2 * Copyright (c) 2013 Seppo Tomperi
3 * Copyright (c) 2013 - 2014 Pierre-Edouard Lepere
6 * This file is part of ffmpeg.
8 * ffmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * ffmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with ffmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #include "libavutil/cpu.h"
25 #include "libavutil/x86/asm.h"
26 #include "libavutil/x86/cpu.h"
27 #include "libavcodec/get_bits.h" /* required for hevcdsp.h GetBitContext */
28 #include "libavcodec/hevcdsp.h"
29 #include "libavcodec/x86/hevcdsp.h"
32 #define LFC_FUNC(DIR, DEPTH, OPT) \
33 void ff_hevc_ ## DIR ## _loop_filter_chroma_ ## DEPTH ## _ ## OPT(uint8_t *_pix, ptrdiff_t _stride, int *_tc, uint8_t *_no_p, uint8_t *_no_q);
35 #define LFL_FUNC(DIR, DEPTH, OPT) \
36 void ff_hevc_ ## DIR ## _loop_filter_luma_ ## DEPTH ## _ ## OPT(uint8_t *_pix, ptrdiff_t stride, int *_beta, int *_tc, \
37 uint8_t *_no_p, uint8_t *_no_q);
39 #define LFC_FUNCS(type, depth, opt) \
40 LFC_FUNC(h, depth, opt) \
41 LFC_FUNC(v, depth, opt)
43 #define LFL_FUNCS(type, depth, opt) \
44 LFL_FUNC(h, depth, opt) \
45 LFL_FUNC(v, depth, opt)
47 LFC_FUNCS(uint8_t, 8, sse2)
48 LFC_FUNCS(uint8_t, 10, sse2)
49 LFL_FUNCS(uint8_t, 8, sse2)
50 LFL_FUNCS(uint8_t, 10, sse2)
51 LFL_FUNCS(uint8_t, 8, ssse3)
52 LFL_FUNCS(uint8_t, 10, ssse3)
54 #if HAVE_SSE2_EXTERNAL
55 void ff_hevc_idct32_dc_add_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
57 ff_hevc_idct16_dc_add_8_sse2(dst, coeffs, stride);
58 ff_hevc_idct16_dc_add_8_sse2(dst+16, coeffs, stride);
59 ff_hevc_idct16_dc_add_8_sse2(dst+16*stride, coeffs, stride);
60 ff_hevc_idct16_dc_add_8_sse2(dst+16*stride+16, coeffs, stride);
63 void ff_hevc_idct16_dc_add_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
65 ff_hevc_idct8_dc_add_10_sse2(dst, coeffs, stride);
66 ff_hevc_idct8_dc_add_10_sse2(dst+16, coeffs, stride);
67 ff_hevc_idct8_dc_add_10_sse2(dst+8*stride, coeffs, stride);
68 ff_hevc_idct8_dc_add_10_sse2(dst+8*stride+16, coeffs, stride);
71 void ff_hevc_idct32_dc_add_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
73 ff_hevc_idct16_dc_add_10_sse2(dst, coeffs, stride);
74 ff_hevc_idct16_dc_add_10_sse2(dst+32, coeffs, stride);
75 ff_hevc_idct16_dc_add_10_sse2(dst+16*stride, coeffs, stride);
76 ff_hevc_idct16_dc_add_10_sse2(dst+16*stride+32, coeffs, stride);
78 #endif //HAVE_SSE2_EXTERNAL
80 void ff_hevc_idct16_dc_add_10_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
82 ff_hevc_idct8_dc_add_10_avx(dst, coeffs, stride);
83 ff_hevc_idct8_dc_add_10_avx(dst+16, coeffs, stride);
84 ff_hevc_idct8_dc_add_10_avx(dst+8*stride, coeffs, stride);
85 ff_hevc_idct8_dc_add_10_avx(dst+8*stride+16, coeffs, stride);
88 void ff_hevc_idct32_dc_add_10_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
90 ff_hevc_idct16_dc_add_10_avx(dst, coeffs, stride);
91 ff_hevc_idct16_dc_add_10_avx(dst+32, coeffs, stride);
92 ff_hevc_idct16_dc_add_10_avx(dst+16*stride, coeffs, stride);
93 ff_hevc_idct16_dc_add_10_avx(dst+16*stride+32, coeffs, stride);
95 #endif //HAVE_AVX_EXTERNAL
