FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
swscale_template.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 #undef REAL_MOVNTQ
22 #undef MOVNTQ
23 #undef MOVNTQ2
24 #undef PREFETCH
25 
26 #if COMPILE_TEMPLATE_MMXEXT
27 #define PREFETCH "prefetchnta"
28 #else
29 #define PREFETCH " # nop"
30 #endif
31 
32 #if COMPILE_TEMPLATE_MMXEXT
33 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
34 #define MOVNTQ2 "movntq "
35 #else
36 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
37 #define MOVNTQ2 "movq "
38 #endif
39 #define MOVNTQ(a,b) REAL_MOVNTQ(a,b)
40 
41 #if !COMPILE_TEMPLATE_MMXEXT
42 static av_always_inline void
43 dither_8to16(const uint8_t *srcDither, int rot)
44 {
45  if (rot) {
46  __asm__ volatile("pxor %%mm0, %%mm0\n\t"
47  "movq (%0), %%mm3\n\t"
48  "movq %%mm3, %%mm4\n\t"
49  "psrlq $24, %%mm3\n\t"
50  "psllq $40, %%mm4\n\t"
51  "por %%mm4, %%mm3\n\t"
52  "movq %%mm3, %%mm4\n\t"
53  "punpcklbw %%mm0, %%mm3\n\t"
54  "punpckhbw %%mm0, %%mm4\n\t"
55  :: "r"(srcDither)
56  );
57  } else {
58  __asm__ volatile("pxor %%mm0, %%mm0\n\t"
59  "movq (%0), %%mm3\n\t"
60  "movq %%mm3, %%mm4\n\t"
61  "punpcklbw %%mm0, %%mm3\n\t"
62  "punpckhbw %%mm0, %%mm4\n\t"
63  :: "r"(srcDither)
64  );
65  }
66 }
67 #endif
68 
69 static void RENAME(yuv2yuvX)(const int16_t *filter, int filterSize,
70  const int16_t **src, uint8_t *dest, int dstW,
71  const uint8_t *dither, int offset)
72 {
74  filterSize--;
75  __asm__ volatile(
76  "movd %0, %%mm1\n\t"
77  "punpcklwd %%mm1, %%mm1\n\t"
78  "punpckldq %%mm1, %%mm1\n\t"
79  "psllw $3, %%mm1\n\t"
80  "paddw %%mm1, %%mm3\n\t"
81  "paddw %%mm1, %%mm4\n\t"
82  "psraw $4, %%mm3\n\t"
83  "psraw $4, %%mm4\n\t"
84  ::"m"(filterSize)
85  );
86 
87  __asm__ volatile(\
88  "movq %%mm3, %%mm6\n\t"
89  "movq %%mm4, %%mm7\n\t"
90  "movl %3, %%ecx\n\t"
91  "mov %0, %%"REG_d" \n\t"\
92  "mov (%%"REG_d"), %%"REG_S" \n\t"\
93  ".p2align 4 \n\t" /* FIXME Unroll? */\
94  "1: \n\t"\
95  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
96  "movq (%%"REG_S", %%"REG_c", 2), %%mm2 \n\t" /* srcData */\
97  "movq 8(%%"REG_S", %%"REG_c", 2), %%mm5 \n\t" /* srcData */\
98  "add $16, %%"REG_d" \n\t"\
99  "mov (%%"REG_d"), %%"REG_S" \n\t"\
100  "test %%"REG_S", %%"REG_S" \n\t"\
101  "pmulhw %%mm0, %%mm2 \n\t"\
102  "pmulhw %%mm0, %%mm5 \n\t"\
103  "paddw %%mm2, %%mm3 \n\t"\
104  "paddw %%mm5, %%mm4 \n\t"\
105  " jnz 1b \n\t"\
106  "psraw $3, %%mm3 \n\t"\
107  "psraw $3, %%mm4 \n\t"\
108  "packuswb %%mm4, %%mm3 \n\t"
109  MOVNTQ2 " %%mm3, (%1, %%"REG_c")\n\t"
110  "add $8, %%"REG_c" \n\t"\
111  "cmp %2, %%"REG_c" \n\t"\
112  "movq %%mm6, %%mm3\n\t"
113  "movq %%mm7, %%mm4\n\t"
114  "mov %0, %%"REG_d" \n\t"\
115  "mov (%%"REG_d"), %%"REG_S" \n\t"\
116  "jb 1b \n\t"\
117  :: "g" (filter),
118  "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
119  : "%"REG_d, "%"REG_S, "%"REG_c
120  );
121 }
122 
123 #define YSCALEYUV2PACKEDX_UV \
124  __asm__ volatile(\
125  "xor %%"REG_a", %%"REG_a" \n\t"\
126  ".p2align 4 \n\t"\
127  "nop \n\t"\
128  "1: \n\t"\
129  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
130  "mov (%%"REG_d"), %%"REG_S" \n\t"\
131  "movq "VROUNDER_OFFSET"(%0), %%mm3 \n\t"\
132  "movq %%mm3, %%mm4 \n\t"\
133  ".p2align 4 \n\t"\
134  "2: \n\t"\
135  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
136  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
137  "add %6, %%"REG_S" \n\t" \
138  "movq (%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\
139  "add $16, %%"REG_d" \n\t"\
140  "mov (%%"REG_d"), %%"REG_S" \n\t"\
141  "pmulhw %%mm0, %%mm2 \n\t"\
142  "pmulhw %%mm0, %%mm5 \n\t"\
143  "paddw %%mm2, %%mm3 \n\t"\
144  "paddw %%mm5, %%mm4 \n\t"\
145  "test %%"REG_S", %%"REG_S" \n\t"\
146  " jnz 2b \n\t"\
147 
148 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
149  "lea "offset"(%0), %%"REG_d" \n\t"\
150  "mov (%%"REG_d"), %%"REG_S" \n\t"\
151  "movq "VROUNDER_OFFSET"(%0), "#dst1" \n\t"\
152  "movq "#dst1", "#dst2" \n\t"\
153  ".p2align 4 \n\t"\
154  "2: \n\t"\
155  "movq 8(%%"REG_d"), "#coeff" \n\t" /* filterCoeff */\
156  "movq (%%"REG_S", %%"REG_a", 2), "#src1" \n\t" /* Y1srcData */\
157  "movq 8(%%"REG_S", %%"REG_a", 2), "#src2" \n\t" /* Y2srcData */\
158  "add $16, %%"REG_d" \n\t"\
159  "mov (%%"REG_d"), %%"REG_S" \n\t"\
160  "pmulhw "#coeff", "#src1" \n\t"\
161  "pmulhw "#coeff", "#src2" \n\t"\
162  "paddw "#src1", "#dst1" \n\t"\
163  "paddw "#src2", "#dst2" \n\t"\
164  "test %%"REG_S", %%"REG_S" \n\t"\
165  " jnz 2b \n\t"\
166 
167 #define YSCALEYUV2PACKEDX \
168  YSCALEYUV2PACKEDX_UV \
169  YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
170 
171 #define YSCALEYUV2PACKEDX_END \
172  :: "r" (&c->redDither), \
173  "m" (dummy), "m" (dummy), "m" (dummy),\
174  "r" (dest), "m" (dstW_reg), "m"(uv_off) \
175  : "%"REG_a, "%"REG_d, "%"REG_S \
176  );
177 
178 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
179  __asm__ volatile(\
180  "xor %%"REG_a", %%"REG_a" \n\t"\
181  ".p2align 4 \n\t"\
182  "nop \n\t"\
183  "1: \n\t"\
184  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
185  "mov (%%"REG_d"), %%"REG_S" \n\t"\
186  "pxor %%mm4, %%mm4 \n\t"\
187  "pxor %%mm5, %%mm5 \n\t"\
188  "pxor %%mm6, %%mm6 \n\t"\
189  "pxor %%mm7, %%mm7 \n\t"\
190  ".p2align 4 \n\t"\
191  "2: \n\t"\
192  "movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\
193  "add %6, %%"REG_S" \n\t" \
194  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\
195  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
196  "movq (%%"REG_S", %%"REG_a"), %%mm1 \n\t" /* UsrcData */\
197  "movq %%mm0, %%mm3 \n\t"\
198  "punpcklwd %%mm1, %%mm0 \n\t"\
199  "punpckhwd %%mm1, %%mm3 \n\t"\
200  "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1 \n\t" /* filterCoeff */\
201  "pmaddwd %%mm1, %%mm0 \n\t"\
202  "pmaddwd %%mm1, %%mm3 \n\t"\
203  "paddd %%mm0, %%mm4 \n\t"\
204  "paddd %%mm3, %%mm5 \n\t"\
205  "add %6, %%"REG_S" \n\t" \
206  "movq (%%"REG_S", %%"REG_a"), %%mm3 \n\t" /* VsrcData */\
207  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
208  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
209  "test %%"REG_S", %%"REG_S" \n\t"\
210  "movq %%mm2, %%mm0 \n\t"\
211  "punpcklwd %%mm3, %%mm2 \n\t"\
212  "punpckhwd %%mm3, %%mm0 \n\t"\
213  "pmaddwd %%mm1, %%mm2 \n\t"\
214  "pmaddwd %%mm1, %%mm0 \n\t"\
215  "paddd %%mm2, %%mm6 \n\t"\
216  "paddd %%mm0, %%mm7 \n\t"\
217  " jnz 2b \n\t"\
218  "psrad $16, %%mm4 \n\t"\
219  "psrad $16, %%mm5 \n\t"\
220  "psrad $16, %%mm6 \n\t"\
221  "psrad $16, %%mm7 \n\t"\
222  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
223  "packssdw %%mm5, %%mm4 \n\t"\
224  "packssdw %%mm7, %%mm6 \n\t"\
225  "paddw %%mm0, %%mm4 \n\t"\
226  "paddw %%mm0, %%mm6 \n\t"\
227  "movq %%mm4, "U_TEMP"(%0) \n\t"\
228  "movq %%mm6, "V_TEMP"(%0) \n\t"\
229 
230 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
231  "lea "offset"(%0), %%"REG_d" \n\t"\
232  "mov (%%"REG_d"), %%"REG_S" \n\t"\
233  "pxor %%mm1, %%mm1 \n\t"\
234  "pxor %%mm5, %%mm5 \n\t"\
235  "pxor %%mm7, %%mm7 \n\t"\
236  "pxor %%mm6, %%mm6 \n\t"\
237  ".p2align 4 \n\t"\
238  "2: \n\t"\
239  "movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\
240  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\
241  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
242  "movq (%%"REG_S", %%"REG_a", 2), %%mm4 \n\t" /* Y1srcData */\
243  "movq %%mm0, %%mm3 \n\t"\
244  "punpcklwd %%mm4, %%mm0 \n\t"\
245  "punpckhwd %%mm4, %%mm3 \n\t"\
246  "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4 \n\t" /* filterCoeff */\
247  "pmaddwd %%mm4, %%mm0 \n\t"\
248  "pmaddwd %%mm4, %%mm3 \n\t"\
249  "paddd %%mm0, %%mm1 \n\t"\
