[Octopus-users] broken pipe? failed and skipped parallel tests in make check
David Strubbe
dstrubbe at berkeley.edu
Thu Feb 4 18:49:04 WET 2010
Emily,
1) The 'broken pipe' messages are more or less meaningless and are due to a
stupidity of 'grep' and 'cat' (or how we use them).
2) If it can't find the file 'out', it is probably because the preceding
calculation failed. I don't know why there would be a problem finding 'could
not find input file: ./finite_systems_3d/21-lobpcg.02-spinors.inp'.
3) Tests are skipped when the appropriate executable is not present, which
is not an error. Some tests are only run for the single-precision
executable, or in serial, so if you haven't built those executables the
tests will be skipped.
David
On Thu, Feb 4, 2010 at 7:18 AM, Emily <listemily at eml.cc> wrote:
> Hi Octopus Users,
>
> Now that my configure, make, and make install commands seem to go okay,
> I have a question about make check.
>
> I'm installing the parallel version of Octopus 3.2.0. The non-parallel
> version installed fine, and passed all the tests in make check.
>
> The parallel version passes 34/38 tests on make check. It has a bunch
> of broken pipe error messages. It fails 2 of the testcases of
> finite_systems_3d/21-lobpcg.test. In the /04-propagate_eigenstate.test
> it can't find the file "out". It skips 3 tests because it says it can't
> find executables.
> (./finite_systems_3d/02-curvilinear_coordinates.test,
> ./optimal_control/01-asym_doublewell.test,
> ./real_time/04-etrs_single.test)
>
> The broken pipe messages seem to be interspersed randomly (not in the
> tests that failed), but if I run make check again, they occur in exactly
> the same places.
>
> Anybody know how to fix these problems?
>
> I am running the make check from my home directory that every node can
> see and write to, rather than my usual way of copying the executable and
> inp to a /tmp directory on the relevant node. Is this the correct thing
> to do? Will I have to run octopus_mpi this way also? Is there any more
> in-depth information about the way the parallelization works that I
> should read other than the brief overview in the manual and the variable
> reference?
>
> Here are some relevant snippets of the output of my makecheck:
>
> No valid executable found for
> ./finite_systems_3d/02-curvilinear_coordinates.test
> Skipping ...
>
> ***** Propagate eigenstate of 1D square well *****
>
> Using workdir :
> /home/et/octopus/Oct3/octopus-3.2.0/tmp/octopus.7Qt65G
> Using executable :
> /home/et/octopus/Oct3/octopus-3.2.0/testsuite/../src/main/octopus_mpi
> Using test file : ./open_systems/04-propagate_eigenstate.test
> cat: /home/et/octopus/Oct3/octopus-3.2.0/tmp/octopus.7Qt65G/out: No such
> file or directory
>
> ***** Asymmetric Double Well *****
>
> No valid executable found for ./optimal_control/01-asym_doublewell.test
> Skipping ...
>
> No valid executable found for ./real_time/04-etrs_single.test
> Skipping ...
>
> ***** LOBPCG (parallel) *****
>
> Using workdir :
> /home/et/octopus/Oct3/octopus-3.2.0/tmp/octopus.lXurFV
> Using executable :
> /home/et/octopus/Oct3/octopus-3.2.0/testsuite/../src/main/octopus_mpi
> Using test file : ./finite_systems_3d/21-lobpcg.test
>
> Using input file : ./finite_systems_3d/21-lobpcg.01-spin_polarized.inp
>
> Starting test run ...
> Executing: cd /home/et/octopus/Oct3/octopus-3.2.0/tmp/octopus.lXurFV;
> /usr/local/intel/mpich/64/bin/mpirun -n 2
> /home/et/octopus/Oct3/octopus-3.2.0/testsuite/../src/main/octopus_mpi >
> out
> Finished test run.
>
> Eigenvalue 1 up: [ OK ]
> Eigenvalue 1 dn: [ OK ]
> Eigenvalue 2 up: [ OK ]
> Eigenvalue 2 dn: [ OK ]
> Eigenvalue 3 up: [ OK ]
> Eigenvalue 3 dn: [ OK ]
> Eigenvalue 4 up: [ OK ]
> could not find input file: ./finite_systems_3d/21-lobpcg.02-spinors.inp
> Eigenvalue 4 dn: [ OK ]
> _______________________________________________
> Octopus-users mailing list
> Octopus-users at tddft.org
> http://www.tddft.org/mailman/listinfo/octopus-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.tddft.org/pipermail/octopus-users/attachments/20100204/6df32bc6/attachment.html
More information about the Octopus-users
mailing list