1 |
#include <stdio.h> |
2 |
#include <math.h> |
3 |
#include <stdlib.h> |
4 |
#include "MatVec3.h" |
5 |
|
6 |
/* |
7 |
* Contains various utilities for dealing with 3x3 matrices and |
8 |
* length 3 vectors |
9 |
*/ |
10 |
|
11 |
void identityMat3(double A[3][3]) { |
12 |
int i; |
13 |
for (i = 0; i < 3; i++) { |
14 |
A[i][0] = A[i][1] = A[i][2] = 0.0; |
15 |
A[i][i] = 1.0; |
16 |
} |
17 |
} |
18 |
|
19 |
void swapVectors3(double v1[3], double v2[3]) { |
20 |
int i; |
21 |
for (i = 0; i < 3; i++) { |
22 |
double tmp = v1[i]; |
23 |
v1[i] = v2[i]; |
24 |
v2[i] = tmp; |
25 |
} |
26 |
} |
27 |
|
28 |
double normalize3(double x[3]) { |
29 |
double den; |
30 |
int i; |
31 |
if ( (den = norm3(x)) != 0.0 ) { |
32 |
for (i=0; i < 3; i++) |
33 |
{ |
34 |
x[i] /= den; |
35 |
} |
36 |
} |
37 |
return den; |
38 |
} |
39 |
|
40 |
void matMul3(double a[3][3], double b[3][3], double c[3][3]) { |
41 |
double r00, r01, r02, r10, r11, r12, r20, r21, r22; |
42 |
|
43 |
r00 = a[0][0]*b[0][0] + a[0][1]*b[1][0] + a[0][2]*b[2][0]; |
44 |
r01 = a[0][0]*b[0][1] + a[0][1]*b[1][1] + a[0][2]*b[2][1]; |
45 |
r02 = a[0][0]*b[0][2] + a[0][1]*b[1][2] + a[0][2]*b[2][2]; |
46 |
|
47 |
r10 = a[1][0]*b[0][0] + a[1][1]*b[1][0] + a[1][2]*b[2][0]; |
48 |
r11 = a[1][0]*b[0][1] + a[1][1]*b[1][1] + a[1][2]*b[2][1]; |
49 |
r12 = a[1][0]*b[0][2] + a[1][1]*b[1][2] + a[1][2]*b[2][2]; |
50 |
|
51 |
r20 = a[2][0]*b[0][0] + a[2][1]*b[1][0] + a[2][2]*b[2][0]; |
52 |
r21 = a[2][0]*b[0][1] + a[2][1]*b[1][1] + a[2][2]*b[2][1]; |
53 |
r22 = a[2][0]*b[0][2] + a[2][1]*b[1][2] + a[2][2]*b[2][2]; |
54 |
|
55 |
c[0][0] = r00; c[0][1] = r01; c[0][2] = r02; |
56 |
c[1][0] = r10; c[1][1] = r11; c[1][2] = r12; |
57 |
c[2][0] = r20; c[2][1] = r21; c[2][2] = r22; |
58 |
} |
59 |
|
60 |
void matVecMul3(double m[3][3], double inVec[3], double outVec[3]) { |
61 |
double a0, a1, a2; |
62 |
|
63 |
a0 = inVec[0]; a1 = inVec[1]; a2 = inVec[2]; |
64 |
|
65 |
outVec[0] = m[0][0]*a0 + m[0][1]*a1 + m[0][2]*a2; |
66 |
outVec[1] = m[1][0]*a0 + m[1][1]*a1 + m[1][2]*a2; |
67 |
outVec[2] = m[2][0]*a0 + m[2][1]*a1 + m[2][2]*a2; |
68 |
} |
69 |
|
70 |
double matDet3(double a[3][3]) { |
71 |
int i, j, k; |
72 |
double determinant; |
73 |
|
74 |
determinant = 0.0; |
75 |
|
76 |
for(i = 0; i < 3; i++) { |
77 |
j = (i+1)%3; |
78 |
k = (i+2)%3; |
79 |
|
80 |
determinant += a[0][i] * (a[1][j]*a[2][k] - a[1][k]*a[2][j]); |
81 |
} |
82 |
|
83 |
return determinant; |
84 |
} |
85 |
|
86 |
void invertMat3(double a[3][3], double b[3][3]) { |
87 |
|
88 |
int i, j, k, l, m, n; |
89 |
double determinant; |
90 |
|
91 |
determinant = matDet3( a ); |
92 |
|
93 |
if (determinant == 0.0) { |
94 |
sprintf( painCave.errMsg, |