250  "paddd %%mm3, %%mm5 \n\t"\
251  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3 \n\t" /* Y2srcData */\
252  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
253  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
254  "test %%"REG_S", %%"REG_S" \n\t"\
255  "movq %%mm2, %%mm0 \n\t"\
256  "punpcklwd %%mm3, %%mm2 \n\t"\
257  "punpckhwd %%mm3, %%mm0 \n\t"\
258  "pmaddwd %%mm4, %%mm2 \n\t"\
259  "pmaddwd %%mm4, %%mm0 \n\t"\
260  "paddd %%mm2, %%mm7 \n\t"\
261  "paddd %%mm0, %%mm6 \n\t"\
262  " jnz 2b \n\t"\
263  "psrad $16, %%mm1 \n\t"\
264  "psrad $16, %%mm5 \n\t"\
265  "psrad $16, %%mm7 \n\t"\
266  "psrad $16, %%mm6 \n\t"\
267  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
268  "packssdw %%mm5, %%mm1 \n\t"\
269  "packssdw %%mm6, %%mm7 \n\t"\
270  "paddw %%mm0, %%mm1 \n\t"\
271  "paddw %%mm0, %%mm7 \n\t"\
272  "movq "U_TEMP"(%0), %%mm3 \n\t"\
273  "movq "V_TEMP"(%0), %%mm4 \n\t"\
274 
275 #define YSCALEYUV2PACKEDX_ACCURATE \
276  YSCALEYUV2PACKEDX_ACCURATE_UV \
277  YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
278 
279 #define YSCALEYUV2RGBX \
280  "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\
281  "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\
282  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
283  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
284  "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\
285  "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\
286  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
287  "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\
288  "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\
289  "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\
290  "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\
291  "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\
292  "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\
293  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
294  "paddw %%mm3, %%mm4 \n\t"\
295  "movq %%mm2, %%mm0 \n\t"\
296  "movq %%mm5, %%mm6 \n\t"\
297  "movq %%mm4, %%mm3 \n\t"\
298  "punpcklwd %%mm2, %%mm2 \n\t"\
299  "punpcklwd %%mm5, %%mm5 \n\t"\
300  "punpcklwd %%mm4, %%mm4 \n\t"\
301  "paddw %%mm1, %%mm2 \n\t"\
302  "paddw %%mm1, %%mm5 \n\t"\
303  "paddw %%mm1, %%mm4 \n\t"\
304  "punpckhwd %%mm0, %%mm0 \n\t"\
305  "punpckhwd %%mm6, %%mm6 \n\t"\
306  "punpckhwd %%mm3, %%mm3 \n\t"\
307  "paddw %%mm7, %%mm0 \n\t"\
308  "paddw %%mm7, %%mm6 \n\t"\
309  "paddw %%mm7, %%mm3 \n\t"\
310  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
311  "packuswb %%mm0, %%mm2 \n\t"\
312  "packuswb %%mm6, %%mm5 \n\t"\
313  "packuswb %%mm3, %%mm4 \n\t"\
314 
315 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
316  "movq "#b", "#q2" \n\t" /* B */\
317  "movq "#r", "#t" \n\t" /* R */\
318  "punpcklbw "#g", "#b" \n\t" /* GBGBGBGB 0 */\
319  "punpcklbw "#a", "#r" \n\t" /* ARARARAR 0 */\
320  "punpckhbw "#g", "#q2" \n\t" /* GBGBGBGB 2 */\
321  "punpckhbw "#a", "#t" \n\t" /* ARARARAR 2 */\
322  "movq "#b", "#q0" \n\t" /* GBGBGBGB 0 */\
323  "movq "#q2", "#q3" \n\t" /* GBGBGBGB 2 */\
324  "punpcklwd "#r", "#q0" \n\t" /* ARGBARGB 0 */\
325  "punpckhwd "#r", "#b" \n\t" /* ARGBARGB 1 */\
326  "punpcklwd "#t", "#q2" \n\t" /* ARGBARGB 2 */\
327  "punpckhwd "#t", "#q3" \n\t" /* ARGBARGB 3 */\
328 \
329  MOVNTQ( q0, (dst, index, 4))\
330  MOVNTQ( b, 8(dst, index, 4))\
331  MOVNTQ( q2, 16(dst, index, 4))\
332  MOVNTQ( q3, 24(dst, index, 4))\
333 \
334  "add $8, "#index" \n\t"\
335  "cmp "#dstw", "#index" \n\t"\
336  " jb 1b \n\t"
337 #define WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t)
338 
339 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
340  const int16_t **lumSrc, int lumFilterSize,
341  const int16_t *chrFilter, const int16_t **chrUSrc,
342  const int16_t **chrVSrc,
343  int chrFilterSize, const int16_t **alpSrc,
344  uint8_t *dest, int dstW, int dstY)
345 {
346  x86_reg dummy=0;
347  x86_reg dstW_reg = dstW;
348  x86_reg uv_off = c->uv_offx2;
349 
350  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
353  "movq %%mm2, "U_TEMP"(%0) \n\t"
354  "movq %%mm4, "V_TEMP"(%0) \n\t"
355  "movq %%mm5, "Y_TEMP"(%0) \n\t"
357  "movq "Y_TEMP"(%0), %%mm5 \n\t"
358  "psraw $3, %%mm1 \n\t"
359  "psraw $3, %%mm7 \n\t"
360  "packuswb %%mm7, %%mm1 \n\t"
361  WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
363  } else {
366  "pcmpeqd %%mm7, %%mm7 \n\t"
367  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
369  }
370 }
371 
372 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
373  const int16_t **lumSrc, int lumFilterSize,
374  const int16_t *chrFilter, const int16_t **chrUSrc,
375  const int16_t **chrVSrc,
376  int chrFilterSize, const int16_t **alpSrc,
377  uint8_t *dest, int dstW, int dstY)
378 {
379  x86_reg dummy=0;
380  x86_reg dstW_reg = dstW;
381  x86_reg uv_off = c->uv_offx2;
382 
383  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
386  YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
387  "psraw $3, %%mm1 \n\t"
388  "psraw $3, %%mm7 \n\t"
389  "packuswb %%mm7, %%mm1 \n\t"
390  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
392  } else {
395  "pcmpeqd %%mm7, %%mm7 \n\t"
396  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
398  }
399 }
400 
401 #define REAL_WRITERGB16(dst, dstw, index) \
402  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
403  "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\
404  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
405  "psrlq $3, %%mm2 \n\t"\
406 \
407  "movq %%mm2, %%mm1 \n\t"\
408  "movq %%mm4, %%mm3 \n\t"\
409 \
410  "punpcklbw %%mm7, %%mm3 \n\t"\
411  "punpcklbw %%mm5, %%mm2 \n\t"\
412  "punpckhbw %%mm7, %%mm4 \n\t"\
413  "punpckhbw %%mm5, %%mm1 \n\t"\
414 \
415  "psllq $3, %%mm3 \n\t"\
416  "psllq $3, %%mm4 \n\t"\
417 \
418  "por %%mm3, %%mm2 \n\t"\
419  "por %%mm4, %%mm1 \n\t"\
420 \
421  MOVNTQ(%%mm2, (dst, index, 2))\
422  MOVNTQ(%%mm1, 8(dst, index, 2))\
423 \
424  "add $8, "#index" \n\t"\
425  "cmp "#dstw", "#index" \n\t"\
426  " jb 1b \n\t"
427 #define WRITERGB16(dst, dstw, index) REAL_WRITERGB16(dst, dstw, index)
428 
429 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
430  const int16_t **lumSrc, int lumFilterSize,
431  const int16_t *chrFilter, const int16_t **chrUSrc,
432  const int16_t **chrVSrc,
433  int chrFilterSize, const int16_t **alpSrc,
434  uint8_t *dest, int dstW, int dstY)
435 {
436  x86_reg dummy=0;
437  x86_reg dstW_reg = dstW;
438  x86_reg uv_off = c->uv_offx2;
439 
442  "pxor %%mm7, %%mm7 \n\t"
443  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
444 #ifdef DITHER1XBPP
445  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
446  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
447  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
448 #endif
449  WRITERGB16(%4, %5, %%REGa)
451 }
452 
453 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
454  const int16_t **lumSrc, int lumFilterSize,
455  const int16_t *chrFilter, const int16_t **chrUSrc,
456  const int16_t **chrVSrc,
457  int chrFilterSize, const int16_t **alpSrc,
458  uint8_t *dest, int dstW, int dstY)
459 {
460  x86_reg dummy=0;
461  x86_reg dstW_reg = dstW;
462  x86_reg uv_off = c->uv_offx2;
463 
466  "pxor %%mm7, %%mm7 \n\t"
467  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
468 #ifdef DITHER1XBPP
469  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
470  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
471  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
472 #endif
473  WRITERGB16(%4, %5, %%REGa)
475 }
476 
477 #define REAL_WRITERGB15(dst, dstw, index) \
478  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