97 #if HAVE_AVX2_EXTERNAL
99 void ff_hevc_idct32_dc_add_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
101 ff_hevc_idct16_dc_add_10_avx2(dst, coeffs, stride);
102 ff_hevc_idct16_dc_add_10_avx2(dst+32, coeffs, stride);
103 ff_hevc_idct16_dc_add_10_avx2(dst+16*stride, coeffs, stride);
104 ff_hevc_idct16_dc_add_10_avx2(dst+16*stride+32, coeffs, stride);
106 #endif //HAVE_AVX2_EXTERNAL
108 #define mc_rep_func(name, bitd, step, W, opt) \
109 void ff_hevc_put_hevc_##name##W##_##bitd##_##opt(int16_t *_dst, ptrdiff_t dststride, \
110 uint8_t *_src, ptrdiff_t _srcstride, int height, \
111 intptr_t mx, intptr_t my, int width) \
116 for (i = 0; i < W; i += step) { \
117 src = _src + (i * ((bitd + 7) / 8)); \
119 ff_hevc_put_hevc_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride, height, mx, my, width); \
122 #define mc_rep_uni_func(name, bitd, step, W, opt) \
123 void ff_hevc_put_hevc_uni_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, \
124 uint8_t *_src, ptrdiff_t _srcstride, int height, \
125 intptr_t mx, intptr_t my, int width) \
130 for (i = 0; i < W; i += step) { \
131 src = _src + (i * ((bitd + 7) / 8)); \
132 dst = _dst + (i * ((bitd + 7) / 8)); \
133 ff_hevc_put_hevc_uni_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride, \
134 height, mx, my, width); \
137 #define mc_rep_bi_func(name, bitd, step, W, opt) \
138 void ff_hevc_put_hevc_bi_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, uint8_t *_src, \
139 ptrdiff_t _srcstride, int16_t* _src2, ptrdiff_t _src2stride, \
140 int height, intptr_t mx, intptr_t my, int width) \
146 for (i = 0; i < W ; i += step) { \
147 src = _src + (i * ((bitd + 7) / 8)); \
148 dst = _dst + (i * ((bitd + 7) / 8)); \
150 ff_hevc_put_hevc_bi_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride, src2, \
151 _src2stride, height, mx, my, width); \
155 #define mc_rep_funcs(name, bitd, step, W, opt) \
156 mc_rep_func(name, bitd, step, W, opt); \
157 mc_rep_uni_func(name, bitd, step, W, opt); \
158 mc_rep_bi_func(name, bitd, step, W, opt)
161 #if ARCH_X86_64 && HAVE_SSE4_EXTERNAL
163 mc_rep_funcs(pel_pixels, 8, 16, 64, sse4);
164 mc_rep_funcs(pel_pixels, 8, 16, 48, sse4);
165 mc_rep_funcs(pel_pixels, 8, 16, 32, sse4);
166 mc_rep_funcs(pel_pixels, 8, 8, 24, sse4);
168 mc_rep_funcs(pel_pixels,10, 8, 64, sse4);
169 mc_rep_funcs(pel_pixels,10, 8, 48, sse4);
170 mc_rep_funcs(pel_pixels,10, 8, 32, sse4);
171 mc_rep_funcs(pel_pixels,10, 8, 24, sse4);
172 mc_rep_funcs(pel_pixels,10, 8, 16, sse4);
173 mc_rep_funcs(pel_pixels,10, 4, 12, sse4);
175 mc_rep_funcs(epel_h, 8, 16, 64, sse4);
176 mc_rep_funcs(epel_h, 8, 16, 48, sse4);
177 mc_rep_funcs(epel_h, 8, 16, 32, sse4);
178 mc_rep_funcs(epel_h, 8, 8, 24, sse4);
179 mc_rep_funcs(epel_h,10, 8, 64, sse4);
180 mc_rep_funcs(epel_h,10, 8, 48, sse4);
181 mc_rep_funcs(epel_h,10, 8, 32, sse4);
182 mc_rep_funcs(epel_h,10, 8, 24, sse4);
183 mc_rep_funcs(epel_h,10, 8, 16, sse4);
184 mc_rep_funcs(epel_h,10, 4, 12, sse4);
185 mc_rep_funcs(epel_v, 8, 16, 64, sse4);
186 mc_rep_funcs(epel_v, 8, 16, 48, sse4);
