libopenjpeg/dwt.c

   1 /*
   2  * Copyright (c) 2002-2007, Communications and Remote Sensing Laboratory, Universite catholique de Louvain (UCL), Belgium
   3  * Copyright (c) 2002-2007, Professor Benoit Macq
   4  * Copyright (c) 2001-2003, David Janssens
   5  * Copyright (c) 2002-2003, Yannick Verschueren
   6  * Copyright (c) 2003-2007, Francois-Olivier Devaux and Antonin Descampe
   7  * Copyright (c) 2005, Herve Drolon, FreeImage Team
   8  * Copyright (c) 2007, Jonathan Ballard <dzonatas@dzonux.net>
   9  * Copyright (c) 2007, Callum Lerwick <seg@haxxed.com>
  10  * All rights reserved.
  11  *
  12  * Redistribution and use in source and binary forms, with or without
  13  * modification, are permitted provided that the following conditions
  14  * are met:
  15  * 1. Redistributions of source code must retain the above copyright
  16  *    notice, this list of conditions and the following disclaimer.
  17  * 2. Redistributions in binary form must reproduce the above copyright
  18  *    notice, this list of conditions and the following disclaimer in the
  19  *    documentation and/or other materials provided with the distribution.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS `AS IS'
  22  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  25  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  26  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  27  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  28  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  29  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  30  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  31  * POSSIBILITY OF SUCH DAMAGE.
  32  */
  33
  34 #ifdef __SSE__
  35 #include <xmmintrin.h>
  36 #endif
  37
  38 #include "opj_includes.h"
  39
  40 /** @defgroup DWT DWT - Implementation of a discrete wavelet transform */
  41 /*@{*/
  42
  43 #define WS(i) v->mem[(i)*2]
  44 #define WD(i) v->mem[(1+(i)*2)]
  45
  46 /** @name Local data structures */
  47 /*@{*/
  48
  49 typedef struct dwt_local {
  50         int* mem;
  51         int dn;
  52         int sn;
  53         int cas;
  54 } dwt_t;
  55
  56 typedef union {
  57         float   f[4];
  58 } v4;
  59
  60 typedef struct v4dwt_local {
  61         v4*     wavelet ;
  62         int             dn ;
  63         int             sn ;
  64         int             cas ;
  65 } v4dwt_t ;
  66
  67 static const float dwt_alpha =  1.586134342f; //  12994
  68 static const float dwt_beta  =  0.052980118f; //    434
  69 static const float dwt_gamma = -0.882911075f; //  -7233
  70 static const float dwt_delta = -0.443506852f; //  -3633
  71
  72 static const float K      = 1.230174105f; //  10078
  73 /* FIXME: What is this constant? */
  74 static const float c13318 = 1.625732422f;
  75
  76 /*@}*/
  77
  78 /**
  79 Virtual function type for wavelet transform in 1-D
  80 */
  81 typedef void (*DWT1DFN)(dwt_t* v);
  82
  83 /** @name Local static functions */
  84 /*@{*/
  85
  86 /**
  87 Forward lazy transform (horizontal)
  88 */
  89 static void dwt_deinterleave_h(int *a, int *b, int dn, int sn, int cas);
  90 /**
  91 Forward lazy transform (vertical)
  92 */
  93 static void dwt_deinterleave_v(int *a, int *b, int dn, int sn, int x, int cas);
  94 /**
  95 Inverse lazy transform (horizontal)
  96 */
  97 static void dwt_interleave_h(dwt_t* h, int *a);
  98 /**
  99 Inverse lazy transform (vertical)
 100 */
 101 static void dwt_interleave_v(dwt_t* v, int *a, int x);
 102 /**
 103 Forward 5-3 wavelet transform in 1-D
 104 */
 105 static void dwt_encode_1(int *a, int dn, int sn, int cas);
 106 /**
 107 Inverse 5-3 wavelet transform in 1-D
 108 */
 109 static void dwt_decode_1(dwt_t *v);
 110 /**
 111 Forward 9-7 wavelet transform in 1-D
 112 */
 113 static void dwt_encode_1_real(int *a, int dn, int sn, int cas);
 114 /**
 115 Explicit calculation of the Quantization Stepsizes
 116 */
 117 static void dwt_encode_stepsize(int stepsize, int numbps, opj_stepsize_t *bandno_stepsize);
 118 /**
 119 Inverse wavelet transform in 2-D.
 120 */
 121 #ifdef OPJ_V1
 122 static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int i, DWT1DFN fn);
 123 #endif
 124 static opj_bool dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i, DWT1DFN fn);
