• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #include <stdint.h>
22 
23 #include "libavutil/x86/asm.h"
24 #include "libswscale/swscale_internal.h"
25 
26 #undef REAL_MOVNTQ
27 #undef MOVNTQ
28 #undef MOVNTQ2
29 #undef PREFETCH
30 
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 #define YSCALEYUV2PACKEDX_UV \
42     __asm__ volatile(\
43         "xor                %%"FF_REG_a", %%"FF_REG_a"  \n\t"\
44         ".p2align                      4                \n\t"\
45         "nop                                            \n\t"\
46         "1:                                             \n\t"\
47         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"FF_REG_d"  \n\t"\
48         "mov              (%%"FF_REG_d"), %%"FF_REG_S"  \n\t"\
49         "movq      "VROUNDER_OFFSET"(%0), %%mm3         \n\t"\
50         "movq                      %%mm3, %%mm4         \n\t"\
51         ".p2align                      4                \n\t"\
52         "2:                                             \n\t"\
53         "movq            8(%%"FF_REG_d"), %%mm0         \n\t" /* filterCoeff */\
54         "movq  (%%"FF_REG_S", %%"FF_REG_a"), %%mm2      \n\t" /* UsrcData */\
55         "add                          %6, %%"FF_REG_S"  \n\t" \
56         "movq  (%%"FF_REG_S", %%"FF_REG_a"), %%mm5      \n\t" /* VsrcData */\
57         "add                         $16, %%"FF_REG_d"  \n\t"\
58         "mov              (%%"FF_REG_d"), %%"FF_REG_S"  \n\t"\
59         "pmulhw                    %%mm0, %%mm2         \n\t"\
60         "pmulhw                    %%mm0, %%mm5         \n\t"\
61         "paddw                     %%mm2, %%mm3         \n\t"\
62         "paddw                     %%mm5, %%mm4         \n\t"\
63         "test               %%"FF_REG_S", %%"FF_REG_S"  \n\t"\
64         " jnz                         2b                \n\t"\
65 
66 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
67     "lea                "offset"(%0), %%"FF_REG_d"  \n\t"\
68     "mov              (%%"FF_REG_d"), %%"FF_REG_S"  \n\t"\
69     "movq      "VROUNDER_OFFSET"(%0), "#dst1"       \n\t"\
70     "movq                    "#dst1", "#dst2"       \n\t"\
71     ".p2align                      4                \n\t"\
72     "2:                                             \n\t"\
73     "movq            8(%%"FF_REG_d"), "#coeff"      \n\t" /* filterCoeff */\
74     "movq  (%%"FF_REG_S", %%"FF_REG_a", 2), "#src1" \n\t" /* Y1srcData */\
75     "movq 8(%%"FF_REG_S", %%"FF_REG_a", 2), "#src2" \n\t" /* Y2srcData */\
76     "add                         $16, %%"FF_REG_d"  \n\t"\
77     "mov              (%%"FF_REG_d"), %%"FF_REG_S"  \n\t"\
78     "pmulhw                 "#coeff", "#src1"       \n\t"\
79     "pmulhw                 "#coeff", "#src2"       \n\t"\
80     "paddw                   "#src1", "#dst1"       \n\t"\
81     "paddw                   "#src2", "#dst2"       \n\t"\
82     "test               %%"FF_REG_S", %%"FF_REG_S"  \n\t"\
83     " jnz                         2b                \n\t"\
84 
85 #define YSCALEYUV2PACKEDX \
86     YSCALEYUV2PACKEDX_UV \
87     YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
88 
89 #define YSCALEYUV2PACKEDX_END                     \
90         :: "r" (&c->redDither),                   \
91             "m" (dummy), "m" (dummy), "m" (dummy),\
92             "r" (dest), "m" (dstW_reg), "m"(uv_off) \
93             NAMED_CONSTRAINTS_ADD(bF8,bFC) \
94         : "%"FF_REG_a, "%"FF_REG_d, "%"FF_REG_S            \
95     );
96 
97 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
98     __asm__ volatile(\
99         "xor %%"FF_REG_a", %%"FF_REG_a"                 \n\t"\
100         ".p2align                      4                \n\t"\
101         "nop                                            \n\t"\
102         "1:                                             \n\t"\
103         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"FF_REG_d"  \n\t"\
104         "mov              (%%"FF_REG_d"), %%"FF_REG_S"  \n\t"\
105         "pxor                      %%mm4, %%mm4         \n\t"\
106         "pxor                      %%mm5, %%mm5         \n\t"\
107         "pxor                      %%mm6, %%mm6         \n\t"\
108         "pxor                      %%mm7, %%mm7         \n\t"\
109         ".p2align                      4                \n\t"\
110         "2:                                             \n\t"\
111         "movq  (%%"FF_REG_S", %%"FF_REG_a"), %%mm0      \n\t" /* UsrcData */\
112         "add                          %6, %%"FF_REG_S"  \n\t" \
113         "movq  (%%"FF_REG_S", %%"FF_REG_a"), %%mm2      \n\t" /* VsrcData */\
114         "mov "STR(APCK_PTR2)"(%%"FF_REG_d"), %%"FF_REG_S" \n\t"\
115         "movq  (%%"FF_REG_S", %%"FF_REG_a"), %%mm1      \n\t" /* UsrcData */\
116         "movq                      %%mm0, %%mm3         \n\t"\
117         "punpcklwd                 %%mm1, %%mm0         \n\t"\
118         "punpckhwd                 %%mm1, %%mm3         \n\t"\
119         "movq "STR(APCK_COEF)"(%%"FF_REG_d"),%%mm1      \n\t" /* filterCoeff */\
120         "pmaddwd                   %%mm1, %%mm0         \n\t"\
121         "pmaddwd                   %%mm1, %%mm3         \n\t"\
122         "paddd                     %%mm0, %%mm4         \n\t"\
123         "paddd                     %%mm3, %%mm5         \n\t"\
124         "add                          %6, %%"FF_REG_S"  \n\t" \
125         "movq  (%%"FF_REG_S", %%"FF_REG_a"), %%mm3      \n\t" /* VsrcData */\
126         "mov "STR(APCK_SIZE)"(%%"FF_REG_d"), %%"FF_REG_S" \n\t"\
127         "add           $"STR(APCK_SIZE)", %%"FF_REG_d"  \n\t"\
128         "test               %%"FF_REG_S", %%"FF_REG_S"  \n\t"\
129         "movq                      %%mm2, %%mm0         \n\t"\
130         "punpcklwd                 %%mm3, %%mm2         \n\t"\
131         "punpckhwd                 %%mm3, %%mm0         \n\t"\
132         "pmaddwd                   %%mm1, %%mm2         \n\t"\
133         "pmaddwd                   %%mm1, %%mm0         \n\t"\
134         "paddd                     %%mm2, %%mm6         \n\t"\
135         "paddd                     %%mm0, %%mm7         \n\t"\
136         " jnz                         2b                \n\t"\
137         "psrad                       $16, %%mm4         \n\t"\
138         "psrad                       $16, %%mm5         \n\t"\
139         "psrad                       $16, %%mm6         \n\t"\
140         "psrad                       $16, %%mm7         \n\t"\
141         "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
142         "packssdw                  %%mm5, %%mm4         \n\t"\
143         "packssdw                  %%mm7, %%mm6         \n\t"\
144         "paddw                     %%mm0, %%mm4         \n\t"\
145         "paddw                     %%mm0, %%mm6         \n\t"\
146         "movq                      %%mm4, "U_TEMP"(%0)  \n\t"\
147         "movq                      %%mm6, "V_TEMP"(%0)  \n\t"\
148 
149 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
150     "lea                "offset"(%0), %%"FF_REG_d"      \n\t"\
151     "mov                 (%%"FF_REG_d"), %%"FF_REG_S"   \n\t"\
152     "pxor                      %%mm1, %%mm1         \n\t"\
153     "pxor                      %%mm5, %%mm5         \n\t"\
154     "pxor                      %%mm7, %%mm7         \n\t"\
155     "pxor                      %%mm6, %%mm6         \n\t"\
156     ".p2align                      4                \n\t"\
157     "2:                                             \n\t"\
158     "movq  (%%"FF_REG_S", %%"FF_REG_a", 2), %%mm0       \n\t" /* Y1srcData */\
159     "movq 8(%%"FF_REG_S", %%"FF_REG_a", 2), %%mm2       \n\t" /* Y2srcData */\
160     "mov "STR(APCK_PTR2)"(%%"FF_REG_d"), %%"FF_REG_S"   \n\t"\
161     "movq  (%%"FF_REG_S", %%"FF_REG_a", 2), %%mm4       \n\t" /* Y1srcData */\
162     "movq                      %%mm0, %%mm3         \n\t"\
163     "punpcklwd                 %%mm4, %%mm0         \n\t"\
164     "punpckhwd                 %%mm4, %%mm3         \n\t"\
165     "movq "STR(APCK_COEF)"(%%"FF_REG_d"), %%mm4     \n\t" /* filterCoeff */\
166     "pmaddwd                   %%mm4, %%mm0         \n\t"\
167     "pmaddwd                   %%mm4, %%mm3         \n\t"\
168     "paddd                     %%mm0, %%mm1         \n\t"\
169     "paddd                     %%mm3, %%mm5         \n\t"\
170     "movq 8(%%"FF_REG_S", %%"FF_REG_a", 2), %%mm3   \n\t" /* Y2srcData */\
171     "mov "STR(APCK_SIZE)"(%%"FF_REG_d"), %%"FF_REG_S" \n\t"\
172     "add           $"STR(APCK_SIZE)", %%"FF_REG_d"  \n\t"\
173     "test               %%"FF_REG_S", %%"FF_REG_S"  \n\t"\
174     "movq                      %%mm2, %%mm0         \n\t"\
175     "punpcklwd                 %%mm3, %%mm2         \n\t"\
176     "punpckhwd                 %%mm3, %%mm0         \n\t"\
177     "pmaddwd                   %%mm4, %%mm2         \n\t"\
178     "pmaddwd                   %%mm4, %%mm0         \n\t"\
179     "paddd                     %%mm2, %%mm7         \n\t"\
180     "paddd                     %%mm0, %%mm6         \n\t"\
181     " jnz                         2b                \n\t"\
