1/*
2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * This file is part of Libav.
5 *
6 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21#include <inttypes.h>
22#include "config.h"
23#include "libswscale/swscale.h"
24#include "libswscale/swscale_internal.h"
25#include "libavutil/intreadwrite.h"
26#include "libavutil/x86_cpu.h"
27#include "libavutil/cpu.h"
28#include "libavutil/pixdesc.h"
29
30DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
31DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
32DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
33DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
34DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
35DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
36DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
37DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
38
39const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
40    0x0103010301030103LL,
41    0x0200020002000200LL,};
42
43const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
44    0x0602060206020602LL,
45    0x0004000400040004LL,};
46
47DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
48DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
49DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
50DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
51DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
52DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
53
54DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
55DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
56DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
57
58#ifdef FAST_BGR2YV12
59DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
60DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
61DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
62#else
63DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
64DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
65DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
66#endif /* FAST_BGR2YV12 */
67DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
68DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
69DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
70
71DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
72DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
73DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
74DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
75DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
76
77DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
78    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
79    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
80};
81
82DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
83
84//MMX versions
85#if HAVE_MMX
86#undef RENAME
87#define COMPILE_TEMPLATE_MMX2 0
88#define RENAME(a) a ## _MMX
89#include "swscale_template.c"
90#endif
91
92//MMX2 versions
93#if HAVE_MMX2
94#undef RENAME
95#undef COMPILE_TEMPLATE_MMX2
96#define COMPILE_TEMPLATE_MMX2 1
97#define RENAME(a) a ## _MMX2
98#include "swscale_template.c"
99#endif
100
101void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex,
102                           int lastInLumBuf, int lastInChrBuf)
103{
104    const int dstH= c->dstH;
105    const int flags= c->flags;
106    int16_t **lumPixBuf= c->lumPixBuf;
107    int16_t **chrUPixBuf= c->chrUPixBuf;
108    int16_t **alpPixBuf= c->alpPixBuf;
109    const int vLumBufSize= c->vLumBufSize;
110    const int vChrBufSize= c->vChrBufSize;
111    int32_t *vLumFilterPos= c->vLumFilterPos;
112    int32_t *vChrFilterPos= c->vChrFilterPos;
113    int16_t *vLumFilter= c->vLumFilter;
114    int16_t *vChrFilter= c->vChrFilter;
115    int32_t *lumMmxFilter= c->lumMmxFilter;
116    int32_t *chrMmxFilter= c->chrMmxFilter;
117    int32_t av_unused *alpMmxFilter= c->alpMmxFilter;
118    const int vLumFilterSize= c->vLumFilterSize;
119    const int vChrFilterSize= c->vChrFilterSize;
120    const int chrDstY= dstY>>c->chrDstVSubSample;
121    const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input
122    const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input
123
124    c->blueDither= ff_dither8[dstY&1];
125    if (c->dstFormat == PIX_FMT_RGB555 || c->dstFormat == PIX_FMT_BGR555)
126        c->greenDither= ff_dither8[dstY&1];
127    else
128        c->greenDither= ff_dither4[dstY&1];
129    c->redDither= ff_dither8[(dstY+1)&1];
130    if (dstY < dstH - 2) {
131        const int16_t **lumSrcPtr= (const int16_t **) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
132        const int16_t **chrUSrcPtr= (const int16_t **) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
