forked from bartvdbraak/blender
Brecht Van Lommel
0b12e61040
This helps to improve the accuracy of UV unwrapping and laplacian deform for high poly meshes, which could get warped quite badly. It's not much slower, doubles are pretty fast on modern CPUs, but it does double memory usage. This seems acceptable as otherwise high poly meshes would not work correctly anyway. Fixes T39004.
241 lines
9.4 KiB
C
241 lines
9.4 KiB
C
/** \file opennl/superlu/ssp_defs.h
|
|
* \ingroup opennl
|
|
*/
|
|
|
|
/*
|
|
* -- SuperLU routine (version 3.0) --
|
|
* Univ. of California Berkeley, Xerox Palo Alto Research Center,
|
|
* and Lawrence Berkeley National Lab.
|
|
* October 15, 2003
|
|
*
|
|
*/
|
|
#ifndef __SUPERLU_sSP_DEFS /* allow multiple inclusions */
|
|
#define __SUPERLU_sSP_DEFS
|
|
|
|
/*
|
|
* File name: ssp_defs.h
|
|
* Purpose: Sparse matrix types and function prototypes
|
|
* History:
|
|
*/
|
|
|
|
#ifdef _CRAY
|
|
#include <fortran.h>
|
|
#include <string.h>
|
|
#endif
|
|
|
|
/* Define my integer type int_t */
|
|
typedef int int_t; /* default */
|
|
|
|
#include "Cnames.h"
|
|
#include "supermatrix.h"
|
|
#include "util.h"
|
|
|
|
|
|
/*
|
|
* Global data structures used in LU factorization -
|
|
*
|
|
* nsuper: #supernodes = nsuper + 1, numbered [0, nsuper].
|
|
* (xsup,supno): supno[i] is the supernode no to which i belongs;
|
|
* xsup(s) points to the beginning of the s-th supernode.
|
|
* e.g. supno 0 1 2 2 3 3 3 4 4 4 4 4 (n=12)
|
|
* xsup 0 1 2 4 7 12
|
|
* Note: dfs will be performed on supernode rep. relative to the new
|
|
* row pivoting ordering
|
|
*
|
|
* (xlsub,lsub): lsub[*] contains the compressed subscript of
|
|
* rectangular supernodes; xlsub[j] points to the starting
|
|
* location of the j-th column in lsub[*]. Note that xlsub
|
|
* is indexed by column.
|
|
* Storage: original row subscripts
|
|
*
|
|
* During the course of sparse LU factorization, we also use
|
|
* (xlsub,lsub) for the purpose of symmetric pruning. For each
|
|
* supernode {s,s+1,...,t=s+r} with first column s and last
|
|
* column t, the subscript set
|
|
* lsub[j], j=xlsub[s], .., xlsub[s+1]-1
|
|
* is the structure of column s (i.e. structure of this supernode).
|
|
* It is used for the storage of numerical values.
|
|
* Furthermore,
|
|
* lsub[j], j=xlsub[t], .., xlsub[t+1]-1
|
|
* is the structure of the last column t of this supernode.
|
|
* It is for the purpose of symmetric pruning. Therefore, the
|
|
* structural subscripts can be rearranged without making physical
|
|
* interchanges among the numerical values.
|
|
*
|
|
* However, if the supernode has only one column, then we
|
|
* only keep one set of subscripts. For any subscript interchange
|
|
* performed, similar interchange must be done on the numerical
|
|
* values.
|
|
*
|
|
* The last column structures (for pruning) will be removed
|
|
* after the numercial LU factorization phase.
|
|
*
|
|
* (xlusup,lusup): lusup[*] contains the numerical values of the
|
|
* rectangular supernodes; xlusup[j] points to the starting
|
|
* location of the j-th column in storage vector lusup[*]
|
|
* Note: xlusup is indexed by column.
|
|
* Each rectangular supernode is stored by column-major
|
|
* scheme, consistent with Fortran 2-dim array storage.
|
|
*
|
|
* (xusub,ucol,usub): ucol[*] stores the numerical values of
|
|
* U-columns outside the rectangular supernodes. The row
|
|
* subscript of nonzero ucol[k] is stored in usub[k].
|
|
* xusub[i] points to the starting location of column i in ucol.
