Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(616)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: media/base/yuv_row_mac.cc

Issue 208030: Mac YUV assembly (Closed) Base URL: svn://chrome-svn/chrome/trunk/src/
Patch Set: Created 11 years, 3 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
« no previous file with comments | « media/base/yuv_row.h ('k') | no next file » | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 // Copyright (c) 2009 The Chromium Authors. All rights reserved. 1 // Copyright (c) 2009 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be 2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file. 3 // found in the LICENSE file.
4 4
5 #include "media/base/yuv_row.h" 5 #include "media/base/yuv_row.h"
6 6
7 #ifdef _DEBUG
8 #include "base/logging.h"
9 #else
10 #define DCHECK(a)
11 #endif
12
13 // TODO(fbarchard): Make MMX work in DLLs. Currently only works in unittests.
14 // TODO(fbarchard): Do 64 bit version. 7 // TODO(fbarchard): Do 64 bit version.
15 8
16 extern "C" { 9 extern "C" {
17 #if USE_MMX
18 10
19 #define RGBY(i) { \ 11 #define RGBY(i) { \
20 static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \ 12 static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
21 static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \ 13 static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
22 static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \ 14 static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
23 0 \ 15 0 \
24 } 16 }
25 17
26 #define RGBU(i) { \ 18 #define RGBU(i) { \
27 static_cast<int16>(2.018 * 64 * (i - 128) + 0.5), \ 19 static_cast<int16>(2.018 * 64 * (i - 128) + 0.5), \
28 static_cast<int16>(-0.391 * 64 * (i - 128) + 0.5), \ 20 static_cast<int16>(-0.391 * 64 * (i - 128) + 0.5), \
29 0, \ 21 0, \
30 static_cast<int16>(256 * 64 - 1) \ 22 static_cast<int16>(256 * 64 - 1) \
31 } 23 }
32 24
33 #define RGBV(i) { \ 25 #define RGBV(i) { \
34 0, \ 26 0, \
35 static_cast<int16>(-0.813 * 64 * (i - 128) + 0.5), \ 27 static_cast<int16>(-0.813 * 64 * (i - 128) + 0.5), \
36 static_cast<int16>(1.596 * 64 * (i - 128) + 0.5), \ 28 static_cast<int16>(1.596 * 64 * (i - 128) + 0.5), \
37 0 \ 29 0 \
38 } 30 }
39 31
40 #define MMX_ALIGNED(var) var __attribute__((aligned(16))) 32 #define MMX_ALIGNED(var) \
33 var __attribute__ ((section ("__TEXT,__text"))) __attribute__ ((aligned(16)))
41 34
42 MMX_ALIGNED(int16 kCoefficientsRgbY[256][4]) = { 35 MMX_ALIGNED(int16 kCoefficientsRgbY[256][4]) = {
43 RGBY(0x00), RGBY(0x01), RGBY(0x02), RGBY(0x03), 36 RGBY(0x00), RGBY(0x01), RGBY(0x02), RGBY(0x03),
44 RGBY(0x04), RGBY(0x05), RGBY(0x06), RGBY(0x07), 37 RGBY(0x04), RGBY(0x05), RGBY(0x06), RGBY(0x07),
45 RGBY(0x08), RGBY(0x09), RGBY(0x0A), RGBY(0x0B), 38 RGBY(0x08), RGBY(0x09), RGBY(0x0A), RGBY(0x0B),
46 RGBY(0x0C), RGBY(0x0D), RGBY(0x0E), RGBY(0x0F), 39 RGBY(0x0C), RGBY(0x0D), RGBY(0x0E), RGBY(0x0F),
47 RGBY(0x10), RGBY(0x11), RGBY(0x12), RGBY(0x13), 40 RGBY(0x10), RGBY(0x11), RGBY(0x12), RGBY(0x13),
48 RGBY(0x14), RGBY(0x15), RGBY(0x16), RGBY(0x17), 41 RGBY(0x14), RGBY(0x15), RGBY(0x16), RGBY(0x17),
49 RGBY(0x18), RGBY(0x19), RGBY(0x1A), RGBY(0x1B), 42 RGBY(0x18), RGBY(0x19), RGBY(0x1A), RGBY(0x1B),
50 RGBY(0x1C), RGBY(0x1D), RGBY(0x1E), RGBY(0x1F), 43 RGBY(0x1C), RGBY(0x1D), RGBY(0x1E), RGBY(0x1F),
