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  __asm__ volatile(\
75  "psraw $4, %%mm3\n\t"
76  "psraw $4, %%mm4\n\t"
77  "movq %%mm3, %%mm6\n\t"
78  "movq %%mm4, %%mm7\n\t"
79  "movl %3, %%ecx\n\t"
80  "mov %0, %%"REG_d" \n\t"\
81  "mov (%%"REG_d"), %%"REG_S" \n\t"\
82  ".p2align 4 \n\t" /* FIXME Unroll? */\
83  "1: \n\t"\
84  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
85  "movq (%%"REG_S", %%"REG_c", 2), %%mm2 \n\t" /* srcData */\
86  "movq 8(%%"REG_S", %%"REG_c", 2), %%mm5 \n\t" /* srcData */\
87  "add $16, %%"REG_d" \n\t"\
88  "mov (%%"REG_d"), %%"REG_S" \n\t"\
89  "test %%"REG_S", %%"REG_S" \n\t"\
90  "pmulhw %%mm0, %%mm2 \n\t"\
91  "pmulhw %%mm0, %%mm5 \n\t"\
92  "paddw %%mm2, %%mm3 \n\t"\
93  "paddw %%mm5, %%mm4 \n\t"\
94  " jnz 1b \n\t"\
95  "psraw $3, %%mm3 \n\t"\
96  "psraw $3, %%mm4 \n\t"\
97  "packuswb %%mm4, %%mm3 \n\t"
98  MOVNTQ2 " %%mm3, (%1, %%"REG_c")\n\t"
99  "add $8, %%"REG_c" \n\t"\
100  "cmp %2, %%"REG_c" \n\t"\
101  "movq %%mm6, %%mm3\n\t"
102  "movq %%mm7, %%mm4\n\t"
103  "mov %0, %%"REG_d" \n\t"\
104  "mov (%%"REG_d"), %%"REG_S" \n\t"\
105  "jb 1b \n\t"\
106  :: "g" (filter),
107  "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
108  : "%"REG_d, "%"REG_S, "%"REG_c
109  );
110 }
111 
112 #define YSCALEYUV2PACKEDX_UV \
113  __asm__ volatile(\
114  "xor %%"REG_a", %%"REG_a" \n\t"\
115  ".p2align 4 \n\t"\
116  "nop \n\t"\
117  "1: \n\t"\
118  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
119  "mov (%%"REG_d"), %%"REG_S" \n\t"\
120  "movq "VROUNDER_OFFSET"(%0), %%mm3 \n\t"\
121  "movq %%mm3, %%mm4 \n\t"\
122  ".p2align 4 \n\t"\
123  "2: \n\t"\
124  "movq 8(%%"REG_d"), %%mm0 \n\t" /* filterCoeff */\
125  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* UsrcData */\
126  "add %6, %%"REG_S" \n\t" \
127  "movq (%%"REG_S", %%"REG_a"), %%mm5 \n\t" /* VsrcData */\
128  "add $16, %%"REG_d" \n\t"\
129  "mov (%%"REG_d"), %%"REG_S" \n\t"\
130  "pmulhw %%mm0, %%mm2 \n\t"\
131  "pmulhw %%mm0, %%mm5 \n\t"\
132  "paddw %%mm2, %%mm3 \n\t"\
133  "paddw %%mm5, %%mm4 \n\t"\
134  "test %%"REG_S", %%"REG_S" \n\t"\
135  " jnz 2b \n\t"\
136 
137 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
138  "lea "offset"(%0), %%"REG_d" \n\t"\
139  "mov (%%"REG_d"), %%"REG_S" \n\t"\
140  "movq "VROUNDER_OFFSET"(%0), "#dst1" \n\t"\
141  "movq "#dst1", "#dst2" \n\t"\
142  ".p2align 4 \n\t"\
143  "2: \n\t"\
144  "movq 8(%%"REG_d"), "#coeff" \n\t" /* filterCoeff */\
145  "movq (%%"REG_S", %%"REG_a", 2), "#src1" \n\t" /* Y1srcData */\
146  "movq 8(%%"REG_S", %%"REG_a", 2), "#src2" \n\t" /* Y2srcData */\
147  "add $16, %%"REG_d" \n\t"\
148  "mov (%%"REG_d"), %%"REG_S" \n\t"\
149  "pmulhw "#coeff", "#src1" \n\t"\
150  "pmulhw "#coeff", "#src2" \n\t"\
151  "paddw "#src1", "#dst1" \n\t"\
152  "paddw "#src2", "#dst2" \n\t"\
153  "test %%"REG_S", %%"REG_S" \n\t"\
154  " jnz 2b \n\t"\
155 
156 #define YSCALEYUV2PACKEDX \
157  YSCALEYUV2PACKEDX_UV \
158  YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
159 
160 #define YSCALEYUV2PACKEDX_END \
161  :: "r" (&c->redDither), \
162  "m" (dummy), "m" (dummy), "m" (dummy),\
163  "r" (dest), "m" (dstW_reg), "m"(uv_off) \
164  : "%"REG_a, "%"REG_d, "%"REG_S \
165  );
166 
167 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
168  __asm__ volatile(\
169  "xor %%"REG_a", %%"REG_a" \n\t"\
170  ".p2align 4 \n\t"\
171  "nop \n\t"\
172  "1: \n\t"\
173  "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d" \n\t"\
174  "mov (%%"REG_d"), %%"REG_S" \n\t"\
175  "pxor %%mm4, %%mm4 \n\t"\
176  "pxor %%mm5, %%mm5 \n\t"\
177  "pxor %%mm6, %%mm6 \n\t"\
178  "pxor %%mm7, %%mm7 \n\t"\
179  ".p2align 4 \n\t"\
180  "2: \n\t"\
181  "movq (%%"REG_S", %%"REG_a"), %%mm0 \n\t" /* UsrcData */\
182  "add %6, %%"REG_S" \n\t" \
183  "movq (%%"REG_S", %%"REG_a"), %%mm2 \n\t" /* VsrcData */\
184  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
185  "movq (%%"REG_S", %%"REG_a"), %%mm1 \n\t" /* UsrcData */\
186  "movq %%mm0, %%mm3 \n\t"\
187  "punpcklwd %%mm1, %%mm0 \n\t"\
188  "punpckhwd %%mm1, %%mm3 \n\t"\
189  "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1 \n\t" /* filterCoeff */\
190  "pmaddwd %%mm1, %%mm0 \n\t"\
191  "pmaddwd %%mm1, %%mm3 \n\t"\
192  "paddd %%mm0, %%mm4 \n\t"\
193  "paddd %%mm3, %%mm5 \n\t"\
194  "add %6, %%"REG_S" \n\t" \
195  "movq (%%"REG_S", %%"REG_a"), %%mm3 \n\t" /* VsrcData */\
196  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
197  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
198  "test %%"REG_S", %%"REG_S" \n\t"\
199  "movq %%mm2, %%mm0 \n\t"\
200  "punpcklwd %%mm3, %%mm2 \n\t"\
201  "punpckhwd %%mm3, %%mm0 \n\t"\
202  "pmaddwd %%mm1, %%mm2 \n\t"\
203  "pmaddwd %%mm1, %%mm0 \n\t"\
204  "paddd %%mm2, %%mm6 \n\t"\
205  "paddd %%mm0, %%mm7 \n\t"\
206  " jnz 2b \n\t"\
207  "psrad $16, %%mm4 \n\t"\
208  "psrad $16, %%mm5 \n\t"\
209  "psrad $16, %%mm6 \n\t"\
210  "psrad $16, %%mm7 \n\t"\
211  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
212  "packssdw %%mm5, %%mm4 \n\t"\
213  "packssdw %%mm7, %%mm6 \n\t"\
214  "paddw %%mm0, %%mm4 \n\t"\
215  "paddw %%mm0, %%mm6 \n\t"\
216  "movq %%mm4, "U_TEMP"(%0) \n\t"\
217  "movq %%mm6, "V_TEMP"(%0) \n\t"\
218 
219 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
220  "lea "offset"(%0), %%"REG_d" \n\t"\
221  "mov (%%"REG_d"), %%"REG_S" \n\t"\
222  "pxor %%mm1, %%mm1 \n\t"\
223  "pxor %%mm5, %%mm5 \n\t"\
224  "pxor %%mm7, %%mm7 \n\t"\
225  "pxor %%mm6, %%mm6 \n\t"\
226  ".p2align 4 \n\t"\
227  "2: \n\t"\
228  "movq (%%"REG_S", %%"REG_a", 2), %%mm0 \n\t" /* Y1srcData */\
229  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2 \n\t" /* Y2srcData */\
230  "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S" \n\t"\
231  "movq (%%"REG_S", %%"REG_a", 2), %%mm4 \n\t" /* Y1srcData */\
232  "movq %%mm0, %%mm3 \n\t"\
233  "punpcklwd %%mm4, %%mm0 \n\t"\
234  "punpckhwd %%mm4, %%mm3 \n\t"\
235  "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4 \n\t" /* filterCoeff */\
236  "pmaddwd %%mm4, %%mm0 \n\t"\
237  "pmaddwd %%mm4, %%mm3 \n\t"\
238  "paddd %%mm0, %%mm1 \n\t"\
239  "paddd %%mm3, %%mm5 \n\t"\
240  "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3 \n\t" /* Y2srcData */\
241  "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S" \n\t"\
242  "add $"STR(APCK_SIZE)", %%"REG_d" \n\t"\
243  "test %%"REG_S", %%"REG_S" \n\t"\
244  "movq %%mm2, %%mm0 \n\t"\
245  "punpcklwd %%mm3, %%mm2 \n\t"\
246  "punpckhwd %%mm3, %%mm0 \n\t"\
247  "pmaddwd %%mm4, %%mm2 \n\t"\
248  "pmaddwd %%mm4, %%mm0 \n\t"\
249  "paddd %%mm2, %%mm7 \n\t"\
250  "paddd %%mm0, %%mm6 \n\t"\
251  " jnz 2b \n\t"\
252  "psrad $16, %%mm1 \n\t"\
253  "psrad $16, %%mm5 \n\t"\
254  "psrad $16, %%mm7 \n\t"\
255  "psrad $16, %%mm6 \n\t"\
256  "movq "VROUNDER_OFFSET"(%0), %%mm0 \n\t"\
257  "packssdw %%mm5, %%mm1 \n\t"\
258  "packssdw %%mm6, %%mm7 \n\t"\
259  "paddw %%mm0, %%mm1 \n\t"\
260  "paddw %%mm0, %%mm7 \n\t"\
261  "movq "U_TEMP"(%0), %%mm3 \n\t"\
262  "movq "V_TEMP"(%0), %%mm4 \n\t"\
263 
264 #define YSCALEYUV2PACKEDX_ACCURATE \
265  YSCALEYUV2PACKEDX_ACCURATE_UV \
266  YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
267 
268 #define YSCALEYUV2RGBX \
269  "psubw "U_OFFSET"(%0), %%mm3 \n\t" /* (U-128)8*/\
270  "psubw "V_OFFSET"(%0), %%mm4 \n\t" /* (V-128)8*/\
271  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
272  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