 125
 126 /**
 127 Inverse wavelet transform in 2-D.
 128 */
 129 static opj_bool dwt_decode_tile_v2(opj_tcd_tilecomp_v2_t* tilec, OPJ_UINT32 i, DWT1DFN fn);
 130
 131 /*@}*/
 132
 133 /*@}*/
 134
 135 #define S(i) a[(i)*2]
 136 #define D(i) a[(1+(i)*2)]
 137 #define S_(i) ((i)<0?S(0):((i)>=sn?S(sn-1):S(i)))
 138 #define D_(i) ((i)<0?D(0):((i)>=dn?D(dn-1):D(i)))
 139 /* new */
 140 #define SS_(i) ((i)<0?S(0):((i)>=dn?S(dn-1):S(i)))
 141 #define DD_(i) ((i)<0?D(0):((i)>=sn?D(sn-1):D(i)))
 142
 143 /* <summary>                                                              */
 144 /* This table contains the norms of the 5-3 wavelets for different bands. */
 145 /* </summary>                                                             */
 146 static const double dwt_norms[4][10] = {
 147         {1.000, 1.500, 2.750, 5.375, 10.68, 21.34, 42.67, 85.33, 170.7, 341.3},
 148         {1.038, 1.592, 2.919, 5.703, 11.33, 22.64, 45.25, 90.48, 180.9},
 149         {1.038, 1.592, 2.919, 5.703, 11.33, 22.64, 45.25, 90.48, 180.9},
 150         {.7186, .9218, 1.586, 3.043, 6.019, 12.01, 24.00, 47.97, 95.93}
 151 };
 152
 153 /* <summary>                                                              */
 154 /* This table contains the norms of the 9-7 wavelets for different bands. */
 155 /* </summary>                                                             */
 156 static const double dwt_norms_real[4][10] = {
 157         {1.000, 1.965, 4.177, 8.403, 16.90, 33.84, 67.69, 135.3, 270.6, 540.9},
 158         {2.022, 3.989, 8.355, 17.04, 34.27, 68.63, 137.3, 274.6, 549.0},
 159         {2.022, 3.989, 8.355, 17.04, 34.27, 68.63, 137.3, 274.6, 549.0},
 160         {2.080, 3.865, 8.307, 17.18, 34.71, 69.59, 139.3, 278.6, 557.2}
 161 };
 162
 163 /*
 164 ==========================================================
 165    local functions
 166 ==========================================================
 167 */
 168
 169 /* <summary>                                     */
 170 /* Forward lazy transform (horizontal).  */
 171 /* </summary>                            */
 172 static void dwt_deinterleave_h(int *a, int *b, int dn, int sn, int cas) {
 173         int i;
 174     for (i=0; i<sn; i++) b[i]=a[2*i+cas];
 175     for (i=0; i<dn; i++) b[sn+i]=a[(2*i+1-cas)];
 176 }
 177
 178 /* <summary>                             */
 179 /* Forward lazy transform (vertical).    */
 180 /* </summary>                            */
 181 static void dwt_deinterleave_v(int *a, int *b, int dn, int sn, int x, int cas) {
 182     int i;
 183     for (i=0; i<sn; i++) b[i*x]=a[2*i+cas];
 184     for (i=0; i<dn; i++) b[(sn+i)*x]=a[(2*i+1-cas)];
 185 }
 186
 187 /* <summary>                             */
 188 /* Inverse lazy transform (horizontal).  */
 189 /* </summary>                            */
 190 static void dwt_interleave_h(dwt_t* h, int *a) {
 191     int *ai = a;
 192     int *bi = h->mem + h->cas;
 193     int  i      = h->sn;
 194     while( i-- ) {
 195       *bi = *(ai++);
 196           bi += 2;
 197     }
 198     ai  = a + h->sn;
 199     bi  = h->mem + 1 - h->cas;
 200     i   = h->dn ;
 201     while( i-- ) {
 202       *bi = *(ai++);
 203           bi += 2;
 204     }
 205 }
 206
 207 /* <summary>                             */
 208 /* Inverse lazy transform (vertical).    */
 209 /* </summary>                            */
 210 static void dwt_interleave_v(dwt_t* v, int *a, int x) {
 211     int *ai = a;
 212     int *bi = v->mem + v->cas;
 213     int  i = v->sn;
 214     while( i-- ) {
 215       *bi = *ai;
 216           bi += 2;
 217           ai += x;
 218     }
 219     ai = a + (v->sn * x);
 220     bi = v->mem + 1 - v->cas;
 221     i = v->dn ;
 222     while( i-- ) {
 223       *bi = *ai;
 224           bi += 2;
 225           ai += x;
 226     }
 227 }
 228
 229
 230 /* <summary>                            */
 231 /* Forward 5-3 wavelet transform in 1-D. */
 232 /* </summary>                           */
 233 static void dwt_encode_1(int *a, int dn, int sn, int cas) {
 234         int i;
 235
 236         if (!cas) {
 237                 if ((dn > 0) || (sn > 1)) {     /* NEW :  CASE ONE ELEMENT */
 238                         for (i = 0; i < dn; i++) D(i) -= (S_(i) + S_(i + 1)) >> 1;
 239                         for (i = 0; i < sn; i++) S(i) += (D_(i - 1) + D_(i) + 2) >> 2;
 240                 }
 241         } else {
 242                 if (!sn && dn == 1)                 /* NEW :  CASE ONE ELEMENT */
 243                         S(0) *= 2;
 244                 else {
 245                         for (i = 0; i < dn; i++) S(i) -= (DD_(i) + DD_(i - 1)) >> 1;
 246                         for (i = 0; i < sn; i++) D(i) += (SS_(i) + SS_(i + 1) + 2) >> 2;
 247                 }
 248         }
 249 }
 250
 251 /* <summary>                            */
 252 /* Inverse 5-3 wavelet transform in 1-D. */
 253 /* </summary>                           */
 254 static void dwt_decode_1_(int *a, int dn, int sn, int cas) {
 255         int i;
 256
 257         if (!cas) {
 258                 if ((dn > 0) || (sn > 1)) { /* NEW :  CASE ONE ELEMENT */
 259                         for (i = 0; i < sn; i++) S(i) -= (D_(i - 1) + D_(i) + 2) >> 2;
 260                         for (i = 0; i < dn; i++) D(i) += (S_(i) + S_(i + 1)) >> 1;
 261                 }
 262         } else {
 263                 if (!sn  && dn == 1)          /* NEW :  CASE ONE ELEMENT */
 264                         S(0) /= 2;
 265                 else {
 266                         for (i = 0; i < sn; i++) D(i) -= (SS_(i) + SS_(i + 1) + 2) >> 2;
 267                         for (i = 0; i < dn; i++) S(i) += (DD_(i) + DD_(i - 1)) >> 1;
 268                 }
 269         }
 270 }
 271
 272 /* <summary>                            */
 273 /* Inverse 5-3 wavelet transform in 1-D. */
 274 /* </summary>                           */
 275 static void dwt_decode_1(dwt_t *v) {
 276         dwt_decode_1_(v->mem, v->dn, v->sn, v->cas);
 277 }
 278
 279 /* <summary>                             */
 280 /* Forward 9-7 wavelet transform in 1-D. */
