders Logg wrote:
>>>> On Tue, Sep 15, 2009 at 05:15:39PM +0200, Anders Logg wrote:
>>>>> We're getting closer. All demos are running in parallel now (or
>>>>> exiting gracefully).
>>>>>
>>>>> It only remains to
>>>>>
>>>>>   1. Measure speedup
>>>> Done.
>>>>
>>>> We need to discuss the results further but I don't think it's a
>>>> blocker for a release.
>>>>
>>>>>   2. Get the buildbot green
>>>> In progress.
>>> I suppose it is no use running the parallel tests for Windows or on the Mac?
>>> The latter seems to lack parmetis?
>>
>> Yes, it should be enough for now to run these on one of the Linux
>> platforms.
>>
>
> I've just pushed a change which fixes the Trilinos Python demo, so all
> the demos should run ok on hardy-i386 and jaunty-amd64. Demos are
> failing in parallel on linux64-exp because PETSc does not have a
> parallel LU solver on that machine
>
> I've relaunched hardy-i386 and jaunty-amd64, and all going will they
> will be green so we'll be ready for a release.

Apart from the memory failures reported by valgrind, the only test
that stops the buildbot from turning green seems to be the
parallel-assembly-solve system test. Here is the output I get when I
run test/system/parallel-assembly-solve/solver.py:

Checking results
----------------
(unitsquare.xml.gz, 1):	OK (norm = 9.547454087327647, reference =
9.547454087327344, diff = 3.037570195374428e-13)
(unitsquare.xml.gz, 2):	OK (norm = 18.42366670418048, reference =
18.42366670418527, diff = 4.785505325344275e-12)
(unitsquare.xml.gz, 3):	OK (norm = 27.29583104739728, reference =
27.29583104741712, diff = 1.98419058961008e-11)
(unitsquare.xml.gz, 4):	*** ERROR (norm = 36.16867128111434, reference
= 36.1686712809094, diff = 2.049418412752857e-10)
(unitcube.xml.gz, 1):	OK (norm = 8.876490653853772, reference =
8.876490653853809, diff = 3.730349362740526e-14)
(unitcube.xml.gz, 2):	OK (norm = 19.99081167299559, reference =
19.99081167299566, diff = 7.105427357601002e-14)
(unitcube.xml.gz, 3):	OK (norm = 33.85477561286741, reference =
33.85477561286852, diff = 1.108446667785756e-12)
(unitcube.xml.gz, 4):	OK (norm = 49.97357666763575, reference =
49.97357666762962, diff = 6.131983809609665e-12)

Johannes

> Garth
>
>
>>> We have defined the Python function has_mpi() and has_parmetis(), if PyDOLFIN
>>> has been compiled with these. We can include a test for these in the test
>>> script ruling out any unintentional test runs.
>>
>> Nice.
>>
>> --
>> Anders
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> DOLFIN-dev mailing list
>> DOLFIN-dev@xxxxxxxxxx
>> http://www.fenics.org/mailman/listinfo/dolfin-dev
>
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev
>