| 10 |
|
#include "copyright.h" |
| 11 |
|
|
| 12 |
|
#include "ray.h" |
| 13 |
– |
|
| 13 |
|
#include "ambient.h" |
| 15 |
– |
|
| 14 |
|
#include "random.h" |
| 15 |
|
|
| 16 |
+ |
#ifdef NEWAMB |
| 17 |
|
|
| 18 |
+ |
extern void SDsquare2disk(double ds[2], double seedx, double seedy); |
| 19 |
+ |
|
| 20 |
+ |
typedef struct { |
| 21 |
+ |
RAY *rp; /* originating ray sample */ |
| 22 |
+ |
FVECT ux, uy; /* tangent axis directions */ |
| 23 |
+ |
int ns; /* number of samples per axis */ |
| 24 |
+ |
COLOR acoef; /* division contribution coefficient */ |
| 25 |
+ |
struct s_ambsamp { |
| 26 |
+ |
COLOR v; /* hemisphere sample value */ |
| 27 |
+ |
float p[3]; /* intersection point */ |
| 28 |
+ |
} sa[1]; /* sample array (extends struct) */ |
| 29 |
+ |
} AMBHEMI; /* ambient sample hemisphere */ |
| 30 |
+ |
|
| 31 |
+ |
#define ambsamp(h,i,j) (h)->sa[(i)*(h)->ns + (j)] |
| 32 |
+ |
|
| 33 |
+ |
|
| 34 |
+ |
static AMBHEMI * |
| 35 |
+ |
inithemi( /* initialize sampling hemisphere */ |
| 36 |
+ |
COLOR ac, |
| 37 |
+ |
RAY *r, |
| 38 |
+ |
double wt |
| 39 |
+ |
) |
| 40 |
+ |
{ |
| 41 |
+ |
AMBHEMI *hp; |
| 42 |
+ |
double d; |
| 43 |
+ |
int n, i; |
| 44 |
+ |
/* set number of divisions */ |
| 45 |
+ |
if (ambacc <= FTINY && |
| 46 |
+ |
wt > (d = 0.8*intens(ac)*r->rweight/(ambdiv*minweight))) |
| 47 |
+ |
wt = d; /* avoid ray termination */ |
| 48 |
+ |
n = sqrt(ambdiv * wt) + 0.5; |
| 49 |
+ |
i = 1 + 4*(ambacc > FTINY); /* minimum number of samples */ |
| 50 |
+ |
if (n < i) |
| 51 |
+ |
n = i; |
| 52 |
+ |
/* allocate sampling array */ |
| 53 |
+ |
hp = (AMBHEMI *)malloc(sizeof(AMBHEMI) + |
| 54 |
+ |
sizeof(struct s_ambsamp)*(n*n - 1)); |
| 55 |
+ |
if (hp == NULL) |
| 56 |
+ |
return(NULL); |
| 57 |
+ |
hp->rp = r; |
| 58 |
+ |
hp->ns = n; |
| 59 |
+ |
/* assign coefficient */ |
| 60 |
+ |
copycolor(hp->acoef, ac); |
| 61 |
+ |
d = 1.0/(n*n); |
| 62 |
+ |
scalecolor(hp->acoef, d); |
| 63 |
+ |
/* make tangent axes */ |
| 64 |
+ |
hp->uy[0] = hp->uy[1] = hp->uy[2] = 0.0; |
| 65 |
+ |
for (i = 0; i < 3; i++) |
| 66 |
+ |
if (r->rn[i] < 0.6 && r->rn[i] > -0.6) |
| 67 |
+ |
break; |
| 68 |
+ |
if (i >= 3) |
| 69 |
+ |
error(CONSISTENCY, "bad ray direction in inithemi()"); |
| 70 |
+ |
hp->uy[i] = 1.0; |
| 71 |
+ |
VCROSS(hp->ux, hp->uy, r->rn); |
| 72 |
+ |
normalize(hp->ux); |
| 73 |
+ |
VCROSS(hp->uy, r->rn, hp->ux); |
| 74 |
+ |
/* we're ready to sample */ |
| 75 |
+ |
return(hp); |
| 76 |
+ |
} |
| 77 |
+ |
|
| 78 |
+ |
|
| 79 |
+ |
static int |
| 80 |
+ |
ambsample( /* sample an ambient direction */ |
