FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
swscale.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #include <inttypes.h>
22 #include "config.h"
23 #include "libswscale/swscale.h"
25 #include "libavutil/attributes.h"
26 #include "libavutil/avassert.h"
27 #include "libavutil/intreadwrite.h"
28 #include "libavutil/x86/asm.h"
29 #include "libavutil/x86/cpu.h"
30 #include "libavutil/cpu.h"
31 #include "libavutil/pixdesc.h"
32 
33 #if HAVE_INLINE_ASM
34 
35 #define DITHER1XBPP
36 
37 DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
38 DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
39 DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
40 DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
41 
42 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
43  0x0103010301030103LL,
44  0x0200020002000200LL,};
45 
46 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
47  0x0602060206020602LL,
48  0x0004000400040004LL,};
49 
50 DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
51 DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
52 DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
53 DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
54 DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
55 DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
56 
57 DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
58 DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
59 DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
60 
61 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
62 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
63 DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
64 
65 
66 //MMX versions
67 #if HAVE_MMX_INLINE
68 #undef RENAME
69 #define COMPILE_TEMPLATE_MMXEXT 0
70 #define RENAME(a) a ## _mmx
71 #include "swscale_template.c"
72 #endif
73 
74 // MMXEXT versions
75 #if HAVE_MMXEXT_INLINE
76 #undef RENAME
77 #undef COMPILE_TEMPLATE_MMXEXT
78 #define COMPILE_TEMPLATE_MMXEXT 1
79 #define RENAME(a) a ## _mmxext
80 #include "swscale_template.c"
81 #endif
82 
84  int lastInLumBuf, int lastInChrBuf)
85 {
86  const int dstH= c->dstH;
87  const int flags= c->flags;
88  int16_t **lumPixBuf= c->lumPixBuf;
89  int16_t **chrUPixBuf= c->chrUPixBuf;
90  int16_t **alpPixBuf= c->alpPixBuf;
91  const int vLumBufSize= c->vLumBufSize;
92  const int vChrBufSize= c->vChrBufSize;
95  int16_t *vLumFilter= c->vLumFilter;
96  int16_t *vChrFilter= c->vChrFilter;
100  const int vLumFilterSize= c->vLumFilterSize;
101  const int vChrFilterSize= c->vChrFilterSize;
102  const int chrDstY= dstY>>c->chrDstVSubSample;
103  const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input
104  const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input
105 
106  c->blueDither= ff_dither8[dstY&1];
108  c->greenDither= ff_dither8[dstY&1];
109  else
110  c->greenDither= ff_dither4[dstY&1];
111  c->redDither= ff_dither8[(dstY+1)&1];
112  if (dstY < dstH - 2) {
113  const int16_t **lumSrcPtr= (const int16_t **)(void*) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
114  const int16_t **chrUSrcPtr= (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
