"cusparse" - A matrix type providing triangular solvers for seq matrices on a single GPU of type, seqaijcusparse, aijcusparse, or seqaijcusp, aijcusp. Currently supported algorithms are ILU(k) and ICC(k). Typically, deeper factorizations (larger k) results in poorer performance in the triangular solves. Full LU, and Cholesky decompositions can be solved through the CUSPARSE triangular solve algorithm. However, the performance can be quite poor and thus these algorithms are not recommended. This class does NOT support direct solver operations. ./configure --download-txpetscgpu to install PETSc to use CUSPARSE
Consult CUSPARSE documentation for more information about the matrix storage formats
which correspond to the options database keys below.
Options Database Keys
-mat_cusparse_solve_storage_format csr -sets the storage format matrices (for factors in MatSolve) during a call to MatSetFromOptions(). Other options include ell (ellpack) or hyb (hybrid). Only available with 'txpetscgpu' package.
PCFactorSetMatSolverPackage(), MatSolverPackage, MatCreateSeqAIJCUSPARSE(), MATAIJCUSPARSE, MatCreateAIJCUSPARSE(), MatCUSPARSESetFormat(), MatCUSPARSEStorageFormat, MatCUSPARSEFormatOperation
Index of all Mat routines
Table of Contents for all manual pages
Index of all manual pages