| 81 |
+ |
AMBHEMI *hp, |
| 82 |
+ |
int i, |
| 83 |
+ |
int j, |
| 84 |
+ |
) |
| 85 |
+ |
{ |
| 86 |
+ |
struct s_ambsamp *ap = &ambsamp(hp,i,j); |
| 87 |
+ |
RAY ar; |
| 88 |
+ |
int hlist[3]; |
| 89 |
+ |
double spt[2], dz; |
| 90 |
+ |
int ii; |
| 91 |
+ |
/* ambient coefficient for weight */ |
| 92 |
+ |
if (ambacc > FTINY) |
| 93 |
+ |
setcolor(ar.rcoef, AVGREFL, AVGREFL, AVGREFL); |
| 94 |
+ |
else |
| 95 |
+ |
copycolor(ar.rcoef, hp->acoef); |
| 96 |
+ |
if (rayorigin(&ar, AMBIENT, hp->rp, ar.rcoef) < 0) { |
| 97 |
+ |
setcolor(ap->v, 0., 0., 0.); |
| 98 |
+ |
ap->r = 0.; |
| 99 |
+ |
return(0); /* no sample taken */ |
| 100 |
+ |
} |
| 101 |
+ |
if (ambacc > FTINY) { |
| 102 |
+ |
multcolor(ar.rcoef, hp->acoef); |
| 103 |
+ |
scalecolor(ar.rcoef, 1./AVGREFL); |
| 104 |
+ |
} |
| 105 |
+ |
/* generate hemispherical sample */ |
| 106 |
+ |
SDsquare2disk(spt, (i+frandom())/hp->ns, (j+frandom())/hp->ns); |
| 107 |
+ |
zd = sqrt(1. - spt[0]*spt[0] - spt[1]*spt[1]); |
| 108 |
+ |
for (ii = 3; ii--; ) |
| 109 |
+ |
ar.rdir[ii] = spt[0]*hp->ux[ii] + |
| 110 |
+ |
spt[1]*hp->uy[ii] + |
| 111 |
+ |
zd*hp->rp->ron[ii]; |
| 112 |
+ |
checknorm(ar.rdir); |
| 113 |
+ |
dimlist[ndims++] = i*hp->ns + j + 90171; |
| 114 |
+ |
rayvalue(&ar); /* evaluate ray */ |
| 115 |
+ |
ndims--; |
| 116 |
+ |
multcolor(ar.rcol, ar.rcoef); /* apply coefficient */ |
| 117 |
+ |
copycolor(ap->v, ar.rcol); |
| 118 |
+ |
if (ar.rt > 20.0*maxarad) /* limit vertex distance */ |
| 119 |
+ |
ar.rt = 20.0*maxarad; |
| 120 |
+ |
VSUM(ap->p, ar.rorg, ar.rdir, ar.rt); |
| 121 |
+ |
return(1); |
| 122 |
+ |
} |
| 123 |
+ |
|
| 124 |
+ |
|
| 125 |
+ |
static void |
| 126 |
+ |
ambHessian( /* anisotropic radii & pos. gradient */ |
| 127 |
+ |
AMBHEMI *hp, |
| 128 |
+ |
FVECT uv[2], /* returned */ |
| 129 |
+ |
float ra[2], /* returned */ |
| 130 |
+ |
float pg[2] /* returned */ |
| 131 |
+ |
) |
| 132 |
+ |
{ |
| 133 |
+ |
if (ra != NULL) { /* compute Hessian-derived radii */ |
| 134 |
+ |
} else { /* else copy original tangent axes */ |
| 135 |
+ |
VCOPY(uv[0], hp->ux); |
| 136 |
+ |
VCOPY(uv[1], hp->uy); |
| 137 |
+ |
} |
| 138 |
+ |
if (pg == NULL) /* no position gradient requested? */ |
| 139 |
+ |
return; |
| 140 |
+ |
} |
| 141 |
+ |
|
| 142 |
+ |
int |
| 143 |
+ |
doambient( /* compute ambient component */ |
| 144 |
+ |
COLOR rcol, /* input/output color */ |
| 145 |
+ |
RAY *r, |
| 146 |
+ |
double wt, |
| 147 |
+ |
FVECT uv[2], /* returned */ |
| 148 |
+ |
float ra[2], /* returned */ |
| 149 |
+ |
float pg[2], /* returned */ |
| 150 |
+ |