182     "psrad                       $16, %%mm1         \n\t"\
183     "psrad                       $16, %%mm5         \n\t"\
184     "psrad                       $16, %%mm7         \n\t"\
185     "psrad                       $16, %%mm6         \n\t"\
186     "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
187     "packssdw                  %%mm5, %%mm1         \n\t"\
188     "packssdw                  %%mm6, %%mm7         \n\t"\
189     "paddw                     %%mm0, %%mm1         \n\t"\
190     "paddw                     %%mm0, %%mm7         \n\t"\
191     "movq               "U_TEMP"(%0), %%mm3         \n\t"\
192     "movq               "V_TEMP"(%0), %%mm4         \n\t"\
193 
194 #define YSCALEYUV2PACKEDX_ACCURATE \
195     YSCALEYUV2PACKEDX_ACCURATE_UV \
196     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
197 
198 #define YSCALEYUV2RGBX \
199     "psubw  "U_OFFSET"(%0), %%mm3       \n\t" /* (U-128)8*/\
200     "psubw  "V_OFFSET"(%0), %%mm4       \n\t" /* (V-128)8*/\
201     "movq            %%mm3, %%mm2       \n\t" /* (U-128)8*/\
202     "movq            %%mm4, %%mm5       \n\t" /* (V-128)8*/\
203     "pmulhw "UG_COEFF"(%0), %%mm3       \n\t"\
204     "pmulhw "VG_COEFF"(%0), %%mm4       \n\t"\
205     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
206     "pmulhw "UB_COEFF"(%0), %%mm2       \n\t"\
207     "pmulhw "VR_COEFF"(%0), %%mm5       \n\t"\
208     "psubw  "Y_OFFSET"(%0), %%mm1       \n\t" /* 8(Y-16)*/\
209     "psubw  "Y_OFFSET"(%0), %%mm7       \n\t" /* 8(Y-16)*/\
210     "pmulhw  "Y_COEFF"(%0), %%mm1       \n\t"\
211     "pmulhw  "Y_COEFF"(%0), %%mm7       \n\t"\
212     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
213     "paddw           %%mm3, %%mm4       \n\t"\
214     "movq            %%mm2, %%mm0       \n\t"\
215     "movq            %%mm5, %%mm6       \n\t"\
216     "movq            %%mm4, %%mm3       \n\t"\
217     "punpcklwd       %%mm2, %%mm2       \n\t"\
218     "punpcklwd       %%mm5, %%mm5       \n\t"\
219     "punpcklwd       %%mm4, %%mm4       \n\t"\
220     "paddw           %%mm1, %%mm2       \n\t"\
221     "paddw           %%mm1, %%mm5       \n\t"\
222     "paddw           %%mm1, %%mm4       \n\t"\
223     "punpckhwd       %%mm0, %%mm0       \n\t"\
224     "punpckhwd       %%mm6, %%mm6       \n\t"\
225     "punpckhwd       %%mm3, %%mm3       \n\t"\
226     "paddw           %%mm7, %%mm0       \n\t"\
227     "paddw           %%mm7, %%mm6       \n\t"\
228     "paddw           %%mm7, %%mm3       \n\t"\
229     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
230     "packuswb        %%mm0, %%mm2       \n\t"\
231     "packuswb        %%mm6, %%mm5       \n\t"\
232     "packuswb        %%mm3, %%mm4       \n\t"\
233 
234 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
235     "movq       "#b", "#q2"     \n\t" /* B */\
236     "movq       "#r", "#t"      \n\t" /* R */\
237     "punpcklbw  "#g", "#b"      \n\t" /* GBGBGBGB 0 */\
238     "punpcklbw  "#a", "#r"      \n\t" /* ARARARAR 0 */\
239     "punpckhbw  "#g", "#q2"     \n\t" /* GBGBGBGB 2 */\
240     "punpckhbw  "#a", "#t"      \n\t" /* ARARARAR 2 */\
241     "movq       "#b", "#q0"     \n\t" /* GBGBGBGB 0 */\
242     "movq      "#q2", "#q3"     \n\t" /* GBGBGBGB 2 */\
243     "punpcklwd  "#r", "#q0"     \n\t" /* ARGBARGB 0 */\
244     "punpckhwd  "#r", "#b"      \n\t" /* ARGBARGB 1 */\
245     "punpcklwd  "#t", "#q2"     \n\t" /* ARGBARGB 2 */\
246     "punpckhwd  "#t", "#q3"     \n\t" /* ARGBARGB 3 */\
247 \
248     MOVNTQ(   q0,   (dst, index, 4))\
249     MOVNTQ(    b,  8(dst, index, 4))\
250     MOVNTQ(   q2, 16(dst, index, 4))\
251     MOVNTQ(   q3, 24(dst, index, 4))\
252 \
253     "add      $8, "#index"      \n\t"\
254     "cmp  "dstw", "#index"      \n\t"\
255     " jb      1b                \n\t"
256 #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)
257 
RENAME(yuv2rgb32_X_ar)258 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
259                                    const int16_t **lumSrc, int lumFilterSize,
260                                    const int16_t *chrFilter, const int16_t **chrUSrc,
261                                    const int16_t **chrVSrc,
262                                    int chrFilterSize, const int16_t **alpSrc,
263                                    uint8_t *dest, int dstW, int dstY)
264 {
265     x86_reg dummy=0;
266     x86_reg dstW_reg = dstW;
267     x86_reg uv_off = c->uv_offx2;
268 
269     if (CONFIG_SWSCALE_ALPHA && c->needAlpha) {
270         YSCALEYUV2PACKEDX_ACCURATE
271         YSCALEYUV2RGBX
272         "movq                      %%mm2, "U_TEMP"(%0)  \n\t"
273         "movq                      %%mm4, "V_TEMP"(%0)  \n\t"
274         "movq                      %%mm5, "Y_TEMP"(%0)  \n\t"
275         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
276         "movq               "Y_TEMP"(%0), %%mm5         \n\t"
277         "psraw                        $3, %%mm1         \n\t"
278         "psraw                        $3, %%mm7         \n\t"
279         "packuswb                  %%mm7, %%mm1         \n\t"
280         WRITEBGR32(%4, "%5", %%FF_REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
281         YSCALEYUV2PACKEDX_END
282     } else {
283         YSCALEYUV2PACKEDX_ACCURATE
284         YSCALEYUV2RGBX
285         "pcmpeqd %%mm7, %%mm7 \n\t"
286         WRITEBGR32(%4, "%5", %%FF_REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
287         YSCALEYUV2PACKEDX_END
288     }
289 }
290 
RENAME(yuv2rgb32_X)291 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
292                                 const int16_t **lumSrc, int lumFilterSize,
293                                 const int16_t *chrFilter, const int16_t **chrUSrc,
294                                 const int16_t **chrVSrc,
295                                 int chrFilterSize, const int16_t **alpSrc,
296                                 uint8_t *dest, int dstW, int dstY)
297 {
298     x86_reg dummy=0;
299     x86_reg dstW_reg = dstW;
300     x86_reg uv_off = c->uv_offx2;
301 
302     if (CONFIG_SWSCALE_ALPHA && c->needAlpha) {
303         YSCALEYUV2PACKEDX
304         YSCALEYUV2RGBX
305         YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
306         "psraw                        $3, %%mm1         \n\t"
307         "psraw                        $3, %%mm7         \n\t"
308         "packuswb                  %%mm7, %%mm1         \n\t"
309         WRITEBGR32(%4, "%5", %%FF_REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
310         YSCALEYUV2PACKEDX_END
311     } else {
312         YSCALEYUV2PACKEDX
313         YSCALEYUV2RGBX
314         "pcmpeqd %%mm7, %%mm7 \n\t"
315         WRITEBGR32(%4, "%5", %%FF_REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
316         YSCALEYUV2PACKEDX_END
317     }
318 }
319 
RENAME(yuv2bgr32_X)320 static void RENAME(yuv2bgr32_X)(SwsContext *c, const int16_t *lumFilter,
321                                 const int16_t **lumSrc, int lumFilterSize,
322                                 const int16_t *chrFilter, const int16_t **chrUSrc,
323                                 const int16_t **chrVSrc,
324                                 int chrFilterSize, const int16_t **alpSrc,
325                                 uint8_t *dest, int dstW, int dstY)
326 {
327     x86_reg dummy=0;
328     x86_reg dstW_reg = dstW;
329     x86_reg uv_off = c->uv_offx2;
330 
331     if (CONFIG_SWSCALE_ALPHA && c->needAlpha) {
332         YSCALEYUV2PACKEDX
333         YSCALEYUV2RGBX
334         YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
335         "psraw                        $3, %%mm1         \n\t"
336         "psraw                        $3, %%mm7         \n\t"
337         "packuswb                  %%mm7, %%mm1         \n\t"
338         WRITEBGR32(%4, "%5", %%FF_REGa, %%mm5, %%mm4, %%mm2, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
339         YSCALEYUV2PACKEDX_END
340     } else {
341         YSCALEYUV2PACKEDX
342         YSCALEYUV2RGBX
343         "pcmpeqd %%mm7, %%mm7 \n\t"
344         WRITEBGR32(%4, "%5", %%FF_REGa, %%mm5, %%mm4, %%mm2, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
345         YSCALEYUV2PACKEDX_END
346     }
347 }
348 
349 #define REAL_WRITERGB16(dst, dstw, index) \
350     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
351     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\
352     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
353     "psrlq           $3, %%mm2  \n\t"\
354 \
355     "movq         %%mm2, %%mm1  \n\t"\
356     "movq         %%mm4, %%mm3  \n\t"\
357 \
358     "punpcklbw    %%mm7, %%mm3  \n\t"\
359     "punpcklbw    %%mm5, %%mm2  \n\t"\
360     "punpckhbw    %%mm7, %%mm4  \n\t"\
361     "punpckhbw    %%mm5, %%mm1  \n\t"\
362 \
363     "psllq           $3, %%mm3  \n\t"\
364     "psllq           $3, %%mm4  \n\t"\
365 \
366     "por          %%mm3, %%mm2  \n\t"\
367     "por          %%mm4, %%mm1  \n\t"\
368 \
369     MOVNTQ(%%mm2,  (dst, index, 2))\
370     MOVNTQ(%%mm1, 8(dst, index, 2))\
371 \
372     "add             $8, "#index"   \n\t"\
373     "cmp         "dstw", "#index"   \n\t"\
374     " jb             1b             \n\t"