187 mc_rep_funcs(epel_v, 8, 16, 32, sse4);
188 mc_rep_funcs(epel_v, 8, 8, 24, sse4);
189 mc_rep_funcs(epel_v,10, 8, 64, sse4);
190 mc_rep_funcs(epel_v,10, 8, 48, sse4);
191 mc_rep_funcs(epel_v,10, 8, 32, sse4);
192 mc_rep_funcs(epel_v,10, 8, 24, sse4);
193 mc_rep_funcs(epel_v,10, 8, 16, sse4);
194 mc_rep_funcs(epel_v,10, 4, 12, sse4);
195 mc_rep_funcs(epel_hv, 8, 8, 64, sse4);
196 mc_rep_funcs(epel_hv, 8, 8, 48, sse4);
197 mc_rep_funcs(epel_hv, 8, 8, 32, sse4);
198 mc_rep_funcs(epel_hv, 8, 8, 24, sse4);
199 mc_rep_funcs(epel_hv, 8, 8, 16, sse4);
200 mc_rep_funcs(epel_hv, 8, 4, 12, sse4);
201 mc_rep_funcs(epel_hv,10, 8, 64, sse4);
202 mc_rep_funcs(epel_hv,10, 8, 48, sse4);
203 mc_rep_funcs(epel_hv,10, 8, 32, sse4);
204 mc_rep_funcs(epel_hv,10, 8, 24, sse4);
205 mc_rep_funcs(epel_hv,10, 8, 16, sse4);
206 mc_rep_funcs(epel_hv,10, 4, 12, sse4);
208 mc_rep_funcs(qpel_h, 8, 16, 64, sse4);
209 mc_rep_funcs(qpel_h, 8, 16, 48, sse4);
210 mc_rep_funcs(qpel_h, 8, 16, 32, sse4);
211 mc_rep_funcs(qpel_h, 8, 8, 24, sse4);
212 mc_rep_funcs(qpel_h,10, 8, 64, sse4);
213 mc_rep_funcs(qpel_h,10, 8, 48, sse4);
214 mc_rep_funcs(qpel_h,10, 8, 32, sse4);
215 mc_rep_funcs(qpel_h,10, 8, 24, sse4);
216 mc_rep_funcs(qpel_h,10, 8, 16, sse4);
217 mc_rep_funcs(qpel_h,10, 4, 12, sse4);
218 mc_rep_funcs(qpel_v, 8, 16, 64, sse4);
219 mc_rep_funcs(qpel_v, 8, 16, 48, sse4);
220 mc_rep_funcs(qpel_v, 8, 16, 32, sse4);
221 mc_rep_funcs(qpel_v, 8, 8, 24, sse4);
222 mc_rep_funcs(qpel_v,10, 8, 64, sse4);
223 mc_rep_funcs(qpel_v,10, 8, 48, sse4);
224 mc_rep_funcs(qpel_v,10, 8, 32, sse4);
225 mc_rep_funcs(qpel_v,10, 8, 24, sse4);
226 mc_rep_funcs(qpel_v,10, 8, 16, sse4);
227 mc_rep_funcs(qpel_v,10, 4, 12, sse4);
228 mc_rep_funcs(qpel_hv, 8, 8, 64, sse4);
229 mc_rep_funcs(qpel_hv, 8, 8, 48, sse4);
230 mc_rep_funcs(qpel_hv, 8, 8, 32, sse4);
231 mc_rep_funcs(qpel_hv, 8, 8, 24, sse4);
232 mc_rep_funcs(qpel_hv, 8, 8, 16, sse4);
233 mc_rep_funcs(qpel_hv, 8, 4, 12, sse4);
234 mc_rep_funcs(qpel_hv,10, 8, 64, sse4);
235 mc_rep_funcs(qpel_hv,10, 8, 48, sse4);
236 mc_rep_funcs(qpel_hv,10, 8, 32, sse4);
237 mc_rep_funcs(qpel_hv,10, 8, 24, sse4);
238 mc_rep_funcs(qpel_hv,10, 8, 16, sse4);
239 mc_rep_funcs(qpel_hv,10, 4, 12, sse4);
241 #define mc_rep_uni_w(bitd, step, W, opt) \
242 void ff_hevc_put_hevc_uni_w##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, int16_t *_src, ptrdiff_t _srcstride,\
243 int height, int denom, int _wx, int _ox) \
248 for (i = 0; i < W; i += step) { \
250 dst= _dst + (i * ((bitd + 7) / 8)); \
251 ff_hevc_put_hevc_uni_w##step##_##bitd##_##opt(dst, dststride, src, _srcstride, \
252 height, denom, _wx, _ox); \
256 mc_rep_uni_w(8, 6, 12, sse4);
257 mc_rep_uni_w(8, 8, 16, sse4);
258 mc_rep_uni_w(8, 8, 24, sse4);
259 mc_rep_uni_w(8, 8, 32, sse4);
260 mc_rep_uni_w(8, 8, 48, sse4);
261 mc_rep_uni_w(8, 8, 64, sse4);
263 mc_rep_uni_w(10, 6, 12, sse4);
264 mc_rep_uni_w(10, 8, 16, sse4);
265 mc_rep_uni_w(10, 8, 24, sse4);
266 mc_rep_uni_w(10, 8, 32, sse4);
267 mc_rep_uni_w(10, 8, 48, sse4);
