Mercurial > mplayer.hg

comparison libswscale/x86/yuv2rgb_template2.c @ 31078:6502a6b24f9b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
alternative LGPL-licensed, MMX-optimized YUV to RGB conversion routines written by Kostya Shishkov
author diego
date Tue, 04 May 2010 09:11:01 +0000
parents
children b3c85aa7adbf
comparison
equal deleted inserted replaced
31077:dd7f15a3fb1b 31078:6502a6b24f9b
1 /*
2 * software YUV to RGB converter
3 *
4 * Copyright (C) 2001-2007 Michael Niedermayer
5 * (c) 2010 Konstantin Shishkov
6 *
7 * This file is part of FFmpeg.
8 *
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
13 *
14 * FFmpeg 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 GNU
17 * Lesser General Public License for more details.
18 *
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 */
23
24 #undef MOVNTQ
25 #undef EMMS
26 #undef SFENCE
27
28 #if HAVE_AMD3DNOW
29 /* On K6 femms is faster than emms. On K7 femms is directly mapped to emms. */
30 #define EMMS "femms"
31 #else
32 #define EMMS "emms"
33 #endif
34
35 #if HAVE_MMX2
36 #define MOVNTQ "movntq"
37 #define SFENCE "sfence"
38 #else
39 #define MOVNTQ "movq"
40 #define SFENCE " # nop"
41 #endif
42
43 #define REG_BLUE "0"
44 #define REG_RED "1"
45 #define REG_GREEN "2"
46 #define REG_ALPHA "3"
47
48 #define YUV2RGB_LOOP(depth) \
49 h_size = (c->dstW + 7) & ~7; \
50 if (h_size * depth > FFABS(dstStride[0])) \
51 h_size -= 8; \
52 \
53 if (c->srcFormat == PIX_FMT_YUV422P) { \
54 srcStride[1] *= 2; \
55 srcStride[2] *= 2; \
56 } \
57 \
58 __asm__ volatile ("pxor %mm4, %mm4\n\t"); \
59 for (y = 0; y < srcSliceH; y++) { \
60 uint8_t *image = dst[0] + (y + srcSliceY) * dstStride[0]; \
61 const uint8_t *py = src[0] + y * srcStride[0]; \
62 const uint8_t *pu = src[1] + (y >> 1) * srcStride[1]; \
63 const uint8_t *pv = src[2] + (y >> 1) * srcStride[2]; \
64 x86_reg index = -h_size / 2; \
65
66 #define YUV2RGB_INITIAL_LOAD \
67 __asm__ volatile ( \
68 "movq (%5, %0, 2), %%mm6\n\t" \
69 "movd (%2, %0), %%mm0\n\t" \
70 "movd (%3, %0), %%mm1\n\t" \
71 "1: \n\t" \
72
73 /* YUV2RGB core
74 * Conversion is performed in usual way:
75 * R = Y' * Ycoef + Vred * V'
76 * G = Y' * Ycoef + Vgreen * V' + Ugreen * U'
77 * B = Y' * Ycoef + Ublue * U'
78 *
79 * where X' = X * 8 - Xoffset (multiplication is performed to increase
80 * precision a bit).
81 * Since it operates in YUV420 colorspace, Y component is additionally
82 * split into Y1 and Y2 for even and odd pixels.