 281 /* </summary>                            */
 282 static void dwt_encode_1_real(int *a, int dn, int sn, int cas) {
 283         int i;
 284         if (!cas) {
 285                 if ((dn > 0) || (sn > 1)) {     /* NEW :  CASE ONE ELEMENT */
 286                         for (i = 0; i < dn; i++)
 287                                 D(i) -= fix_mul(S_(i) + S_(i + 1), 12993);
 288                         for (i = 0; i < sn; i++)
 289                                 S(i) -= fix_mul(D_(i - 1) + D_(i), 434);
 290                         for (i = 0; i < dn; i++)
 291                                 D(i) += fix_mul(S_(i) + S_(i + 1), 7233);
 292                         for (i = 0; i < sn; i++)
 293                                 S(i) += fix_mul(D_(i - 1) + D_(i), 3633);
 294                         for (i = 0; i < dn; i++)
 295                                 D(i) = fix_mul(D(i), 5038);     /*5038 */
 296                         for (i = 0; i < sn; i++)
 297                                 S(i) = fix_mul(S(i), 6659);     /*6660 */
 298                 }
 299         } else {
 300                 if ((sn > 0) || (dn > 1)) {     /* NEW :  CASE ONE ELEMENT */
 301                         for (i = 0; i < dn; i++)
 302                                 S(i) -= fix_mul(DD_(i) + DD_(i - 1), 12993);
 303                         for (i = 0; i < sn; i++)
 304                                 D(i) -= fix_mul(SS_(i) + SS_(i + 1), 434);
 305                         for (i = 0; i < dn; i++)
 306                                 S(i) += fix_mul(DD_(i) + DD_(i - 1), 7233);
 307                         for (i = 0; i < sn; i++)
 308                                 D(i) += fix_mul(SS_(i) + SS_(i + 1), 3633);
 309                         for (i = 0; i < dn; i++)
 310                                 S(i) = fix_mul(S(i), 5038);     /*5038 */
 311                         for (i = 0; i < sn; i++)
 312                                 D(i) = fix_mul(D(i), 6659);     /*6660 */
 313                 }
 314         }
 315 }
 316
 317 static void dwt_encode_stepsize(int stepsize, int numbps, opj_stepsize_t *bandno_stepsize) {
 318         int p, n;
 319         p = int_floorlog2(stepsize) - 13;
 320         n = 11 - int_floorlog2(stepsize);
 321         bandno_stepsize->mant = (n < 0 ? stepsize >> -n : stepsize << n) & 0x7ff;
 322         bandno_stepsize->expn = numbps - p;
 323 }
 324
 325 /*
 326 ==========================================================
 327    DWT interface
 328 ==========================================================
 329 */
 330
 331 /* <summary>                            */
 332 /* Forward 5-3 wavelet transform in 2-D. */
 333 /* </summary>                           */
 334 void dwt_encode(opj_tcd_tilecomp_t * tilec) {
 335         int i, j, k;
 336         int *a = NULL;
 337         int *aj = NULL;
 338         int *bj = NULL;
 339         int w, l;
 340
 341         w = tilec->x1-tilec->x0;
 342         l = tilec->numresolutions-1;
 343         a = tilec->data;
 344
 345         for (i = 0; i < l; i++) {
 346                 int rw;                 /* width of the resolution level computed                                                           */
 347                 int rh;                 /* height of the resolution level computed                                                          */
 348                 int rw1;                /* width of the resolution level once lower than computed one                                       */
 349                 int rh1;                /* height of the resolution level once lower than computed one                                      */
 350                 int cas_col;    /* 0 = non inversion on horizontal filtering 1 = inversion between low-pass and high-pass filtering */
 351                 int cas_row;    /* 0 = non inversion on vertical filtering 1 = inversion between low-pass and high-pass filtering   */
 352                 int dn, sn;
 353
 354                 rw = tilec->resolutions[l - i].x1 - tilec->resolutions[l - i].x0;
 355                 rh = tilec->resolutions[l - i].y1 - tilec->resolutions[l - i].y0;
 356                 rw1= tilec->resolutions[l - i - 1].x1 - tilec->resolutions[l - i - 1].x0;
 357                 rh1= tilec->resolutions[l - i - 1].y1 - tilec->resolutions[l - i - 1].y0;
 358
 359                 cas_row = tilec->resolutions[l - i].x0 % 2;
 360                 cas_col = tilec->resolutions[l - i].y0 % 2;
 361
 362                 sn = rh1;
 363                 dn = rh - rh1;
 364                 bj = (int*)opj_malloc(rh * sizeof(int));
 365                 for (j = 0; j < rw; j++) {
 366                         aj = a + j;
 367                         for (k = 0; k < rh; k++)  bj[k] = aj[k*w];
 368                         dwt_encode_1(bj, dn, sn, cas_col);
 369                         dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col);
 370                 }
 371                 opj_free(bj);
 372
 373                 sn = rw1;
 374                 dn = rw - rw1;
 375                 bj = (int*)opj_malloc(rw * sizeof(int));
 376                 for (j = 0; j < rh; j++) {
 377                         aj = a + j * w;
 378                         for (k = 0; k < rw; k++)  bj[k] = aj[k];
 379                         dwt_encode_1(bj, dn, sn, cas_row);
 380                         dwt_deinterleave_h(bj, aj, dn, sn, cas_row);
 381                 }
 382                 opj_free(bj);
 383         }
 384 }
 385
 386 #ifdef OPJ_V1
 387 /* <summary>                            */
 388 /* Inverse 5-3 wavelet transform in 2-D. */
 389 /* </summary>                           */
 390 void dwt_decode(opj_tcd_tilecomp_t* tilec, int numres) {
 391         dwt_decode_tile(tilec, numres, &dwt_decode_1);
 392 }
 393 #endif
 394
 395 /* <summary>                            */
 396 /* Inverse 5-3 wavelet transform in 2-D. */
 397 /* </summary>                           */
 398 opj_bool dwt_decode(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres) {
