src/lib/openjp2/sparse_array.c

   1 /*
   2  * The copyright in this software is being made available under the 2-clauses
   3  * BSD License, included below. This software may be subject to other third
   4  * party and contributor rights, including patent rights, and no such rights
   5  * are granted under this license.
   6  *
   7  * Copyright (c) 2017, IntoPix SA <contact@intopix.com>
   8  * All rights reserved.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  * 1. Redistributions of source code must retain the above copyright
  14  *    notice, this list of conditions and the following disclaimer.
  15  * 2. Redistributions in binary form must reproduce the above copyright
  16  *    notice, this list of conditions and the following disclaimer in the
  17  *    documentation and/or other materials provided with the distribution.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS `AS IS'
  20  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  23  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  29  * POSSIBILITY OF SUCH DAMAGE.
  30  */
  31
  32 #include "opj_includes.h"
  33
  34
  35 struct opj_sparse_array_int32 {
  36     OPJ_UINT32 width;
  37     OPJ_UINT32 height;
  38     OPJ_UINT32 block_width;
  39     OPJ_UINT32 block_height;
  40     OPJ_UINT32 block_count_hor;
  41     OPJ_UINT32 block_count_ver;
  42     OPJ_INT32** data_blocks;
  43 };
  44
  45 opj_sparse_array_int32_t* opj_sparse_array_int32_create(OPJ_UINT32 width,
  46         OPJ_UINT32 height,
  47         OPJ_UINT32 block_width,
  48         OPJ_UINT32 block_height)
  49 {
  50     opj_sparse_array_int32_t* sa;
  51
  52     if (width == 0 || height == 0 || block_width == 0 || block_height == 0) {
  53         return NULL;
  54     }
  55     if (block_width > ((OPJ_UINT32)~0U) / block_height / sizeof(OPJ_INT32)) {
  56         return NULL;
  57     }
  58
  59     sa = opj_calloc(1, sizeof(opj_sparse_array_int32_t));
  60     sa->width = width;
  61     sa->height = height;
  62     sa->block_width = block_width;
  63     sa->block_height = block_height;
  64     sa->block_count_hor = opj_uint_ceildiv(width, block_width);
  65     sa->block_count_ver = opj_uint_ceildiv(height, block_height);
  66     if (sa->block_count_hor > ((OPJ_UINT32)~0U) / sa->block_count_ver) {
  67         opj_free(sa);
  68         return NULL;
  69     }
  70     sa->data_blocks = opj_calloc(sizeof(OPJ_INT32*),
  71                                  sa->block_count_hor * sa->block_count_ver);
  72     if (sa->data_blocks == NULL) {
  73         opj_free(sa);
  74         return NULL;
  75     }
  76
  77     return sa;
  78 }
  79
  80 void opj_sparse_array_int32_free(opj_sparse_array_int32_t* sa)
  81 {
  82     if (sa) {
  83         OPJ_UINT32 i;
  84         for (i = 0; i < sa->block_count_hor * sa->block_count_ver; i++) {
  85             if (sa->data_blocks[i]) {
  86                 opj_free(sa->data_blocks[i]);
  87             }
  88         }
  89         opj_free(sa->data_blocks);
  90         opj_free(sa);
  91     }
  92 }
  93
  94 OPJ_BOOL opj_sparse_array_is_region_valid(const opj_sparse_array_int32_t* sa,
  95         OPJ_UINT32 x0,
  96         OPJ_UINT32 y0,
  97         OPJ_UINT32 x1,
  98         OPJ_UINT32 y1)
  99 {
 100     return !(x0 >= sa->width || x1 <= x0 || x1 > sa->width ||
 101              y0 >= sa->height || y1 <= y0 || y1 > sa->height);
 102 }
 103
 104 static OPJ_BOOL opj_sparse_array_int32_read_or_write(
 105     const opj_sparse_array_int32_t* sa,
 106     OPJ_UINT32 x0,
 107     OPJ_UINT32 y0,
 108     OPJ_UINT32 x1,
 109     OPJ_UINT32 y1,
 110     OPJ_INT32* buf,
 111     OPJ_UINT32 buf_col_stride,
 112     OPJ_UINT32 buf_line_stride,
 113     OPJ_BOOL forgiving,
 114     OPJ_BOOL is_read_op)
 115 {
 116     OPJ_UINT32 y, block_y;
 117     OPJ_UINT32 y_incr = 0;
 118     const OPJ_UINT32 block_width = sa->block_width;
 119
 120     if (!opj_sparse_array_is_region_valid(sa, x0, y0, x1, y1)) {
 121         return forgiving;
 122     }
 123
 124     block_y = y0 / sa->block_height;
