search

openmc-dev / openmc

OpenMC Monte Carlo Code
https://docs.openmc.org
Other
764 stars 492 forks source link

CMFD test run on BG/Q #205

Closed bhermanmit closed 10 years ago

bhermanmit commented 11 years ago

Current, CMFD slab test only works when OpenMC is compiled in debug and linked with a debug version of PETSc. Unknown errors occur when using other combinations of OpenMC compiled with optimization and/or an optimized version of PETSc linked in.

bhermanmit commented 11 years ago

After looking into this further, PETSc is not causing this error. The error occurs in neutron_balance subroutine of cmfd_data module and only occurs when optimization is on. Code runs fine without optimization. We can just add a compiler directive to hide this subroutine when compiling on BG/Q.

paulromano commented 11 years ago

What error do you run across in that subroutine?

bhermanmit commented 11 years ago

It just says termination on signal 11 and gives a core dump file that tells me nothing. It seems to error out on the nested loop statements, doesn't make sense.

paulromano commented 11 years ago

Did you try to bgq_stack the core dump?

bhermanmit commented 11 years ago

I did try that, here is the result:

+++ID Rank: 0, TGID: 1, Core: 0, HWTID:0 TID: 1 State: RUN

0000000002a6a518 abort /bgsys/drivers/V1R2M0/ppc64/toolchain/gnu/glibc-2.12.2/stdlib/abort.c:77

00000000023254a0 MPID_Abort /bgsys/source/srcV1R2M0.14091/comm/lib/dev/mpich2/src/mpid/pamid/src/misc/mpid_abort.c:98

00000000022ae908 PMPI_Abort /bgsys/source/srcV1R2M0.14091/comm/lib/dev/mpich2/src/mpi/init/abort.c:118

0000000001475ea0 PetscSignalHandlerDefault /gpfs/vesta-home/herman/opt/petsc_src/src/sys/error/signal.c:161

0000000001476c80 PetscSignalHandler_Private /gpfs/vesta-home/herman/opt/petsc_src/src/sys/error/signal.c:49

0000000002f36000 ?? ??:0

0000000001024b0c 00000012.long_branch_r2off.feenableexcept+0 :0

000000000102ad98 00000012.long_branch_r2off.feenableexcept+0 :0

000000000103c41c 00000012.long_branch_r2off.feenableexcept+0 :0

000000000106e7e4 00000012.long_branch_r2off.feenableexcept+0 :0

000000000106f244 00000012.long_branch_r2off.feenableexcept+0 :0

00000000010dc848 00000012.long_branch_r2off.feenableexcept+0 :0

0000000002a60d08 generic_start_main /bgsys/drivers/V1R2M0/ppc64/toolchain/gnu/glibc-2.12.2/csu/../csu/libc-start.c:226

0000000002a61004 __libc_start_main /bgsys/drivers/V1R2M0/ppc64/toolchain/gnu/glibc-2.12.2/csu/../sysdeps/unix/sysv/linux/powerpc/libc-start.c:194

paulromano commented 10 years ago

Closing this unless there is further information that this is a continuing problem. The fact that we now have PETSc-less CMFD probably obviates the problem.