The nvmath.sparse module is experimental and potentially subject to future changes.

DirectSolverPlanPreferences#

class nvmath.sparse.advanced.DirectSolverPlanPreferences( *, host_nthreads: int | None = None, reordering_algorithm: ReorderingAlg | None = None, pivot_type: int | None = None, pivot_threshold: float | None = None, max_nnz: int | None = None, matching_algorithm: MatchingAlg | None = None, nd_min_levels: int | None = None, use_superpanels: bool | None = None, )[source]#

A data class for providing plan preferences to the direct_solver() function.

host_nthreads#

The number of host threads to use.

Type:: int | None

reordering_algorithm#

The reordering algorithm to use.

Type:: nvmath.bindings.cudss.ReorderingAlg | None

pivot_type#

The pivot type to use.

Type:: int | None

pivot_threshold#

The pivot threshold to use.

Type:: float | None

max_nnz#

The maximum number of non-zeros in the LU factorization.

Type:: int | None

matching_algorithm#

The matching algorithm to use.

Type:: nvmath.bindings.cudss.MatchingAlg | None

nd_min_levels#

The number of levels in the nested dissection.

Type:: int | None

use_superpanels#

Whether to use superpanels.

Type:: bool | None