133        const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
134        int i;
135
136        if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->srcH) {
137            const int16_t **tmpY = (const int16_t **) lumPixBuf + 2 * vLumBufSize;
138            int neg = -firstLumSrcY, i, end = FFMIN(c->srcH - firstLumSrcY, vLumFilterSize);
139            for (i = 0; i < neg;            i++)
140                tmpY[i] = lumSrcPtr[neg];
141            for (     ; i < end;            i++)
142                tmpY[i] = lumSrcPtr[i];
143            for (     ; i < vLumFilterSize; i++)
144                tmpY[i] = tmpY[i-1];
145            lumSrcPtr = tmpY;
146
147            if (alpSrcPtr) {
148                const int16_t **tmpA = (const int16_t **) alpPixBuf + 2 * vLumBufSize;
149                for (i = 0; i < neg;            i++)
150                    tmpA[i] = alpSrcPtr[neg];
151                for (     ; i < end;            i++)
152                    tmpA[i] = alpSrcPtr[i];
153                for (     ; i < vLumFilterSize; i++)
154                    tmpA[i] = tmpA[i - 1];
155                alpSrcPtr = tmpA;
156            }
157        }
158        if (firstChrSrcY < 0 || firstChrSrcY + vChrFilterSize > c->chrSrcH) {
159            const int16_t **tmpU = (const int16_t **) chrUPixBuf + 2 * vChrBufSize;
160            int neg = -firstChrSrcY, i, end = FFMIN(c->chrSrcH - firstChrSrcY, vChrFilterSize);
161            for (i = 0; i < neg;            i++) {
162                tmpU[i] = chrUSrcPtr[neg];
163            }
164            for (     ; i < end;            i++) {
165                tmpU[i] = chrUSrcPtr[i];
166            }
167            for (     ; i < vChrFilterSize; i++) {
168                tmpU[i] = tmpU[i - 1];
169            }
170            chrUSrcPtr = tmpU;
171        }
172
173        if (flags & SWS_ACCURATE_RND) {
174            int s= APCK_SIZE / 8;
175            for (i=0; i<vLumFilterSize; i+=2) {
176                *(const void**)&lumMmxFilter[s*i              ]= lumSrcPtr[i  ];
177                *(const void**)&lumMmxFilter[s*i+APCK_PTR2/4  ]= lumSrcPtr[i+(vLumFilterSize>1)];
178                lumMmxFilter[s*i+APCK_COEF/4  ]=
179                lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i    ]
180                + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);
181                if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
182                    *(const void**)&alpMmxFilter[s*i              ]= alpSrcPtr[i  ];
183                    *(const void**)&alpMmxFilter[s*i+APCK_PTR2/4  ]= alpSrcPtr[i+(vLumFilterSize>1)];
184                    alpMmxFilter[s*i+APCK_COEF/4  ]=
185                    alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4  ];
186                }
187            }
188            for (i=0; i<vChrFilterSize; i+=2) {
189                *(const void**)&chrMmxFilter[s*i              ]= chrUSrcPtr[i  ];
190                *(const void**)&chrMmxFilter[s*i+APCK_PTR2/4  ]= chrUSrcPtr[i+(vChrFilterSize>1)];
191                chrMmxFilter[s*i+APCK_COEF/4  ]=
192                chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i    ]
193                + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);
194            }
195        } else {
196            for (i=0; i<vLumFilterSize; i++) {
197                *(const void**)&lumMmxFilter[4*i+0]= lumSrcPtr[i];
198                lumMmxFilter[4*i+2]=
199                lumMmxFilter[4*i+3]=
200                ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;
201                if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
202                    *(const void**)&alpMmxFilter[4*i+0]= alpSrcPtr[i];
203                    alpMmxFilter[4*i+2]=
204                    alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2];
205                }
206            }
207            for (i=0; i<vChrFilterSize; i++) {
208                *(const void**)&chrMmxFilter[4*i+0]= chrUSrcPtr[i];
209                chrMmxFilter[4*i+2]=
210                chrMmxFilter[4*i+3]=
211                ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;
212            }
213        }
214    }
215}
216
217#define SCALE_FUNC(filter_n, from_bpc, to_bpc, opt) \
218extern void ff_hscale ## from_bpc ## to ## to_bpc ## _ ## filter_n ## _ ## opt( \
219                                                SwsContext *c, int16_t *data, \
220                                                int dstW, const uint8_t *src, \
221                                                const int16_t *filter, \
222                                                const int32_t *filterPos, int filterSize)