(...skipping 187 matching lines...) Expand 10 before | Expand all | Expand 10 after
238 RGBV(0xF4), RGBV(0xF5), RGBV(0xF6), RGBV(0xF7), 231 RGBV(0xF4), RGBV(0xF5), RGBV(0xF6), RGBV(0xF7),
239 RGBV(0xF8), RGBV(0xF9), RGBV(0xFA), RGBV(0xFB), 232 RGBV(0xF8), RGBV(0xF9), RGBV(0xFA), RGBV(0xFB),
240 RGBV(0xFC), RGBV(0xFD), RGBV(0xFE), RGBV(0xFF), 233 RGBV(0xFC), RGBV(0xFD), RGBV(0xFE), RGBV(0xFF),
241 }; 234 };
242 235
243 #undef RGBY 236 #undef RGBY
244 #undef RGBU 237 #undef RGBU
245 #undef RGBV 238 #undef RGBV
246 #undef MMX_ALIGNED 239 #undef MMX_ALIGNED
247 240
248 // TODO(fbarchard): Use the following function instead of 241 extern void MacConvertYUVToRGB32Row(const uint8* y_buf,
249 // pure assembly to help make code more portable to 64 bit 242 const uint8* u_buf,
250 // and Mac, which has different labels. 243 const uint8* v_buf,
251 // no-gcse eliminates the frame pointer, freeing up ebp. 244 uint8* rgb_buf,
252 245 int width,
253 #if defined(FUTURE_64BIT_VERSION) 246 int16 *kCoefficientsRgbY);
254 void __attribute__((optimize("O2", "no-gcse")))
255 NewFastConvertYUVToRGB32Row(const uint8* y_buf,
256 const uint8* u_buf,
257 const uint8* v_buf,
258 uint8* rgb_buf,
259 int width) {
260 asm(
261 "shr %4\n"
262 "1:\n"
263 "movzb (%1),%%eax\n"
264 "add $0x1,%1\n"
265 "movzb (%2),%%ebx\n"
266 "add $0x1,%2\n"
267 "movq kCoefficientsRgbU(,%%eax,8),%%mm0\n"
268 "movzb (%0),%%eax\n"
269 "paddsw kCoefficientsRgbV(,%%ebx,8),%%mm0\n"
270 "movzb 0x1(%0),%%ebx\n"
271 "movq kCoefficientsRgbY(,%%eax,8),%%mm1\n"
272 "add $0x2,%0\n"
273 "movq kCoefficientsRgbY(,%%ebx,8),%%mm2\n"
274 "paddsw %%mm0,%%mm1\n"
275 "paddsw %%mm0,%%mm2\n"
276 "psraw $0x6,%%mm1\n"
277 "psraw $0x6,%%mm2\n"
278 "packuswb %%mm2,%%mm1\n"
279 "movntq %%mm1,0x0(%3)\n"
280 "add $0x8,%3\n"
281 "sub $0x1,%4\n"
282 "jne 1b\n"
283 : : "r"(y_buf),"r"(u_buf),"r"(v_buf),"r"(rgb_buf),"r"(width)
284 : "eax","ebx");
285 }
286 #endif
287
288 extern void FastConvertYUVToRGB32Row(const uint8* y_buf,
289 const uint8* u_buf,
290 const uint8* v_buf,
291 uint8* rgb_buf,
292 int width);
293
294 __asm__( 247 __asm__(
295 " .globl _FastConvertYUVToRGB32Row\n" 248 " .globl _MacConvertYUVToRGB32Row\n"
296 "_FastConvertYUVToRGB32Row:\n" 249 "_MacConvertYUVToRGB32Row:\n"
297 "pusha\n" 250 "pusha\n"
298 "mov 0x24(%esp),%edx\n" 251 "mov 0x24(%esp),%edx\n"
299 "mov 0x28(%esp),%edi\n" 252 "mov 0x28(%esp),%edi\n"
300 "mov 0x2c(%esp),%esi\n" 253 "mov 0x2c(%esp),%esi\n"
301 "mov 0x30(%esp),%ebp\n" 254 "mov 0x30(%esp),%ebp\n"
302 "mov 0x34(%esp),%ecx\n" 255 "mov 0x38(%esp),%ecx\n"
303 "jmp convertend\n"
304 256
305 "convertloop:" 257 "jmp Lconvertend\n"
258
259 "Lconvertloop:"
306 "movzbl (%edi),%eax\n" 260 "movzbl (%edi),%eax\n"
307 "add $0x1,%edi\n" 261 "add $0x1,%edi\n"
308 "movzbl (%esi),%ebx\n" 262 "movzbl (%esi),%ebx\n"
309 "add $0x1,%esi\n" 263 "add $0x1,%esi\n"
310 "movq _kCoefficientsRgbU(,%eax,8),%mm0\n" 264 "movq 2048(%ecx,%eax,8),%mm0\n"
311 "movzbl (%edx),%eax\n" 265 "movzbl (%edx),%eax\n"
312 "paddsw _kCoefficientsRgbV(,%ebx,8),%mm0\n" 266 "paddsw 4096(%ecx,%ebx,8),%mm0\n"
313 "movzbl 0x1(%edx),%ebx\n" 267 "movzbl 0x1(%edx),%ebx\n"
314 "movq _kCoefficientsRgbY(,%eax,8),%mm1\n" 268 "movq 0(%ecx,%eax,8),%mm1\n"
315 "add $0x2,%edx\n" 269 "add $0x2,%edx\n"
316 "movq _kCoefficientsRgbY(,%ebx,8),%mm2\n" 270 "movq 0(%ecx,%ebx,8),%mm2\n"
