/* glplux.h (LU-factorization, bignum arithmetic) */ /*********************************************************************** * This code is part of GLPK (GNU Linear Programming Kit). * * Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, * 2009, 2010 Andrew Makhorin, Department for Applied Informatics, * Moscow Aviation Institute, Moscow, Russia. All rights reserved. * E-mail: . * * GLPK is free software: you can redistribute it and/or modify it * under the terms of the GNU General Public License as published by * the Free Software Foundation, either version 3 of the License, or * (at your option) any later version. * * GLPK is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public * License for more details. * * You should have received a copy of the GNU General Public License * along with GLPK. If not, see . ***********************************************************************/ #ifndef GLPLUX_H #define GLPLUX_H #include "glpdmp.h" #include "glpgmp.h" /*---------------------------------------------------------------------- // The structure LUX defines LU-factorization of a square matrix A, // which is the following quartet: // // [A] = (F, V, P, Q), (1) // // where F and V are such matrices that // // A = F * V, (2) // // and P and Q are such permutation matrices that the matrix // // L = P * F * inv(P) (3) // // is lower triangular with unity diagonal, and the matrix // // U = P * V * Q (4) // // is upper triangular. All the matrices have the order n. // // The matrices F and V are stored in row/column-wise sparse format as // row and column linked lists of non-zero elements. Unity elements on // the main diagonal of the matrix F are not stored. Pivot elements of // the matrix V (that correspond to diagonal elements of the matrix U) // are also missing from the row and column lists and stored separately // in an ordinary array. // // The permutation matrices P and Q are stored as ordinary arrays using // both row- and column-like formats. // // The matrices L and U being completely defined by the matrices F, V, // P, and Q are not stored explicitly. // // It is easy to show that the factorization (1)-(3) is some version of // LU-factorization. Indeed, from (3) and (4) it follows that: // // F = inv(P) * L * P, // // V = inv(P) * U * inv(Q), // // and substitution into (2) gives: // // A = F * V = inv(P) * L * U * inv(Q). // // For more details see the program documentation. */ typedef struct LUX LUX; typedef struct LUXELM LUXELM; typedef struct LUXWKA LUXWKA; struct LUX { /* LU-factorization of a square matrix */ int n; /* the order of matrices A, F, V, P, Q */ DMP *pool; /* memory pool for elements of matrices F and V */ LUXELM **F_row; /* LUXELM *F_row[1+n]; */ /* F_row[0] is not used; F_row[i], 1 <= i <= n, is a pointer to the list of elements in i-th row of matrix F (diagonal elements are not stored) */ LUXELM **F_col; /* LUXELM *F_col[1+n]; */ /* F_col[0] is not used; F_col[j], 1 <= j <= n, is a pointer to the list of elements in j-th column of matrix F (diagonal elements are not stored) */ mpq_t *V_piv; /* mpq_t V_piv[1+n]; */ /* V_piv[0] is not used; V_piv[p], 1 <= p <= n, is a pivot element v[p,q] corresponding to a diagonal element u[k,k] of matrix U = P*V*Q (used on k-th elimination step, k = 1, 2, ..., n) */ LUXELM **V_row; /* LUXELM *V_row[1+n]; */ /* V_row[0] is not used; V_row[i], 1 <= i <= n, is a pointer to the list of elements in i-th row of matrix V (except pivot elements) */ LUXELM **V_col; /* LUXELM *V_col[1+n]; */ /* V_col[0] is not used; V_col[j], 1 <= j <= n, is a pointer to the list of elements in j-th column of matrix V (except pivot elements) */ int *P_row; /* int P_row[1+n]; */ /* P_row[0] is not used; P_row[i] = j means that p[i,j] = 1, where p[i,j] is an element of permutation matrix P */ int *P_col; /* int P_col[1+n]; */ /* P_col[0] is not used; P_col[j] = i means that p[i,j] = 1, where p[i,j] is an element of permutation matrix P */ /* if i-th row or column of matrix F is i'-th row or column of matrix L = P*F*inv(P), or if i-th row of matrix V is i'-th row of matrix U = P*V*Q, then P_row[i'] = i and P_col[i] = i' */ int *Q_row; /* int Q_row[1+n]; */ /* Q_row[0] is not used; Q_row[i] = j means that q[i,j] = 1, where q[i,j] is an element of permutation matrix Q */ int *Q_col; /* int Q_col[1+n]; */ /* Q_col[0] is not used; Q_col[j] = i means that q[i,j] = 1, where q[i,j] is an element of permutation matrix Q */ /* if j-th column of matrix V is j'-th column of matrix U = P*V*Q, then Q_row[j] = j' and Q_col[j'] = j */ int rank; /* the (exact) rank of matrices A and V */ }; struct LUXELM { /* element of matrix F or V */ int i; /* row index, 1 <= i <= m */ int j; /* column index, 1 <= j <= n */ mpq_t val; /* numeric (non-zero) element value */ LUXELM *r_prev; /* pointer to previous element in the same row */ LUXELM *r_next; /* pointer to next element in the same row */ LUXELM *c_prev; /* pointer to previous element in the same column */ LUXELM *c_next; /* pointer to next element in the same column */ }; struct LUXWKA { /* working area (used only during factorization) */ /* in order to efficiently implement Markowitz strategy and Duff search technique there are two families {R[0], R[1], ..., R[n]} and {C[0], C[1], ..., C[n]}; member R[k] is a set of active rows of matrix V having k non-zeros, and member C[k] is a set of active columns of matrix V having k non-zeros (in the active submatrix); each set R[k] and C[k] is implemented as a separate doubly linked list */ int *R_len; /* int R_len[1+n]; */ /* R_len[0] is not used; R_len[i], 1 <= i <= n, is the number of non-zero elements in i-th row of matrix V (that is the length of i-th row) */ int *R_head; /* int R_head[1+n]; */ /* R_head[k], 0 <= k <= n, is the number of a first row, which is active and whose length is k */ int *R_prev; /* int R_prev[1+n]; */ /* R_prev[0] is not used; R_prev[i], 1 <= i <= n, is the number of a previous row, which is active and has the same length as i-th row */ int *R_next; /* int R_next[1+n]; */ /* R_prev[0] is not used; R_prev[i], 1 <= i <= n, is the number of a next row, which is active and has the same length as i-th row */ int *C_len; /* int C_len[1+n]; */ /* C_len[0] is not used; C_len[j], 1 <= j <= n, is the number of non-zero elements in j-th column of the active submatrix of matrix V (that is the length of j-th column in the active submatrix) */ int *C_head; /* int C_head[1+n]; */ /* C_head[k], 0 <= k <= n, is the number of a first column, which is active and whose length is k */ int *C_prev; /* int C_prev[1+n]; */ /* C_prev[0] is not used; C_prev[j], 1 <= j <= n, is the number of a previous column, which is active and has the same length as j-th column */ int *C_next; /* int C_next[1+n]; */ /* C_next[0] is not used; C_next[j], 1 <= j <= n, is the number of a next column, which is active and has the same length as j-th column */ }; #define lux_create _glp_lux_create #define lux_decomp _glp_lux_decomp #define lux_f_solve _glp_lux_f_solve #define lux_v_solve _glp_lux_v_solve #define lux_solve _glp_lux_solve #define lux_delete _glp_lux_delete LUX *lux_create(int n); /* create LU-factorization */ int lux_decomp(LUX *lux, int (*col)(void *info, int j, int ind[], mpq_t val[]), void *info); /* compute LU-factorization */ void lux_f_solve(LUX *lux, int tr, mpq_t x[]); /* solve system F*x = b or F'*x = b */ void lux_v_solve(LUX *lux, int tr, mpq_t x[]); /* solve system V*x = b or V'*x = b */ void lux_solve(LUX *lux, int tr, mpq_t x[]); /* solve system A*x = b or A'*x = b */ void lux_delete(LUX *lux); /* delete LU-factorization */ #endif /* eof */