115  const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
116  int i;
117 
118  if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->srcH) {
119  const int16_t **tmpY = (const int16_t **) lumPixBuf + 2 * vLumBufSize;
120  int neg = -firstLumSrcY, i, end = FFMIN(c->srcH - firstLumSrcY, vLumFilterSize);
121  for (i = 0; i < neg; i++)
122  tmpY[i] = lumSrcPtr[neg];
123  for ( ; i < end; i++)
124  tmpY[i] = lumSrcPtr[i];
125  for ( ; i < vLumFilterSize; i++)
126  tmpY[i] = tmpY[i-1];
127  lumSrcPtr = tmpY;
128 
129  if (alpSrcPtr) {
130  const int16_t **tmpA = (const int16_t **) alpPixBuf + 2 * vLumBufSize;
131  for (i = 0; i < neg; i++)
132  tmpA[i] = alpSrcPtr[neg];
133  for ( ; i < end; i++)
134  tmpA[i] = alpSrcPtr[i];
135  for ( ; i < vLumFilterSize; i++)
136  tmpA[i] = tmpA[i - 1];
137  alpSrcPtr = tmpA;
138  }
139  }
140  if (firstChrSrcY < 0 || firstChrSrcY + vChrFilterSize > c->chrSrcH) {
141  const int16_t **tmpU = (const int16_t **) chrUPixBuf + 2 * vChrBufSize;
142  int neg = -firstChrSrcY, i, end = FFMIN(c->chrSrcH - firstChrSrcY, vChrFilterSize);
143  for (i = 0; i < neg; i++) {
144  tmpU[i] = chrUSrcPtr[neg];
145  }
146  for ( ; i < end; i++) {
147  tmpU[i] = chrUSrcPtr[i];
148  }
149  for ( ; i < vChrFilterSize; i++) {
150  tmpU[i] = tmpU[i - 1];
151  }
152  chrUSrcPtr = tmpU;
153  }
154 
155  if (flags & SWS_ACCURATE_RND) {
156  int s= APCK_SIZE / 8;
157  for (i=0; i<vLumFilterSize; i+=2) {
158  *(const void**)&lumMmxFilter[s*i ]= lumSrcPtr[i ];
159  *(const void**)&lumMmxFilter[s*i+APCK_PTR2/4 ]= lumSrcPtr[i+(vLumFilterSize>1)];
160  lumMmxFilter[s*i+APCK_COEF/4 ]=
161  lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i ]
162  + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);
163  if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
164  *(const void**)&alpMmxFilter[s*i ]= alpSrcPtr[i ];
165  *(const void**)&alpMmxFilter[s*i+APCK_PTR2/4 ]= alpSrcPtr[i+(vLumFilterSize>1)];
166  alpMmxFilter[s*i+APCK_COEF/4 ]=
167  alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4 ];
168  }
169  }
170  for (i=0; i<vChrFilterSize; i+=2) {
171  *(const void**)&chrMmxFilter[s*i ]= chrUSrcPtr[i ];
172  *(const void**)&chrMmxFilter[s*i+APCK_PTR2/4 ]= chrUSrcPtr[i+(vChrFilterSize>1)];
173  chrMmxFilter[s*i+APCK_COEF/4 ]=
174  chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i ]
175  + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);
176  }
177  } else {
178  for (i=0; i<vLumFilterSize; i++) {
179  *(const void**)&lumMmxFilter[4*i+0]= lumSrcPtr[i];
180  lumMmxFilter[4*i+2]=
181  lumMmxFilter[4*i+3]=
182  ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001U;
183  if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
184  *(const void**)&alpMmxFilter[4*i+0]= alpSrcPtr[i];
185  alpMmxFilter[4*i+2]=
186  alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2];
187  }
188  }
189  for (i=0; i<vChrFilterSize; i++) {
190  *(const void**)&chrMmxFilter[4*i+0]= chrUSrcPtr[i];
191  chrMmxFilter[4*i+2]=
192  chrMmxFilter[4*i+3]=
193  ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001U;
194  }
195  }
196  }
197 }
198 
199 #if HAVE_MMXEXT
200 static void yuv2yuvX_sse3(const int16_t *filter, int filterSize,