479  "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\
480  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
481  "psrlq $3, %%mm2 \n\t"\
482  "psrlq $1, %%mm5 \n\t"\
483 \
484  "movq %%mm2, %%mm1 \n\t"\
485  "movq %%mm4, %%mm3 \n\t"\
486 \
487  "punpcklbw %%mm7, %%mm3 \n\t"\
488  "punpcklbw %%mm5, %%mm2 \n\t"\
489  "punpckhbw %%mm7, %%mm4 \n\t"\
490  "punpckhbw %%mm5, %%mm1 \n\t"\
491 \
492  "psllq $2, %%mm3 \n\t"\
493  "psllq $2, %%mm4 \n\t"\
494 \
495  "por %%mm3, %%mm2 \n\t"\
496  "por %%mm4, %%mm1 \n\t"\
497 \
498  MOVNTQ(%%mm2, (dst, index, 2))\
499  MOVNTQ(%%mm1, 8(dst, index, 2))\
500 \
501  "add $8, "#index" \n\t"\
502  "cmp "#dstw", "#index" \n\t"\
503  " jb 1b \n\t"
504 #define WRITERGB15(dst, dstw, index) REAL_WRITERGB15(dst, dstw, index)
505 
506 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
507  const int16_t **lumSrc, int lumFilterSize,
508  const int16_t *chrFilter, const int16_t **chrUSrc,
509  const int16_t **chrVSrc,
510  int chrFilterSize, const int16_t **alpSrc,
511  uint8_t *dest, int dstW, int dstY)
512 {
513  x86_reg dummy=0;
514  x86_reg dstW_reg = dstW;
515  x86_reg uv_off = c->uv_offx2;
516 
519  "pxor %%mm7, %%mm7 \n\t"
520  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
521 #ifdef DITHER1XBPP
522  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
523  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
524  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
525 #endif
526  WRITERGB15(%4, %5, %%REGa)
528 }
529 
530 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
531  const int16_t **lumSrc, int lumFilterSize,
532  const int16_t *chrFilter, const int16_t **chrUSrc,
533  const int16_t **chrVSrc,
534  int chrFilterSize, const int16_t **alpSrc,
535  uint8_t *dest, int dstW, int dstY)
536 {
537  x86_reg dummy=0;
538  x86_reg dstW_reg = dstW;
539  x86_reg uv_off = c->uv_offx2;
540 
543  "pxor %%mm7, %%mm7 \n\t"
544  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
545 #ifdef DITHER1XBPP
546  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
547  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
548  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
549 #endif
550  WRITERGB15(%4, %5, %%REGa)
552 }
553 
554 #define WRITEBGR24MMX(dst, dstw, index) \
555  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
556  "movq %%mm2, %%mm1 \n\t" /* B */\
557  "movq %%mm5, %%mm6 \n\t" /* R */\
558  "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\
559  "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\
560  "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\
561  "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\
562  "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\
563  "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\
564  "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\
565  "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\
566  "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\
567  "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\
568 \
569  "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\
570  "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\
571  "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\
572  "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\
573 \
574  "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\
575  "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\
576  "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\
577  "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\
578 \
579  "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\
580  "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\
581  "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\
582  "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\
583 \
584  "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\
585  "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\
586  "psllq $40, %%mm2 \n\t" /* GB000000 1 */\
587  "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\
588  MOVNTQ(%%mm0, (dst))\
589 \
590  "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\
591  "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\
592  "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\
593  "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\
594  MOVNTQ(%%mm6, 8(dst))\
595 \
596  "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\
597  "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\
598  "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\
599  MOVNTQ(%%mm5, 16(dst))\
600 \
601  "add $24, "#dst" \n\t"\
602 \
603  "add $8, "#index" \n\t"\
604  "cmp "#dstw", "#index" \n\t"\
605  " jb 1b \n\t"
606 
607 #define WRITEBGR24MMXEXT(dst, dstw, index) \
608  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
609  "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
610  "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
611  "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\
612  "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\
613  "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\
614 \
615  "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\
616  "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\
617  "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\
618 \
619  "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\
620  "por %%mm1, %%mm6 \n\t"\
621  "por %%mm3, %%mm6 \n\t"\
622  MOVNTQ(%%mm6, (dst))\
623 \
624  "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\
625  "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\
626  "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\
627  "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\
628 \
629  "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5 B4 B3 */\
630  "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\
631  "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\
632 \
633  "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\
634  "por %%mm3, %%mm6 \n\t"\
635  MOVNTQ(%%mm6, 8(dst))\
636 \
637  "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\
638  "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\
639  "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\
640 \
641  "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\
642  "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\
643  "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7 R6 R5 */\
644 \
645  "por %%mm1, %%mm3 \n\t"\
646  "por %%mm3, %%mm6 \n\t"\
647  MOVNTQ(%%mm6, 16(dst))\
648 \
649  "add $24, "#dst" \n\t"\
650 \
651  "add $8, "#index" \n\t"\
652  "cmp "#dstw", "#index" \n\t"\
653  " jb 1b \n\t"
654 
655 #if COMPILE_TEMPLATE_MMXEXT
656 #undef WRITEBGR24
657 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMXEXT(dst, dstw, index)
658 #else
659 #undef WRITEBGR24
660 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index)
661 #endif
662 
663 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
664  const int16_t **lumSrc, int lumFilterSize,
665  const int16_t *chrFilter, const int16_t **chrUSrc,
666  const int16_t **chrVSrc,
667  int chrFilterSize, const int16_t **alpSrc,
668  uint8_t *dest, int dstW, int dstY)
669 {
670  x86_reg dummy=0;
671  x86_reg dstW_reg = dstW;
672  x86_reg uv_off = c->uv_offx2;
673 
676  "pxor %%mm7, %%mm7 \n\t"
677  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
678  "add %4, %%"REG_c" \n\t"
679  WRITEBGR24(%%REGc, %5, %%REGa)
680  :: "r" (&c->redDither),
681  "m" (dummy), "m" (dummy), "m" (dummy),
682  "r" (dest), "m" (dstW_reg), "m"(uv_off)
683  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
684  );
685 }
686 
687 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