375 #define WRITERGB16(dst, dstw, index)  REAL_WRITERGB16(dst, dstw, index)
376 
RENAME(yuv2rgb565_X_ar)377 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
378                                     const int16_t **lumSrc, int lumFilterSize,
379                                     const int16_t *chrFilter, const int16_t **chrUSrc,
380                                     const int16_t **chrVSrc,
381                                     int chrFilterSize, const int16_t **alpSrc,
382                                     uint8_t *dest, int dstW, int dstY)
383 {
384     x86_reg dummy=0;
385     x86_reg dstW_reg = dstW;
386     x86_reg uv_off = c->uv_offx2;
387 
388     YSCALEYUV2PACKEDX_ACCURATE
389     YSCALEYUV2RGBX
390     "pxor %%mm7, %%mm7 \n\t"
391     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
392 #ifdef DITHER1XBPP
393     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
394     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
395     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
396 #endif
397     WRITERGB16(%4, "%5", %%FF_REGa)
398     YSCALEYUV2PACKEDX_END
399 }
400 
RENAME(yuv2rgb565_X)401 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
402                                  const int16_t **lumSrc, int lumFilterSize,
403                                  const int16_t *chrFilter, const int16_t **chrUSrc,
404                                  const int16_t **chrVSrc,
405                                  int chrFilterSize, const int16_t **alpSrc,
406                                  uint8_t *dest, int dstW, int dstY)
407 {
408     x86_reg dummy=0;
409     x86_reg dstW_reg = dstW;
410     x86_reg uv_off = c->uv_offx2;
411 
412     YSCALEYUV2PACKEDX
413     YSCALEYUV2RGBX
414     "pxor %%mm7, %%mm7 \n\t"
415     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
416 #ifdef DITHER1XBPP
417     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
418     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
419     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
420 #endif
421     WRITERGB16(%4, "%5", %%FF_REGa)
422     YSCALEYUV2PACKEDX_END
423 }
424 
425 #define REAL_WRITERGB15(dst, dstw, index) \
426     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
427     "pand "MANGLE(bF8)", %%mm4  \n\t" /* G */\
428     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
429     "psrlq           $3, %%mm2  \n\t"\
430     "psrlq           $1, %%mm5  \n\t"\
431 \
432     "movq         %%mm2, %%mm1  \n\t"\
433     "movq         %%mm4, %%mm3  \n\t"\
434 \
435     "punpcklbw    %%mm7, %%mm3  \n\t"\
436     "punpcklbw    %%mm5, %%mm2  \n\t"\
437     "punpckhbw    %%mm7, %%mm4  \n\t"\
438     "punpckhbw    %%mm5, %%mm1  \n\t"\
439 \
440     "psllq           $2, %%mm3  \n\t"\
441     "psllq           $2, %%mm4  \n\t"\
442 \
443     "por          %%mm3, %%mm2  \n\t"\
444     "por          %%mm4, %%mm1  \n\t"\
445 \
446     MOVNTQ(%%mm2,  (dst, index, 2))\
447     MOVNTQ(%%mm1, 8(dst, index, 2))\
448 \
449     "add             $8, "#index"   \n\t"\
450     "cmp         "dstw", "#index"   \n\t"\
451     " jb             1b             \n\t"
452 #define WRITERGB15(dst, dstw, index)  REAL_WRITERGB15(dst, dstw, index)
453 
RENAME(yuv2rgb555_X_ar)454 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
455                                     const int16_t **lumSrc, int lumFilterSize,
456                                     const int16_t *chrFilter, const int16_t **chrUSrc,
457                                     const int16_t **chrVSrc,
458                                     int chrFilterSize, const int16_t **alpSrc,
459                                     uint8_t *dest, int dstW, int dstY)
460 {
461     x86_reg dummy=0;
462     x86_reg dstW_reg = dstW;
463     x86_reg uv_off = c->uv_offx2;
464 
465     YSCALEYUV2PACKEDX_ACCURATE
466     YSCALEYUV2RGBX
467     "pxor %%mm7, %%mm7 \n\t"
468     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
469 #ifdef DITHER1XBPP
470     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
471     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
472     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
473 #endif
474     WRITERGB15(%4, "%5", %%FF_REGa)
475     YSCALEYUV2PACKEDX_END
476 }
477 
RENAME(yuv2rgb555_X)478 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
479                                  const int16_t **lumSrc, int lumFilterSize,
480                                  const int16_t *chrFilter, const int16_t **chrUSrc,
481                                  const int16_t **chrVSrc,
482                                  int chrFilterSize, const int16_t **alpSrc,
483                                  uint8_t *dest, int dstW, int dstY)
484 {
485     x86_reg dummy=0;
486     x86_reg dstW_reg = dstW;
487     x86_reg uv_off = c->uv_offx2;
488 
489     YSCALEYUV2PACKEDX
490     YSCALEYUV2RGBX
491     "pxor %%mm7, %%mm7 \n\t"
492     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
493 #ifdef DITHER1XBPP
494     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
495     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
496     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
497 #endif
498     WRITERGB15(%4, "%5", %%FF_REGa)
499     YSCALEYUV2PACKEDX_END
500 }
501 
502 #define WRITEBGR24MMX(dst, dstw, index) \
503     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
504     "movq      %%mm2, %%mm1     \n\t" /* B */\
505     "movq      %%mm5, %%mm6     \n\t" /* R */\
506     "punpcklbw %%mm4, %%mm2     \n\t" /* GBGBGBGB 0 */\
507     "punpcklbw %%mm7, %%mm5     \n\t" /* 0R0R0R0R 0 */\
508     "punpckhbw %%mm4, %%mm1     \n\t" /* GBGBGBGB 2 */\
509     "punpckhbw %%mm7, %%mm6     \n\t" /* 0R0R0R0R 2 */\
510     "movq      %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */\
511     "movq      %%mm1, %%mm3     \n\t" /* GBGBGBGB 2 */\
512     "punpcklwd %%mm5, %%mm0     \n\t" /* 0RGB0RGB 0 */\
513     "punpckhwd %%mm5, %%mm2     \n\t" /* 0RGB0RGB 1 */\
514     "punpcklwd %%mm6, %%mm1     \n\t" /* 0RGB0RGB 2 */\
515     "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */\
516 \
517     "movq      %%mm0, %%mm4     \n\t" /* 0RGB0RGB 0 */\
518     "movq      %%mm2, %%mm6     \n\t" /* 0RGB0RGB 1 */\
519     "movq      %%mm1, %%mm5     \n\t" /* 0RGB0RGB 2 */\
520     "movq      %%mm3, %%mm7     \n\t" /* 0RGB0RGB 3 */\
521 \
522     "psllq       $40, %%mm0     \n\t" /* RGB00000 0 */\
523     "psllq       $40, %%mm2     \n\t" /* RGB00000 1 */\
524     "psllq       $40, %%mm1     \n\t" /* RGB00000 2 */\
525     "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */\
526 \
527     "punpckhdq %%mm4, %%mm0     \n\t" /* 0RGBRGB0 0 */\
528     "punpckhdq %%mm6, %%mm2     \n\t" /* 0RGBRGB0 1 */\
529     "punpckhdq %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */\
530     "punpckhdq %%mm7, %%mm3     \n\t" /* 0RGBRGB0 3 */\
531 \
532     "psrlq        $8, %%mm0     \n\t" /* 00RGBRGB 0 */\
533     "movq      %%mm2, %%mm6     \n\t" /* 0RGBRGB0 1 */\
534     "psllq       $40, %%mm2     \n\t" /* GB000000 1 */\
535     "por       %%mm2, %%mm0     \n\t" /* GBRGBRGB 0 */\
536     MOVNTQ(%%mm0, (dst))\
537 \
538     "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */\
539     "movq      %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */\
540     "psllq       $24, %%mm1     \n\t" /* BRGB0000 2 */\
541     "por       %%mm1, %%mm6     \n\t" /* BRGBRGBR 1 */\
542     MOVNTQ(%%mm6, 8(dst))\
543 \
544     "psrlq       $40, %%mm5     \n\t" /* 000000RG 2 */\
545     "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */\
546     "por       %%mm3, %%mm5     \n\t" /* RGBRGBRG 2 */\
547     MOVNTQ(%%mm5, 16(dst))\
548 \
549     "add         $24, "#dst"    \n\t"\
550 \
551     "add          $8, "#index"  \n\t"\
552     "cmp      "dstw", "#index"  \n\t"\
553     " jb          1b            \n\t"
554 
555 #define WRITEBGR24MMXEXT(dst, dstw, index) \
556     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
557     "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
558     "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
559     "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\
560     "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\
561     "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\
562 \
563     "pand   %%mm0, %%mm1        \n\t" /*    B2        B1       B0 */\
564     "pand   %%mm0, %%mm3        \n\t" /*    G2        G1       G0 */\
565     "pand   %%mm7, %%mm6        \n\t" /*       R1        R0       */\
566 \
567     "psllq     $8, %%mm3        \n\t" /* G2        G1       G0    */\
568     "por    %%mm1, %%mm6        \n\t"\
569     "por    %%mm3, %%mm6        \n\t"\
570     MOVNTQ(%%mm6, (dst))\
571 \
572     "psrlq     $8, %%mm4        \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\
573     "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\
574     "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\
575     "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\
576 \
577     "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5       B4        B3    */\
578     "pand   %%mm7, %%mm3        \n\t" /*       G4        G3       */\
