MPQC 3.0.0-alpha
Loading...
Searching...
No Matches
misc.h
1
2//
3// Levenberg - Marquardt non-linear minimization algorithm
4// Copyright (C) 2004 Manolis Lourakis (lourakis@ics.forth.gr)
5// Institute of Computer Science, Foundation for Research & Technology - Hellas
6// Heraklion, Crete, Greece.
7//
8// This program is free software; you can redistribute it and/or modify
9// it under the terms of the GNU General Public License as published by
10// the Free Software Foundation; either version 2 of the License, or
11// (at your option) any later version.
12//
13// This program is distributed in the hope that it will be useful,
14// but WITHOUT ANY WARRANTY; without even the implied warranty of
15// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16// GNU General Public License for more details.
17//
19
20#ifndef _MISC_H_
21#define _MISC_H_
22
23/* common prefix for BLAS subroutines */
24#define LM_BLAS_PREFIX // none
25//#define LM_BLAS_PREFIX f2c_ // f2c'd BLAS
26//#define LM_BLAS_PREFIX cblas_ // C interface to BLAS
27
28#define LCAT_(a, b) #a b
29#define LCAT(a, b) LCAT_(a, b) // force substitution
30#define RCAT_(a, b) a #b
31#define RCAT(a, b) RCAT_(a, b) // force substitution
32
33#define __BLOCKSZ__ 32 /* block size for cache-friendly matrix-matrix multiply. It should be
34 * such that __BLOCKSZ__^2*sizeof(LM_REAL) is smaller than the CPU (L1)
35 * data cache size. Notice that a value of 32 when LM_REAL=double assumes
36 * an 8Kb L1 data cache (32*32*8=8K). This is a concervative choice since
37 * newer Pentium 4s have a L1 data cache of size 16K, capable of holding
38 * up to 45x45 double blocks.
39 */
40#define __BLOCKSZ__SQ (__BLOCKSZ__)*(__BLOCKSZ__)
41
42#ifdef _MSC_VER
43#define inline __inline //MSVC
44#elif !defined(__GNUC__)
45#define inline //other than MSVC, GCC: define empty
46#endif
47
48/* add a prefix in front of a token */
49#define LM_CAT__(a, b) a ## b
50#define LM_CAT_(a, b) LM_CAT__(a, b) // force substitution
51#define LM_ADD_PREFIX(s) LM_CAT_(LM_PREFIX, s)
52
53#ifdef __cplusplus
54extern "C" {
55#endif
56
57/* blocking-based matrix multiply */
58extern void strans_mat_mat_mult(float *a, float *b, int n, int m);
59extern void dtrans_mat_mat_mult(double *a, double *b, int n, int m);
60
61/* forward finite differences */
62extern void sfdif_forw_jac_approx(void (*func)(float *p, float *hx, int m, int n, void *adata),
63 float *p, float *hx, float *hxx, float delta,
64 float *jac, int m, int n, void *adata);
65extern void dfdif_forw_jac_approx(void (*func)(double *p, double *hx, int m, int n, void *adata),
66 double *p, double *hx, double *hxx, double delta,
67 double *jac, int m, int n, void *adata);
68
69/* central finite differences */
70extern void sfdif_cent_jac_approx(void (*func)(float *p, float *hx, int m, int n, void *adata),
71 float *p, float *hxm, float *hxp, float delta,
72 float *jac, int m, int n, void *adata);
73extern void dfdif_cent_jac_approx(void (*func)(double *p, double *hx, int m, int n, void *adata),
74 double *p, double *hxm, double *hxp, double delta,
75 double *jac, int m, int n, void *adata);
76
77/* covariance of LS fit */
78extern int slevmar_covar(float *JtJ, float *C, float sumsq, int m, int n);
79extern int dlevmar_covar(double *JtJ, double *C, double sumsq, int m, int n);
80
81#ifdef __cplusplus
82}
83#endif
84
85#endif /* _MISC_H */

Generated at Wed Sep 25 2024 02:45:30 for MPQC 3.0.0-alpha using the documentation package Doxygen 1.12.0.