268 mc_rep_uni_w(10, 8, 64, sse4);
270 #define mc_rep_bi_w(bitd, step, W, opt) \
271 void ff_hevc_put_hevc_bi_w##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, int16_t *_src, ptrdiff_t _srcstride, \
272 int16_t *_src2, ptrdiff_t _src2stride, int height, \
273 int denom, int _wx0, int _wx1, int _ox0, int _ox1) \
279 for (i = 0; i < W; i += step) { \
282 dst = _dst + (i * ((bitd + 7) / 8)); \
283 ff_hevc_put_hevc_bi_w##step##_##bitd##_##opt(dst, dststride, src, _srcstride, src2, _src2stride, \
284 height, denom, _wx0, _wx1, _ox0, _ox1); \
288 mc_rep_bi_w(8, 6, 12, sse4);
289 mc_rep_bi_w(8, 8, 16, sse4);
290 mc_rep_bi_w(8, 8, 24, sse4);
291 mc_rep_bi_w(8, 8, 32, sse4);
292 mc_rep_bi_w(8, 8, 48, sse4);
293 mc_rep_bi_w(8, 8, 64, sse4);
295 mc_rep_bi_w(10, 6, 12, sse4);
296 mc_rep_bi_w(10, 8, 16, sse4);
297 mc_rep_bi_w(10, 8, 24, sse4);
298 mc_rep_bi_w(10, 8, 32, sse4);
299 mc_rep_bi_w(10, 8, 48, sse4);
300 mc_rep_bi_w(10, 8, 64, sse4);
302 #define mc_uni_w_func(name, bitd, W, opt) \
303 void ff_hevc_put_hevc_uni_w_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t _dststride, \
304 uint8_t *_src, ptrdiff_t _srcstride, \
305 int height, int denom, \
307 intptr_t mx, intptr_t my, int width) \
309 LOCAL_ALIGNED_16(int16_t, temp, [71 * 64]); \
310 ff_hevc_put_hevc_##name##W##_##bitd##_##opt(temp, 64, _src, _srcstride, height, mx, my, width); \
311 ff_hevc_put_hevc_uni_w##W##_##bitd##_##opt(_dst, _dststride, temp, 64, height, denom, _wx, _ox);\
314 #define mc_uni_w_funcs(name, bitd, opt) \
315 mc_uni_w_func(name, bitd, 4, opt); \
316 mc_uni_w_func(name, bitd, 8, opt); \
317 mc_uni_w_func(name, bitd, 12, opt); \
318 mc_uni_w_func(name, bitd, 16, opt); \
319 mc_uni_w_func(name, bitd, 24, opt); \
320 mc_uni_w_func(name, bitd, 32, opt); \
321 mc_uni_w_func(name, bitd, 48, opt); \
322 mc_uni_w_func(name, bitd, 64, opt)
324 mc_uni_w_funcs(pel_pixels, 8, sse4);
325 mc_uni_w_func(pel_pixels, 8, 6, sse4);
326 mc_uni_w_funcs(epel_h, 8, sse4);
327 mc_uni_w_func(epel_h, 8, 6, sse4);
328 mc_uni_w_funcs(epel_v, 8, sse4);
329 mc_uni_w_func(epel_v, 8, 6, sse4);
330 mc_uni_w_funcs(epel_hv, 8, sse4);
331 mc_uni_w_func(epel_hv, 8, 6, sse4);
332 mc_uni_w_funcs(qpel_h, 8, sse4);
333 mc_uni_w_funcs(qpel_v, 8, sse4);
334 mc_uni_w_funcs(qpel_hv, 8, sse4);
336 mc_uni_w_funcs(pel_pixels, 10, sse4);
337 mc_uni_w_func(pel_pixels, 10, 6, sse4);
338 mc_uni_w_funcs(epel_h, 10, sse4);
339 mc_uni_w_func(epel_h, 10, 6, sse4);
340 mc_uni_w_funcs(epel_v, 10, sse4);
341 mc_uni_w_func(epel_v, 10, 6, sse4);
342 mc_uni_w_funcs(epel_hv, 10, sse4);
343 mc_uni_w_func(epel_hv, 10, 6, sse4);
344 mc_uni_w_funcs(qpel_h, 10, sse4);
345 mc_uni_w_funcs(qpel_v, 10, sse4);
346 mc_uni_w_funcs(qpel_hv, 10, sse4);
349 #define mc_bi_w_func(name, bitd, W, opt) \
350 void ff_hevc_put_hevc_bi_w_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t _dststride, \
351 uint8_t *_src, ptrdiff_t _srcstride, \
352 int16_t *_src2, ptrdiff_t _src2stride, \
353 int height, int denom, \
354 int _wx0, int _wx1, int _ox0, int _ox1, \
355 intptr_t mx, intptr_t my, int width) \
