alpar@9: /* glpmat.h (linear algebra routines) */
alpar@9: 
alpar@9: /***********************************************************************
alpar@9: *  This code is part of GLPK (GNU Linear Programming Kit).
alpar@9: *
alpar@9: *  Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008,
alpar@9: *  2009, 2010, 2011 Andrew Makhorin, Department for Applied Informatics,
alpar@9: *  Moscow Aviation Institute, Moscow, Russia. All rights reserved.
alpar@9: *  E-mail: <mao@gnu.org>.
alpar@9: *
alpar@9: *  GLPK is free software: you can redistribute it and/or modify it
alpar@9: *  under the terms of the GNU General Public License as published by
alpar@9: *  the Free Software Foundation, either version 3 of the License, or
alpar@9: *  (at your option) any later version.
alpar@9: *
alpar@9: *  GLPK is distributed in the hope that it will be useful, but WITHOUT
alpar@9: *  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
alpar@9: *  or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
alpar@9: *  License for more details.
alpar@9: *
alpar@9: *  You should have received a copy of the GNU General Public License
alpar@9: *  along with GLPK. If not, see <http://www.gnu.org/licenses/>.
alpar@9: ***********************************************************************/
alpar@9: 
alpar@9: #ifndef GLPMAT_H
alpar@9: #define GLPMAT_H
alpar@9: 
alpar@9: /***********************************************************************
alpar@9: *  FULL-VECTOR STORAGE
alpar@9: * 
alpar@9: *  For a sparse vector x having n elements, ne of which are non-zero,
alpar@9: *  the full-vector storage format uses two arrays x_ind and x_vec, which
alpar@9: *  are set up as follows:
alpar@9: * 
alpar@9: *  x_ind is an integer array of length [1+ne]. Location x_ind[0] is
alpar@9: *  not used, and locations x_ind[1], ..., x_ind[ne] contain indices of
alpar@9: *  non-zero elements in vector x.
alpar@9: * 
alpar@9: *  x_vec is a floating-point array of length [1+n]. Location x_vec[0]
alpar@9: *  is not used, and locations x_vec[1], ..., x_vec[n] contain numeric
alpar@9: *  values of ALL elements in vector x, including its zero elements.
alpar@9: * 
alpar@9: *  Let, for example, the following sparse vector x be given:
alpar@9: * 
alpar@9: *     (0, 1, 0, 0, 2, 3, 0, 4)
alpar@9: * 
alpar@9: *  Then the arrays are:
alpar@9: * 
alpar@9: *     x_ind = { X; 2, 5, 6, 8 }
alpar@9: * 
alpar@9: *     x_vec = { X; 0, 1, 0, 0, 2, 3, 0, 4 }
alpar@9: * 
alpar@9: *  COMPRESSED-VECTOR STORAGE
alpar@9: * 
alpar@9: *  For a sparse vector x having n elements, ne of which are non-zero,
alpar@9: *  the compressed-vector storage format uses two arrays x_ind and x_vec,
alpar@9: *  which are set up as follows:
alpar@9: * 
alpar@9: *  x_ind is an integer array of length [1+ne]. Location x_ind[0] is
alpar@9: *  not used, and locations x_ind[1], ..., x_ind[ne] contain indices of
alpar@9: *  non-zero elements in vector x.
alpar@9: * 
alpar@9: *  x_vec is a floating-point array of length [1+ne]. Location x_vec[0]
alpar@9: *  is not used, and locations x_vec[1], ..., x_vec[ne] contain numeric
alpar@9: *  values of corresponding non-zero elements in vector x.
alpar@9: * 
alpar@9: *  Let, for example, the following sparse vector x be given:
alpar@9: * 
alpar@9: *     (0, 1, 0, 0, 2, 3, 0, 4)
alpar@9: * 
alpar@9: *  Then the arrays are:
alpar@9: *
alpar@9: *     x_ind = { X; 2, 5, 6, 8 }
alpar@9: * 
alpar@9: *     x_vec = { X; 1, 2, 3, 4 }
alpar@9: * 
alpar@9: *  STORAGE-BY-ROWS
alpar@9: * 
alpar@9: *  For a sparse matrix A, which has m rows, n columns, and ne non-zero
alpar@9: *  elements the storage-by-rows format uses three arrays A_ptr, A_ind,
alpar@9: *  and A_val, which are set up as follows:
alpar@9: * 
alpar@9: *  A_ptr is an integer array of length [1+m+1] also called "row pointer
alpar@9: *  array". It contains the relative starting positions of each row of A
alpar@9: *  in the arrays A_ind and A_val, i.e. element A_ptr[i], 1 <= i <= m,
alpar@9: *  indicates where row i begins in the arrays A_ind and A_val. If all
alpar@9: *  elements in row i are zero, then A_ptr[i] = A_ptr[i+1]. Location
alpar@9: *  A_ptr[0] is not used, location A_ptr[1] must contain 1, and location
alpar@9: *  A_ptr[m+1] must contain ne+1 that indicates the position after the
alpar@9: *  last element in the arrays A_ind and A_val.
alpar@9: * 
alpar@9: *  A_ind is an integer array of length [1+ne]. Location A_ind[0] is not
alpar@9: *  used, and locations A_ind[1], ..., A_ind[ne] contain column indices
alpar@9: *  of (non-zero) elements in matrix A.
alpar@9: *
alpar@9: *  A_val is a floating-point array of length [1+ne]. Location A_val[0]
alpar@9: *  is not used, and locations A_val[1], ..., A_val[ne] contain numeric
alpar@9: *  values of non-zero elements in matrix A.