273  "pmulhw "UG_COEFF"(%0), %%mm3 \n\t"\
274  "pmulhw "VG_COEFF"(%0), %%mm4 \n\t"\
275  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
276  "pmulhw "UB_COEFF"(%0), %%mm2 \n\t"\
277  "pmulhw "VR_COEFF"(%0), %%mm5 \n\t"\
278  "psubw "Y_OFFSET"(%0), %%mm1 \n\t" /* 8(Y-16)*/\
279  "psubw "Y_OFFSET"(%0), %%mm7 \n\t" /* 8(Y-16)*/\
280  "pmulhw "Y_COEFF"(%0), %%mm1 \n\t"\
281  "pmulhw "Y_COEFF"(%0), %%mm7 \n\t"\
282  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
283  "paddw %%mm3, %%mm4 \n\t"\
284  "movq %%mm2, %%mm0 \n\t"\
285  "movq %%mm5, %%mm6 \n\t"\
286  "movq %%mm4, %%mm3 \n\t"\
287  "punpcklwd %%mm2, %%mm2 \n\t"\
288  "punpcklwd %%mm5, %%mm5 \n\t"\
289  "punpcklwd %%mm4, %%mm4 \n\t"\
290  "paddw %%mm1, %%mm2 \n\t"\
291  "paddw %%mm1, %%mm5 \n\t"\
292  "paddw %%mm1, %%mm4 \n\t"\
293  "punpckhwd %%mm0, %%mm0 \n\t"\
294  "punpckhwd %%mm6, %%mm6 \n\t"\
295  "punpckhwd %%mm3, %%mm3 \n\t"\
296  "paddw %%mm7, %%mm0 \n\t"\
297  "paddw %%mm7, %%mm6 \n\t"\
298  "paddw %%mm7, %%mm3 \n\t"\
299  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
300  "packuswb %%mm0, %%mm2 \n\t"\
301  "packuswb %%mm6, %%mm5 \n\t"\
302  "packuswb %%mm3, %%mm4 \n\t"\
303 
304 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
305  "movq "#b", "#q2" \n\t" /* B */\
306  "movq "#r", "#t" \n\t" /* R */\
307  "punpcklbw "#g", "#b" \n\t" /* GBGBGBGB 0 */\
308  "punpcklbw "#a", "#r" \n\t" /* ARARARAR 0 */\
309  "punpckhbw "#g", "#q2" \n\t" /* GBGBGBGB 2 */\
310  "punpckhbw "#a", "#t" \n\t" /* ARARARAR 2 */\
311  "movq "#b", "#q0" \n\t" /* GBGBGBGB 0 */\
312  "movq "#q2", "#q3" \n\t" /* GBGBGBGB 2 */\
313  "punpcklwd "#r", "#q0" \n\t" /* ARGBARGB 0 */\
314  "punpckhwd "#r", "#b" \n\t" /* ARGBARGB 1 */\
315  "punpcklwd "#t", "#q2" \n\t" /* ARGBARGB 2 */\
316  "punpckhwd "#t", "#q3" \n\t" /* ARGBARGB 3 */\
317 \
318  MOVNTQ( q0, (dst, index, 4))\
319  MOVNTQ( b, 8(dst, index, 4))\
320  MOVNTQ( q2, 16(dst, index, 4))\
321  MOVNTQ( q3, 24(dst, index, 4))\
322 \
323  "add $8, "#index" \n\t"\
324  "cmp "#dstw", "#index" \n\t"\
325  " jb 1b \n\t"
326 #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)
327 
328 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
329  const int16_t **lumSrc, int lumFilterSize,
330  const int16_t *chrFilter, const int16_t **chrUSrc,
331  const int16_t **chrVSrc,
332  int chrFilterSize, const int16_t **alpSrc,
333  uint8_t *dest, int dstW, int dstY)
334 {
335  x86_reg dummy=0;
336  x86_reg dstW_reg = dstW;
337  x86_reg uv_off = c->uv_offx2;
338 
339  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
342  "movq %%mm2, "U_TEMP"(%0) \n\t"
343  "movq %%mm4, "V_TEMP"(%0) \n\t"
344  "movq %%mm5, "Y_TEMP"(%0) \n\t"
346  "movq "Y_TEMP"(%0), %%mm5 \n\t"
347  "psraw $3, %%mm1 \n\t"
348  "psraw $3, %%mm7 \n\t"
349  "packuswb %%mm7, %%mm1 \n\t"
350  WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
352  } else {
355  "pcmpeqd %%mm7, %%mm7 \n\t"
356  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
358  }
359 }
360 
361 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
362  const int16_t **lumSrc, int lumFilterSize,
363  const int16_t *chrFilter, const int16_t **chrUSrc,
364  const int16_t **chrVSrc,
365  int chrFilterSize, const int16_t **alpSrc,
366  uint8_t *dest, int dstW, int dstY)
367 {
368  x86_reg dummy=0;
369  x86_reg dstW_reg = dstW;
370  x86_reg uv_off = c->uv_offx2;
371 
372  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
375  YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
376  "psraw $3, %%mm1 \n\t"
377  "psraw $3, %%mm7 \n\t"
378  "packuswb %%mm7, %%mm1 \n\t"
379  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
381  } else {
384  "pcmpeqd %%mm7, %%mm7 \n\t"
385  WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
387  }
388 }
389 
390 #define REAL_WRITERGB16(dst, dstw, index) \
391  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
392  "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\
393  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
394  "psrlq $3, %%mm2 \n\t"\
395 \
396  "movq %%mm2, %%mm1 \n\t"\
397  "movq %%mm4, %%mm3 \n\t"\
398 \
399  "punpcklbw %%mm7, %%mm3 \n\t"\
400  "punpcklbw %%mm5, %%mm2 \n\t"\
401  "punpckhbw %%mm7, %%mm4 \n\t"\
402  "punpckhbw %%mm5, %%mm1 \n\t"\
403 \
404  "psllq $3, %%mm3 \n\t"\
405  "psllq $3, %%mm4 \n\t"\
406 \
407  "por %%mm3, %%mm2 \n\t"\
408  "por %%mm4, %%mm1 \n\t"\
409 \
410  MOVNTQ(%%mm2, (dst, index, 2))\
411  MOVNTQ(%%mm1, 8(dst, index, 2))\
412 \
413  "add $8, "#index" \n\t"\
414  "cmp "#dstw", "#index" \n\t"\
415  " jb 1b \n\t"
416 #define WRITERGB16(dst, dstw, index) REAL_WRITERGB16(dst, dstw, index)
417 
418 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
419  const int16_t **lumSrc, int lumFilterSize,
420  const int16_t *chrFilter, const int16_t **chrUSrc,
421  const int16_t **chrVSrc,
422  int chrFilterSize, const int16_t **alpSrc,
423  uint8_t *dest, int dstW, int dstY)
424 {
425  x86_reg dummy=0;
426  x86_reg dstW_reg = dstW;
427  x86_reg uv_off = c->uv_offx2;
428 
431  "pxor %%mm7, %%mm7 \n\t"
432  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
433 #ifdef DITHER1XBPP
434  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
435  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
436  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
437 #endif
438  WRITERGB16(%4, %5, %%REGa)
440 }
441 
442 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
443  const int16_t **lumSrc, int lumFilterSize,
444  const int16_t *chrFilter, const int16_t **chrUSrc,
445  const int16_t **chrVSrc,
446  int chrFilterSize, const int16_t **alpSrc,
447  uint8_t *dest, int dstW, int dstY)
448 {
449  x86_reg dummy=0;
450  x86_reg dstW_reg = dstW;
451  x86_reg uv_off = c->uv_offx2;
452 
455  "pxor %%mm7, %%mm7 \n\t"
456  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
457 #ifdef DITHER1XBPP
458  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
459  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
460  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
461 #endif
462  WRITERGB16(%4, %5, %%REGa)
464 }
465 
466 #define REAL_WRITERGB15(dst, dstw, index) \
467  "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\
468  "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\
469  "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\
470  "psrlq $3, %%mm2 \n\t"\
471  "psrlq $1, %%mm5 \n\t"\
472 \
473  "movq %%mm2, %%mm1 \n\t"\
474  "movq %%mm4, %%mm3 \n\t"\
475 \
476  "punpcklbw %%mm7, %%mm3 \n\t"\
477  "punpcklbw %%mm5, %%mm2 \n\t"\
478  "punpckhbw %%mm7, %%mm4 \n\t"\
479  "punpckhbw %%mm5, %%mm1 \n\t"\
480 \
481  "psllq $2, %%mm3 \n\t"\
482  "psllq $2, %%mm4 \n\t"\
483 \
484  "por %%mm3, %%mm2 \n\t"\
485  "por %%mm4, %%mm1 \n\t"\
486 \
487  MOVNTQ(%%mm2, (dst, index, 2))\
488  MOVNTQ(%%mm1, 8(dst, index, 2))\
489 \
490  "add $8, "#index" \n\t"\
491  "cmp "#dstw", "#index" \n\t"\
492  " jb 1b \n\t"
493 #define WRITERGB15(dst, dstw, index) REAL_WRITERGB15(dst, dstw, index)
494 
495 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
496  const int16_t **lumSrc, int lumFilterSize,
497  const int16_t *chrFilter, const int16_t **chrUSrc,
498  const int16_t **chrVSrc,
499  int chrFilterSize, const int16_t **alpSrc,
500  uint8_t *dest, int dstW, int dstY)
501 {
502  x86_reg dummy=0;
503  x86_reg dstW_reg = dstW;
504  x86_reg uv_off = c->uv_offx2;
505 
508  "pxor %%mm7, %%mm7 \n\t"