 125     for (y = y0; y < y1; block_y ++, y += y_incr) {
 126         OPJ_UINT32 x, block_x;
 127         OPJ_UINT32 x_incr = 0;
 128         OPJ_UINT32 block_y_offset;
 129         y_incr = (y == y0) ? sa->block_height - (y0 % sa->block_height) :
 130                  sa->block_height;
 131         block_y_offset = sa->block_height - y_incr;
 132         y_incr = opj_uint_min(y_incr, y1 - y);
 133         block_x = x0 / block_width;
 134         for (x = x0; x < x1; block_x ++, x += x_incr) {
 135             OPJ_UINT32 j;
 136             OPJ_UINT32 block_x_offset;
 137             OPJ_INT32* src_block;
 138             x_incr = (x == x0) ? block_width - (x0 % block_width) : block_width;
 139             block_x_offset = block_width - x_incr;
 140             x_incr = opj_uint_min(x_incr, x1 - x);
 141             src_block = sa->data_blocks[block_y * sa->block_count_hor + block_x];
 142             if (is_read_op) {
 143                 if (src_block == NULL) {
 144                     if (buf_col_stride == 1) {
 145                         OPJ_INT32* dest_ptr = buf + (y - y0) * (size_t)buf_line_stride +
 146                                               (x - x0) * buf_col_stride;
 147                         for (j = 0; j < y_incr; j++) {
 148                             memset(dest_ptr, 0, sizeof(OPJ_INT32) * x_incr);
 149                             dest_ptr += buf_line_stride;
 150                         }
 151                     } else {
 152                         OPJ_INT32* dest_ptr = buf + (y - y0) * (size_t)buf_line_stride +
 153                                               (x - x0) * buf_col_stride;
 154                         for (j = 0; j < y_incr; j++) {
 155                             OPJ_UINT32 k;
 156                             for (k = 0; k < x_incr; k++) {
 157                                 dest_ptr[k * buf_col_stride] = 0;
 158                             }
 159                             dest_ptr += buf_line_stride;
 160                         }
 161                     }
 162                 } else {
 163                     const OPJ_INT32* OPJ_RESTRICT src_ptr = src_block + block_y_offset *
 164                                                             (size_t)block_width + block_x_offset;
 165                     if (buf_col_stride == 1) {
 166                         OPJ_INT32* OPJ_RESTRICT dest_ptr = buf + (y - y0) * (size_t)buf_line_stride +
 167                                                            (x - x0) * buf_col_stride;
 168                         if (x_incr == 4) {
 169                             // Same code as general branch, but the compiler
 170                             // can have an efficient memcpy()
 171                             for (j = 0; j < y_incr; j++) {
 172                                 memcpy(dest_ptr, src_ptr, sizeof(OPJ_INT32) * x_incr);
 173                                 dest_ptr += buf_line_stride;
 174                                 src_ptr += block_width;
 175                             }
 176                         } else {
 177                             for (j = 0; j < y_incr; j++) {
 178                                 memcpy(dest_ptr, src_ptr, sizeof(OPJ_INT32) * x_incr);
 179                                 dest_ptr += buf_line_stride;
 180                                 src_ptr += block_width;
 181                             }
 182                         }
 183                     } else {
 184                         OPJ_INT32* OPJ_RESTRICT dest_ptr = buf + (y - y0) * (size_t)buf_line_stride +
 185                                                            (x - x0) * buf_col_stride;
 186                         if (x_incr == 1) {
 187                             for (j = 0; j < y_incr; j++) {
 188                                 *dest_ptr = *src_ptr;
 189                                 dest_ptr += buf_line_stride;
 190                                 src_ptr += block_width;
 191                             }
 192                         } else if (y_incr == 1 && buf_col_stride == 2) {
