MOAB: Mesh Oriented datABase
(version 5.4.1)
|
00001 #include <stdio.h> 00002 #include <stdlib.h> 00003 #include <stdarg.h> 00004 #include <math.h> /* for cos, fabs */ 00005 #include <float.h> 00006 #include <string.h> /* for memcpy */ 00007 00008 #include "moab/FindPtFuncs.h" 00009 00010 const unsigned opt_no_constraints_2 = 3 + 1; 00011 const unsigned opt_no_constraints_3 = 9 + 3 + 1; 00012 00013 /*-------------------------------------------------------------------------- 00014 Lobatto Polynomial Bounds 00015 00016 Needed inputs are the Gauss-Lobatto quadrature nodes and weights: 00017 unsigned nr = ..., ns = ...; 00018 realType zr[nr], wr[nr]; 00019 realType zs[ns], ws[ns]; 00020 00021 lobatto_nodes(zr,nr); lobatto_weights(zr,wr,nr); 00022 lobatto_nodes(zs,ns); lobatto_weights(zs,ws,ns); 00023 00024 The number of points in the constructed piecewise (bi-)linear bounds 00025 is a parameter; more points give tighter bounds 00026 00027 unsigned mr = 2*nr, ms = 2*ns; 00028 00029 The necessary setup is accomplished via: 00030 lob_bnd_base b_data_r; 00031 lob_bnd_ext e_data_s; 00032 00033 lob_bnd_base_alloc(&b_data_r,nr,mr); 00034 lob_bnd_base_setup(&b_data_r,zr,wr); 00035 lob_bnd_ext_alloc(&e_data_s,ns,ms); 00036 lob_bnd_ext_setup(&e_data_s,zs,ws); 00037 00038 Bounds may then be computed via: 00039 realType work1r[2*mr], work1s[2*ms], work2[2*mr + 2*mr*ns + 2*mr*ms]; 00040 realType ur[nr], us[ns]; // 1-d polynomials on the zr[] and zs[] nodes 00041 realType u[ns][nr]; // 2-d polynomial on zr[] (x) zs[] 00042 realType bound[2]; // = { min, max } (to be computed) 00043 00044 lob_bnd_1(&b_data_r ,ur,bound,work1r); // compute bounds on ur 00045 lob_bnd_1(&e_data_s.b,us,bound,work1s); // compute bounds on us 00046 lob_bnd_2(&b_data_r, &e_data_s, 00047 (const double*)&u[0][0],bound,work2); // compute bounds on u 00048 The above routines access the zr,zs arrays passed to *_setup 00049 (so do not delete them between calls) 00050 00051 Memory allocated in *_setup is freed with 00052 lob_bnd_base_free(&b_data_r); 00053 lob_bnd_ext_free(&e_data_s); 00054 00055 --------------------------------------------------------------------------*/ 00056 00057 typedef struct 00058 { 00059 unsigned n; /* number of Lobatto nodes in input */ 00060 unsigned m; /* number of Chebyshev nodes used to calculate bounds */ 00061 realType *Q0, *Q1; /* Q0[n], Q1[n] -- first two rows of change of basis matrix 00062 from Lobatto node Lagrangian to Legendre */ 00063 const realType* z; /* z[n] -- external; Lobatto nodes */ 00064 realType* h; /* h[m] -- Chebyshev nodes */ 00065 realType *uv, *ov; /* uv[n][m], ov[n][m] -- 00066 uv[j][:] is a piecewise linear function in the nodal 00067 basis with nodes h[m] that is everywhere less 00068 than or equal to the jth Lagrangian basis 00069 function (with the Lobatto nodes z[n]) 00070 ov[j][:] is everywhere greater than or equal */ 00071 } lob_bnd_base; 00072 00073 typedef struct 00074 { 00075 lob_bnd_base b; 00076 realType *uvp, *uvn, *ovp, *ovn; /* [n][m] -- uv and ov split into 00077 positive and negative parts */ 00078 } lob_bnd_ext; 00079 00080 static void lob_bnd_base_alloc( lob_bnd_base* p, unsigned n, unsigned m ) 00081 { 00082 p->n = n, p->m = m; 00083 p->Q0 = tmalloc( realType, 2 * n + m + 2 * n * m ); 00084 p->Q1 = p->Q0 + n; 00085 p->h = p->Q1 + n; 00086 p->uv = p->h + m; 00087 p->ov = p->uv + n * m; 00088 } 00089 00090 static void lob_bnd_base_free( lob_bnd_base* p ) 00091 { 00092 free( p->Q0 ); 00093 } 00094 00095 static void lob_bnd_ext_alloc( lob_bnd_ext* p, unsigned n, unsigned m ) 00096 { 00097 p->b.n = n, p->b.m = m; 00098 p->b.Q0 = tmalloc( realType, 2 * n + m + 6 * n * m ); 00099 p->b.Q1 = p->b.Q0 + n; 00100 p->b.h = p->b.Q1 + n; 00101 p->b.uv = p->b.h + m; 00102 p->b.ov = p->b.uv + n * m; 00103 p->uvp = p->b.ov + n * m; 00104 p->uvn = p->uvp + n * m; 00105 p->ovp = p->uvn + n * m; 00106 p->ovn = p->ovp + n * m; 00107 } 00108 00109 static void lob_bnd_ext_free( lob_bnd_ext* p ) 00110 { 00111 free( p->b.Q0 ); 00112 } 00113 00114 static void lob_bnd_base_setup( lob_bnd_base* p, const realType* z, const realType* w ) 00115 { 00116 unsigned i, j, m = p->m, n = p->n, mm = 2 * m - 1; 00117 realType *q = tmalloc( realType, ( 2 * n + 1 ) * mm + 6 * n ), *J = q + mm, *D = J + n * mm, *work = D + n * mm; 00118 p->z = z; 00119 for( i = 0; i < n; ++i ) 00120 p->Q0[i] = w[i] / 2, p->Q1[i] = 3 * p->Q0[i] * z[i]; 00121 p->h[0] = -1, p->h[m - 1] = 1; 00122 for( j = 1; j < m - 1; ++j ) 00123 p->h[j] = mbcos( ( m - j - 1 ) * MOAB_POLY_PI / ( m - 1 ) ); 00124 for( j = 0; j < m - 1; ++j ) 00125 q[2 * j] = p->h[j], q[2 * j + 1] = ( p->h[j] + p->h[j + 1] ) / 2; 00126 q[mm - 1] = p->h[m - 1]; 00127 lagrange_weights_deriv( z, n, q, mm, J, D, work ); 00128 for( i = 0; i < n; ++i ) 00129 { 00130 realType *uv = p->uv + i * m, *ov = p->ov + i * m; 00131 ov[0] = uv[0] = J[i]; 00132 ov[m - 1] = uv[m - 1] = J[( mm - 1 ) * n + i]; 00133 for( j = 1; j < m - 1; ++j ) 00134 { 00135 unsigned jj = 2 * j; 00136 realType c0 = J[( jj - 1 ) * n + i] + ( q[jj] - q[jj - 1] ) * D[( jj - 1 ) * n + i]; 00137 realType c1 = J[( jj + 1 ) * n + i] + ( q[jj] - q[jj + 1] ) * D[( jj + 1 ) * n + i]; 00138 realType c2 = J[jj * n + i]; 00139 rminmax_3( &uv[j], &ov[j], c0, c1, c2 ); 00140 } 00141 } 00142 free( q ); 00143 } 00144 00145 static void lob_bnd_ext_setup( lob_bnd_ext* p, const realType* z, const realType* w ) 00146 { 00147 unsigned i, mn = p->b.m * p->b.n; 00148 lob_bnd_base_setup( &p->b, z, w ); 00149 for( i = 0; i < mn; ++i ) 00150 { 00151 realType uvi = p->b.uv[i], ovi = p->b.ov[i]; 00152 p->uvp[i] = p->uvn[i] = p->ovp[i] = p->ovn[i] = 0; 00153 if( uvi > 0 ) 00154 p->uvp[i] = uvi; 00155 else 00156 p->uvn[i] = uvi; 00157 if( ovi > 0 ) 00158 p->ovp[i] = ovi; 00159 else 00160 p->ovn[i] = ovi; 00161 } 00162 } 00163 00164 static void lob_bnd_lines( const lob_bnd_base* p, const realType* u, realType* a, realType* b ) 00165 { 00166 unsigned i, j; 00167 realType a0 = 0, a1 = 0; 00168 const realType *uv = p->uv, *ov = p->ov; 00169 for( i = 0; i < p->n; ++i ) 00170 a0 += p->Q0[i] * u[i], a1 += p->Q1[i] * u[i]; 00171 for( j = 0; j < p->m; ++j ) 00172 b[j] = a[j] = a0 + a1 * p->h[j]; 00173 for( i = 0; i < p->n; ++i ) 00174 { 00175 realType w = u[i] - ( a0 + a1 * p->z[i] ); 00176 if( w >= 0 ) 00177 for( j = 0; j < p->m; ++j ) 00178 a[j] += w * ( *uv++ ), b[j] += w * ( *ov++ ); 00179 else 00180 for( j = 0; j < p->m; ++j ) 00181 a[j] += w * ( *ov++ ), b[j] += w * ( *uv++ ); 00182 } 00183 } 00184 00185 /* work holds p->m * 2 doubles */ 00186 static void lob_bnd_1( const lob_bnd_base* p, const realType* u, realType bnd[2], realType* work ) 00187 { 00188 unsigned j; 00189 realType *a = work, *b = work + p->m; 00190 lob_bnd_lines( p, u, a, b ); 00191 bnd[0] = a[0], bnd[1] = b[0]; 00192 for( j = 1; j < p->m; ++j ) 00193 { 00194 if( a[j] < bnd[0] ) bnd[0] = a[j]; 00195 if( b[j] > bnd[1] ) bnd[1] = b[j]; 00196 } 00197 } 00198 00199 /* work holds 2*mr + 2*mr*ns + 2*mr*ms doubles */ 00200 static void lob_bnd_2( const lob_bnd_base* pr, 00201 const lob_bnd_ext* ps, 00202 const realType* u, 00203 realType bnd[2], 00204 realType* work ) 00205 { 00206 unsigned nr = pr->n, mr = pr->m, ns = ps->b.n, ms = ps->b.m; 00207 realType *a0 = work, *a1 = a0 + mr, *ar_ = a1 + mr, *ar = ar_, *br_ = ar + mr * ns, *br = br_, *a_ = br + mr * ns, 00208 *a = a_, *b_ = a + mr * ms, *b = b_, *uvp, *ovp, *uvn, *ovn; 00209 realType b0, b1; 00210 unsigned i, j, k; 00211 for( i = 0; i < mr; ++i ) 00212 a0[i] = a1[i] = 0; 00213 for( j = 0; j < ns; ++j, u += nr ) 00214 { 00215 realType q0 = ps->b.Q0[j], q1 = ps->b.Q1[j]; 00216 lob_bnd_lines( pr, u, ar, br ); 00217 for( i = 0; i < mr; ++i ) 00218 { 00219 realType t = ( *ar++ + *br++ ) / 2; 00220 a0[i] += q0 * t, a1[i] += q1 * t; 00221 } 00222 } 00223 for( i = 0; i < mr; ++i ) 00224 { 00225 realType a0i = a0[i], a1i = a1[i]; 00226 for( k = 0; k < ms; ++k ) 00227 *b++ = *a++ = a0i + a1i * ps->b.h[k]; 00228 } 00229 ar = ar_, br = br_; 00230 uvp = ps->uvp, ovp = ps->ovp, uvn = ps->uvn, ovn = ps->ovn; 00231 for( j = 0; j < ns; ++j, uvp += ms, uvn += ms, ovp += ms, ovn += ms ) 00232 { 00233 realType zj = ps->b.z[j]; 00234 a = a_, b = b_; 00235 for( i = 0; i < mr; ++i ) 00236 { 00237 realType t = a0[i] + a1[i] * zj; 00238 realType uw = *ar++ - t; 00239 realType ow = *br++ - t; 00240 if( uw >= 0 ) /* 0 <= uw <= ow */ 00241 for( k = 0; k < ms; ++k ) 00242 *a++ += uw * uvp[k] + ow * uvn[k], *b++ += ow * ovp[k] + uw * ovn[k]; 00243 else if( ow <= 0 ) /* uw <= ow <= 0 */ 00244 for( k = 0; k < ms; ++k ) 00245 *a++ += uw * ovp[k] + ow * ovn[k], *b++ += ow * uvp[k] + uw * uvn[k]; 00246 else /* uw < 0 < ow */ 00247 for( k = 0; k < ms; ++k ) 00248 *a++ += uw * ovp[k] + ow * uvn[k], *b++ += ow * ovp[k] + uw * uvn[k]; 00249 } 00250 } 00251 b0 = a_[0], b1 = b_[0]; 00252 for( i = 1; i < mr * ms; ++i ) 00253 { 00254 if( a_[i] < b0 ) b0 = a_[i]; 00255 if( b_[i] > b1 ) b1 = b_[i]; 00256 } 00257 bnd[0] = b0, bnd[1] = b1; 00258 } 00259 00260 /*-------------------------------------------------------------------------- 00261 Small Matrix Inverse 00262 --------------------------------------------------------------------------*/ 00263 00264 static void mat_inv_2( const realType A[4], realType inv[4] ) 00265 { 00266 const realType idet = 1 / ( A[0] * A[3] - A[1] * A[2] ); 00267 inv[0] = idet * A[3]; 00268 inv[1] = -( idet * A[1] ); 00269 inv[2] = -( idet * A[2] ); 00270 inv[3] = idet * A[0]; 00271 } 00272 00273 static void mat_inv_3( const realType A[9], realType inv[9] ) 00274 { 00275 const realType a = A[4] * A[8] - A[5] * A[7], b = A[5] * A[6] - A[3] * A[8], c = A[3] * A[7] - A[4] * A[6], 00276 idet = 1 / ( A[0] * a + A[1] * b + A[2] * c ); 00277 inv[0] = idet * a; 00278 inv[1] = idet * ( A[2] * A[7] - A[1] * A[8] ); 00279 inv[2] = idet * ( A[1] * A[5] - A[2] * A[4] ); 00280 inv[3] = idet * b; 00281 inv[4] = idet * ( A[0] * A[8] - A[2] * A[6] ); 00282 inv[5] = idet * ( A[2] * A[3] - A[0] * A[5] ); 00283 inv[6] = idet * c; 00284 inv[7] = idet * ( A[1] * A[6] - A[0] * A[7] ); 00285 inv[8] = idet * ( A[0] * A[4] - A[1] * A[3] ); 00286 } 00287 00288 static void mat_app_2r( realType y[2], const realType A[4], const realType x[2] ) 00289 { 00290 y[0] = A[0] * x[0] + A[1] * x[1]; 00291 y[1] = A[2] * x[0] + A[3] * x[1]; 00292 } 00293 00294 static void mat_app_2c( realType y[2], const realType A[4], const realType x[2] ) 00295 { 00296 y[0] = A[0] * x[0] + A[2] * x[1]; 00297 y[1] = A[1] * x[0] + A[3] * x[1]; 00298 } 00299 00300 static void mat_app_3r( realType y[3], const realType A[9], const realType x[3] ) 00301 { 00302 y[0] = A[0] * x[0] + A[1] * x[1] + A[2] * x[2]; 00303 y[1] = A[3] * x[0] + A[4] * x[1] + A[5] * x[2]; 00304 y[2] = A[6] * x[0] + A[7] * x[1] + A[8] * x[2]; 00305 } 00306 00307 static void mat_app_3c( realType y[3], const realType A[9], const realType x[3] ) 00308 { 00309 y[0] = A[0] * x[0] + A[3] * x[1] + A[6] * x[2]; 00310 y[1] = A[1] * x[0] + A[4] * x[1] + A[7] * x[2]; 00311 y[2] = A[2] * x[0] + A[5] * x[1] + A[8] * x[2]; 00312 } 00313 00314 static void tinyla_solve_2( realType x[2], const realType A[4], const realType b[2] ) 00315 { 00316 realType inv[4]; 00317 mat_inv_2( A, inv ); 00318 mat_app_2r( x, inv, b ); 00319 } 00320 00321 static void tinyla_solve_3( realType x[3], const realType A[9], const realType b[3] ) 00322 { 00323 realType inv[9]; 00324 mat_inv_3( A, inv ); 00325 mat_app_3r( x, inv, b ); 00326 } 00327 00328 /* solve 00329 A[0] x0 + A[2] x1 = b0, 00330 A[2] x0 + A[1] x1 = b1 00331 */ 00332 static void tinyla_solve_sym_2( realType* x0, realType* x1, const realType A[3], realType b0, realType b1 ) 00333 { 00334 const realType idet = 1 / ( A[0] * A[1] - A[2] * A[2] ); 00335 *x0 = idet * ( A[1] * b0 - A[2] * b1 ); 00336 *x1 = idet * ( A[0] * b1 - A[2] * b0 ); 00337 } 00338 00339 /*-------------------------------------------------------------------------- 00340 Oriented Bounding Box 00341 00342 Suffixes on names are _2 for 2-d and _3 for 3-d 00343 00344 Needed inputs are the Gauss-Lobatto quadrature nodes and weights: 00345 unsigned nr = ..., ns = ...; 00346 realType zr[nr], wr[nr]; 00347 realType zs[ns], ws[ns]; 00348 00349 lobatto_nodes(zr,nr); lobatto_weights(zr,wr,nr); 00350 lobatto_nodes(zs,ns); lobatto_weights(zs,ws,ns); 00351 00352 The number of points in the constructed piecewise (bi-)linear bounds 00353 for the boundaries is a parameter; more points give tighter bounds 00354 00355 unsigned mr = 2*nr, ms = 2*ns; 00356 00357 Bounding boxes are increased by a relative amount as a parameter 00358 00359 realType tol = 0.01; 00360 00361 Setup is accomplished via: 00362 00363 const realType *z[2] = {zr,zs}, *w[2] = {wr,ws}; 00364 const unsigned n[2] = {nr,ns}, m[2] = {mr,ms}; 00365 obbox_data_2 *data = obbox_setup_2(z,w,n,m); 00366 00367 Bounding box data may then be computed: 00368 00369 obbox_2 box; // will store bounding box information 00370 realType xm[ns][nr], ym[ns][nr]; // x, y coordinates of the element nodes 00371 00372 obbox_calc_2(data, tol, (const realType *)&xm[0][0], 00373 (const realType *)&ym[0][0], &box); 00374 00375 A point may be tested: 00376 00377 const realType x[2]; // point to test 00378 realType r[2]; 00379 00380 if( obbox_axis_test_2(&box, x) ) 00381 ... // x failed axis-aligned bounding box test 00382 00383 if( obbox_test_2(&box, x, r) ) 00384 ... // x failed oriented bounding box test 00385 else 00386 ... // r suitable as initial guess for parametric coords 00387 00388 Once all bounding box information has been computed 00389 00390 obbox_free_2(data); 00391 00392 to free the memory allocated with obbox_setup_2. 00393 00394 --------------------------------------------------------------------------*/ 00395 00396 typedef struct 00397 { 00398 lob_bnd_base dr, ds; 00399 realType *Jr0, *Dr0, *Js0, *Ds0, *work; 00400 } obbox_data_2; 00401 00402 typedef struct 00403 { 00404 lob_bnd_base dr; 00405 lob_bnd_ext ds, dt; 00406 realType *Jr0, *Dr0, *Js0, *Ds0, *Jt0, *Dt0, *work; 00407 } obbox_data_3; 00408 00409 static void obbox_data_alloc_2( obbox_data_2* p, const unsigned n[2], const unsigned m[2] ) 00410 { 00411 const unsigned max_npm = umax_2( n[0] + m[0], n[1] + m[1] ); 00412 lob_bnd_base_alloc( &p->dr, n[0], m[0] ); 00413 lob_bnd_base_alloc( &p->ds, n[1], m[1] ); 00414 p->Jr0 = tmalloc( realType, 2 * n[0] + 2 * n[1] + 2 * max_npm ); 00415 p->Dr0 = p->Jr0 + n[0]; 00416 p->Js0 = p->Dr0 + n[0]; 00417 p->Ds0 = p->Js0 + n[1]; 00418 p->work = p->Ds0 + n[1]; 00419 } 00420 00421 static void obbox_data_free_2( obbox_data_2* p ) 00422 { 00423 lob_bnd_base_free( &p->dr ); 00424 lob_bnd_base_free( &p->ds ); 00425 free( p->Jr0 ); 00426 } 00427 00428 static void obbox_data_alloc_3( obbox_data_3* p, const unsigned n[3], const unsigned m[3] ) 00429 { 00430 const unsigned wk1 = 3 * n[0] * n[1] + 2 * m[0] + 2 * m[0] * n[1] + 2 * m[0] * m[1]; 00431 const unsigned wk2 = 3 * n[0] * n[2] + 2 * m[0] + 2 * m[0] * n[2] + 2 * m[0] * m[2]; 00432 const unsigned wk3 = 3 * n[1] * n[2] + 2 * m[1] + 2 * m[1] * n[2] + 2 * m[1] * m[2]; 00433 const unsigned wk_max = umax_3( wk1, wk2, wk3 ); 00434 lob_bnd_base_alloc( &p->dr, n[0], m[0] ); 00435 lob_bnd_ext_alloc( &p->ds, n[1], m[1] ); 00436 lob_bnd_ext_alloc( &p->dt, n[2], m[2] ); 00437 p->Jr0 = tmalloc( realType, 2 * n[0] + 2 * n[1] + 2 * n[2] + wk_max ); 00438 p->Dr0 = p->Jr0 + n[0]; 00439 p->Js0 = p->Dr0 + n[0]; 00440 p->Ds0 = p->Js0 + n[1]; 00441 p->Jt0 = p->Ds0 + n[1]; 00442 p->Dt0 = p->Jt0 + n[2]; 00443 p->work = p->Dt0 + n[2]; 00444 } 00445 00446 static void obbox_data_free_3( obbox_data_3* p ) 00447 { 00448 lob_bnd_base_free( &p->dr ); 00449 lob_bnd_ext_free( &p->ds ); 00450 lob_bnd_ext_free( &p->dt ); 00451 free( p->Jr0 ); 00452 } 00453 00454 static obbox_data_2* obbox_setup_2( const realType* const z[2], 00455 const realType* const w[2], 00456 const unsigned n[2], 00457 const unsigned m[2] ) 00458 { 00459 const realType zero = 0; 00460 realType* work; 00461 obbox_data_2* p = tmalloc( obbox_data_2, 1 ); 00462 obbox_data_alloc_2( p, n, m ); 00463 lob_bnd_base_setup( &p->dr, z[0], w[0] ); 00464 lob_bnd_base_setup( &p->ds, z[1], w[1] ); 00465 work = tmalloc( realType, 6 * umax_2( n[0], n[1] ) ); 00466 lagrange_weights_deriv( z[0], n[0], &zero, 1, p->Jr0, p->Dr0, work ); 00467 lagrange_weights_deriv( z[1], n[1], &zero, 1, p->Js0, p->Ds0, work ); 00468 free( work ); 00469 return p; 00470 } 00471 00472 static obbox_data_3* obbox_setup_3( const realType* const z[3], 00473 const realType* const w[3], 00474 const unsigned n[3], 00475 const unsigned m[3] ) 00476 { 00477 const realType zero = 0; 00478 realType* work; 00479 obbox_data_3* p = tmalloc( obbox_data_3, 1 ); 00480 obbox_data_alloc_3( p, n, m ); 00481 lob_bnd_base_setup( &p->dr, z[0], w[0] ); 00482 lob_bnd_ext_setup( &p->ds, z[1], w[1] ); 00483 lob_bnd_ext_setup( &p->dt, z[2], w[2] ); 00484 work = tmalloc( realType, 6 * umax_3( n[0], n[1], n[2] ) ); 00485 lagrange_weights_deriv( z[0], n[0], &zero, 1, p->Jr0, p->Dr0, work ); 00486 lagrange_weights_deriv( z[1], n[1], &zero, 1, p->Js0, p->Ds0, work ); 00487 lagrange_weights_deriv( z[2], n[2], &zero, 1, p->Jt0, p->Dt0, work ); 00488 free( work ); 00489 return p; 00490 } 00491 00492 static void obbox_free_2( obbox_data_2* p ) 00493 { 00494 obbox_data_free_2( p ); 00495 free( p ); 00496 } 00497 00498 static void obbox_free_3( obbox_data_3* p ) 00499 { 00500 obbox_data_free_3( p ); 00501 free( p ); 00502 } 00503 00504 int obbox_axis_test_2( const obbox_2* p, const realType x[2] ) 00505 { 00506 return ( x[0] < p->axis_bnd[0] || x[0] > p->axis_bnd[1] || x[1] < p->axis_bnd[2] || x[1] > p->axis_bnd[3] ); 00507 } 00508 00509 int obbox_axis_test_3( const obbox_3* p, const realType x[3] ) 00510 { 00511 return ( x[0] < p->axis_bnd[0] || x[0] > p->axis_bnd[1] || x[1] < p->axis_bnd[2] || x[1] > p->axis_bnd[3] || 00512 x[2] < p->axis_bnd[4] || x[2] > p->axis_bnd[5] ); 00513 } 00514 00515 int obbox_test_2( const obbox_2* p, const realType x[2], realType r[2] ) 00516 { 00517 realType xt[2]; 00518 xt[0] = x[0] - p->x[0]; 00519 xt[1] = x[1] - p->x[1]; 00520 r[0] = p->A[0] * xt[0] + p->A[1] * xt[1]; 00521 if( mbabs( r[0] ) > 1 ) return 1; 00522 r[1] = p->A[2] * xt[0] + p->A[3] * xt[1]; 00523 return mbabs( r[1] ) > 1; 00524 } 00525 00526 int obbox_test_3( const obbox_3* p, const realType x[3], realType r[3] ) 00527 { 00528 realType xt[3]; 00529 xt[0] = x[0] - p->x[0]; 00530 xt[1] = x[1] - p->x[1]; 00531 xt[2] = x[2] - p->x[2]; 00532 r[0] = p->A[0] * xt[0] + p->A[1] * xt[1] + p->A[2] * xt[2]; 00533 if( mbabs( r[0] ) > 1 ) return 1; 00534 r[1] = p->A[3] * xt[0] + p->A[4] * xt[1] + p->A[5] * xt[2]; 00535 if( mbabs( r[1] ) > 1 ) return 1; 00536 r[2] = p->A[6] * xt[0] + p->A[7] * xt[1] + p->A[8] * xt[2]; 00537 return mbabs( r[2] ) > 1; 00538 } 00539 00540 void obbox_calc_tfm_2( const realType* x, 00541 const