579     "pand   %%mm0, %%mm6        \n\t" /*    R4        R3       R2 */\
580 \
581     "por    %%mm1, %%mm3        \n\t" /* B5    G4 B4     G3 B3    */\
582     "por    %%mm3, %%mm6        \n\t"\
583     MOVNTQ(%%mm6, 8(dst))\
584 \
585     "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\
586     "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\
587     "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\
588 \
589     "pand   %%mm7, %%mm1        \n\t" /*       B7        B6       */\
590     "pand   %%mm0, %%mm3        \n\t" /*    G7        G6       G5 */\
591     "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7       R6        R5    */\
592 \
593     "por    %%mm1, %%mm3        \n\t"\
594     "por    %%mm3, %%mm6        \n\t"\
595     MOVNTQ(%%mm6, 16(dst))\
596 \
597     "add      $24, "#dst"       \n\t"\
598 \
599     "add       $8, "#index"     \n\t"\
600     "cmp   "dstw", "#index"     \n\t"\
601     " jb       1b               \n\t"
602 
603 #if COMPILE_TEMPLATE_MMXEXT
604 #undef WRITEBGR24
605 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMXEXT(dst, dstw, index)
606 #else
607 #undef WRITEBGR24
608 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX(dst, dstw, index)
609 #endif
610 
611 #if HAVE_6REGS
RENAME(yuv2bgr24_X_ar)612 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
613                                    const int16_t **lumSrc, int lumFilterSize,
614                                    const int16_t *chrFilter, const int16_t **chrUSrc,
615                                    const int16_t **chrVSrc,
616                                    int chrFilterSize, const int16_t **alpSrc,
617                                    uint8_t *dest, int dstW, int dstY)
618 {
619     x86_reg dummy=0;
620     x86_reg dstW_reg = dstW;
621     x86_reg uv_off = c->uv_offx2;
622 
623     YSCALEYUV2PACKEDX_ACCURATE
624     YSCALEYUV2RGBX
625     "pxor %%mm7, %%mm7 \n\t"
626     "lea (%%"FF_REG_a", %%"FF_REG_a", 2), %%"FF_REG_c"\n\t" //FIXME optimize
627     "add %4, %%"FF_REG_c"                        \n\t"
628     WRITEBGR24(%%FF_REGc, "%5", %%FF_REGa)
629     :: "r" (&c->redDither),
630        "m" (dummy), "m" (dummy), "m" (dummy),
631        "r" (dest), "m" (dstW_reg), "m"(uv_off)
632        NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
633     : "%"FF_REG_a, "%"FF_REG_c, "%"FF_REG_d, "%"FF_REG_S
634     );
635 }
636 
RENAME(yuv2bgr24_X)637 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
638                                 const int16_t **lumSrc, int lumFilterSize,
639                                 const int16_t *chrFilter, const int16_t **chrUSrc,
640                                 const int16_t **chrVSrc,
641                                 int chrFilterSize, const int16_t **alpSrc,
642                                 uint8_t *dest, int dstW, int dstY)
643 {
644     x86_reg dummy=0;
645     x86_reg dstW_reg = dstW;
646     x86_reg uv_off = c->uv_offx2;
647 
648     YSCALEYUV2PACKEDX
649     YSCALEYUV2RGBX
650     "pxor                    %%mm7, %%mm7              \n\t"
651     "lea (%%"FF_REG_a", %%"FF_REG_a", 2), %%"FF_REG_c" \n\t" //FIXME optimize
652     "add                        %4, %%"FF_REG_c"       \n\t"
653     WRITEBGR24(%%FF_REGc, "%5", %%FF_REGa)
654     :: "r" (&c->redDither),
655        "m" (dummy), "m" (dummy), "m" (dummy),
656        "r" (dest),  "m" (dstW_reg), "m"(uv_off)
657        NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
658     : "%"FF_REG_a, "%"FF_REG_c, "%"FF_REG_d, "%"FF_REG_S
659     );
660 }
661 #endif /* HAVE_6REGS */
662 
663 #define REAL_WRITEYUY2(dst, dstw, index) \
664     "packuswb  %%mm3, %%mm3     \n\t"\
665     "packuswb  %%mm4, %%mm4     \n\t"\
666     "packuswb  %%mm7, %%mm1     \n\t"\
667     "punpcklbw %%mm4, %%mm3     \n\t"\
668     "movq      %%mm1, %%mm7     \n\t"\
669     "punpcklbw %%mm3, %%mm1     \n\t"\
670     "punpckhbw %%mm3, %%mm7     \n\t"\
671 \
672     MOVNTQ(%%mm1, (dst, index, 2))\
673     MOVNTQ(%%mm7, 8(dst, index, 2))\
674 \
675     "add          $8, "#index"  \n\t"\
676     "cmp      "dstw", "#index"  \n\t"\
677     " jb          1b            \n\t"
678 #define WRITEYUY2(dst, dstw, index)  REAL_WRITEYUY2(dst, dstw, index)
679 
RENAME(yuv2yuyv422_X_ar)680 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
681                                      const int16_t **lumSrc, int lumFilterSize,
682                                      const int16_t *chrFilter, const int16_t **chrUSrc,
683                                      const int16_t **chrVSrc,
684                                      int chrFilterSize, const int16_t **alpSrc,
685                                      uint8_t *dest, int dstW, int dstY)
686 {
687     x86_reg dummy=0;
688     x86_reg dstW_reg = dstW;
689     x86_reg uv_off = c->uv_offx2;
690 
691     YSCALEYUV2PACKEDX_ACCURATE
692     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
693     "psraw $3, %%mm3    \n\t"
694     "psraw $3, %%mm4    \n\t"
695     "psraw $3, %%mm1    \n\t"
696     "psraw $3, %%mm7    \n\t"
697     WRITEYUY2(%4, "%5", %%FF_REGa)
698     YSCALEYUV2PACKEDX_END
699 }
700 
RENAME(yuv2yuyv422_X)701 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
702                                   const int16_t **lumSrc, int lumFilterSize,
703                                   const int16_t *chrFilter, const int16_t **chrUSrc,
704                                   const int16_t **chrVSrc,
705                                   int chrFilterSize, const int16_t **alpSrc,
706                                   uint8_t *dest, int dstW, int dstY)
707 {
708     x86_reg dummy=0;
709     x86_reg dstW_reg = dstW;
710     x86_reg uv_off = c->uv_offx2;
711 
712     YSCALEYUV2PACKEDX
713     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
714     "psraw $3, %%mm3    \n\t"
715     "psraw $3, %%mm4    \n\t"
716     "psraw $3, %%mm1    \n\t"
717     "psraw $3, %%mm7    \n\t"
718     WRITEYUY2(%4, "%5", %%FF_REGa)
719     YSCALEYUV2PACKEDX_END
720 }
721 
722 #define REAL_YSCALEYUV2RGB_UV(index, c) \
723     "xor            "#index", "#index"  \n\t"\
724     ".p2align              4            \n\t"\
725     "1:                                 \n\t"\
726     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
727     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
728     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
729     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
730     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
731     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
732     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
733     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
734     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
735     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
736     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
737     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
738     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
739     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
740     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
741     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
742     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
743     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
744     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
745     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
746     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
747     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
748 
749 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
750     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
751     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
752     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
753     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
754     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
755     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
756     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
757     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
758     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
759     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
760     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
761     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
762 
763 #define REAL_YSCALEYUV2RGB_COEFF(c) \
764     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
765     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
766     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
767     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
768     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
769     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
770     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
771     "paddw             %%mm3, %%mm4     \n\t"\
772     "movq              %%mm2, %%mm0     \n\t"\
773     "movq              %%mm5, %%mm6     \n\t"\
774     "movq              %%mm4, %%mm3     \n\t"\
775     "punpcklwd         %%mm2, %%mm2     \n\t"\
776     "punpcklwd         %%mm5, %%mm5     \n\t"\