509  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
510 #ifdef DITHER1XBPP
511  "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
512  "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
513  "paddusb "RED_DITHER"(%0), %%mm5\n\t"
514 #endif
515  WRITERGB15(%4, %5, %%REGa)
517 }
518 
519 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
520  const int16_t **lumSrc, int lumFilterSize,
521  const int16_t *chrFilter, const int16_t **chrUSrc,
522  const int16_t **chrVSrc,
523  int chrFilterSize, const int16_t **alpSrc,
524  uint8_t *dest, int dstW, int dstY)
525 {
526  x86_reg dummy=0;
527  x86_reg dstW_reg = dstW;
528  x86_reg uv_off = c->uv_offx2;
529 
532  "pxor %%mm7, %%mm7 \n\t"
533  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
534 #ifdef DITHER1XBPP
535  "paddusb "BLUE_DITHER"(%0), %%mm2 \n\t"
536  "paddusb "GREEN_DITHER"(%0), %%mm4 \n\t"
537  "paddusb "RED_DITHER"(%0), %%mm5 \n\t"
538 #endif
539  WRITERGB15(%4, %5, %%REGa)
541 }
542 
543 #define WRITEBGR24MMX(dst, dstw, index) \
544  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
545  "movq %%mm2, %%mm1 \n\t" /* B */\
546  "movq %%mm5, %%mm6 \n\t" /* R */\
547  "punpcklbw %%mm4, %%mm2 \n\t" /* GBGBGBGB 0 */\
548  "punpcklbw %%mm7, %%mm5 \n\t" /* 0R0R0R0R 0 */\
549  "punpckhbw %%mm4, %%mm1 \n\t" /* GBGBGBGB 2 */\
550  "punpckhbw %%mm7, %%mm6 \n\t" /* 0R0R0R0R 2 */\
551  "movq %%mm2, %%mm0 \n\t" /* GBGBGBGB 0 */\
552  "movq %%mm1, %%mm3 \n\t" /* GBGBGBGB 2 */\
553  "punpcklwd %%mm5, %%mm0 \n\t" /* 0RGB0RGB 0 */\
554  "punpckhwd %%mm5, %%mm2 \n\t" /* 0RGB0RGB 1 */\
555  "punpcklwd %%mm6, %%mm1 \n\t" /* 0RGB0RGB 2 */\
556  "punpckhwd %%mm6, %%mm3 \n\t" /* 0RGB0RGB 3 */\
557 \
558  "movq %%mm0, %%mm4 \n\t" /* 0RGB0RGB 0 */\
559  "movq %%mm2, %%mm6 \n\t" /* 0RGB0RGB 1 */\
560  "movq %%mm1, %%mm5 \n\t" /* 0RGB0RGB 2 */\
561  "movq %%mm3, %%mm7 \n\t" /* 0RGB0RGB 3 */\
562 \
563  "psllq $40, %%mm0 \n\t" /* RGB00000 0 */\
564  "psllq $40, %%mm2 \n\t" /* RGB00000 1 */\
565  "psllq $40, %%mm1 \n\t" /* RGB00000 2 */\
566  "psllq $40, %%mm3 \n\t" /* RGB00000 3 */\
567 \
568  "punpckhdq %%mm4, %%mm0 \n\t" /* 0RGBRGB0 0 */\
569  "punpckhdq %%mm6, %%mm2 \n\t" /* 0RGBRGB0 1 */\
570  "punpckhdq %%mm5, %%mm1 \n\t" /* 0RGBRGB0 2 */\
571  "punpckhdq %%mm7, %%mm3 \n\t" /* 0RGBRGB0 3 */\
572 \
573  "psrlq $8, %%mm0 \n\t" /* 00RGBRGB 0 */\
574  "movq %%mm2, %%mm6 \n\t" /* 0RGBRGB0 1 */\
575  "psllq $40, %%mm2 \n\t" /* GB000000 1 */\
576  "por %%mm2, %%mm0 \n\t" /* GBRGBRGB 0 */\
577  MOVNTQ(%%mm0, (dst))\
578 \
579  "psrlq $24, %%mm6 \n\t" /* 0000RGBR 1 */\
580  "movq %%mm1, %%mm5 \n\t" /* 0RGBRGB0 2 */\
581  "psllq $24, %%mm1 \n\t" /* BRGB0000 2 */\
582  "por %%mm1, %%mm6 \n\t" /* BRGBRGBR 1 */\
583  MOVNTQ(%%mm6, 8(dst))\
584 \
585  "psrlq $40, %%mm5 \n\t" /* 000000RG 2 */\
586  "psllq $8, %%mm3 \n\t" /* RGBRGB00 3 */\
587  "por %%mm3, %%mm5 \n\t" /* RGBRGBRG 2 */\
588  MOVNTQ(%%mm5, 16(dst))\
589 \
590  "add $24, "#dst" \n\t"\
591 \
592  "add $8, "#index" \n\t"\
593  "cmp "#dstw", "#index" \n\t"\
594  " jb 1b \n\t"
595 
596 #define WRITEBGR24MMXEXT(dst, dstw, index) \
597  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
598  "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
599  "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
600  "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */\
601  "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */\
602  "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */\
603 \
604  "pand %%mm0, %%mm1 \n\t" /* B2 B1 B0 */\
605  "pand %%mm0, %%mm3 \n\t" /* G2 G1 G0 */\
606  "pand %%mm7, %%mm6 \n\t" /* R1 R0 */\
607 \
608  "psllq $8, %%mm3 \n\t" /* G2 G1 G0 */\
609  "por %%mm1, %%mm6 \n\t"\
610  "por %%mm3, %%mm6 \n\t"\
611  MOVNTQ(%%mm6, (dst))\
612 \
613  "psrlq $8, %%mm4 \n\t" /* 00 G7 G6 G5 G4 G3 G2 G1 */\
614  "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4 B3 B2 B3 B2 */\
615  "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */\
616  "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */\
617 \
618  "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5 B4 B3 */\
619  "pand %%mm7, %%mm3 \n\t" /* G4 G3 */\
620  "pand %%mm0, %%mm6 \n\t" /* R4 R3 R2 */\
621 \
622  "por %%mm1, %%mm3 \n\t" /* B5 G4 B4 G3 B3 */\
623  "por %%mm3, %%mm6 \n\t"\
624  MOVNTQ(%%mm6, 8(dst))\
625 \
626  "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6 B7 B6 B6 B7 */\
627  "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7 G6 G5 G6 G5 */\
628  "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6 R5 R4 R5 R4 */\
629 \
630  "pand %%mm7, %%mm1 \n\t" /* B7 B6 */\
631  "pand %%mm0, %%mm3 \n\t" /* G7 G6 G5 */\
632  "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7 R6 R5 */\
633 \
634  "por %%mm1, %%mm3 \n\t"\
635  "por %%mm3, %%mm6 \n\t"\
636  MOVNTQ(%%mm6, 16(dst))\
637 \
638  "add $24, "#dst" \n\t"\
639 \
640  "add $8, "#index" \n\t"\
641  "cmp "#dstw", "#index" \n\t"\
642  " jb 1b \n\t"
643 
644 #if COMPILE_TEMPLATE_MMXEXT
645 #undef WRITEBGR24
646 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMXEXT(dst, dstw, index)
647 #else
648 #undef WRITEBGR24
649 #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index)
650 #endif
651 
652 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
653  const int16_t **lumSrc, int lumFilterSize,
654  const int16_t *chrFilter, const int16_t **chrUSrc,
655  const int16_t **chrVSrc,
656  int chrFilterSize, const int16_t **alpSrc,
657  uint8_t *dest, int dstW, int dstY)
658 {
659  x86_reg dummy=0;
660  x86_reg dstW_reg = dstW;
661  x86_reg uv_off = c->uv_offx2;
662 
665  "pxor %%mm7, %%mm7 \n\t"
666  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
667  "add %4, %%"REG_c" \n\t"
668  WRITEBGR24(%%REGc, %5, %%REGa)
669  :: "r" (&c->redDither),
670  "m" (dummy), "m" (dummy), "m" (dummy),
671  "r" (dest), "m" (dstW_reg), "m"(uv_off)
672  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
673  );
674 }
675 
676 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
677  const int16_t **lumSrc, int lumFilterSize,
678  const int16_t *chrFilter, const int16_t **chrUSrc,
679  const int16_t **chrVSrc,
680  int chrFilterSize, const int16_t **alpSrc,
681  uint8_t *dest, int dstW, int dstY)
682 {
683  x86_reg dummy=0;
684  x86_reg dstW_reg = dstW;
685  x86_reg uv_off = c->uv_offx2;
686 
689  "pxor %%mm7, %%mm7 \n\t"
690  "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c" \n\t" //FIXME optimize
691  "add %4, %%"REG_c" \n\t"
692  WRITEBGR24(%%REGc, %5, %%REGa)
693  :: "r" (&c->redDither),
694  "m" (dummy), "m" (dummy), "m" (dummy),
695  "r" (dest), "m" (dstW_reg), "m"(uv_off)
696  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
697  );
698 }
699 
700 #define REAL_WRITEYUY2(dst, dstw, index) \
701  "packuswb %%mm3, %%mm3 \n\t"\
702  "packuswb %%mm4, %%mm4 \n\t"\
703  "packuswb %%mm7, %%mm1 \n\t"\
704  "punpcklbw %%mm4, %%mm3 \n\t"\
705  "movq %%mm1, %%mm7 \n\t"\
706  "punpcklbw %%mm3, %%mm1 \n\t"\
707  "punpckhbw %%mm3, %%mm7 \n\t"\
708 \
709  MOVNTQ(%%mm1, (dst, index, 2))\
710  MOVNTQ(%%mm7, 8(dst, index, 2))\
711 \
712  "add $8, "#index" \n\t"\
713  "cmp "#dstw", "#index" \n\t"\
714  " jb 1b \n\t"
715 #define WRITEYUY2(dst, dstw, index) REAL_WRITEYUY2(dst, dstw, index)
716 
717 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
718  const int16_t **lumSrc, int lumFilterSize,
719  const int16_t *chrFilter, const int16_t **chrUSrc,
720  const int16_t **chrVSrc,
