NullEvol_InitialData.F90 triggers Fortran exception

Issue #939 closed
Roland Haas created an issue

I ran the SphericalHarmonicReconn testsuite (regression_test.par) with a Lovelace executable compiled via simfactory build --debug and get Fortran out-of bounds exceptions from NullEvolve when running the test (2 processes 2 threads each):

INFO (NullInterp):  the guard point shell is set for a max. stencil size of           3
INFO (NullSHRExtract):  CCE world-tube index range is           4           4   20.404040404040405        20.404040404040405
INFO (NullEvolve): Null Initial Data
At line 269 of file /mnt/data/rhaas/postdoc/gr/ET_Lovelace/arrangements/PITTNullCode/NullEvolve/src/NullEvol_InitialData.F90
Fortran runtime error: Array bound mismatch for dimension 1 of array 'jcn_rad' (15/31)
At line 269 of file /mnt/data/rhaas/postdoc/gr/ET_Lovelace/arrangements/PITTNullCode/NullEvolve/src/NullEvol_InitialData.F90
Fortran runtime error: Array bound mismatch for dimension 1 of array 'jcn_rad' (16/31)

This happens when compiling with gcc 4.6.3 but does not seem to happen when using the intel compiler (though I am not quite sure how to turn on all the bounds checks in it).

Keyword: NullEvolve

Comments (3)

  1. Roland Haas reporter
    • changed status to open
    • removed comment

    The attached patch to NullEvolve fixes the issue for me. Someone who understands how the PITTNull code parallelizes its domain should look at at. I am basing it mostly on the fact that one the arrays slice that is the LHS of the assignment is a "jcn[radial_points] type=array dim=2" and the last index is floating (":").

  2. Roland Haas reporter
    • changed status to resolved
    • removed comment

    Bela Szilagyi said please apply (private communication :-) ). Applied as rev 9 of of NullEvolve.

  3. Log in to comment