777     "punpcklwd         %%mm4, %%mm4     \n\t"\
778     "paddw             %%mm1, %%mm2     \n\t"\
779     "paddw             %%mm1, %%mm5     \n\t"\
780     "paddw             %%mm1, %%mm4     \n\t"\
781     "punpckhwd         %%mm0, %%mm0     \n\t"\
782     "punpckhwd         %%mm6, %%mm6     \n\t"\
783     "punpckhwd         %%mm3, %%mm3     \n\t"\
784     "paddw             %%mm7, %%mm0     \n\t"\
785     "paddw             %%mm7, %%mm6     \n\t"\
786     "paddw             %%mm7, %%mm3     \n\t"\
787     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
788     "packuswb          %%mm0, %%mm2     \n\t"\
789     "packuswb          %%mm6, %%mm5     \n\t"\
790     "packuswb          %%mm3, %%mm4     \n\t"\
791 
792 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
793 
794 #define YSCALEYUV2RGB(index, c) \
795     REAL_YSCALEYUV2RGB_UV(index, c) \
796     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
797     REAL_YSCALEYUV2RGB_COEFF(c)
798 
799 /**
800  * vertical bilinear scale YV12 to RGB
801  */
RENAME(yuv2rgb32_2)802 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
803                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
804                                 const int16_t *abuf[2], uint8_t *dest,
805                                 int dstW, int yalpha, int uvalpha, int y)
806 {
807     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
808                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
809 
810     if (CONFIG_SWSCALE_ALPHA && c->needAlpha) {
811         const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
812 #if ARCH_X86_64
813         __asm__ volatile(
814             YSCALEYUV2RGB(%%r8, %5)
815             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
816             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
817             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
818             "packuswb            %%mm7, %%mm1       \n\t"
819             WRITEBGR32(%4, DSTW_OFFSET"(%5)", %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
820             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
821                "a" (&c->redDither),
822                "r" (abuf0), "r" (abuf1)
823             : "%r8"
824         );
825 #else
826         c->u_temp=(intptr_t)abuf0;
827         c->v_temp=(intptr_t)abuf1;
828         __asm__ volatile(
829             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
830             "mov        %4, %%"FF_REG_b"            \n\t"
831             "push %%"FF_REG_BP"                     \n\t"
832             YSCALEYUV2RGB(%%FF_REGBP, %5)
833             "push                   %0              \n\t"
834             "push                   %1              \n\t"
835             "mov          "U_TEMP"(%5), %0          \n\t"
836             "mov          "V_TEMP"(%5), %1          \n\t"
837             YSCALEYUV2RGB_YA(%%FF_REGBP, %5, %0, %1)
838             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
839             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
840             "packuswb            %%mm7, %%mm1       \n\t"
841             "pop                    %1              \n\t"
842             "pop                    %0              \n\t"
843             WRITEBGR32(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
844             "pop %%"FF_REG_BP"                      \n\t"
845             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
846             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
847                "a" (&c->redDither)
848         );
849 #endif
850     } else {
851         __asm__ volatile(
852             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
853             "mov        %4, %%"FF_REG_b"            \n\t"
854             "push %%"FF_REG_BP"                     \n\t"
855             YSCALEYUV2RGB(%%FF_REGBP, %5)
856             "pcmpeqd %%mm7, %%mm7                   \n\t"
857             WRITEBGR32(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
858             "pop %%"FF_REG_BP"                      \n\t"
859             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
860             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
861                "a" (&c->redDither)
862         );
863     }
864 }
865 
RENAME(yuv2bgr24_2)866 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
867                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
868                                 const int16_t *abuf[2], uint8_t *dest,
869                                 int dstW, int yalpha, int uvalpha, int y)
870 {
871     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
872                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
873 
874     __asm__ volatile(
875         "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
876         "mov           %4, %%"FF_REG_b"         \n\t"
877         "push %%"FF_REG_BP"                     \n\t"
878         YSCALEYUV2RGB(%%FF_REGBP, %5)
879         "pxor    %%mm7, %%mm7                   \n\t"
880         WRITEBGR24(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
881         "pop %%"FF_REG_BP"                      \n\t"
882         "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
883         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
884            "a" (&c->redDither)
885            NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
886     );
887 }
888 
RENAME(yuv2rgb555_2)889 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
890                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
891                                  const int16_t *abuf[2], uint8_t *dest,
892                                  int dstW, int yalpha, int uvalpha, int y)
893 {
894     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
895                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
896 
897     __asm__ volatile(
898         "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
899         "mov        %4, %%"FF_REG_b"            \n\t"
900         "push %%"FF_REG_BP"                     \n\t"
901         YSCALEYUV2RGB(%%FF_REGBP, %5)
902         "pxor    %%mm7, %%mm7                   \n\t"
903         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
904 #ifdef DITHER1XBPP
905         "paddusb "BLUE_DITHER"(%5), %%mm2       \n\t"
906         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
907         "paddusb "RED_DITHER"(%5), %%mm5        \n\t"
908 #endif
909         WRITERGB15(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
910         "pop %%"FF_REG_BP"                      \n\t"
911         "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
912         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
913            "a" (&c->redDither)
914            NAMED_CONSTRAINTS_ADD(bF8)
915     );
916 }
917 
RENAME(yuv2rgb565_2)918 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
919                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
920                                  const int16_t *abuf[2], uint8_t *dest,
921                                  int dstW, int yalpha, int uvalpha, int y)
922 {
923     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
924                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
925 
926     __asm__ volatile(
927         "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
928         "mov           %4, %%"FF_REG_b"         \n\t"
929         "push %%"FF_REG_BP"                     \n\t"
930         YSCALEYUV2RGB(%%FF_REGBP, %5)
931         "pxor    %%mm7, %%mm7                   \n\t"
932         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
933 #ifdef DITHER1XBPP
934         "paddusb "BLUE_DITHER"(%5), %%mm2       \n\t"
935         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
936         "paddusb "RED_DITHER"(%5), %%mm5        \n\t"
937 #endif
938         WRITERGB16(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
939         "pop %%"FF_REG_BP"                      \n\t"
940         "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
941         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
942            "a" (&c->redDither)
943            NAMED_CONSTRAINTS_ADD(bF8,bFC)
944     );
945 }
946 
947 #define REAL_YSCALEYUV2PACKED(index, c) \
948     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\
949     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\
950     "psraw                $3, %%mm0                           \n\t"\
951     "psraw                $3, %%mm1                           \n\t"\
952     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
953     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
954     "xor            "#index", "#index"                        \n\t"\
955     ".p2align              4            \n\t"\
956     "1:                                 \n\t"\
957     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
958     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
959     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
960     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
961     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