357 LOCAL_ALIGNED_16(int16_t, temp, [71 * 64]); \
358 ff_hevc_put_hevc_##name##W##_##bitd##_##opt(temp, 64, _src, _srcstride, height, mx, my, width); \
359 ff_hevc_put_hevc_bi_w##W##_##bitd##_##opt(_dst, _dststride, temp, 64, _src2, _src2stride, \
360 height, denom, _wx0, _wx1, _ox0, _ox1); \
363 #define mc_bi_w_funcs(name, bitd, opt) \
364 mc_bi_w_func(name, bitd, 4, opt); \
365 mc_bi_w_func(name, bitd, 8, opt); \
366 mc_bi_w_func(name, bitd, 12, opt); \
367 mc_bi_w_func(name, bitd, 16, opt); \
368 mc_bi_w_func(name, bitd, 24, opt); \
369 mc_bi_w_func(name, bitd, 32, opt); \
370 mc_bi_w_func(name, bitd, 48, opt); \
371 mc_bi_w_func(name, bitd, 64, opt)
373 mc_bi_w_funcs(pel_pixels, 8, sse4);
374 mc_bi_w_func(pel_pixels, 8, 6, sse4);
375 mc_bi_w_funcs(epel_h, 8, sse4);
376 mc_bi_w_func(epel_h, 8, 6, sse4);
377 mc_bi_w_funcs(epel_v, 8, sse4);
378 mc_bi_w_func(epel_v, 8, 6, sse4);
379 mc_bi_w_funcs(epel_hv, 8, sse4);
380 mc_bi_w_func(epel_hv, 8, 6, sse4);
381 mc_bi_w_funcs(qpel_h, 8, sse4);
382 mc_bi_w_funcs(qpel_v, 8, sse4);
383 mc_bi_w_funcs(qpel_hv, 8, sse4);
385 mc_bi_w_funcs(pel_pixels, 10, sse4);
386 mc_bi_w_func(pel_pixels, 10, 6, sse4);
387 mc_bi_w_funcs(epel_h, 10, sse4);
388 mc_bi_w_func(epel_h, 10, 6, sse4);
389 mc_bi_w_funcs(epel_v, 10, sse4);
390 mc_bi_w_func(epel_v, 10, 6, sse4);
391 mc_bi_w_funcs(epel_hv, 10, sse4);
392 mc_bi_w_func(epel_hv, 10, 6, sse4);
393 mc_bi_w_funcs(qpel_h, 10, sse4);
394 mc_bi_w_funcs(qpel_v, 10, sse4);
395 mc_bi_w_funcs(qpel_hv, 10, sse4);
397 #endif //ARCH_X86_64 && HAVE_SSE4_EXTERNAL
400 #define EPEL_LINKS(pointer, my, mx, fname, bitd, opt ) \
401 PEL_LINK(pointer, 1, my , mx , fname##4 , bitd, opt ); \
402 PEL_LINK(pointer, 2, my , mx , fname##6 , bitd, opt ); \
403 PEL_LINK(pointer, 3, my , mx , fname##8 , bitd, opt ); \
404 PEL_LINK(pointer, 4, my , mx , fname##12, bitd, opt ); \
405 PEL_LINK(pointer, 5, my , mx , fname##16, bitd, opt ); \
406 PEL_LINK(pointer, 6, my , mx , fname##24, bitd, opt ); \
407 PEL_LINK(pointer, 7, my , mx , fname##32, bitd, opt ); \
408 PEL_LINK(pointer, 8, my , mx , fname##48, bitd, opt ); \
409 PEL_LINK(pointer, 9, my , mx , fname##64, bitd, opt )
410 #define QPEL_LINKS(pointer, my, mx, fname, bitd, opt) \
411 PEL_LINK(pointer, 1, my , mx , fname##4 , bitd, opt ); \
412 PEL_LINK(pointer, 3, my , mx , fname##8 , bitd, opt ); \
413 PEL_LINK(pointer, 4, my , mx , fname##12, bitd, opt ); \
414 PEL_LINK(pointer, 5, my , mx , fname##16, bitd, opt ); \
415 PEL_LINK(pointer, 6, my , mx , fname##24, bitd, opt ); \
416 PEL_LINK(pointer, 7, my , mx , fname##32, bitd, opt ); \
417 PEL_LINK(pointer, 8, my , mx , fname##48, bitd, opt ); \
418 PEL_LINK(pointer, 9, my , mx , fname##64, bitd, opt )
421 void ff_hevcdsp_init_x86(HEVCDSPContext *c, const int bit_depth)
423 int mm_flags = av_get_cpu_flags();
425 if (bit_depth == 8) {
426 if (EXTERNAL_MMXEXT(mm_flags)) {
427 c->transform_dc_add[0] = ff_hevc_idct4_dc_add_8_mmxext;
428 c->transform_dc_add[1] = ff_hevc_idct8_dc_add_8_mmxext;
431 if (EXTERNAL_SSE2(mm_flags)) {