95 |
"Can't invert a matrix with a zero determinant!\n"); |
96 |
painCave.isFatal = 1; |
97 |
simError(); |
98 |
} |
99 |
|
100 |
for (i=0; i < 3; i++) { |
101 |
j = (i+1)%3; |
102 |
k = (i+2)%3; |
103 |
for(l = 0; l < 3; l++) { |
104 |
m = (l+1)%3; |
105 |
n = (l+2)%3; |
106 |
|
107 |
b[l][i] = (a[j][m]*a[k][n] - a[j][n]*a[k][m]) / determinant; |
108 |
} |
109 |
} |
110 |
} |
111 |
|
112 |
void transposeMat3(double in[3][3], double out[3][3]) { |
113 |
double temp[3][3]; |
114 |
int i, j; |
115 |
|
116 |
for (i = 0; i < 3; i++) { |
117 |
for (j = 0; j < 3; j++) { |
118 |
temp[j][i] = in[i][j]; |
119 |
} |
120 |
} |
121 |
for (i = 0; i < 3; i++) { |
122 |
for (j = 0; j < 3; j++) { |
123 |
out[i][j] = temp[i][j]; |
124 |
} |
125 |
} |
126 |
} |
127 |
|
128 |
void printMat3(double A[3][3] ){ |
129 |
|
130 |
fprintf(stderr, "[ %g, %g, %g ]\n[ %g, %g, %g ]\n[ %g, %g, %g ]\n", |
131 |
A[0][0] , A[0][1] , A[0][2], |
132 |
A[1][0] , A[1][1] , A[1][2], |
133 |
A[2][0] , A[2][1] , A[2][2]) ; |
134 |
} |
135 |
|
136 |
void printMat9(double A[9] ){ |
137 |
|
138 |
fprintf(stderr, "[ %g, %g, %g ]\n[ %g, %g, %g ]\n[ %g, %g, %g ]\n", |
139 |
A[0], A[1], A[2], |
140 |
A[3], A[4], A[5], |
141 |
A[6], A[7], A[8]); |
142 |
} |
143 |
|
144 |
double matTrace3(double m[3][3]){ |
145 |
double trace; |
146 |
trace = m[0][0] + m[1][1] + m[2][2]; |
147 |
|
148 |
return trace; |
149 |
} |
150 |
|
151 |
void crossProduct3(double a[3],double b[3], double out[3]){ |
152 |
|
153 |
out[0] = a[1] * b[2] - a[2] * b[1]; |
154 |
out[1] = a[2] * b[0] - a[0] * b[2] ; |
155 |
out[2] = a[0] * b[1] - a[1] * b[0]; |
156 |
|
157 |
} |
158 |
|
159 |
double dotProduct3(double a[3], double b[3]){ |
160 |
return a[0]*b[0] + a[1]*b[1]+ a[2]*b[2]; |
161 |
} |
162 |
|
163 |
//---------------------------------------------------------------------------- |
164 |
// Extract the eigenvalues and eigenvectors from a 3x3 matrix. |
165 |
// The eigenvectors (the columns of V) will be normalized. |
166 |
// The eigenvectors are aligned optimally with the x, y, and z |
167 |
// axes respectively. |
168 |
|
169 |
void diagonalize3x3(const double A[3][3], double w[3], double V[3][3]) { |
170 |
int i,j,k,maxI; |
171 |
double tmp, maxVal; |
172 |
|
173 |
// do the matrix[3][3] to **matrix conversion for Jacobi |
174 |
double C[3][3]; |
175 |
double *ATemp[3],*VTemp[3]; |
176 |
for (i = 0; i < 3; i++) |
177 |
{ |
178 |
C[i][0] = A[i][0]; |
179 |
C[i][1] = A[i][1]; |
180 |
C[i][2] = A[i][2]; |
181 |
ATemp[i] = C[i]; |
182 |
VTemp[i] = V[i]; |
183 |
} |
184 |
|
185 |
// diagonalize using Jacobi |
186 |
JacobiN(ATemp,3,w,VTemp); |
187 |
|
188 |