688  const int16_t **lumSrc, int lumFilterSize,
689  const int16_t *chrFilter, const int16_t **chrUSrc,
690  const int16_t **chrVSrc,
691  int chrFilterSize, const int16_t **alpSrc,
692  uint8_t *dest, int dstW, int dstY)
693 {
694  x86_reg dummy=0;
695  x86_reg dstW_reg = dstW;
696  x86_reg uv_off = c->uv_offx2;
697 
700  "pxor %%mm7, %%mm7 \n\t"
701  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c" \n\t" //FIXME optimize
702  "add %4, %%"REG_c" \n\t"
703  WRITEBGR24(%%REGc, %5, %%REGa)
704  :: "r" (&c->redDither),
705  "m" (dummy), "m" (dummy), "m" (dummy),
706  "r" (dest), "m" (dstW_reg), "m"(uv_off)
707  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
708  );
709 }
710 
711 #define REAL_WRITEYUY2(dst, dstw, index) \
712  "packuswb %%mm3, %%mm3 \n\t"\
713  "packuswb %%mm4, %%mm4 \n\t"\
714  "packuswb %%mm7, %%mm1 \n\t"\
715  "punpcklbw %%mm4, %%mm3 \n\t"\
716  "movq %%mm1, %%mm7 \n\t"\
717  "punpcklbw %%mm3, %%mm1 \n\t"\
718  "punpckhbw %%mm3, %%mm7 \n\t"\
719 \
720  MOVNTQ(%%mm1, (dst, index, 2))\
721  MOVNTQ(%%mm7, 8(dst, index, 2))\
722 \
723  "add $8, "#index" \n\t"\
724  "cmp "#dstw", "#index" \n\t"\
725  " jb 1b \n\t"
726 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index)
727 
728 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
729  const int16_t **lumSrc, int lumFilterSize,
730  const int16_t *chrFilter, const int16_t **chrUSrc,
731  const int16_t **chrVSrc,
732  int chrFilterSize, const int16_t **alpSrc,
733  uint8_t *dest, int dstW, int dstY)
734 {
735  x86_reg dummy=0;
736  x86_reg dstW_reg = dstW;
737  x86_reg uv_off = c->uv_offx2;
738 
740  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
741  "psraw $3, %%mm3 \n\t"
742  "psraw $3, %%mm4 \n\t"
743  "psraw $3, %%mm1 \n\t"
744  "psraw $3, %%mm7 \n\t"
745  WRITEYUY2(%4, %5, %%REGa)
747 }
748 
749 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
750  const int16_t **lumSrc, int lumFilterSize,
751  const int16_t *chrFilter, const int16_t **chrUSrc,
752  const int16_t **chrVSrc,
753  int chrFilterSize, const int16_t **alpSrc,
754  uint8_t *dest, int dstW, int dstY)
755 {
756  x86_reg dummy=0;
757  x86_reg dstW_reg = dstW;
758  x86_reg uv_off = c->uv_offx2;
759 
761  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
762  "psraw $3, %%mm3 \n\t"
763  "psraw $3, %%mm4 \n\t"
764  "psraw $3, %%mm1 \n\t"
765  "psraw $3, %%mm7 \n\t"
766  WRITEYUY2(%4, %5, %%REGa)
768 }
769 
770 #define REAL_YSCALEYUV2RGB_UV(index, c) \
771  "xor "#index", "#index" \n\t"\
772  ".p2align 4 \n\t"\
773  "1: \n\t"\
774  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
775  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
776  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
777  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
778  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
779  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
780  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
781  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
782  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
783  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
784  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
785  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
786  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
787  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
788  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
789  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
790  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
791  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
792  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
793  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
794  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
795  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
796 
797 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
798  "movq ("#b1", "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
799  "movq ("#b2", "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
800  "movq 8("#b1", "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
801  "movq 8("#b2", "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
802  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
803  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
804  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
805  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
806  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
807  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
808  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
809  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
810 
811 #define REAL_YSCALEYUV2RGB_COEFF(c) \
812  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
813  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
814  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
815  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
816  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
817  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
818  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
819  "paddw %%mm3, %%mm4 \n\t"\
820  "movq %%mm2, %%mm0 \n\t"\
821  "movq %%mm5, %%mm6 \n\t"\
822  "movq %%mm4, %%mm3 \n\t"\
823  "punpcklwd %%mm2, %%mm2 \n\t"\
824  "punpcklwd %%mm5, %%mm5 \n\t"\
825  "punpcklwd %%mm4, %%mm4 \n\t"\
826  "paddw %%mm1, %%mm2 \n\t"\
827  "paddw %%mm1, %%mm5 \n\t"\
828  "paddw %%mm1, %%mm4 \n\t"\
829  "punpckhwd %%mm0, %%mm0 \n\t"\
830  "punpckhwd %%mm6, %%mm6 \n\t"\
831  "punpckhwd %%mm3, %%mm3 \n\t"\
832  "paddw %%mm7, %%mm0 \n\t"\
833  "paddw %%mm7, %%mm6 \n\t"\
834  "paddw %%mm7, %%mm3 \n\t"\
835  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
836  "packuswb %%mm0, %%mm2 \n\t"\
837  "packuswb %%mm6, %%mm5 \n\t"\
838  "packuswb %%mm3, %%mm4 \n\t"\
839 
840 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
841 
842 #define YSCALEYUV2RGB(index, c) \
843  REAL_YSCALEYUV2RGB_UV(index, c) \
844  REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
845  REAL_YSCALEYUV2RGB_COEFF(c)
846 
847 /**
848  * vertical bilinear scale YV12 to RGB
849  */
850 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
851  const int16_t *ubuf[2], const int16_t *vbuf[2],
852  const int16_t *abuf[2], uint8_t *dest,
853  int dstW, int yalpha, int uvalpha, int y)
854 {
855  const int16_t *buf0 = buf[0], *buf1 = buf[1],
856  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
857 
858  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
859  const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
860 #if ARCH_X86_64
861  __asm__ volatile(
862  YSCALEYUV2RGB(%%r8, %5)
863  YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
864  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
865  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
866  "packuswb %%mm7, %%mm1 \n\t"
867  WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
868  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
869  "a" (&c->redDither),
870  "r" (abuf0), "r" (abuf1)
871  : "%r8"
872  );
873 #else
874  c->u_temp=(intptr_t)abuf0;
875  c->v_temp=(intptr_t)abuf1;
876  __asm__ volatile(
877  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
878  "mov %4, %%"REG_b" \n\t"
879  "push %%"REG_BP" \n\t"
880  YSCALEYUV2RGB(%%REGBP, %5)
881  "push %0 \n\t"
882  "push %1 \n\t"
883  "mov "U_TEMP"(%5), %0 \n\t"
884  "mov "V_TEMP"(%5), %1 \n\t"
885  YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
886  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
887  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
888  "packuswb %%mm7, %%mm1 \n\t"