realType* y, 00542 unsigned n, 00543 unsigned s, 00544 const realType c0[2], 00545 const realType A[4], 00546 realType* u ) 00547 { 00548 unsigned i; 00549 realType* v = u + n; 00550 for( i = 0; i < n; ++i, x += s, y += s ) 00551 { 00552 const realType xt = *x - c0[0], yt = *y - c0[1]; 00553 u[i] = A[0] * xt + A[1] * yt; 00554 v[i] = A[2] * xt + A[3] * yt; 00555 } 00556 } 00557 00558 void obbox_calc_tfm_3( const realType* x, 00559 const realType* y, 00560 const realType* z, 00561 unsigned nr, 00562 unsigned sr, 00563 unsigned ns, 00564 unsigned ss, 00565 const realType c0[3], 00566 const realType A[9], 00567 realType* u ) 00568 { 00569 unsigned i, j; 00570 realType *v = u + nr * ns, *w = v + nr * ns; 00571 for( j = 0; j < ns; ++j, x += ss, y += ss, z += ss ) 00572 { 00573 for( i = 0; i < nr; ++i, x += sr, y += sr, z += sr ) 00574 { 00575 const realType xt = *x - c0[0], yt = *y - c0[1], zt = *z - c0[2]; 00576 *u++ = A[0] * xt + A[1] * yt + A[2] * zt; 00577 *v++ = A[3] * xt + A[4] * yt + A[5] * zt; 00578 *w++ = A[6] * xt + A[7] * yt + A[8] * zt; 00579 } 00580 } 00581 } 00582 00583 void obbox_merge_2( realType* b, const realType* ob ) 00584 { 00585 if( ob[0] < b[0] ) b[0] = ob[0]; 00586 if( ob[1] > b[1] ) b[1] = ob[1]; 00587 if( ob[2] < b[2] ) b[2] = ob[2]; 00588 if( ob[3] > b[3] ) b[3] = ob[3]; 00589 } 00590 00591 void obbox_merge_3( realType* b, const realType* ob ) 00592 { 00593 if( ob[0] < b[0] ) b[0] = ob[0]; 00594 if( ob[1] > b[1] ) b[1] = ob[1]; 00595 if( ob[2] < b[2] ) b[2] = ob[2]; 00596 if( ob[3] > b[3] ) b[3] = ob[3]; 00597 if( ob[4] < b[4] ) b[4] = ob[4]; 00598 if( ob[5] > b[5] ) b[5] = ob[5]; 00599 } 00600 00601 /* work holds 2*n + 2*m realTypes */ 00602 void obbox_side_2( const realType* x, 00603 const realType* y, 00604 unsigned n, 00605 unsigned s, 00606 const realType c0[2], 00607 const realType A[4], 00608 realType* work, 00609 const lob_bnd_base* lbd, 00610 realType bnd[4] ) 00611 { 00612 obbox_calc_tfm_2( x, y, n, s, c0, A, work ); 00613 lob_bnd_1( lbd, work, bnd, work + 2 * n ); 00614 lob_bnd_1( lbd, work + n, bnd + 2, work + 2 * n ); 00615 } 00616 00617 /* work holds 3*nr*ns + 2*mr + 2*mr*ns + 2*mr*ms realTypes */ 00618 void obbox_side_3( const realType* x, 00619 const realType* y, 00620 const realType* z, 00621 unsigned nr, 00622 unsigned sr, 00623 unsigned ns, 00624 unsigned ss, 00625 const realType c0[3], 00626 const realType A[9], 00627 realType* work, 00628 const lob_bnd_base* dr, 00629 const lob_bnd_ext* ds, 00630 realType bnd[6] ) 00631 { 00632 obbox_calc_tfm_3( x, y, z, nr, sr, ns, ss, c0, A, work ); 00633 lob_bnd_2( dr, ds, work, bnd, work + 3 * nr * ns ); 00634 lob_bnd_2( dr, ds, work + nr * ns, bnd + 2, work + 3 * nr * ns ); 00635 lob_bnd_2( dr, ds, work + 2 * nr * ns, bnd + 4, work + 3 * nr * ns ); 00636 } 00637 00638 /* return bounds on u = A (x - c0) 00639 bnd[0] <= u_0 <= bnd[1] 00640 bnd[2] <= u_1 <= bnd[3] */ 00641 void obbox_bnd_2( const obbox_data_2* p, 00642 const realType* x, 00643 const realType* y, 00644 const realType c0[2], 00645 const realType A[4], 00646 realType bnd[4] ) 00647 { 00648 unsigned i, nr = p->dr.n, ns = p->ds.n; 00649 realType obnd[4]; 00650 00651 i = nr * ( ns - 1 ); 00652 obbox_side_2( x, y, nr, 1, c0, A, p->work, &p->dr, bnd ); 00653 obbox_side_2( x + i, y + i, nr, 1, c0, A, p->work, &p->dr, obnd ); 00654 obbox_merge_2( bnd, obnd ); 00655 00656 i = nr - 1; 00657 obbox_side_2( x, y, ns, nr, c0, A, p->work, &p->ds, obnd ); 00658 obbox_merge_2( bnd, obnd ); 00659 obbox_side_2( x + i, y + i, nr, nr, c0, A, p->work, &p->ds, obnd ); 00660 obbox_merge_2( bnd, obnd ); 00661 } 00662 00663 /* return bounds on u = A (x - c0) 00664 bnd[0] <= u_0 <= bnd[1] 00665 bnd[2] <= u_1 <= bnd[3] 00666 bnd[4] <= u_2 <= bnd[5] */ 00667 void obbox_bnd_3( const obbox_data_3* p, 00668 const realType* x, 00669 const realType* y, 00670 const realType* z, 00671 const realType c0[3], 00672 const realType A[9], 00673 realType bnd[6] ) 00674 { 00675 unsigned i, nr = p->dr.n, ns = p->ds.b.n, nt = p->dt.b.n; 00676 realType obnd[6]; 00677 00678 i = nr * ns * ( nt - 1 ); 00679 obbox_side_3( x, y, z, nr, 1, ns, 0, c0, A, p->work, &p->dr, &p->ds, bnd ); 00680 obbox_side_3( x + i, y + i, z + i, nr, 1, ns, 0, c0, A, p->work, &p->dr, &p->ds, obnd ); 00681 obbox_merge_3( bnd, obnd ); 00682 00683 i = nr * ( ns - 1 ); 00684 obbox_side_3( x, y, z, nr, 1, nt, i, c0, A, p->work, &p->dr, &p->dt, obnd ); 00685 obbox_merge_3( bnd, obnd ); 00686 obbox_side_3( x + i, y + i, z + i, nr, 1, nt, i, c0, A, p->work, &p->dr, &p->dt, obnd ); 00687 obbox_merge_3( bnd, obnd ); 00688 00689 i = nr - 1; 00690 obbox_side_3( x, y, z, ns, nr, nt, 0, c0, A, p->work, &p->ds.b, &p->dt, obnd ); 00691 obbox_merge_3( bnd, obnd ); 00692 obbox_side_3( x + i, y + i, z + i, ns, nr, nt, 0, c0, A, p->work, &p->ds.b, &p->dt, obnd ); 00693 obbox_merge_3( bnd, obnd ); 00694 } 00695 00696 void obbox_calc_2( const obbox_data_2* p, realType tol, const realType* x, const realType* y, obbox_2* b ) 00697 { 00698 const realType zero[2] = { 0, 0 }, id[4] = { 1, 0, 0, 1 }; 00699 realType c0[2], jac[4], inv[4], bnd[4], u0[2], d[2]; 00700 00701 obbox_bnd_2( p, x, y, zero, id, b->axis_bnd ); 00702 d[0] = b->axis_bnd[1] - b->axis_bnd[0]; 00703 d[1] = b->axis_bnd[3] - b->axis_bnd[2]; 00704 b->axis_bnd[0] -= tol * d[0], b->axis_bnd[1] += tol * d[0]; 00705 b->axis_bnd[2] -= tol * d[1], b->axis_bnd[3] += tol * d[1]; 00706 00707 c0[0] = tensor_ig2( p->Jr0, p->Dr0, p->dr.n, p->Js0, p->Ds0, p->ds.n, x, jac, p->work ); 00708 c0[1] = tensor_ig2( p->Jr0, p->Dr0, p->dr.n, p->Js0, p->Ds0, p->ds.n, y, jac + 2, p->work ); 00709 mat_inv_2( jac, inv ); 00710 00711 obbox_bnd_2( p, x, y, c0, inv, bnd ); 00712 00713 u0[0] = ( bnd[0] + bnd[1] ) / 2; 00714 u0[1] = ( bnd[2] + bnd[3] ) / 2; 00715 d[0] = 2 / ( ( 1 + tol ) * ( bnd[1] - bnd[0] ) ); 00716 d[1] = 2 / ( ( 1 + tol ) * ( bnd[3] - bnd[2] ) ); 00717 b->x[0] = c0[0] + jac[0] * u0[0] + jac[1] * u0[1]; 00718 b->x[1] = c0[1] + jac[2] * u0[0] + jac[3] * u0[1]; 00719 b->A[0] = d[0] * inv[0], b->A[1] = d[0] * inv[1]; 00720 b->A[2] = d[1] * inv[2], b->A[3] = d[1] * inv[3]; 00721 } 00722 00723 void obbox_calc_3( const obbox_data_3* p, 00724 realType tol, 00725 const realType* x, 00726 const realType* y, 00727 const realType* z, 00728 obbox_3* b ) 00729 { 00730 const realType zero[3] = { 0, 0 }, id[9] = { 1, 0, 0, 0, 1, 0, 0, 0, 1 }; 00731 realType c0[3], jac[9], inv[9], bnd[6], u0[3], d[3]; 00732 00733 obbox_bnd_3( p, x, y, z, zero, id, b->axis_bnd ); 00734 d[0] = b->axis_bnd[1] - b->axis_bnd[0]; 00735 d[1] = b->axis_bnd[3] - b->axis_bnd[2]; 00736 d[2] = b->axis_bnd[5] - b->axis_bnd[4]; 00737 b->axis_bnd[0] -= tol * d[0], b->axis_bnd[1] += tol * d[0]; 00738 b->axis_bnd[2] -= tol * d[1], b->axis_bnd[3] += tol * d[1]; 00739 b->axis_bnd[4] -= tol * d[2], b->axis_bnd[5] += tol * d[2]; 00740 00741 c0[0] = 00742 tensor_ig3( p->Jr0, p->Dr0, p->dr.n, p->Js0, p->Ds0, p->ds.b.n, p->Jt0, p->Dt0, p->dt.b.n, x, jac, p->work ); 00743 c0[1] = tensor_ig3( p->Jr0, p->Dr0, p->dr.n, p->Js0, p->Ds0, p->ds.b.n, p->Jt0, p->Dt0, p->dt.b.n, y, jac + 3, 00744 p->work ); 00745 c0[2] = tensor_ig3( p->Jr0, p->Dr0, p->dr.n, p->Js0, p->Ds0, p->ds.b.n, p->Jt0, p->Dt0, p->dt.b.n, z, jac + 6, 00746 p->work ); 00747 mat_inv_3( jac, inv ); 00748 00749 obbox_bnd_3( p, x, y, z, c0, inv, bnd ); 00750 00751 u0[0] = ( bnd[0] + bnd[1] ) / 2; 00752 u0[1] = ( bnd[2] + bnd[3] ) / 2; 00753 u0[2] = ( bnd[4] + bnd[5] ) / 2; 00754 d[0] = 2 / ( ( 1 + tol ) * ( bnd[1] - bnd[0] ) ); 00755 d[1] = 2 / ( ( 1 + tol ) * ( bnd[3] - bnd[2] ) ); 00756 d[2] = 2 / ( ( 1 + tol ) * ( bnd[5] - bnd[4] ) ); 00757 b->x[0] = c0[0] + jac[0] * u0[0] + jac[1] * u0[1] + jac[2] * u0[2]; 00758 b->x[1] = c0[1] + jac[3] * u0[0] + jac[4] * u0[1] + jac[5] * u0[2]; 00759 b->x[2] = c0[2] + jac[6] * u0[0] + jac[7] * u0[1] + jac[8] * u0[2]; 00760 b->A[0] = d[0] * inv[0], b->A[1] = d[0] * inv[1], b->A[2] = d[0] * inv[2]; 00761 b->A[3] = d[1] * inv[3], b->A[4] = d[1] * inv[4], b->A[5] = d[1] * inv[5]; 00762 b->A[6] = d[2] * inv[6], b->A[7] = d[2] * inv[7], b->A[8] = d[2] * inv[8]; 00763 } 00764 00765 /*-------------------------------------------------------------------------- 00766 Point to Possible Elements Hashing 00767 00768 Initializing the data: 00769 unsigned nel; // number of elements 00770 const unsigned n[3]; // number of nodes in r, s, t directions 00771 const realType *xm[3]; // n[0]*n[1]*n[2]*nel x,y,z coordinates 00772 realType tol = 0.01; // how far point is allowed to be outside element 00773 // relative to element size 00774 unsigned max_size = n[0]*n[1]*n[2]*nel; // maximum size of hash table 00775 00776 hash_data_3 data; 00777 hash_build_3(&data, xm, n, nel, max_size, tol); 00778 00779 Using the data: 00780 realType x[3]; // point to find 00781 00782 unsigned index = hash_index_3(&data, x); 00783 unsigned i, b = data.offset[index], e = data.offset[index+1]; 00784 00785 // point may be in elements 00786 // data.offset[b], data.offset[b+1], ... , data.offset[e-1] 00787 // 00788 // list has maximum size data.max (e.g., e-b <= data.max) 00789 00790 for(i=b; i!=e; ++i) { 00791 unsigned el = data.offset[i]; 00792 const obbox_3 *obb = &data.obb[el]; // bounding box data for element el 00793 ... 