float dg[2] /* returned */ |
| 151 |
+ |
) |
| 152 |
+ |
{ |
| 153 |
+ |
int cnt = 0; |
| 154 |
+ |
FVECT my_uv[2]; |
| 155 |
+ |
AMBHEMI *hp; |
| 156 |
+ |
double d, acol[3]; |
| 157 |
+ |
struct s_ambsamp *ap; |
| 158 |
+ |
int i, j; |
| 159 |
+ |
/* initialize */ |
| 160 |
+ |
if ((hp = inithemi(rcol, r, wt)) == NULL) |
| 161 |
+ |
return(0); |
| 162 |
+ |
if (uv != NULL) |
| 163 |
+ |
memset(uv, 0, sizeof(FVECT)*2); |
| 164 |
+ |
if (ra != NULL) |
| 165 |
+ |
ra[0] = ra[1] = 0.0; |
| 166 |
+ |
if (pg != NULL) |
| 167 |
+ |
pg[0] = pg[1] = 0.0; |
| 168 |
+ |
if (dg != NULL) |
| 169 |
+ |
dg[0] = dg[1] = 0.0; |
| 170 |
+ |
/* sample the hemisphere */ |
| 171 |
+ |
acol[0] = acol[1] = acol[2] = 0.0; |
| 172 |
+ |
for (i = hemi.ns; i--; ) |
| 173 |
+ |
for (j = hemi.ns; j--; ) |
| 174 |
+ |
if (ambsample(hp, i, j)) { |
| 175 |
+ |
ap = &ambsamp(hp,i,j); |
| 176 |
+ |
addcolor(acol, ap->v); |
| 177 |
+ |
++cnt; |
| 178 |
+ |
} |
| 179 |
+ |
if (!cnt) { |
| 180 |
+ |
setcolor(rcol, 0.0, 0.0, 0.0); |
| 181 |
+ |
free(hp); |
| 182 |
+ |
return(0); /* no valid samples */ |
| 183 |
+ |
} |
| 184 |
+ |
d = 1.0 / cnt; /* final indirect irradiance/PI */ |
| 185 |
+ |
acol[0] *= d; acol[1] *= d; acol[2] *= d; |
| 186 |
+ |
copycolor(rcol, acol); |
| 187 |
+ |
if (cnt < hp->ns*hp->ns || /* incomplete sampling? */ |
| 188 |
+ |
(ra == NULL) & (pg == NULL) & (dg == NULL)) { |
| 189 |
+ |
free(hp); |
| 190 |
+ |
return(-1); /* no radius or gradient calc. */ |
| 191 |
+ |
} |
| 192 |
+ |
d = 0.01 * bright(rcol); /* add in 1% before Hessian comp. */ |
| 193 |
+ |
if (d < FTINY) d = FTINY; |
| 194 |
+ |
ap = hp->sa; /* using Y channel from here on... */ |
| 195 |
+ |
for (i = hp->ns*hp->ns; i--; ap++) |
| 196 |
+ |
colval(ap->v,CIEY) = bright(ap->v) + d; |
| 197 |
+ |
|
| 198 |
+ |
if (uv == NULL) /* make sure we have axis pointers */ |
| 199 |
+ |
uv = my_uv; |
| 200 |
+ |
/* compute radii & pos. gradient */ |
| 201 |
+ |
ambHessian(hp, uv, ra, pg); |
| 202 |
+ |
if (dg != NULL) /* compute direction gradient */ |
| 203 |
+ |
ambdirgrad(hp, uv, dg); |
| 204 |
+ |
if (ra != NULL) { /* adjust/clamp radii */ |
| 205 |
+ |
d = pow(wt, -0.25); |
| 206 |
+ |
if ((ra[0] *= d) > maxarad) |
| 207 |
+ |
ra[0] = maxarad; |
| 208 |
+ |
if ((ra[1] *= d) > 2.0*ra[0]) |
| 209 |
+ |
ra[1] = 2.0*ra[0]; |
| 210 |
+ |
} |
| 211 |
+ |
free(hp); /* clean up and return */ |
| 212 |
+ |
return(1); |
| 213 |
+ |
} |
| 214 |
+ |
|
| 215 |
+ |
|
| 216 |
+ |
#else /* ! NEWAMB */ |
| 217 |
+ |
|
| 218 |
+ |
|
| 219 |
|
void |
| 220 |
|
inithemi( /* initialize sampling hemisphere */ |