// if all the eigenvalues are the same, return identity matrix |
189 |
if (w[0] == w[1] && w[0] == w[2]) |
190 |
{ |
191 |
identityMat3(V); |
192 |
return; |
193 |
} |
194 |
|
195 |
// transpose temporarily, it makes it easier to sort the eigenvectors |
196 |
transposeMat3(V,V); |
197 |
|
198 |
// if two eigenvalues are the same, re-orthogonalize to optimally line |
199 |
// up the eigenvectors with the x, y, and z axes |
200 |
for (i = 0; i < 3; i++) |
201 |
{ |
202 |
if (w[(i+1)%3] == w[(i+2)%3]) // two eigenvalues are the same |
203 |
{ |
204 |
// find maximum element of the independant eigenvector |
205 |
maxVal = fabs(V[i][0]); |
206 |
maxI = 0; |
207 |
for (j = 1; j < 3; j++) |
208 |
{ |
209 |
if (maxVal < (tmp = fabs(V[i][j]))) |
210 |
{ |
211 |
maxVal = tmp; |
212 |
maxI = j; |
213 |
} |
214 |
} |
215 |
// swap the eigenvector into its proper position |
216 |
if (maxI != i) |
217 |
{ |
218 |
tmp = w[maxI]; |
219 |
w[maxI] = w[i]; |
220 |
w[i] = tmp; |
221 |
swapVectors3(V[i],V[maxI]); |
222 |
} |
223 |
// maximum element of eigenvector should be positive |
224 |
if (V[maxI][maxI] < 0) |
225 |
{ |
226 |
V[maxI][0] = -V[maxI][0]; |
227 |
V[maxI][1] = -V[maxI][1]; |
228 |
V[maxI][2] = -V[maxI][2]; |
229 |
} |
230 |
|
231 |
// re-orthogonalize the other two eigenvectors |
232 |
j = (maxI+1)%3; |
233 |
k = (maxI+2)%3; |
234 |
|
235 |
V[j][0] = 0.0; |
236 |
V[j][1] = 0.0; |
237 |
V[j][2] = 0.0; |
238 |
V[j][j] = 1.0; |
239 |
crossProduct3(V[maxI],V[j],V[k]); |
240 |
normalize3(V[k]); |
241 |
crossProduct3(V[k],V[maxI],V[j]); |
242 |
|
243 |
// transpose vectors back to columns |
244 |
transposeMat3(V,V); |
245 |
return; |
246 |
} |
247 |
} |
248 |
|
249 |
// the three eigenvalues are different, just sort the eigenvectors |
250 |
// to align them with the x, y, and z axes |
251 |
|
252 |
// find the vector with the largest x element, make that vector |
253 |
// the first vector |
254 |
maxVal = fabs(V[0][0]); |
255 |
maxI = 0; |
256 |
for (i = 1; i < 3; i++) |
257 |
{ |
258 |
if (maxVal < (tmp = fabs(V[i][0]))) |
259 |
{ |
260 |
maxVal = tmp; |
261 |
maxI = i; |
262 |
} |
263 |
} |
264 |
// swap eigenvalue and eigenvector |
265 |
if (maxI != 0) |
266 |
{ |
267 |
tmp = w[maxI]; |
268 |
w[maxI] = w[0]; |
269 |
w[0] = tmp; |
270 |
swapVectors3(V[maxI],V[0]); |
271 |
} |
272 |
// do the same for the y element |
273 |
if (fabs(V[1][1]) < fabs(V[2][1])) |
274 |
{ |
275 |
tmp = w[2]; |
276 |
w[2] = w[1]; |
277 |
w[1] = tmp; |
278 |
swapVectors3(V[2],V[1]); |
279 |
} |
280 |
|
281 |
// ensure that the sign of the eigenvectors is correct |
282 |
for (i = 0; i < 2; i++) |
283 |
{ |
284 |
if (V[i][i] < 0) |
285 |
{ |
286 |
V[i][0] = -V[i][0]; |