317 "paddsw %mm0,%mm1\n" 271 "paddsw %mm0,%mm1\n"
318 "paddsw %mm0,%mm2\n" 272 "paddsw %mm0,%mm2\n"
319 "psraw $0x6,%mm1\n" 273 "psraw $0x6,%mm1\n"
320 "psraw $0x6,%mm2\n" 274 "psraw $0x6,%mm2\n"
321 "packuswb %mm2,%mm1\n" 275 "packuswb %mm2,%mm1\n"
322 "movntq %mm1,0x0(%ebp)\n" 276 "movntq %mm1,0x0(%ebp)\n"
323 "add $0x8,%ebp\n" 277 "add $0x8,%ebp\n"
324 "convertend:" 278 "Lconvertend:"
325 "sub $0x2,%ecx\n" 279 "sub $0x2,0x34(%esp)\n"
326 "jns convertloop\n" 280 "jns Lconvertloop\n"
327 281
328 "and $0x1,%ecx\n" 282 "and $0x1,0x34(%esp)\n"
329 "je convertdone\n" 283 "je Lconvertdone\n"
330 284
331 "movzbl (%edi),%eax\n" 285 "movzbl (%edi),%eax\n"
332 "movq _kCoefficientsRgbU(,%eax,8),%mm0\n" 286 "movq 2048(%ecx,%eax,8),%mm0\n"
333 "movzbl (%esi),%eax\n" 287 "movzbl (%esi),%eax\n"
334 "paddsw _kCoefficientsRgbV(,%eax,8),%mm0\n" 288 "paddsw 4096(%ecx,%eax,8),%mm0\n"
335 "movzbl (%edx),%eax\n" 289 "movzbl (%edx),%eax\n"
336 "movq _kCoefficientsRgbY(,%eax,8),%mm1\n" 290 "movq 0(%ecx,%eax,8),%mm1\n"
337 "paddsw %mm0,%mm1\n" 291 "paddsw %mm0,%mm1\n"
338 "psraw $0x6,%mm1\n" 292 "psraw $0x6,%mm1\n"
339 "packuswb %mm1,%mm1\n" 293 "packuswb %mm1,%mm1\n"
340 "movd %mm1,0x0(%ebp)\n" 294 "movd %mm1,0x0(%ebp)\n"
341 "convertdone:" 295 "Lconvertdone:\n"
342 "popa\n" 296 "popa\n"
343 "ret\n" 297 "ret\n"
344 ); 298 );
345 299
300 void FastConvertYUVToRGB32Row(const uint8* y_buf,
301 const uint8* u_buf,
302 const uint8* v_buf,
303 uint8* rgb_buf,
304 int width) {
305 MacConvertYUVToRGB32Row(y_buf, u_buf, v_buf, rgb_buf, width,
306 &kCoefficientsRgbY[0][0]);
307 }
346 308
347 extern void ScaleYUVToRGB32Row(const uint8* y_buf, 309 extern void MacScaleYUVToRGB32Row(const uint8* y_buf,
348 const uint8* u_buf, 310 const uint8* u_buf,
349 const uint8* v_buf, 311 const uint8* v_buf,
350 uint8* rgb_buf, 312 uint8* rgb_buf,
351 int width, 313 int width,
352 int scaled_dx); 314 int scaled_dx,
315 int16 *kCoefficientsRgbY);
353 316
354 __asm__( 317 __asm__(
355 " .globl _ScaleYUVToRGB32Row\n" 318 " .globl _MacScaleYUVToRGB32Row\n"
356 "_ScaleYUVToRGB32Row:\n" 319 "_MacScaleYUVToRGB32Row:\n"
357 "pusha\n" 320 "pusha\n"
358 "mov 0x24(%esp),%edx\n" 321 "mov 0x24(%esp),%edx\n"
359 "mov 0x28(%esp),%edi\n" 322 "mov 0x28(%esp),%edi\n"
360 "mov 0x2c(%esp),%esi\n" 323 "mov 0x2c(%esp),%esi\n"
361 "mov 0x30(%esp),%ebp\n" 324 "mov 0x30(%esp),%ebp\n"
362 "mov 0x34(%esp),%ecx\n" 325 "mov 0x3c(%esp),%ecx\n"
363 "xor %ebx,%ebx\n" 326 "xor %ebx,%ebx\n"
364 "jmp scaleend\n" 327 "jmp Lscaleend\n"
365 328
366 "scaleloop:" 329 "Lscaleloop:"
367 "mov %ebx,%eax\n" 330 "mov %ebx,%eax\n"
368 "sar $0x5,%eax\n" 331 "sar $0x5,%eax\n"
369 "movzbl (%edi,%eax,1),%eax\n" 332 "movzbl (%edi,%eax,1),%eax\n"
370 "movq kCoefficientsRgbU(,%eax,8),%mm0\n" 333 "movq 2048(%ecx,%eax,8),%mm0\n"
371 "mov %ebx,%eax\n" 334 "mov %ebx,%eax\n"
372 "sar $0x5,%eax\n" 335 "sar $0x5,%eax\n"
373 "movzbl (%esi,%eax,1),%eax\n" 336 "movzbl (%esi,%eax,1),%eax\n"
374 "paddsw kCoefficientsRgbV(,%eax,8),%mm0\n" 337 "paddsw 4096(%ecx,%eax,8),%mm0\n"
375 "mov %ebx,%eax\n" 338 "mov %ebx,%eax\n"
376 "add 0x38(%esp),%ebx\n" 339 "add 0x38(%esp),%ebx\n"
377 "sar $0x4,%eax\n" 340 "sar $0x4,%eax\n"
378 "movzbl (%edx,%eax,1),%eax\n" 341 "movzbl (%edx,%eax,1),%eax\n"
379 "movq kCoefficientsRgbY(,%eax,8),%mm1\n" 342 "movq 0(%ecx,%eax,8),%mm1\n"
380 "mov %ebx,%eax\n" 343 "mov %ebx,%eax\n"