| 221 |
< |
register AMBHEMI *hp, |
| 221 |
> |
AMBHEMI *hp, |
| 222 |
|
COLOR ac, |
| 223 |
|
RAY *r, |
| 224 |
|
double wt |
| 225 |
|
) |
| 226 |
|
{ |
| 227 |
|
double d; |
| 228 |
< |
register int i; |
| 228 |
> |
int i; |
| 229 |
|
/* set number of divisions */ |
| 230 |
|
if (ambacc <= FTINY && |
| 231 |
< |
wt > (d = 0.8*bright(ac)*r->rweight/(ambdiv*minweight))) |
| 231 |
> |
wt > (d = 0.8*intens(ac)*r->rweight/(ambdiv*minweight))) |
| 232 |
|
wt = d; /* avoid ray termination */ |
| 233 |
|
hp->nt = sqrt(ambdiv * wt / PI) + 0.5; |
| 234 |
|
i = ambacc > FTINY ? 3 : 1; /* minimum number of samples */ |
| 258 |
|
|
| 259 |
|
int |
| 260 |
|
divsample( /* sample a division */ |
| 261 |
< |
register AMBSAMP *dp, |
| 261 |
> |
AMBSAMP *dp, |
| 262 |
|
AMBHEMI *h, |
| 263 |
|
RAY *r |
| 264 |
|
) |
| 269 |
|
double xd, yd, zd; |
| 270 |
|
double b2; |
| 271 |
|
double phi; |
| 272 |
< |
register int i; |
| 272 |
> |
int i; |
| 273 |
|
/* ambient coefficient for weight */ |
| 274 |
|
if (ambacc > FTINY) |
| 275 |
|
setcolor(ar.rcoef, AVGREFL, AVGREFL, AVGREFL); |
| 294 |
|
ar.rdir[i] = xd*h->ux[i] + |
| 295 |
|
yd*h->uy[i] + |
| 296 |
|
zd*h->uz[i]; |
| 297 |
+ |
checknorm(ar.rdir); |
| 298 |
|
dimlist[ndims++] = dp->t*h->np + dp->p + 90171; |
| 299 |
|
rayvalue(&ar); |
| 300 |
|
ndims--; |
| 339 |
|
{ |
| 340 |
|
const AMBSAMP *d1 = (const AMBSAMP *)p1; |
| 341 |
|
const AMBSAMP *d2 = (const AMBSAMP *)p2; |
| 342 |
< |
register int c; |
| 342 |
> |
int c; |
| 343 |
|
|
| 344 |
|
if ( (c = d1->t - d2->t) ) |
| 345 |
|
return(c); |
| 349 |
|
|
| 350 |
|
double |
| 351 |
|
doambient( /* compute ambient component */ |
| 352 |
< |
COLOR acol, |
| 352 |
> |
COLOR rcol, |
| 353 |
|
RAY *r, |
| 354 |
|
double wt, |
| 355 |
|
FVECT pg, |
| 356 |
|
FVECT dg |
| 357 |
|
) |
| 358 |
|
{ |
| 359 |
< |
double b, d; |
| 359 |
> |
double b, d=0; |
| 360 |
|
AMBHEMI hemi; |
| 361 |
|
AMBSAMP *div; |
| 362 |
|
AMBSAMP dnew; |
| 363 |
< |
register AMBSAMP *dp; |
| 363 |
> |
double acol[3]; |
| 364 |
> |
AMBSAMP *dp; |
| 365 |
|
double arad; |
| 366 |
< |
int ndivs; |
| 367 |
< |
register int i, j; |
| 366 |
> |
int divcnt; |
| 367 |
> |
int i, j; |
| 368 |
|
/* initialize hemisphere */ |
| 369 |
< |
inithemi(&hemi, acol, r, wt); |
| 370 |
< |
ndivs = hemi.nt * hemi.np; |
| 369 |
> |
inithemi(&hemi, rcol, r, wt); |
| 370 |
> |
divcnt = hemi.nt * hemi.np; |
| 371 |
|
/* initialize */ |
| 372 |
|
if (pg != NULL) |
| 373 |
|
pg[0] = pg[1] = pg[2] = 0.0; |
| 374 |
|
if (dg != NULL) |
| 375 |
|
dg[0] = dg[1] = dg[2] = 0.0; |