962     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
963     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
964     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
965     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
966     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
967     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
968     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
969     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
970     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
971     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
972     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
973     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
974     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
975     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
976     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
977     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
978     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
979     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
980     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
981     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
982     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
983     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
984 
985 #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)
986 
RENAME(yuv2yuyv422_2)987 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
988                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
989                                   const int16_t *abuf[2], uint8_t *dest,
990                                   int dstW, int yalpha, int uvalpha, int y)
991 {
992     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
993                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
994 
995     __asm__ volatile(
996         "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
997         "mov           %4, %%"FF_REG_b"         \n\t"
998         "push %%"FF_REG_BP"                     \n\t"
999         YSCALEYUV2PACKED(%%FF_REGBP, %5)
1000         WRITEYUY2(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1001         "pop %%"FF_REG_BP"                      \n\t"
1002         "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1003         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1004            "a" (&c->redDither)
1005     );
1006 }
1007 
1008 #define REAL_YSCALEYUV2RGB1(index, c) \
1009     "xor            "#index", "#index"  \n\t"\
1010     ".p2align              4            \n\t"\
1011     "1:                                 \n\t"\
1012     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1013     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1014     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1015     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1016     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1017     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1018     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1019     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1020     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1021     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1022     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1023     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1024     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1025     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1026     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1027     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1028     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1029     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1030     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1031     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1032     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1033     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1034     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1035     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1036     "paddw             %%mm3, %%mm4     \n\t"\
1037     "movq              %%mm2, %%mm0     \n\t"\
1038     "movq              %%mm5, %%mm6     \n\t"\
1039     "movq              %%mm4, %%mm3     \n\t"\
1040     "punpcklwd         %%mm2, %%mm2     \n\t"\
1041     "punpcklwd         %%mm5, %%mm5     \n\t"\
1042     "punpcklwd         %%mm4, %%mm4     \n\t"\
1043     "paddw             %%mm1, %%mm2     \n\t"\
1044     "paddw             %%mm1, %%mm5     \n\t"\
1045     "paddw             %%mm1, %%mm4     \n\t"\
1046     "punpckhwd         %%mm0, %%mm0     \n\t"\
1047     "punpckhwd         %%mm6, %%mm6     \n\t"\
1048     "punpckhwd         %%mm3, %%mm3     \n\t"\
1049     "paddw             %%mm7, %%mm0     \n\t"\
1050     "paddw             %%mm7, %%mm6     \n\t"\
1051     "paddw             %%mm7, %%mm3     \n\t"\
1052     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1053     "packuswb          %%mm0, %%mm2     \n\t"\
1054     "packuswb          %%mm6, %%mm5     \n\t"\
1055     "packuswb          %%mm3, %%mm4     \n\t"\
1056 
1057 #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)
1058 
1059 // do vertical chrominance interpolation
1060 #define REAL_YSCALEYUV2RGB1b(index, c) \
1061     "xor            "#index", "#index"  \n\t"\
1062     ".p2align              4            \n\t"\
1063     "1:                                 \n\t"\
1064     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1065     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1066     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1067     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1068     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1069     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1070     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1071     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1072     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\
1073     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\
1074     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1075     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1076     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1077     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1078     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1079     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1080     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1081     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1082     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1083     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1084     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1085     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1086     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1087     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1088     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1089     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1090     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1091     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1092     "paddw             %%mm3, %%mm4     \n\t"\
1093     "movq              %%mm2, %%mm0     \n\t"\
1094     "movq              %%mm5, %%mm6     \n\t"\
1095     "movq              %%mm4, %%mm3     \n\t"\
1096     "punpcklwd         %%mm2, %%mm2     \n\t"\
1097     "punpcklwd         %%mm5, %%mm5     \n\t"\
1098     "punpcklwd         %%mm4, %%mm4     \n\t"\
1099     "paddw             %%mm1, %%mm2     \n\t"\
1100     "paddw             %%mm1, %%mm5     \n\t"\
1101     "paddw             %%mm1, %%mm4     \n\t"\
1102     "punpckhwd         %%mm0, %%mm0     \n\t"\
1103     "punpckhwd         %%mm6, %%mm6     \n\t"\
1104     "punpckhwd         %%mm3, %%mm3     \n\t"\
1105     "paddw             %%mm7, %%mm0     \n\t"\
1106     "paddw             %%mm7, %%mm6     \n\t"\
1107     "paddw             %%mm7, %%mm3     \n\t"\
1108     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1109     "packuswb          %%mm0, %%mm2     \n\t"\
1110     "packuswb          %%mm6, %%mm5     \n\t"\
1111     "packuswb          %%mm3, %%mm4     \n\t"\
1112 
1113 #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)
1114 
1115 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1116     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\
1117     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\
1118     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\
1119     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\
1120     "packuswb          %%mm1, %%mm7     \n\t"
1121 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1122 
1123 /**
1124  * YV12 to RGB without scaling or interpolating
1125  */
RENAME(yuv2rgb32_1)1126 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1127                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1128                                 const int16_t *abuf0, uint8_t *dest,