889  "pop %1 \n\t"
890  "pop %0 \n\t"
891  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
892  "pop %%"REG_BP" \n\t"
893  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
894  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
895  "a" (&c->redDither)
896  );
897 #endif
898  } else {
899  __asm__ volatile(
900  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
901  "mov %4, %%"REG_b" \n\t"
902  "push %%"REG_BP" \n\t"
903  YSCALEYUV2RGB(%%REGBP, %5)
904  "pcmpeqd %%mm7, %%mm7 \n\t"
905  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
906  "pop %%"REG_BP" \n\t"
907  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
908  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
909  "a" (&c->redDither)
910  );
911  }
912 }
913 
914 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
915  const int16_t *ubuf[2], const int16_t *vbuf[2],
916  const int16_t *abuf[2], uint8_t *dest,
917  int dstW, int yalpha, int uvalpha, int y)
918 {
919  const int16_t *buf0 = buf[0], *buf1 = buf[1],
920  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
921 
922  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
923  __asm__ volatile(
924  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
925  "mov %4, %%"REG_b" \n\t"
926  "push %%"REG_BP" \n\t"
927  YSCALEYUV2RGB(%%REGBP, %5)
928  "pxor %%mm7, %%mm7 \n\t"
929  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
930  "pop %%"REG_BP" \n\t"
931  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
932  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
933  "a" (&c->redDither)
934  );
935 }
936 
937 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
938  const int16_t *ubuf[2], const int16_t *vbuf[2],
939  const int16_t *abuf[2], uint8_t *dest,
940  int dstW, int yalpha, int uvalpha, int y)
941 {
942  const int16_t *buf0 = buf[0], *buf1 = buf[1],
943  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
944 
945  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
946  __asm__ volatile(
947  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
948  "mov %4, %%"REG_b" \n\t"
949  "push %%"REG_BP" \n\t"
950  YSCALEYUV2RGB(%%REGBP, %5)
951  "pxor %%mm7, %%mm7 \n\t"
952  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
953 #ifdef DITHER1XBPP
954  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
955  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
956  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
957 #endif
958  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
959  "pop %%"REG_BP" \n\t"
960  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
961  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
962  "a" (&c->redDither)
963  );
964 }
965 
966 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
967  const int16_t *ubuf[2], const int16_t *vbuf[2],
968  const int16_t *abuf[2], uint8_t *dest,
969  int dstW, int yalpha, int uvalpha, int y)
970 {
971  const int16_t *buf0 = buf[0], *buf1 = buf[1],
972  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
973 
974  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
975  __asm__ volatile(
976  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
977  "mov %4, %%"REG_b" \n\t"
978  "push %%"REG_BP" \n\t"
979  YSCALEYUV2RGB(%%REGBP, %5)
980  "pxor %%mm7, %%mm7 \n\t"
981  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
982 #ifdef DITHER1XBPP
983  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
984  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
985  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
986 #endif
987  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
988  "pop %%"REG_BP" \n\t"
989  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
990  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
991  "a" (&c->redDither)
992  );
993 }
994 
995 #define REAL_YSCALEYUV2PACKED(index, c) \
996  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
997  "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1 \n\t"\
998  "psraw $3, %%mm0 \n\t"\
999  "psraw $3, %%mm1 \n\t"\
1000  "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1001  "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1002  "xor "#index", "#index" \n\t"\
1003  ".p2align 4 \n\t"\
1004  "1: \n\t"\
1005  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1006  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1007  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1008  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1009  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1010  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1011  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1012  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1013  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
1014  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1015  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1016  "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1017  "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1018  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1019  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1020  "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
1021  "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
1022  "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
1023  "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
1024  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
1025  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
1026  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1027  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1028  "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1029  "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1030  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1031  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1032 
1033 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c)
1034 
1035 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
1036  const int16_t *ubuf[2], const int16_t *vbuf[2],
1037  const int16_t *abuf[2], uint8_t *dest,
1038  int dstW, int yalpha, int uvalpha, int y)
1039 {
1040  const int16_t *buf0 = buf[0], *buf1 = buf[1],
1041  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1042 
1043  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1044  __asm__ volatile(
1045  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1046  "mov %4, %%"REG_b" \n\t"
1047  "push %%"REG_BP" \n\t"
1048  YSCALEYUV2PACKED(%%REGBP, %5)
1049  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1050  "pop %%"REG_BP" \n\t"
1051  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1052  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1053  "a" (&c->redDither)
1054  );
1055 }
1056 
1057 #define REAL_YSCALEYUV2RGB1(index, c) \
1058  "xor "#index", "#index" \n\t"\
1059  ".p2align 4 \n\t"\
1060  "1: \n\t"\
1061  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1062  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1063  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1064  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1065  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1066  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1067  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1068  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1069  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1070  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1071  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1072  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1073  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1074  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1075  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1076  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1077  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1078  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1079  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1080  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1081  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1082  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1083  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1084  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1085  "paddw %%mm3, %%mm4 \n\t"\
