bls1.h

求矩阵奇异分解svd算法

 /***************************************************************
 * Sparse SVD Via Hybrid Block Lanczos Procedure for Equivalent *
 * 2-Cyclic Eigensystems.                                       *
 *                                                              *
 * Global variables and common areas used by bls1 and its       *
 * functions.                                                   *
 ****************************************************************/

int    *pointr ,      /* pointer to column start array         */
	*rowind ;      /* pointer to row indices array          */
float  *value  ;      /* pointer to nonzero values array       */

int    mxvcount,      /* matrix-vector multiplications counter */ 
        mtxvcount,     /* transposed matrix-vector mult. counter*/
	iconv,         /* converged vector counter              */
	nn,            /* current subspace size                 */
	iter;          /* iteration counter                     */

/* memory for the following arrays are allocated in bls1.c and  *
 * blklan1.c (see blklan1.c for dimensions of these arrays)     */

float *alpha,         /* diagonal elements of bidiagonal matrix*
		        * from single vector Lanczos (inner)    *
			* recursion                             */
       *beta,          /* super-diagonals of bidiagonal matrix  *
			* from single vector Lanczos (inner)    *
			* recursion                             */
       *p,             /* work array                            */
       *q,             /* work array                            */
       *t,             /* work array                            */
       *z;             /* work array                            */
float *tres,          /* temporary residual array              */
       *y,             /* work array                            */
       *temp,          /* temporary work space                  */
       *uu,            /* work array                            */
       *vv,            /* work array                            */
       *u0,            /* array of converged left S-vectors     */
       *v0,            /* array of converged right S-vectors    */
       *uvtmp,         /* temporary work space                  */
       *pp,            /* left S-vectors of block upper bi-     *
			* diagonal matrix from outer recursion  */ 
       *qq;            /* right S-vectors of block upper bi-    *
			* diagonal matrix from outer recursion  */

float **uup,          /* corresponding 2-dimensional           */
       **yp,           /* ...array representation of the above  */
       **vvp,          /* ...linear arrays                      */
       **uvtmpp, 
       **ppp, 
       **qqp;