83 *
84 * Input:
85 * mm0 - U (4 elems), mm1 - V (4 elems), mm6 - Y (8 elems), mm4 - zero register
86 * Output:
87 * mm1 - R, mm2 - G, mm0 - B
88 */
89 #define YUV2RGB \
90 /* convert Y, U, V into Y1', Y2', U', V' */ \
91 "movq %%mm6, %%mm7\n\t" \
92 "punpcklbw %%mm4, %%mm0\n\t" \
93 "punpcklbw %%mm4, %%mm1\n\t" \
94 "pand "MANGLE(mmx_00ffw)", %%mm6\n\t" \
95 "psrlw $8, %%mm7\n\t" \
96 "psllw $3, %%mm0\n\t" \
97 "psllw $3, %%mm1\n\t" \
98 "psllw $3, %%mm6\n\t" \
99 "psllw $3, %%mm7\n\t" \
100 "psubsw "U_OFFSET"(%4), %%mm0\n\t" \
101 "psubsw "V_OFFSET"(%4), %%mm1\n\t" \
102 "psubw "Y_OFFSET"(%4), %%mm6\n\t" \
103 "psubw "Y_OFFSET"(%4), %%mm7\n\t" \
104 \
105 /* multiply by coefficients */ \
106 "movq %%mm0, %%mm2\n\t" \
107 "movq %%mm1, %%mm3\n\t" \
108 "pmulhw "UG_COEFF"(%4), %%mm2\n\t" \
109 "pmulhw "VG_COEFF"(%4), %%mm3\n\t" \
110 "pmulhw "Y_COEFF" (%4), %%mm6\n\t" \
111 "pmulhw "Y_COEFF" (%4), %%mm7\n\t" \
112 "pmulhw "UB_COEFF"(%4), %%mm0\n\t" \
113 "pmulhw "VR_COEFF"(%4), %%mm1\n\t" \
114 "paddsw %%mm3, %%mm2\n\t" \
115 /* now: mm0 = UB, mm1 = VR, mm2 = CG */ \
116 /* mm6 = Y1, mm7 = Y2 */ \
117 \
118 /* produce RGB */ \
119 "movq %%mm7, %%mm3\n\t" \
120 "movq %%mm7, %%mm5\n\t" \
121 "paddsw %%mm0, %%mm3\n\t" \
122 "paddsw %%mm1, %%mm5\n\t" \
123 "paddsw %%mm2, %%mm7\n\t" \
124 "paddsw %%mm6, %%mm0\n\t" \
125 "paddsw %%mm6, %%mm1\n\t" \
126 "paddsw %%mm6, %%mm2\n\t" \
127 \
128 /* pack and interleave even/odd pixels */ \
129 "packuswb %%mm0, %%mm0\n\t" \
130 "packuswb %%mm1, %%mm1\n\t" \
131 "packuswb %%mm2, %%mm2\n\t" \
132 "packuswb %%mm3, %%mm3\n\t" \
133 "packuswb %%mm5, %%mm5\n\t" \
134 "packuswb %%mm7, %%mm7\n\t" \
135 "punpcklbw %%mm3, %%mm0\n\t" \
136 "punpcklbw %%mm5, %%mm1\n\t" \
137 "punpcklbw %%mm7, %%mm2\n\t" \
138
139 #define YUV2RGB_ENDLOOP(depth) \
140 "movq 8 (%5, %0, 2), %%mm6\n\t" \
141 "movd 4 (%3, %0), %%mm1\n\t" \
142 "movd 4 (%2, %0), %%mm0\n\t" \
143 "add $"AV_STRINGIFY(depth * 8)", %1\n\t" \
144 "add $4, %0\n\t" \
145 "js 1b\n\t" \
146
147 #define YUV2RGB_OPERANDS \
148 : "+r" (index), "+r" (image) \
149 : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), \
150 "r" (py - 2*index) \
151 ); \
152 } \
153
154 #define YUV2RGB_OPERANDS_ALPHA \
155 : "+r" (index), "+r" (image) \
156 : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), \
157 "r" (py - 2*index), "r" (pa - 2*index) \
158 ); \
159 } \
160
161 #define YUV2RGB_ENDFUNC \
162 __asm__ volatile (SFENCE"\n\t"EMMS); \
163 return srcSliceH; \
164
165
166 #define RGB_PACK16(gmask, gshift, rshift) \