 399         return dwt_decode_tile(tilec, numres, &dwt_decode_1);
 400 }
 401
 402 /* <summary>                            */
 403 /* Inverse 5-3 wavelet transform in 2-D. */
 404 /* </summary>                           */
 405 opj_bool dwt_decode_v2(opj_tcd_tilecomp_v2_t* tilec, OPJ_UINT32 numres) {
 406         return dwt_decode_tile_v2(tilec, numres, &dwt_decode_1);
 407 }
 408
 409
 410 /* <summary>                          */
 411 /* Get gain of 5-3 wavelet transform. */
 412 /* </summary>                         */
 413 int dwt_getgain(int orient) {
 414         if (orient == 0)
 415                 return 0;
 416         if (orient == 1 || orient == 2)
 417                 return 1;
 418         return 2;
 419 }
 420
 421 /* <summary>                          */
 422 /* Get gain of 5-3 wavelet transform. */
 423 /* </summary>                         */
 424 OPJ_UINT32 dwt_getgain_v2(OPJ_UINT32 orient) {
 425         if (orient == 0)
 426                 return 0;
 427         if (orient == 1 || orient == 2)
 428                 return 1;
 429         return 2;
 430 }
 431
 432 /* <summary>                */
 433 /* Get norm of 5-3 wavelet. */
 434 /* </summary>               */
 435 double dwt_getnorm(int level, int orient) {
 436         return dwt_norms[orient][level];
 437 }
 438
 439 /* <summary>                             */
 440 /* Forward 9-7 wavelet transform in 2-D. */
 441 /* </summary>                            */
 442
 443 void dwt_encode_real(opj_tcd_tilecomp_t * tilec) {
 444         int i, j, k;
 445         int *a = NULL;
 446         int *aj = NULL;
 447         int *bj = NULL;
 448         int w, l;
 449
 450         w = tilec->x1-tilec->x0;
 451         l = tilec->numresolutions-1;
 452         a = tilec->data;
 453
 454         for (i = 0; i < l; i++) {
 455                 int rw;                 /* width of the resolution level computed                                                     */
 456                 int rh;                 /* height of the resolution level computed                                                    */
 457                 int rw1;                /* width of the resolution level once lower than computed one                                 */
 458                 int rh1;                /* height of the resolution level once lower than computed one                                */
 459                 int cas_col;    /* 0 = non inversion on horizontal filtering 1 = inversion between low-pass and high-pass filtering */
 460                 int cas_row;    /* 0 = non inversion on vertical filtering 1 = inversion between low-pass and high-pass filtering   */
 461                 int dn, sn;
 462
 463                 rw = tilec->resolutions[l - i].x1 - tilec->resolutions[l - i].x0;
 464                 rh = tilec->resolutions[l - i].y1 - tilec->resolutions[l - i].y0;
 465                 rw1= tilec->resolutions[l - i - 1].x1 - tilec->resolutions[l - i - 1].x0;
 466                 rh1= tilec->resolutions[l - i - 1].y1 - tilec->resolutions[l - i - 1].y0;
 467
 468                 cas_row = tilec->resolutions[l - i].x0 % 2;
 469                 cas_col = tilec->resolutions[l - i].y0 % 2;
 470
 471                 sn = rh1;
 472                 dn = rh - rh1;
 473                 bj = (int*)opj_malloc(rh * sizeof(int));
 474                 for (j = 0; j < rw; j++) {
 475                         aj = a + j;
 476                         for (k = 0; k < rh; k++)  bj[k] = aj[k*w];
 477                         dwt_encode_1_real(bj, dn, sn, cas_col);
 478                         dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col);
 479                 }
 480                 opj_free(bj);
 481
 482                 sn = rw1;
 483                 dn = rw - rw1;
 484                 bj = (int*)opj_malloc(rw * sizeof(int));
 485                 for (j = 0; j < rh; j++) {
 486                         aj = a + j * w;
 487                         for (k = 0; k < rw; k++)  bj[k] = aj[k];
 488                         dwt_encode_1_real(bj, dn, sn, cas_row);
 489                         dwt_deinterleave_h(bj, aj, dn, sn, cas_row);
 490                 }
 491                 opj_free(bj);
 492         }
 493 }
 494
 495
 496 /* <summary>                          */
 497 /* Get gain of 9-7 wavelet transform. */
 498 /* </summary>                         */
 499 int dwt_getgain_real(int orient) {
 500         (void)orient;
 501         return 0;
 502 }
 503
 504 /* <summary>                          */
 505 /* Get gain of 9-7 wavelet transform. */
 506 /* </summary>                         */
 507 OPJ_UINT32 dwt_getgain_real_v2(OPJ_UINT32 orient) {
 508         (void)orient;
 509         return 0;
 510 }
 511
 512 /* <summary>                */
 513 /* Get norm of 9-7 wavelet. */
 514 /* </summary>               */
 515 double dwt_getnorm_real(int level, int orient) {
 516         return dwt_norms_real[orient][level];
 517 }
 518
 519 void dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, int prec) {
 520         int numbands, bandno;
 521         numbands = 3 * tccp->numresolutions - 2;
 522         for (bandno = 0; bandno < numbands; bandno++) {
 523                 double stepsize;
 524                 int resno, level, orient, gain;
 525
 526                 resno = (bandno == 0) ? 0 : ((bandno - 1) / 3 + 1);
 527                 orient = (bandno == 0) ? 0 : ((bandno - 1) % 3 + 1);
 528                 level = tccp->numresolutions - 1 - resno;
 529                 gain = (tccp->qmfbid == 0) ? 0 : ((orient == 0) ? 0 : (((orient == 1) || (orient == 2)) ? 1 : 2));
 530                 if (tccp->qntsty == J2K_CCP_QNTSTY_NOQNT) {
 531                         stepsize = 1.0;
 532                 } else {
 533                         double norm = dwt_norms_real[orient][level];
 534                         stepsize = (1 << (gain)) / norm;
 535                 }
 536                 dwt_encode_stepsize((int) floor(stepsize * 8192.0), prec + gain, &tccp->stepsizes[bandno]);
 537         }
 538 }
 539
 540 #ifdef OPJ_V1
