| 1 | /* |
| 2 | * Texture Filtering |
| 3 | * Version: 1.0 |
| 4 | * |
| 5 | * Copyright (C) 2007 Hiroshi Morii All Rights Reserved. |
| 6 | * Email koolsmoky(at)users.sourceforge.net |
| 7 | * Web http://www.3dfxzone.it/koolsmoky |
| 8 | * |
| 9 | * this is free software; you can redistribute it and/or modify |
| 10 | * it under the terms of the GNU General Public License as published by |
| 11 | * the Free Software Foundation; either version 2, or (at your option) |
| 12 | * any later version. |
| 13 | * |
| 14 | * this is distributed in the hope that it will be useful, |
| 15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 17 | * GNU General Public License for more details. |
| 18 | * |
| 19 | * You should have received a copy of the GNU General Public License |
| 20 | * along with GNU Make; see the file COPYING. If not, write to |
| 21 | * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. |
| 22 | */ |
| 23 | |
| 24 | #ifdef __MSC__ |
| 25 | #pragma warning(disable: 4786) |
| 26 | #endif |
| 27 | |
| 28 | #ifndef NO_FILTER_THREAD |
| 29 | #include <functional> |
| 30 | #include <thread> |
| 31 | #endif |
| 32 | |
| 33 | /* NOTE: The codes are not optimized. They can be made faster. */ |
| 34 | |
| 35 | #include "TxQuantize.h" |
| 36 | |
| 37 | TxQuantize::TxQuantize() |
| 38 | { |
| 39 | _txUtil = new TxUtil(); |
| 40 | |
| 41 | /* get number of CPU cores. */ |
| 42 | _numcore = _txUtil->getNumberofProcessors(); |
| 43 | |
| 44 | /* get dxtn extensions */ |
| 45 | _tx_compress_fxt1 = TxLoadLib::getInstance()->getfxtCompressTexFuncExt(); |
| 46 | _tx_compress_dxtn_rgba = TxLoadLib::getInstance()->getdxtCompressTexFuncExt(); |
| 47 | } |
| 48 | |
| 49 | |
| 50 | TxQuantize::~TxQuantize() |
| 51 | { |
| 52 | delete _txUtil; |
| 53 | } |
| 54 | |
| 55 | void |
| 56 | TxQuantize::ARGB1555_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 57 | { |
| 58 | #if 1 |
| 59 | int siz = (width * height) >> 1; |
| 60 | int i; |
| 61 | for (i = 0; i < siz; i++) { |
| 62 | *dest = (((*src & 0x00008000) ? 0xff000000 : 0x00000000) | |
| 63 | ((*src & 0x00007c00) << 9) | ((*src & 0x00007000) << 4) | |
| 64 | ((*src & 0x000003e0) << 6) | ((*src & 0x00000380) << 1) | |
| 65 | ((*src & 0x0000001f) << 3) | ((*src & 0x0000001c) >> 2)); |
| 66 | dest++; |
| 67 | *dest = (((*src & 0x80000000) ? 0xff000000 : 0x00000000) | |
| 68 | ((*src & 0x7c000000) >> 7) | ((*src & 0x70000000) >> 12) | |
| 69 | ((*src & 0x03e00000) >> 10) | ((*src & 0x03800000) >> 15) | |
| 70 | ((*src & 0x001f0000) >> 13) | ((*src & 0x001c0000) >> 18)); |
| 71 | dest++; |
| 72 | src++; |
| 73 | } |
| 74 | #else |
| 75 | int siz = (width * height) >> 1; |
| 76 | |
| 77 | __asm { |
| 78 | push ebx; |
| 79 | push esi; |
| 80 | push edi; |
| 81 | |
| 82 | mov esi, dword ptr [src]; |
| 83 | mov edi, dword ptr [dest]; |
| 84 | mov ecx, dword ptr [siz]; |
| 85 | |
| 86 | tc1_loop: |
| 87 | mov eax, dword ptr [esi]; |
| 88 | add esi, 4; |
| 89 | |
| 90 | // arrr rrgg gggb bbbb |
| 91 | // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 92 | mov edx, eax; // edx = arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 93 | mov ebx, 0x00000000; |
| 94 | and eax, 0x00008000; // eax = 00000000 00000000 a0000000 00000000 |
| 95 | jz transparent1; |
| 96 | mov ebx, 0xff000000; // ebx = aaaaaaaa 00000000 00000000 00000000 |
| 97 | |
| 98 | transparent1: |
| 99 | mov eax, edx; // eax = arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 100 | and edx, 0x00007c00; // edx = 00000000 00000000 0rrrrr00 00000000 |
| 101 | shl edx, 4; // edx = 00000000 00000rrr rr000000 00000000 |
| 102 | or ebx, edx; // ebx = aaaaaaaa 00000rrr rr000000 00000000 |
| 103 | shl edx, 5; // edx = 00000000 rrrrr000 00000000 00000000 |
| 104 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr rr000000 00000000 |
| 105 | and ebx, 0xffff0000; // ebx = aaaaaaaa rrrrrrrr 00000000 00000000 |
| 106 | mov edx, eax; |
| 107 | and edx, 0x000003e0; // edx = 00000000 00000000 000000gg ggg00000 |
| 108 | shl edx, 1; // edx = 00000000 00000000 00000ggg gg000000 |
| 109 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr 00000ggg gg000000 |
| 110 | shl edx, 5; // edx = 00000000 00000000 ggggg000 00000000 |
| 111 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr gggggggg gg000000 |
| 112 | and ebx, 0xffffff00; // ebx = aaaaaaaa rrrrrrrr gggggggg 00000000 |
| 113 | mov edx, eax; |
| 114 | and edx, 0x0000001f; // edx = 00000000 00000000 00000000 000bbbbb |
| 115 | shl edx, 3; // edx = 00000000 00000000 00000000 bbbbb000 |
| 116 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr gggggggg bbbbb000 |
| 117 | shr edx, 5; // edx = 00000000 00000000 00000000 00000bbb |
| 118 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 119 | |
| 120 | mov dword ptr [edi], ebx; |
| 121 | add edi, 4; |
| 122 | |
| 123 | shr eax, 16; // eax = 00000000 00000000 arrrrrgg gggbbbbb |
| 124 | mov edx, eax; // edx = 00000000 00000000 arrrrrgg gggbbbbb |
| 125 | mov ebx, 0x00000000; |
| 126 | and eax, 0x00008000; // eax = 00000000 00000000 a0000000 00000000 |
| 127 | jz transparent2; |
| 128 | mov ebx, 0xff000000; // ebx = aaaaaaaa 00000000 00000000 00000000 |
| 129 | |
| 130 | transparent2: |
| 131 | mov eax, edx; // eax = 00000000 00000000 arrrrrgg gggbbbbb |
| 132 | and edx, 0x00007c00; // edx = 00000000 00000000 0rrrrr00 00000000 |
| 133 | shl edx, 4; // edx = 00000000 00000rrr rr000000 00000000 |
| 134 | or ebx, edx; // ebx = aaaaaaaa 00000rrr rr000000 00000000 |
| 135 | shl edx, 5; // edx = 00000000 rrrrr000 00000000 00000000 |
| 136 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr rr000000 00000000 |
| 137 | and ebx, 0xffff0000; // ebx = aaaaaaaa rrrrrrrr 00000000 00000000 |
| 138 | mov edx, eax; |
| 139 | and edx, 0x000003e0; // edx = 00000000 00000000 000000gg ggg00000 |
| 140 | shl edx, 1; // edx = 00000000 00000000 00000ggg gg000000 |
| 141 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr 00000ggg gg000000 |
| 142 | shl edx, 5; // edx = 00000000 00000000 ggggg000 00000000 |
| 143 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr gggggggg gg000000 |
| 144 | and ebx, 0xffffff00; // ebx = aaaaaaaa rrrrrrrr gggggggg 00000000 |
| 145 | mov edx, eax; |
| 146 | and edx, 0x0000001f; // edx = 00000000 00000000 00000000 000bbbbb |
| 147 | shl edx, 3; // edx = 00000000 00000000 00000000 bbbbb000 |
| 148 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr gggggggg bbbbb000 |
| 149 | shr edx, 5; // edx = 00000000 00000000 00000000 00000bbb |
| 150 | or ebx, edx; // ebx = aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 151 | |
| 152 | mov dword ptr [edi], ebx; |
| 153 | add edi, 4; |
| 154 | |
| 155 | dec ecx; |
| 156 | jnz tc1_loop; |
| 157 | |
| 158 | pop edi; |
| 159 | pop esi; |
| 160 | pop ebx; |
| 161 | } |
| 162 | #endif |
| 163 | } |
| 164 | |
| 165 | void |
| 166 | TxQuantize::ARGB4444_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 167 | { |
| 168 | #if 1 |
| 169 | int siz = (width * height) >> 1; |
| 170 | int i; |
| 171 | for (i = 0; i < siz; i++) { |
| 172 | *dest = ((*src & 0x0000f000) << 12) | |
| 173 | ((*src & 0x00000f00) << 8) | |
| 174 | ((*src & 0x000000f0) << 4) | |
| 175 | (*src & 0x0000000f); |
| 176 | *dest |= (*dest << 4); |
| 177 | dest++; |
| 178 | *dest = ((*src & 0xf0000000) | |
| 179 | ((*src & 0x0f000000) >> 4) | |
| 180 | ((*src & 0x00f00000) >> 8) | |
| 181 | ((*src & 0x000f0000) >> 12)); |
| 182 | *dest |= (*dest >> 4); |
| 183 | dest++; |
| 184 | src++; |
| 185 | } |
| 186 | #else |
| 187 | int siz = (width * height) >> 1; |
| 188 | |
| 189 | __asm { |
| 190 | push ebx; |
| 191 | push esi; |
| 192 | push edi; |
| 193 | |
| 194 | mov esi, dword ptr [src]; |
| 195 | mov edi, dword ptr [dest]; |
| 196 | mov ecx, dword ptr [siz]; |
| 197 | |
| 198 | tc1_loop: |
| 199 | mov eax, dword ptr [esi]; |
| 200 | add esi, 4; |
| 201 | |
| 202 | // aaaa rrrr gggg bbbb |
| 203 | // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 204 | mov edx, eax; |
| 205 | and eax, 0x0000ffff; |
| 206 | mov ebx, eax; // 00000000 00000000 aaaarrrr ggggbbbb |
| 207 | and ebx, 0x0000f000; // 00000000 00000000 aaaa0000 00000000 |
| 208 | shl ebx, 12; // 0000aaaa 00000000 00000000 00000000 |
| 209 | or eax, ebx; // 0000aaaa 00000000 aaaarrrr ggggbbbb |
| 210 | mov ebx, eax; |
| 211 | and ebx, 0x00000f00; // 00000000 00000000 0000rrrr 00000000 |
| 212 | shl ebx, 8; // 00000000 0000rrrr 00000000 00000000 |
| 213 | or eax, ebx; // 0000aaaa 0000rrrr aaaarrrr ggggbbbb |
| 214 | mov ebx, eax; |
| 215 | and ebx, 0x000000f0; // 00000000 00000000 00000000 gggg0000 |
| 216 | shl ebx, 4; // 00000000 00000000 0000gggg 00000000 |
| 217 | and eax, 0x0f0f000f; // 0000aaaa 0000rrrr 00000000 0000bbbb |
| 218 | or eax, ebx; // 0000aaaa 0000rrrr 0000gggg 0000bbbb |
| 219 | mov ebx, eax; |
| 220 | shl ebx, 4; // aaaa0000 rrrr0000 gggg0000 bbbb0000 |
| 221 | or eax, ebx; // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 222 | |
| 223 | mov dword ptr [edi], eax; |
| 224 | add edi, 4; |
| 225 | |
| 226 | shr edx, 16; |
| 227 | mov ebx, edx; // 00000000 00000000 aaaarrrr ggggbbbb |
| 228 | and ebx, 0x0000f000; // 00000000 00000000 aaaa0000 00000000 |
| 229 | shl ebx, 12; // 0000aaaa 00000000 00000000 00000000 |
| 230 | or edx, ebx; // 0000aaaa 00000000 aaaarrrr ggggbbbb |
| 231 | mov ebx, edx; |
| 232 | and ebx, 0x00000f00; // 00000000 00000000 0000rrrr 00000000 |
| 233 | shl ebx, 8; // 00000000 0000rrrr 00000000 00000000 |
| 234 | or edx, ebx; // 0000aaaa 0000rrrr aaaarrrr ggggbbbb |
| 235 | mov ebx, edx; |
| 236 | and ebx, 0x000000f0; // 00000000 00000000 00000000 gggg0000 |
| 237 | shl ebx, 4; // 00000000 00000000 0000gggg 00000000 |
| 238 | and edx, 0x0f0f000f; // 0000aaaa 0000rrrr 00000000 0000bbbb |
| 239 | or edx, ebx; // 0000aaaa 0000rrrr 0000gggg 0000bbbb |
| 240 | mov ebx, edx; |
| 241 | shl ebx, 4; // aaaa0000 rrrr0000 gggg0000 bbbb0000 |
| 242 | or edx, ebx; // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 243 | |
| 244 | mov dword ptr [edi], edx; |
| 245 | add edi, 4; |
| 246 | |
| 247 | dec ecx; |
| 248 | jnz tc1_loop; |
| 249 | |
| 250 | pop edi; |
| 251 | pop esi; |
| 252 | pop ebx; |
| 253 | } |
| 254 | #endif |
| 255 | } |
| 256 | |
| 257 | void |
| 258 | TxQuantize::RGB565_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 259 | { |
| 260 | #if 1 |
| 261 | int siz = (width * height) >> 1; |
| 262 | int i; |
| 263 | for (i = 0; i < siz; i++) { |
| 264 | *dest = (0xff000000 | |
| 265 | ((*src & 0x0000f800) << 8) | ((*src & 0x0000e000) << 3) | |
| 266 | ((*src & 0x000007e0) << 5) | ((*src & 0x00000600) >> 1) | |
| 267 | ((*src & 0x0000001f) << 3) | ((*src & 0x0000001c) >> 2)); |
| 268 | dest++; |
| 269 | *dest = (0xff000000 | |
| 270 | ((*src & 0xf8000000) >> 8) | ((*src & 0xe0000000) >> 13) | |
| 271 | ((*src & 0x07e00000) >> 11) | ((*src & 0x06000000) >> 17) | |
| 272 | ((*src & 0x001f0000) >> 13) | ((*src & 0x001c0000) >> 18)); |
| 273 | dest++; |
| 274 | src++; |