alpar@9: * 
alpar@9: *  Non-zero elements of matrix A are stored contiguously, and the rows
alpar@9: *  of matrix A are stored consecutively from 1 to m in the arrays A_ind
alpar@9: *  and A_val. The elements in each row of A may be stored in any order
alpar@9: *  in A_ind and A_val. Note that elements with duplicate column indices
alpar@9: *  are not allowed.
alpar@9: * 
alpar@9: *  Let, for example, the following sparse matrix A be given:
alpar@9: * 
alpar@9: *     | 11  . 13  .  .  . |
alpar@9: *     | 21 22  . 24  .  . |
alpar@9: *     |  . 32 33  .  .  . |
alpar@9: *     |  .  . 43 44  . 46 |
alpar@9: *     |  .  .  .  .  .  . |
alpar@9: *     | 61 62  .  .  . 66 |
alpar@9: * 
alpar@9: *  Then the arrays are:
alpar@9: * 
alpar@9: *     A_ptr = { X; 1, 3, 6, 8, 11, 11; 14 }
alpar@9: *
alpar@9: *     A_ind = { X;  1,  3;  4,  2,  1;  2,  3;  4,  3,  6;  1,  2,  6 }
alpar@9: * 
alpar@9: *     A_val = { X; 11, 13; 24, 22, 21; 32, 33; 44, 43, 46; 61, 62, 66 }
alpar@9: * 
alpar@9: *  PERMUTATION MATRICES
alpar@9: * 
alpar@9: *  Let P be a permutation matrix of the order n. It is represented as
alpar@9: *  an integer array P_per of length [1+n+n] as follows: if p[i,j] = 1,
alpar@9: *  then P_per[i] = j and P_per[n+j] = i. Location P_per[0] is not used.
alpar@9: * 
alpar@9: *  Let A' = P*A. If i-th row of A corresponds to i'-th row of A', then
alpar@9: *  P_per[i'] = i and P_per[n+i] = i'.
alpar@9: * 
alpar@9: *  References:
alpar@9: * 
alpar@9: *  1. Gustavson F.G. Some basic techniques for solving sparse systems of
alpar@9: *     linear equations. In Rose and Willoughby (1972), pp. 41-52.
alpar@9: * 
alpar@9: *  2. Basic Linear Algebra Subprograms Technical (BLAST) Forum Standard.
alpar@9: *     University of Tennessee (2001). */
alpar@9: 
alpar@9: #define check_fvs _glp_mat_check_fvs
alpar@9: int check_fvs(int n, int nnz, int ind[], double vec[]);
alpar@9: /* check sparse vector in full-vector storage format */
alpar@9: 
alpar@9: #define check_pattern _glp_mat_check_pattern
alpar@9: int check_pattern(int m, int n, int A_ptr[], int A_ind[]);
alpar@9: /* check pattern of sparse matrix */
alpar@9: 
alpar@9: #define transpose _glp_mat_transpose
alpar@9: void transpose(int m, int n, int A_ptr[], int A_ind[], double A_val[],
alpar@9:       int AT_ptr[], int AT_ind[], double AT_val[]);
alpar@9: /* transpose sparse matrix */
alpar@9: 
alpar@9: #define adat_symbolic _glp_mat_adat_symbolic
alpar@9: int *adat_symbolic(int m, int n, int P_per[], int A_ptr[], int A_ind[],
alpar@9:       int S_ptr[]);
alpar@9: /* compute S = P*A*D*A'*P' (symbolic phase) */
alpar@9: 
alpar@9: #define adat_numeric _glp_mat_adat_numeric
alpar@9: void adat_numeric(int m, int n, int P_per[],
alpar@9:       int A_ptr[], int A_ind[], double A_val[], double D_diag[],
alpar@9:       int S_ptr[], int S_ind[], double S_val[], double S_diag[]);
alpar@9: /* compute S = P*A*D*A'*P' (numeric phase) */
alpar@9: 
alpar@9: #define min_degree _glp_mat_min_degree
alpar@9: void min_degree(int n, int A_ptr[], int A_ind[], int P_per[]);
alpar@9: /* minimum degree ordering */
alpar@9: 
alpar@9: #define amd_order1 _glp_mat_amd_order1
alpar@9: void amd_order1(int n, int A_ptr[], int A_ind[], int P_per[]);
alpar@9: /* approximate minimum degree ordering (AMD) */
alpar@9: 
alpar@9: #define symamd_ord _glp_mat_symamd_ord
alpar@9: void symamd_ord(int n, int A_ptr[], int A_ind[], int P_per[]);
alpar@9: /* approximate minimum degree ordering (SYMAMD) */
alpar@9: 
alpar@9: #define chol_symbolic _glp_mat_chol_symbolic
alpar@9: int *chol_symbolic(int n, int A_ptr[], int A_ind[], int U_ptr[]);
alpar@9: /* compute Cholesky factorization (symbolic phase) */
alpar@9: 
alpar@9: #define chol_numeric _glp_mat_chol_numeric
alpar@9: int chol_numeric(int n,
alpar@9:       int A_ptr[], int A_ind[], double A_val[], double A_diag[],
alpar@9:       int U_ptr[], int U_ind[], double U_val[], double U_diag[]);
alpar@9: /* compute Cholesky factorization (numeric phase) */
alpar@9: 
alpar@9: #define u_solve _glp_mat_u_solve
alpar@9: void u_solve(int n, int U_ptr[], int U_ind[], double U_val[],
alpar@9:       double U_diag[], double x[]);
alpar@9: /* solve upper triangular system U*x = b */
alpar@9: 
alpar@9: #define ut_solve _glp_mat_ut_solve
alpar@9: void ut_solve(int n, int U_ptr[], int U_ind[], double U_val[],
alpar@9:       double U_diag[], double x[]);
alpar@9: /* solve lower triangular system U'*x = b */
alpar@9: 
alpar@9: #endif
alpar@9: 
alpar@9: /* eof */