It's possible that our PBS max_procs function has bit-rotted...?
Have a look in lib/MTT/Values/Functions.pm:pbs_max_procs(). See if it
calculates the wrong thing under your PBS jobs...
On Oct 2, 2013, at 6:42 AM, Christoph Niethammer wrote:
> Hello.
>
> The &env_max_procs() does not seem to work in my setup.
> MTT is run from within a PBS job which should be supported.
> If I set np in the configuration to a fixed value, tests run propperly. :/
>
> Here the config section:
>
> [Test run]
> pass = &and(&cmd_wifexited(), &eq(&cmd_wexitstatus(), 0))
> timeout = &max(30, &multiply(10, &test_np()))
> save_stdout_on_pass = 1
> merge_stdout_stderr = 1
> stdout_save_lines = 1000
> np = &env_max_procs()
> #np = 64
> specify_module = Simple
>
> [Test run: mpi_test_suite ]
> test_build = hlrs_mpi_test_suite
> include_section = Test run
> specify_module = Simple
> simple_pass:tests = mpi_test_suite
>
> And here the error message where np is empty during execution:
>
> Running with [ompi-nightly-v1.6] / [1.6.6a1r29300] / [gcc warnings]
> Using MPI Details [open mpi] with MPI Install [gcc warnings]
Using group_reports
> ###
> ### Total tests to run in this section:
> ### 1 test executable(s)
> ### 1 np value(s)
> ### 1 argv value(s)
> ### 1 test variant(s)
> ### 1 total mpirun command(s) to run
> ###
> *** WARNING: Test: mpi_test_suite -t "All,^io,^one-sided", np=, variant=1:
>FAILED
>
> Any idea where this problem comes from?
>
>
> Best regards
> Christoph Niethammer
> ___
> mtt-users mailing list
> mtt-us...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/mtt-users
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/