223
224#define SCALE_FUNCS(filter_n, opt) \
225    SCALE_FUNC(filter_n,  8, 15, opt); \
226    SCALE_FUNC(filter_n,  9, 15, opt); \
227    SCALE_FUNC(filter_n, 10, 15, opt); \
228    SCALE_FUNC(filter_n, 16, 15, opt); \
229    SCALE_FUNC(filter_n,  8, 19, opt); \
230    SCALE_FUNC(filter_n,  9, 19, opt); \
231    SCALE_FUNC(filter_n, 10, 19, opt); \
232    SCALE_FUNC(filter_n, 16, 19, opt)
233
234#define SCALE_FUNCS_MMX(opt) \
235    SCALE_FUNCS(4, opt); \
236    SCALE_FUNCS(8, opt); \
237    SCALE_FUNCS(X, opt)
238
239#define SCALE_FUNCS_SSE(opt) \
240    SCALE_FUNCS(4, opt); \
241    SCALE_FUNCS(8, opt); \
242    SCALE_FUNCS(X4, opt); \
243    SCALE_FUNCS(X8, opt)
244
245#if ARCH_X86_32
246SCALE_FUNCS_MMX(mmx);
247#endif
248SCALE_FUNCS_SSE(sse2);
249SCALE_FUNCS_SSE(ssse3);
250SCALE_FUNCS_SSE(sse4);
251
252#define VSCALEX_FUNC(size, opt) \
253extern void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filterSize, \
254                                               const int16_t **src, uint8_t *dest, int dstW, \
255                                               const uint8_t *dither, int offset)
256#define VSCALEX_FUNCS(opt) \
257    VSCALEX_FUNC(8,  opt); \
258    VSCALEX_FUNC(9,  opt); \
259    VSCALEX_FUNC(10, opt)
260
261#if ARCH_X86_32
262VSCALEX_FUNCS(mmx2);
263#endif
264VSCALEX_FUNCS(sse2);
265VSCALEX_FUNCS(sse4);
266VSCALEX_FUNC(16, sse4);
267VSCALEX_FUNCS(avx);
268
269#define VSCALE_FUNC(size, opt) \
270extern void ff_yuv2plane1_ ## size ## _ ## opt(const int16_t *src, uint8_t *dst, int dstW, \
271                                               const uint8_t *dither, int offset)
272#define VSCALE_FUNCS(opt1, opt2) \
273    VSCALE_FUNC(8,  opt1); \
274    VSCALE_FUNC(9,  opt2); \
275    VSCALE_FUNC(10, opt2); \
276    VSCALE_FUNC(16, opt1)
277
278#if ARCH_X86_32
279VSCALE_FUNCS(mmx, mmx2);
280#endif
281VSCALE_FUNCS(sse2, sse2);
282VSCALE_FUNC(16, sse4);
283VSCALE_FUNCS(avx, avx);
284
285#define INPUT_UV_FUNC(fmt, opt) \
286extern void ff_ ## fmt ## ToUV_ ## opt(uint8_t *dstU, uint8_t *dstV, \
287                                       const uint8_t *src, const uint8_t *unused1, \
288                                       int w, uint32_t *unused2)
289#define INPUT_FUNC(fmt, opt) \
290extern void ff_ ## fmt ## ToY_  ## opt(uint8_t *dst, const uint8_t *src, \
291                                       int w, uint32_t *unused); \
292    INPUT_UV_FUNC(fmt, opt)
293#define INPUT_FUNCS(opt) \
294    INPUT_FUNC(uyvy, opt); \
295    INPUT_FUNC(yuyv, opt); \
296    INPUT_UV_FUNC(nv12, opt); \
297    INPUT_UV_FUNC(nv21, opt)
298
299#if ARCH_X86_32
300INPUT_FUNCS(mmx);
301#endif
302INPUT_FUNCS(sse2);
303INPUT_FUNCS(avx);
304
305void ff_sws_init_swScale_mmx(SwsContext *c)
306{
307    int cpu_flags = av_get_cpu_flags();
308
309    if (cpu_flags & AV_CPU_FLAG_MMX)
310        sws_init_swScale_MMX(c);
311#if HAVE_MMX2
312    if (cpu_flags & AV_CPU_FLAG_MMX2)
313        sws_init_swScale_MMX2(c);
314#endif
315
316#if HAVE_YASM
317#define ASSIGN_SCALE_FUNC2(hscalefn, filtersize, opt1, opt2) do { \
318    if (c->srcBpc == 8) { \
319        hscalefn = c->dstBpc <= 10 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \
320                                     ff_hscale8to19_ ## filtersize ## _ ## opt1; \
321    } else if (c->srcBpc == 9) { \
322        hscalefn = c->dstBpc <= 10 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \
323                                     ff_hscale9to19_ ## filtersize ## _ ## opt1; \
324    } else if (c->srcBpc == 10) { \
325        hscalefn = c->dstBpc <= 10 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \
326                                     ff_hscale10to19_ ## filtersize ## _ ## opt1; \
327    } else /* c->srcBpc == 16 */ { \
328        hscalefn = c->dstBpc <= 10 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \
329                                     ff_hscale16to19_ ## filtersize ## _ ## opt1; \
330    } \
331} while (0)
332#define ASSIGN_MMX_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
333    switch (filtersize) { \
334    case 4:  ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
335    case 8:  ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
336    default: ASSIGN_SCALE_FUNC2(hscalefn, X, opt1, opt2); break; \
337    }
338#define ASSIGN_VSCALEX_FUNC(vscalefn, opt, do_16_case) \
339switch(c->dstBpc){ \
340    case 16:                          do_16_case;                          break; \
341    case 10: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_10_ ## opt; break; \
342    case 9:  if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_9_  ## opt; break; \
343    default:                          vscalefn = ff_yuv2planeX_8_  ## opt; break; \