721  int chrFilterSize, const int16_t **alpSrc,
722  uint8_t *dest, int dstW, int dstY)
723 {
724  x86_reg dummy=0;
725  x86_reg dstW_reg = dstW;
726  x86_reg uv_off = c->uv_offx2;
727 
729  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
730  "psraw $3, %%mm3 \n\t"
731  "psraw $3, %%mm4 \n\t"
732  "psraw $3, %%mm1 \n\t"
733  "psraw $3, %%mm7 \n\t"
734  WRITEYUY2(%4, %5, %%REGa)
736 }
737 
738 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
739  const int16_t **lumSrc, int lumFilterSize,
740  const int16_t *chrFilter, const int16_t **chrUSrc,
741  const int16_t **chrVSrc,
742  int chrFilterSize, const int16_t **alpSrc,
743  uint8_t *dest, int dstW, int dstY)
744 {
745  x86_reg dummy=0;
746  x86_reg dstW_reg = dstW;
747  x86_reg uv_off = c->uv_offx2;
748 
750  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
751  "psraw $3, %%mm3 \n\t"
752  "psraw $3, %%mm4 \n\t"
753  "psraw $3, %%mm1 \n\t"
754  "psraw $3, %%mm7 \n\t"
755  WRITEYUY2(%4, %5, %%REGa)
757 }
758 
759 #define REAL_YSCALEYUV2RGB_UV(index, c) \
760  "xor "#index", "#index" \n\t"\
761  ".p2align 4 \n\t"\
762  "1: \n\t"\
763  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
764  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
765  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
766  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
767  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
768  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
769  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
770  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
771  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
772  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
773  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
774  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
775  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
776  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
777  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
778  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
779  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
780  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
781  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
782  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
783  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
784  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
785 
786 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
787  "movq ("#b1", "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
788  "movq ("#b2", "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
789  "movq 8("#b1", "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
790  "movq 8("#b2", "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
791  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
792  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
793  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
794  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
795  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
796  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
797  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
798  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
799 
800 #define REAL_YSCALEYUV2RGB_COEFF(c) \
801  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
802  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
803  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
804  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
805  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
806  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
807  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
808  "paddw %%mm3, %%mm4 \n\t"\
809  "movq %%mm2, %%mm0 \n\t"\
810  "movq %%mm5, %%mm6 \n\t"\
811  "movq %%mm4, %%mm3 \n\t"\
812  "punpcklwd %%mm2, %%mm2 \n\t"\
813  "punpcklwd %%mm5, %%mm5 \n\t"\
814  "punpcklwd %%mm4, %%mm4 \n\t"\
815  "paddw %%mm1, %%mm2 \n\t"\
816  "paddw %%mm1, %%mm5 \n\t"\
817  "paddw %%mm1, %%mm4 \n\t"\
818  "punpckhwd %%mm0, %%mm0 \n\t"\
819  "punpckhwd %%mm6, %%mm6 \n\t"\
820  "punpckhwd %%mm3, %%mm3 \n\t"\
821  "paddw %%mm7, %%mm0 \n\t"\
822  "paddw %%mm7, %%mm6 \n\t"\
823  "paddw %%mm7, %%mm3 \n\t"\
824  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
825  "packuswb %%mm0, %%mm2 \n\t"\
826  "packuswb %%mm6, %%mm5 \n\t"\
827  "packuswb %%mm3, %%mm4 \n\t"\
828 
829 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
830 
831 #define YSCALEYUV2RGB(index, c) \
832  REAL_YSCALEYUV2RGB_UV(index, c) \
833  REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
834  REAL_YSCALEYUV2RGB_COEFF(c)
835 
836 /**
837  * vertical bilinear scale YV12 to RGB
838  */
839 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
840  const int16_t *ubuf[2], const int16_t *vbuf[2],
841  const int16_t *abuf[2], uint8_t *dest,
842  int dstW, int yalpha, int uvalpha, int y)
843 {
844  const int16_t *buf0 = buf[0], *buf1 = buf[1],
845  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
846 
847  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
848  const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
849 #if ARCH_X86_64
850  __asm__ volatile(
851  YSCALEYUV2RGB(%%r8, %5)
852  YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
853  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
854  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
855  "packuswb %%mm7, %%mm1 \n\t"
856  WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
857  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
858  "a" (&c->redDither),
859  "r" (abuf0), "r" (abuf1)
860  : "%r8"
861  );
862 #else
863  c->u_temp=(intptr_t)abuf0;
864  c->v_temp=(intptr_t)abuf1;
865  __asm__ volatile(
866  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
867  "mov %4, %%"REG_b" \n\t"
868  "push %%"REG_BP" \n\t"
869  YSCALEYUV2RGB(%%REGBP, %5)
870  "push %0 \n\t"
871  "push %1 \n\t"
872  "mov "U_TEMP"(%5), %0 \n\t"
873  "mov "V_TEMP"(%5), %1 \n\t"
874  YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
875  "psraw $3, %%mm1 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
876  "psraw $3, %%mm7 \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
877  "packuswb %%mm7, %%mm1 \n\t"
878  "pop %1 \n\t"
879  "pop %0 \n\t"
880  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
881  "pop %%"REG_BP" \n\t"
882  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
883  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
884  "a" (&c->redDither)
885  );
886 #endif
887  } else {
888  __asm__ volatile(
889  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
890  "mov %4, %%"REG_b" \n\t"
891  "push %%"REG_BP" \n\t"
892  YSCALEYUV2RGB(%%REGBP, %5)
893  "pcmpeqd %%mm7, %%mm7 \n\t"
894  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
895  "pop %%"REG_BP" \n\t"
896  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
897  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
898  "a" (&c->redDither)
899  );
900  }
901 }
902 
903 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
904  const int16_t *ubuf[2], const int16_t *vbuf[2],
905  const int16_t *abuf[2], uint8_t *dest,
906  int dstW, int yalpha, int uvalpha, int y)
907 {
908  const int16_t *buf0 = buf[0], *buf1 = buf[1],
909  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
910 