1086  "movq %%mm2, %%mm0 \n\t"\
1087  "movq %%mm5, %%mm6 \n\t"\
1088  "movq %%mm4, %%mm3 \n\t"\
1089  "punpcklwd %%mm2, %%mm2 \n\t"\
1090  "punpcklwd %%mm5, %%mm5 \n\t"\
1091  "punpcklwd %%mm4, %%mm4 \n\t"\
1092  "paddw %%mm1, %%mm2 \n\t"\
1093  "paddw %%mm1, %%mm5 \n\t"\
1094  "paddw %%mm1, %%mm4 \n\t"\
1095  "punpckhwd %%mm0, %%mm0 \n\t"\
1096  "punpckhwd %%mm6, %%mm6 \n\t"\
1097  "punpckhwd %%mm3, %%mm3 \n\t"\
1098  "paddw %%mm7, %%mm0 \n\t"\
1099  "paddw %%mm7, %%mm6 \n\t"\
1100  "paddw %%mm7, %%mm3 \n\t"\
1101  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1102  "packuswb %%mm0, %%mm2 \n\t"\
1103  "packuswb %%mm6, %%mm5 \n\t"\
1104  "packuswb %%mm3, %%mm4 \n\t"\
1105 
1106 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c)
1107 
1108 // do vertical chrominance interpolation
1109 #define REAL_YSCALEYUV2RGB1b(index, c) \
1110  "xor "#index", "#index" \n\t"\
1111  ".p2align 4 \n\t"\
1112  "1: \n\t"\
1113  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1114  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1115  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1116  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1117  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1118  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1119  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1120  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1121  "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\
1122  "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\
1123  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1124  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1125  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1126  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1127  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1128  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1129  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1130  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1131  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1132  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1133  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1134  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1135  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1136  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1137  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1138  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1139  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1140  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1141  "paddw %%mm3, %%mm4 \n\t"\
1142  "movq %%mm2, %%mm0 \n\t"\
1143  "movq %%mm5, %%mm6 \n\t"\
1144  "movq %%mm4, %%mm3 \n\t"\
1145  "punpcklwd %%mm2, %%mm2 \n\t"\
1146  "punpcklwd %%mm5, %%mm5 \n\t"\
1147  "punpcklwd %%mm4, %%mm4 \n\t"\
1148  "paddw %%mm1, %%mm2 \n\t"\
1149  "paddw %%mm1, %%mm5 \n\t"\
1150  "paddw %%mm1, %%mm4 \n\t"\
1151  "punpckhwd %%mm0, %%mm0 \n\t"\
1152  "punpckhwd %%mm6, %%mm6 \n\t"\
1153  "punpckhwd %%mm3, %%mm3 \n\t"\
1154  "paddw %%mm7, %%mm0 \n\t"\
1155  "paddw %%mm7, %%mm6 \n\t"\
1156  "paddw %%mm7, %%mm3 \n\t"\
1157  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1158  "packuswb %%mm0, %%mm2 \n\t"\
1159  "packuswb %%mm6, %%mm5 \n\t"\
1160  "packuswb %%mm3, %%mm4 \n\t"\
1161 
1162 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c)
1163 
1164 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1165  "movq (%1, "#index", 2), %%mm7 \n\t" /* abuf0[index ] */\
1166  "movq 8(%1, "#index", 2), %%mm1 \n\t" /* abuf0[index+4] */\
1167  "psraw $7, %%mm7 \n\t" /* abuf0[index ] >>7 */\
1168  "psraw $7, %%mm1 \n\t" /* abuf0[index+4] >>7 */\
1169  "packuswb %%mm1, %%mm7 \n\t"
1170 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1171 
1172 /**
1173  * YV12 to RGB without scaling or interpolating
1174  */
1175 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1176  const int16_t *ubuf[2], const int16_t *vbuf[2],
1177  const int16_t *abuf0, uint8_t *dest,
1178  int dstW, int uvalpha, int y)
1179 {
1180  const int16_t *ubuf0 = ubuf[0];
1181  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1182 
1183  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1184  const int16_t *ubuf1 = ubuf[0];
1185  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1186  __asm__ volatile(
1187  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1188  "mov %4, %%"REG_b" \n\t"
1189  "push %%"REG_BP" \n\t"
1190  YSCALEYUV2RGB1(%%REGBP, %5)
1191  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1192  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1193  "pop %%"REG_BP" \n\t"
1194  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1195  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1196  "a" (&c->redDither)
1197  );
1198  } else {
1199  __asm__ volatile(
1200  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1201  "mov %4, %%"REG_b" \n\t"
1202  "push %%"REG_BP" \n\t"
1203  YSCALEYUV2RGB1(%%REGBP, %5)
1204  "pcmpeqd %%mm7, %%mm7 \n\t"
1205  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1206  "pop %%"REG_BP" \n\t"
1207  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1208  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1209  "a" (&c->redDither)
1210  );
1211  }
1212  } else {
1213  const int16_t *ubuf1 = ubuf[1];
1214  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1215  __asm__ volatile(
1216  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1217  "mov %4, %%"REG_b" \n\t"
1218  "push %%"REG_BP" \n\t"
1219  YSCALEYUV2RGB1b(%%REGBP, %5)
1220  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1221  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1222  "pop %%"REG_BP" \n\t"
1223  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1224  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1225  "a" (&c->redDither)
1226  );
1227  } else {
1228  __asm__ volatile(
1229  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1230  "mov %4, %%"REG_b" \n\t"
1231  "push %%"REG_BP" \n\t"
1232  YSCALEYUV2RGB1b(%%REGBP, %5)
1233  "pcmpeqd %%mm7, %%mm7 \n\t"
1234  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1235  "pop %%"REG_BP" \n\t"
1236  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1237  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1238  "a" (&c->redDither)
1239  );
1240  }
1241  }
1242 }
1243 
1244 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1245  const int16_t *ubuf[2], const int16_t *vbuf[2],
1246  const int16_t *abuf0, uint8_t *dest,
1247  int dstW, int uvalpha, int y)
1248 {
1249  const int16_t *ubuf0 = ubuf[0];
1250  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1251 
1252  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1253  const int16_t *ubuf1 = ubuf[0];
1254  __asm__ volatile(
1255  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1256  "mov %4, %%"REG_b" \n\t"
1257  "push %%"REG_BP" \n\t"
1258  YSCALEYUV2RGB1(%%REGBP, %5)
1259  "pxor %%mm7, %%mm7 \n\t"