| 275 | } |
| 276 | #else |
| 277 | int siz = (width * height) >> 1; |
| 278 | |
| 279 | __asm { |
| 280 | push ebx; |
| 281 | push esi; |
| 282 | push edi; |
| 283 | |
| 284 | mov esi, dword ptr [src]; |
| 285 | mov edi, dword ptr [dest]; |
| 286 | mov ecx, dword ptr [siz]; |
| 287 | |
| 288 | tc1_loop: |
| 289 | mov eax, dword ptr [esi]; |
| 290 | add esi, 4; |
| 291 | |
| 292 | // rrrr rggg gggb bbbb |
| 293 | // 11111111 rrrrrrrr gggggggg bbbbbbbb |
| 294 | mov edx, eax; |
| 295 | and eax, 0x0000ffff; |
| 296 | mov ebx, eax; // 00000000 00000000 rrrrrggg gggbbbbb |
| 297 | and ebx, 0x0000f800; // 00000000 00000000 rrrrr000 00000000 |
| 298 | shl ebx, 5; // 00000000 000rrrrr 00000000 00000000 |
| 299 | or eax, ebx; // 00000000 000rrrrr rrrrrggg gggbbbbb |
| 300 | mov ebx, eax; |
| 301 | and ebx, 0x000007e0; // 00000000 00000000 00000ggg ggg00000 |
| 302 | shl ebx, 5; // 00000000 00000000 gggggg00 00000000 |
| 303 | and eax, 0x001F001F; // 00000000 000rrrrr 00000000 000bbbbb |
| 304 | shl eax, 3; // 00000000 rrrrr000 00000000 bbbbb000 |
| 305 | or eax, ebx; // 00000000 rrrrr000 gggggg00 bbbbb000 |
| 306 | mov ebx, eax; |
| 307 | shr ebx, 5; // 00000000 00000rrr rr000ggg ggg00bbb |
| 308 | and ebx, 0x00070007; // 00000000 00000rrr 00000000 00000bbb |
| 309 | or eax, ebx; // 00000000 rrrrrrrr gggggg00 bbbbbbbb |
| 310 | mov ebx, eax; |
| 311 | shr ebx, 6; |
| 312 | and ebx, 0x00000300; // 00000000 00000000 000000gg 00000000 |
| 313 | or eax, ebx // 00000000 rrrrrrrr gggggggg bbbbbbbb |
| 314 | or eax, 0xff000000; // 11111111 rrrrrrrr gggggggg bbbbbbbb |
| 315 | |
| 316 | mov dword ptr [edi], eax; |
| 317 | add edi, 4; |
| 318 | |
| 319 | shr edx, 16; |
| 320 | mov eax, edx; // 00000000 00000000 rrrrrggg gggbbbbb |
| 321 | and eax, 0x0000ffff; |
| 322 | mov ebx, eax; // 00000000 00000000 rrrrrggg gggbbbbb |
| 323 | and ebx, 0x0000f800; // 00000000 00000000 rrrrr000 00000000 |
| 324 | shl ebx, 5; // 00000000 000rrrrr 00000000 00000000 |
| 325 | or eax, ebx; // 00000000 000rrrrr rrrrrggg gggbbbbb |
| 326 | mov ebx, eax; |
| 327 | and ebx, 0x000007e0; // 00000000 00000000 00000ggg ggg00000 |
| 328 | shl ebx, 5; // 00000000 00000000 gggggg00 00000000 |
| 329 | and eax, 0x001F001F; // 00000000 000rrrrr 00000000 000bbbbb |
| 330 | shl eax, 3; // 00000000 rrrrr000 00000000 bbbbb000 |
| 331 | or eax, ebx; // 00000000 rrrrr000 gggggg00 bbbbb000 |
| 332 | mov ebx, eax; |
| 333 | shr ebx, 5; // 00000000 00000rrr rr000ggg ggg00bbb |
| 334 | and ebx, 0x00070007; // 00000000 00000rrr 00000000 00000bbb |
| 335 | or eax, ebx; // 00000000 rrrrrrrr gggggg00 bbbbbbbb |
| 336 | mov ebx, eax; |
| 337 | shr ebx, 6; |
| 338 | and ebx, 0x00000300; // 00000000 00000000 000000gg 00000000 |
| 339 | or eax, ebx // 00000000 rrrrrrrr gggggggg bbbbbbbb |
| 340 | or eax, 0xff000000; // 11111111 rrrrrrrr gggggggg bbbbbbbb |
| 341 | |
| 342 | mov dword ptr [edi], eax; |
| 343 | add edi, 4; |
| 344 | |
| 345 | dec ecx; |
| 346 | jnz tc1_loop; |
| 347 | |
| 348 | pop edi; |
| 349 | pop esi; |
| 350 | pop ebx; |
| 351 | } |
| 352 | #endif |
| 353 | } |
| 354 | |
| 355 | void |
| 356 | TxQuantize::A8_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 357 | { |
| 358 | #if 1 |
| 359 | int siz = (width * height) >> 2; |
| 360 | int i; |
| 361 | for (i = 0; i < siz; i++) { |
| 362 | *dest = (*src & 0x000000ff); |
| 363 | *dest |= (*dest << 8); |
| 364 | *dest |= (*dest << 16); |
| 365 | dest++; |
| 366 | *dest = (*src & 0x0000ff00); |
| 367 | *dest |= (*dest >> 8); |
| 368 | *dest |= (*dest << 16); |
| 369 | dest++; |
| 370 | *dest = (*src & 0x00ff0000); |
| 371 | *dest |= (*dest << 8); |
| 372 | *dest |= (*dest >> 16); |
| 373 | dest++; |
| 374 | *dest = (*src & 0xff000000); |
| 375 | *dest |= (*dest >> 8); |
| 376 | *dest |= (*dest >> 16); |
| 377 | dest++; |
| 378 | src++; |
| 379 | } |
| 380 | #else |
| 381 | int siz = (width * height) >> 2; |
| 382 | |
| 383 | __asm { |
| 384 | push ebx; |
| 385 | push esi; |
| 386 | push edi; |
| 387 | |
| 388 | mov esi, dword ptr [src]; |
| 389 | mov edi, dword ptr [dest]; |
| 390 | mov ecx, dword ptr [siz]; |
| 391 | |
| 392 | tc1_loop: |
| 393 | mov eax, dword ptr [esi]; |
| 394 | add esi, 4; |
| 395 | |
| 396 | // aaaaaaaa |
| 397 | // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 398 | mov edx, eax; |
| 399 | and eax, 0x000000ff; |
| 400 | mov ebx, eax; // 00000000 00000000 00000000 aaaaaaaa |
| 401 | shl ebx, 8; // 00000000 00000000 aaaaaaaa 00000000 |
| 402 | or eax, ebx; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 403 | mov ebx, eax; |
| 404 | shl ebx, 16; // aaaaaaaa aaaaaaaa 00000000 00000000 |
| 405 | or eax, ebx; // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 406 | |
| 407 | mov dword ptr [edi], eax; |
| 408 | add edi, 4; |
| 409 | |
| 410 | mov eax, edx; |
| 411 | and eax, 0x0000ff00; |
| 412 | mov ebx, eax; // 00000000 00000000 aaaaaaaa 00000000 |
| 413 | shr ebx, 8; // 00000000 00000000 00000000 aaaaaaaa |
| 414 | or eax, ebx; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 415 | mov ebx, eax; |
| 416 | shl ebx, 16; // aaaaaaaa aaaaaaaa 00000000 00000000 |
| 417 | or eax, ebx; // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 418 | |
| 419 | mov dword ptr [edi], eax; |
| 420 | add edi, 4; |
| 421 | |
| 422 | mov eax, edx; |
| 423 | and eax, 0x00ff0000; |
| 424 | mov ebx, eax; // 00000000 aaaaaaaa 00000000 00000000 |
| 425 | shl ebx, 8; // aaaaaaaa 00000000 00000000 00000000 |
| 426 | or eax, ebx; // aaaaaaaa aaaaaaaa 00000000 00000000 |
| 427 | mov ebx, eax; |
| 428 | shr ebx, 16; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 429 | or eax, ebx; // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 430 | |
| 431 | mov dword ptr [edi], eax; |
| 432 | add edi, 4; |
| 433 | |
| 434 | mov eax, edx; |
| 435 | and eax, 0xff000000; |
| 436 | mov ebx, eax; // aaaaaaaa 00000000 00000000 00000000 |
| 437 | shr ebx, 8; // 00000000 aaaaaaaa 00000000 00000000 |
| 438 | or eax, ebx; // aaaaaaaa aaaaaaaa 00000000 00000000 |
| 439 | mov ebx, eax; |
| 440 | shr ebx, 16; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 441 | or eax, ebx; // aaaaaaaa rrrrrrrr gggggggg bbbbbbbb |
| 442 | |
| 443 | mov dword ptr [edi], eax; |
| 444 | add edi, 4; |
| 445 | |
| 446 | dec ecx; |
| 447 | jnz tc1_loop; |
| 448 | |
| 449 | pop edi; |
| 450 | pop esi; |
| 451 | pop ebx; |
| 452 | } |
| 453 | #endif |
| 454 | } |
| 455 | |
| 456 | void |
| 457 | TxQuantize::AI44_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 458 | { |
| 459 | #if 1 |
| 460 | int siz = (width * height) >> 2; |
| 461 | int i; |
| 462 | for (i = 0; i < siz; i++) { |
| 463 | *dest = (*src & 0x0000000f); |
| 464 | *dest |= ((*dest << 8) | (*dest << 16)); |
| 465 | *dest |= ((*src & 0x000000f0) << 20); |
| 466 | *dest |= (*dest << 4); |
| 467 | dest++; |
| 468 | *dest = (*src & 0x00000f00); |
| 469 | *dest |= ((*dest << 8) | (*dest >> 8)); |
| 470 | *dest |= ((*src & 0x0000f000) << 12); |
| 471 | *dest |= (*dest << 4); |
| 472 | dest++; |
| 473 | *dest = (*src & 0x000f0000); |
| 474 | *dest |= ((*dest >> 8) | (*dest >> 16)); |
| 475 | *dest |= ((*src & 0x00f00000) << 4); |
| 476 | *dest |= (*dest << 4); |
| 477 | dest++; |
| 478 | *dest = ((*src & 0x0f000000) >> 4); |
| 479 | *dest |= ((*dest >> 8) | (*dest >> 16)); |
| 480 | *dest |= (*src & 0xf0000000); |
| 481 | *dest |= (*dest >> 4); |
| 482 | dest++; |
| 483 | src++; |
| 484 | } |
| 485 | #else |
| 486 | int siz = (width * height) >> 2; |
| 487 | |
| 488 | __asm { |
| 489 | push ebx; |
| 490 | push esi; |
| 491 | push edi; |
| 492 | |
| 493 | mov esi, dword ptr [src]; |
| 494 | mov edi, dword ptr [dest]; |
| 495 | mov ecx, dword ptr [siz]; |
| 496 | |
| 497 | tc1_loop: |
| 498 | mov eax, dword ptr [esi]; |
| 499 | add esi, 4; |
| 500 | |
| 501 | // aaaaiiii |
| 502 | // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 503 | mov edx, eax; |
| 504 | and eax, 0x000000f0; // 00000000 00000000 00000000 aaaa0000 |
| 505 | mov ebx, edx; |
| 506 | shl eax, 20; // 0000aaaa 00000000 00000000 00000000 |
| 507 | and ebx, 0x0000000f; // 00000000 00000000 00000000 0000iiii |
| 508 | or eax, ebx; // 0000aaaa 00000000 00000000 0000iiii |
| 509 | shl ebx, 8; // 00000000 00000000 0000iiii 00000000 |
| 510 | or eax, ebx; // 0000aaaa 00000000 0000iiii 0000iiii |
| 511 | shl ebx, 8; // 00000000 0000iiii 00000000 00000000 |
| 512 | or eax, ebx; // 0000aaaa 0000iiii 0000iiii 0000iiii |
| 513 | mov ebx, eax; |
| 514 | shl ebx, 4; // aaaa0000 iiii0000 iiii0000 iiii0000 |
| 515 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 516 | |
| 517 | mov dword ptr [edi], eax; |
| 518 | add edi, 4; |
| 519 | |
| 520 | mov eax, edx; |
| 521 | and eax, 0x0000f000; // 00000000 00000000 aaaa0000 00000000 |
| 522 | mov ebx, edx; |
| 523 | shl eax, 12; // 0000aaaa 00000000 00000000 00000000 |
| 524 | and ebx, 0x00000f00; // 00000000 00000000 0000iiii 00000000 |
| 525 | or eax, ebx; // 0000aaaa 00000000 0000iiii 00000000 |
| 526 | shr ebx, 8; // 00000000 00000000 00000000 0000iiii |
| 527 | or eax, ebx; // 0000aaaa 00000000 0000iiii 0000iiii |
| 528 | shl ebx, 16; // 00000000 0000iiii 00000000 00000000 |
| 529 | or eax, ebx; // 0000aaaa 0000iiii 0000iiii 0000iiii |
| 530 | mov ebx, eax; |
| 531 | shl ebx, 4; // aaaa0000 iiii0000 iiii0000 iiii0000 |
| 532 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 533 | |
| 534 | mov dword ptr [edi], eax; |
| 535 | add edi, 4; |
| 536 | |
| 537 | mov eax, edx; |
| 538 | and eax, 0x00f00000; // 00000000 aaaa0000 00000000 00000000 |
| 539 | mov ebx, edx; |
| 540 | shl eax, 4; // 0000aaaa 00000000 00000000 00000000 |
| 541 | and ebx, 0x000f0000; // 00000000 0000iiii 00000000 00000000 |
| 542 | or eax, ebx; // 0000aaaa 0000iiii 00000000 00000000 |
| 543 | shr ebx, 8; // 00000000 00000000 0000iiii 00000000 |
| 544 | or eax, ebx; // 0000aaaa 0000iiii 0000iiii 00000000 |
| 545 | shr ebx, 8; // 00000000 00000000 00000000 0000iiii |
| 546 | or eax, ebx; // 0000aaaa 0000iiii 0000iiii 0000iiii |
| 547 | mov ebx, eax; |
| 548 | shl ebx, 4; // aaaa0000 iiii0000 iiii0000 iiii0000 |
| 549 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 550 | |
| 551 | mov dword ptr [edi], eax; |
| 552 | add edi, 4; |
| 553 | |
| 554 | mov eax, edx; |
| 555 | and eax, 0xf0000000; // aaaa0000 00000000 00000000 00000000 |
| 556 | mov ebx, edx; |
| 557 | and ebx, 0x0f000000; // 0000iiii 00000000 00000000 00000000 |
| 558 | shr ebx, 4; // 00000000 iiii0000 00000000 00000000 |
| 559 | or eax, ebx; // aaaa0000 iiii0000 00000000 00000000 |
| 560 | shr ebx, 8; // 00000000 00000000 iiii0000 00000000 |
| 561 | or eax, ebx; // aaaa0000 iiii0000 iiii0000 00000000 |
| 562 | shr ebx, 8; // 00000000 00000000 00000000 iiii0000 |
| 563 | or eax, ebx; // aaaa0000 iiii0000 iiii0000 iiii0000 |
| 564 | mov ebx, eax; |
| 565 | shr ebx, 4; // 0000aaaa 0000iiii 0000iiii 0000iiii |
| 566 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 567 | |
| 568 | mov dword ptr [edi], eax; |