381 "add 0x38(%esp),%ebx\n" 344 "add 0x38(%esp),%ebx\n"
382 "sar $0x4,%eax\n" 345 "sar $0x4,%eax\n"
383 "movzbl (%edx,%eax,1),%eax\n" 346 "movzbl (%edx,%eax,1),%eax\n"
384 "movq kCoefficientsRgbY(,%eax,8),%mm2\n" 347 "movq 0(%ecx,%eax,8),%mm2\n"
385 "paddsw %mm0,%mm1\n" 348 "paddsw %mm0,%mm1\n"
386 "paddsw %mm0,%mm2\n" 349 "paddsw %mm0,%mm2\n"
387 "psraw $0x6,%mm1\n" 350 "psraw $0x6,%mm1\n"
388 "psraw $0x6,%mm2\n" 351 "psraw $0x6,%mm2\n"
389 "packuswb %mm2,%mm1\n" 352 "packuswb %mm2,%mm1\n"
390 "movntq %mm1,0x0(%ebp)\n" 353 "movntq %mm1,0x0(%ebp)\n"
391 "add $0x8,%ebp\n" 354 "add $0x8,%ebp\n"
392 "scaleend:" 355 "Lscaleend:"
393 "sub $0x2,%ecx\n" 356 "sub $0x2,0x34(%esp)\n"
394 "jns scaleloop\n" 357 "jns Lscaleloop\n"
395 358
396 "and $0x1,%ecx\n" 359 "and $0x1,0x34(%esp)\n"
397 "je scaledone\n" 360 "je Lscaledone\n"
398 361
399 "mov %ebx,%eax\n" 362 "mov %ebx,%eax\n"
400 "sar $0x5,%eax\n" 363 "sar $0x5,%eax\n"
401 "movzbl (%edi,%eax,1),%eax\n" 364 "movzbl (%edi,%eax,1),%eax\n"
402 "movq kCoefficientsRgbU(,%eax,8),%mm0\n" 365 "movq 2048(%ecx,%eax,8),%mm0\n"
403 "mov %ebx,%eax\n" 366 "mov %ebx,%eax\n"
404 "sar $0x5,%eax\n" 367 "sar $0x5,%eax\n"
405 "movzbl (%esi,%eax,1),%eax\n" 368 "movzbl (%esi,%eax,1),%eax\n"
406 "paddsw kCoefficientsRgbV(,%eax,8),%mm0\n" 369 "paddsw 4096(%ecx,%eax,8),%mm0\n"
407 "mov %ebx,%eax\n" 370 "mov %ebx,%eax\n"
408 "sar $0x4,%eax\n" 371 "sar $0x4,%eax\n"
409 "movzbl (%edx,%eax,1),%eax\n" 372 "movzbl (%edx,%eax,1),%eax\n"
410 "movq kCoefficientsRgbY(,%eax,8),%mm1\n" 373 "movq 0(%ecx,%eax,8),%mm1\n"
411 "mov %ebx,%eax\n" 374 "mov %ebx,%eax\n"
412 "sar $0x4,%eax\n" 375 "sar $0x4,%eax\n"
413 "movzbl (%edx,%eax,1),%eax\n" 376 "movzbl (%edx,%eax,1),%eax\n"
414 "movq kCoefficientsRgbY(,%eax,8),%mm2\n" 377 "movq 0(%ecx,%eax,8),%mm2\n"
415 "paddsw %mm0,%mm1\n" 378 "paddsw %mm0,%mm1\n"
416 "paddsw %mm0,%mm2\n" 379 "paddsw %mm0,%mm2\n"
417 "psraw $0x6,%mm1\n" 380 "psraw $0x6,%mm1\n"
418 "psraw $0x6,%mm2\n" 381 "psraw $0x6,%mm2\n"
419 "packuswb %mm2,%mm1\n" 382 "packuswb %mm2,%mm1\n"
420 "movd %mm1,0x0(%ebp)\n" 383 "movd %mm1,0x0(%ebp)\n"
421 384
422 "scaledone:" 385 "Lscaledone:"
423 "popa\n" 386 "popa\n"
424 "ret\n" 387 "ret\n"
425 ); 388 );
426 389
427 #else // USE_MMX
428 390
429 // Reference version of YUV converter.
430 static const int kClipTableSize = 256;
431 static const int kClipOverflow = 288; // Cb max is 535.
432
433 static uint8 kRgbClipTable[kClipOverflow +
434 kClipTableSize +
435 kClipOverflow] = {
436 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 288 underflow values
437 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // clipped to 0.
438 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
439 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
440 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
441 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
442 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
443 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
444 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
445 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
446 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
447 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
448 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
449 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