1129                                 int dstW, int uvalpha, int y)
1130 {
1131     const int16_t *ubuf0 = ubuf[0];
1132     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1133 
1134     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1135         const int16_t *ubuf1 = ubuf[0];
1136         if (CONFIG_SWSCALE_ALPHA && c->needAlpha) {
1137             __asm__ volatile(
1138                 "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1139                 "mov           %4, %%"FF_REG_b"         \n\t"
1140                 "push %%"FF_REG_BP"                     \n\t"
1141                 YSCALEYUV2RGB1(%%FF_REGBP, %5)
1142                 YSCALEYUV2RGB1_ALPHA(%%FF_REGBP)
1143                 WRITEBGR32(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1144                 "pop %%"FF_REG_BP"                      \n\t"
1145                 "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1146                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1147                    "a" (&c->redDither)
1148             );
1149         } else {
1150             __asm__ volatile(
1151                 "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1152                 "mov           %4, %%"FF_REG_b"         \n\t"
1153                 "push %%"FF_REG_BP"                     \n\t"
1154                 YSCALEYUV2RGB1(%%FF_REGBP, %5)
1155                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1156                 WRITEBGR32(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1157                 "pop %%"FF_REG_BP"                      \n\t"
1158                 "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1159                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1160                    "a" (&c->redDither)
1161             );
1162         }
1163     } else {
1164         const int16_t *ubuf1 = ubuf[1];
1165         if (CONFIG_SWSCALE_ALPHA && c->needAlpha) {
1166             __asm__ volatile(
1167                 "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1168                 "mov           %4, %%"FF_REG_b"         \n\t"
1169                 "push %%"FF_REG_BP"                     \n\t"
1170                 YSCALEYUV2RGB1b(%%FF_REGBP, %5)
1171                 YSCALEYUV2RGB1_ALPHA(%%FF_REGBP)
1172                 WRITEBGR32(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1173                 "pop %%"FF_REG_BP"                      \n\t"
1174                 "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1175                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1176                    "a" (&c->redDither)
1177             );
1178         } else {
1179             __asm__ volatile(
1180                 "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1181                 "mov           %4, %%"FF_REG_b"         \n\t"
1182                 "push %%"FF_REG_BP"                     \n\t"
1183                 YSCALEYUV2RGB1b(%%FF_REGBP, %5)
1184                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1185                 WRITEBGR32(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1186                 "pop %%"FF_REG_BP"                      \n\t"
1187                 "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1188                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1189                    "a" (&c->redDither)
1190             );
1191         }
1192     }
1193 }
1194 
RENAME(yuv2bgr24_1)1195 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1196                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1197                                 const int16_t *abuf0, uint8_t *dest,
1198                                 int dstW, int uvalpha, int y)
1199 {
1200     const int16_t *ubuf0 = ubuf[0];
1201     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1202 
1203     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1204         const int16_t *ubuf1 = ubuf[0];
1205         __asm__ volatile(
1206             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1207             "mov           %4, %%"FF_REG_b"         \n\t"
1208             "push %%"FF_REG_BP"                     \n\t"
1209             YSCALEYUV2RGB1(%%FF_REGBP, %5)
1210             "pxor    %%mm7, %%mm7                   \n\t"
1211             WRITEBGR24(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1212             "pop %%"FF_REG_BP"                      \n\t"
1213             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1214             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1215                "a" (&c->redDither)
1216                NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
1217         );
1218     } else {
1219         const int16_t *ubuf1 = ubuf[1];
1220         __asm__ volatile(
1221             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1222             "mov           %4, %%"FF_REG_b"         \n\t"
1223             "push %%"FF_REG_BP"                     \n\t"
1224             YSCALEYUV2RGB1b(%%FF_REGBP, %5)
1225             "pxor    %%mm7, %%mm7                   \n\t"
1226             WRITEBGR24(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1227             "pop %%"FF_REG_BP"                      \n\t"
1228             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1229             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1230                "a" (&c->redDither)
1231                NAMED_CONSTRAINTS_ADD(ff_M24A,ff_M24C,ff_M24B)
1232         );
1233     }
1234 }
1235 
RENAME(yuv2rgb555_1)1236 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1237                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1238                                  const int16_t *abuf0, uint8_t *dest,
1239                                  int dstW, int uvalpha, int y)
1240 {
1241     const int16_t *ubuf0 = ubuf[0];
1242     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1243 
1244     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1245         const int16_t *ubuf1 = ubuf[0];
1246         __asm__ volatile(
1247             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1248             "mov           %4, %%"FF_REG_b"         \n\t"
1249             "push %%"FF_REG_BP"                     \n\t"
1250             YSCALEYUV2RGB1(%%FF_REGBP, %5)
1251             "pxor    %%mm7, %%mm7                   \n\t"
1252             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1253 #ifdef DITHER1XBPP
1254             "paddusb "BLUE_DITHER"(%5), %%mm2       \n\t"
1255             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1256             "paddusb "RED_DITHER"(%5), %%mm5        \n\t"
1257 #endif
1258             WRITERGB15(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1259             "pop %%"FF_REG_BP"                      \n\t"
1260             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1261             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1262                "a" (&c->redDither)
1263                NAMED_CONSTRAINTS_ADD(bF8)
1264         );
1265     } else {
1266         const int16_t *ubuf1 = ubuf[1];
1267         __asm__ volatile(
1268             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1269             "mov           %4, %%"FF_REG_b"         \n\t"
1270             "push %%"FF_REG_BP"                     \n\t"
1271             YSCALEYUV2RGB1b(%%FF_REGBP, %5)
1272             "pxor    %%mm7, %%mm7                   \n\t"
1273             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1274 #ifdef DITHER1XBPP
1275             "paddusb "BLUE_DITHER"(%5), %%mm2       \n\t"
1276             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1277             "paddusb "RED_DITHER"(%5), %%mm5        \n\t"
1278 #endif
1279             WRITERGB15(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1280             "pop %%"FF_REG_BP"                      \n\t"
1281             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1282             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1283                "a" (&c->redDither)
1284                NAMED_CONSTRAINTS_ADD(bF8)
1285         );
1286     }
1287 }
1288 
RENAME(yuv2rgb565_1)1289 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1290                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1291                                  const int16_t *abuf0, uint8_t *dest,
1292                                  int dstW, int uvalpha, int y)
1293 {
1294     const int16_t *ubuf0 = ubuf[0];
1295     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1296 
1297     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1298         const int16_t *ubuf1 = ubuf[0];
1299         __asm__ volatile(
1300             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1301             "mov           %4, %%"FF_REG_b"         \n\t"
1302             "push %%"FF_REG_BP"                     \n\t"
1303             YSCALEYUV2RGB1(%%FF_REGBP, %5)
1304             "pxor    %%mm7, %%mm7                   \n\t"
1305             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1306 #ifdef DITHER1XBPP