911  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
912  __asm__ volatile(
913  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
914  "mov %4, %%"REG_b" \n\t"
915  "push %%"REG_BP" \n\t"
916  YSCALEYUV2RGB(%%REGBP, %5)
917  "pxor %%mm7, %%mm7 \n\t"
918  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
919  "pop %%"REG_BP" \n\t"
920  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
921  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
922  "a" (&c->redDither)
923  );
924 }
925 
926 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
927  const int16_t *ubuf[2], const int16_t *vbuf[2],
928  const int16_t *abuf[2], uint8_t *dest,
929  int dstW, int yalpha, int uvalpha, int y)
930 {
931  const int16_t *buf0 = buf[0], *buf1 = buf[1],
932  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
933 
934  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
935  __asm__ volatile(
936  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
937  "mov %4, %%"REG_b" \n\t"
938  "push %%"REG_BP" \n\t"
939  YSCALEYUV2RGB(%%REGBP, %5)
940  "pxor %%mm7, %%mm7 \n\t"
941  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
942 #ifdef DITHER1XBPP
943  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
944  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
945  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
946 #endif
947  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
948  "pop %%"REG_BP" \n\t"
949  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
950  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
951  "a" (&c->redDither)
952  );
953 }
954 
955 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
956  const int16_t *ubuf[2], const int16_t *vbuf[2],
957  const int16_t *abuf[2], uint8_t *dest,
958  int dstW, int yalpha, int uvalpha, int y)
959 {
960  const int16_t *buf0 = buf[0], *buf1 = buf[1],
961  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
962 
963  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
964  __asm__ volatile(
965  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
966  "mov %4, %%"REG_b" \n\t"
967  "push %%"REG_BP" \n\t"
968  YSCALEYUV2RGB(%%REGBP, %5)
969  "pxor %%mm7, %%mm7 \n\t"
970  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
971 #ifdef DITHER1XBPP
972  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
973  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
974  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
975 #endif
976  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
977  "pop %%"REG_BP" \n\t"
978  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
979  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
980  "a" (&c->redDither)
981  );
982 }
983 
984 #define REAL_YSCALEYUV2PACKED(index, c) \
985  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
986  "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1 \n\t"\
987  "psraw $3, %%mm0 \n\t"\
988  "psraw $3, %%mm1 \n\t"\
989  "movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
990  "movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
991  "xor "#index", "#index" \n\t"\
992  ".p2align 4 \n\t"\
993  "1: \n\t"\
994  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
995  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
996  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
997  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
998  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
999  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1000  "psubw %%mm3, %%mm2 \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1001  "psubw %%mm4, %%mm5 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1002  "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t"\
1003  "pmulhw %%mm0, %%mm2 \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1004  "pmulhw %%mm0, %%mm5 \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1005  "psraw $7, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1006  "psraw $7, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1007  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1008  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1009  "movq (%0, "#index", 2), %%mm0 \n\t" /*buf0[eax]*/\
1010  "movq (%1, "#index", 2), %%mm1 \n\t" /*buf1[eax]*/\
1011  "movq 8(%0, "#index", 2), %%mm6 \n\t" /*buf0[eax]*/\
1012  "movq 8(%1, "#index", 2), %%mm7 \n\t" /*buf1[eax]*/\
1013  "psubw %%mm1, %%mm0 \n\t" /* buf0[eax] - buf1[eax]*/\
1014  "psubw %%mm7, %%mm6 \n\t" /* buf0[eax] - buf1[eax]*/\
1015  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1016  "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6 \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1017  "psraw $7, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1018  "psraw $7, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1019  "paddw %%mm0, %%mm1 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1020  "paddw %%mm6, %%mm7 \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1021 
1022 #define YSCALEYUV2PACKED(index, c) REAL_YSCALEYUV2PACKED(index, c)
1023 
1024 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
1025  const int16_t *ubuf[2], const int16_t *vbuf[2],
1026  const int16_t *abuf[2], uint8_t *dest,
1027  int dstW, int yalpha, int uvalpha, int y)
1028 {
1029  const int16_t *buf0 = buf[0], *buf1 = buf[1],
1030  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1031 
1032  //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1033  __asm__ volatile(
1034  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1035  "mov %4, %%"REG_b" \n\t"
1036  "push %%"REG_BP" \n\t"
1037  YSCALEYUV2PACKED(%%REGBP, %5)
1038  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1039  "pop %%"REG_BP" \n\t"
1040  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1041  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1042  "a" (&c->redDither)
1043  );
1044 }
1045 
1046 #define REAL_YSCALEYUV2RGB1(index, c) \
1047  "xor "#index", "#index" \n\t"\
1048  ".p2align 4 \n\t"\
1049  "1: \n\t"\
1050  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1051  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1052  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1053  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1054  "psraw $4, %%mm3 \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1055  "psraw $4, %%mm4 \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1056  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1057  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1058  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1059  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1060  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1061  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1062  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1063  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1064  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1065  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1066  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1067  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1068  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1069  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1070  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1071  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1072  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1073  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1074  "paddw %%mm3, %%mm4 \n\t"\