 541 /* <summary>                             */
 542 /* Determine maximum computed resolution level for inverse wavelet transform */
 543 /* </summary>                            */
 544 static int dwt_decode_max_resolution(opj_tcd_resolution_t* restrict r, int i) {
 545         int mr  = 1;
 546         int w;
 547         while( --i ) {
 548                 r++;
 549                 if( mr < ( w = r->x1 - r->x0 ) )
 550                         mr = w ;
 551                 if( mr < ( w = r->y1 - r->y0 ) )
 552                         mr = w ;
 553         }
 554         return mr ;
 555 }
 556 #endif
 557 /* <summary>                             */
 558 /* Determine maximum computed resolution level for inverse wavelet transform */
 559 /* </summary>                            */
 560 static OPJ_UINT32 dwt_max_resolution(opj_tcd_resolution_t* restrict r, OPJ_UINT32 i) {
 561         OPJ_UINT32 mr   = 0;
 562         OPJ_UINT32 w;
 563         while( --i ) {
 564                 ++r;
 565                 if( mr < ( w = r->x1 - r->x0 ) )
 566                         mr = w ;
 567                 if( mr < ( w = r->y1 - r->y0 ) )
 568                         mr = w ;
 569         }
 570         return mr ;
 571 }
 572
 573 /* <summary>                             */
 574 /* Determine maximum computed resolution level for inverse wavelet transform */
 575 /* </summary>                            */
 576 static OPJ_UINT32 dwt_max_resolution_v2(opj_tcd_resolution_v2_t* restrict r, OPJ_UINT32 i) {
 577         OPJ_UINT32 mr   = 0;
 578         OPJ_UINT32 w;
 579         while( --i ) {
 580                 ++r;
 581                 if( mr < ( w = r->x1 - r->x0 ) )
 582                         mr = w ;
 583                 if( mr < ( w = r->y1 - r->y0 ) )
 584                         mr = w ;
 585         }
 586         return mr ;
 587 }
 588
 589 #ifdef OPJ_V1
 590 /* <summary>                            */
 591 /* Inverse wavelet transform in 2-D.     */
 592 /* </summary>                           */
 593 static void dwt_decode_tile(opj_tcd_tilecomp_t* tilec, int numres, DWT1DFN dwt_1D) {
 594         dwt_t h;
 595         dwt_t v;
 596
 597         opj_tcd_resolution_t* tr = tilec->resolutions;
 598
 599         int rw = tr->x1 - tr->x0;       /* width of the resolution level computed */
 600         int rh = tr->y1 - tr->y0;       /* height of the resolution level computed */
 601
 602         int w = tilec->x1 - tilec->x0;
 603
 604         h.mem = (int*)opj_aligned_malloc(dwt_max_resolution(tr, numres) * sizeof(int));
 605         v.mem = h.mem;
 606
 607         while( --numres) {
 608                 int * restrict tiledp = tilec->data;
 609                 int j;
 610
 611                 ++tr;
 612                 h.sn = rw;
 613                 v.sn = rh;
 614
 615                 rw = tr->x1 - tr->x0;
 616                 rh = tr->y1 - tr->y0;
 617
 618                 h.dn = rw - h.sn;
 619                 h.cas = tr->x0 % 2;
 620
 621                 for(j = 0; j < rh; ++j) {
 622                         dwt_interleave_h(&h, &tiledp[j*w]);
 623                         (dwt_1D)(&h);
 624                         memcpy(&tiledp[j*w], h.mem, rw * sizeof(int));
 625                 }
 626
 627                 v.dn = rh - v.sn;
 628                 v.cas = tr->y0 % 2;
 629
 630                 for(j = 0; j < rw; ++j){
 631                         int k;
 632                         dwt_interleave_v(&v, &tiledp[j], w);
 633                         (dwt_1D)(&v);
 634                         for(k = 0; k < rh; ++k) {
 635                                 tiledp[k * w + j] = v.mem[k];
 636                         }
 637                 }
 638         }
 639         opj_aligned_free(h.mem);
 640 }
 641 #endif
 642
 643 /* <summary>                            */
 644 /* Inverse wavelet transform in 2-D.     */
 645 /* </summary>                           */
 646 static opj_bool dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres, DWT1DFN dwt_1D) {
 647         dwt_t h;
 648         dwt_t v;
 649
 650         opj_tcd_resolution_t* tr = tilec->resolutions;
 651
 652         OPJ_UINT32 rw = tr->x1 - tr->x0;        /* width of the resolution level computed */
 653         OPJ_UINT32 rh = tr->y1 - tr->y0;        /* height of the resolution level computed */
 654
 655         OPJ_UINT32 w = tilec->x1 - tilec->x0;
 656
 657         h.mem = (OPJ_INT32*)
 658         opj_aligned_malloc(dwt_max_resolution(tr, numres) * sizeof(OPJ_INT32));
 659         if
 660                 (! h.mem)
 661         {
 662                 return OPJ_FALSE;
 663         }
 664
 665         v.mem = h.mem;
 666
 667         while( --numres) {
 668                 OPJ_INT32 * restrict tiledp = tilec->data;
 669                 OPJ_UINT32 j;
 670
 671                 ++tr;
 672                 h.sn = rw;
 673                 v.sn = rh;
 674
 675                 rw = tr->x1 - tr->x0;
 676                 rh = tr->y1 - tr->y0;
 677
 678                 h.dn = rw - h.sn;
 679                 h.cas = tr->x0 % 2;
 680
 681                 for(j = 0; j < rh; ++j) {
 682                         dwt_interleave_h(&h, &tiledp[j*w]);
 683                         (dwt_1D)(&h);
 684                         memcpy(&tiledp[j*w], h.mem, rw * sizeof(OPJ_INT32));
 685                 }
 686
 687                 v.dn = rh - v.sn;
 688                 v.cas = tr->y0 % 2;
 689
 690                 for(j = 0; j < rw; ++j){
 691                         OPJ_UINT32 k;
 692                         dwt_interleave_v(&v, &tiledp[j], w);
 693                         (dwt_1D)(&v);
 694                         for(k = 0; k < rh; ++k) {
 695                                 tiledp[k * w + j] = v.mem[k];
 696                         }
 697                 }
 698         }
 699         opj_aligned_free(h.mem);
 700         return OPJ_TRUE;
 701 }
 702
 703 /* <summary>                            */
 704 /* Inverse wavelet transform in 2-D.     */
 705 /* </summary>                           */