| 569 | add edi, 4; |
| 570 | |
| 571 | dec ecx; |
| 572 | jnz tc1_loop; |
| 573 | |
| 574 | pop edi; |
| 575 | pop esi; |
| 576 | pop ebx; |
| 577 | } |
| 578 | #endif |
| 579 | } |
| 580 | |
| 581 | void |
| 582 | TxQuantize::AI88_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 583 | { |
| 584 | #if 1 |
| 585 | int siz = (width * height) >> 1; |
| 586 | int i; |
| 587 | for (i = 0; i < siz; i++) { |
| 588 | *dest = (*src & 0x000000ff); |
| 589 | *dest |= ((*dest << 8) | (*dest << 16)); |
| 590 | *dest |= ((*src & 0x0000ff00) << 16); |
| 591 | dest++; |
| 592 | *dest = (*src & 0x00ff0000); |
| 593 | *dest |= ((*dest >> 8) | (*dest >> 16)); |
| 594 | *dest |= (*src & 0xff000000); |
| 595 | dest++; |
| 596 | src++; |
| 597 | } |
| 598 | #else |
| 599 | int siz = (width * height) >> 1; |
| 600 | |
| 601 | __asm { |
| 602 | push ebx; |
| 603 | push esi; |
| 604 | push edi; |
| 605 | |
| 606 | mov esi, dword ptr [src]; |
| 607 | mov edi, dword ptr [dest]; |
| 608 | mov ecx, dword ptr [siz]; |
| 609 | |
| 610 | tc1_loop: |
| 611 | mov eax, dword ptr [esi]; |
| 612 | add esi, 4; |
| 613 | |
| 614 | // aaaaaaaa iiiiiiii |
| 615 | // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 616 | mov edx, eax; |
| 617 | and eax, 0x0000ffff; // 00000000 00000000 aaaaaaaa iiiiiiii |
| 618 | mov ebx, eax; // 00000000 00000000 aaaaaaaa iiiiiiii |
| 619 | shl eax, 16; // aaaaaaaa iiiiiiii 00000000 00000000 |
| 620 | and ebx, 0x000000ff; // 00000000 00000000 00000000 iiiiiiii |
| 621 | or eax, ebx; // aaaaaaaa iiiiiiii 00000000 iiiiiiii |
| 622 | shl ebx, 8; // 00000000 00000000 iiiiiiii 00000000 |
| 623 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 624 | |
| 625 | mov dword ptr [edi], eax; |
| 626 | add edi, 4; |
| 627 | |
| 628 | mov eax, edx; |
| 629 | and eax, 0xffff0000; // aaaaaaaa iiiiiiii 00000000 00000000 |
| 630 | mov ebx, eax; // aaaaaaaa iiiiiiii 00000000 00000000 |
| 631 | and ebx, 0x00ff0000; // 00000000 iiiiiiii 00000000 00000000 |
| 632 | shr ebx, 8; // 00000000 00000000 iiiiiiii 00000000 |
| 633 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii 00000000 |
| 634 | shr ebx, 8; // 00000000 00000000 00000000 iiiiiiii |
| 635 | or eax, ebx; // aaaaaaaa iiiiiiii iiiiiiii iiiiiiii |
| 636 | |
| 637 | mov dword ptr [edi], eax; |
| 638 | add edi, 4; |
| 639 | |
| 640 | dec ecx; |
| 641 | jnz tc1_loop; |
| 642 | |
| 643 | pop edi; |
| 644 | pop esi; |
| 645 | pop ebx; |
| 646 | } |
| 647 | #endif |
| 648 | } |
| 649 | |
| 650 | void |
| 651 | TxQuantize::ARGB8888_ARGB1555(uint32* src, uint32* dest, int width, int height) |
| 652 | { |
| 653 | #if 1 |
| 654 | int siz = (width * height) >> 1; |
| 655 | int i; |
| 656 | for (i = 0; i < siz; i++) { |
| 657 | *dest = ((*src & 0xff000000) ? 0x00008000 : 0x00000000); |
| 658 | *dest |= (((*src & 0x00f80000) >> 9) | |
| 659 | ((*src & 0x0000f800) >> 6) | |
| 660 | ((*src & 0x000000f8) >> 3)); |
| 661 | src++; |
| 662 | *dest |= ((*src & 0xff000000) ? 0x80000000 : 0x00000000); |
| 663 | *dest |= (((*src & 0x00f80000) << 7) | |
| 664 | ((*src & 0x0000f800) << 10) | |
| 665 | ((*src & 0x000000f8) << 13)); |
| 666 | src++; |
| 667 | dest++; |
| 668 | } |
| 669 | #else |
| 670 | int siz = (width * height) >> 1; |
| 671 | |
| 672 | __asm { |
| 673 | push ebx; |
| 674 | push esi; |
| 675 | push edi; |
| 676 | |
| 677 | mov esi, dword ptr [src]; |
| 678 | mov edi, dword ptr [dest]; |
| 679 | mov ecx, dword ptr [siz]; |
| 680 | |
| 681 | tc1_loop: |
| 682 | mov eax, dword ptr [esi]; |
| 683 | add esi, 4; |
| 684 | |
| 685 | #if 1 |
| 686 | mov edx, eax; |
| 687 | and eax, 0xff000000; // aaaa0000 00000000 00000000 00000000 |
| 688 | jz transparent1; |
| 689 | mov eax, 0x00008000; // 00000000 00000000 a0000000 00000000 |
| 690 | |
| 691 | transparent1: |
| 692 | mov ebx, edx; |
| 693 | and ebx, 0x00f80000; // 00000000 rrrrr000 00000000 00000000 |
| 694 | shr ebx, 9; // 00000000 00000000 0rrrrr00 00000000 |
| 695 | or eax, ebx; // 00000000 00000000 arrrrr00 00000000 |
| 696 | mov ebx, edx; |
| 697 | and ebx, 0x0000f800; // 00000000 00000000 ggggg000 00000000 |
| 698 | shr ebx, 6; // 00000000 00000000 000000gg ggg00000 |
| 699 | or eax, ebx; // 00000000 00000000 arrrrrgg ggg00000 |
| 700 | and edx, 0x000000f8; // 00000000 00000000 00000000 bbbbb000 |
| 701 | shr edx, 3; // 00000000 00000000 00000000 000bbbbb |
| 702 | or edx, eax; // 00000000 00000000 arrrrrgg gggbbbbb |
| 703 | |
| 704 | mov eax, dword ptr [esi]; |
| 705 | add esi, 4; |
| 706 | |
| 707 | mov ebx, eax; |
| 708 | and eax, 0xff000000; // aaaa0000 00000000 00000000 00000000 |
| 709 | jz transparent2; |
| 710 | or edx, 0x80000000; // a0000000 00000000 arrrrrgg gggbbbbb |
| 711 | |
| 712 | transparent2: |
| 713 | mov eax, ebx; |
| 714 | and ebx, 0x00f80000; // 00000000 rrrrr000 00000000 00000000 |
| 715 | shl ebx, 7; // 0rrrrr00 00000000 00000000 00000000 |
| 716 | or edx, ebx; // arrrrr00 00000000 arrrrrgg gggbbbbb |
| 717 | mov ebx, eax; |
| 718 | and ebx, 0x0000f800; // 00000000 00000000 ggggg000 00000000 |
| 719 | shl ebx, 10; // 000000gg ggg00000 00000000 00000000 |
| 720 | or edx, ebx; // arrrrrgg ggg00000 arrrrrgg gggbbbbb |
| 721 | and eax, 0x000000f8; // 00000000 00000000 00000000 bbbbb000 |
| 722 | shl eax, 13; // 00000000 000bbbbb 00000000 00000000 |
| 723 | or edx, eax; // arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 724 | |
| 725 | mov dword ptr [edi], edx; |
| 726 | add edi, 4; |
| 727 | #else |
| 728 | mov edx, eax; |
| 729 | and edx, 0x01000000; // 0000000a 00000000 00000000 00000000 |
| 730 | shr edx, 9; // 00000000 00000000 a0000000 00000000 |
| 731 | mov ebx, eax; |
| 732 | and ebx, 0x00f80000; // 00000000 rrrrr000 00000000 00000000 |
| 733 | shr ebx, 9; // 00000000 00000000 0rrrrr00 00000000 |
| 734 | or edx, ebx; // 00000000 00000000 arrrrr00 00000000 |
| 735 | mov ebx, eax; |
| 736 | and ebx, 0x0000f800; // 00000000 00000000 ggggg000 00000000 |
| 737 | shr ebx, 6; // 00000000 00000000 000000gg ggg00000 |
| 738 | or edx, ebx; // 00000000 00000000 arrrrrgg ggg00000 |
| 739 | and eax, 0x000000f8; // 00000000 00000000 00000000 bbbbb000 |
| 740 | shr eax, 3; // 00000000 00000000 00000000 000bbbbb |
| 741 | or edx, eax; // 00000000 00000000 arrrrrgg gggbbbbb |
| 742 | |
| 743 | mov eax, dword ptr [esi]; |
| 744 | add esi, 4; |
| 745 | |
| 746 | mov ebx, eax; |
| 747 | and ebx, 0x80000000; // a0000000 00000000 00000000 00000000 |
| 748 | or edx, ebx; // a0000000 00000000 arrrrrgg gggbbbbb |
| 749 | mov ebx, eax; |
| 750 | and ebx, 0x00f80000; // 00000000 rrrrr000 00000000 00000000 |
| 751 | shl ebx, 7; // 0rrrrr00 00000000 00000000 00000000 |
| 752 | or edx, ebx; // arrrrr00 00000000 arrrrrgg gggbbbbb |
| 753 | mov ebx, eax; |
| 754 | and ebx, 0x0000f800; // 00000000 00000000 ggggg000 00000000 |
| 755 | shl ebx, 10; // 000000gg ggg00000 00000000 00000000 |
| 756 | or edx, ebx; // arrrrrgg ggg00000 arrrrrgg gggbbbbb |
| 757 | and eax, 0x000000f8; // 00000000 00000000 00000000 bbbbb000 |
| 758 | shl eax, 13; // 00000000 000bbbbb 00000000 00000000 |
| 759 | or edx, eax; // arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 760 | |
| 761 | mov dword ptr [edi], edx; |
| 762 | add edi, 4; |
| 763 | #endif |
| 764 | dec ecx; |
| 765 | jnz tc1_loop; |
| 766 | |
| 767 | pop edi; |
| 768 | pop esi; |
| 769 | pop ebx; |
| 770 | } |
| 771 | #endif |
| 772 | } |
| 773 | |
| 774 | void |
| 775 | TxQuantize::ARGB8888_ARGB4444(uint32* src, uint32* dest, int width, int height) |
| 776 | { |
| 777 | #if 1 |
| 778 | int siz = (width * height) >> 1; |
| 779 | int i; |
| 780 | for (i = 0; i < siz; i++) { |
| 781 | *dest = (((*src & 0xf0000000) >> 16) | |
| 782 | ((*src & 0x00f00000) >> 12) | |
| 783 | ((*src & 0x0000f000) >> 8) | |
| 784 | ((*src & 0x000000f0) >> 4)); |
| 785 | src++; |
| 786 | *dest |= ((*src & 0xf0000000) | |
| 787 | ((*src & 0x00f00000) << 4) | |
| 788 | ((*src & 0x0000f000) << 8) | |
| 789 | ((*src & 0x000000f0) << 12)); |
| 790 | src++; |
| 791 | dest++; |
| 792 | } |
| 793 | #else |
| 794 | int siz = (width * height) >> 1; |
| 795 | |
| 796 | __asm { |
| 797 | push ebx; |
| 798 | push esi; |
| 799 | push edi; |
| 800 | |
| 801 | mov esi, dword ptr [src]; |
| 802 | mov edi, dword ptr [dest]; |
| 803 | mov ecx, dword ptr [siz]; |
| 804 | |
| 805 | tc1_loop: |
| 806 | mov eax, dword ptr [esi]; |
| 807 | add esi, 4; |
| 808 | |
| 809 | mov edx, eax; |
| 810 | and edx, 0xf0000000; // aaaa0000 00000000 00000000 00000000 |
| 811 | shr edx, 16; // 00000000 00000000 aaaa0000 00000000 |
| 812 | mov ebx, eax; |
| 813 | and ebx, 0x00f00000; // 00000000 rrrr0000 00000000 00000000 |
| 814 | shr ebx, 12; // 00000000 00000000 0000rrrr 00000000 |
| 815 | or edx, ebx; // 00000000 00000000 aaaarrrr 00000000 |
| 816 | mov ebx, eax; |
| 817 | and ebx, 0x0000f000; // 00000000 00000000 gggg0000 00000000 |
| 818 | shr ebx, 8; // 00000000 00000000 00000000 gggg0000 |
| 819 | or edx, ebx; // 00000000 00000000 aaaarrrr gggg0000 |
| 820 | and eax, 0x000000f0; // 00000000 00000000 00000000 bbbb0000 |
| 821 | shr eax, 4; // 00000000 00000000 00000000 0000bbbb |
| 822 | or edx, eax; // 00000000 00000000 aaaarrrr ggggbbbb |
| 823 | |
| 824 | mov eax, dword ptr [esi]; |
| 825 | add esi, 4; |
| 826 | |
| 827 | mov ebx, eax; |
| 828 | and ebx, 0xf0000000; // aaaa0000 00000000 00000000 00000000 |
| 829 | or edx, ebx; // aaaa0000 00000000 aaaarrrr ggggbbbb |
| 830 | mov ebx, eax; |
| 831 | and ebx, 0x00f00000; // 00000000 rrrr0000 00000000 00000000 |
| 832 | shl ebx, 4; // 0000rrrr 00000000 00000000 00000000 |
| 833 | or edx, ebx; // aaaarrrr 00000000 aaaarrrr ggggbbbb |
| 834 | mov ebx, eax; |
| 835 | and ebx, 0x0000f000; // 00000000 00000000 gggg0000 00000000 |
| 836 | shl ebx, 8; // 00000000 gggg0000 00000000 00000000 |
| 837 | or edx, ebx; // aaaarrrr gggg0000 aaaarrrr ggggbbbb |
| 838 | and eax, 0x000000f0; // 00000000 00000000 00000000 bbbb0000 |
| 839 | shl eax, 12; // 00000000 0000bbbb 00000000 00000000 |
| 840 | or edx, eax; // arrrrrgg ggggbbbb aaaarrrr ggggbbbb |
| 841 | |
| 842 | mov dword ptr [edi], edx; |
| 843 | add edi, 4; |
| 844 | |
| 845 | dec ecx; |
| 846 | jnz tc1_loop; |
| 847 | |
| 848 | pop edi; |
| 849 | pop esi; |
| 850 | pop ebx; |
| 851 | } |
| 852 | #endif |
| 853 | } |
| 854 | |
| 855 | void |
| 856 | TxQuantize::ARGB8888_RGB565(uint32* src, uint32* dest, int width, int height) |
| 857 | { |
| 858 | #if 1 |
| 859 | int siz = (width * height) >> 1; |
| 860 | int i; |
| 861 | for (i = 0; i < siz; i++) { |
| 862 | *dest = (((*src & 0x000000f8) >> 3) | |
| 863 | ((*src & 0x0000fc00) >> 5) | |
| 864 | ((*src & 0x00f80000) >> 8)); |
| 865 | src++; |
| 866 | *dest |= (((*src & 0x000000f8) << 13) | |
| 867 | ((*src & 0x0000fc00) << 11) | |
| 868 | ((*src & 0x00f80000) << 8)); |
| 869 | src++; |
| 870 | dest++; |
| 871 | } |
| 872 | #else |
| 873 | int siz = (width * height) >> 1; |
| 874 | |
| 875 | __asm { |
| 876 | push ebx; |
| 877 | push esi; |
| 878 | push edi; |
| 879 | |