|
|
* Storage: new row subscripts; that is subscripts of PA.
|
|
*/
|
|
typedef struct {
|
|
int *xsup; /* supernode and column mapping */
|
|
int *supno;
|
|
int *lsub; /* compressed L subscripts */
|
|
int *xlsub;
|
|
double *lusup; /* L supernodes */
|
|
int *xlusup;
|
|
double *ucol; /* U columns */
|
|
int *usub;
|
|
int *xusub;
|
|
int nzlmax; /* current max size of lsub */
|
|
int nzumax; /* " " " ucol */
|
|
int nzlumax; /* " " " lusup */
|
|
int n; /* number of columns in the matrix */
|
|
LU_space_t MemModel; /* 0 - system malloc'd; 1 - user provided */
|
|
} GlobalLU_t;
|
|
|
|
typedef struct {
|
|
double for_lu;
|
|
double total_needed;
|
|
int expansions;
|
|
} mem_usage_t;
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
/* Driver routines */
|
|
extern void
|
|
sgssv(superlu_options_t *, SuperMatrix *, int *, int *, SuperMatrix *,
|
|
SuperMatrix *, SuperMatrix *, SuperLUStat_t *, int *);
|
|
extern void
|
|
sgssvx(superlu_options_t *, SuperMatrix *, int *, int *, int *,
|
|
char *, double *, double *, SuperMatrix *, SuperMatrix *,
|
|
void *, int, SuperMatrix *, SuperMatrix *,
|
|
double *, double *, double *, double *,
|
|
mem_usage_t *, SuperLUStat_t *, int *);
|
|
|
|
/* Supernodal LU factor related */
|
|
extern void
|
|
sCreate_CompCol_Matrix(SuperMatrix *, int, int, int, double *,
|
|
int *, int *, Stype_t, Dtype_t, Mtype_t);
|
|
extern void
|
|
sCreate_CompRow_Matrix(SuperMatrix *, int, int, int, double *,
|
|
int *, int *, Stype_t, Dtype_t, Mtype_t);
|
|
extern void
|
|
sCopy_CompCol_Matrix(SuperMatrix *, SuperMatrix *);
|
|
extern void
|
|
sCreate_Dense_Matrix(SuperMatrix *, int, int, double *, int,
|
|
Stype_t, Dtype_t, Mtype_t);
|
|
extern void
|
|
sCreate_SuperNode_Matrix(SuperMatrix *, int, int, int, double *,
|
|
int *, int *, int *, int *, int *,
|
|
Stype_t, Dtype_t, Mtype_t);
|
|
extern void
|
|
sCopy_Dense_Matrix(int, int, double *, int, double *, int);
|
|
|
|
extern void countnz (const int, int *, int *, int *, GlobalLU_t *);
|
|
extern void fixupL (const int, const int *, GlobalLU_t *);
|
|
|
|
extern void sallocateA (int, int, double **, int **, int **);
|
|
extern void sgstrf (superlu_options_t*, SuperMatrix*,
|
|
int, int, int*, void *, int, int *, int *,
|
|
SuperMatrix *, SuperMatrix *, SuperLUStat_t*, int *);
|
|
extern int ssnode_dfs (const int, const int, const int *, const int *,
|
|
const int *, int *, int *, GlobalLU_t *);
|
|
extern int ssnode_bmod (const int, const int, double *,
|
|
double *, GlobalLU_t *, SuperLUStat_t*);
|
|
extern void spanel_dfs (const int, const int, const int, SuperMatrix *,
|
|
int *, int *, double *, int *, int *, int *,
|
|
int *, int *, int *, int *, GlobalLU_t *);
|
|
extern void spanel_bmod (const int, const int, const int, const int,
|
|
double *, double *, int *, int *,
|
|
GlobalLU_t *, SuperLUStat_t*);
|
|
extern int scolumn_dfs (const int, const int, int *, int *, int *, int *,
|
|
int *, int *, int *, int *, int *, GlobalLU_t *);
|
|
extern int scolumn_bmod (const int, const int, double *,
|
|
double *, int *, int *, int,
|
|
GlobalLU_t *, SuperLUStat_t*);
|
|
extern int scopy_to_ucol (int, int, int *, int *, int *,
|
|
double *, GlobalLU_t *);
|
|
extern int spivotL (const int, const double, int *, int *,