 706 static opj_bool dwt_decode_tile_v2(opj_tcd_tilecomp_v2_t* tilec, OPJ_UINT32 numres, DWT1DFN dwt_1D) {
 707         dwt_t h;
 708         dwt_t v;
 709
 710         opj_tcd_resolution_v2_t* tr = tilec->resolutions;
 711
 712         OPJ_UINT32 rw = tr->x1 - tr->x0;        /* width of the resolution level computed */
 713         OPJ_UINT32 rh = tr->y1 - tr->y0;        /* height of the resolution level computed */
 714
 715         OPJ_UINT32 w = tilec->x1 - tilec->x0;
 716
 717         h.mem = (OPJ_INT32*)
 718         opj_aligned_malloc(dwt_max_resolution_v2(tr, numres) * sizeof(OPJ_INT32));
 719         if
 720                 (! h.mem)
 721         {
 722                 return OPJ_FALSE;
 723         }
 724
 725         v.mem = h.mem;
 726
 727         while( --numres) {
 728                 OPJ_INT32 * restrict tiledp = tilec->data;
 729                 OPJ_UINT32 j;
 730
 731                 ++tr;
 732                 h.sn = rw;
 733                 v.sn = rh;
 734
 735                 rw = tr->x1 - tr->x0;
 736                 rh = tr->y1 - tr->y0;
 737
 738                 h.dn = rw - h.sn;
 739                 h.cas = tr->x0 % 2;
 740
 741                 for(j = 0; j < rh; ++j) {
 742                         dwt_interleave_h(&h, &tiledp[j*w]);
 743                         (dwt_1D)(&h);
 744                         memcpy(&tiledp[j*w], h.mem, rw * sizeof(OPJ_INT32));
 745                 }
 746
 747                 v.dn = rh - v.sn;
 748                 v.cas = tr->y0 % 2;
 749
 750                 for(j = 0; j < rw; ++j){
 751                         OPJ_UINT32 k;
 752                         dwt_interleave_v(&v, &tiledp[j], w);
 753                         (dwt_1D)(&v);
 754                         for(k = 0; k < rh; ++k) {
 755                                 tiledp[k * w + j] = v.mem[k];
 756                         }
 757                 }
 758         }
 759         opj_aligned_free(h.mem);
 760         return OPJ_TRUE;
 761 }
 762
 763 static void v4dwt_interleave_h(v4dwt_t* restrict w, float* restrict a, int x, int size){
 764         float* restrict bi = (float*) (w->wavelet + w->cas);
 765         int count = w->sn;
 766         int i, k;
 767
 768         for(k = 0; k < 2; ++k){
 769                 if (count + 3 * x < size && ((size_t) a & 0x0f) == 0 &&
 770                                 ((size_t) bi & 0x0f) == 0 && (x & 0x0f) == 0) {
 771                         /* Fast code path */
 772                         for(i = 0; i < count; ++i){
 773                                 int j = i;
 774                                 bi[i*8    ] = a[j];
 775                                 j += x;
 776                                 bi[i*8 + 1] = a[j];
 777                                 j += x;
 778                                 bi[i*8 + 2] = a[j];
 779                                 j += x;
 780                                 bi[i*8 + 3] = a[j];
 781                         }
 782                 }
 783                 else {
 784                         /* Slow code path */
 785                         for(i = 0; i < count; ++i){
 786                                 int j = i;
 787                                 bi[i*8    ] = a[j];
 788                                 j += x;
 789                                 if(j > size) continue;
 790                                 bi[i*8 + 1] = a[j];
 791                                 j += x;
 792                                 if(j > size) continue;
 793                                 bi[i*8 + 2] = a[j];
 794                                 j += x;
 795                                 if(j > size) continue;
 796                                 bi[i*8 + 3] = a[j];
 797                         }
 798                 }
 799
 800                 bi = (float*) (w->wavelet + 1 - w->cas);
 801                 a += w->sn;
 802                 size -= w->sn;
 803                 count = w->dn;
 804         }
 805 }
 806
 807 static void v4dwt_interleave_v(v4dwt_t* restrict v , float* restrict a , int x){
 808         v4* restrict bi = v->wavelet + v->cas;
 809         int i;
 810         for(i = 0; i < v->sn; ++i){
 811                 memcpy(&bi[i*2], &a[i*x], 4 * sizeof(float));
 812         }
 813         a += v->sn * x;
 814         bi = v->wavelet + 1 - v->cas;
 815         for(i = 0; i < v->dn; ++i){
 816                 memcpy(&bi[i*2], &a[i*x], 4 * sizeof(float));
 817         }
 818 }
 819
 820 #ifdef __SSE__
 821
 822 static void v4dwt_decode_step1_sse(v4* w, int count, const __m128 c){
 823         __m128* restrict vw = (__m128*) w;
 824         int i;
 825         /* 4x unrolled loop */
 826         for(i = 0; i < count >> 2; ++i){
 827                 *vw = _mm_mul_ps(*vw, c);
 828                 vw += 2;
 829                 *vw = _mm_mul_ps(*vw, c);
 830                 vw += 2;
 831                 *vw = _mm_mul_ps(*vw, c);
 832                 vw += 2;
 833                 *vw = _mm_mul_ps(*vw, c);
 834                 vw += 2;
 835         }
 836         count &= 3;
 837         for(i = 0; i < count; ++i){
 838                 *vw = _mm_mul_ps(*vw, c);
 839                 vw += 2;
 840         }
 841 }
 842
 843 static void v4dwt_decode_step2_sse(v4* l, v4* w, int k, int m, __m128 c){
 844         __m128* restrict vl = (__m128*) l;
 845         __m128* restrict vw = (__m128*) w;
 846         int i;
 847         __m128 tmp1, tmp2, tmp3;
 848         tmp1 = vl[0];
 849         for(i = 0; i < m; ++i){
 850                 tmp2 = vw[-1];
 851                 tmp3 = vw[ 0];
 852                 vw[-1] = _mm_add_ps(tmp2, _mm_mul_ps(_mm_add_ps(tmp1, tmp3), c));
 853                 tmp1 = tmp3;
 854                 vw += 2;
 855         }
 856         vl = vw - 2;
 857         if(m >= k){
 858                 return;
 859         }
 860         c = _mm_add_ps(c, c);
 861         c = _mm_mul_ps(c, vl[0]);
 862         for(; m < k; ++m){
 863                 __m128 tmp = vw[-1];
 864                 vw[-1] = _mm_add_ps(tmp, c);
 865                 vw += 2;
 866         }
 867 }
 868
 869 #else
 870