1260  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1261  "pop %%"REG_BP" \n\t"
1262  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1263  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1264  "a" (&c->redDither)
1265  );
1266  } else {
1267  const int16_t *ubuf1 = ubuf[1];
1268  __asm__ volatile(
1269  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1270  "mov %4, %%"REG_b" \n\t"
1271  "push %%"REG_BP" \n\t"
1272  YSCALEYUV2RGB1b(%%REGBP, %5)
1273  "pxor %%mm7, %%mm7 \n\t"
1274  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1275  "pop %%"REG_BP" \n\t"
1276  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1277  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1278  "a" (&c->redDither)
1279  );
1280  }
1281 }
1282 
1283 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1284  const int16_t *ubuf[2], const int16_t *vbuf[2],
1285  const int16_t *abuf0, uint8_t *dest,
1286  int dstW, int uvalpha, int y)
1287 {
1288  const int16_t *ubuf0 = ubuf[0];
1289  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1290 
1291  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1292  const int16_t *ubuf1 = ubuf[0];
1293  __asm__ volatile(
1294  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1295  "mov %4, %%"REG_b" \n\t"
1296  "push %%"REG_BP" \n\t"
1297  YSCALEYUV2RGB1(%%REGBP, %5)
1298  "pxor %%mm7, %%mm7 \n\t"
1299  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1300 #ifdef DITHER1XBPP
1301  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1302  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1303  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1304 #endif
1305  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1306  "pop %%"REG_BP" \n\t"
1307  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1308  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1309  "a" (&c->redDither)
1310  );
1311  } else {
1312  const int16_t *ubuf1 = ubuf[1];
1313  __asm__ volatile(
1314  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1315  "mov %4, %%"REG_b" \n\t"
1316  "push %%"REG_BP" \n\t"
1317  YSCALEYUV2RGB1b(%%REGBP, %5)
1318  "pxor %%mm7, %%mm7 \n\t"
1319  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1320 #ifdef DITHER1XBPP
1321  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1322  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1323  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1324 #endif
1325  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1326  "pop %%"REG_BP" \n\t"
1327  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1328  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1329  "a" (&c->redDither)
1330  );
1331  }
1332 }
1333 
1334 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1335  const int16_t *ubuf[2], const int16_t *vbuf[2],
1336  const int16_t *abuf0, uint8_t *dest,
1337  int dstW, int uvalpha, int y)
1338 {
1339  const int16_t *ubuf0 = ubuf[0];
1340  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1341 
1342  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1343  const int16_t *ubuf1 = ubuf[0];
1344  __asm__ volatile(
1345  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1346  "mov %4, %%"REG_b" \n\t"
1347  "push %%"REG_BP" \n\t"
1348  YSCALEYUV2RGB1(%%REGBP, %5)
1349  "pxor %%mm7, %%mm7 \n\t"
1350  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1351 #ifdef DITHER1XBPP
1352  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1353  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1354  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1355 #endif
1356  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1357  "pop %%"REG_BP" \n\t"
1358  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1359  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1360  "a" (&c->redDither)
1361  );
1362  } else {
1363  const int16_t *ubuf1 = ubuf[1];
1364  __asm__ volatile(
1365  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1366  "mov %4, %%"REG_b" \n\t"
1367  "push %%"REG_BP" \n\t"
1368  YSCALEYUV2RGB1b(%%REGBP, %5)
1369  "pxor %%mm7, %%mm7 \n\t"
1370  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1371 #ifdef DITHER1XBPP
1372  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1373  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1374  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1375 #endif
1376  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1377  "pop %%"REG_BP" \n\t"
1378  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1379  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1380  "a" (&c->redDither)
1381  );
1382  }
1383 }
1384 
1385 #define REAL_YSCALEYUV2PACKED1(index, c) \
1386  "xor "#index", "#index" \n\t"\
1387  ".p2align 4 \n\t"\
1388  "1: \n\t"\
1389  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1390  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1391  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1392  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1393  "psraw $7, %%mm3 \n\t" \
1394  "psraw $7, %%mm4 \n\t" \
1395  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1396  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1397  "psraw $7, %%mm1 \n\t" \
1398  "psraw $7, %%mm7 \n\t" \
1399 
1400 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c)
1401 
1402 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1403  "xor "#index", "#index" \n\t"\
1404  ".p2align 4 \n\t"\
1405  "1: \n\t"\
1406  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1407  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1408  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1409  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1410  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1411  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1412  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1413  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1414  "psrlw $8, %%mm3 \n\t" \
1415  "psrlw $8, %%mm4 \n\t" \
1416  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1417  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1418  "psraw $7, %%mm1 \n\t" \
1419  "psraw $7, %%mm7 \n\t"
1420 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c)
1421 
1422 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1423  const int16_t *ubuf[2], const int16_t *vbuf[2],
1424  const int16_t *abuf0, uint8_t *dest,
1425  int dstW, int uvalpha, int y)
1426 {
1427  const int16_t *ubuf0 = ubuf[0];
1428  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1429 
1430  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1431  const int16_t *ubuf1 = ubuf[0];
1432  __asm__ volatile(
1433  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1434  "mov %4, %%"REG_b" \n\t"
1435  "push %%"REG_BP" \n\t"
1436  YSCALEYUV2PACKED1(%%REGBP, %5)
1437  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1438  "pop %%"REG_BP" \n\t"
1439  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1440  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1441  "a" (&c->redDither)
1442  );
1443  } else {
1444  const int16_t *ubuf1 = ubuf[1];
1445  __asm__ volatile(
1446  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1447  "mov %4, %%"REG_b" \n\t"
1448  "push %%"REG_BP" \n\t"
1449  YSCALEYUV2PACKED1b(%%REGBP, %5)
1450  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1451  "pop %%"REG_BP" \n\t"
1452  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1453  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1454  "a" (&c->redDither)
1455  );
1456  }
1457 }
1458 
1459 #if COMPILE_TEMPLATE_MMXEXT
1460 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