201  const int16_t **src, uint8_t *dest, int dstW,
202  const uint8_t *dither, int offset)
203 {
204  if(((int)dest) & 15){
205  yuv2yuvX_mmxext(filter, filterSize, src, dest, dstW, dither, offset);
206  return;
207  }
208  if (offset) {
209  __asm__ volatile("movq (%0), %%xmm3\n\t"
210  "movdqa %%xmm3, %%xmm4\n\t"
211  "psrlq $24, %%xmm3\n\t"
212  "psllq $40, %%xmm4\n\t"
213  "por %%xmm4, %%xmm3\n\t"
214  :: "r"(dither)
215  );
216  } else {
217  __asm__ volatile("movq (%0), %%xmm3\n\t"
218  :: "r"(dither)
219  );
220  }
221  filterSize--;
222  __asm__ volatile(
223  "pxor %%xmm0, %%xmm0\n\t"
224  "punpcklbw %%xmm0, %%xmm3\n\t"
225  "movd %0, %%xmm1\n\t"
226  "punpcklwd %%xmm1, %%xmm1\n\t"
227  "punpckldq %%xmm1, %%xmm1\n\t"
228  "punpcklqdq %%xmm1, %%xmm1\n\t"
229  "psllw $3, %%xmm1\n\t"
230  "paddw %%xmm1, %%xmm3\n\t"
231  "psraw $4, %%xmm3\n\t"
232  ::"m"(filterSize)
233  );
234  __asm__ volatile(
235  "movdqa %%xmm3, %%xmm4\n\t"
236  "movdqa %%xmm3, %%xmm7\n\t"
237  "movl %3, %%ecx\n\t"
238  "mov %0, %%"REG_d" \n\t"\
239  "mov (%%"REG_d"), %%"REG_S" \n\t"\
240  ".p2align 4 \n\t" /* FIXME Unroll? */\
241  "1: \n\t"\
242  "movddup 8(%%"REG_d"), %%xmm0 \n\t" /* filterCoeff */\
243  "movdqa (%%"REG_S", %%"REG_c", 2), %%xmm2 \n\t" /* srcData */\
244  "movdqa 16(%%"REG_S", %%"REG_c", 2), %%xmm5 \n\t" /* srcData */\
245  "add $16, %%"REG_d" \n\t"\
246  "mov (%%"REG_d"), %%"REG_S" \n\t"\
247  "test %%"REG_S", %%"REG_S" \n\t"\
248  "pmulhw %%xmm0, %%xmm2 \n\t"\
249  "pmulhw %%xmm0, %%xmm5 \n\t"\
250  "paddw %%xmm2, %%xmm3 \n\t"\
251  "paddw %%xmm5, %%xmm4 \n\t"\
252  " jnz 1b \n\t"\
253  "psraw $3, %%xmm3 \n\t"\
254  "psraw $3, %%xmm4 \n\t"\
255  "packuswb %%xmm4, %%xmm3 \n\t"
256  "movntdq %%xmm3, (%1, %%"REG_c")\n\t"
257  "add $16, %%"REG_c" \n\t"\
258  "cmp %2, %%"REG_c" \n\t"\
259  "movdqa %%xmm7, %%xmm3\n\t"
260  "movdqa %%xmm7, %%xmm4\n\t"
261  "mov %0, %%"REG_d" \n\t"\
262  "mov (%%"REG_d"), %%"REG_S" \n\t"\
263  "jb 1b \n\t"\
264  :: "g" (filter),
265  "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
266  : XMM_CLOBBERS("%xmm0" , "%xmm1" , "%xmm2" , "%xmm3" , "%xmm4" , "%xmm5" , "%xmm7" ,)
267  "%"REG_d, "%"REG_S, "%"REG_c
268  );
269 }
270 #endif
271 
272 #endif /* HAVE_INLINE_ASM */
273 
274 #define SCALE_FUNC(filter_n, from_bpc, to_bpc, opt) \
275 void ff_hscale ## from_bpc ## to ## to_bpc ## _ ## filter_n ## _ ## opt( \
276  SwsContext *c, int16_t *data, \
277  int dstW, const uint8_t *src, \
278  const int16_t *filter, \
279  const int32_t *filterPos, int filterSize)
280 
281 #define SCALE_FUNCS(filter_n, opt) \
282  SCALE_FUNC(filter_n, 8, 15, opt); \
283  SCALE_FUNC(filter_n, 9, 15, opt); \
284  SCALE_FUNC(filter_n, 10, 15, opt); \
285  SCALE_FUNC(filter_n, 12, 15, opt); \
286  SCALE_FUNC(filter_n, 14, 15, opt); \
287  SCALE_FUNC(filter_n, 16, 15, opt); \
288  SCALE_FUNC(filter_n, 8, 19, opt); \
289  SCALE_FUNC(filter_n, 9, 19, opt); \
290  SCALE_FUNC(filter_n, 10, 19, opt); \
291  SCALE_FUNC(filter_n, 12, 19, opt); \
292  SCALE_FUNC(filter_n, 14, 19, opt); \
293  SCALE_FUNC(filter_n, 16, 19, opt)
294 
295 #define SCALE_FUNCS_MMX(opt) \
296  SCALE_FUNCS(4, opt); \
297  SCALE_FUNCS(8, opt); \
298  SCALE_FUNCS(X, opt)