1307             "paddusb "BLUE_DITHER"(%5), %%mm2       \n\t"
1308             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1309             "paddusb "RED_DITHER"(%5), %%mm5        \n\t"
1310 #endif
1311             WRITERGB16(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1312             "pop %%"FF_REG_BP"                      \n\t"
1313             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1314             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1315                "a" (&c->redDither)
1316                NAMED_CONSTRAINTS_ADD(bF8,bFC)
1317         );
1318     } else {
1319         const int16_t *ubuf1 = ubuf[1];
1320         __asm__ volatile(
1321             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1322             "mov           %4, %%"FF_REG_b"         \n\t"
1323             "push %%"FF_REG_BP"                     \n\t"
1324             YSCALEYUV2RGB1b(%%FF_REGBP, %5)
1325             "pxor    %%mm7, %%mm7                   \n\t"
1326             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1327 #ifdef DITHER1XBPP
1328             "paddusb "BLUE_DITHER"(%5), %%mm2       \n\t"
1329             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1330             "paddusb "RED_DITHER"(%5), %%mm5        \n\t"
1331 #endif
1332             WRITERGB16(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1333             "pop %%"FF_REG_BP"                      \n\t"
1334             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1335             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1336                "a" (&c->redDither)
1337                NAMED_CONSTRAINTS_ADD(bF8,bFC)
1338         );
1339     }
1340 }
1341 
1342 #define REAL_YSCALEYUV2PACKED1(index, c) \
1343     "xor            "#index", "#index"  \n\t"\
1344     ".p2align              4            \n\t"\
1345     "1:                                 \n\t"\
1346     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1347     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1348     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1349     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1350     "psraw                $7, %%mm3     \n\t" \
1351     "psraw                $7, %%mm4     \n\t" \
1352     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1353     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1354     "psraw                $7, %%mm1     \n\t" \
1355     "psraw                $7, %%mm7     \n\t" \
1356 
1357 #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)
1358 
1359 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1360     "xor "#index", "#index"             \n\t"\
1361     ".p2align              4            \n\t"\
1362     "1:                                 \n\t"\
1363     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1364     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1365     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1366     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1367     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1368     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1369     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1370     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1371     "psrlw                $8, %%mm3     \n\t" \
1372     "psrlw                $8, %%mm4     \n\t" \
1373     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1374     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1375     "psraw                $7, %%mm1     \n\t" \
1376     "psraw                $7, %%mm7     \n\t"
1377 #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)
1378 
RENAME(yuv2yuyv422_1)1379 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1380                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
1381                                   const int16_t *abuf0, uint8_t *dest,
1382                                   int dstW, int uvalpha, int y)
1383 {
1384     const int16_t *ubuf0 = ubuf[0];
1385     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1386 
1387     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1388         const int16_t *ubuf1 = ubuf[0];
1389         __asm__ volatile(
1390             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1391             "mov           %4, %%"FF_REG_b"         \n\t"
1392             "push %%"FF_REG_BP"                     \n\t"
1393             YSCALEYUV2PACKED1(%%FF_REGBP, %5)
1394             WRITEYUY2(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1395             "pop %%"FF_REG_BP"                      \n\t"
1396             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1397             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1398                "a" (&c->redDither)
1399         );
1400     } else {
1401         const int16_t *ubuf1 = ubuf[1];
1402         __asm__ volatile(
1403             "mov %%"FF_REG_b", "ESP_OFFSET"(%5)     \n\t"
1404             "mov           %4, %%"FF_REG_b"         \n\t"
1405             "push %%"FF_REG_BP"                     \n\t"
1406             YSCALEYUV2PACKED1b(%%FF_REGBP, %5)
1407             WRITEYUY2(%%FF_REGb, DSTW_OFFSET"(%5)", %%FF_REGBP)
1408             "pop %%"FF_REG_BP"                      \n\t"
1409             "mov "ESP_OFFSET"(%5), %%"FF_REG_b"     \n\t"
1410             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1411                "a" (&c->redDither)
1412         );
1413     }
1414 }
RENAME(sws_init_swscale)1415 static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
1416 {
1417     enum AVPixelFormat dstFormat = c->dstFormat;
1418 
1419     c->use_mmx_vfilter= 0;
1420     if (!is16BPS(dstFormat) && !isNBPS(dstFormat) && !isSemiPlanarYUV(dstFormat)
1421         && dstFormat != AV_PIX_FMT_GRAYF32BE && dstFormat != AV_PIX_FMT_GRAYF32LE
1422         && !(c->flags & SWS_BITEXACT)) {
1423             if (c->flags & SWS_ACCURATE_RND) {
1424                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1425                     switch (c->dstFormat) {
1426                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;
1427 #if HAVE_6REGS
1428                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;
1429 #endif
1430                     case AV_PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;
1431                     case AV_PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;
1432                     case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1433                     default: break;
1434                     }
1435                 }
1436             } else {
1437                 c->use_mmx_vfilter= 1;
1438                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1439                     switch (c->dstFormat) {
1440                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;
1441                     case AV_PIX_FMT_BGR32:   c->yuv2packedX = RENAME(yuv2bgr32_X);   break;
1442 #if HAVE_6REGS
1443                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;
1444 #endif
1445                     case AV_PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;
1446                     case AV_PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;
1447                     case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1448                     default: break;
1449                     }
1450                 }
1451             }
1452         if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1453             switch (c->dstFormat) {
1454             case AV_PIX_FMT_RGB32:
1455                 c->yuv2packed1 = RENAME(yuv2rgb32_1);
1456                 c->yuv2packed2 = RENAME(yuv2rgb32_2);
1457                 break;
1458             case AV_PIX_FMT_BGR24:
1459                 c->yuv2packed1 = RENAME(yuv2bgr24_1);
1460                 c->yuv2packed2 = RENAME(yuv2bgr24_2);
1461                 break;
1462             case AV_PIX_FMT_RGB555:
1463                 c->yuv2packed1 = RENAME(yuv2rgb555_1);
1464                 c->yuv2packed2 = RENAME(yuv2rgb555_2);
1465                 break;
1466             case AV_PIX_FMT_RGB565:
1467                 c->yuv2packed1 = RENAME(yuv2rgb565_1);
1468                 c->yuv2packed2 = RENAME(yuv2rgb565_2);
1469                 break;
1470             case AV_PIX_FMT_YUYV422:
1471                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1472                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1473                 break;
1474             default:
1475                 break;
1476             }
1477         }
1478     }
1479 
1480     if (c->srcBpc == 8 && c->dstBpc <= 14) {
1481     // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1482 #if COMPILE_TEMPLATE_MMXEXT
1483     if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1484         c->hyscale_fast = ff_hyscale_fast_mmxext;
1485         c->hcscale_fast = ff_hcscale_fast_mmxext;
1486     } else {
1487 #endif /* COMPILE_TEMPLATE_MMXEXT */
1488         c->hyscale_fast = NULL;
1489         c->hcscale_fast = NULL;
1490 #if COMPILE_TEMPLATE_MMXEXT
1491     }
1492 #endif /* COMPILE_TEMPLATE_MMXEXT */
1493     }
1494 }
1495