| 376 |
< |
setcolor(acol, 0.0, 0.0, 0.0); |
| 377 |
< |
if (ndivs == 0) |
| 376 |
> |
setcolor(rcol, 0.0, 0.0, 0.0); |
| 377 |
> |
if (divcnt == 0) |
| 378 |
|
return(0.0); |
| 379 |
|
/* allocate super-samples */ |
| 380 |
|
if (hemi.ns > 0 || pg != NULL || dg != NULL) { |
| 381 |
< |
div = (AMBSAMP *)malloc(ndivs*sizeof(AMBSAMP)); |
| 381 |
> |
div = (AMBSAMP *)malloc(divcnt*sizeof(AMBSAMP)); |
| 382 |
|
if (div == NULL) |
| 383 |
|
error(SYSTEM, "out of memory in doambient"); |
| 384 |
|
} else |
| 385 |
|
div = NULL; |
| 386 |
|
/* sample the divisions */ |
| 387 |
|
arad = 0.0; |
| 388 |
+ |
acol[0] = acol[1] = acol[2] = 0.0; |
| 389 |
|
if ((dp = div) == NULL) |
| 390 |
|
dp = &dnew; |
| 391 |
+ |
divcnt = 0; |
| 392 |
|
for (i = 0; i < hemi.nt; i++) |
| 393 |
|
for (j = 0; j < hemi.np; j++) { |
| 394 |
|
dp->t = i; dp->p = j; |
| 396 |
|
dp->r = 0.0; |
| 397 |
|
dp->n = 0; |
| 398 |
|
if (divsample(dp, &hemi, r) < 0) { |
| 399 |
< |
if (div == NULL) continue; |
| 400 |
< |
dp++; |
| 197 |
< |
hemi.ns = 0; /* incomplete sampling */ |
| 198 |
< |
pg = dg = NULL; |
| 399 |
> |
if (div != NULL) |
| 400 |
> |
dp++; |
| 401 |
|
continue; |
| 402 |
|
} |
| 403 |
|
arad += dp->r; |
| 404 |
+ |
divcnt++; |
| 405 |
|
if (div != NULL) |
| 406 |
|
dp++; |
| 407 |
|
else |
| 408 |
|
addcolor(acol, dp->v); |
| 409 |
|
} |
| 410 |
< |
if (hemi.ns > 0 && arad > FTINY && ndivs/arad < minarad) |
| 410 |
> |
if (!divcnt) { |
| 411 |
> |
if (div != NULL) |
| 412 |
> |
free((void *)div); |
| 413 |
> |
return(0.0); /* no samples taken */ |
| 414 |
> |
} |
| 415 |
> |
if (divcnt < hemi.nt*hemi.np) { |
| 416 |
> |
pg = dg = NULL; /* incomplete sampling */ |
| 417 |
> |
hemi.ns = 0; |
| 418 |
> |
} else if (arad > FTINY && divcnt/arad < minarad) { |
| 419 |
|
hemi.ns = 0; /* close enough */ |
| 420 |
< |
else if (hemi.ns > 0) { /* else perform super-sampling */ |
| 420 |
> |
} else if (hemi.ns > 0) { /* else perform super-sampling? */ |
| 421 |
|
comperrs(div, &hemi); /* compute errors */ |
| 422 |
< |
qsort(div, ndivs, sizeof(AMBSAMP), ambcmp); /* sort divs */ |
| 422 |
> |
qsort(div, divcnt, sizeof(AMBSAMP), ambcmp); /* sort divs */ |
| 423 |
|
/* super-sample */ |
| 424 |
|
for (i = hemi.ns; i > 0; i--) { |
| 425 |
|
dnew = *div; |
| 428 |
|
continue; |
| 429 |
|
} |
| 430 |
|
dp = div; /* reinsert */ |
| 431 |
< |
j = ndivs < i ? ndivs : i; |
| 431 |
> |
j = divcnt < i ? divcnt : i; |
| 432 |
|
while (--j > 0 && dnew.k < dp[1].k) { |
| 433 |
|
*dp = *(dp+1); |
| 434 |
|
dp++; |
| 436 |
|
*dp = dnew; |
| 437 |
|
} |
| 438 |
|