287 |
V[i][1] = -V[i][1]; |
288 |
V[i][2] = -V[i][2]; |
289 |
} |
290 |
} |
291 |
// set sign of final eigenvector to ensure that determinant is positive |
292 |
if (matDet3(V) < 0) |
293 |
{ |
294 |
V[2][0] = -V[2][0]; |
295 |
V[2][1] = -V[2][1]; |
296 |
V[2][2] = -V[2][2]; |
297 |
} |
298 |
|
299 |
// transpose the eigenvectors back again |
300 |
transposeMat3(V,V); |
301 |
} |
302 |
|
303 |
|
304 |
#define MAT_ROTATE(a,i,j,k,l) g=a[i][j];h=a[k][l];a[i][j]=g-s*(h+g*tau); a[k][l]=h+s*(g-h*tau); |
305 |
|
306 |
#define MAX_ROTATIONS 20 |
307 |
|
308 |
// Jacobi iteration for the solution of eigenvectors/eigenvalues of a nxn |
309 |
// real symmetric matrix. Square nxn matrix a; size of matrix in n; |
310 |
// output eigenvalues in w; and output eigenvectors in v. Resulting |
311 |
// eigenvalues/vectors are sorted in decreasing order; eigenvectors are |
312 |
// normalized. |
313 |
int JacobiN(double **a, int n, double *w, double **v) { |
314 |
|
315 |
int i, j, k, iq, ip, numPos; |
316 |
int ceil_half_n; |
317 |
double tresh, theta, tau, t, sm, s, h, g, c, tmp; |
318 |
double bspace[4], zspace[4]; |
319 |
double *b = bspace; |
320 |
double *z = zspace; |
321 |
|
322 |
|
323 |
// only allocate memory if the matrix is large |
324 |
if (n > 4) |
325 |
{ |
326 |
b = (double *) calloc(n, sizeof(double)); |
327 |
z = (double *) calloc(n, sizeof(double)); |
328 |
} |
329 |
|
330 |
// initialize |
331 |
for (ip=0; ip<n; ip++) |
332 |
{ |
333 |
for (iq=0; iq<n; iq++) |
334 |
{ |
335 |
v[ip][iq] = 0.0; |
336 |
} |
337 |
v[ip][ip] = 1.0; |
338 |
} |
339 |
for (ip=0; ip<n; ip++) |
340 |
{ |
341 |
b[ip] = w[ip] = a[ip][ip]; |
342 |
z[ip] = 0.0; |
343 |
} |
344 |
|
345 |
// begin rotation sequence |
346 |
for (i=0; i<MAX_ROTATIONS; i++) |
347 |
{ |
348 |
sm = 0.0; |
349 |
for (ip=0; ip<n-1; ip++) |
350 |
{ |
351 |
for (iq=ip+1; iq<n; iq++) |
352 |
{ |
353 |
sm += fabs(a[ip][iq]); |
354 |
} |
355 |
} |
356 |
if (sm == 0.0) |
357 |
{ |
358 |
break; |
359 |
} |
360 |
|
361 |
if (i < 3) // first 3 sweeps |
362 |
{ |
363 |
tresh = 0.2*sm/(n*n); |
364 |
} |
365 |
else |
366 |
{ |
367 |
tresh = 0.0; |
368 |
} |
369 |
|
370 |
for (ip=0; ip<n-1; ip++) |
371 |
{ |
372 |
for (iq=ip+1; iq<n; iq++) |
373 |
{ |
374 |
g = 100.0*fabs(a[ip][iq]); |
375 |
|
376 |
// after 4 sweeps |
377 |
if (i > 3 && (fabs(w[ip])+g) == fabs(w[ip]) |
378 |
&& (fabs(w[iq])+g) == fabs(w[iq])) |
379 |
{ |
380 |
a[ip][iq] = 0.0; |
381 |
} |
382 |
else if (fabs(a[ip][iq]) > tresh) |
383 |
{ |
384 |
h = w[iq] - w[ip]; |
385 |
if ( (fabs(h)+g) == fabs(h)) |
386 |
{ |
387 |
t = (a[ip][iq]) / h; |
388 |
} |
389 |
else |
390 |
{ |
391 |
theta = 0.5*h / (a[ip][iq]); |