| 880 | mov esi, dword ptr [src]; |
| 881 | mov edi, dword ptr [dest]; |
| 882 | mov ecx, dword ptr [siz]; |
| 883 | |
| 884 | tc1_loop: |
| 885 | mov eax, dword ptr [esi]; |
| 886 | add esi, 4; |
| 887 | |
| 888 | mov edx, eax; |
| 889 | and edx, 0x000000F8; // 00000000 00000000 00000000 bbbbb000 |
| 890 | shr edx, 3; // 00000000 00000000 00000000 000bbbbb |
| 891 | mov ebx, eax; |
| 892 | and ebx, 0x0000FC00; // 00000000 00000000 gggggg00 00000000 |
| 893 | shr ebx, 5; // 00000000 00000000 00000ggg ggg00000 |
| 894 | or edx, ebx; // 00000000 00000000 00000ggg gggbbbbb |
| 895 | mov ebx, eax; |
| 896 | and ebx, 0x00F80000; // 00000000 rrrrr000 00000000 00000000 |
| 897 | shr ebx, 8; // 00000000 00000000 rrrrr000 00000000 |
| 898 | or edx, ebx; // 00000000 00000000 rrrrrggg gggbbbbb |
| 899 | |
| 900 | mov eax, dword ptr [esi]; |
| 901 | add esi, 4; |
| 902 | |
| 903 | mov ebx, eax; |
| 904 | and ebx, 0x000000F8; // 00000000 00000000 00000000 bbbbb000 |
| 905 | shl ebx, 13; // 00000000 000bbbbb 00000000 00000000 |
| 906 | or edx, ebx; // 00000000 000bbbbb rrrrrggg gggbbbbb |
| 907 | mov ebx, eax; |
| 908 | and ebx, 0x0000FC00; // 00000000 00000000 gggggg00 00000000 |
| 909 | shl ebx, 11; // 00000ggg ggg00000 00000000 00000000 |
| 910 | or edx, ebx; // 00000ggg gggbbbbb rrrrrggg gggbbbbb |
| 911 | mov ebx, eax; |
| 912 | and ebx, 0x00F80000; // 00000000 rrrrr000 00000000 00000000 |
| 913 | shl ebx, 8; // rrrrr000 00000000 00000000 00000000 |
| 914 | or edx, ebx; // rrrrrggg gggbbbbb rrrrrggg gggbbbbb |
| 915 | |
| 916 | mov dword ptr [edi], edx; |
| 917 | add edi, 4; |
| 918 | |
| 919 | dec ecx; |
| 920 | jnz tc1_loop; |
| 921 | |
| 922 | pop edi; |
| 923 | pop esi; |
| 924 | pop ebx; |
| 925 | } |
| 926 | #endif |
| 927 | } |
| 928 | |
| 929 | void |
| 930 | TxQuantize::ARGB8888_A8(uint32* src, uint32* dest, int width, int height) |
| 931 | { |
| 932 | #if 1 |
| 933 | int siz = (width * height) >> 2; |
| 934 | int i; |
| 935 | for (i = 0; i < siz; i++) { |
| 936 | *dest = (*src & 0x0000ff00) >> 8; |
| 937 | src++; |
| 938 | *dest |= (*src & 0x0000ff00); |
| 939 | src++; |
| 940 | *dest |= ((*src & 0x0000ff00) << 8); |
| 941 | src++; |
| 942 | *dest |= ((*src & 0x0000ff00) << 16); |
| 943 | src++; |
| 944 | dest++; |
| 945 | } |
| 946 | #else |
| 947 | int siz = (width * height) >> 2; |
| 948 | |
| 949 | __asm { |
| 950 | push ebx; |
| 951 | push esi; |
| 952 | push edi; |
| 953 | |
| 954 | mov esi, dword ptr [src]; |
| 955 | mov edi, dword ptr [dest]; |
| 956 | mov ecx, dword ptr [siz]; |
| 957 | |
| 958 | tc1_loop: |
| 959 | mov eax, dword ptr [esi]; |
| 960 | add esi, 4; |
| 961 | |
| 962 | #if 0 |
| 963 | mov edx, eax; // we'll use A comp for every pixel |
| 964 | and edx, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 965 | shr edx, 24; // 00000000 00000000 00000000 aaaaaaaa |
| 966 | |
| 967 | mov eax, dword ptr [esi]; |
| 968 | add esi, 4; |
| 969 | |
| 970 | and eax, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 971 | shr eax, 16; // 00000000 00000000 aaaaaaaa 00000000 |
| 972 | or edx, eax; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 973 | |
| 974 | mov eax, dword ptr [esi]; |
| 975 | add esi, 4; |
| 976 | |
| 977 | and eax, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 978 | shr eax, 8; // 00000000 aaaaaaaa 00000000 00000000 |
| 979 | or edx, eax; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 980 | |
| 981 | mov eax, dword ptr [esi]; |
| 982 | add esi, 4; |
| 983 | |
| 984 | and eax, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 985 | or edx, eax; // aaaaaaaa aaaaaaaa aaaaaaaa aaaaaaaa |
| 986 | #endif |
| 987 | |
| 988 | #if 1 |
| 989 | mov edx, eax; // we'll use G comp for every pixel |
| 990 | and edx, 0x0000FF00; // 00000000 00000000 aaaaaaaa 00000000 |
| 991 | shr edx, 8; // 00000000 00000000 00000000 aaaaaaaa |
| 992 | |
| 993 | mov eax, dword ptr [esi]; |
| 994 | add esi, 4; |
| 995 | |
| 996 | and eax, 0x0000FF00; // 00000000 00000000 aaaaaaaa 00000000 |
| 997 | or edx, eax; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 998 | |
| 999 | mov eax, dword ptr [esi]; |
| 1000 | add esi, 4; |
| 1001 | |
| 1002 | and eax, 0x0000FF00; // 00000000 00000000 aaaaaaaa 00000000 |
| 1003 | shl eax, 8; // 00000000 aaaaaaaa 00000000 00000000 |
| 1004 | or edx, eax; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 1005 | |
| 1006 | mov eax, dword ptr [esi]; |
| 1007 | add esi, 4; |
| 1008 | |
| 1009 | and eax, 0x0000FF00; // 00000000 00000000 aaaaaaaa 00000000 |
| 1010 | shl eax, 16; // aaaaaaaa 00000000 00000000 00000000 |
| 1011 | or edx, eax; // aaaaaaaa aaaaaaaa aaaaaaaa aaaaaaaa |
| 1012 | #endif |
| 1013 | |
| 1014 | #if 0 |
| 1015 | mov edx, eax; |
| 1016 | and edx, 0x000000FF; // 00000000 00000000 00000000 aaaaaaaa |
| 1017 | |
| 1018 | mov eax, dword ptr [esi]; |
| 1019 | add esi, 4; |
| 1020 | |
| 1021 | and eax, 0x0000FF00; // 00000000 00000000 aaaaaaaa 00000000 |
| 1022 | or edx, eax; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 1023 | |
| 1024 | mov eax, dword ptr [esi]; |
| 1025 | add esi, 4; |
| 1026 | |
| 1027 | and eax, 0x00FF0000; // 00000000 aaaaaaaa 00000000 00000000 |
| 1028 | or edx, eax; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 1029 | |
| 1030 | mov eax, dword ptr [esi]; |
| 1031 | add esi, 4; |
| 1032 | |
| 1033 | and eax, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 1034 | or edx, eax; // aaaaaaaa aaaaaaaa aaaaaaaa aaaaaaaa |
| 1035 | #endif |
| 1036 | mov dword ptr [edi], edx; |
| 1037 | add edi, 4; |
| 1038 | |
| 1039 | dec ecx; |
| 1040 | jnz tc1_loop; |
| 1041 | |
| 1042 | pop edi; |
| 1043 | pop esi; |
| 1044 | pop ebx; |
| 1045 | } |
| 1046 | #endif |
| 1047 | } |
| 1048 | |
| 1049 | void |
| 1050 | TxQuantize::ARGB8888_AI44(uint32* src, uint32* dest, int width, int height) |
| 1051 | { |
| 1052 | #if 1 |
| 1053 | int siz = (width * height) >> 2; |
| 1054 | int i; |
| 1055 | for (i = 0; i < siz; i++) { |
| 1056 | *dest = (((*src & 0xf0000000) >> 24) | ((*src & 0x0000f000) >> 12)); |
| 1057 | src++; |
| 1058 | *dest |= (((*src & 0xf0000000) >> 16) | ((*src & 0x0000f000) >> 4)); |
| 1059 | src++; |
| 1060 | *dest |= (((*src & 0xf0000000) >> 8) | ((*src & 0x0000f000) << 4)); |
| 1061 | src++; |
| 1062 | *dest |= ((*src & 0xf0000000) | ((*src & 0x0000f000) << 12)); |
| 1063 | src++; |
| 1064 | dest++; |
| 1065 | } |
| 1066 | #else |
| 1067 | int siz = (width * height) >> 2; |
| 1068 | |
| 1069 | __asm { |
| 1070 | push ebx; |
| 1071 | push esi; |
| 1072 | push edi; |
| 1073 | |
| 1074 | mov esi, dword ptr [src]; |
| 1075 | mov edi, dword ptr [dest]; |
| 1076 | mov ecx, dword ptr [siz]; |
| 1077 | |
| 1078 | tc1_loop: |
| 1079 | mov eax, dword ptr [esi]; |
| 1080 | add esi, 4; |
| 1081 | |
| 1082 | mov edx, eax; // use A and G comps MSB |
| 1083 | and edx, 0xF0000000; // aaaa0000 00000000 00000000 00000000 |
| 1084 | mov ebx, eax; |
| 1085 | shr edx, 24; // 00000000 00000000 00000000 aaaa0000 |
| 1086 | and ebx, 0x0000F000; // 00000000 00000000 iiii0000 00000000 |
| 1087 | shr ebx, 12; // 00000000 00000000 00000000 0000iiii |
| 1088 | or edx, ebx; // 00000000 00000000 00000000 aaaaiiii |
| 1089 | |
| 1090 | mov eax, dword ptr [esi]; |
| 1091 | add esi, 4; |
| 1092 | |
| 1093 | mov ebx, eax; |
| 1094 | and eax, 0xF0000000; // aaaa0000 00000000 00000000 00000000 |
| 1095 | shr eax, 16; // 00000000 00000000 aaaa0000 00000000 |
| 1096 | and ebx, 0x0000F000; // 00000000 00000000 iiii0000 00000000 |
| 1097 | shr ebx, 4; // 00000000 00000000 0000iiii 00000000 |
| 1098 | or eax, ebx; // 00000000 00000000 aaaaiiii 00000000 |
| 1099 | or edx, eax; // 00000000 00000000 aaaaiiii aaaaiiii |
| 1100 | |
| 1101 | mov eax, dword ptr [esi]; |
| 1102 | add esi, 4; |
| 1103 | |
| 1104 | mov ebx, eax; |
| 1105 | and eax, 0xF0000000; // aaaa0000 00000000 00000000 00000000 |
| 1106 | shr eax, 8; // 00000000 aaaa0000 00000000 00000000 |
| 1107 | and ebx, 0x0000F000; // 00000000 00000000 iiii0000 00000000 |
| 1108 | shl ebx, 4; // 00000000 0000iiii 00000000 00000000 |
| 1109 | or eax, ebx; // 00000000 aaaaiiii 00000000 00000000 |
| 1110 | or edx, eax; // 00000000 aaaaiiii aaaaiiii aaaaiiii |
| 1111 | |
| 1112 | mov eax, dword ptr [esi]; |
| 1113 | add esi, 4; |
| 1114 | |
| 1115 | mov ebx, eax; |
| 1116 | and eax, 0xF0000000; // aaaa0000 00000000 00000000 00000000 |
| 1117 | and ebx, 0x0000F000; // 00000000 00000000 iiii0000 00000000 |
| 1118 | shl ebx, 12; // 0000iiii 00000000 00000000 00000000 |
| 1119 | or eax, ebx; // aaaaiiii 00000000 00000000 00000000 |
| 1120 | or edx, eax; // aaaaiiii aaaaiiii aaaaiiii aaaaiiii |
| 1121 | |
| 1122 | mov dword ptr [edi], edx; |
| 1123 | add edi, 4; |
| 1124 | |
| 1125 | dec ecx; |
| 1126 | jnz tc1_loop; |
| 1127 | |
| 1128 | pop edi; |
| 1129 | pop esi; |
| 1130 | pop ebx; |
| 1131 | } |
| 1132 | #endif |
| 1133 | } |
| 1134 | |
| 1135 | void |
| 1136 | TxQuantize::ARGB8888_AI88(uint32* src, uint32* dest, int width, int height) |
| 1137 | { |
| 1138 | #if 1 |
| 1139 | int siz = (width * height) >> 1; |
| 1140 | int i; |
| 1141 | for (i = 0; i < siz; i++) { |
| 1142 | *dest = (((*src & 0xff000000) >> 16) | ((*src & 0x0000ff00) >> 8)); |
| 1143 | src++; |
| 1144 | *dest |= ((*src & 0xff000000) | ((*src & 0x0000ff00) << 8)); |
| 1145 | src++; |
| 1146 | dest++; |
| 1147 | } |
| 1148 | #else |
| 1149 | int siz = (width * height) >> 1; |
| 1150 | |
| 1151 | __asm { |
| 1152 | push ebx; |
| 1153 | push esi; |
| 1154 | push edi; |
| 1155 | |
| 1156 | mov esi, dword ptr [src]; |
| 1157 | mov edi, dword ptr [dest]; |
| 1158 | mov ecx, dword ptr [siz]; |
| 1159 | |
| 1160 | tc1_loop: |
| 1161 | mov eax, dword ptr [esi]; |
| 1162 | add esi, 4; |
| 1163 | |
| 1164 | mov edx, eax; |
| 1165 | and edx, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 1166 | mov ebx, eax; |
| 1167 | shr edx, 16; // 00000000 00000000 aaaaaaaa 00000000 |
| 1168 | and ebx, 0x0000FF00; // 00000000 00000000 iiiiiiii 00000000 |
| 1169 | shr ebx, 8; // 00000000 00000000 00000000 iiiiiiii |
| 1170 | or edx, ebx; // 00000000 00000000 aaaaaaaa iiiiiiii |
| 1171 | |
| 1172 | mov eax, dword ptr [esi]; |
| 1173 | add esi, 4; |
| 1174 | |
| 1175 | mov ebx, eax; |
| 1176 | and eax, 0xFF000000; // aaaaaaaa 00000000 00000000 00000000 |
| 1177 | and ebx, 0x0000FF00; // 00000000 00000000 iiiiiiii 00000000 |
| 1178 | shl ebx, 8; // 00000000 iiiiiiii 00000000 00000000 |
| 1179 | or eax, ebx; // aaaaaaaa iiiiiiii 00000000 00000000 |
| 1180 | or edx, eax; // aaaaaaaa iiiiiiii aaaaaaaa iiiiiiii |
| 1181 | |
| 1182 | mov dword ptr [edi], edx; |
| 1183 | add edi, 4; |
| 1184 | |
| 1185 | dec ecx; |
| 1186 | jnz tc1_loop; |
| 1187 | |
| 1188 | pop edi; |
| 1189 | pop esi; |
| 1190 | pop ebx; |
| 1191 | } |
| 1192 | #endif |
| 1193 | } |
| 1194 | |
| 1195 | /* R.W. Floyd and L. Steinberg, An adaptive algorithm |
| 1196 | * for spatial grey scale, Proceedings of the Society |
| 1197 | * of Information Display 17, pp75-77, 1976 |
| 1198 | */ |