00794 } 00795 00796 When done: 00797 hash_free_3(&data); 00798 00799 --------------------------------------------------------------------------*/ 00800 00801 static int ifloor( realType x ) 00802 { 00803 /* 00804 int y = x; 00805 return (double)y > x ? y-1 : y; 00806 */ 00807 return mbfloor( x ); 00808 } 00809 00810 static int iceil( realType x ) 00811 { 00812 /* 00813 int y = x; 00814 return (double)y < x ? y+1 : y; 00815 */ 00816 return mbceil( x ); 00817 } 00818 00819 static unsigned hash_index_helper( realType low, realType fac, unsigned n, realType x ) 00820 { 00821 const int i = ifloor( ( x - low ) * fac ); 00822 if( i < 0 ) return 0; 00823 return umin_2( i, n - 1 ); 00824 } 00825 00826 static uint hash_index_2( const hash_data_2* p, const realType x[2] ) 00827 { 00828 const unsigned n = p->hash_n; 00829 return (uint)hash_index_helper( p->bnd[2], p->fac[1], n, x[1] ) * n + 00830 hash_index_helper( p->bnd[0], p->fac[0], n, x[0] ); 00831 } 00832 00833 static uint hash_index_3( const hash_data_3* p, const realType x[3] ) 00834 { 00835 const unsigned n = p->hash_n; 00836 return ( (uint)hash_index_helper( p->bnd[4], p->fac[2], n, x[2] ) * n + 00837 hash_index_helper( p->bnd[2], p->fac[1], n, x[1] ) ) * 00838 n + 00839 hash_index_helper( p->bnd[0], p->fac[0], n, x[0] ); 00840 } 00841 00842 static void hash_setfac_2( hash_data_2* p, unsigned n ) 00843 { 00844 p->hash_n = n; 00845 p->fac[0] = n / ( p->bnd[1] - p->bnd[0] ); 00846 p->fac[1] = n / ( p->bnd[3] - p->bnd[2] ); 00847 } 00848 00849 static void hash_setfac_3( hash_data_3* p, unsigned n ) 00850 { 00851 p->hash_n = n; 00852 p->fac[0] = n / ( p->bnd[1] - p->bnd[0] ); 00853 p->fac[1] = n / ( p->bnd[3] - p->bnd[2] ); 00854 p->fac[2] = n / ( p->bnd[5] - p->bnd[4] ); 00855 } 00856 00857 static void hash_range_2( const hash_data_2* p, uint i, unsigned d, unsigned* ia, unsigned* ib ) 00858 { 00859 const realType a = p->obb[i].axis_bnd[d * 2]; 00860 const realType b = p->obb[i].axis_bnd[d * 2 + 1]; 00861 const int i0 = ifloor( ( a - p->bnd[d * 2] ) * p->fac[d] ); 00862 const unsigned i1 = iceil( ( b - p->bnd[d * 2] ) * p->fac[d] ); 00863 *ia = imax_2( 0, i0 ); 00864 *ib = imin_2( i1, p->hash_n ); 00865 if( *ib == *ia ) ++( *ib ); 00866 } 00867 00868 static void hash_range_3( const hash_data_3* p, uint i, unsigned d, unsigned* ia, unsigned* ib ) 00869 { 00870 const realType a = p->obb[i].axis_bnd[d * 2]; 00871 const realType b = p->obb[i].axis_bnd[d * 2 + 1]; 00872 const int i0 = ifloor( ( a - p->bnd[d * 2] ) * p->fac[d] ); 00873 const unsigned i1 = iceil( ( b - p->bnd[d * 2] ) * p->fac[d] ); 00874 *ia = imax_2( 0, i0 ); 00875 *ib = imin_2( i1, p->hash_n ); 00876 if( *ib == *ia ) ++( *ib ); 00877 } 00878 00879 static uint hash_count_2( hash_data_2* p, uint nel, unsigned n ) 00880 { 00881 uint i, count = 0; 00882 hash_setfac_2( p, n ); 00883 for( i = 0; i < nel; ++i ) 00884 { 00885 unsigned ia, ib; 00886 uint ci; 00887 hash_range_2( p, i, 0, &ia, &ib ); 00888 ci = ib - ia; 00889 hash_range_2( p, i, 1, &ia, &ib ); 00890 ci *= ib - ia; 00891 count += ci; 00892 } 00893 return count; 00894 } 00895 00896 static uint hash_count_3( hash_data_3* p, uint nel, unsigned n ) 00897 { 00898 uint i, count = 0; 00899 hash_setfac_3( p, n ); 00900 for( i = 0; i < nel; ++i ) 00901 { 00902 unsigned ia, ib; 00903 uint ci; 00904 hash_range_3( p, i, 0, &ia, &ib ); 00905 ci = ib - ia; 00906 hash_range_3( p, i, 1, &ia, &ib ); 00907 ci *= ib - ia; 00908 hash_range_3( p, i, 2, &ia, &ib ); 00909 ci *= ib - ia; 00910 count += ci; 00911 } 00912 return count; 00913 } 00914 00915 static uint hash_opt_size_2( hash_data_2* p, uint nel, uint max_size ) 00916 { 00917 unsigned nl = 1, nu = ceil( sqrt( max_size - nel ) ); 00918 uint size_low = 2 + nel; 00919 while( nu - nl > 1 ) 00920 { 00921 unsigned nm = nl + ( nu - nl ) / 2; 00922 uint size = (uint)nm * nm + 1 + hash_count_2( p, nel, nm ); 00923 if( size <= max_size ) 00924 nl = nm, size_low = size; 00925 else 00926 nu = nm; 00927 } 00928 hash_setfac_2( p, nl ); 00929 return size_low; 00930 } 00931 00932 static uint hash_opt_size_3( hash_data_3* p, uint nel, uint max_size ) 00933 { 00934 unsigned nl = 1, nu = ceil( pow( max_size - nel, 1.0 / 3 ) ); 00935 uint size_low = 2 + nel; 00936 while( nu - nl > 1 ) 00937 { 00938 unsigned nm = nl + ( nu - nl ) / 2; 00939 uint size = (uint)nm * nm * nm + 1 + hash_count_3( p, nel, nm ); 00940 if( size <= max_size ) 00941 nl = nm, size_low = size; 00942 else 00943 nu = nm; 00944 } 00945 hash_setfac_3( p, nl ); 00946 return size_low; 00947 } 00948 00949 static void hash_getbb_2( hash_data_2* p, const realType* const elx[2], const unsigned n[2], uint nel, realType tol ) 00950 { 00951 obbox_data_2* data; 00952 realType *z[2], *w[2]; 00953 uint i; 00954 unsigned d; 00955 const unsigned nn = n[0] * n[1]; 00956 unsigned int m[2]; 00957 const realType* x[2]; 00958 00959 for( i = 0; i < 2; ++i ) 00960 { 00961 x[i] = elx[i]; 00962 m[i] = 2 * n[i]; 00963 } 00964 00965 z[0] = tmalloc( realType, 2 * ( n[0] + n[1] ) ); 00966 w[0] = z[0] + n[0]; 00967 z[1] = w[0] + n[0], w[1] = z[1] + n[1]; 00968 for( d = 0; d < 2; ++d ) 00969 lobatto_nodes( z[d], n[d] ), lobatto_weights( z[d], w[d], n[d] ); 00970 data = obbox_setup_2( (const realType* const*)z, (const realType* const*)w, n, m ); 00971 obbox_calc_2( data, tol, x[0], x[1], &p->obb[0] ); 00972 memcpy( &p->bnd[0], (const realType*)&p->obb[0].axis_bnd[0], 4 * sizeof( realType ) ); 00973 for( i = 0; i < nel; ++i, x[0] += nn, x[1] += nn ) 00974 { 00975 obbox_calc_2( data, tol, x[0], x[1], &p->obb[i] ); 00976 obbox_merge_2( &p->bnd[0], (const realType*)&p->obb[i].axis_bnd[0] ); 00977 } 00978 obbox_free_2( data ); 00979 free( z[0] ); 00980 } 00981 00982 static void hash_getbb_3( hash_data_3* p, const realType* const elx[3], const unsigned n[3], uint nel, realType tol ) 00983 { 00984 obbox_data_3* data; 00985 const realType* x[3]; 00986 realType *z[3], *w[3]; 00987 uint i; 00988 unsigned d; 00989 const unsigned nn = n[0] * n[1] * n[2]; 00990 unsigned int m[3]; 00991 00992 for( i = 0; i < 3; ++i ) 00993 { 00994 x[i] = elx[i]; 00995 m[i] = 2 * n[i]; 00996 } 00997 00998 z[0] = tmalloc( realType, 2 * ( n[0] + n[1] + n[2] ) ); 00999 w[0] = z[0] + n[0]; 01000 for( d = 1; d < 3; ++d ) 01001 z[d] = w[d - 1] + n[d - 1], w[d] = z[d] + n[d]; 01002 for( d = 0; d < 3; ++d ) 01003 lobatto_nodes( z[d], n[d] ), lobatto_weights( z[d], w[d], n[d] ); 01004 data = obbox_setup_3( (const realType* const*)z, (const realType* const*)w, n, m ); 01005 obbox_calc_3( data, tol, x[0], x[1], x[2], &p->obb[0] ); 01006 memcpy( &p->bnd[0], (const realType*)&p->obb[0].axis_bnd[0], 6 * sizeof( realType ) ); 01007 for( i = 0; i < nel; ++i, x[0] += nn, x[1] += nn, x[2] += nn ) 01008 { 01009 obbox_calc_3( data, tol, x[0], x[1], x[2], &p->obb[i] ); 01010 obbox_merge_3( &p->bnd[0], (const realType*)&p->obb[i].axis_bnd[0] ); 01011 } 01012 obbox_free_3( data ); 01013 free( z[0] ); 01014 } 01015 01016 static void hash_build_2( hash_data_2* p, 01017 const realType* const x[2], 01018 const unsigned n[2], 01019 uint nel, 01020 uint max_hash_size, 01021 realType tol ) 01022 { 01023 uint i, el, size, hn2, sum; 01024 unsigned hn; 01025 unsigned* count; 01026 p->obb = tmalloc( obbox_2, nel ); 01027 hash_getbb_2( p, x, n, nel, tol ); 01028 size = hash_opt_size_2( p, nel, max_hash_size ); 01029 p->offset = tmalloc( uint, size ); 01030 hn = p->hash_n; 01031 hn2 = (uint)hn * hn; 01032 count = tcalloc( unsigned, hn2 ); 01033 for( el = 0; el < nel; ++el ) 01034 { 01035 unsigned ia, ib, j, ja, jb; 01036 hash_range_2( p, el, 0, &ia, &ib ); 01037 hash_range_2( p, el, 1, &ja, &jb ); 01038 for( j = ja; j < jb; ++j ) 01039 for( i = ia; i < ib; ++i ) 01040 ++count[(uint)j * hn + i]; 01041 } 01042 sum = hn2 + 1, p->max = count[0]; 01043 p->offset[0] = sum; 01044 for( i = 0; i < hn2; ++i ) 01045 { 01046 if( count[i] > p->max ) p->max = count[i]; 01047 sum += count[i]; 01048 p->offset[i + 1] = sum; 01049 } 01050 for( el = 0; el < nel; ++el ) 01051 { 01052 unsigned ia, ib, j, ja, jb; 01053 hash_range_2( p, el, 0, &ia, &ib ); 01054 hash_range_2( p, el, 1, &ja, &jb ); 01055 for( j = ja; j < jb; ++j ) 01056 for( i = ia; i < ib; ++i ) 01057 { 01058 uint arrindex = (uint)j * hn + i; 01059 p->offset[p->offset[arrindex + 1] - count[arrindex]] = el; 01060 --count[arrindex]; 01061 } 01062 } 01063 free( count ); 01064 } 01065 01066 static void hash_build_3( hash_data_3* p, 01067 const realType* const x[3], 01068 const unsigned n[3], 01069 uint nel, 01070 uint max_hash_size, 01071 realType tol ) 01072 { 01073 uint i, el, size, hn3, sum; 01074 unsigned hn; 01075 unsigned* count; 01076 p->obb = tmalloc( obbox_3, nel ); 01077 hash_getbb_3( p, x, n, nel, tol ); 01078 size = hash_opt_size_3( p, nel, max_hash_size ); 01079 p->offset = tmalloc( uint, size ); 01080 hn = p->hash_n; 01081 hn3 = (uint)hn * hn * hn; 01082 count = tcalloc( unsigned, hn3 ); 01083 for( el = 0; el < nel; ++el ) 01084 { 01085 unsigned ia, ib, j, ja, jb, k, ka, kb; 01086 hash_range_3( p, el, 0, &ia, &ib ); 01087 hash_range_3( p, el, 1, &ja, &jb ); 01088 hash_range_3( p, el, 2, &ka, &kb ); 01089 for( k = ka; k < kb; ++k ) 01090 for( j = ja; j < jb; ++j ) 01091 for( i = ia; i < ib; ++i ) 01092 ++count[( (uint)k * hn + j ) * hn + i]; 01093 } 01094 sum = hn3 + 1, p->max = count[0]; 01095 p->offset[0] = sum; 01096 for( i = 0; i < hn3; ++i ) 01097 { 01098 if( count[i] > p->max ) p->max = count[i]; 01099 sum += count[i]; 01100 p->offset[i + 1] = sum; 01101 } 01102 for( el = 0; el < nel; ++el ) 01103 { 01104 unsigned ia, ib, j, ja, jb, k, ka, kb; 01105 hash_range_3( p, el, 0, &ia, &ib ); 01106 hash_range_3( p, el, 1, &ja, &jb ); 01107 hash_range_3( p, el, 2, &ka, &kb ); 01108 for( k = ka; k < kb; ++k ) 01109 for( j = ja; j < jb; ++j ) 01110 for( i = ia; i < ib; ++i ) 01111 { 01112 uint arrindex = ( (uint)k * hn + j ) * hn + i; 01113 p->offset[p->offset[arrindex + 1] - count[arrindex]] = el; 01114 --count[arrindex]; 01115 } 01116 } 01117 free( count ); 01118 } 01119 01120 static void hash_free_2( hash_data_2* p ) 01121 { 01122 free( p->obb ); 01123 free( p->offset ); 01124 } 01125 01126 static void hash_free_3( hash_data_3* p ) 01127 { 01128 free( p->obb ); 01129 free( p->offset ); 01130 } 01131 01132 /*-------------------------------------------------------------------------- 01133 Optimization algorithm to find a point within an element 01134 01135 Given x(r) (as values of x,y,z at all Lobatto nodes) and x_star, 01136 find the r that minimizes || x_star - x(r) ||_2 01137 01138 As a minimization problem, the Newton step is 01139 01140 __ 3 01141 [ J^T J - >_ d=1 resid_d H_d ] dr = J^t resid 01142 01143 where resid = x_star - x(r), J = [ dx_i/dr_j ], 01144 and H_d = [ d^2 x_d/dr_i dr_j ]. 