1075  "movq %%mm2, %%mm0 \n\t"\
1076  "movq %%mm5, %%mm6 \n\t"\
1077  "movq %%mm4, %%mm3 \n\t"\
1078  "punpcklwd %%mm2, %%mm2 \n\t"\
1079  "punpcklwd %%mm5, %%mm5 \n\t"\
1080  "punpcklwd %%mm4, %%mm4 \n\t"\
1081  "paddw %%mm1, %%mm2 \n\t"\
1082  "paddw %%mm1, %%mm5 \n\t"\
1083  "paddw %%mm1, %%mm4 \n\t"\
1084  "punpckhwd %%mm0, %%mm0 \n\t"\
1085  "punpckhwd %%mm6, %%mm6 \n\t"\
1086  "punpckhwd %%mm3, %%mm3 \n\t"\
1087  "paddw %%mm7, %%mm0 \n\t"\
1088  "paddw %%mm7, %%mm6 \n\t"\
1089  "paddw %%mm7, %%mm3 \n\t"\
1090  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1091  "packuswb %%mm0, %%mm2 \n\t"\
1092  "packuswb %%mm6, %%mm5 \n\t"\
1093  "packuswb %%mm3, %%mm4 \n\t"\
1094 
1095 #define YSCALEYUV2RGB1(index, c) REAL_YSCALEYUV2RGB1(index, c)
1096 
1097 // do vertical chrominance interpolation
1098 #define REAL_YSCALEYUV2RGB1b(index, c) \
1099  "xor "#index", "#index" \n\t"\
1100  ".p2align 4 \n\t"\
1101  "1: \n\t"\
1102  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1103  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1104  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1105  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1106  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1107  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1108  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1109  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1110  "psrlw $5, %%mm3 \n\t" /*FIXME might overflow*/\
1111  "psrlw $5, %%mm4 \n\t" /*FIXME might overflow*/\
1112  "psubw "U_OFFSET"("#c"), %%mm3 \n\t" /* (U-128)8*/\
1113  "psubw "V_OFFSET"("#c"), %%mm4 \n\t" /* (V-128)8*/\
1114  "movq %%mm3, %%mm2 \n\t" /* (U-128)8*/\
1115  "movq %%mm4, %%mm5 \n\t" /* (V-128)8*/\
1116  "pmulhw "UG_COEFF"("#c"), %%mm3 \n\t"\
1117  "pmulhw "VG_COEFF"("#c"), %%mm4 \n\t"\
1118  /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1119  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1120  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1121  "psraw $4, %%mm1 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1122  "psraw $4, %%mm7 \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1123  "pmulhw "UB_COEFF"("#c"), %%mm2 \n\t"\
1124  "pmulhw "VR_COEFF"("#c"), %%mm5 \n\t"\
1125  "psubw "Y_OFFSET"("#c"), %%mm1 \n\t" /* 8(Y-16)*/\
1126  "psubw "Y_OFFSET"("#c"), %%mm7 \n\t" /* 8(Y-16)*/\
1127  "pmulhw "Y_COEFF"("#c"), %%mm1 \n\t"\
1128  "pmulhw "Y_COEFF"("#c"), %%mm7 \n\t"\
1129  /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1130  "paddw %%mm3, %%mm4 \n\t"\
1131  "movq %%mm2, %%mm0 \n\t"\
1132  "movq %%mm5, %%mm6 \n\t"\
1133  "movq %%mm4, %%mm3 \n\t"\
1134  "punpcklwd %%mm2, %%mm2 \n\t"\
1135  "punpcklwd %%mm5, %%mm5 \n\t"\
1136  "punpcklwd %%mm4, %%mm4 \n\t"\
1137  "paddw %%mm1, %%mm2 \n\t"\
1138  "paddw %%mm1, %%mm5 \n\t"\
1139  "paddw %%mm1, %%mm4 \n\t"\
1140  "punpckhwd %%mm0, %%mm0 \n\t"\
1141  "punpckhwd %%mm6, %%mm6 \n\t"\
1142  "punpckhwd %%mm3, %%mm3 \n\t"\
1143  "paddw %%mm7, %%mm0 \n\t"\
1144  "paddw %%mm7, %%mm6 \n\t"\
1145  "paddw %%mm7, %%mm3 \n\t"\
1146  /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1147  "packuswb %%mm0, %%mm2 \n\t"\
1148  "packuswb %%mm6, %%mm5 \n\t"\
1149  "packuswb %%mm3, %%mm4 \n\t"\
1150 
1151 #define YSCALEYUV2RGB1b(index, c) REAL_YSCALEYUV2RGB1b(index, c)
1152 
1153 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1154  "movq (%1, "#index", 2), %%mm7 \n\t" /* abuf0[index ] */\
1155  "movq 8(%1, "#index", 2), %%mm1 \n\t" /* abuf0[index+4] */\
1156  "psraw $7, %%mm7 \n\t" /* abuf0[index ] >>7 */\
1157  "psraw $7, %%mm1 \n\t" /* abuf0[index+4] >>7 */\
1158  "packuswb %%mm1, %%mm7 \n\t"
1159 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1160 
1161 /**
1162  * YV12 to RGB without scaling or interpolating
1163  */
1164 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1165  const int16_t *ubuf[2], const int16_t *vbuf[2],
1166  const int16_t *abuf0, uint8_t *dest,
1167  int dstW, int uvalpha, int y)
1168 {
1169  const int16_t *ubuf0 = ubuf[0];
1170  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1171 
1172  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1173  const int16_t *ubuf1 = ubuf[0];
1174  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1175  __asm__ volatile(
1176  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1177  "mov %4, %%"REG_b" \n\t"
1178  "push %%"REG_BP" \n\t"
1179  YSCALEYUV2RGB1(%%REGBP, %5)
1180  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1181  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1182  "pop %%"REG_BP" \n\t"
1183  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1184  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1185  "a" (&c->redDither)
1186  );
1187  } else {
1188  __asm__ volatile(
1189  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1190  "mov %4, %%"REG_b" \n\t"
1191  "push %%"REG_BP" \n\t"
1192  YSCALEYUV2RGB1(%%REGBP, %5)
1193  "pcmpeqd %%mm7, %%mm7 \n\t"
1194  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1195  "pop %%"REG_BP" \n\t"
1196  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1197  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1198  "a" (&c->redDither)
1199  );
1200  }
1201  } else {
1202  const int16_t *ubuf1 = ubuf[1];
1203  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1204  __asm__ volatile(
1205  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1206  "mov %4, %%"REG_b" \n\t"
1207  "push %%"REG_BP" \n\t"
1208  YSCALEYUV2RGB1b(%%REGBP, %5)
1209  YSCALEYUV2RGB1_ALPHA(%%REGBP)
1210  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1211  "pop %%"REG_BP" \n\t"
1212  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1213  :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1214  "a" (&c->redDither)
1215  );
1216  } else {
1217  __asm__ volatile(
1218  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1219  "mov %4, %%"REG_b" \n\t"
1220  "push %%"REG_BP" \n\t"
1221  YSCALEYUV2RGB1b(%%REGBP, %5)
1222  "pcmpeqd %%mm7, %%mm7 \n\t"
1223  WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1224  "pop %%"REG_BP" \n\t"
1225  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1226  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1227  "a" (&c->redDither)
1228  );
1229  }
1230  }
1231 }
1232 
1233 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1234  const int16_t *ubuf[2], const int16_t *vbuf[2],
1235  const int16_t *abuf0, uint8_t *dest,
1236  int dstW, int uvalpha, int y)
1237 {
1238  const int16_t *ubuf0 = ubuf[0];
1239  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1240 
1241  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1242  const int16_t *ubuf1 = ubuf[0];
1243  __asm__ volatile(
1244  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1245  "mov %4, %%"REG_b" \n\t"
1246  "push %%"REG_BP" \n\t"
1247  YSCALEYUV2RGB1(%%REGBP, %5)
1248  "pxor %%mm7, %%mm7 \n\t"
1249  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1250  "pop %%"REG_BP" \n\t"
1251  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1252  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1253  "a" (&c->redDither)
1254  );
1255  } else {
1256  const int16_t *ubuf1 = ubuf[1];
1257  __asm__ volatile(
1258  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1259  "mov %4, %%"REG_b" \n\t"