if (pg != NULL || dg != NULL) /* restore order */ |
| 439 |
< |
qsort(div, ndivs, sizeof(AMBSAMP), ambnorm); |
| 439 |
> |
qsort(div, divcnt, sizeof(AMBSAMP), ambnorm); |
| 440 |
|
} |
| 441 |
|
/* compute returned values */ |
| 442 |
|
if (div != NULL) { |
| 443 |
< |
arad = 0.0; |
| 444 |
< |
for (i = ndivs, dp = div; i-- > 0; dp++) { |
| 443 |
> |
arad = 0.0; /* note: divcnt may be < nt*np */ |
| 444 |
> |
for (i = hemi.nt*hemi.np, dp = div; i-- > 0; dp++) { |
| 445 |
|
arad += dp->r; |
| 446 |
|
if (dp->n > 1) { |
| 447 |
|
b = 1.0/dp->n; |
| 467 |
|
} |
| 468 |
|
free((void *)div); |
| 469 |
|
} |
| 470 |
+ |
copycolor(rcol, acol); |
| 471 |
|
if (arad <= FTINY) |
| 472 |
|
arad = maxarad; |
| 473 |
|
else |
| 474 |
< |
arad = (ndivs+hemi.ns)/arad; |
| 474 |
> |
arad = (divcnt+hemi.ns)/arad; |
| 475 |
|
if (pg != NULL) { /* reduce radius if gradient large */ |
| 476 |
|
d = DOT(pg,pg); |
| 477 |
|
if (d*arad*arad > 1.0) |
| 494 |
|
void |
| 495 |
|
comperrs( /* compute initial error estimates */ |
| 496 |
|
AMBSAMP *da, /* assumes standard ordering */ |
| 497 |
< |
register AMBHEMI *hp |
| 497 |
> |
AMBHEMI *hp |
| 498 |
|
) |
| 499 |
|
{ |
| 500 |
|
double b, b2; |
| 501 |
|
int i, j; |
| 502 |
< |
register AMBSAMP *dp; |
| 502 |
> |
AMBSAMP *dp; |
| 503 |
|
/* sum differences from neighbors */ |
| 504 |
|
dp = da; |
| 505 |
|
for (i = 0; i < hp->nt; i++) |
| 547 |
|
posgradient( /* compute position gradient */ |
| 548 |
|
FVECT gv, |
| 549 |
|
AMBSAMP *da, /* assumes standard ordering */ |
| 550 |
< |
register AMBHEMI *hp |
| 550 |
> |
AMBHEMI *hp |
| 551 |
|
) |
| 552 |
|
{ |
| 553 |
< |
register int i, j; |
| 553 |
> |
int i, j; |
| 554 |
|
double nextsine, lastsine, b, d; |
| 555 |
|
double mag0, mag1; |
| 556 |
|
double phi, cosp, sinp, xd, yd; |
| 557 |
< |
register AMBSAMP *dp; |
| 557 |
> |
AMBSAMP *dp; |
| 558 |
|
|
| 559 |
|
xd = yd = 0.0; |
| 560 |
|
for (j = 0; j < hp->np; j++) { |
| 605 |
|
dirgradient( /* compute direction gradient */ |
| 606 |
|
FVECT gv, |
| 607 |
|
AMBSAMP *da, /* assumes standard ordering */ |
| 608 |
< |
register AMBHEMI *hp |
| 608 |
> |
AMBHEMI *hp |
| 609 |
|
) |
| 610 |
|
{ |
| 611 |
< |
register int i, j; |
| 611 |
> |
int i, j; |
| 612 |
|
double mag; |
| 613 |
|
double phi, xd, yd; |
| 614 |
< |
register AMBSAMP *dp; |
| 614 |
> |
AMBSAMP *dp; |
| 615 |
|
|
| 616 |
|
xd = yd = 0.0; |
| 617 |
|
for (j = 0; j < hp->np; j++) { |
| 634 |
|
for (i = 0; i < 3; i++) |
| 635 |
|
gv[i] = xd*hp->ux[i] + yd*hp->uy[i]; |
| 636 |
|
} |
| 637 |
+ |
|
| 638 |
+ |
#endif /* ! NEWAMB */ |