344    }
345#define ASSIGN_VSCALE_FUNC(vscalefn, opt1, opt2, opt2chk) \
346    switch(c->dstBpc){ \
347    case 16: if (!isBE(c->dstFormat))            vscalefn = ff_yuv2plane1_16_ ## opt1; break; \
348    case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_10_ ## opt2; break; \
349    case 9:  if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_9_  ## opt2;  break; \
350    default:                                     vscalefn = ff_yuv2plane1_8_  ## opt1;  break; \
351    }
352#if ARCH_X86_32
353    if (cpu_flags & AV_CPU_FLAG_MMX) {
354        ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
355        ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
356        ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2);
357
358        switch (c->srcFormat) {
359        case PIX_FMT_Y400A:
360            c->lumToYV12 = ff_yuyvToY_mmx;
361            if (c->alpPixBuf)
362                c->alpToYV12 = ff_uyvyToY_mmx;
363            break;
364        case PIX_FMT_YUYV422:
365            c->lumToYV12 = ff_yuyvToY_mmx;
366            c->chrToYV12 = ff_yuyvToUV_mmx;
367            break;
368        case PIX_FMT_UYVY422:
369            c->lumToYV12 = ff_uyvyToY_mmx;
370            c->chrToYV12 = ff_uyvyToUV_mmx;
371            break;
372        case PIX_FMT_NV12:
373            c->chrToYV12 = ff_nv12ToUV_mmx;
374            break;
375        case PIX_FMT_NV21:
376            c->chrToYV12 = ff_nv21ToUV_mmx;
377            break;
378        default:
379            break;
380        }
381    }
382    if (cpu_flags & AV_CPU_FLAG_MMX2) {
383        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx2,);
384    }
385#endif
386#define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \
387    switch (filtersize) { \
388    case 4:  ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \
389    case 8:  ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \
390    default: if (filtersize & 4) ASSIGN_SCALE_FUNC2(hscalefn, X4, opt1, opt2); \
391             else                ASSIGN_SCALE_FUNC2(hscalefn, X8, opt1, opt2); \
392             break; \
393    }
394    if (cpu_flags & AV_CPU_FLAG_SSE2) {
395        ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse2, sse2);
396        ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2);
397        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2,);
398        ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2, sse2, 1);
399
400        switch (c->srcFormat) {
401        case PIX_FMT_Y400A:
402            c->lumToYV12 = ff_yuyvToY_sse2;
403            if (c->alpPixBuf)
404                c->alpToYV12 = ff_uyvyToY_sse2;
405            break;
406        case PIX_FMT_YUYV422:
407            c->lumToYV12 = ff_yuyvToY_sse2;
408            c->chrToYV12 = ff_yuyvToUV_sse2;
409            break;
410        case PIX_FMT_UYVY422:
411            c->lumToYV12 = ff_uyvyToY_sse2;
412            c->chrToYV12 = ff_uyvyToUV_sse2;
413            break;
414        case PIX_FMT_NV12:
415            c->chrToYV12 = ff_nv12ToUV_sse2;
416            break;
417        case PIX_FMT_NV21:
418            c->chrToYV12 = ff_nv21ToUV_sse2;
419            break;
420        }
421    }
422    if (cpu_flags & AV_CPU_FLAG_SSSE3) {
423        ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3);
424        ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, ssse3, ssse3);
425    }
426    if (cpu_flags & AV_CPU_FLAG_SSE4) {
427        /* Xto15 don't need special sse4 functions */
428        ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse4, ssse3);
429        ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse4, ssse3);
430        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse4,
431                            if (!isBE(c->dstFormat)) c->yuv2planeX = ff_yuv2planeX_16_sse4);
432        if (c->dstBpc == 16 && !isBE(c->dstFormat))
433            c->yuv2plane1 = ff_yuv2plane1_16_sse4;
434    }
435
436    if (cpu_flags & AV_CPU_FLAG_AVX) {
437        ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx,);
438        ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx, avx, 1);
439
440        switch (c->srcFormat) {
441        case PIX_FMT_YUYV422:
442            c->chrToYV12 = ff_yuyvToUV_avx;
443            break;
444        case PIX_FMT_UYVY422:
445            c->chrToYV12 = ff_uyvyToUV_avx;
446            break;
447        case PIX_FMT_NV12:
448            c->chrToYV12 = ff_nv12ToUV_avx;
449            break;
450        case PIX_FMT_NV21:
451            c->chrToYV12 = ff_nv21ToUV_avx;
452            break;
453        default:
454            break;
455        }
456    }
457#endif
458}
459