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