167 "pand "MANGLE(mmx_redmask)", %%mm0\n\t" \
168 "pand "MANGLE(mmx_redmask)", %%mm1\n\t" \
169 "psrlw $3, %%mm0\n\t" \
170 "pand "MANGLE(gmask)", %%mm2\n\t" \
171 "movq %%mm0, %%mm5\n\t" \
172 "movq %%mm1, %%mm6\n\t" \
173 "movq %%mm2, %%mm7\n\t" \
174 "punpcklbw %%mm4, %%mm0\n\t" \
175 "punpcklbw %%mm4, %%mm1\n\t" \
176 "punpcklbw %%mm4, %%mm2\n\t" \
177 "punpckhbw %%mm4, %%mm5\n\t" \
178 "punpckhbw %%mm4, %%mm6\n\t" \
179 "punpckhbw %%mm4, %%mm7\n\t" \
180 "psllw $"rshift", %%mm1\n\t" \
181 "psllw $"rshift", %%mm6\n\t" \
182 "psllw $"gshift", %%mm2\n\t" \
183 "psllw $"gshift", %%mm7\n\t" \
184 "por %%mm1, %%mm0\n\t" \
185 "por %%mm6, %%mm5\n\t" \
186 "por %%mm2, %%mm0\n\t" \
187 "por %%mm7, %%mm5\n\t" \
188 MOVNTQ " %%mm0, (%1)\n\t" \
189 MOVNTQ " %%mm5, 8(%1)\n\t" \
190
191 #define DITHER_RGB \
192 "paddusb "BLUE_DITHER"(%4), %%mm0\n\t" \
193 "paddusb "GREEN_DITHER"(%4), %%mm2\n\t" \
194 "paddusb "RED_DITHER"(%4), %%mm1\n\t" \
195
196 static inline int RENAME(yuv420_rgb15)(SwsContext *c, const uint8_t *src[],
197 int srcStride[],
198 int srcSliceY, int srcSliceH,
199 uint8_t *dst[], int dstStride[])
200 {
201 int y, h_size;
202
203 YUV2RGB_LOOP(2)
204
205 #ifdef DITHER1XBPP
206 c->blueDither = ff_dither8[y & 1];
207 c->greenDither = ff_dither8[y & 1];
208 c->redDither = ff_dither8[(y + 1) & 1];
209 #endif
210
211 YUV2RGB_INITIAL_LOAD
212 YUV2RGB
213 #ifdef DITHER1XBPP
214 DITHER_RGB
215 #endif
216 RGB_PACK16(mmx_redmask, "2", "7")
217
218 YUV2RGB_ENDLOOP(2)
219 YUV2RGB_OPERANDS
220 YUV2RGB_ENDFUNC
221 }
222
223 static inline int RENAME(yuv420_rgb16)(SwsContext *c, const uint8_t *src[],
224 int srcStride[],
225 int srcSliceY, int srcSliceH,
226 uint8_t *dst[], int dstStride[])
227 {
228 int y, h_size;
229
230 YUV2RGB_LOOP(2)
231
232 #ifdef DITHER1XBPP
233 c->blueDither = ff_dither8[y & 1];
234 c->greenDither = ff_dither4[y & 1];
235 c->redDither = ff_dither8[(y + 1) & 1];
236 #endif
237
238 YUV2RGB_INITIAL_LOAD
239 YUV2RGB
240 #ifdef DITHER1XBPP
241 DITHER_RGB
242 #endif
243 RGB_PACK16(mmx_grnmask, "3", "8")
244
245 YUV2RGB_ENDLOOP(2)
246 YUV2RGB_OPERANDS
247 YUV2RGB_ENDFUNC
248 }
249
250
251 #define RGB_PACK24(red, blue) \
252 /* generate first packed RGB octet */ \
253 "movq %%mm2, %%mm5\n\t" \
254 "movq %%mm"blue", %%mm6\n\t" \
255 "movq %%mm"red", %%mm7\n\t" \
256 "punpcklbw %%mm5, %%mm6\n\t" \
257 "punpcklbw %%mm4, %%mm7\n\t" \
258 "movq %%mm6, %%mm3\n\t" \
259 "punpcklwd %%mm7, %%mm6\n\t" \
260 "psrlq $32, %%mm3\n\t" \
261 "movq %%mm6, %%mm5\n\t" \
262 "psllq $40, %%mm6\n\t" \
263 "psllq $48, %%mm3\n\t" \