|
|
int *, int *, int *, GlobalLU_t *, SuperLUStat_t*);
|
|
extern void spruneL (const int, const int *, const int, const int,
|
|
const int *, const int *, int *, GlobalLU_t *);
|
|
extern void sreadmt (int *, int *, int *, double **, int **, int **);
|
|
extern void sGenXtrue (int, int, double *, int);
|
|
extern void sFillRHS (trans_t, int, double *, int, SuperMatrix *,
|
|
SuperMatrix *);
|
|
extern void sgstrs (trans_t, SuperMatrix *, SuperMatrix *, int *, int *,
|
|
SuperMatrix *, SuperLUStat_t*, int *);
|
|
|
|
|
|
/* Driver related */
|
|
|
|
extern void sgsequ (SuperMatrix *, double *, double *, double *,
|
|
double *, double *, int *);
|
|
extern void slaqgs (SuperMatrix *, double *, double *, double,
|
|
double, double, char *);
|
|
extern void sgscon (char *, SuperMatrix *, SuperMatrix *,
|
|
double, double *, SuperLUStat_t*, int *);
|
|
extern double sPivotGrowth(int, SuperMatrix *, int *,
|
|
SuperMatrix *, SuperMatrix *);
|
|
extern void sgsrfs (trans_t, SuperMatrix *, SuperMatrix *,
|
|
SuperMatrix *, int *, int *, char *, double *,
|
|
double *, SuperMatrix *, SuperMatrix *,
|
|
double *, double *, SuperLUStat_t*, int *);
|
|
|
|
extern int sp_strsv (char *, char *, char *, SuperMatrix *,
|
|
SuperMatrix *, double *, SuperLUStat_t*, int *);
|
|
extern int sp_sgemv (char *, double, SuperMatrix *, double *,
|
|
int, double, double *, int);
|
|
|
|
extern int sp_sgemm (char *, int, double,
|
|
SuperMatrix *, double *, int, double,
|
|
double *, int);
|
|
|
|
/* Memory-related */
|
|
extern int sLUMemInit (fact_t, void *, int, int, int, int, int,
|
|
SuperMatrix *, SuperMatrix *,
|
|
GlobalLU_t *, int **, double **);
|
|
extern void sSetRWork (int, int, double *, double **, double **);
|
|
extern void sLUWorkFree (int *, double *, GlobalLU_t *);
|
|
extern int sLUMemXpand (int, int, MemType, int *, GlobalLU_t *);
|
|
|
|
extern double *doubleMalloc(int);
|
|
extern double *doubleCalloc(int);
|
|
extern int smemory_usage(const int, const int, const int, const int);
|
|
extern int sQuerySpace (SuperMatrix *, SuperMatrix *, mem_usage_t *);
|
|
|
|
/* Auxiliary routines */
|
|
extern void sreadhb(int *, int *, int *, double **, int **, int **);
|
|
extern void sCompRow_to_CompCol(int, int, int, double*, int*, int*,
|
|
double **, int **, int **);
|
|
extern void sfill (double *, int, double);
|
|
extern void sinf_norm_error (int, SuperMatrix *, double *);
|
|
extern void PrintPerf (SuperMatrix *, SuperMatrix *, mem_usage_t *,
|
|
double, double, double *, double *, char *);
|
|
|
|
/* Routines for debugging */
|
|
extern void sPrint_CompCol_Matrix(char *, SuperMatrix *);
|
|
extern void sPrint_SuperNode_Matrix(char *, SuperMatrix *);
|
|
extern void sPrint_Dense_Matrix(char *, SuperMatrix *);
|
|
extern void print_lu_col(char *, int, int, int *, GlobalLU_t *);
|
|
extern void check_tempv(int, double *);
|
|
extern int print_int_vec(char *what, int n, int *vec);
|
|
|
|
extern int sp_symetree(int *acolst, int *acolend, int *arow, int n, int *parent);
|
|
extern void sprint_lu_col(char *msg, int jcol, int pivrow, int *xprune, GlobalLU_t *Glu); // added to build with debug for blender - campbell
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|
|
|
|
#endif /* __SUPERLU_sSP_DEFS */
|
|
|