299 
300 #define SCALE_FUNCS_SSE(opt) \
301  SCALE_FUNCS(4, opt); \
302  SCALE_FUNCS(8, opt); \
303  SCALE_FUNCS(X4, opt); \
304  SCALE_FUNCS(X8, opt)
305 
306 #if ARCH_X86_32
307 SCALE_FUNCS_MMX(mmx);
308 #endif
309 SCALE_FUNCS_SSE(sse2);
310 SCALE_FUNCS_SSE(ssse3);
311 SCALE_FUNCS_SSE(sse4);
312 
313 #define VSCALEX_FUNC(size, opt) \
314 void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filterSize, \
315  const int16_t **src, uint8_t *dest, int dstW, \
316  const uint8_t *dither, int offset)
317 #define VSCALEX_FUNCS(opt) \
318  VSCALEX_FUNC(8, opt); \
319  VSCALEX_FUNC(9, opt); \
320  VSCALEX_FUNC(10, opt)
321 
322 #if ARCH_X86_32
323 VSCALEX_FUNCS(mmxext);
324 #endif
325 VSCALEX_FUNCS(sse2);
326 VSCALEX_FUNCS(sse4);
327 VSCALEX_FUNC(16, sse4);
328 VSCALEX_FUNCS(avx);
329 
330 #define VSCALE_FUNC(size, opt) \
331 void ff_yuv2plane1_ ## size ## _ ## opt(const int16_t *src, uint8_t *dst, int dstW, \
332  const uint8_t *dither, int offset)
333 #define VSCALE_FUNCS(opt1, opt2) \
334  VSCALE_FUNC(8, opt1); \
335  VSCALE_FUNC(9, opt2); \
336  VSCALE_FUNC(10, opt2); \
337  VSCALE_FUNC(16, opt1)
338 
339 #if ARCH_X86_32
340 VSCALE_FUNCS(mmx, mmxext);
341 #endif
342 VSCALE_FUNCS(sse2, sse2);
343 VSCALE_FUNC(16, sse4);
344 VSCALE_FUNCS(avx, avx);
345 
346 #define INPUT_Y_FUNC(fmt, opt) \
347 void ff_ ## fmt ## ToY_ ## opt(uint8_t *dst, const uint8_t *src, \
348  const uint8_t *unused1, const uint8_t *unused2, \
349  int w, uint32_t *unused)
350 #define INPUT_UV_FUNC(fmt, opt) \
351 void ff_ ## fmt ## ToUV_ ## opt(uint8_t *dstU, uint8_t *dstV, \
352  const uint8_t *unused0, \
353  const uint8_t *src1, \
354  const uint8_t *src2, \
355  int w, uint32_t *unused)
356 #define INPUT_FUNC(fmt, opt) \
357  INPUT_Y_FUNC(fmt, opt); \
358  INPUT_UV_FUNC(fmt, opt)
359 #define INPUT_FUNCS(opt) \
360  INPUT_FUNC(uyvy, opt); \
361  INPUT_FUNC(yuyv, opt); \
362  INPUT_UV_FUNC(nv12, opt); \
363  INPUT_UV_FUNC(nv21, opt); \
364  INPUT_FUNC(rgba, opt); \
365  INPUT_FUNC(bgra, opt); \
366  INPUT_FUNC(argb, opt); \
367  INPUT_FUNC(abgr, opt); \
368  INPUT_FUNC(rgb24, opt); \
369  INPUT_FUNC(bgr24, opt)
370 
371 #if ARCH_X86_32
372 INPUT_FUNCS(mmx);
373 #endif
374 INPUT_FUNCS(sse2);
375 INPUT_FUNCS(ssse3);
376 INPUT_FUNCS(avx);
377 
379 {
380  int cpu_flags = av_get_cpu_flags();
381 
382 #if HAVE_MMX_INLINE
383  if (INLINE_MMX(cpu_flags))
384  sws_init_swscale_mmx(c);
385 #endif
386 #if HAVE_MMXEXT_INLINE
387  if (INLINE_MMXEXT(cpu_flags))
388  sws_init_swscale_mmxext(c);
389  if (cpu_flags & AV_CPU_FLAG_SSE3){
390  if(c->use_mmx_vfilter && !(c->flags & SWS_ACCURATE_RND))
391  c->yuv2planeX = yuv2yuvX_sse3;
392  }
393 #endif
394 
395 #define ASSIGN_SCALE_FUNC2(hscalefn, filtersize, opt1, opt2) do { \
396  if (c->srcBpc == 8) { \
397  hscalefn = c->dstBpc <= 14 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \
398  ff_hscale8to19_ ## filtersize ## _ ## opt1; \
399  } else if (c->srcBpc == 9) { \
400  hscalefn = c->dstBpc <= 14 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \
401  ff_hscale9to19_ ## filtersize ## _ ## opt1; \
402  } else if (c->srcBpc == 10) { \
403  hscalefn = c->dstBpc <= 14 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \
404  ff_hscale10to19_ ## filtersize ## _ ## opt1; \
405  } else if (c->srcBpc == 12) { \
406  hscalefn = c->dstBpc <= 14 ? ff_hscale12to15_ ## filtersize ## _ ## opt2 : \
407  ff_hscale12to19_ ## filtersize ## _ ## opt1; \
408  } else if (c->srcBpc == 14 || ((c->srcFormat==AV_PIX_FMT_PAL8||isAnyRGB(c->srcFormat)) && av_pix_fmt_desc_get(c->srcFormat)->comp[0].depth_minus1<15)) { \
409  hscalefn = c->dstBpc <= 14 ? ff_hscale14to15_ ## filtersize ## _ ## opt2 : \
410  ff_hscale14to19_ ## filtersize ## _ ## opt1; \
411  } else { /* c->srcBpc == 16 */ \
412  av_assert0(c->srcBpc == 16);\
413  hscalefn = c->dstBpc <= 14 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \
414  ff_hscale16to19_ ## filtersize ## _ ## opt1; \
415  } \
416 } while (0)
417 #define ASSIGN_MMX_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
418  switch (filtersize) { \
419  case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
420  case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
421  default: ASSIGN_SCALE_FUNC2(hscalefn, X, opt1, opt2); break; \
422  }
423 #define ASSIGN_VSCALEX_FUNC(vscalefn, opt, do_16_case, condition_8bit) \
424 switch(c->dstBpc){ \
425  case 16: do_16_case; break; \
426  case 10: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_10_ ## opt; break; \
427  case 9: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_9_ ## opt; break; \
428  default: if (condition_8bit) /*vscalefn = ff_yuv2planeX_8_ ## opt;*/ break; \
429  }
430 #define ASSIGN_VSCALE_FUNC(vscalefn, opt1, opt2, opt2chk) \
431  switch(c->dstBpc){ \
432  case 16: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2plane1_16_ ## opt1; break; \
433  case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_10_ ## opt2; break; \
434  case 9: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_9_ ## opt2; break; \
435  case 8: vscalefn = ff_yuv2plane1_8_ ## opt1; break; \
436  default: av_assert0(c->dstBpc>8); \
437  }
438 #define case_rgb(x, X, opt) \
439  case AV_PIX_FMT_ ## X: \
440  c->lumToYV12 = ff_ ## x ## ToY_ ## opt; \
441  if (!c->chrSrcHSubSample) \
442  c->chrToYV12 = ff_ ## x ## ToUV_ ## opt; \
443  break
444 #if ARCH_X86_32
445  if (EXTERNAL_MMX(cpu_flags)) {
446  ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
447  ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
448  ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmxext, cpu_flags & AV_CPU_FLAG_MMXEXT);
449 
450  switch (c->srcFormat) {
451  case AV_PIX_FMT_YA8:
452  c->lumToYV12 = ff_yuyvToY_mmx;
453  if (c->alpPixBuf)
454  c->alpToYV12 = ff_uyvyToY_mmx;
455  break;
456  case AV_PIX_FMT_YUYV422:
457  c->lumToYV12 = ff_yuyvToY_mmx;
458  c->chrToYV12 = ff_yuyvToUV_mmx;
459  break;
460  case AV_PIX_FMT_UYVY422:
461  c->lumToYV12 = ff_uyvyToY_mmx;
462  c->chrToYV12 = ff_uyvyToUV_mmx;
463  break;
464  case AV_PIX_FMT_NV12:
465  c->chrToYV12 = ff_nv12ToUV_mmx;
466  break;
467  case AV_PIX_FMT_NV21:
468  c->chrToYV12 = ff_nv21ToUV_mmx;
469  break;
470  case_rgb(rgb24, RGB24, mmx);
471  case_rgb(bgr24, BGR24, mmx);
472  case_rgb(bgra, BGRA, mmx);
473  case_rgb(rgba, RGBA, mmx);
474  case_rgb(abgr, ABGR, mmx);
475  case_rgb(argb, ARGB, mmx);
476  default:
477  break;