1260  "push %%"REG_BP" \n\t"
1261  YSCALEYUV2RGB1b(%%REGBP, %5)
1262  "pxor %%mm7, %%mm7 \n\t"
1263  WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1264  "pop %%"REG_BP" \n\t"
1265  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1266  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1267  "a" (&c->redDither)
1268  );
1269  }
1270 }
1271 
1272 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1273  const int16_t *ubuf[2], const int16_t *vbuf[2],
1274  const int16_t *abuf0, uint8_t *dest,
1275  int dstW, int uvalpha, int y)
1276 {
1277  const int16_t *ubuf0 = ubuf[0];
1278  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1279 
1280  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1281  const int16_t *ubuf1 = ubuf[0];
1282  __asm__ volatile(
1283  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1284  "mov %4, %%"REG_b" \n\t"
1285  "push %%"REG_BP" \n\t"
1286  YSCALEYUV2RGB1(%%REGBP, %5)
1287  "pxor %%mm7, %%mm7 \n\t"
1288  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1289 #ifdef DITHER1XBPP
1290  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1291  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1292  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1293 #endif
1294  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1295  "pop %%"REG_BP" \n\t"
1296  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1297  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1298  "a" (&c->redDither)
1299  );
1300  } else {
1301  const int16_t *ubuf1 = ubuf[1];
1302  __asm__ volatile(
1303  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1304  "mov %4, %%"REG_b" \n\t"
1305  "push %%"REG_BP" \n\t"
1306  YSCALEYUV2RGB1b(%%REGBP, %5)
1307  "pxor %%mm7, %%mm7 \n\t"
1308  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1309 #ifdef DITHER1XBPP
1310  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1311  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1312  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1313 #endif
1314  WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1315  "pop %%"REG_BP" \n\t"
1316  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1317  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1318  "a" (&c->redDither)
1319  );
1320  }
1321 }
1322 
1323 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1324  const int16_t *ubuf[2], const int16_t *vbuf[2],
1325  const int16_t *abuf0, uint8_t *dest,
1326  int dstW, int uvalpha, int y)
1327 {
1328  const int16_t *ubuf0 = ubuf[0];
1329  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1330 
1331  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1332  const int16_t *ubuf1 = ubuf[0];
1333  __asm__ volatile(
1334  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1335  "mov %4, %%"REG_b" \n\t"
1336  "push %%"REG_BP" \n\t"
1337  YSCALEYUV2RGB1(%%REGBP, %5)
1338  "pxor %%mm7, %%mm7 \n\t"
1339  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1340 #ifdef DITHER1XBPP
1341  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1342  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1343  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1344 #endif
1345  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1346  "pop %%"REG_BP" \n\t"
1347  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1348  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1349  "a" (&c->redDither)
1350  );
1351  } else {
1352  const int16_t *ubuf1 = ubuf[1];
1353  __asm__ volatile(
1354  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1355  "mov %4, %%"REG_b" \n\t"
1356  "push %%"REG_BP" \n\t"
1357  YSCALEYUV2RGB1b(%%REGBP, %5)
1358  "pxor %%mm7, %%mm7 \n\t"
1359  /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1360 #ifdef DITHER1XBPP
1361  "paddusb "BLUE_DITHER"(%5), %%mm2 \n\t"
1362  "paddusb "GREEN_DITHER"(%5), %%mm4 \n\t"
1363  "paddusb "RED_DITHER"(%5), %%mm5 \n\t"
1364 #endif
1365  WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1366  "pop %%"REG_BP" \n\t"
1367  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1368  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1369  "a" (&c->redDither)
1370  );
1371  }
1372 }
1373 
1374 #define REAL_YSCALEYUV2PACKED1(index, c) \
1375  "xor "#index", "#index" \n\t"\
1376  ".p2align 4 \n\t"\
1377  "1: \n\t"\
1378  "movq (%2, "#index"), %%mm3 \n\t" /* uvbuf0[eax]*/\
1379  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1380  "movq (%2, "#index"), %%mm4 \n\t" /* uvbuf0[eax+2048]*/\
1381  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1382  "psraw $7, %%mm3 \n\t" \
1383  "psraw $7, %%mm4 \n\t" \
1384  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1385  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1386  "psraw $7, %%mm1 \n\t" \
1387  "psraw $7, %%mm7 \n\t" \
1388 
1389 #define YSCALEYUV2PACKED1(index, c) REAL_YSCALEYUV2PACKED1(index, c)
1390 
1391 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1392  "xor "#index", "#index" \n\t"\
1393  ".p2align 4 \n\t"\
1394  "1: \n\t"\
1395  "movq (%2, "#index"), %%mm2 \n\t" /* uvbuf0[eax]*/\
1396  "movq (%3, "#index"), %%mm3 \n\t" /* uvbuf1[eax]*/\
1397  "add "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1398  "movq (%2, "#index"), %%mm5 \n\t" /* uvbuf0[eax+2048]*/\
1399  "movq (%3, "#index"), %%mm4 \n\t" /* uvbuf1[eax+2048]*/\
1400  "sub "UV_OFF_BYTE"("#c"), "#index" \n\t" \
1401  "paddw %%mm2, %%mm3 \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1402  "paddw %%mm5, %%mm4 \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1403  "psrlw $8, %%mm3 \n\t" \
1404  "psrlw $8, %%mm4 \n\t" \
1405  "movq (%0, "#index", 2), %%mm1 \n\t" /*buf0[eax]*/\
1406  "movq 8(%0, "#index", 2), %%mm7 \n\t" /*buf0[eax]*/\
1407  "psraw $7, %%mm1 \n\t" \
1408  "psraw $7, %%mm7 \n\t"
1409 #define YSCALEYUV2PACKED1b(index, c) REAL_YSCALEYUV2PACKED1b(index, c)
1410 
1411 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1412  const int16_t *ubuf[2], const int16_t *vbuf[2],
1413  const int16_t *abuf0, uint8_t *dest,
1414  int dstW, int uvalpha, int y)
1415 {
1416  const int16_t *ubuf0 = ubuf[0];
1417  const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1418 
1419  if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1420  const int16_t *ubuf1 = ubuf[0];
1421  __asm__ volatile(
1422  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1423  "mov %4, %%"REG_b" \n\t"
1424  "push %%"REG_BP" \n\t"
1425  YSCALEYUV2PACKED1(%%REGBP, %5)
1426  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1427  "pop %%"REG_BP" \n\t"
1428  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1429  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1430  "a" (&c->redDither)
1431  );
1432  } else {
1433  const int16_t *ubuf1 = ubuf[1];
1434  __asm__ volatile(
1435  "mov %%"REG_b", "ESP_OFFSET"(%5) \n\t"
1436  "mov %4, %%"REG_b" \n\t"
1437  "push %%"REG_BP" \n\t"
1438  YSCALEYUV2PACKED1b(%%REGBP, %5)
1439  WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1440  "pop %%"REG_BP" \n\t"
1441  "mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"
1442  :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1443  "a" (&c->redDither)
1444  );
1445  }
1446 }
1447 
1448 #if COMPILE_TEMPLATE_MMXEXT
1449 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
1450  int dstWidth, const uint8_t *src,
1451  int srcW, int xInc)
1452 {
1453  int32_t *filterPos = c->hLumFilterPos;