264 "psrlq $32, %%mm5\n\t" \
265 "psrlq $40, %%mm6\n\t" \
266 "psllq $24, %%mm5\n\t" \
267 "por %%mm3, %%mm6\n\t" \
268 "por %%mm5, %%mm6\n\t" \
269 MOVNTQ " %%mm6, (%1)\n\t" \
270 \
271 /* generate second packed RGB octet */ \
272 "movq %%mm"red", %%mm7\n\t" \
273 "movq %%mm2, %%mm5\n\t" \
274 "movq %%mm"blue", %%mm6\n\t" \
275 "punpcklbw %%mm4, %%mm7\n\t" \
276 "punpcklbw %%mm5, %%mm6\n\t" \
277 "movq %%mm7, %%mm3\n\t" \
278 "punpckhwd %%mm7, %%mm6\n\t" \
279 "psllq $16, %%mm3\n\t" \
280 "psrlq $32, %%mm6\n\t" \
281 "psrlq $48, %%mm3\n\t" \
282 "psllq $8, %%mm6\n\t" \
283 "movq %%mm"red", %%mm7\n\t" \
284 "por %%mm6, %%mm3\n\t" \
285 "movq %%mm"blue", %%mm6\n\t" \
286 "movq %%mm2, %%mm5\n\t" \
287 "punpckhbw %%mm4, %%mm7\n\t" \
288 "punpckhbw %%mm5, %%mm6\n\t" \
289 "movq %%mm6, %%mm5\n\t" \
290 "punpcklwd %%mm7, %%mm6\n\t" \
291 "psrlq $16, %%mm5\n\t" \
292 "psllq $56, %%mm5\n\t" \
293 "por %%mm5, %%mm3\n\t" \
294 "psllq $32, %%mm6\n\t" \
295 "por %%mm6, %%mm3\n\t" \
296 MOVNTQ " %%mm3, 8(%1)\n\t" \
297 \
298 /* generate third packed RGB octet */ \
299 "movq %%mm"red", %%mm7\n\t" \
300 "movq %%mm2, %%mm5\n\t" \
301 "movq %%mm2, %%mm3\n\t" \
302 "movq %%mm"blue", %%mm6\n\t" \
303 "punpckhbw %%mm"red", %%mm3\n\t" \
304 "punpckhbw %%mm4, %%mm7\n\t" \
305 "psllq $32, %%mm3\n\t" \
306 "punpckhbw %%mm5, %%mm6\n\t" \
307 "psrlq $48, %%mm3\n\t" \
308 "punpckhwd %%mm7, %%mm6\n\t" \
309 "movq %%mm6, %%mm7\n\t" \
310 "psrlq $32, %%mm6\n\t" \
311 "psllq $32, %%mm7\n\t" \
312 "psllq $40, %%mm6\n\t" \
313 "psrlq $16, %%mm7\n\t" \
314 "por %%mm6, %%mm3\n\t" \
315 "por %%mm7, %%mm3\n\t" \
316 MOVNTQ " %%mm3, 16(%1)\n\t" \
317
318 static inline int RENAME(yuv420_rgb24)(SwsContext *c, const uint8_t *src[],
319 int srcStride[],
320 int srcSliceY, int srcSliceH,
321 uint8_t *dst[], int dstStride[])
322 {
323 int y, h_size;
324
325 YUV2RGB_LOOP(3)
326
327 YUV2RGB_INITIAL_LOAD
328 YUV2RGB
329 RGB_PACK24(REG_BLUE, REG_RED)
330
331 YUV2RGB_ENDLOOP(3)
332 YUV2RGB_OPERANDS
333 YUV2RGB_ENDFUNC
334 }
335
336 static inline int RENAME(yuv420_bgr24)(SwsContext *c, const uint8_t *src[],
337 int srcStride[],
338 int srcSliceY, int srcSliceH,
339 uint8_t *dst[], int dstStride[])
340 {
341 int y, h_size;
342
343 YUV2RGB_LOOP(3)
344
345 YUV2RGB_INITIAL_LOAD
346 YUV2RGB
347 RGB_PACK24(REG_RED, REG_BLUE)
348
349 YUV2RGB_ENDLOOP(3)
350 YUV2RGB_OPERANDS
351 YUV2RGB_ENDFUNC
352 }
353
354
355 #define SET_EMPTY_ALPHA \
356 "pcmpeqd %%mm"REG_ALPHA", %%mm"REG_ALPHA"\n\t" /* set alpha to 0xFF */ \
357
358 #define LOAD_ALPHA \