01145 01146 This is the appropriate step to take whenever constraints are active, 01147 and the current iterate is on a boundary of the element. When the current 01148 iterate is inside, J is square ( dim r = dim x ), resid will become small, 01149 and the step 01150 01151 J dr = resid 01152 01153 may be used instead, still giving quadratic convergence. 01154 01155 01156 Names use a _3 suffix for 3-d and _2 for 2-d. 01157 The routines require an initialized lagrange_data array as input: 01158 unsigned d, n[3] = { ... }; 01159 realType *z[3] = { tmalloc(realType, n[0]), ... }; 01160 for(d=0;d<3;++d) lobatto_nodes(z[d],n[d]); 01161 01162 lagrange_data ld[3]; 01163 for(d=0;d<3;++d) lagrange_setup(&ld[d],z[d],n[d]); 01164 01165 Initialization: 01166 opt_data_3 data; 01167 opt_alloc_3(&data, ld); 01168 01169 Use: 01170 const realType *xm[3]; // 3 pointers, each to n[0]*n[1]*n[2] realTypes 01171 // giving the nodal x, y, or z coordinates 01172 01173 const realType x_star[3] = { ... }; // point to find 01174 realType r[3] = { 0,0,0 }; // initial guess with 01175 unsigned c = opt_no_constraints_3; // these constraints active 01176 01177 realType dist = opt_findpt_3(&data,xm,x_star,r,&c); 01178 // minimizer is r with constraints c; 2-norm of resid is dist 01179 01180 Clean-up: 01181 opt_free_3(&data); 01182 01183 for(d=0;d<3;++d) lagrange_free(&ld[d]); 01184 for(d=0;d<3;++d) free(z[d]); 01185 01186 The constraint number works as follows. Let cr be the constraints 01187 on the r variable: 01188 cr = 0 r fixed at -1 01189 cr = 1 r not fixed 01190 cr = 2 r fixed at 1 01191 Then the constraint number is (ct*3+cs)*3+cr 01192 01193 --------------------------------------------------------------------------*/ 01194 01195 /* how many directions are constrained? */ 01196 static const char opt_constr_num_2[9] = { 2, 1, 2, 1, 0, 1, 2, 1, 2 }; 01197 static const char opt_constr_num_3[27] = { 3, 2, 3, 2, 1, 2, 3, 2, 3, 2, 1, 2, 1, 0, 01198 1, 2, 1, 2, 3, 2, 3, 2, 1, 2, 3, 2, 3 }; 01199 01200 /* which direction is constrained? */ 01201 /* static const char opt_constr_dir_2[9] = {-1, 1,-1, 0,-1, 0, -1, 1,-1}; */ 01202 static const char opt_constr_dir_3[27] = { -1, -1, -1, -1, 2, -1, -1, -1, -1, -1, 1, -1, 0, -1, 01203 0, -1, 1, -1, -1, -1, -1, -1, 2, -1, -1, -1, -1 }; 01204 01205 /* which direction is not constrained? */ 01206 static const char opt_constr_not[27] = { -1, 0, -1, 1, -1, 1, -1, 0, -1, 2, -1, 2, -1, -1, 01207 -1, 2, -1, 2, -1, 0, -1, 1, -1, 1, -1, 0, -1 }; 01208 01209 static const char opt_constr_wide[27] = { 0x00, 0x01, 0x02, 0x04, 0x05, 0x06, 0x08, 0x09, 0x0a, 01210 0x10, 0x11, 0x12, 0x14, 0x15, 0x16, 0x18, 0x19, 0x1a, 01211 0x20, 0x21, 0x22, 0x24, 0x25, 0x26, 0x28, 0x29, 0x2a }; 01212 01213 static const unsigned opt_other1_3[3] = { 1, 0, 0 }, opt_other2_3[3] = { 2, 2, 1 }; 01214 01215 static unsigned opt_constr( unsigned constraints, unsigned d ) 01216 { 01217 return ( opt_constr_wide[constraints] >> ( d * 2 ) ) & 3; 01218 } 01219 01220 static void opt_constr_unpack_2( unsigned constraints, unsigned* c ) 01221 { 01222 const char cw = opt_constr_wide[constraints]; 01223 c[0] = cw & 3; 01224 c[1] = cw >> 2; 01225 } 01226 01227 static void opt_constr_unpack_3( unsigned constraints, unsigned* c ) 01228 { 01229 const char cw = opt_constr_wide[constraints]; 01230 c[0] = cw & 3; 01231 c[1] = ( cw >> 2 ) & 3; 01232 c[2] = cw >> 4; 01233 } 01234 01235 static unsigned opt_constr_pack_2( const unsigned* c ) 01236 { 01237 return c[1] * 3 + c[0]; 01238 } 01239 01240 static unsigned opt_constr_pack_3( const unsigned* c ) 01241 { 01242 return ( c[2] * 3 + c[1] ) * 3 + c[0]; 01243 } 01244 01245 /*-------------------------------------------------------------------------- 01246 01247 3 - D 01248 01249 --------------------------------------------------------------------------*/ 01250 01251 void opt_alloc_3( opt_data_3* p, lagrange_data* ld ) 01252 { 01253 const unsigned nr = ld[0].n, ns = ld[1].n, nt = ld[2].n, nf = umax_3( nr * ns, nr * nt, ns * nt ), 01254 ne = umax_3( nr, ns, nt ), nw = 2 * ns * nt + 3 * ns; 01255 p->size[0] = 1; 01256 p->size[1] = nr; 01257 p->size[2] = nr * ns; 01258 p->size[3] = p->size[2] * nt; 01259 p->ld = ld; 01260 p->work = tmalloc( realType, 6 * nf + 9 * ne + nw ); 01261 p->fd.x[0] = p->work + nw; 01262 p->fd.x[1] = p->fd.x[0] + nf; 01263 p->fd.x[2] = p->fd.x[1] + nf; 01264 p->fd.fdn[0] = p->fd.x[2] + nf; 01265 p->fd.fdn[1] = p->fd.fdn[0] + nf; 01266 p->fd.fdn[2] = p->fd.fdn[1] + nf; 01267 p->ed.x[0] = p->fd.fdn[2] + nf; 01268 p->ed.x[1] = p->ed.x[0] + ne; 01269 p->ed.x[2] = p->ed.x[1] + ne; 01270 p->ed.fd1[0] = p->ed.x[2] + ne; 01271 p->ed.fd1[1] = p->ed.fd1[0] + ne; 01272 p->ed.fd1[2] = p->ed.fd1[1] + ne; 01273 p->ed.fd2[0] = p->ed.fd1[2] + ne; 01274 p->ed.fd2[1] = p->ed.fd2[0] + ne; 01275 p->ed.fd2[2] = p->ed.fd2[1] + ne; 01276 } 01277 01278 void opt_free_3( opt_data_3* p ) 01279 { 01280 free( p->work ); 01281 } 01282 01283 static void opt_vol_set_3( opt_data_3* p, const realType r[3] ) 01284 { 01285 lagrange_1( &p->ld[0], r[0] ); 01286 lagrange_1( &p->ld[1], r[1] ); 01287 lagrange_1( &p->ld[2], r[2] ); 01288 } 01289 01290 /* work holds 2*ns*nt + 3*ns realTypes */ 01291 static void opt_vol_intp_3( opt_data_3* p ) 01292 { 01293 unsigned d; 01294 const lagrange_data* ld = p->ld; 01295 01296 for( d = 0; d < 3; ++d ) 01297 p->x[d] = tensor_ig3( ld[0].J, ld[0].D, ld[0].n, ld[1].J, ld[1].D, ld[1].n, ld[2].J, ld[2].D, ld[2].n, 01298 p->elx[d], &p->jac[d * 3], p->work ); 01299 } 01300 01301 void opt_vol_set_intp_3( opt_data_3* p, const realType r[3] ) 01302 { 01303 opt_vol_set_3( p, r ); 01304 opt_vol_intp_3( p ); 01305 } 01306 01307 static void opt_face_proj_3( opt_data_3* p ) 01308 { 01309 unsigned d, off = 0; 01310 const unsigned dn = p->fd.dn, d1 = p->fd.d1, d2 = p->fd.d2, so = p->size[d2] - p->size[d1 + 1], s1 = p->size[d1], 01311 sn = p->size[dn], n1 = p->ld[d1].n, n2 = p->ld[d2].n, nn = p->ld[dn].n; 01312 const realType* D = p->ld[dn].D_z0; 01313 if( opt_constr( p->fd.constraints, dn ) == 2 ) off = p->size[dn + 1] - p->size[dn], D = p->ld[dn].D_zn; 01314 for( d = 0; d < 3; ++d ) 01315 { 01316 unsigned i, j, k, arrindex = 0; 01317 const realType* in = p->elx[d] + off; 01318 for( j = n2; j; --j, in += so ) 01319 for( i = n1; i; --i, ++arrindex, in += s1 ) 01320 { 01321 const realType* ind = in - off; 01322 realType* fdn = &p->fd.fdn[d][arrindex]; 01323 p->fd.x[d][arrindex] = *in; 01324 *fdn = 0; 01325 for( k = 0; k < nn; ++k, ind += sn ) 01326 *fdn += *ind * D[k]; 01327 } 01328 } 01329 } 01330 01331 static void opt_face_set_3( opt_data_3* p, const realType r[3], unsigned constr ) 01332 { 01333 if( p->fd.constraints != constr ) 01334 { 01335 p->fd.constraints = constr; 01336 p->fd.dn = opt_constr_dir_3[constr]; 01337 p->fd.d1 = opt_other1_3[p->fd.dn]; 01338 p->fd.d2 = opt_other2_3[p->fd.dn]; 01339 opt_face_proj_3( p ); 01340 } 01341 lagrange_1( &p->ld[p->fd.d1], r[p->fd.d1] ); 01342 lagrange_1( &p->ld[p->fd.d2], r[p->fd.d2] ); 01343 } 01344 01345 /* work holds 2*ld[d2].n realTypes */ 01346 static void opt_face_intp_3( opt_data_3* p ) 01347 { 01348 unsigned d; 01349 const unsigned dn = p->fd.dn, d1 = p->fd.d1, d2 = p->fd.d2, n1 = p->ld[d1].n, n2 = p->ld[d2].n; 01350 const realType *J1 = p->ld[d1].J, *J2 = p->ld[d2].J, *D1 = p->ld[d1].D, *D2 = p->ld[d2].D; 01351 01352 for( d = 0; d < 3; ++d ) 01353 { 01354 realType g[2]; 01355 p->x[d] = tensor_ig2( J1, D1, n1, J2, D2, n2, p->fd.x[d], &g[0], p->work ); 01356 p->jac[d * 3 + d1] = g[0]; 01357 p->jac[d * 3 + d2] = g[1]; 01358 p->jac[d * 3 + dn] = tensor_i2( J1, n1, J2, n2, p->fd.fdn[d], p->work ); 01359 } 01360 } 01361 01362 static void opt_face_set_intp_3( opt_data_3* p, const realType r[3], unsigned constr ) 01363 { 01364 opt_face_set_3( p, r, constr ); 01365 opt_face_intp_3( p ); 01366 } 01367 01368 static void opt_face_hess_3( opt_data_3* p, realType hess[9] ) 01369 { 01370 unsigned d; 01371 const unsigned d1 = p->fd.d1, d2 = p->fd.d2, n1 = p->ld[d1].n, n2 = p->ld[d2].n; 01372 const realType *J1 = p->ld[d1].J, *J2 = p->ld[d2].J, *D1 = p->ld[d1].D, *D2 = p->ld[d2].D, *S1 = p->ld[d1].D2, 01373 *S2 = p->ld[d2].D2; 01374 01375 lagrange_2u( &p->ld[d1] ); 01376 lagrange_2u( &p->ld[d2] ); 01377 01378 for( d = 0; d < 3; ++d ) 01379 { 01380 (void)tensor_ig2( J1, S1, n1, J2, S2, n2, p->fd.x[d], hess + d * 3, p->work ); 01381 hess[d * 3 + 0] = tensor_i2( S1, n1, J2, n2, p->fd.x[d], p->work ); 01382 hess[d * 3 + 1] = tensor_i2( J1, n1, S2, n2, p->fd.x[d], p->work ); 01383 hess[d * 3 + 2] = tensor_i2( D1, n1, D2, n2, p->fd.x[d], p->work ); 01384 } 01385 } 01386 01387 static void opt_edge_proj_3( opt_data_3* p ) 01388 { 01389 unsigned d, off, off1 = 0, off2 = 0; 01390 const unsigned de = p->ed.de, d1 = p->ed.d1, d2 = p->ed.d2, se = p->size[de], s1 = p->size[d1], s2 = p->size[d2], 01391 ne = p->ld[de].n, n1 = p->ld[d1].n, n2 = p->ld[d2].n; 01392 const realType *fD1, *fD2; 01393 if( opt_constr( p->ed.constraints, d1 ) == 0 ) 01394 fD1 = p->ld[d1].D_z0; 01395 else 01396 fD1 = p->ld[d1].D_zn, off1 = p->size[d1 + 1] - p->size[d1]; 01397 if( opt_constr( p->ed.constraints, d2 ) == 0 ) 01398 fD2 = p->ld[d2].D_z0; 01399 else 01400 fD2 = p->ld[d2].D_zn, off2 = p->size[d2 + 1] - p->size[d2]; 01401 off = off1 + off2; 01402 for( d = 0; d < 3; ++d ) 01403 { 01404 unsigned i, j; 01405 const realType* in = p->elx[d] + off; 01406 for( i = 0; i < ne; ++i, in += se ) 01407 { 01408 const realType *in1 = in - off1, *in2 = in - off2; 01409 realType *fd1 = &p->ed.fd1[d][i], *fd2 = &p->ed.fd2[d][i]; 01410 p->ed.x[d][i] = *in; 01411 *fd1 = *fd2 = 0; 01412 for( j = 0; j < n1; ++j, in1 += s1 ) 01413 *fd1 += *in1 * fD1[j]; 01414 for( j = 0; j < n2; ++j, in2 += s2 ) 01415 *fd2 += *in2 * fD2[j]; 01416 } 01417 } 01418 } 01419 01420 static void opt_edge_set_3( opt_data_3* p, const realType r[3], unsigned constr ) 01421 { 01422 if( p->ed.constraints != constr ) 01423 { 01424 p->ed.constraints = constr; 01425 p->ed.de = opt_constr_not[constr]; 01426 p->ed.d1 = opt_other1_3[p->ed.de]; 01427 p->ed.d2 = opt_other2_3[p->ed.de]; 01428 opt_edge_proj_3( p ); 01429 } 01430 lagrange_1( &p->ld[p->ed.de], r[p->ed.de] ); 01431 } 01432 01433 static void opt_edge_intp_3( opt_data_3* p ) 01434 { 01435 unsigned d; 01436 const unsigned de = p->ed.de, d1 = p->ed.d1, d2 = p->ed.d2, n = p->ld[de].n; 01437 const realType *J = p->ld[de].J, *D = p->ld[de].D; 01438 01439 for( d = 0; d < 3; ++d ) 01440 { 01441 p->x[d] = tensor_ig1( J, D, n, p->ed.x[d], &p->jac[d * 3 + de] ); 01442 p->jac[d * 3 + d1] = tensor_i1( J, n, p->ed.fd1[d] ); 01443 p->jac[d * 3 + d2] = tensor_i1( J, n, p->ed.fd2[d] ); 01444 } 01445 } 01446 01447 static void opt_edge_set_intp_3( opt_data_3* p, const realType r[3], unsigned constr ) 01448 { 01449 opt_edge_set_3( p, r, constr ); 01450 opt_edge_intp_3( p ); 01451 } 01452 01453 static void opt_edge_hess_3( opt_data_3* p, realType hess[3] ) 01454 { 01455 unsigned d; 01456 const unsigned de = p->ed.de, n = p->ld[de].n; 01457 const realType* D2 = p->ld[de].D2; 01458 lagrange_2u( &p->ld[de] ); 01459 for( d = 0; d < 3; ++d ) 01460 hess[d] = tensor_i1( D2, n, p->ed.x[d] ); 01461 } 01462 01463 static void opt_point_proj_3( opt_data_3* p ) 01464 { 01465 unsigned off[3], offt, d, c[3]; 01466 const realType* fD[3]; 01467 opt_constr_unpack_3( p->pd.constraints, c ); 01468 for( d = 0; d < 3; ++d ) 01469 if( c[d] == 0 ) 01470 fD[d] = p->ld[d].D_z0, off[d] = 0; 01471 else 01472 fD[d] = p->ld[d].D_zn, off[d] = p->size[d + 1] - p->size[d]; 01473 offt = off[0] + off[1] + off[2]; 01474 for( d = 0; d < 9; ++d ) 01475 p->pd.jac[d] = 0; 01476 for( d = 0; d < 3; ++d ) 01477 { 01478 unsigned i, j; 01479 p->pd.x[d] = p->elx[d][offt]; 01480 for( i = 0; i < 3; ++i ) 01481 { 01482 const realType* in = p->elx[d] + offt - off[i]; 01483 for( j = 0; j < p->ld[i].n; ++j, in += p->size[i] ) 01484 p->pd.jac[d * 3 + i] += *in * fD[i][j]; 01485 } 01486 } 01487 } 01488 01489 static void opt_point_set_3( opt_data_3* p, unsigned constr ) 01490 { 01491 if( p->pd.constraints != constr ) 01492 { 01493 p->pd.constraints = constr; 01494 opt_point_proj_3( p ); 01495 } 01496 } 01497 01498 static void opt_point_intp_3( opt_data_3* p ) 01499 { 01500 memcpy( p->x, p->pd.x, 3 * sizeof( realType ) ); 01501 memcpy( p->jac, p->pd.jac, 9 * sizeof( realType ) ); 01502 } 01503 01504 static void opt_point_set_intp_3( opt_data_3* p, unsigned constr ) 01505 { 01506 opt_point_set_3( p, constr ); 01507 opt_point_intp_3( p ); 01508 } 01509 01510 #define DIAGNOSTICS 0 01511 01512 double opt_findpt_3( opt_data_3* p, 01513 const realType* const elx[3], 01514 const realType xstar[3], 01515 realType r[3], 01516 unsigned* constr ) 01517 { 01518 realType dr[3], resid[3], steep[3]; 01519 01520 unsigned c = *constr, ac, d, cc[3], step = 0; 01521 01522 p->elx[0] = elx[0], p->elx[1] = elx[1], p->elx[2] = elx[2]; 01523 01524 p->fd.constraints = opt_no_constraints_3; 01525 p->ed.constraints = opt_no_constraints_3; 01526 p->pd.constraints = opt_no_constraints_3; 01527 01528 #if DIAGNOSTICS 01529 printf( "opt_findpt: xstar = %g, %g, %g\n", xstar[0], xstar[1], xstar[2] ); 01530 #endif 01531 01532 do 01533 { 01534 ++step; 01535 if( step == 50 ) /*fail("%s: opt_findpt_3 did not converge\n",__FILE__);*/ 01536 return 1.e+30; 01537 #if DIAGNOSTICS 01538 printf( " iteration %u\n", step ); 01539 printf( " %d constraint(s) active\n", (int)opt_constr_num_3[c] ); 01540 #endif 01541 /* update face/edge/point data if necessary, 01542 and evaluate x(r) as well as the jacobian */ 01543 switch( opt_constr_num_3[c] ) 01544 { 01545 case 0: 01546 opt_vol_set_intp_3( p, r ); 01547 break; 01548 case 1: 01549 opt_face_set_intp_3( p, r, c ); 01550 break; 01551 case 2: 01552 opt_edge_set_intp_3( p, r, c ); 01553 break; 01554 case 3: 01555 opt_point_set_intp_3( p, c ); 01556 break; 01557 } 01558 #if DIAGNOSTICS 01559 printf( " r = %g, %g, %g\n", r[0], r[1], r[2] ); 01560 printf( " x = %g, %g, %g\n", p->x[0], p->x[1], p->x[2] ); 01561 #endif 01562 /* compute residual */ 01563 for( d = 0; d < 3; ++d ) 01564 resid[d] = xstar[d] - p->x[d]; 01565 #if DIAGNOSTICS 01566 printf( " resid = %g, %g, %g\n", resid[0], resid[1], resid[2] ); 01567 printf( " 2-norm = %g\n", r2norm_3( resid[0], resid[1], resid[2] ) ); 01568 #endif 01569 /* check constraints against steepest descent direction */ 01570 ac = c; 01571 if( opt_constr_num_3[c] ) 01572 { 01573 opt_constr_unpack_3( c, cc ); 01574 mat_app_3c( steep, p->jac, resid ); /* steepest descent = J^T r */ 01575 #if DIAGNOSTICS 01576 printf( " steepest descent = %g, %g, %g\n", steep[0], steep[1], steep[2] ); 01577 #endif 01578 for( d = 0; d < 3; ++d ) 01579 if( ( cc[d] == 0 && steep[d] > 0 ) || ( cc[d] == 2 && steep[d] < 0 ) ) cc[d] = 1; 01580 ac = opt_constr_pack_3( cc ); 01581 } 01582 /* update face/edge/point data if necessary */ 01583 if( ac != c ) 01584 { 01585 c = ac; 01586 #if DIAGNOSTICS 01587 printf( " relaxed to %d constraints\n", (int)opt_constr_num_3[c] ); 01588 #endif 01589 switch( opt_constr_num_3[c] ) 01590 { 01591 case 1: 01592 opt_face_set_3( p, r, c ); 01593 break; 01594 case 2: 01595 opt_edge_set_3( p, r, c ); 01596 break; 01597 case 3: 01598 opt_point_set_3( p, c ); 01599 break; 01600 } 01601 } 01602 /* compute Newton step */ 01603 switch( opt_constr_num_3[c] ) 01604 { 01605 case 0: 01606 tinyla_solve_3( dr, p->jac, resid ); 01607 break; 01608 case 1: { 01609 const unsigned dn = p->fd.dn, d1 = p->fd.d1, d2 = p->fd.d2; 01610 realType A[4], H[9]; 01611 const realType* J = p->jac; 01612 opt_face_hess_3( p, H ); 01613 A[0] = J[d1] * J[d1] + J[3 + d1] * J[3 + d1] + J[6 + d1] * J[6 + d1]; 01614 A[1] = J[d2] * J[d2] + J[3 + d2] * J[3 + d2] + J[6 + d2] * J[6 + d2]; 01615 A[2] = J[d1] * J[d2] + J[3 + d1] * J[3 + d2] + J[6 + d1] * J[6 + d2]; 01616 A[0] -= resid[0] * H[0] + resid[1] * H[3] + resid[2] * H[6]; 01617 A[1] -= resid[0] * H[1] + resid[1] * H[4] + resid[2] * H[7]; 01618 A[2] -= resid[0] * H[2] + resid[1] * H[5] + resid[2] * H[8]; 01619 tinyla_solve_sym_2( &dr[d1], &dr[d2], A, steep[d1], steep[d2] ); 01620 dr[dn] = 0; 01621 } 01622 break; 01623 case 2: { 01624 const unsigned de = p->ed.de, d1 = p->ed.d1, d2 = p->ed.d2; 01625 realType fac, H[3]; 01626 const realType* J = p->jac + de; 01627 opt_edge_hess_3( p, H ); 01628 fac = J[0] * J[0] + J[3] * J[3] + J[6] * J[6] - ( resid[0] * H[0] + resid[1] * H[1] + resid[2] * H[2] ); 01629 dr[de] = steep[de] / fac; 01630 dr[d1] = 0, dr[d2] = 0; 01631 } 01632 break; 01633 case 3: 01634 dr[0] = dr[1] = dr[2] = 0; 01635 break; 01636 } 01637 #if DIAGNOSTICS 01638 printf( " dr = %g, %g, %g\n", dr[0], dr[1], dr[2] ); 01639 #endif 01640 /* project new iteration onto [-1,1]^3 */ 01641 opt_constr_unpack_3( c, cc ); 01642 for( d = 0; d < 3; ++d ) 01643 { 01644 if( cc[d] != 1 ) continue; 01645 r[d] += dr[d]; 01646 if( r[d] <= -1 ) 01647 dr[d] -= r[d] + 1, r[d] = -1, cc[d] = 0; 01648 else if( r[d] >= 1 ) 01649 dr[d] -= r[d] - 1, r[d] = 1, cc[d] = 2; 01650 } 01651 c = opt_constr_pack_3( cc ); 01652 } while( r1norm_3( dr[0], dr[1], dr[2] ) > 3 * MOAB_POLY_EPS ); 01653 *constr = c; 01654 #if 0 01655 printf("opt_findpt_3 converged in %u iterations\n", step); 01656 #endif 01657 return r2norm_3( resid[0], resid[1], resid[2] ); 01658 } 01659 01660 #undef DIAGNOSTICS 01661 01662 void opt_alloc_2( opt_data_2* p, lagrange_data* ld ) 01663 { 01664 const unsigned nr = ld[0].n, ns = ld[1].n, ne = umax_2( nr, ns ), nw = 2 * ns; 01665 p->size[0] = 1; 01666 p->size[1] = nr; 01667 p->size[2] = nr * ns; 01668 p->ld = ld; 01669 p->work = tmalloc( realType, 4 * ne + nw ); 01670 p->ed.x[0] = p->work + nw; 01671 p->ed.x[1] = p->ed.x[0] + ne; 01672 p->ed.fd1[0] = p->ed.x[1] + ne; 01673 p->ed.fd1[1] = p->ed.fd1[0] + ne; 01674 } 01675 