| 1199 | void |
| 1200 | TxQuantize::ARGB8888_RGB565_ErrD(uint32* src, uint32* dst, int width, int height) |
| 1201 | { |
| 1202 | /* Floyd-Steinberg error-diffusion halftoning */ |
| 1203 | |
| 1204 | int i, x, y; |
| 1205 | int qr, qg, qb; /* quantized incoming values */ |
| 1206 | int ir, ig, ib; /* incoming values */ |
| 1207 | int t; |
| 1208 | int *errR = new int[width]; |
| 1209 | int *errG = new int[width]; |
| 1210 | int *errB = new int[width]; |
| 1211 | |
| 1212 | uint16 *dest = (uint16 *)dst; |
| 1213 | |
| 1214 | for (i = 0; i < width; i++) errR[i] = errG[i] = errB[i] = 0; |
| 1215 | |
| 1216 | for (y = 0; y < height; y++) { |
| 1217 | qr = qg = qb = 0; |
| 1218 | for (x = 0; x < width; x++) { |
| 1219 | /* incoming pixel values */ |
| 1220 | ir = ((*src >> 16) & 0xFF) * 10000; |
| 1221 | ig = ((*src >> 8) & 0xFF) * 10000; |
| 1222 | ib = ((*src ) & 0xFF) * 10000; |
| 1223 | |
| 1224 | /* quantize pixel values. |
| 1225 | * qr * 0.4375 is the error from the pixel to the left, |
| 1226 | * errR is the error from the pixel to the top, top left, and top right */ |
| 1227 | /* qr * 0.4375 is the error distribution to the EAST in |
| 1228 | * the previous loop */ |
| 1229 | ir += errR[x] + qr * 4375 / 10000; |
| 1230 | ig += errG[x] + qg * 4375 / 10000; |
| 1231 | ib += errB[x] + qb * 4375 / 10000; |
| 1232 | |
| 1233 | /* error distribution to the SOUTH-EAST in the previous loop |
| 1234 | * can't calculate in the previous loop because it steps on |
| 1235 | * the above quantization */ |
| 1236 | errR[x] = qr * 625 / 10000; |
| 1237 | errG[x] = qg * 625 / 10000; |
| 1238 | errB[x] = qb * 625 / 10000; |
| 1239 | |
| 1240 | qr = ir; |
| 1241 | qg = ig; |
| 1242 | qb = ib; |
| 1243 | |
| 1244 | /* clamp */ |
| 1245 | if (qr < 0) qr = 0; else if (qr > 2550000) qr = 2550000; |
| 1246 | if (qg < 0) qg = 0; else if (qg > 2550000) qg = 2550000; |
| 1247 | if (qb < 0) qb = 0; else if (qb > 2550000) qb = 2550000; |
| 1248 | |
| 1249 | /* convert to RGB565 */ |
| 1250 | qr = qr * 0x1F / 2550000; |
| 1251 | qg = qg * 0x3F / 2550000; |
| 1252 | qb = qb * 0x1F / 2550000; |
| 1253 | |
| 1254 | /* this is the dithered pixel */ |
| 1255 | t = (qr << 11) | (qg << 5) | qb; |
| 1256 | |
| 1257 | /* compute the errors */ |
| 1258 | qr = ((qr << 3) | (qr >> 2)) * 10000; |
| 1259 | qg = ((qg << 2) | (qg >> 4)) * 10000; |
| 1260 | qb = ((qb << 3) | (qb >> 2)) * 10000; |
| 1261 | qr = ir - qr; |
| 1262 | qg = ig - qg; |
| 1263 | qb = ib - qb; |
| 1264 | |
| 1265 | /* compute the error distributions */ |
| 1266 | /* Floyd-Steinberg filter |
| 1267 | * 7/16 (=0.4375) to the EAST |
| 1268 | * 5/16 (=0.3125) to the SOUTH |
| 1269 | * 1/16 (=0.0625) to the SOUTH-EAST |
| 1270 | * 3/16 (=0.1875) to the SOUTH-WEST |
| 1271 | * |
| 1272 | * x 7/16 |
| 1273 | * 3/16 5/16 1/16 |
| 1274 | */ |
| 1275 | /* SOUTH-WEST */ |
| 1276 | if (x > 1) { |
| 1277 | errR[x - 1] += qr * 1875 / 10000; |
| 1278 | errG[x - 1] += qg * 1875 / 10000; |
| 1279 | errB[x - 1] += qb * 1875 / 10000; |
| 1280 | } |
| 1281 | |
| 1282 | /* SOUTH */ |
| 1283 | errR[x] += qr * 3125 / 10000; |
| 1284 | errG[x] += qg * 3125 / 10000; |
| 1285 | errB[x] += qb * 3125 / 10000; |
| 1286 | |
| 1287 | *dest = (t & 0xFFFF); |
| 1288 | |
| 1289 | dest++; |
| 1290 | src++; |
| 1291 | } |
| 1292 | } |
| 1293 | |
| 1294 | delete [] errR; |
| 1295 | delete [] errG; |
| 1296 | delete [] errB; |
| 1297 | } |
| 1298 | |
| 1299 | |
| 1300 | void |
| 1301 | TxQuantize::ARGB8888_ARGB1555_ErrD(uint32* src, uint32* dst, int width, int height) |
| 1302 | { |
| 1303 | /* Floyd-Steinberg error-diffusion halftoning */ |
| 1304 | |
| 1305 | int i, x, y; |
| 1306 | int qr, qg, qb; /* quantized incoming values */ |
| 1307 | int ir, ig, ib; /* incoming values */ |
| 1308 | int t; |
| 1309 | int *errR = new int[width]; |
| 1310 | int *errG = new int[width]; |
| 1311 | int *errB = new int[width]; |
| 1312 | |
| 1313 | uint16 *dest = (uint16 *)dst; |
| 1314 | |
| 1315 | for (i = 0; i < width; i++) errR[i] = errG[i] = errB[i] = 0; |
| 1316 | |
| 1317 | for (y = 0; y < height; y++) { |
| 1318 | qr = qg = qb = 0; |
| 1319 | for (x = 0; x < width; x++) { |
| 1320 | /* incoming pixel values */ |
| 1321 | ir = ((*src >> 16) & 0xFF) * 10000; |
| 1322 | ig = ((*src >> 8) & 0xFF) * 10000; |
| 1323 | ib = ((*src ) & 0xFF) * 10000; |
| 1324 | |
| 1325 | /* quantize pixel values. |
| 1326 | * qr * 0.4375 is the error from the pixel to the left, |
| 1327 | * errR is the error from the pixel to the top, top left, and top right */ |
| 1328 | /* qr * 0.4375 is the error distribution to the EAST in |
| 1329 | * the previous loop */ |
| 1330 | ir += errR[x] + qr * 4375 / 10000; |
| 1331 | ig += errG[x] + qg * 4375 / 10000; |
| 1332 | ib += errB[x] + qb * 4375 / 10000; |
| 1333 | |
| 1334 | /* error distribution to the SOUTH-EAST of the previous loop. |
| 1335 | * cannot calculate in the previous loop because it steps on |
| 1336 | * the above quantization */ |
| 1337 | errR[x] = qr * 625 / 10000; |
| 1338 | errG[x] = qg * 625 / 10000; |
| 1339 | errB[x] = qb * 625 / 10000; |
| 1340 | |
| 1341 | qr = ir; |
| 1342 | qg = ig; |
| 1343 | qb = ib; |
| 1344 | |
| 1345 | /* clamp */ |
| 1346 | if (qr < 0) qr = 0; else if (qr > 2550000) qr = 2550000; |
| 1347 | if (qg < 0) qg = 0; else if (qg > 2550000) qg = 2550000; |
| 1348 | if (qb < 0) qb = 0; else if (qb > 2550000) qb = 2550000; |
| 1349 | |
| 1350 | /* convert to RGB555 */ |
| 1351 | qr = qr * 0x1F / 2550000; |
| 1352 | qg = qg * 0x1F / 2550000; |
| 1353 | qb = qb * 0x1F / 2550000; |
| 1354 | |
| 1355 | /* this is the dithered pixel */ |
| 1356 | t = (qr << 10) | (qg << 5) | qb; |
| 1357 | t |= ((*src >> 24) ? 0x8000 : 0); |
| 1358 | |
| 1359 | /* compute the errors */ |
| 1360 | qr = ((qr << 3) | (qr >> 2)) * 10000; |
| 1361 | qg = ((qg << 3) | (qg >> 2)) * 10000; |
| 1362 | qb = ((qb << 3) | (qb >> 2)) * 10000; |
| 1363 | qr = ir - qr; |
| 1364 | qg = ig - qg; |
| 1365 | qb = ib - qb; |
| 1366 | |
| 1367 | /* compute the error distributions */ |
| 1368 | /* Floyd-Steinberg filter |
| 1369 | * 7/16 (=0.4375) to the EAST |
| 1370 | * 5/16 (=0.3125) to the SOUTH |
| 1371 | * 1/16 (=0.0625) to the SOUTH-EAST |
| 1372 | * 3/16 (=0.1875) to the SOUTH-WEST |
| 1373 | * |
| 1374 | * x 7/16 |
| 1375 | * 3/16 5/16 1/16 |
| 1376 | */ |
| 1377 | /* SOUTH-WEST */ |
| 1378 | if (x > 1) { |
| 1379 | errR[x - 1] += qr * 1875 / 10000; |
| 1380 | errG[x - 1] += qg * 1875 / 10000; |
| 1381 | errB[x - 1] += qb * 1875 / 10000; |
| 1382 | } |
| 1383 | |
| 1384 | /* SOUTH */ |
| 1385 | errR[x] += qr * 3125 / 10000; |
| 1386 | errG[x] += qg * 3125 / 10000; |
| 1387 | errB[x] += qb * 3125 / 10000; |
| 1388 | |
| 1389 | *dest = (t & 0xFFFF); |
| 1390 | |
| 1391 | dest++; |
| 1392 | src++; |
| 1393 | } |
| 1394 | } |
| 1395 | |
| 1396 | delete [] errR; |
| 1397 | delete [] errG; |
| 1398 | delete [] errB; |
| 1399 | } |
| 1400 | |
| 1401 | void |
| 1402 | TxQuantize::ARGB8888_ARGB4444_ErrD(uint32* src, uint32* dst, int width, int height) |
| 1403 | { |
| 1404 | /* Floyd-Steinberg error-diffusion halftoning */ |
| 1405 | |
| 1406 | /* NOTE: alpha dithering looks better for alpha gradients, but are prone |
| 1407 | * to producing noisy speckles for constant or step level alpha. Output |
| 1408 | * results should always be checked. |
| 1409 | */ |
| 1410 | boolean ditherAlpha = 0; |
| 1411 | |
| 1412 | int i, x, y; |
| 1413 | int qr, qg, qb, qa; /* quantized incoming values */ |
| 1414 | int ir, ig, ib, ia; /* incoming values */ |
| 1415 | int t; |
| 1416 | int *errR = new int[width]; |
| 1417 | int *errG = new int[width]; |
| 1418 | int *errB = new int[width]; |
| 1419 | int *errA = new int[width]; |
| 1420 | |
| 1421 | uint16 *dest = (uint16 *)dst; |
| 1422 | |
| 1423 | for (i = 0; i < width; i++) errR[i] = errG[i] = errB[i] = errA[i] = 0; |
| 1424 | |
| 1425 | for (y = 0; y < height; y++) { |
| 1426 | qr = qg = qb = qa = 0; |
| 1427 | for (x = 0; x < width; x++) { |
| 1428 | /* incoming pixel values */ |
| 1429 | ir = ((*src >> 16) & 0xFF) * 10000; |
| 1430 | ig = ((*src >> 8) & 0xFF) * 10000; |
| 1431 | ib = ((*src ) & 0xFF) * 10000; |
| 1432 | ia = ((*src >> 24) & 0xFF) * 10000; |
| 1433 | |
| 1434 | /* quantize pixel values. |
| 1435 | * qr * 0.4375 is the error from the pixel to the left, |
| 1436 | * errR is the error from the pixel to the top, top left, and top right */ |
| 1437 | /* qr * 0.4375 is the error distribution to the EAST in |
| 1438 | * the previous loop */ |
| 1439 | ir += errR[x] + qr * 4375 / 10000; |
| 1440 | ig += errG[x] + qg * 4375 / 10000; |
| 1441 | ib += errB[x] + qb * 4375 / 10000; |
| 1442 | ia += errA[x] + qa * 4375 / 10000; |
| 1443 | |
| 1444 | /* error distribution to the SOUTH-EAST of the previous loop. |
| 1445 | * cannot calculate in the previous loop because it steps on |
| 1446 | * the above quantization */ |
| 1447 | errR[x] = qr * 625 / 10000; |
| 1448 | errG[x] = qg * 625 / 10000; |
| 1449 | errB[x] = qb * 625 / 10000; |
| 1450 | errA[x] = qa * 625 / 10000; |
| 1451 | |
| 1452 | qr = ir; |
| 1453 | qg = ig; |
| 1454 | qb = ib; |
| 1455 | qa = ia; |
| 1456 | |
| 1457 | /* clamp */ |
| 1458 | if (qr < 0) qr = 0; else if (qr > 2550000) qr = 2550000; |
| 1459 | if (qg < 0) qg = 0; else if (qg > 2550000) qg = 2550000; |
| 1460 | if (qb < 0) qb = 0; else if (qb > 2550000) qb = 2550000; |
| 1461 | if (qa < 0) qa = 0; else if (qa > 2550000) qa = 2550000; |
| 1462 | |
| 1463 | /* convert to RGB444 */ |
| 1464 | qr = qr * 0xF / 2550000; |
| 1465 | qg = qg * 0xF / 2550000; |
| 1466 | qb = qb * 0xF / 2550000; |
| 1467 | qa = qa * 0xF / 2550000; |
| 1468 | |
| 1469 | /* this is the value to be returned */ |
| 1470 | if (ditherAlpha) { |
| 1471 | t = (qa << 12) | (qr << 8) | (qg << 4) | qb; |
| 1472 | } else { |
| 1473 | t = (qr << 8) | (qg << 4) | qb; |
| 1474 | t |= (*src >> 16) & 0xF000; |
| 1475 | } |
| 1476 | |
| 1477 | /* compute the errors */ |
| 1478 | qr = ((qr << 4) | qr) * 10000; |
| 1479 | qg = ((qg << 4) | qg) * 10000; |
| 1480 | qb = ((qb << 4) | qb) * 10000; |
| 1481 | qa = ((qa << 4) | qa) * 10000; |
| 1482 | qr = ir - qr; |
| 1483 | qg = ig - qg; |
| 1484 | qb = ib - qb; |
| 1485 | qa = ia - qa; |
| 1486 | |
| 1487 | /* compute the error distributions */ |
| 1488 | /* Floyd-Steinberg filter |
| 1489 | * 7/16 (=0.4375) to the EAST |
| 1490 | * 5/16 (=0.3125) to the SOUTH |
| 1491 | * 1/16 (=0.0625) to the SOUTH-EAST |
| 1492 | * 3/16 (=0.1875) to the SOUTH-WEST |
| 1493 | * |
| 1494 | * x 7/16 |
| 1495 | * 3/16 5/16 1/16 |
| 1496 | */ |
| 1497 | /* SOUTH-WEST */ |
| 1498 | if (x > 1) { |
| 1499 | errR[x - 1] += qr * 1875 / 10000; |
| 1500 | errG[x - 1] += qg * 1875 / 10000; |
| 1501 | errB[x - 1] += qb * 1875 / 10000; |
| 1502 | errA[x - 1] += qa * 1875 / 10000; |
| 1503 | } |
| 1504 | |
| 1505 | /* SOUTH */ |
| 1506 | errR[x] += qr * 3125 / 10000; |
| 1507 | errG[x] += qg * 3125 / 10000; |
| 1508 | errB[x] += qb * 3125 / 10000; |