 193                             OPJ_UINT32 k;
 194                             for (k = 0; k < (x_incr & ~3U); k += 4) {
 195                                 dest_ptr[k * buf_col_stride] = src_ptr[k];
 196                                 dest_ptr[(k + 1) * buf_col_stride] = src_ptr[k + 1];
 197                                 dest_ptr[(k + 2) * buf_col_stride] = src_ptr[k + 2];
 198                                 dest_ptr[(k + 3) * buf_col_stride] = src_ptr[k + 3];
 199                             }
 200                             for (; k < x_incr; k++) {
 201                                 dest_ptr[k * buf_col_stride] = src_ptr[k];
 202                             }
 203                         } else if (x_incr >= 8 && buf_col_stride == 8) {
 204                             for (j = 0; j < y_incr; j++) {
 205                                 OPJ_UINT32 k;
 206                                 for (k = 0; k < (x_incr & ~3U); k += 4) {
 207                                     dest_ptr[k * buf_col_stride] = src_ptr[k];
 208                                     dest_ptr[(k + 1) * buf_col_stride] = src_ptr[k + 1];
 209                                     dest_ptr[(k + 2) * buf_col_stride] = src_ptr[k + 2];
 210                                     dest_ptr[(k + 3) * buf_col_stride] = src_ptr[k + 3];
 211                                 }
 212                                 for (; k < x_incr; k++) {
 213                                     dest_ptr[k * buf_col_stride] = src_ptr[k];
 214                                 }
 215                                 dest_ptr += buf_line_stride;
 216                                 src_ptr += block_width;
 217                             }
 218                         } else {
 219                             /* General case */
 220                             for (j = 0; j < y_incr; j++) {
 221                                 OPJ_UINT32 k;
 222                                 for (k = 0; k < x_incr; k++) {
 223                                     dest_ptr[k * buf_col_stride] = src_ptr[k];
 224                                 }
 225                                 dest_ptr += buf_line_stride;
 226                                 src_ptr += block_width;
 227                             }
 228                         }
 229                     }
 230                 }
 231             } else {
 232                 if (src_block == NULL) {
 233                     src_block = opj_calloc(1,
 234                                            sa->block_width * sa->block_height * sizeof(OPJ_INT32));
 235                     if (src_block == NULL) {
 236                         return OPJ_FALSE;
 237                     }
 238                     sa->data_blocks[block_y * sa->block_count_hor + block_x] = src_block;
 239                 }
 240
 241                 if (buf_col_stride == 1) {
 242                     OPJ_INT32* OPJ_RESTRICT dest_ptr = src_block + block_y_offset *
 243                                                        (size_t)block_width + block_x_offset;
 244                     const OPJ_INT32* OPJ_RESTRICT src_ptr = buf + (y - y0) *
 245                                                             (size_t)buf_line_stride + (x - x0) * buf_col_stride;
 246                     if (x_incr == 4) {
 247                         // Same code as general branch, but the compiler
 248                         // can have an efficient memcpy()
 249                         for (j = 0; j < y_incr; j++) {
 250                             memcpy(dest_ptr, src_ptr, sizeof(OPJ_INT32) * x_incr);
 251                             dest_ptr += block_width;
 252                             src_ptr += buf_line_stride;
 253                         }
 254                     } else {
 255                         for (j = 0; j < y_incr; j++) {
 256                             memcpy(dest_ptr, src_ptr, sizeof(OPJ_INT32) * x_incr);
 257                             dest_ptr += block_width;
 258                             src_ptr += buf_line_stride;
 259                         }