478  }
479  }
480  if (EXTERNAL_MMXEXT(cpu_flags)) {
481  ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmxext, , 1);
482  }
483 #endif /* ARCH_X86_32 */
484 #define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
485  switch (filtersize) { \
486  case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
487  case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
488  default: if (filtersize & 4) ASSIGN_SCALE_FUNC2(hscalefn, X4, opt1, opt2); \
489  else ASSIGN_SCALE_FUNC2(hscalefn, X8, opt1, opt2); \
490  break; \
491  }
492  if (EXTERNAL_SSE2(cpu_flags)) {
493  ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse2, sse2);
494  ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
495  ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2, ,
496  HAVE_ALIGNED_STACK || ARCH_X86_64);
497  ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2, sse2, 1);
498 
499  switch (c->srcFormat) {
500  case AV_PIX_FMT_YA8:
501  c->lumToYV12 = ff_yuyvToY_sse2;
502  if (c->alpPixBuf)
503  c->alpToYV12 = ff_uyvyToY_sse2;
504  break;
505  case AV_PIX_FMT_YUYV422:
506  c->lumToYV12 = ff_yuyvToY_sse2;
507  c->chrToYV12 = ff_yuyvToUV_sse2;
508  break;
509  case AV_PIX_FMT_UYVY422:
510  c->lumToYV12 = ff_uyvyToY_sse2;
511  c->chrToYV12 = ff_uyvyToUV_sse2;
512  break;
513  case AV_PIX_FMT_NV12:
514  c->chrToYV12 = ff_nv12ToUV_sse2;
515  break;
516  case AV_PIX_FMT_NV21:
517  c->chrToYV12 = ff_nv21ToUV_sse2;
518  break;
519  case_rgb(rgb24, RGB24, sse2);
520  case_rgb(bgr24, BGR24, sse2);
521  case_rgb(bgra, BGRA, sse2);
522  case_rgb(rgba, RGBA, sse2);
523  case_rgb(abgr, ABGR, sse2);
524  case_rgb(argb, ARGB, sse2);
525  default:
526  break;
527  }
528  }
529  if (EXTERNAL_SSSE3(cpu_flags)) {
530  ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3);
531  ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, ssse3, ssse3);
532  switch (c->srcFormat) {
533  case_rgb(rgb24, RGB24, ssse3);
534  case_rgb(bgr24, BGR24, ssse3);
535  default:
536  break;
537  }
538  }
539  if (EXTERNAL_SSE4(cpu_flags)) {
540  /* Xto15 don't need special sse4 functions */
541  ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse4, ssse3);
542  ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse4, ssse3);
544  if (!isBE(c->dstFormat)) c->yuv2planeX = ff_yuv2planeX_16_sse4,
545  HAVE_ALIGNED_STACK || ARCH_X86_64);
546  if (c->dstBpc == 16 && !isBE(c->dstFormat))
547  c->yuv2plane1 = ff_yuv2plane1_16_sse4;
548  }
549 
550  if (EXTERNAL_AVX(cpu_flags)) {
552  HAVE_ALIGNED_STACK || ARCH_X86_64);
553  ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx, avx, 1);
554 
555  switch (c->srcFormat) {
556  case AV_PIX_FMT_YUYV422:
557  c->chrToYV12 = ff_yuyvToUV_avx;
558  break;
559  case AV_PIX_FMT_UYVY422:
560  c->chrToYV12 = ff_uyvyToUV_avx;
561  break;
562  case AV_PIX_FMT_NV12:
563  c->chrToYV12 = ff_nv12ToUV_avx;
564  break;
565  case AV_PIX_FMT_NV21:
566  c->chrToYV12 = ff_nv21ToUV_avx;
567  break;
568  case_rgb(rgb24, RGB24, avx);
569  case_rgb(bgr24, BGR24, avx);
570  case_rgb(bgra, BGRA, avx);
571  case_rgb(rgba, RGBA, avx);
572  case_rgb(abgr, ABGR, avx);
573  case_rgb(argb, ARGB, avx);
574  default:
575  break;
576  }
577  }
578 }