| 1509 | errA[x] += qa * 3125 / 10000; |
| 1510 | |
| 1511 | *dest = (t & 0xFFFF); |
| 1512 | |
| 1513 | dest++; |
| 1514 | src++; |
| 1515 | } |
| 1516 | } |
| 1517 | |
| 1518 | delete [] errR; |
| 1519 | delete [] errG; |
| 1520 | delete [] errB; |
| 1521 | delete [] errA; |
| 1522 | } |
| 1523 | |
| 1524 | void |
| 1525 | TxQuantize::ARGB8888_AI44_ErrD(uint32* src, uint32* dst, int width, int height) |
| 1526 | { |
| 1527 | /* Floyd-Steinberg error-diffusion halftoning */ |
| 1528 | |
| 1529 | /* NOTE: alpha dithering looks better for alpha gradients, but are prone |
| 1530 | * to producing noisy speckles for constant or step level alpha. Output |
| 1531 | * results should always be checked. |
| 1532 | */ |
| 1533 | boolean ditherAlpha = 0; |
| 1534 | |
| 1535 | int i, x, y; |
| 1536 | int qi, qa; /* quantized incoming values */ |
| 1537 | int ii, ia; /* incoming values */ |
| 1538 | int t; |
| 1539 | int *errI = new int[width]; |
| 1540 | int *errA = new int[width]; |
| 1541 | |
| 1542 | uint8 *dest = (uint8 *)dst; |
| 1543 | |
| 1544 | for (i = 0; i < width; i++) errI[i] = errA[i] = 0; |
| 1545 | |
| 1546 | for (y = 0; y < height; y++) { |
| 1547 | qi = qa = 0; |
| 1548 | for (x = 0; x < width; x++) { |
| 1549 | /* 3dfx style Intensity = R * 0.299 + G * 0.587 + B * 0.114 */ |
| 1550 | ii = ((*src >> 16) & 0xFF) * 2990 + |
| 1551 | ((*src >> 8) & 0xFF) * 5870 + |
| 1552 | ((*src ) & 0xFF) * 1140; |
| 1553 | ia = ((*src >> 24) & 0xFF) * 10000; |
| 1554 | |
| 1555 | /* quantize pixel values. |
| 1556 | * qi * 0.4375 is the error from the pixel to the left, |
| 1557 | * errI is the error from the pixel to the top, top left, and top right */ |
| 1558 | /* qi * 0.4375 is the error distrtibution to the EAST in |
| 1559 | * the previous loop */ |
| 1560 | ii += errI[x] + qi * 4375 / 10000; |
| 1561 | ia += errA[x] + qa * 4375 / 10000; |
| 1562 | |
| 1563 | /* error distribution to the SOUTH-EAST in the previous loop. |
| 1564 | * cannot calculate in the previous loop because it steps on |
| 1565 | * the above quantization */ |
| 1566 | errI[x] = qi * 625 / 10000; |
| 1567 | errA[x] = qa * 625 / 10000; |
| 1568 | |
| 1569 | qi = ii; |
| 1570 | qa = ia; |
| 1571 | |
| 1572 | /* clamp */ |
| 1573 | if (qi < 0) qi = 0; else if (qi > 2550000) qi = 2550000; |
| 1574 | if (qa < 0) qa = 0; else if (qa > 2550000) qa = 2550000; |
| 1575 | |
| 1576 | /* convert to I4 */ |
| 1577 | qi = qi * 0xF / 2550000; |
| 1578 | qa = qa * 0xF / 2550000; |
| 1579 | |
| 1580 | /* this is the value to be returned */ |
| 1581 | if (ditherAlpha) { |
| 1582 | t = (qa << 4) | qi; |
| 1583 | } else { |
| 1584 | t = qi; |
| 1585 | t |= ((*src >> 24) & 0xF0); |
| 1586 | } |
| 1587 | |
| 1588 | /* compute the errors */ |
| 1589 | qi = ((qi << 4) | qi) * 10000; |
| 1590 | qa = ((qa << 4) | qa) * 10000; |
| 1591 | qi = ii - qi; |
| 1592 | qa = ia - qa; |
| 1593 | |
| 1594 | /* compute the error distributions */ |
| 1595 | /* Floyd-Steinberg filter |
| 1596 | * 7/16 (=0.4375) to the EAST |
| 1597 | * 5/16 (=0.3125) to the SOUTH |
| 1598 | * 1/16 (=0.0625) to the SOUTH-EAST |
| 1599 | * 3/16 (=0.1875) to the SOUTH-WEST |
| 1600 | * |
| 1601 | * x 7/16 |
| 1602 | * 3/16 5/16 1/16 |
| 1603 | */ |
| 1604 | /* SOUTH-WEST */ |
| 1605 | if (x > 1) { |
| 1606 | errI[x - 1] += qi * 1875 / 10000; |
| 1607 | errA[x - 1] += qa * 1875 / 10000; |
| 1608 | } |
| 1609 | |
| 1610 | /* SOUTH */ |
| 1611 | errI[x] += qi * 3125 / 10000; |
| 1612 | errA[x] += qa * 3125 / 10000; |
| 1613 | |
| 1614 | *dest = t & 0xFF; |
| 1615 | |
| 1616 | dest++; |
| 1617 | src++; |
| 1618 | } |
| 1619 | } |
| 1620 | |
| 1621 | delete [] errI; |
| 1622 | delete [] errA; |
| 1623 | } |
| 1624 | |
| 1625 | void |
| 1626 | TxQuantize::ARGB8888_AI88_Slow(uint32* src, uint32* dst, int width, int height) |
| 1627 | { |
| 1628 | int x, y; |
| 1629 | uint16 *dest = (uint16 *)dst; |
| 1630 | for (y = 0; y < height; y++) { |
| 1631 | for (x = 0; x < width; x++) { |
| 1632 | #if 1 |
| 1633 | /* libpng style grayscale conversion. |
| 1634 | * Reduce RGB files to grayscale with or without alpha |
| 1635 | * using the equation given in Poynton's ColorFAQ at |
| 1636 | * <http://www.inforamp.net/~poynton/> |
| 1637 | * Copyright (c) 1998-01-04 Charles Poynton poynton at inforamp.net |
| 1638 | * |
| 1639 | * Y = 0.212671 * R + 0.715160 * G + 0.072169 * B |
| 1640 | * |
| 1641 | * We approximate this with |
| 1642 | * |
| 1643 | * Y = 0.21268 * R + 0.7151 * G + 0.07217 * B |
| 1644 | * |
| 1645 | * which can be expressed with integers as |
| 1646 | * |
| 1647 | * Y = (6969 * R + 23434 * G + 2365 * B)/32768 |
| 1648 | * |
| 1649 | * The calculation is to be done in a linear colorspace. |
| 1650 | */ |
| 1651 | *dest = (((int)((((*src >> 16) & 0xFF) * 6969 + |
| 1652 | ((*src >> 8) & 0xFF) * 23434 + |
| 1653 | ((*src ) & 0xFF) * 2365) / 32768) & 0xFF) | |
| 1654 | (uint16)((*src >> 16) & 0xFF00)); |
| 1655 | #else |
| 1656 | /* 3dfx style Intensity = R * 0.299 + G * 0.587 + B * 0.114 |
| 1657 | * this is same as the standard NTSC gray scale conversion. */ |
| 1658 | *dest = (((int)((((*src >> 16) & 0xFF) * 299 + |
| 1659 | ((*src >> 8) & 0xFF) * 587 + |
| 1660 | ((*src ) & 0xFF) * 114) / 1000) & 0xFF) | |
| 1661 | (uint16)((*src >> 16) & 0xFF00)); |
| 1662 | #endif |
| 1663 | dest++; |
| 1664 | src++; |
| 1665 | } |
| 1666 | } |
| 1667 | } |
| 1668 | |
| 1669 | void |
| 1670 | TxQuantize::ARGB8888_I8_Slow(uint32* src, uint32* dst, int width, int height) |
| 1671 | { |
| 1672 | int x, y; |
| 1673 | uint8 *dest = (uint8 *)dst; |
| 1674 | for (y = 0; y < height; y++) { |
| 1675 | for (x = 0; x < width; x++) { |
| 1676 | #if 1 |
| 1677 | /* libpng style Intensity = (6969 * R + 23434 * G + 2365 * B)/32768 */ |
| 1678 | *dest = (int)((((*src >> 16) & 0xFF) * 6969 + |
| 1679 | ((*src >> 8) & 0xFF) * 23434 + |
| 1680 | ((*src ) & 0xFF) * 2365) / 32768) & 0xFF; |
| 1681 | #else |
| 1682 | /* 3dfx style Intensity = R * 0.299 + G * 0.587 + B * 0.114 |
| 1683 | * this is same as the standard NTSC gray scale conversion. */ |
| 1684 | *dest = (int)((((*src >>16) & 0xFF) * 299 + |
| 1685 | ((*src >> 8) & 0xFF) * 587 + |
| 1686 | ((*src ) & 0xFF) * 114) / 1000) & 0xFF; |
| 1687 | #endif |
| 1688 | dest++; |
| 1689 | src++; |
| 1690 | } |
| 1691 | } |
| 1692 | } |
| 1693 | |
| 1694 | void |
| 1695 | TxQuantize::P8_16BPP(uint32* src, uint32* dest, int width, int height, uint32* palette) |
| 1696 | { |
| 1697 | /* passed in palette is RGBA5551 format */ |
| 1698 | #if 1 |
| 1699 | int i; |
| 1700 | int size = width * height; |
| 1701 | for (i = 0; i < size; i++) { |
| 1702 | ((uint16*)dest)[i] = ((uint16*)palette)[(int)(((uint8*)src)[i])]; |
| 1703 | ((uint16*)dest)[i] = ((((uint16*)dest)[i] << 15) | (((uint16*)dest)[i] >> 1)); |
| 1704 | } |
| 1705 | #else |
| 1706 | |
| 1707 | /* not finished yet... */ |
| 1708 | |
| 1709 | int siz = (width * height) >> 2; |
| 1710 | |
| 1711 | __asm { |
| 1712 | push ebx; |
| 1713 | push esi; |
| 1714 | push edi; |
| 1715 | |
| 1716 | mov esi, dword ptr [src]; |
| 1717 | mov edi, dword ptr [dest]; |
| 1718 | mov ecx, dword ptr [siz]; |
| 1719 | mov edx, dword ptr [palette]; |
| 1720 | |
| 1721 | tc1_loop: |
| 1722 | mov eax, dword ptr [esi]; |
| 1723 | add esi, 4; |
| 1724 | |
| 1725 | dec ecx; |
| 1726 | jnz tc1_loop; |
| 1727 | |
| 1728 | pop edi; |
| 1729 | pop esi; |
| 1730 | pop ebx; |
| 1731 | } |
| 1732 | #endif |
| 1733 | } |
| 1734 | |
| 1735 | boolean |
| 1736 | TxQuantize::quantize(uint8* src, uint8* dest, int width, int height, uint16 srcformat, uint16 destformat, boolean fastQuantizer) |
| 1737 | { |
| 1738 | typedef void (TxQuantize::*quantizerFunc)(uint32* src, uint32* dest, int width, int height); |
| 1739 | quantizerFunc quantizer; |
| 1740 | int bpp_shift = 0; |
| 1741 | |
| 1742 | if (destformat == GR_TEXFMT_ARGB_8888) { |
| 1743 | switch (srcformat) { |
| 1744 | case GR_TEXFMT_ARGB_1555: |
| 1745 | quantizer = &TxQuantize::ARGB1555_ARGB8888; |
| 1746 | bpp_shift = 1; |
| 1747 | break; |
| 1748 | case GR_TEXFMT_ARGB_4444: |
| 1749 | quantizer = &TxQuantize::ARGB4444_ARGB8888; |
| 1750 | bpp_shift = 1; |
| 1751 | break; |
| 1752 | case GR_TEXFMT_RGB_565: |
| 1753 | quantizer = &TxQuantize::RGB565_ARGB8888; |
| 1754 | bpp_shift = 1; |
| 1755 | break; |
| 1756 | case GR_TEXFMT_ALPHA_8: |
| 1757 | quantizer = &TxQuantize::A8_ARGB8888; |
| 1758 | bpp_shift = 2; |
| 1759 | break; |
| 1760 | case GR_TEXFMT_ALPHA_INTENSITY_44: |
| 1761 | quantizer = &TxQuantize::AI44_ARGB8888; |
| 1762 | bpp_shift = 2; |
| 1763 | break; |
| 1764 | case GR_TEXFMT_ALPHA_INTENSITY_88: |
| 1765 | quantizer = &TxQuantize::AI88_ARGB8888; |
| 1766 | bpp_shift = 1; |
| 1767 | break; |
| 1768 | default: |
| 1769 | return 0; |
| 1770 | } |
| 1771 | |
| 1772 | #if !defined(NO_FILTER_THREAD) |
| 1773 | unsigned int numcore = _numcore; |
| 1774 | unsigned int blkrow = 0; |
| 1775 | while (numcore > 1 && blkrow == 0) { |
| 1776 | blkrow = (height >> 2) / numcore; |
| 1777 | numcore--; |
| 1778 | } |
| 1779 | if (blkrow > 0 && numcore > 1) { |
| 1780 | std::thread *thrd[MAX_NUMCORE]; |
| 1781 | unsigned int i; |
| 1782 | int blkheight = blkrow << 2; |
| 1783 | unsigned int srcStride = (width * blkheight) << (2 - bpp_shift); |
| 1784 | unsigned int destStride = srcStride << bpp_shift; |
| 1785 | for (i = 0; i < numcore - 1; i++) { |
| 1786 | thrd[i] = new std::thread(std::bind(quantizer, |
| 1787 | this, |
| 1788 | (uint32*)src, |
| 1789 | (uint32*)dest, |
| 1790 | width, |
| 1791 | blkheight)); |
| 1792 | src += srcStride; |
| 1793 | dest += destStride; |
| 1794 | } |
| 1795 | thrd[i] = new std::thread(std::bind(quantizer, |
| 1796 | this, |
| 1797 | (uint32*)src, |
| 1798 | (uint32*)dest, |
| 1799 | width, |
| 1800 | height - blkheight * i)); |
| 1801 | for (i = 0; i < numcore; i++) { |
| 1802 | thrd[i]->join(); |
| 1803 | delete thrd[i]; |
| 1804 | } |
| 1805 | } else { |
| 1806 | (*this.*quantizer)((uint32*)src, (uint32*)dest, width, height); |
| 1807 | } |
| 1808 | #else |
| 1809 | (*this.*quantizer)((uint32*)src, (uint32*)dest, width, height); |
| 1810 | #endif |
| 1811 | |
| 1812 | } else if (srcformat == GR_TEXFMT_ARGB_8888) { |
| 1813 | switch (destformat) { |
| 1814 | case GR_TEXFMT_ARGB_1555: |
| 1815 | quantizer = fastQuantizer ? &TxQuantize::ARGB8888_ARGB1555 : &TxQuantize::ARGB8888_ARGB1555_ErrD; |
| 1816 | bpp_shift = 1; |
| 1817 | break; |
| 1818 | case GR_TEXFMT_ARGB_4444: |
| 1819 | quantizer = fastQuantizer ? &TxQuantize::ARGB8888_ARGB4444 : &TxQuantize::ARGB8888_ARGB4444_ErrD; |
| 1820 | bpp_shift = 1; |
| 1821 | break; |
| 1822 | case GR_TEXFMT_RGB_565: |
| 1823 | quantizer = fastQuantizer ? &TxQuantize::ARGB8888_RGB565 : &TxQuantize::ARGB8888_RGB565_ErrD; |
| 1824 | bpp_shift = 1; |
| 1825 | break; |
| 1826 | case GR_TEXFMT_ALPHA_8: |
| 1827 | case GR_TEXFMT_INTENSITY_8: |
| 1828 | quantizer = fastQuantizer ? &TxQuantize::ARGB8888_A8 : &TxQuantize::ARGB8888_I8_Slow; |
| 1829 | bpp_shift = 2; |
