Skip to first unread message
Lukasz Kaczmraczyk
Nov 13, 2018, 3:21:15 PM11/13/18
to mofem...@googlegroups.com
Hello All,
If you are encountering problems with direct solver mumps, it crushing code with a segmentation fault and little warning this is a very likely problem with memory and matrix factorisation. The solution is to force natural ordering and increasing memory available to mumps. The problem will more likely to happen if you have higher polynomials orders and heterogenous approximation,i.e. higher order at the crack tip than in the rest of the mesh.
If you are encountering problems with direct solver mumps, it crushing code with a segmentation fault and little warning this is a very likely problem with memory and matrix factorisation. The solution is to force natural ordering and increasing memory available to mumps. The problem will more likely to happen if you have higher polynomials orders and heterogenous approximation,i.e. higher order at the crack tip than in the rest of the mesh.
In case of fracture module add this at the end of petsc_param.file to fix the problem:
-pc_factor_mat_ordering_type natural
-material_pc_factor_mat_ordering_type natural
-elastic_pc_factor_mat_ordering_type natural
-propagation_pc_factor_mat_ordering_type natural
-propagation_fieldsplit_0_mg_coarse_pc_factor_mat_ordering_type natural
-propagation_fieldsplit_0_pc_factor_mat_ordering_type natural
-propagation_fieldsplit_1_pc_factor_mat_ordering_type natural
-material_mat_mumps_icntl_14 10000
-elastic_mat_mumps_icntl_14 10000
-propagation_mat_mumps_icntl_14 10000
-propagation_fieldsplit_0_mat_mumps_icntl_14 10000
-propagation_fieldsplit_1_mat_mumps_icntl_14 10000
-mat_mumps_icntl_14 10000
Regards,
Lukasz
Reply all
Reply to author
Forward
0 new messages