cholmod-1.5-win-x64 Release Include及Lib库上传

/* ========================================================================== */ /* === Include/cholmod_core.h =============================================== */ /* ========================================================================== */ /* ----------------------------------------------------------------------------- * CHOLMOD/Include/cholmod_core.h. * Copyright (C) 2005-2013, Univ. of Florida. Author: Timothy A. Davis * -------------------------------------------------------------------------- */ /* CHOLMOD Core module: basic CHOLMOD objects and routines. * Required by all CHOLMOD modules. Requires no other module or package. * * The CHOLMOD modules are: * * Core basic data structures and definitions * Check check/print the 5 CHOLMOD objects, & 3 types of integer vectors * Cholesky sparse Cholesky factorization * Modify sparse Cholesky update/downdate/row-add/row-delete * MatrixOps sparse matrix functions (add, multiply, norm, ...) * Supernodal supernodal sparse Cholesky factorization * Partition graph-partitioning based orderings * * The CHOLMOD objects: * -------------------- * * cholmod_common parameters, statistics, and workspace * cholmod_sparse a sparse matrix in compressed column form * cholmod_factor an LL' or LDL' factorization * cholmod_dense a dense matrix * cholmod_triplet a sparse matrix in "triplet" form * * The Core module described here defines the CHOLMOD data structures, and * basic operations on them. To create and solve a sparse linear system Ax=b, * the user must create A and b, populate them with values, and then pass them * to the routines in the CHOLMOD Cholesky module. There are two primary * methods for creating A: (1) allocate space for a column-oriented sparse * matrix and fill it with pattern and values, or (2) create a triplet form * matrix and convert it to a sparse matrix. The latter option is simpler. * * The matrices b and x are typically dense matrices, but can also be sparse. * You can allocate and free them as dense matrices with the * cholmod_allocate_dense and cholmod_free_dense routines. * * The cholmod_factor object contains the symbolic and numeric LL' or LDL' * factorization of sparse symmetric matrix. The matrix must be positive * definite for an LL' factorization. It need only be symmetric and have well- * conditioned leading submatrices for it to have an LDL' factorization * (CHOLMOD does not pivot for numerical stability). It is typically created * with the cholmod_factorize routine in the Cholesky module, but can also * be initialized to L=D=I in the Core module and then modified by the Modify * module. It must be freed with cholmod_free_factor, defined below. * * The Core routines for each object are described below. Each list is split * into two parts: the primary routines and secondary routines. * * ============================================================================ * === cholmod_common ========================================================= * ============================================================================ * * The Common object contains control parameters, statistics, and * You must call cholmod_start before calling any other CHOLMOD routine, and * must call cholmod_finish as your last call to CHOLMOD, with two exceptions: * you may call cholmod_print_common and cholmod_check_common in the Check * module after calling cholmod_finish. * * cholmod_start first call to CHOLMOD * cholmod_finish last call to CHOLMOD * ----------------------------- * cholmod_defaults restore default parameters * cholmod_maxrank maximum rank for update/downdate * cholmod_allocate_work allocate workspace in Common * cholmod_free_work free workspace in Common * cholmod_clear_flag clear Flag workspace in Common * cholmod_error called when CHOLMOD encounters an error * cholmod_dbound for internal use in CHOLMOD only * cholmod_hypot compute sqrt (x*x + y*y) accurately * cholmod_divcomplex complex division, c = a/b * * ============================================================================ * === cholmod_sparse ========================================================= * ============================================================================ * * A sparse matrix is held in compressed column form. In the basic type * ("packed", which corresponds to a MATLAB sparse matrix), an n-by-n matrix * with nz entries is held in three arrays: p of size n+1, i of size nz, and x * of size nz. Row indices of column j are held in i [p [j] ... p [j+1]-1] and * in the same locations in x. There may be no duplicate entries in a column. * Row indices in each column may be sorted or unsorted (CHOLMOD keeps track). * A->stype determines the storage mode: 0 if both upper/lower parts are stored, * -1 if A is symmetric and just tril(A) is stored, +1 if symmetric and triu(A) * is stored. * * cholmod_allocate_sparse allocate a sparse matrix * cholmod_free_sparse free a sparse matrix * ----------------------------- * cholmod_reallocate_sparse change the size (# entries) of sparse matrix * cholmod_nnz number of nonzeros in a sparse matrix * cholmod_speye sparse identity matrix * cholmod_spzeros sparse zero matrix * cholmod_transpose transpose a sparse matrix * cholmod_ptranspose transpose/permute a sparse matrix * cholmod_transpose_unsym transpose/permute an unsymmetric sparse matrix * cholmod_transpose_sym transpose/permute a symmetric sparse matrix * cholmod_sort sort row indices in each column of sparse matrix * cholmod_band C = tril (triu (A,k1), k2) * cholmod_band_inplace A = tril (triu (A,k1), k2) * cholmod_aat C = A*A' * cholmod_copy_sparse C = A, create an exact copy of a sparse matrix * cholmod_copy C = A, with possible change of stype * cholmod_add C = alpha*A + beta*B * cholmod_sparse_xtype change the xtype of a sparse matrix * * ============================================================================ * === cholmod_factor ========================================================= * ============================================================================ * * The data structure for an LL' or LDL' factorization is too complex to * describe in one sentence. This object can hold the symbolic analysis alone, * or in combination with a "simplicial" (similar to a sparse matrix) or * "supernodal" form of the numerical factorization. Only the routine to free * a factor is primary, since a factor object is created by the factorization * routine (cholmod_factorize). It must be freed with cholmod_free_factor. * * cholmod_free_factor free a factor * ----------------------------- * cholmod_allocate_factor allocate a factor (LL' or LDL') * cholmod_reallocate_factor change the # entries in a factor * cholmod_change_factor change the type of factor (e.g., LDL' to LL') * cholmod_pack_factor pack the columns of a factor * cholmod_reallocate_column resize a single column of a factor * cholmod_factor_to_sparse create a sparse matrix copy of a factor * cholmod_copy_factor create a copy of a factor * cholmod_factor_xtype change the xtype of a factor * * Note that there is no cholmod_sparse_to_factor routine to create a factor * as a copy of a sparse matrix. It could be done, after a fashion, but a * lower triangular sparse matrix would not necessarily have a chordal graph, * which would break the many CHOLMOD routines that rely on this property. * * ============================================================================ * === cholmod_dense ========================================================== * ============================================================================ * * The solve routines and some of the MatrixOps and Modify routines use dense * matrices as inputs. These ar