 871 static void v4dwt_decode_step1(v4* w, int count, const float c){
 872         float* restrict fw = (float*) w;
 873         int i;
 874         for(i = 0; i < count; ++i){
 875                 float tmp1 = fw[i*8    ];
 876                 float tmp2 = fw[i*8 + 1];
 877                 float tmp3 = fw[i*8 + 2];
 878                 float tmp4 = fw[i*8 + 3];
 879                 fw[i*8    ] = tmp1 * c;
 880                 fw[i*8 + 1] = tmp2 * c;
 881                 fw[i*8 + 2] = tmp3 * c;
 882                 fw[i*8 + 3] = tmp4 * c;
 883         }
 884 }
 885
 886 static void v4dwt_decode_step2(v4* l, v4* w, int k, int m, float c){
 887         float* restrict fl = (float*) l;
 888         float* restrict fw = (float*) w;
 889         int i;
 890         for(i = 0; i < m; ++i){
 891                 float tmp1_1 = fl[0];
 892                 float tmp1_2 = fl[1];
 893                 float tmp1_3 = fl[2];
 894                 float tmp1_4 = fl[3];
 895                 float tmp2_1 = fw[-4];
 896                 float tmp2_2 = fw[-3];
 897                 float tmp2_3 = fw[-2];
 898                 float tmp2_4 = fw[-1];
 899                 float tmp3_1 = fw[0];
 900                 float tmp3_2 = fw[1];
 901                 float tmp3_3 = fw[2];
 902                 float tmp3_4 = fw[3];
 903                 fw[-4] = tmp2_1 + ((tmp1_1 + tmp3_1) * c);
 904                 fw[-3] = tmp2_2 + ((tmp1_2 + tmp3_2) * c);
 905                 fw[-2] = tmp2_3 + ((tmp1_3 + tmp3_3) * c);
 906                 fw[-1] = tmp2_4 + ((tmp1_4 + tmp3_4) * c);
 907                 fl = fw;
 908                 fw += 8;
 909         }
 910         if(m < k){
 911                 float c1;
 912                 float c2;
 913                 float c3;
 914                 float c4;
 915                 c += c;
 916                 c1 = fl[0] * c;
 917                 c2 = fl[1] * c;
 918                 c3 = fl[2] * c;
 919                 c4 = fl[3] * c;
 920                 for(; m < k; ++m){
 921                         float tmp1 = fw[-4];
 922                         float tmp2 = fw[-3];
 923                         float tmp3 = fw[-2];
 924                         float tmp4 = fw[-1];
 925                         fw[-4] = tmp1 + c1;
 926                         fw[-3] = tmp2 + c2;
 927                         fw[-2] = tmp3 + c3;
 928                         fw[-1] = tmp4 + c4;
 929                         fw += 8;
 930                 }
 931         }
 932 }
 933
 934 #endif
 935
 936 /* <summary>                             */
 937 /* Inverse 9-7 wavelet transform in 1-D. */
 938 /* </summary>                            */
 939 static void v4dwt_decode(v4dwt_t* restrict dwt){
 940         int a, b;
 941         if(dwt->cas == 0) {
 942                 if(!((dwt->dn > 0) || (dwt->sn > 1))){
 943                         return;
 944                 }
 945                 a = 0;
 946                 b = 1;
 947         }else{
 948                 if(!((dwt->sn > 0) || (dwt->dn > 1))) {
 949                         return;
 950                 }
 951                 a = 1;
 952                 b = 0;
 953         }
 954 #ifdef __SSE__
 955         v4dwt_decode_step1_sse(dwt->wavelet+a, dwt->sn, _mm_set1_ps(K));
 956         v4dwt_decode_step1_sse(dwt->wavelet+b, dwt->dn, _mm_set1_ps(c13318));
 957         v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(dwt_delta));
 958         v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(dwt_gamma));
 959         v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(dwt_beta));
 960         v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(dwt_alpha));
 961 #else
 962         v4dwt_decode_step1(dwt->wavelet+a, dwt->sn, K);
 963         v4dwt_decode_step1(dwt->wavelet+b, dwt->dn, c13318);
 964         v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), dwt_delta);
 965         v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), dwt_gamma);
 966         v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), dwt_beta);
 967         v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), dwt_alpha);
 968 #endif
 969 }
 970
 971
 972 // KEEP TRUNK VERSION + return type of v2 because rev557
 973 /* <summary>                             */
 974 /* Inverse 9-7 wavelet transform in 2-D. */
 975 /* </summary>                            */
 976 // V1 void dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, int numres){
 977 opj_bool dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, int numres){
 978         v4dwt_t h;
 979         v4dwt_t v;
 980
 981         opj_tcd_resolution_t* res = tilec->resolutions;
 982
 983         int rw = res->x1 - res->x0;     /* width of the resolution level computed */
 984         int rh = res->y1 - res->y0;     /* height of the resolution level computed */
 985
 986         int w = tilec->x1 - tilec->x0;
 987
 988         h.wavelet = (v4*) opj_aligned_malloc((dwt_max_resolution(res, numres)+5) * sizeof(v4));
 989         v.wavelet = h.wavelet;
 990
 991         while( --numres) {
 992                 float * restrict aj = (float*) tilec->data;
 993                 int bufsize = (tilec->x1 - tilec->x0) * (tilec->y1 - tilec->y0);
 994                 int j;
 995
 996                 h.sn = rw;
 997                 v.sn = rh;
 998
 999                 ++res;
1000
1001                 rw = res->x1 - res->x0; /* width of the resolution level computed */
1002                 rh = res->y1 - res->y0; /* height of the resolution level computed */
1003
1004                 h.dn = rw - h.sn;
1005                 h.cas = res->x0 % 2;
1006
1007                 for(j = rh; j > 3; j -= 4){
1008                         int k;
1009                         v4dwt_interleave_h(&h, aj, w, bufsize);
1010                         v4dwt_decode(&h);
1011                                 for(k = rw; --k >= 0;){
1012                                         aj[k    ] = h.wavelet[k].f[0];
1013                                         aj[k+w  ] = h.wavelet[k].f[1];
1014                                         aj[k+w*2] = h.wavelet[k].f[2];