392 |
t = 1.0 / (fabs(theta)+sqrt(1.0+theta*theta)); |
393 |
if (theta < 0.0) |
394 |
{ |
395 |
t = -t; |
396 |
} |
397 |
} |
398 |
c = 1.0 / sqrt(1+t*t); |
399 |
s = t*c; |
400 |
tau = s/(1.0+c); |
401 |
h = t*a[ip][iq]; |
402 |
z[ip] -= h; |
403 |
z[iq] += h; |
404 |
w[ip] -= h; |
405 |
w[iq] += h; |
406 |
a[ip][iq]=0.0; |
407 |
|
408 |
// ip already shifted left by 1 unit |
409 |
for (j = 0;j <= ip-1;j++) |
410 |
{ |
411 |
MAT_ROTATE(a,j,ip,j,iq) |
412 |
} |
413 |
// ip and iq already shifted left by 1 unit |
414 |
for (j = ip+1;j <= iq-1;j++) |
415 |
{ |
416 |
MAT_ROTATE(a,ip,j,j,iq) |
417 |
} |
418 |
// iq already shifted left by 1 unit |
419 |
for (j=iq+1; j<n; j++) |
420 |
{ |
421 |
MAT_ROTATE(a,ip,j,iq,j) |
422 |
} |
423 |
for (j=0; j<n; j++) |
424 |
{ |
425 |
MAT_ROTATE(v,j,ip,j,iq) |
426 |
} |
427 |
} |
428 |
} |
429 |
} |
430 |
|
431 |
for (ip=0; ip<n; ip++) |
432 |
{ |
433 |
b[ip] += z[ip]; |
434 |
w[ip] = b[ip]; |
435 |
z[ip] = 0.0; |
436 |
} |
437 |
} |
438 |
|
439 |
//// this is NEVER called |
440 |
if ( i >= MAX_ROTATIONS ) |
441 |
{ |
442 |
sprintf( painCave.errMsg, |
443 |
"Jacobi: Error extracting eigenfunctions!\n"); |
444 |
painCave.isFatal = 1; |
445 |
simError(); |
446 |
return 0; |
447 |
} |
448 |
|
449 |
// sort eigenfunctions these changes do not affect accuracy |
450 |
for (j=0; j<n-1; j++) // boundary incorrect |
451 |
{ |
452 |
k = j; |
453 |
tmp = w[k]; |
454 |
for (i=j+1; i<n; i++) // boundary incorrect, shifted already |
455 |
{ |
456 |
if (w[i] >= tmp) // why exchage if same? |
457 |
{ |
458 |
k = i; |
459 |
tmp = w[k]; |
460 |
} |
461 |
} |
462 |
if (k != j) |
463 |
{ |
464 |
w[k] = w[j]; |
465 |
w[j] = tmp; |
466 |
for (i=0; i<n; i++) |
467 |
{ |
468 |
tmp = v[i][j]; |
469 |
v[i][j] = v[i][k]; |
470 |
v[i][k] = tmp; |
471 |
} |
472 |
} |
473 |
} |
474 |
// insure eigenvector consistency (i.e., Jacobi can compute vectors that |
475 |
// are negative of one another (.707,.707,0) and (-.707,-.707,0). This can |
476 |
// reek havoc in hyperstreamline/other stuff. We will select the most |
477 |
// positive eigenvector. |
478 |
ceil_half_n = (n >> 1) + (n & 1); |
479 |
for (j=0; j<n; j++) |
480 |
{ |
481 |
for (numPos=0, i=0; i<n; i++) |
482 |
{ |
483 |
if ( v[i][j] >= 0.0 ) |
484 |
{ |
485 |
numPos++; |
486 |
} |
487 |
} |
488 |
// if ( numPos < ceil(double(n)/double(2.0)) ) |
489 |
if ( numPos < ceil_half_n) |
490 |
{ |
491 |
for(i=0; i<n; i++) |
492 |
{ |
493 |
v[i][j] *= -1.0; |
494 |
} |
495 |
} |
496 |
} |
497 |
|
498 |
if (n > 4) |
499 |
{ |
500 |
free(b); |
501 |
free(z); |
502 |
} |
503 |
return 1; |
504 |
} |
505 |
|
506 |
#undef MAT_ROTATE |
507 |
#undef MAX_ROTATIONS |