450 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
451 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
452 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
453 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
454 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
455 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
456 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
457 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
458 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
459 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
460 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
461 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
462 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
463 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
464 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
465 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
466 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
467 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
468 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
469 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
470 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
471 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
472 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, // Unclipped values.
473 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
474 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
475 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F,
476 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
477 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F,
478 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
479 0x38, 0x39, 0x3A, 0x3B, 0x3C, 0x3D, 0x3E, 0x3F,
480 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47,
481 0x48, 0x49, 0x4A, 0x4B, 0x4C, 0x4D, 0x4E, 0x4F,
482 0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57,
483 0x58, 0x59, 0x5A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F,
484 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
485 0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F,
486 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
487 0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F,
488 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
489 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
490 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97,
491 0x98, 0x99, 0x9A, 0x9B, 0x9C, 0x9D, 0x9E, 0x9F,
492 0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7,
493 0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF,
494 0xB0, 0xB1, 0xB2, 0xB3, 0xB4, 0xB5, 0xB6, 0xB7,
495 0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF,
496 0xC0, 0xC1, 0xC2, 0xC3, 0xC4, 0xC5, 0xC6, 0xC7,
497 0xC8, 0xC9, 0xCA, 0xCB, 0xCC, 0xCD, 0xCE, 0xCF,
498 0xD0, 0xD1, 0xD2, 0xD3, 0xD4, 0xD5, 0xD6, 0xD7,
499 0xD8, 0xD9, 0xDA, 0xDB, 0xDC, 0xDD, 0xDE, 0xDF,
500 0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7,
501 0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF,
502 0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7,
503 0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xFF,
504 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 288 overflow values
505 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // clipped to 255.