1461  int dstWidth, const uint8_t *src,
1462  int srcW, int xInc)
1463 {
1464  int32_t *filterPos = c->hLumFilterPos;
1465  int16_t *filter = c->hLumFilter;
1466  void *mmxextFilterCode = c->lumMmxextFilterCode;
1467  int i;
1468 #if defined(PIC)
1469  uint64_t ebxsave;
1470 #endif
1471 #if ARCH_X86_64
1472  uint64_t retsave;
1473 #endif
1474 
1475  __asm__ volatile(
1476 #if defined(PIC)
1477  "mov %%"REG_b", %5 \n\t"
1478 #if ARCH_X86_64
1479  "mov -8(%%rsp), %%"REG_a" \n\t"
1480  "mov %%"REG_a", %6 \n\t"
1481 #endif
1482 #else
1483 #if ARCH_X86_64
1484  "mov -8(%%rsp), %%"REG_a" \n\t"
1485  "mov %%"REG_a", %5 \n\t"
1486 #endif
1487 #endif
1488  "pxor %%mm7, %%mm7 \n\t"
1489  "mov %0, %%"REG_c" \n\t"
1490  "mov %1, %%"REG_D" \n\t"
1491  "mov %2, %%"REG_d" \n\t"
1492  "mov %3, %%"REG_b" \n\t"
1493  "xor %%"REG_a", %%"REG_a" \n\t" // i
1494  PREFETCH" (%%"REG_c") \n\t"
1495  PREFETCH" 32(%%"REG_c") \n\t"
1496  PREFETCH" 64(%%"REG_c") \n\t"
1497 
1498 #if ARCH_X86_64
1499 #define CALL_MMXEXT_FILTER_CODE \
1500  "movl (%%"REG_b"), %%esi \n\t"\
1501  "call *%4 \n\t"\
1502  "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\
1503  "add %%"REG_S", %%"REG_c" \n\t"\
1504  "add %%"REG_a", %%"REG_D" \n\t"\
1505  "xor %%"REG_a", %%"REG_a" \n\t"\
1506 
1507 #else
1508 #define CALL_MMXEXT_FILTER_CODE \
1509  "movl (%%"REG_b"), %%esi \n\t"\
1510  "call *%4 \n\t"\
1511  "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
1512  "add %%"REG_a", %%"REG_D" \n\t"\
1513  "xor %%"REG_a", %%"REG_a" \n\t"\
1514 
1515 #endif /* ARCH_X86_64 */
1516 
1517  CALL_MMXEXT_FILTER_CODE
1518  CALL_MMXEXT_FILTER_CODE
1519  CALL_MMXEXT_FILTER_CODE
1520  CALL_MMXEXT_FILTER_CODE
1521  CALL_MMXEXT_FILTER_CODE
1522  CALL_MMXEXT_FILTER_CODE
1523  CALL_MMXEXT_FILTER_CODE
1524  CALL_MMXEXT_FILTER_CODE
1525 
1526 #if defined(PIC)
1527  "mov %5, %%"REG_b" \n\t"
1528 #if ARCH_X86_64
1529  "mov %6, %%"REG_a" \n\t"
1530  "mov %%"REG_a", -8(%%rsp) \n\t"
1531 #endif
1532 #else
1533 #if ARCH_X86_64
1534  "mov %5, %%"REG_a" \n\t"
1535  "mov %%"REG_a", -8(%%rsp) \n\t"
1536 #endif
1537 #endif
1538  :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
1539  "m" (mmxextFilterCode)
1540 #if defined(PIC)
1541  ,"m" (ebxsave)
1542 #endif
1543 #if ARCH_X86_64
1544  ,"m"(retsave)
1545 #endif
1546  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1547 #if !defined(PIC)
1548  ,"%"REG_b
1549 #endif
1550  );
1551 
1552  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
1553  dst[i] = src[srcW-1]*128;
1554 }
1555 
1556 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
1557  int dstWidth, const uint8_t *src1,
1558  const uint8_t *src2, int srcW, int xInc)
1559 {
1560  int32_t *filterPos = c->hChrFilterPos;
1561  int16_t *filter = c->hChrFilter;
1562  void *mmxextFilterCode = c->chrMmxextFilterCode;
1563  int i;
1564 #if defined(PIC)
1565  DECLARE_ALIGNED(8, uint64_t, ebxsave);
1566 #endif
1567 #if ARCH_X86_64
1568  DECLARE_ALIGNED(8, uint64_t, retsave);
1569 #endif
1570 
1571  __asm__ volatile(
1572 #if defined(PIC)
1573  "mov %%"REG_b", %7 \n\t"
1574 #if ARCH_X86_64
1575  "mov -8(%%rsp), %%"REG_a" \n\t"
1576  "mov %%"REG_a", %8 \n\t"
1577 #endif
1578 #else
1579 #if ARCH_X86_64
1580  "mov -8(%%rsp), %%"REG_a" \n\t"
1581  "mov %%"REG_a", %7 \n\t"
1582 #endif
1583 #endif
1584  "pxor %%mm7, %%mm7 \n\t"
1585  "mov %0, %%"REG_c" \n\t"
1586  "mov %1, %%"REG_D" \n\t"
1587  "mov %2, %%"REG_d" \n\t"
1588  "mov %3, %%"REG_b" \n\t"
1589  "xor %%"REG_a", %%"REG_a" \n\t" // i
1590  PREFETCH" (%%"REG_c") \n\t"
1591  PREFETCH" 32(%%"REG_c") \n\t"
1592  PREFETCH" 64(%%"REG_c") \n\t"
1593 
1594  CALL_MMXEXT_FILTER_CODE
1595  CALL_MMXEXT_FILTER_CODE
1596  CALL_MMXEXT_FILTER_CODE
1597  CALL_MMXEXT_FILTER_CODE
1598  "xor %%"REG_a", %%"REG_a" \n\t" // i
1599  "mov %5, %%"REG_c" \n\t" // src
1600  "mov %6, %%"REG_D" \n\t" // buf2
1601  PREFETCH" (%%"REG_c") \n\t"
1602  PREFETCH" 32(%%"REG_c") \n\t"
1603  PREFETCH" 64(%%"REG_c") \n\t"
1604 
1605  CALL_MMXEXT_FILTER_CODE
1606  CALL_MMXEXT_FILTER_CODE
1607  CALL_MMXEXT_FILTER_CODE
1608  CALL_MMXEXT_FILTER_CODE
1609 
1610 #if defined(PIC)
1611  "mov %7, %%"REG_b" \n\t"
1612 #if ARCH_X86_64
1613  "mov %8, %%"REG_a" \n\t"
1614  "mov %%"REG_a", -8(%%rsp) \n\t"
1615 #endif
1616 #else
1617 #if ARCH_X86_64
1618  "mov %7, %%"REG_a" \n\t"
1619  "mov %%"REG_a", -8(%%rsp) \n\t"
1620 #endif
1621 #endif
1622  :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
1623  "m" (mmxextFilterCode), "m" (src2), "m"(dst2)
1624 #if defined(PIC)
1625  ,"m" (ebxsave)
1626 #endif
1627 #if ARCH_X86_64
1628  ,"m"(retsave)
1629 #endif
1630  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1631 #if !defined(PIC)
1632  ,"%"REG_b
1633 #endif
1634  );
1635 
1636  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
1637  dst1[i] = src1[srcW-1]*128;
1638  dst2[i] = src2[srcW-1]*128;
1639  }
1640 }
1641 #endif /* COMPILE_TEMPLATE_MMXEXT */
1642 
1643 static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
1644 {
1645  enum AVPixelFormat dstFormat = c->dstFormat;
1646 
1647  c->use_mmx_vfilter= 0;
1648  if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != AV_PIX_FMT_NV12
1649  && dstFormat != AV_PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
1650  if (c->flags & SWS_ACCURATE_RND) {
1651  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1652  switch (c->dstFormat) {
1653  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X_ar); break;
1654  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X_ar); break;
1655  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X_ar); break;
1656  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X_ar); break;
1657  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1658  default: break;
1659  }
1660  }
1661  } else {
1662  c->use_mmx_vfilter= 1;
1663  c->yuv2planeX = RENAME(yuv2yuvX );
1664  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1665  switch (c->dstFormat) {
1666  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X); break;
1667  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X); break;
1668  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X); break;
1669  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X); break;
1670  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1671  default: break;
1672  }
1673  }
1674  }
1675  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1676  switch (c->dstFormat) {
1677  case AV_PIX_FMT_RGB32:
1678  c->yuv2packed1 = RENAME(yuv2rgb32_1);
1679  c->yuv2packed2 = RENAME(yuv2rgb32_2);
1680  break;
1681  case AV_PIX_FMT_BGR24:
1682  c->yuv2packed1 = RENAME(yuv2bgr24_1);
1683  c->yuv2packed2 = RENAME(yuv2bgr24_2);
1684  break;
1685  case AV_PIX_FMT_RGB555:
1686  c->yuv2packed1 = RENAME(yuv2rgb555_1);
1687  c->yuv2packed2 = RENAME(yuv2rgb555_2);
1688  break;
1689  case AV_PIX_FMT_RGB565:
1690  c->yuv2packed1 = RENAME(yuv2rgb565_1);
1691  c->yuv2packed2 = RENAME(yuv2rgb565_2);
1692  break;
1693  case AV_PIX_FMT_YUYV422:
1694  c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1695  c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1696  break;
1697  default:
1698  break;
1699  }
1700  }
1701  }
1702 
1703  if (c->srcBpc == 8 && c->dstBpc <= 14) {
1704  // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1705 #if COMPILE_TEMPLATE_MMXEXT
1706  if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1707  c->hyscale_fast = RENAME(hyscale_fast);
1708  c->hcscale_fast = RENAME(hcscale_fast);
1709  } else {
1710 #endif /* COMPILE_TEMPLATE_MMXEXT */
1711  c->hyscale_fast = NULL;
1712  c->hcscale_fast = NULL;
1713 #if COMPILE_TEMPLATE_MMXEXT
1714  }
1715 #endif /* COMPILE_TEMPLATE_MMXEXT */
1716  }
1717 }