359 "movq (%6, %0, 2), %%mm"REG_ALPHA"\n\t" \
360
361 #define RGB_PACK32(red, green, blue, alpha) \
362 "movq %%mm"blue", %%mm5\n\t" \
363 "movq %%mm"red", %%mm6\n\t" \
364 "punpckhbw %%mm"green", %%mm5\n\t" \
365 "punpcklbw %%mm"green", %%mm"blue"\n\t" \
366 "punpckhbw %%mm"alpha", %%mm6\n\t" \
367 "punpcklbw %%mm"alpha", %%mm"red"\n\t" \
368 "movq %%mm"blue", %%mm"green"\n\t" \
369 "movq %%mm5, %%mm"alpha"\n\t" \
370 "punpcklwd %%mm"red", %%mm"blue"\n\t" \
371 "punpckhwd %%mm"red", %%mm"green"\n\t" \
372 "punpcklwd %%mm6, %%mm5\n\t" \
373 "punpckhwd %%mm6, %%mm"alpha"\n\t" \
374 MOVNTQ " %%mm"blue", 0(%1)\n\t" \
375 MOVNTQ " %%mm"green", 8(%1)\n\t" \
376 MOVNTQ " %%mm5, 16(%1)\n\t" \
377 MOVNTQ " %%mm"alpha", 24(%1)\n\t" \
378
379 static inline int RENAME(yuv420_rgb32)(SwsContext *c, const uint8_t *src[],
380 int srcStride[],
381 int srcSliceY, int srcSliceH,
382 uint8_t *dst[], int dstStride[])
383 {
384 int y, h_size;
385
386 YUV2RGB_LOOP(4)
387
388 YUV2RGB_INITIAL_LOAD
389 YUV2RGB
390 SET_EMPTY_ALPHA
391 RGB_PACK32(REG_RED, REG_GREEN, REG_BLUE, REG_ALPHA)
392
393 YUV2RGB_ENDLOOP(4)
394 YUV2RGB_OPERANDS
395 YUV2RGB_ENDFUNC
396 }
397
398 static inline int RENAME(yuva420_rgb32)(SwsContext *c, const uint8_t *src[],
399 int srcStride[],
400 int srcSliceY, int srcSliceH,
401 uint8_t *dst[], int dstStride[])
402 {
403 #if HAVE_7REGS
404 int y, h_size;
405
406 YUV2RGB_LOOP(4)
407
408 const uint8_t *pa = src[3] + y * srcStride[3];
409 YUV2RGB_INITIAL_LOAD
410 YUV2RGB
411 LOAD_ALPHA
412 RGB_PACK32(REG_RED, REG_GREEN, REG_BLUE, REG_ALPHA)
413
414 YUV2RGB_ENDLOOP(4)
415 YUV2RGB_OPERANDS_ALPHA
416 YUV2RGB_ENDFUNC
417 #endif
418 }
419
420 static inline int RENAME(yuv420_bgr32)(SwsContext *c, const uint8_t *src[],
421 int srcStride[],
422 int srcSliceY, int srcSliceH,
423 uint8_t *dst[], int dstStride[])
424 {
425 int y, h_size;
426
427 YUV2RGB_LOOP(4)
428
429 YUV2RGB_INITIAL_LOAD
430 YUV2RGB
431 SET_EMPTY_ALPHA
432 RGB_PACK32(REG_BLUE, REG_GREEN, REG_RED, REG_ALPHA)
433
434 YUV2RGB_ENDLOOP(4)
435 YUV2RGB_OPERANDS
436 YUV2RGB_ENDFUNC
437 }
438
439 static inline int RENAME(yuva420_bgr32)(SwsContext *c, const uint8_t *src[],
440 int srcStride[],
441 int srcSliceY, int srcSliceH,
442 uint8_t *dst[], int dstStride[])
443 {
444 #if HAVE_7REGS
445 int y, h_size;
446
447 YUV2RGB_LOOP(4)
448
449 const uint8_t *pa = src[3] + y * srcStride[3];
450 YUV2RGB_INITIAL_LOAD
451 YUV2RGB
452 LOAD_ALPHA
453 RGB_PACK32(REG_BLUE, REG_GREEN, REG_RED, REG_ALPHA)
454
455 YUV2RGB_ENDLOOP(4)
456 YUV2RGB_OPERANDS_ALPHA
457 YUV2RGB_ENDFUNC
458 #endif
459 }