1015                                         aj[k+w*3] = h.wavelet[k].f[3];
1016                                 }
1017                         aj += w*4;
1018                         bufsize -= w*4;
1019                 }
1020                 if (rh & 0x03) {
1021                                 int k;
1022                         j = rh & 0x03;
1023                         v4dwt_interleave_h(&h, aj, w, bufsize);
1024                         v4dwt_decode(&h);
1025                                 for(k = rw; --k >= 0;){
1026                                         switch(j) {
1027                                                 case 3: aj[k+w*2] = h.wavelet[k].f[2];
1028                                                 case 2: aj[k+w  ] = h.wavelet[k].f[1];
1029                                                 case 1: aj[k    ] = h.wavelet[k].f[0];
1030                                         }
1031                                 }
1032                         }
1033
1034                 v.dn = rh - v.sn;
1035                 v.cas = res->y0 % 2;
1036
1037                 aj = (float*) tilec->data;
1038                 for(j = rw; j > 3; j -= 4){
1039                         int k;
1040                         v4dwt_interleave_v(&v, aj, w);
1041                         v4dwt_decode(&v);
1042                                 for(k = 0; k < rh; ++k){
1043                                         memcpy(&aj[k*w], &v.wavelet[k], 4 * sizeof(float));
1044                                 }
1045                         aj += 4;
1046                 }
1047                 if (rw & 0x03){
1048                                 int k;
1049                         j = rw & 0x03;
1050                         v4dwt_interleave_v(&v, aj, w);
1051                         v4dwt_decode(&v);
1052                                 for(k = 0; k < rh; ++k){
1053                                         memcpy(&aj[k*w], &v.wavelet[k], j * sizeof(float));
1054                                 }
1055                         }
1056         }
1057
1058         opj_aligned_free(h.wavelet);
1059         return OPJ_TRUE;
1060 }
1061
1062
1063 /* <summary>                             */
1064 /* Inverse 9-7 wavelet transform in 2-D. */
1065 /* </summary>                            */
1066 opj_bool dwt_decode_real_v2(opj_tcd_tilecomp_v2_t* restrict tilec, OPJ_UINT32 numres){
1067         v4dwt_t h;
1068         v4dwt_t v;
1069
1070         opj_tcd_resolution_v2_t* res = tilec->resolutions;
1071
1072         OPJ_UINT32 rw = res->x1 - res->x0;      /* width of the resolution level computed */
1073         OPJ_UINT32 rh = res->y1 - res->y0;      /* height of the resolution level computed */
1074
1075         OPJ_UINT32 w = tilec->x1 - tilec->x0;
1076
1077         h.wavelet = (v4*) opj_aligned_malloc((dwt_max_resolution_v2(res, numres)+5) * sizeof(v4));
1078         v.wavelet = h.wavelet;
1079
1080         while( --numres) {
1081                 OPJ_FLOAT32 * restrict aj = (OPJ_FLOAT32*) tilec->data;
1082                 OPJ_UINT32 bufsize = (tilec->x1 - tilec->x0) * (tilec->y1 - tilec->y0);
1083                 OPJ_INT32 j;
1084
1085                 h.sn = rw;
1086                 v.sn = rh;
1087
1088                 ++res;
1089
1090                 rw = res->x1 - res->x0; /* width of the resolution level computed */
1091                 rh = res->y1 - res->y0; /* height of the resolution level computed */
1092
1093                 h.dn = rw - h.sn;
1094                 h.cas = res->x0 % 2;
1095
1096                 for(j = rh; j > 3; j -= 4) {
1097                         OPJ_INT32 k;
1098                         v4dwt_interleave_h(&h, aj, w, bufsize);
1099                         v4dwt_decode(&h);
1100
1101                         for(k = rw; --k >= 0;){
1102                                 aj[k    ] = h.wavelet[k].f[0];
1103                                 aj[k+w  ] = h.wavelet[k].f[1];
1104                                 aj[k+w*2] = h.wavelet[k].f[2];
1105                                 aj[k+w*3] = h.wavelet[k].f[3];
1106                         }
1107
1108                         aj += w*4;
1109                         bufsize -= w*4;
1110                 }
1111
1112                 if (rh & 0x03) {
1113                         int k;
1114                         j = rh & 0x03;
1115                         v4dwt_interleave_h(&h, aj, w, bufsize);
1116                         v4dwt_decode(&h);
1117                         for(k = rw; --k >= 0;){
1118                                 switch(j) {
1119                                         case 3: aj[k+w*2] = h.wavelet[k].f[2];
1120                                         case 2: aj[k+w  ] = h.wavelet[k].f[1];
1121                                         case 1: aj[k    ] = h.wavelet[k].f[0];
1122                                 }
1123                         }
1124                 }
1125
1126                 v.dn = rh - v.sn;
1127                 v.cas = res->y0 % 2;
1128
1129                 aj = (OPJ_FLOAT32*) tilec->data;
1130                 for(j = rw; j > 3; j -= 4){
1131                         OPJ_INT32 k;
1132
1133                         v4dwt_interleave_v(&v, aj, w);
1134                         v4dwt_decode(&v);
1135
1136                         for(k = 0; k < rh; ++k){
1137                                 memcpy(&aj[k*w], &v.wavelet[k], 4 * sizeof(OPJ_FLOAT32));
1138                         }
1139                         aj += 4;
1140                 }
1141
1142                 if (rw & 0x03){
1143                         OPJ_INT32 k;
1144
1145                         j = rw & 0x03;
1146
1147                         v4dwt_interleave_v(&v, aj, w);
1148                         v4dwt_decode(&v);
1149
1150                         for(k = 0; k < rh; ++k){
1151                                 memcpy(&aj[k*w], &v.wavelet[k], j * sizeof(OPJ_FLOAT32));
1152                         }
1153                 }
1154         }
1155
1156         opj_aligned_free(h.wavelet);
1157         return OPJ_TRUE;
1158 }