1454  int16_t *filter = c->hLumFilter;
1455  void *mmxextFilterCode = c->lumMmxextFilterCode;
1456  int i;
1457 #if defined(PIC)
1458  uint64_t ebxsave;
1459 #endif
1460 #if ARCH_X86_64
1461  uint64_t retsave;
1462 #endif
1463 
1464  __asm__ volatile(
1465 #if defined(PIC)
1466  "mov %%"REG_b", %5 \n\t"
1467 #if ARCH_X86_64
1468  "mov -8(%%rsp), %%"REG_a" \n\t"
1469  "mov %%"REG_a", %6 \n\t"
1470 #endif
1471 #else
1472 #if ARCH_X86_64
1473  "mov -8(%%rsp), %%"REG_a" \n\t"
1474  "mov %%"REG_a", %5 \n\t"
1475 #endif
1476 #endif
1477  "pxor %%mm7, %%mm7 \n\t"
1478  "mov %0, %%"REG_c" \n\t"
1479  "mov %1, %%"REG_D" \n\t"
1480  "mov %2, %%"REG_d" \n\t"
1481  "mov %3, %%"REG_b" \n\t"
1482  "xor %%"REG_a", %%"REG_a" \n\t" // i
1483  PREFETCH" (%%"REG_c") \n\t"
1484  PREFETCH" 32(%%"REG_c") \n\t"
1485  PREFETCH" 64(%%"REG_c") \n\t"
1486 
1487 #if ARCH_X86_64
1488 #define CALL_MMXEXT_FILTER_CODE \
1489  "movl (%%"REG_b"), %%esi \n\t"\
1490  "call *%4 \n\t"\
1491  "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\
1492  "add %%"REG_S", %%"REG_c" \n\t"\
1493  "add %%"REG_a", %%"REG_D" \n\t"\
1494  "xor %%"REG_a", %%"REG_a" \n\t"\
1495 
1496 #else
1497 #define CALL_MMXEXT_FILTER_CODE \
1498  "movl (%%"REG_b"), %%esi \n\t"\
1499  "call *%4 \n\t"\
1500  "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
1501  "add %%"REG_a", %%"REG_D" \n\t"\
1502  "xor %%"REG_a", %%"REG_a" \n\t"\
1503 
1504 #endif /* ARCH_X86_64 */
1505 
1506  CALL_MMXEXT_FILTER_CODE
1507  CALL_MMXEXT_FILTER_CODE
1508  CALL_MMXEXT_FILTER_CODE
1509  CALL_MMXEXT_FILTER_CODE
1510  CALL_MMXEXT_FILTER_CODE
1511  CALL_MMXEXT_FILTER_CODE
1512  CALL_MMXEXT_FILTER_CODE
1513  CALL_MMXEXT_FILTER_CODE
1514 
1515 #if defined(PIC)
1516  "mov %5, %%"REG_b" \n\t"
1517 #if ARCH_X86_64
1518  "mov %6, %%"REG_a" \n\t"
1519  "mov %%"REG_a", -8(%%rsp) \n\t"
1520 #endif
1521 #else
1522 #if ARCH_X86_64
1523  "mov %5, %%"REG_a" \n\t"
1524  "mov %%"REG_a", -8(%%rsp) \n\t"
1525 #endif
1526 #endif
1527  :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
1528  "m" (mmxextFilterCode)
1529 #if defined(PIC)
1530  ,"m" (ebxsave)
1531 #endif
1532 #if ARCH_X86_64
1533  ,"m"(retsave)
1534 #endif
1535  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1536 #if !defined(PIC)
1537  ,"%"REG_b
1538 #endif
1539  );
1540 
1541  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
1542  dst[i] = src[srcW-1]*128;
1543 }
1544 
1545 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
1546  int dstWidth, const uint8_t *src1,
1547  const uint8_t *src2, int srcW, int xInc)
1548 {
1549  int32_t *filterPos = c->hChrFilterPos;
1550  int16_t *filter = c->hChrFilter;
1551  void *mmxextFilterCode = c->chrMmxextFilterCode;
1552  int i;
1553 #if defined(PIC)
1554  DECLARE_ALIGNED(8, uint64_t, ebxsave);
1555 #endif
1556 #if ARCH_X86_64
1557  DECLARE_ALIGNED(8, uint64_t, retsave);
1558 #endif
1559 
1560  __asm__ volatile(
1561 #if defined(PIC)
1562  "mov %%"REG_b", %7 \n\t"
1563 #if ARCH_X86_64
1564  "mov -8(%%rsp), %%"REG_a" \n\t"
1565  "mov %%"REG_a", %8 \n\t"
1566 #endif
1567 #else
1568 #if ARCH_X86_64
1569  "mov -8(%%rsp), %%"REG_a" \n\t"
1570  "mov %%"REG_a", %7 \n\t"
1571 #endif
1572 #endif
1573  "pxor %%mm7, %%mm7 \n\t"
1574  "mov %0, %%"REG_c" \n\t"
1575  "mov %1, %%"REG_D" \n\t"
1576  "mov %2, %%"REG_d" \n\t"
1577  "mov %3, %%"REG_b" \n\t"
1578  "xor %%"REG_a", %%"REG_a" \n\t" // i
1579  PREFETCH" (%%"REG_c") \n\t"
1580  PREFETCH" 32(%%"REG_c") \n\t"
1581  PREFETCH" 64(%%"REG_c") \n\t"
1582 
1583  CALL_MMXEXT_FILTER_CODE
1584  CALL_MMXEXT_FILTER_CODE
1585  CALL_MMXEXT_FILTER_CODE
1586  CALL_MMXEXT_FILTER_CODE
1587  "xor %%"REG_a", %%"REG_a" \n\t" // i
1588  "mov %5, %%"REG_c" \n\t" // src
1589  "mov %6, %%"REG_D" \n\t" // buf2
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 
1599 #if defined(PIC)
1600  "mov %7, %%"REG_b" \n\t"
1601 #if ARCH_X86_64
1602  "mov %8, %%"REG_a" \n\t"
1603  "mov %%"REG_a", -8(%%rsp) \n\t"
1604 #endif
1605 #else
1606 #if ARCH_X86_64
1607  "mov %7, %%"REG_a" \n\t"
1608  "mov %%"REG_a", -8(%%rsp) \n\t"
1609 #endif
1610 #endif
1611  :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
1612  "m" (mmxextFilterCode), "m" (src2), "m"(dst2)
1613 #if defined(PIC)
1614  ,"m" (ebxsave)
1615 #endif
1616 #if ARCH_X86_64
1617  ,"m"(retsave)
1618 #endif
1619  : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1620 #if !defined(PIC)
1621  ,"%"REG_b
1622 #endif
1623  );
1624 
1625  for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
1626  dst1[i] = src1[srcW-1]*128;
1627  dst2[i] = src2[srcW-1]*128;
1628  }
1629 }
1630 #endif /* COMPILE_TEMPLATE_MMXEXT */
1631 
1632 static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
1633 {
1634  enum AVPixelFormat dstFormat = c->dstFormat;
1635 
1636  c->use_mmx_vfilter= 0;
1637  if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != AV_PIX_FMT_NV12
1638  && dstFormat != AV_PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
1639  if (c->flags & SWS_ACCURATE_RND) {
1640  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1641  switch (c->dstFormat) {
1642  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X_ar); break;
1643  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X_ar); break;
1644  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X_ar); break;
1645  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X_ar); break;
1646  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1647  default: break;
1648  }
1649  }
1650  } else {
1651  c->use_mmx_vfilter= 1;
1652  c->yuv2planeX = RENAME(yuv2yuvX );
1653  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1654  switch (c->dstFormat) {
1655  case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X); break;
1656  case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X); break;
1657  case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X); break;
1658  case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X); break;
1659  case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1660  default: break;
1661  }
1662  }
1663  }
1664  if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1665  switch (c->dstFormat) {
1666  case AV_PIX_FMT_RGB32:
1667  c->yuv2packed1 = RENAME(yuv2rgb32_1);
1668  c->yuv2packed2 = RENAME(yuv2rgb32_2);
1669  break;
1670  case AV_PIX_FMT_BGR24:
1671  c->yuv2packed1 = RENAME(yuv2bgr24_1);
1672  c->yuv2packed2 = RENAME(yuv2bgr24_2);
1673  break;
1674  case AV_PIX_FMT_RGB555:
1675  c->yuv2packed1 = RENAME(yuv2rgb555_1);
1676  c->yuv2packed2 = RENAME(yuv2rgb555_2);
1677  break;
1678  case AV_PIX_FMT_RGB565:
1679  c->yuv2packed1 = RENAME(yuv2rgb565_1);
1680  c->yuv2packed2 = RENAME(yuv2rgb565_2);
1681  break;
1682  case AV_PIX_FMT_YUYV422:
1683  c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1684  c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1685  break;
1686  default:
1687  break;
1688  }
1689  }
1690  }
1691 
1692  if (c->srcBpc == 8 && c->dstBpc <= 14) {
1693  // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1694 #if COMPILE_TEMPLATE_MMXEXT
1695  if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1696  c->hyscale_fast = RENAME(hyscale_fast);
1697  c->hcscale_fast = RENAME(hcscale_fast);
1698  } else {
1699 #endif /* COMPILE_TEMPLATE_MMXEXT */
1700  c->hyscale_fast = NULL;
1701  c->hcscale_fast = NULL;
1702 #if COMPILE_TEMPLATE_MMXEXT
1703  }
1704 #endif /* COMPILE_TEMPLATE_MMXEXT */
1705  }
1706 }