| 1830 | break; |
| 1831 | case GR_TEXFMT_ALPHA_INTENSITY_44: |
| 1832 | quantizer = fastQuantizer ? &TxQuantize::ARGB8888_AI44 : &TxQuantize::ARGB8888_AI44_ErrD; |
| 1833 | bpp_shift = 2; |
| 1834 | break; |
| 1835 | case GR_TEXFMT_ALPHA_INTENSITY_88: |
| 1836 | quantizer = fastQuantizer ? &TxQuantize::ARGB8888_AI88 : &TxQuantize::ARGB8888_AI88_Slow; |
| 1837 | bpp_shift = 1; |
| 1838 | break; |
| 1839 | default: |
| 1840 | return 0; |
| 1841 | } |
| 1842 | |
| 1843 | #if !defined(NO_FILTER_THREAD) |
| 1844 | unsigned int numcore = _numcore; |
| 1845 | unsigned int blkrow = 0; |
| 1846 | while (numcore > 1 && blkrow == 0) { |
| 1847 | blkrow = (height >> 2) / numcore; |
| 1848 | numcore--; |
| 1849 | } |
| 1850 | if (blkrow > 0 && numcore > 1) { |
| 1851 | std::thread *thrd[MAX_NUMCORE]; |
| 1852 | unsigned int i; |
| 1853 | int blkheight = blkrow << 2; |
| 1854 | unsigned int srcStride = (width * blkheight) << 2; |
| 1855 | unsigned int destStride = srcStride >> bpp_shift; |
| 1856 | for (i = 0; i < numcore - 1; i++) { |
| 1857 | thrd[i] = new std::thread(std::bind(quantizer, |
| 1858 | this, |
| 1859 | (uint32*)src, |
| 1860 | (uint32*)dest, |
| 1861 | width, |
| 1862 | blkheight)); |
| 1863 | src += srcStride; |
| 1864 | dest += destStride; |
| 1865 | } |
| 1866 | thrd[i] = new std::thread(std::bind(quantizer, |
| 1867 | this, |
| 1868 | (uint32*)src, |
| 1869 | (uint32*)dest, |
| 1870 | width, |
| 1871 | height - blkheight * i)); |
| 1872 | for (i = 0; i < numcore; i++) { |
| 1873 | thrd[i]->join(); |
| 1874 | delete thrd[i]; |
| 1875 | } |
| 1876 | } else { |
| 1877 | (*this.*quantizer)((uint32*)src, (uint32*)dest, width, height); |
| 1878 | } |
| 1879 | #else |
| 1880 | (*this.*quantizer)((uint32*)src, (uint32*)dest, width, height); |
| 1881 | #endif |
| 1882 | |
| 1883 | } else { |
| 1884 | return 0; |
| 1885 | } |
| 1886 | |
| 1887 | return 1; |
| 1888 | } |
| 1889 | |
| 1890 | boolean |
| 1891 | TxQuantize::FXT1(uint8 *src, uint8 *dest, |
| 1892 | int srcwidth, int srcheight, uint16 srcformat, |
| 1893 | int *destwidth, int *destheight, uint16 *destformat) |
| 1894 | { |
| 1895 | /* |
| 1896 | * NOTE: src must be in ARGB8888 format, srcformat describes |
| 1897 | * the closest 16bbp representation of src. |
| 1898 | * |
| 1899 | * NOTE: I have modified the dxtn library to use ARGB format |
| 1900 | * which originaly was ABGR format. |
| 1901 | */ |
| 1902 | |
| 1903 | boolean bRet = 0; |
| 1904 | |
| 1905 | if (_tx_compress_fxt1 && |
| 1906 | srcwidth >= 8 && srcheight >= 4) { |
| 1907 | /* compress to fxt1 |
| 1908 | * width and height must be larger than 8 and 4 respectively |
| 1909 | */ |
| 1910 | int dstRowStride = ((srcwidth + 7) & ~7) << 1; |
| 1911 | int srcRowStride = (srcwidth << 2); |
| 1912 | |
| 1913 | #if !defined(NO_FILTER_THREAD) |
| 1914 | unsigned int numcore = _numcore; |
| 1915 | unsigned int blkrow = 0; |
| 1916 | while (numcore > 1 && blkrow == 0) { |
| 1917 | blkrow = (srcheight >> 2) / numcore; |
| 1918 | numcore--; |
| 1919 | } |
| 1920 | if (blkrow > 0 && numcore > 1) { |
| 1921 | std::thread *thrd[MAX_NUMCORE]; |
| 1922 | unsigned int i; |
| 1923 | int blkheight = blkrow << 2; |
| 1924 | unsigned int srcStride = (srcwidth * blkheight) << 2; |
| 1925 | unsigned int destStride = dstRowStride * blkrow; |
| 1926 | for (i = 0; i < numcore - 1; i++) { |
| 1927 | thrd[i] = new std::thread(std::bind(_tx_compress_fxt1, |
| 1928 | srcwidth, |
| 1929 | blkheight, |
| 1930 | 4, |
| 1931 | src, |
| 1932 | srcRowStride, |
| 1933 | dest, |
| 1934 | dstRowStride)); |
| 1935 | src += srcStride; |
| 1936 | dest += destStride; |
| 1937 | } |
| 1938 | thrd[i] = new std::thread(std::bind(_tx_compress_fxt1, |
| 1939 | srcwidth, |
| 1940 | srcheight - blkheight * i, |
| 1941 | 4, |
| 1942 | src, |
| 1943 | srcRowStride, |
| 1944 | dest, |
| 1945 | dstRowStride)); |
| 1946 | for (i = 0; i < numcore; i++) { |
| 1947 | thrd[i]->join(); |
| 1948 | delete thrd[i]; |
| 1949 | } |
| 1950 | } else { |
| 1951 | (*_tx_compress_fxt1)(srcwidth, /* width */ |
| 1952 | srcheight, /* height */ |
| 1953 | 4, /* comps: ARGB8888=4, RGB888=3 */ |
| 1954 | src, /* source */ |
| 1955 | srcRowStride, /* width*comps */ |
| 1956 | dest, /* destination */ |
| 1957 | dstRowStride); /* 16 bytes per 8x4 texel */ |
| 1958 | } |
| 1959 | #else |
| 1960 | (*_tx_compress_fxt1)(srcwidth, /* width */ |
| 1961 | srcheight, /* height */ |
| 1962 | 4, /* comps: ARGB8888=4, RGB888=3 */ |
| 1963 | src, /* source */ |
| 1964 | srcRowStride, /* width*comps */ |
| 1965 | dest, /* destination */ |
| 1966 | dstRowStride); /* 16 bytes per 8x4 texel */ |
| 1967 | #endif |
| 1968 | |
| 1969 | /* dxtn adjusts width and height to M8 and M4 respectively by replication */ |
| 1970 | *destwidth = (srcwidth + 7) & ~7; |
| 1971 | *destheight = (srcheight + 3) & ~3; |
| 1972 | *destformat = GR_TEXFMT_ARGB_CMP_FXT1; |
| 1973 | |
| 1974 | bRet = 1; |
| 1975 | } |
| 1976 | |
| 1977 | return bRet; |
| 1978 | } |
| 1979 | |
| 1980 | boolean |
| 1981 | TxQuantize::DXTn(uint8 *src, uint8 *dest, |
| 1982 | int srcwidth, int srcheight, uint16 srcformat, |
| 1983 | int *destwidth, int *destheight, uint16 *destformat) |
| 1984 | { |
| 1985 | /* |
| 1986 | * NOTE: src must be in ARGB8888 format, srcformat describes |
| 1987 | * the closest 16bbp representation of src. |
| 1988 | * |
| 1989 | * NOTE: I have modified the dxtn library to use ARGB format |
| 1990 | * which originaly was ABGR format. |
| 1991 | */ |
| 1992 | |
| 1993 | boolean bRet = 0; |
| 1994 | |
| 1995 | if (_tx_compress_dxtn_rgba && |
| 1996 | srcwidth >= 4 && srcheight >= 4) { |
| 1997 | /* compress to dxtn |
| 1998 | * width and height must be larger than 4 |
| 1999 | */ |
| 2000 | |
| 2001 | /* skip formats that DXTn won't help in size. */ |
| 2002 | if (srcformat == GR_TEXFMT_ALPHA_8 || |
| 2003 | srcformat == GR_TEXFMT_ALPHA_INTENSITY_44) { |
| 2004 | ; /* shutup compiler */ |
| 2005 | } else { |
| 2006 | int dstRowStride = ((srcwidth + 3) & ~3) << 2; |
| 2007 | int compression = GL_COMPRESSED_RGBA_S3TC_DXT5_EXT; |
| 2008 | |
| 2009 | *destformat = GR_TEXFMT_ARGB_CMP_DXT5; |
| 2010 | |
| 2011 | #if !GLIDE64_DXTN |
| 2012 | /* okay... we are going to disable DXT1 with 1bit alpha |
| 2013 | * for Glide64. some textures have all 0 alpha values. |
| 2014 | * see "N64 Kobe Bryant in NBA Courtside" |
| 2015 | */ |
| 2016 | if (srcformat == GR_TEXFMT_ARGB_1555) { |
| 2017 | dstRowStride >>= 1; |
| 2018 | compression = GL_COMPRESSED_RGBA_S3TC_DXT1_EXT; |
| 2019 | *destformat = GR_TEXFMT_ARGB_CMP_DXT1; |
| 2020 | } else |
| 2021 | #endif |
| 2022 | if (srcformat == GR_TEXFMT_RGB_565 || |
| 2023 | srcformat == GR_TEXFMT_INTENSITY_8) { |
| 2024 | dstRowStride >>= 1; |
| 2025 | compression = GL_COMPRESSED_RGB_S3TC_DXT1_EXT; |
| 2026 | *destformat = GR_TEXFMT_ARGB_CMP_DXT1; |
| 2027 | } |
| 2028 | |
| 2029 | #if !defined(NO_FILTER_THREAD) |
| 2030 | unsigned int numcore = _numcore; |
| 2031 | unsigned int blkrow = 0; |
| 2032 | while (numcore > 1 && blkrow == 0) { |
| 2033 | blkrow = (srcheight >> 2) / numcore; |
| 2034 | numcore--; |
| 2035 | } |
| 2036 | if (blkrow > 0 && numcore > 1) { |
| 2037 | std::thread *thrd[MAX_NUMCORE]; |
| 2038 | unsigned int i; |
| 2039 | int blkheight = blkrow << 2; |
| 2040 | unsigned int srcStride = (srcwidth * blkheight) << 2; |
| 2041 | unsigned int destStride = dstRowStride * blkrow; |
| 2042 | for (i = 0; i < numcore - 1; i++) { |
| 2043 | thrd[i] = new std::thread(std::bind(_tx_compress_dxtn_rgba, |
| 2044 | 4, |
| 2045 | srcwidth, |
| 2046 | blkheight, |
| 2047 | src, |
| 2048 | compression, |
| 2049 | dest, |
| 2050 | dstRowStride)); |
| 2051 | src += srcStride; |
| 2052 | dest += destStride; |
| 2053 | } |
| 2054 | thrd[i] = new std::thread(std::bind(_tx_compress_dxtn_rgba, |
| 2055 | 4, |
| 2056 | srcwidth, |
| 2057 | srcheight - blkheight * i, |
| 2058 | src, |
| 2059 | compression, |
| 2060 | dest, |
| 2061 | dstRowStride)); |
| 2062 | for (i = 0; i < numcore; i++) { |
| 2063 | thrd[i]->join(); |
| 2064 | delete thrd[i]; |
| 2065 | } |
| 2066 | } else { |
| 2067 | (*_tx_compress_dxtn_rgba)(4, /* comps: ARGB8888=4, RGB888=3 */ |
| 2068 | srcwidth, /* width */ |
| 2069 | srcheight, /* height */ |
| 2070 | src, /* source */ |
| 2071 | compression, /* format */ |
| 2072 | dest, /* destination */ |
| 2073 | dstRowStride); /* DXT1 = 8 bytes per 4x4 texel |
| 2074 | * others = 16 bytes per 4x4 texel */ |
| 2075 | } |
| 2076 | #else |
| 2077 | (*_tx_compress_dxtn_rgba)(4, /* comps: ARGB8888=4, RGB888=3 */ |
| 2078 | srcwidth, /* width */ |
| 2079 | srcheight, /* height */ |
| 2080 | src, /* source */ |
| 2081 | compression, /* format */ |
| 2082 | dest, /* destination */ |
| 2083 | dstRowStride); /* DXT1 = 8 bytes per 4x4 texel |
| 2084 | * others = 16 bytes per 4x4 texel */ |
| 2085 | #endif |
| 2086 | |
| 2087 | /* dxtn adjusts width and height to M4 by replication */ |
| 2088 | *destwidth = (srcwidth + 3) & ~3; |
| 2089 | *destheight = (srcheight + 3) & ~3; |
| 2090 | |
| 2091 | bRet = 1; |
| 2092 | } |
| 2093 | } |
| 2094 | |
| 2095 | return bRet; |
| 2096 | } |
| 2097 | |
| 2098 | boolean |
| 2099 | TxQuantize::compress(uint8 *src, uint8 *dest, |
| 2100 | int srcwidth, int srcheight, uint16 srcformat, |
| 2101 | int *destwidth, int *destheight, uint16 *destformat, |
| 2102 | int compressionType) |
| 2103 | { |
| 2104 | boolean bRet = 0; |
| 2105 | |
| 2106 | switch (compressionType) { |
| 2107 | case FXT1_COMPRESSION: |
| 2108 | bRet = FXT1(src, dest, |
| 2109 | srcwidth, srcheight, srcformat, |
| 2110 | destwidth, destheight, destformat); |
| 2111 | break; |
| 2112 | case S3TC_COMPRESSION: |
| 2113 | bRet = DXTn(src, dest, |
| 2114 | srcwidth, srcheight, srcformat, |
| 2115 | destwidth, destheight, destformat); |
| 2116 | break; |
| 2117 | case NCC_COMPRESSION: |
| 2118 | /* TODO: narrow channel compression */ |
| 2119 | ; |
| 2120 | } |
| 2121 | |
| 2122 | return bRet; |
| 2123 | } |
| 2124 | |
| 2125 | #if 0 /* unused */ |
| 2126 | void |
| 2127 | TxQuantize::I8_ARGB8888(uint32* src, uint32* dest, int width, int height) |
| 2128 | { |
| 2129 | int siz = (width * height) >> 2; |
| 2130 | |
| 2131 | __asm { |
| 2132 | push ebx; |
| 2133 | push esi; |
| 2134 | push edi; |
| 2135 | |
| 2136 | mov esi, dword ptr [src]; |
| 2137 | mov edi, dword ptr [dest]; |
| 2138 | mov ecx, dword ptr [siz]; |
| 2139 | |
| 2140 | tc1_loop: |
| 2141 | mov eax, dword ptr [esi]; |
| 2142 | add esi, 4; |
| 2143 | |
| 2144 | // aaaaaaaa |
| 2145 | // 11111111 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2146 | mov edx, eax; |
| 2147 | and eax, 0x000000ff; |
| 2148 | mov ebx, eax; // 00000000 00000000 00000000 aaaaaaaa |
| 2149 | shl ebx, 8; // 00000000 00000000 aaaaaaaa 00000000 |
| 2150 | or eax, ebx; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 2151 | shl ebx, 8; // 00000000 aaaaaaaa 00000000 00000000 |
