/* glplpf.h (LP basis factorization, Schur complement version) */ /*********************************************************************** * 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 GLPLPF_H #define GLPLPF_H #include "glpscf.h" #include "glpluf.h" /*********************************************************************** * The structure LPF defines the factorization of the basis mxm matrix * B, where m is the number of rows in corresponding problem instance. * * This factorization is the following septet: * * [B] = (L0, U0, R, S, C, P, Q), (1) * * and is based on the following main equality: * * ( B F^) ( B0 F ) ( L0 0 ) ( U0 R ) * ( ) = P ( ) Q = P ( ) ( ) Q, (2) * ( G^ H^) ( G H ) ( S I ) ( 0 C ) * * where: * * B is the current basis matrix (not stored); * * F^, G^, H^ are some additional matrices (not stored); * * B0 is some initial basis matrix (not stored); * * F, G, H are some additional matrices (not stored); * * P, Q are permutation matrices (stored in both row- and column-like * formats); * * L0, U0 are some matrices that defines a factorization of the initial * basis matrix B0 = L0 * U0 (stored in an invertable form); * * R is a matrix defined from L0 * R = F, so R = inv(L0) * F (stored in * a column-wise sparse format); * * S is a matrix defined from S * U0 = G, so S = G * inv(U0) (stored in * a row-wise sparse format); * * C is the Schur complement for matrix (B0 F G H). It is defined from * S * R + C = H, so C = H - S * R = H - G * inv(U0) * inv(L0) * F = * = H - G * inv(B0) * F. Matrix C is stored in an invertable form. * * REFERENCES * * 1. M.A.Saunders, "LUSOL: A basis package for constrained optimiza- * tion," SCCM, Stanford University, 2006. * * 2. M.A.Saunders, "Notes 5: Basis Updates," CME 318, Stanford Univer- * sity, Spring 2006. * * 3. M.A.Saunders, "Notes 6: LUSOL---a Basis Factorization Package," * ibid. */ typedef struct LPF LPF; struct LPF { /* LP basis factorization */ int valid; /* the factorization is valid only if this flag is set */ /*--------------------------------------------------------------*/ /* initial basis matrix B0 */ int m0_max; /* maximal value of m0 (increased automatically, if necessary) */ int m0; /* the order of B0 */ LUF *luf; /* LU-factorization of B0 */ /*--------------------------------------------------------------*/ /* current basis matrix B */ int m; /* the order of B */ double *B; /* double B[1+m*m]; */ /* B in dense format stored by rows and used only for debugging; normally this array is not allocated */ /*--------------------------------------------------------------*/ /* augmented matrix (B0 F G H) of the order m0+n */ int n_max; /* maximal number of additional rows and columns */ int n; /* current number of additional rows and columns */ /*--------------------------------------------------------------*/ /* m0xn matrix R in column-wise format */ int *R_ptr; /* int R_ptr[1+n_max]; */ /* R_ptr[j], 1 <= j <= n, is a pointer to j-th column */ int *R_len; /* int R_len[1+n_max]; */ /* R_len[j], 1 <= j <= n, is the length of j-th column */ /*--------------------------------------------------------------*/ /* nxm0 matrix S in row-wise format */ int *S_ptr; /* int S_ptr[1+n_max]; */ /* S_ptr[i], 1 <= i <= n, is a pointer to i-th row */ int *S_len; /* int S_len[1+n_max]; */ /* S_len[i], 1 <= i <= n, is the length of i-th row */ /*--------------------------------------------------------------*/ /* Schur complement C of the order n */ SCF *scf; /* SCF scf[1:n_max]; */ /* factorization of the Schur complement */ /*--------------------------------------------------------------*/ /* matrix P of the order m0+n */ int *P_row; /* int P_row[1+m0_max+n_max]; */ /* P_row[i] = j means that P[i,j] = 1 */ int *P_col; /* int P_col[1+m0_max+n_max]; */ /* P_col[j] = i means that P[i,j] = 1 */ /*--------------------------------------------------------------*/ /* matrix Q of the order m0+n */ int *Q_row; /* int Q_row[1+m0_max+n_max]; */ /* Q_row[i] = j means that Q[i,j] = 1 */ int *Q_col; /* int Q_col[1+m0_max+n_max]; */ /* Q_col[j] = i means that Q[i,j] = 1 */ /*--------------------------------------------------------------*/ /* Sparse Vector Area (SVA) is a set of locations intended to store sparse vectors which represent columns of matrix R and rows of matrix S; each location is a doublet (ind, val), where ind is an index, val is a numerical value of a sparse vector element; in the whole each sparse vector is a set of adjacent locations defined by a pointer to its first element and its length, i.e. the number of its elements */ int v_size; /* the SVA size, in locations; locations are numbered by integers 1, 2, ..., v_size, and location 0 is not used */ int v_ptr; /* pointer to the first available location */ int *v_ind; /* int v_ind[1+v_size]; */ /* v_ind[k], 1 <= k <= v_size, is the index field of location k */ double *v_val; /* double v_val[1+v_size]; */ /* v_val[k], 1 <= k <= v_size, is the value field of location k */ /*--------------------------------------------------------------*/ double *work1; /* double work1[1+m0+n_max]; */ /* working array */ double *work2; /* double work2[1+m0+n_max]; */ /* working array */ }; /* return codes: */ #define LPF_ESING 1 /* singular matrix */ #define LPF_ECOND 2 /* ill-conditioned matrix */ #define LPF_ELIMIT 3 /* update limit reached */ #define lpf_create_it _glp_lpf_create_it LPF *lpf_create_it(void); /* create LP basis factorization */ #define lpf_factorize _glp_lpf_factorize int lpf_factorize(LPF *lpf, int m, const int bh[], int (*col) (void *info, int j, int ind[], double val[]), void *info); /* compute LP basis factorization */ #define lpf_ftran _glp_lpf_ftran void lpf_ftran(LPF *lpf, double x[]); /* perform forward transformation (solve system B*x = b) */ #define lpf_btran _glp_lpf_btran void lpf_btran(LPF *lpf, double x[]); /* perform backward transformation (solve system B'*x = b) */ #define lpf_update_it _glp_lpf_update_it int lpf_update_it(LPF *lpf, int j, int bh, int len, const int ind[], const double val[]); /* update LP basis factorization */ #define lpf_delete_it _glp_lpf_delete_it void lpf_delete_it(LPF *lpf); /* delete LP basis factorization */ #endif /* eof */