 260                     }
 261                 } else {
 262                     OPJ_INT32* OPJ_RESTRICT dest_ptr = src_block + block_y_offset *
 263                                                        (size_t)block_width + block_x_offset;
 264                     const OPJ_INT32* OPJ_RESTRICT src_ptr = buf + (y - y0) *
 265                                                             (size_t)buf_line_stride + (x - x0) * buf_col_stride;
 266                     if (x_incr == 1) {
 267                         for (j = 0; j < y_incr; j++) {
 268                             *dest_ptr = *src_ptr;
 269                             src_ptr += buf_line_stride;
 270                             dest_ptr += block_width;
 271                         }
 272                     } else if (x_incr >= 8 && buf_col_stride == 8) {
 273                         for (j = 0; j < y_incr; j++) {
 274                             OPJ_UINT32 k;
 275                             for (k = 0; k < (x_incr & ~3U); k += 4) {
 276                                 dest_ptr[k] = src_ptr[k * buf_col_stride];
 277                                 dest_ptr[k + 1] = src_ptr[(k + 1) * buf_col_stride];
 278                                 dest_ptr[k + 2] = src_ptr[(k + 2) * buf_col_stride];
 279                                 dest_ptr[k + 3] = src_ptr[(k + 3) * buf_col_stride];
 280                             }
 281                             for (; k < x_incr; k++) {
 282                                 dest_ptr[k] = src_ptr[k * buf_col_stride];
 283                             }
 284                             src_ptr += buf_line_stride;
 285                             dest_ptr += block_width;
 286                         }
 287                     } else {
 288                         /* General case */
 289                         for (j = 0; j < y_incr; j++) {
 290                             OPJ_UINT32 k;
 291                             for (k = 0; k < x_incr; k++) {
 292                                 dest_ptr[k] = src_ptr[k * buf_col_stride];
 293                             }
 294                             src_ptr += buf_line_stride;
 295                             dest_ptr += block_width;
 296                         }
 297                     }
 298                 }
 299             }
 300         }
 301     }
 302
 303     return OPJ_TRUE;
 304 }
 305
 306 OPJ_BOOL opj_sparse_array_int32_read(const opj_sparse_array_int32_t* sa,
 307                                      OPJ_UINT32 x0,
 308                                      OPJ_UINT32 y0,
 309                                      OPJ_UINT32 x1,
 310                                      OPJ_UINT32 y1,
 311                                      OPJ_INT32* dest,
 312                                      OPJ_UINT32 dest_col_stride,
 313                                      OPJ_UINT32 dest_line_stride,
 314                                      OPJ_BOOL forgiving)
 315 {
 316     return opj_sparse_array_int32_read_or_write(
 317                (opj_sparse_array_int32_t*)sa, x0, y0, x1, y1,
 318                dest,
 319                dest_col_stride,
 320                dest_line_stride,
 321                forgiving,
 322                OPJ_TRUE);
 323 }
 324
 325 OPJ_BOOL opj_sparse_array_int32_write(opj_sparse_array_int32_t* sa,
 326                                       OPJ_UINT32 x0,
 327                                       OPJ_UINT32 y0,
 328                                       OPJ_UINT32 x1,
 329                                       OPJ_UINT32 y1,
 330                                       const OPJ_INT32* src,
 331                                       OPJ_UINT32 src_col_stride,
 332                                       OPJ_UINT32 src_line_stride,
 333                                       OPJ_BOOL forgiving)
 334 {
 335     return opj_sparse_array_int32_read_or_write(sa, x0, y0, x1, y1,
 336             (OPJ_INT32*)src,
 337             src_col_stride,
 338             src_line_stride,
 339             forgiving,
 340             OPJ_FALSE);
 341 }