01676 void opt_free_2( opt_data_2* p ) 01677 { 01678 free( p->work ); 01679 } 01680 01681 static void opt_area_set_2( opt_data_2* p, const realType r[2] ) 01682 { 01683 lagrange_1( &p->ld[0], r[0] ); 01684 lagrange_1( &p->ld[1], r[1] ); 01685 } 01686 01687 /* work holds 2*ns realTypes */ 01688 static void opt_area_intp_2( opt_data_2* p ) 01689 { 01690 unsigned d; 01691 const lagrange_data* ld = p->ld; 01692 01693 for( d = 0; d < 2; ++d ) 01694 p->x[d] = 01695 tensor_ig2( ld[0].J, ld[0].D, ld[0].n, ld[1].J, ld[1].D, ld[1].n, p->elx[d], &p->jac[d * 2], p->work ); 01696 } 01697 01698 static void opt_area_set_intp_2( opt_data_2* p, const realType r[2] ) 01699 { 01700 opt_area_set_2( p, r ); 01701 opt_area_intp_2( p ); 01702 } 01703 01704 static void opt_edge_proj_2( opt_data_2* p ) 01705 { 01706 unsigned d, off = 0; 01707 const unsigned de = p->ed.de, d1 = p->ed.d1, se = p->size[de], s1 = p->size[d1], ne = p->ld[de].n, n1 = p->ld[d1].n; 01708 const realType* fD1; 01709 if( opt_constr( p->ed.constraints, d1 ) == 0 ) 01710 fD1 = p->ld[d1].D_z0; 01711 else 01712 fD1 = p->ld[d1].D_zn, off = p->size[d1 + 1] - p->size[d1]; 01713 for( d = 0; d < 2; ++d ) 01714 { 01715 unsigned i, j; 01716 const realType* in = p->elx[d] + off; 01717 for( i = 0; i < ne; ++i, in += se ) 01718 { 01719 const realType* in1 = in - off; 01720 realType* fd1 = &p->ed.fd1[d][i]; 01721 p->ed.x[d][i] = *in; 01722 *fd1 = 0; 01723 for( j = 0; j < n1; ++j, in1 += s1 ) 01724 *fd1 += *in1 * fD1[j]; 01725 } 01726 } 01727 } 01728 01729 static void opt_edge_set_2( opt_data_2* p, const realType r[2], unsigned constr ) 01730 { 01731 if( p->ed.constraints != constr ) 01732 { 01733 p->ed.constraints = constr; 01734 p->ed.de = opt_constr_not[constr]; 01735 p->ed.d1 = 1 - p->ed.de; 01736 opt_edge_proj_2( p ); 01737 } 01738 lagrange_1( &p->ld[p->ed.de], r[p->ed.de] ); 01739 } 01740 01741 static void opt_edge_intp_2( opt_data_2* p ) 01742 { 01743 unsigned d; 01744 const unsigned de = p->ed.de, d1 = p->ed.d1, n = p->ld[de].n; 01745 const realType *J = p->ld[de].J, *D = p->ld[de].D; 01746 for( d = 0; d < 2; ++d ) 01747 { 01748 p->x[d] = tensor_ig1( J, D, n, p->ed.x[d], &p->jac[d * 2 + de] ); 01749 p->jac[d * 2 + d1] = tensor_i1( J, n, p->ed.fd1[d] ); 01750 } 01751 } 01752 01753 static void opt_edge_set_intp_2( opt_data_2* p, const realType r[2], unsigned constr ) 01754 { 01755 opt_edge_set_2( p, r, constr ); 01756 opt_edge_intp_2( p ); 01757 } 01758 01759 static void opt_edge_hess_2( opt_data_2* p, realType hess[2] ) 01760 { 01761 unsigned d; 01762 const unsigned de = p->ed.de, n = p->ld[de].n; 01763 const realType* D2 = p->ld[de].D2; 01764 lagrange_2u( &p->ld[de] ); 01765 for( d = 0; d < 2; ++d ) 01766 hess[d] = tensor_i1( D2, n, p->ed.x[d] ); 01767 } 01768 01769 static void opt_point_proj_2( opt_data_2* p ) 01770 { 01771 unsigned off[2], offt, d, c[2]; 01772 const realType* fD[2]; 01773 opt_constr_unpack_2( p->pd.constraints, c ); 01774 for( d = 0; d < 2; ++d ) 01775 if( c[d] == 0 ) 01776 fD[d] = p->ld[d].D_z0, off[d] = 0; 01777 else 01778 fD[d] = p->ld[d].D_zn, off[d] = p->size[d + 1] - p->size[d]; 01779 offt = off[0] + off[1]; 01780 for( d = 0; d < 4; ++d ) 01781 p->pd.jac[d] = 0; 01782 for( d = 0; d < 2; ++d ) 01783 { 01784 unsigned i, j; 01785 p->pd.x[d] = p->elx[d][offt]; 01786 for( i = 0; i < 2; ++i ) 01787 { 01788 const realType* in = p->elx[d] + offt - off[i]; 01789 for( j = 0; j < p->ld[i].n; ++j, in += p->size[i] ) 01790 p->pd.jac[d * 2 + i] += *in * fD[i][j]; 01791 } 01792 } 01793 } 01794 01795 static void opt_point_set_2( opt_data_2* p, unsigned constr ) 01796 { 01797 if( p->pd.constraints != constr ) 01798 { 01799 p->pd.constraints = constr; 01800 opt_point_proj_2( p ); 01801 } 01802 } 01803 01804 static void opt_point_intp_2( opt_data_2* p ) 01805 { 01806 memcpy( p->x, p->pd.x, 2 * sizeof( realType ) ); 01807 memcpy( p->jac, p->pd.jac, 4 * sizeof( realType ) ); 01808 } 01809 01810 static void opt_point_set_intp_2( opt_data_2* p, unsigned constr ) 01811 { 01812 opt_point_set_2( p, constr ); 01813 opt_point_intp_2( p ); 01814 } 01815 01816 #define DIAGNOSTICS 0 01817 01818 double opt_findpt_2( opt_data_2* p, 01819 const realType* const elx[2], 01820 const realType xstar[2], 01821 realType r[2], 01822 unsigned* constr ) 01823 { 01824 realType dr[2], resid[2], steep[2]; 01825 01826 unsigned c = *constr, ac, d, cc[2], step = 0; 01827 01828 p->elx[0] = elx[0], p->elx[1] = elx[1]; 01829 01830 p->ed.constraints = opt_no_constraints_2; 01831 p->pd.constraints = opt_no_constraints_2; 01832 01833 #if DIAGNOSTICS 01834 printf( "opt_findpt: xstar = %g, %g\n", xstar[0], xstar[1] ); 01835 #endif 01836 01837 do 01838 { 01839 ++step; 01840 if( step == 50 ) /*fail("%s: opt_findpt_2 did not converge\n",__FILE__);*/ 01841 return 1.e+30; 01842 #if DIAGNOSTICS 01843 printf( " iteration %u\n", step ); 01844 printf( " %d constraint(s) active\n", (int)opt_constr_num_2[c] ); 01845 #endif 01846 /* update face/edge/point data if necessary, 01847 and evaluate x(r) as well as the jacobian */ 01848 switch( opt_constr_num_2[c] ) 01849 { 01850 case 0: 01851 opt_area_set_intp_2( p, r ); 01852 break; 01853 case 1: 01854 opt_edge_set_intp_2( p, r, c ); 01855 break; 01856 case 2: 01857 opt_point_set_intp_2( p, c ); 01858 break; 01859 } 01860 #if DIAGNOSTICS 01861 printf( " r = %g, %g\n", r[0], r[1] ); 01862 printf( " x = %g, %g\n", p->x[0], p->x[1] ); 01863 #endif 01864 /* compute residual */ 01865 for( d = 0; d < 2; ++d ) 01866 resid[d] = xstar[d] - p->x[d]; 01867 #if DIAGNOSTICS 01868 printf( " resid = %g, %g\n", resid[0], resid[1] ); 01869 printf( " 2-norm = %g\n", r2norm_2( resid[0], resid[1] ) ); 01870 #endif 01871 /* check constraints against steepest descent direction */ 01872 ac = c; 01873 if( opt_constr_num_2[c] ) 01874 { 01875 opt_constr_unpack_2( c, cc ); 01876 mat_app_2c( steep, p->jac, resid ); /* steepest descent = J^T r */ 01877 #if DIAGNOSTICS 01878 printf( " steepest descent = %g, %g\n", steep[0], steep[1] ); 01879 #endif 01880 for( d = 0; d < 2; ++d ) 01881 if( ( cc[d] == 0 && steep[d] > 0 ) || ( cc[d] == 2 && steep[d] < 0 ) ) cc[d] = 1; 01882 ac = opt_constr_pack_2( cc ); 01883 } 01884 /* update face/edge/point data if necessary */ 01885 if( ac != c ) 01886 { 01887 c = ac; 01888 #if DIAGNOSTICS 01889 printf( " relaxed to %d constraints\n", (int)opt_constr_num_2[c] ); 01890 #endif 01891 switch( opt_constr_num_2[c] ) 01892 { 01893 case 1: 01894 opt_edge_set_2( p, r, c ); 01895 break; 01896 case 2: 01897 opt_point_set_2( p, c ); 01898 break; 01899 } 01900 } 01901 /* compute Newton step */ 01902 switch( opt_constr_num_2[c] ) 01903 { 01904 case 0: 01905 tinyla_solve_2( dr, p->jac, resid ); 01906 break; 01907 case 1: { 01908 const unsigned de = p->ed.de, d1 = p->ed.d1; 01909 realType fac, H[2]; 01910 const realType* J = p->jac + de; 01911 opt_edge_hess_2( p, H ); 01912 fac = J[0] * J[0] + J[2] * J[2] - ( resid[0] * H[0] + resid[1] * H[1] ); 01913 dr[de] = steep[de] / fac; 01914 dr[d1] = 0; 01915 } 01916 break; 01917 case 2: 01918 dr[0] = dr[1] = 0; 01919 break; 01920 } 01921 #if DIAGNOSTICS 01922 printf( " dr = %g, %g\n", dr[0], dr[1] ); 01923 #endif 01924 /* project new iteration onto [-1,1]^2 */ 01925 opt_constr_unpack_2( c, cc ); 01926 for( d = 0; d < 2; ++d ) 01927 { 01928 if( cc[d] != 1 ) continue; 01929 r[d] += dr[d]; 01930 if( r[d] <= -1 ) 01931 dr[d] -= r[d] + 1, r[d] = -1, cc[d] = 0; 01932 else if( r[d] >= 1 ) 01933 dr[d] -= r[d] - 1, r[d] = 1, cc[d] = 2; 01934 } 01935 c = opt_constr_pack_2( cc ); 01936 } while( r1norm_2( dr[0], dr[1] ) > 2 * MOAB_POLY_EPS ); 01937 *constr = c; 01938 return r2norm_2( resid[0], resid[1] ); 01939 } 01940 01941 #undef DIAGNOSTICS 01942 01943 /*-------------------------------------------------------------------------- 01944 Point Finding (interface/top-level) 01945 01946 Initializing the data: 01947 unsigned nel; // number of elements 01948 const unsigned n[3]; // number of nodes in r, s, t directions 01949 const realType *xm[3]; // n[0]*n[1]*n[2]*nel x,y,z coordinates 01950 realType tol = 0.01; // how far point is allowed to be outside element 01951 // relative to element size 01952 unsigned max_size = n[0]*n[1]*n[2]*nel; // maximum size of hash table 01953 01954 findpt_data_3 *data = findpt_setup_3(xm,n,nel,max_size,tol); 01955 01956 Using the data: 01957 realType x[3] = { ... }; // point to find 01958 int el; // element number 01959 realType r[3]; // parametric coordinates 01960 int guess = 0; // do we use (el,r,s,t) as an initial guess? 01961 int code; // 0 : normal, -1 : outside all elements, 01962 // 1 : border, or outside but within tolerance 01963 realType dist; // distance in xyz space from returned (el,r,s,t) to given 01964 // (x,y,z) 01965 01966 code = findpt_3(data, x, guess, &el, r, &dist); 01967 01968 When done: 01969 findpt_free_3(&data); 01970 01971 --------------------------------------------------------------------------*/ 01972 01973 /* heap sort on A[0:n-1] with key A[i]->dist 01974 precondition: n!