MUMPS - a parallel direct solver for sparse systems
Usage:
-mat_type aijmumps -ksp_type preonly -pc_type lu
aij matrix type
-mat_type sbaijmumps -ksp_type preonly -pc_type cholesky
symmetric sbaij matrix type
Options:
-mat_mumps_sym <0,1,2>
0 the matrix is unsymmetric, 1 symmetric positive definite, 2 symmetric
-mat_mumps_icntl_4 <0,1,2,3,4>
print level
-mat_mumps_icntl_6 <0,...,7>
matrix prescaling options (see MUMPS User's Guide)
-mat_mumps_icntl_7 <0,...,7>
matrix orderings (see MUMPS User's Guide)
-mat_mumps_icntl_9 <1,2>
A or A^T x=b to be solved: 1 denotes A, 2 denotes A^T
-mat_mumps_icntl_10 <n>
maximum number of iterative refinements
-mat_mumps_icntl_11 <n>
error analysis, a positive value returns statistics during -ksp_view
-mat_mumps_icntl_12 <n>
efficiency control (see MUMPS User's Guide)
-mat_mumps_icntl_13 <n>
efficiency control (see MUMPS User's Guide)
-mat_mumps_icntl_14 <n>
efficiency control (see MUMPS User's Guide)
-mat_mumps_icntl_15 <n>
efficiency control (see MUMPS User's Guide)
-mat_mumps_cntl_1 <delta>
relative pivoting threshold
-mat_mumps_cntl_2 <tol>
stopping criterion for refinement
-mat_mumps_cntl_3 <adelta>
absolute pivoting threshold