| 2152 | or eax, ebx; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2153 | or eax, 0xff000000; // 11111111 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2154 | |
| 2155 | mov dword ptr [edi], eax; |
| 2156 | add edi, 4; |
| 2157 | |
| 2158 | mov eax, edx; |
| 2159 | and eax, 0x0000ff00; |
| 2160 | mov ebx, eax; // 00000000 00000000 aaaaaaaa 00000000 |
| 2161 | shr ebx, 8; // 00000000 00000000 00000000 aaaaaaaa |
| 2162 | or eax, ebx; // 00000000 00000000 aaaaaaaa aaaaaaaa |
| 2163 | shl ebx, 16; // 00000000 aaaaaaaa 00000000 00000000 |
| 2164 | or eax, ebx; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2165 | or eax, 0xff000000; // 11111111 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2166 | |
| 2167 | mov dword ptr [edi], eax; |
| 2168 | add edi, 4; |
| 2169 | |
| 2170 | mov eax, edx; |
| 2171 | and eax, 0x00ff0000; |
| 2172 | mov ebx, eax; // 00000000 aaaaaaaa 00000000 00000000 |
| 2173 | shr ebx, 8; // 00000000 00000000 aaaaaaaa 00000000 |
| 2174 | or eax, ebx; // 00000000 aaaaaaaa aaaaaaaa 00000000 |
| 2175 | shr ebx, 8; // 00000000 00000000 00000000 aaaaaaaa |
| 2176 | or eax, ebx; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2177 | or eax, 0xff000000; // 11111111 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2178 | |
| 2179 | mov dword ptr [edi], eax; |
| 2180 | add edi, 4; |
| 2181 | |
| 2182 | mov eax, edx; |
| 2183 | and eax, 0xff000000; |
| 2184 | mov ebx, eax; // aaaaaaaa 00000000 00000000 00000000 |
| 2185 | shr ebx, 8; // 00000000 aaaaaaaa 00000000 00000000 |
| 2186 | or eax, ebx; // aaaaaaaa aaaaaaaa 00000000 00000000 |
| 2187 | shr ebx, 8; // 00000000 00000000 aaaaaaaa 00000000 |
| 2188 | or eax, ebx; // aaaaaaaa aaaaaaaa aaaaaaaa 00000000 |
| 2189 | shr eax, 8; // 00000000 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2190 | or eax, 0xff000000; // 11111111 aaaaaaaa aaaaaaaa aaaaaaaa |
| 2191 | |
| 2192 | mov dword ptr [edi], eax; |
| 2193 | add edi, 4; |
| 2194 | |
| 2195 | dec ecx; |
| 2196 | jnz tc1_loop; |
| 2197 | |
| 2198 | pop edi; |
| 2199 | pop esi; |
| 2200 | pop ebx; |
| 2201 | } |
| 2202 | } |
| 2203 | |
| 2204 | void |
| 2205 | TxQuantize::ARGB8888_I8(uint32* src, uint32* dest, int width, int height) |
| 2206 | { |
| 2207 | ARGB8888_A8(src, dest, width, height); |
| 2208 | } |
| 2209 | |
| 2210 | void |
| 2211 | TxQuantize::ARGB1555_ABGR8888(uint32* src, uint32* dest, int width, int height) |
| 2212 | { |
| 2213 | int siz = (width * height) >> 1; |
| 2214 | |
| 2215 | __asm { |
| 2216 | push ebx; |
| 2217 | push esi; |
| 2218 | push edi; |
| 2219 | |
| 2220 | mov esi, dword ptr [src]; |
| 2221 | mov edi, dword ptr [dest]; |
| 2222 | mov ecx, dword ptr [siz]; |
| 2223 | |
| 2224 | tc1_loop: |
| 2225 | mov eax, dword ptr [esi]; |
| 2226 | add esi, 4; |
| 2227 | |
| 2228 | // arrr rrgg gggb bbbb |
| 2229 | // aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2230 | mov edx, eax; // edx = arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 2231 | and ebx, 0x00000000; |
| 2232 | and eax, 0x00008000; // eax = 00000000 00000000 a0000000 00000000 |
| 2233 | jz transparent1; |
| 2234 | or ebx, 0xff000000; // ebx = aaaaaaaa 00000000 00000000 00000000 |
| 2235 | |
| 2236 | transparent1: |
| 2237 | mov eax, edx; // eax = arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 2238 | and edx, 0x0000001f; // edx = 00000000 00000000 00000000 000bbbbb |
| 2239 | shl edx, 14; // edx = 00000000 00000bbb bb000000 00000000 |
| 2240 | or ebx, edx; // ebx = aaaaaaaa 00000bbb bb000000 00000000 |
| 2241 | shl edx, 5; // edx = 00000000 bbbbb000 00000000 00000000 |
| 2242 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb bb000000 00000000 |
| 2243 | and ebx, 0xffff0000; // ebx = aaaaaaaa bbbbbbbb 00000000 00000000 |
| 2244 | mov edx, eax; |
| 2245 | and edx, 0x000003e0; // edx = 00000000 00000000 000000gg ggg00000 |
| 2246 | shl edx, 1; // edx = 00000000 00000000 00000ggg gg000000 |
| 2247 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb 00000ggg gg000000 |
| 2248 | shl edx, 5; // edx = 00000000 00000000 ggggg000 00000000 |
| 2249 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb gggggggg gg000000 |
| 2250 | and ebx, 0xffffff00; // ebx = aaaaaaaa bbbbbbbb gggggggg 00000000 |
| 2251 | mov edx, eax; |
| 2252 | and edx, 0x00007c00; // edx = 00000000 00000000 0rrrrr00 00000000 |
| 2253 | shr edx, 7; // edx = 00000000 00000000 00000000 rrrrr000 |
| 2254 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb gggggggg rrrrr000 |
| 2255 | shr edx, 5; // edx = 00000000 00000000 00000000 00000rrr |
| 2256 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2257 | |
| 2258 | mov dword ptr [edi], ebx; |
| 2259 | add edi, 4; |
| 2260 | |
| 2261 | shr eax, 16; // eax = 00000000 00000000 arrrrrgg gggbbbbb |
| 2262 | mov edx, eax; // edx = 00000000 00000000 arrrrrgg gggbbbbb |
| 2263 | and ebx, 0x00000000; |
| 2264 | and eax, 0x00008000; // eax = 00000000 00000000 a0000000 00000000 |
| 2265 | jz transparent2; |
| 2266 | or ebx, 0xff000000; // ebx = aaaaaaaa 00000000 00000000 00000000 |
| 2267 | |
| 2268 | transparent2: |
| 2269 | mov eax, edx; // eax = arrrrrgg gggbbbbb arrrrrgg gggbbbbb |
| 2270 | and edx, 0x0000001f; // edx = 00000000 00000000 00000000 000bbbbb |
| 2271 | shl edx, 14; // edx = 00000000 00000bbb bb000000 00000000 |
| 2272 | or ebx, edx; // ebx = aaaaaaaa 00000bbb bb000000 00000000 |
| 2273 | shl edx, 5; // edx = 00000000 bbbbb000 00000000 00000000 |
| 2274 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb bb000000 00000000 |
| 2275 | and ebx, 0xffff0000; // ebx = aaaaaaaa bbbbbbbb 00000000 00000000 |
| 2276 | mov edx, eax; |
| 2277 | and edx, 0x000003e0; // edx = 00000000 00000000 000000gg ggg00000 |
| 2278 | shl edx, 1; // edx = 00000000 00000000 00000ggg gg000000 |
| 2279 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb 00000ggg gg000000 |
| 2280 | shl edx, 5; // edx = 00000000 00000000 ggggg000 00000000 |
| 2281 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb gggggggg gg000000 |
| 2282 | and ebx, 0xffffff00; // ebx = aaaaaaaa bbbbbbbb gggggggg 00000000 |
| 2283 | mov edx, eax; |
| 2284 | and edx, 0x00007c00; // edx = 00000000 00000000 0rrrrr00 00000000 |
| 2285 | shr edx, 7; // edx = 00000000 00000000 00000000 rrrrr000 |
| 2286 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb gggggggg rrrrr000 |
| 2287 | shr edx, 5; // edx = 00000000 00000000 00000000 00000rrr |
| 2288 | or ebx, edx; // ebx = aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2289 | |
| 2290 | mov dword ptr [edi], ebx; |
| 2291 | add edi, 4; |
| 2292 | |
| 2293 | dec ecx; |
| 2294 | jnz tc1_loop; |
| 2295 | |
| 2296 | pop edi; |
| 2297 | pop esi; |
| 2298 | pop ebx; |
| 2299 | } |
| 2300 | } |
| 2301 | |
| 2302 | void |
| 2303 | TxQuantize::ARGB4444_ABGR8888(uint32* src, uint32* dest, int width, int height) |
| 2304 | { |
| 2305 | int siz = (width * height) >> 1; |
| 2306 | |
| 2307 | __asm { |
| 2308 | push ebx; |
| 2309 | push esi; |
| 2310 | push edi; |
| 2311 | |
| 2312 | mov esi, dword ptr [src]; |
| 2313 | mov edi, dword ptr [dest]; |
| 2314 | mov ecx, dword ptr [siz]; |
| 2315 | |
| 2316 | tc1_loop: |
| 2317 | mov eax, dword ptr [esi]; |
| 2318 | add esi, 4; |
| 2319 | |
| 2320 | // aaaa rrrr gggg bbbb |
| 2321 | // aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2322 | mov edx, eax; |
| 2323 | and eax, 0x0000ffff; |
| 2324 | mov ebx, eax; // 00000000 00000000 aaaarrrr ggggbbbb |
| 2325 | and ebx, 0x0000f000; // 00000000 00000000 aaaa0000 00000000 |
| 2326 | shl ebx, 12; // 0000aaaa 00000000 00000000 00000000 |
| 2327 | or eax, ebx; // 0000aaaa 00000000 aaaarrrr ggggbbbb |
| 2328 | mov ebx, eax; |
| 2329 | and ebx, 0x0000000f; // 00000000 00000000 00000000 0000bbbb |
| 2330 | shl ebx, 16; // 00000000 0000bbbb 00000000 00000000 |
| 2331 | or eax, ebx; // 0000aaaa 0000bbbb aaaarrrr ggggbbbb |
| 2332 | mov ebx, eax; |
| 2333 | and ebx, 0x00000f00; // 00000000 00000000 0000rrrr 00000000 |
| 2334 | shr ebx, 8; // 00000000 00000000 00000000 0000rrrr |
| 2335 | and eax, 0xfffffff0; |
| 2336 | or eax, ebx; // 0000aaaa 0000bbbb aaaarrrr ggggrrrr |
| 2337 | mov ebx, eax; |
| 2338 | and ebx, 0x000000f0; // 00000000 00000000 00000000 gggg0000 |
| 2339 | shl ebx, 4; // 00000000 00000000 0000gggg 00000000 |
| 2340 | and eax, 0x0f0f000f; // 0000aaaa 0000bbbb 00000000 0000rrrr |
| 2341 | or eax, ebx; // 0000aaaa 0000bbbb 0000gggg 0000rrrr |
| 2342 | mov ebx, eax; |
| 2343 | shl ebx, 4; // aaaa0000 bbbb0000 gggg0000 rrrr0000 |
| 2344 | or eax, ebx; // aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2345 | |
| 2346 | mov dword ptr [edi], eax; |
| 2347 | |
| 2348 | add edi, 4; |
| 2349 | |
| 2350 | shr edx, 16; |
| 2351 | mov ebx, edx; // 00000000 00000000 aaaarrrr ggggbbbb |
| 2352 | and ebx, 0x0000f000; // 00000000 00000000 aaaa0000 00000000 |
| 2353 | shl ebx, 12; // 0000aaaa 00000000 00000000 00000000 |
| 2354 | or edx, ebx; // 0000aaaa 00000000 aaaarrrr ggggbbbb |
| 2355 | mov ebx, edx; |
| 2356 | and ebx, 0x0000000f; // 00000000 00000000 00000000 0000bbbb |
| 2357 | shl ebx, 16; // 00000000 0000bbbb 00000000 00000000 |
| 2358 | or edx, ebx; // 0000aaaa 0000bbbb aaaarrrr ggggbbbb |
| 2359 | mov ebx, edx; |
| 2360 | and ebx, 0x00000f00; // 00000000 00000000 0000rrrr 00000000 |
| 2361 | shr ebx, 8; // 00000000 00000000 00000000 0000rrrr |
| 2362 | and edx, 0xfffffff0; |
| 2363 | or edx, ebx; // 0000aaaa 0000bbbb aaaarrrr ggggrrrr |
| 2364 | mov ebx, edx; |
| 2365 | and ebx, 0x000000f0; // 00000000 00000000 00000000 gggg0000 |
| 2366 | shl ebx, 4; // 00000000 00000000 0000gggg 00000000 |
| 2367 | and edx, 0x0f0f000f; // 0000aaaa 0000bbbb 00000000 0000rrrr |
| 2368 | or edx, ebx; // 0000aaaa 0000bbbb 0000gggg 0000rrrr |
| 2369 | mov ebx, edx; |
| 2370 | shl ebx, 4; // aaaa0000 bbbb0000 gggg0000 rrrr0000 |
| 2371 | or edx, ebx; // aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2372 | |
| 2373 | mov dword ptr [edi], edx; |
| 2374 | add edi, 4; |
| 2375 | |
| 2376 | dec ecx; |
| 2377 | jnz tc1_loop; |
| 2378 | |
| 2379 | pop edi; |
| 2380 | pop esi; |
| 2381 | pop ebx; |
| 2382 | } |
| 2383 | } |
| 2384 | |
| 2385 | void |
| 2386 | TxQuantize::ARGB8888_ABGR8888(uint32* src, uint32* dest, int width, int height) |
| 2387 | { |
| 2388 | int siz = width * height; |
| 2389 | |
| 2390 | __asm { |
| 2391 | push ebx; |
| 2392 | push esi; |
| 2393 | push edi; |
| 2394 | |
| 2395 | mov esi, dword ptr [src]; |
| 2396 | mov edi, dword ptr [dest]; |
| 2397 | mov ecx, dword ptr [siz]; |
| 2398 | |
| 2399 | tc1_loop: |
| 2400 | mov eax, dword ptr [esi]; |
| 2401 | add esi, 4; |
| 2402 | |
| 2403 | // aaaaaaaa bbbbbbbb gggggggg rrrrrrrr |
| 2404 | mov edx, eax; |
| 2405 | bswap edx; |
| 2406 | shr edx, 8; |
| 2407 | and eax, 0xff000000; |
| 2408 | |
| 2409 | or eax, edx; |
| 2410 | |
| 2411 | mov dword ptr [edi], eax; |
| 2412 | add edi, 4; |
| 2413 | |
| 2414 | dec ecx; |
| 2415 | jnz tc1_loop; |
| 2416 | |
| 2417 | pop edi; |
| 2418 | pop esi; |
| 2419 | pop ebx; |
| 2420 | } |
| 2421 | } |
| 2422 | #endif |