506 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
507 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
508 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
509 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
510 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
511 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
512 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
513 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
514 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
515 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
516 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
517 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
518 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
519 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
520 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
521 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
522 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
523 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
524 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
525 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
526 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
527 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
528 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
529 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
530 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
531 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
532 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
533 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
534 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
535 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
536 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
537 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
538 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
539 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
540 };
541
542 // Clip an rgb channel value to 0..255 range.
543 // Source is signed fixed point 8.8.
544 // Table allows for values to underflow or overflow by 128.
545 // Therefore source range is -128 to 384.
546 // Output clips to unsigned 0 to 255.
547 static inline uint32 clip(int32 value) {
548 DCHECK(((value >> 8) + kClipOverflow) >= 0);
549 DCHECK(((value >> 8) + kClipOverflow) <
550 (kClipOverflow + kClipTableSize + kClipOverflow));
551 return static_cast<uint32>(kRgbClipTable[((value) >> 8) + kClipOverflow]);
552 }
553
554 static inline void YuvPixel(uint8 y,
555 uint8 u,
556 uint8 v,
557 uint8* rgb_buf) {
558 int32 d = static_cast<int32>(u) - 128;
559 int32 e = static_cast<int32>(v) - 128;
560
561 int32 cb = (516 * d + 128);
562 int32 cg = (- 100 * d - 208 * e + 128);
563 int32 cr = (409 * e + 128);
564
565 int32 C298a = ((static_cast<int32>(y) - 16) * 298 + 128);
566 *reinterpret_cast<uint32*>(rgb_buf) = (clip(C298a + cb)) |
567 (clip(C298a + cg) << 8) |
568 (clip(C298a + cr) << 16) |
569 (0xff000000);
570 }
571
572 void FastConvertYUVToRGB32Row(const uint8* y_buf,
573 const uint8* u_buf,
574 const uint8* v_buf,
575 uint8* rgb_buf,
576 int width) {
577 for (int x = 0; x < width; x += 2) {
578 uint8 u = u_buf[x >> 1];
579 uint8 v = v_buf[x >> 1];
580 uint8 y0 = y_buf[x];
581 YuvPixel(y0, u, v, rgb_buf);
582 if ((x + 1) < width) {
583 uint8 y1 = y_buf[x + 1];
584 YuvPixel(y1, u, v, rgb_buf + 4);
585 }
586 rgb_buf += 8; // Advance 2 pixels.
587 }
588 }
589
590 // 28.4 fixed point is used. A shift by 4 isolates the integer.
591 // A shift by 5 is used to further subsample the chrominence channels.
592 // & 15 isolates the fixed point fraction. >> 2 to get the upper 2 bits,
593 // for 1/4 pixel accurate interpolation.
594 void ScaleYUVToRGB32Row(const uint8* y_buf, 391 void ScaleYUVToRGB32Row(const uint8* y_buf,
595 const uint8* u_buf, 392 const uint8* u_buf,
596 const uint8* v_buf, 393 const uint8* v_buf,
597 uint8* rgb_buf, 394 uint8* rgb_buf,
598 int width, 395 int width,
599 int scaled_dx) { 396 int scaled_dx) {
600 int scaled_x = 0; 397
601 for (int x = 0; x < width; ++x) { 398 MacScaleYUVToRGB32Row(y_buf, u_buf, v_buf, rgb_buf, width, scaled_dx,
602 uint8 u = u_buf[scaled_x >> 5]; 399 &kCoefficientsRgbY[0][0]);
603 uint8 v = v_buf[scaled_x >> 5];
604 uint8 y0 = y_buf[scaled_x >> 4];
605 YuvPixel(y0, u, v, rgb_buf);
606 rgb_buf += 4;
607 scaled_x += scaled_dx;
608 }
609 } 400 }
610 #endif // USE_MMX 401
611 } // extern "C" 402 } // extern "C"
612 403
OLDNEW
« no previous file with comments | « media/base/yuv_row.h ('k') | no next file » | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698