432 c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2;
433 c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_8_sse2;
435 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_sse2;
436 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_sse2;
439 c->transform_dc_add[2] = ff_hevc_idct16_dc_add_8_sse2;
440 c->transform_dc_add[3] = ff_hevc_idct32_dc_add_8_sse2;
442 if (EXTERNAL_SSSE3(mm_flags) && ARCH_X86_64) {
443 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_ssse3;
444 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_ssse3;
446 if (EXTERNAL_SSE4(mm_flags) && ARCH_X86_64) {
448 EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 8, sse4);
449 EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h, 8, sse4);
450 EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v, 8, sse4);
451 EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv, 8, sse4);
453 QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 8, sse4);
454 QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h, 8, sse4);
455 QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v, 8, sse4);
456 QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv, 8, sse4);
458 if (EXTERNAL_AVX2(mm_flags)) {
459 c->transform_dc_add[3] = ff_hevc_idct32_dc_add_8_avx2;
461 } else if (bit_depth == 10) {
462 if (EXTERNAL_MMXEXT(mm_flags)) {
463 c->transform_dc_add[0] = ff_hevc_idct4_dc_add_10_mmxext;
466 if (EXTERNAL_SSE2(mm_flags)) {
467 c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_sse2;
468 c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_10_sse2;
470 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_sse2;
471 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_sse2;
474 c->transform_dc_add[1] = ff_hevc_idct8_dc_add_10_sse2;
475 c->transform_dc_add[2] = ff_hevc_idct16_dc_add_10_sse2;
476 c->transform_dc_add[3] = ff_hevc_idct32_dc_add_10_sse2;
478 if (EXTERNAL_SSSE3(mm_flags) && ARCH_X86_64) {
479 c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3;
480 c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_ssse3;
482 if (EXTERNAL_SSE4(mm_flags) && ARCH_X86_64) {
484 EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 10, sse4);
485 EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h, 10, sse4);
486 EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v, 10, sse4);
487 EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv, 10, sse4);
489 QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 10, sse4);
490 QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h, 10, sse4);
491 QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v, 10, sse4);
492 QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv, 10, sse4);
494 if (EXTERNAL_AVX(mm_flags)) {
495 c->transform_dc_add[1] = ff_hevc_idct8_dc_add_10_avx;
496 c->transform_dc_add[2] = ff_hevc_idct16_dc_add_10_avx;
497 c->transform_dc_add[3] = ff_hevc_idct32_dc_add_10_avx;
499 if (EXTERNAL_AVX2(mm_flags)) {
500 c->transform_dc_add[2] = ff_hevc_idct16_dc_add_10_avx2;
501 c->transform_dc_add[3] = ff_hevc_idct32_dc_add_10_avx2;