=0 */ 01975 static void findpt_list_sort( findpt_listel** A, unsigned n ) 01976 { 01977 unsigned i; 01978 --A; /* make A have a base index of 1 */ 01979 /* build heap */ 01980 for( i = 2; i <= n; ++i ) 01981 { 01982 findpt_listel* item = A[i]; 01983 unsigned hole = i, parent = hole >> 1; 01984 if( A[parent]->dist >= item->dist ) continue; 01985 do 01986 { 01987 A[hole] = A[parent]; 01988 hole = parent; 01989 parent >>= 1; 01990 } while( parent && A[parent]->dist < item->dist ); 01991 A[hole] = item; 01992 } 01993 /* extract */ 01994 for( i = n - 1; i; --i ) 01995 { 01996 findpt_listel* item = A[i + 1]; 01997 unsigned hole = 1; 01998 A[i + 1] = A[1]; 01999 for( ;; ) 02000 { 02001 unsigned ch = hole << 1, r = ch + 1; 02002 if( r <= i && A[ch]->dist < A[r]->dist ) ch = r; 02003 if( ch > i || item->dist >= A[ch]->dist ) break; 02004 A[hole] = A[ch]; 02005 hole = ch; 02006 } 02007 A[hole] = item; 02008 } 02009 } 02010 02011 findpt_data_2* findpt_setup_2( const realType* const xw[2], 02012 const unsigned n[2], 02013 uint nel, 02014 uint max_hash_size, 02015 realType bbox_tol ) 02016 { 02017 unsigned d; 02018 findpt_data_2* p = tmalloc( findpt_data_2, 1 ); 02019 02020 p->hash = tmalloc( hash_data_2, 1 ); 02021 p->od = tmalloc( opt_data_2, 1 ); 02022 02023 for( d = 0; d < 2; ++d ) 02024 p->xw[d] = xw[d]; 02025 p->nptel = n[0] * n[1]; 02026 02027 hash_build_2( p->hash, xw, n, nel, max_hash_size, bbox_tol ); 02028 02029 for( d = 0; d < 2; ++d ) 02030 { 02031 p->z[d] = tmalloc( realType, n[d] ); 02032 lobatto_nodes( p->z[d], n[d] ); 02033 lagrange_setup( &p->ld[d], p->z[d], n[d] ); 02034 } 02035 02036 p->list = tmalloc( findpt_listel, p->hash->max ); 02037 p->sorted = tmalloc( findpt_listel*, p->hash->max ); 02038 02039 opt_alloc_2( p->od, p->ld ); 02040 p->od_work = p->od->work; 02041 02042 return p; 02043 } 02044 02045 findpt_data_3* findpt_setup_3( const realType* const xw[3], 02046 const unsigned n[3], 02047 uint nel, 02048 uint max_hash_size, 02049 realType bbox_tol ) 02050 { 02051 unsigned d; 02052 findpt_data_3* p = tmalloc( findpt_data_3, 1 ); 02053 02054 p->hash = tmalloc( hash_data_3, 1 ); 02055 p->od = tmalloc( opt_data_3, 1 ); 02056 02057 for( d = 0; d < 3; ++d ) 02058 p->xw[d] = xw[d]; 02059 p->nptel = n[0] * n[1] * n[2]; 02060 02061 hash_build_3( p->hash, xw, n, nel, max_hash_size, bbox_tol ); 02062 02063 for( d = 0; d < 3; ++d ) 02064 { 02065 p->z[d] = tmalloc( realType, n[d] ); 02066 lobatto_nodes( p->z[d], n[d] ); 02067 lagrange_setup( &p->ld[d], p->z[d], n[d] ); 02068 } 02069 02070 p->list = tmalloc( findpt_listel, p->hash->max ); 02071 p->sorted = tmalloc( findpt_listel*, p->hash->max ); 02072 02073 opt_alloc_3( p->od, p->ld ); 02074 p->od_work = p->od->work; 02075 02076 return p; 02077 } 02078 02079 void findpt_free_2( findpt_data_2* p ) 02080 { 02081 unsigned d; 02082 opt_free_2( p->od ); 02083 free( p->od ); 02084 hash_free_2( p->hash ); 02085 free( p->hash ); 02086 free( p->list ); 02087 free( p->sorted ); 02088 for( d = 0; d < 2; ++d ) 02089 free( p->z[d] ); 02090 free( p ); 02091 } 02092 02093 void findpt_free_3( findpt_data_3* p ) 02094 { 02095 unsigned d; 02096 opt_free_3( p->od ); 02097 free( p->od ); 02098 hash_free_3( p->hash ); 02099 free( p->hash ); 02100 free( p->list ); 02101 free( p->sorted ); 02102 for( d = 0; d < 3; ++d ) 02103 free( p->z[d] ); 02104 free( p ); 02105 } 02106 02107 const realType* findpt_allbnd_2( const findpt_data_2* p ) 02108 { 02109 return p->hash->bnd; 02110 } 02111 02112 const realType* findpt_allbnd_3( const findpt_data_3* p ) 02113 { 02114 return p->hash->bnd; 02115 } 02116 02117 static void findpt_hash_2( findpt_data_2* p, const realType x[2] ) 02118 { 02119 findpt_listel *list = p->list, **sorted = p->sorted; 02120 const uint hi = hash_index_2( p->hash, x ); 02121 const uint* offset = p->hash->offset; 02122 uint i; 02123 const uint b = offset[hi], e = offset[hi + 1]; 02124 for( i = b; i != e; ++i ) 02125 { 02126 const uint el = offset[i]; 02127 realType* r = &list->r[0]; 02128 const obbox_2* obb = &p->hash->obb[el]; 02129 if( obbox_axis_test_2( obb, x ) ) continue; 02130 if( obbox_test_2( obb, x, r ) ) continue; 02131 list->el = el; 02132 list->dist = r1norm_2( r[0], r[1] ); 02133 *sorted++ = list++; 02134 } 02135 p->end = sorted; 02136 if( p->end != p->sorted ) findpt_list_sort( p->sorted, p->end - p->sorted ); 02137 } 02138 02139 static void findpt_hash_3( findpt_data_3* p, const realType x[3] ) 02140 { 02141 findpt_listel *list = p->list, **sorted = p->sorted; 02142 const uint hi = hash_index_3( p->hash, x ); 02143 const uint* offset = p->hash->offset; 02144 uint i; 02145 const uint b = offset[hi], e = offset[hi + 1]; 02146 for( i = b; i != e; ++i ) 02147 { 02148 const uint el = offset[i]; 02149 realType* r = &list->r[0]; 02150 const obbox_3* obb = &p->hash->obb[el]; 02151 if( obbox_axis_test_3( obb, x ) ) continue; 02152 if( obbox_test_3( obb, x, r ) ) continue; 02153 list->el = el; 02154 list->dist = r1norm_3( r[0], r[1], r[2] ); 02155 *sorted++ = list++; 02156 } 02157 p->end = sorted; 02158 if( p->end != p->sorted ) findpt_list_sort( p->sorted, p->end - p->sorted ); 02159 } 02160 02161 static int findpt_guess_2( findpt_data_2* p, const realType x[2], uint el, realType r[2], realType* dist ) 02162 { 02163 const uint arrindex = p->nptel * el; 02164 const realType* elx[2]; 02165 realType g[2]; 02166 unsigned c = opt_no_constraints_2; 02167 const obbox_2* obb = &p->hash->obb[el]; 02168 elx[0] = p->xw[0] + arrindex; 02169 elx[1] = p->xw[1] + arrindex; 02170 if( obbox_axis_test_2( obb, x ) || obbox_test_2( obb, x, g ) ) return 0; 02171 *dist = opt_findpt_2( p->od, elx, x, r, &c ); 02172 return c == opt_no_constraints_2; 02173 } 02174 02175 static int findpt_guess_3( findpt_data_3* p, const realType x[3], uint el, realType r[3], realType* dist ) 02176 { 02177 const uint arrindex = p->nptel * el; 02178 const realType* elx[3]; 02179 realType g[3]; 02180 unsigned c = opt_no_constraints_3; 02181 const obbox_3* obb = &p->hash->obb[el]; 02182 elx[0] = p->xw[0] + arrindex; 02183 elx[1] = p->xw[1] + arrindex; 02184 elx[2] = p->xw[2] + arrindex; 02185 if( obbox_axis_test_3( obb, x ) || obbox_test_3( obb, x, g ) ) return 0; 02186 *dist = opt_findpt_3( p->od, elx, x, r, &c ); 02187 return c == opt_no_constraints_3; 02188 } 02189 02190 #define DIAGNOSTICS 0 02191 02192 static int findpt_pass_2( findpt_data_2* p, const realType x[2], uint* el, realType r[2], realType* dist_min ) 02193 { 02194 findpt_listel** qq = p->sorted; 02195 const realType* bnd; 02196 realType dist; 02197 do 02198 { 02199 findpt_listel* q = *qq; 02200 const uint arrindex = p->nptel * q->el; 02201 const realType* elx[2]; 02202 unsigned c = opt_no_constraints_2; 02203 elx[0] = p->xw[0] + arrindex; 02204 elx[1] = p->xw[1] + arrindex; 02205 dist = opt_findpt_2( p->od, elx, x, q->r, &c ); 02206 if( qq == p->sorted || dist < *dist_min || c == opt_no_constraints_2 ) 02207 { 02208 *dist_min = dist; 02209 *el = q->el; 02210 memcpy( r, q->r, 2 * sizeof( realType ) ); 02211 if( c == opt_no_constraints_2 ) return 0; 02212 } 02213 } while( ++qq != p->end ); 02214 bnd = p->hash->obb[*el].axis_bnd; 02215 return *dist_min > r2norm_2( bnd[1] - bnd[0], bnd[3] - bnd[2] ) ? -1 : 1; 02216 } 02217 02218 static int findpt_pass_3( findpt_data_3* p, const realType x[3], uint* el, realType r[3], realType* dist_min ) 02219 { 02220 findpt_listel** qq = p->sorted; 02221 const realType* bnd; 02222 realType dist; 02223 do 02224 { 02225 findpt_listel* q = *qq; 02226 const uint arrindex = p->nptel * q->el; 02227 const realType* elx[3]; 02228 unsigned c = opt_no_constraints_3; 02229 elx[0] = p->xw[0] + arrindex; 02230 elx[1] = p->xw[1] + arrindex; 02231 elx[2] = p->xw[2] + arrindex; 02232 dist = opt_findpt_3( p->od, elx, x, q->r, &c ); 02233 if( qq == p->sorted || dist < *dist_min || c == opt_no_constraints_3 ) 02234 { 02235 *dist_min = dist; 02236 *el = q->el; 02237 memcpy( r, q->r, 3 * sizeof( realType ) ); 02238 if( c == opt_no_constraints_3 ) 02239 { 02240 #if DIAGNOSTICS 02241 printf( "point found in element #%d\n", qq - p->sorted ); 02242 #endif 02243 return 0; 02244 } 02245 } 02246 } while( ++qq != p->end ); 02247 bnd = p->hash->obb[*el].axis_bnd; 02248 return *dist_min > r2norm_3( bnd[1] - bnd[0], bnd[3] - bnd[2], bnd[5] - bnd[4] ) ? -1 : 1; 02249 } 02250 02251 int findpt_2( findpt_data_2* p, const realType x[2], int guess, uint* el, realType r[2], realType* dist ) 02252 { 02253 if( guess && findpt_guess_2( p, x, *el, r, dist ) ) return 0; 02254 findpt_hash_2( p, x ); 02255 if( p->sorted == p->end ) return -1; 02256 return findpt_pass_2( p, x, el, r, dist ); 02257 } 02258 02259 int findpt_3( findpt_data_3* p, const realType x[3], int guess, uint* el, realType r[3], realType* dist ) 02260 { 02261 if( guess && findpt_guess_3( p, x, *el, r, dist ) ) return 0; 02262 findpt_hash_3( p, x ); 02263 #if DIAGNOSTICS 02264 printf( "hashing leaves %d elements to consider\n", p->end - p->sorted ); 02265 #endif 02266 if( p->sorted == p->end ) return -1; 02267 return findpt_pass_3( p, x, el, r, dist ); 02268 } 02269 02270 void findpt_weights_2( findpt_data_2* p, const realType r[2] ) 02271 { 02272 lagrange_0( &p->ld[0], r[0] ); 02273 lagrange_0( &p->ld[1], r[1] ); 02274 } 02275 02276 void findpt_weights_3( findpt_data_3* p, const realType r[3] ) 02277 { 02278 lagrange_0( &p->ld[0], r[0] ); 02279 lagrange_0( &p->ld[1], r[1] ); 02280 lagrange_0( &p->ld[2], r[2] ); 02281 } 02282 02283 double findpt_eval_2( findpt_data_2* p, const realType* u ) 02284 { 02285 return tensor_i2( p->ld[0].J, p->ld[0].n, p->ld[1].J, p->ld[1].n, u, p->od_work ); 02286 } 02287 02288 double findpt_eval_3( findpt_data_3* p, const realType* u ) 02289 { 02290 return tensor_i3( p->ld[0].J, p->ld[0].n, p->ld[1].J, p->ld[1].n, p->ld[2].J, p->ld[2].n, u, p->od_work ); 02291 }