2 * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
4 * This file is part of FFmpeg.
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.
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.
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
21 #include "libavutil/attributes.h"
22 #include "libavutil/cpu.h"
23 #include "libavutil/x86/asm.h"
24 #include "libavutil/x86/cpu.h"
25 #include "libavcodec/h264dsp.h"
27 /***********************************/
29 #define IDCT_ADD_FUNC(NUM, DEPTH, OPT) \
30 void ff_h264_idct ## NUM ## _add_ ## DEPTH ## _ ## OPT(uint8_t *dst, \
34 IDCT_ADD_FUNC(, 8, sse2
)
35 IDCT_ADD_FUNC(, 8, avx
)
36 IDCT_ADD_FUNC(, 10, sse2
)
37 IDCT_ADD_FUNC(_dc
, 8, sse2
)
38 IDCT_ADD_FUNC(_dc
, 8, avx
)
39 IDCT_ADD_FUNC(_dc
, 10, mmxext
)
40 IDCT_ADD_FUNC(8_dc
, 8, mmxext
)
41 IDCT_ADD_FUNC(8_dc
, 10, sse2
)
42 IDCT_ADD_FUNC(8, 8, sse2
)
43 IDCT_ADD_FUNC(8, 10, sse2
)
44 IDCT_ADD_FUNC(, 10, avx
)
45 IDCT_ADD_FUNC(8_dc
, 10, avx
)
46 IDCT_ADD_FUNC(8, 10, avx
)
49 #define IDCT_ADD_REP_FUNC(NUM, REP, DEPTH, OPT) \
50 void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
51 (uint8_t *dst, const int *block_offset, \
52 int16_t *block, int stride, const uint8_t nnzc[5 * 8]);
54 IDCT_ADD_REP_FUNC(8, 4, 8, sse2
)
55 IDCT_ADD_REP_FUNC(8, 4, 10, sse2
)
56 IDCT_ADD_REP_FUNC(8, 4, 10, avx
)
57 IDCT_ADD_REP_FUNC(, 16, 8, sse2
)
58 IDCT_ADD_REP_FUNC(, 16, 10, sse2
)
59 IDCT_ADD_REP_FUNC(, 16intra
, 8, sse2
)
60 IDCT_ADD_REP_FUNC(, 16intra
, 10, sse2
)
61 IDCT_ADD_REP_FUNC(, 16, 10, avx
)
62 IDCT_ADD_REP_FUNC(, 16intra
, 10, avx
)
65 #define IDCT_ADD_REP_FUNC2(NUM, REP, DEPTH, OPT) \
66 void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
67 (uint8_t **dst, const int *block_offset, \
68 int16_t *block, int stride, const uint8_t nnzc[15 * 8]);
70 IDCT_ADD_REP_FUNC2(, 8, 8, sse2
)
71 IDCT_ADD_REP_FUNC2(, 8, 10, sse2
)
72 IDCT_ADD_REP_FUNC2(, 8, 10, avx
)
74 IDCT_ADD_REP_FUNC2(, 8_422
, 8, mmx
)
76 IDCT_ADD_REP_FUNC2(, 8_422
, 10, sse2
)
77 IDCT_ADD_REP_FUNC2(, 8_422
, 10, avx
)
79 void ff_h264_luma_dc_dequant_idct_sse2(int16_t *output
, int16_t *input
, int qmul
);
81 /***********************************/
84 void ff_h264_loop_filter_strength_mmxext(int16_t bS
[2][4][4], uint8_t nnz
[40],
87 int bidir
, int edges
, int step
,
88 int mask_mv0
, int mask_mv1
, int field
);
90 #define LF_FUNC(DIR, TYPE, DEPTH, OPT) \
91 void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix, \
96 #define LF_IFUNC(DIR, TYPE, DEPTH, OPT) \
97 void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix, \
102 #define LF_FUNCS(type, depth) \
103 LF_FUNC(h, luma, depth, sse2) \
104 LF_IFUNC(h, luma_intra, depth, sse2) \
105 LF_FUNC(v, luma, depth, sse2) \
106 LF_IFUNC(v, luma_intra, depth, sse2) \
107 LF_FUNC(h, chroma, depth, sse2) \
108 LF_IFUNC(h, chroma_intra, depth, sse2) \
109 LF_FUNC(h, chroma422, depth, sse2) \
110 LF_IFUNC(h, chroma422_intra, depth, sse2) \
111 LF_FUNC(v, chroma, depth, sse2) \
112 LF_IFUNC(v, chroma_intra, depth, sse2) \
113 LF_FUNC(h, luma, depth, avx) \
114 LF_IFUNC(h, luma_intra, depth, avx) \
115 LF_FUNC(v, luma, depth, avx) \
116 LF_IFUNC(v, luma_intra, depth, avx) \
117 LF_FUNC(h, chroma, depth, avx) \
118 LF_IFUNC(h, chroma_intra, depth, avx) \
119 LF_FUNC(h, chroma422, depth, avx) \
120 LF_IFUNC(h, chroma422_intra, depth, avx) \
121 LF_FUNC(v, chroma, depth, avx) \
122 LF_IFUNC(v, chroma_intra, depth, avx)
124 LF_FUNC(h
, luma_mbaff
, 8, sse2
)
125 LF_FUNC(h
, luma_mbaff
, 8, avx
)
128 LF_FUNCS(uint16_t, 10)
130 LF_FUNC(v
, luma
, 10, mmxext
)
131 LF_FUNC(h
, luma
, 10, mmxext
)
132 LF_IFUNC(v
, luma_intra
, 10, mmxext
)
133 LF_IFUNC(h
, luma_intra
, 10, mmxext
)
135 /***********************************/
136 /* weighted prediction */
138 #define H264_WEIGHT(W, OPT) \
139 void ff_h264_weight_ ## W ## _ ## OPT(uint8_t *dst, ptrdiff_t stride, \
140 int height, int log2_denom, \
141 int weight, int offset);
143 #define H264_BIWEIGHT(W, OPT) \
144 void ff_h264_biweight_ ## W ## _ ## OPT(uint8_t *dst, uint8_t *src, \
145 ptrdiff_t stride, int height, \
146 int log2_denom, int weightd, \
147 int weights, int offset);
149 #define H264_BIWEIGHT_MMX(W) \
150 H264_WEIGHT(W, mmxext) \
151 H264_BIWEIGHT(W, mmxext)
153 #define H264_BIWEIGHT_SSE(W) \
154 H264_WEIGHT(W, sse2) \
155 H264_BIWEIGHT(W, sse2) \
156 H264_BIWEIGHT(W, ssse3)
158 H264_BIWEIGHT_SSE(16)
162 #define H264_WEIGHT_10(W, DEPTH, OPT) \
163 void ff_h264_weight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst, \
170 #define H264_BIWEIGHT_10(W, DEPTH, OPT) \
171 void ff_h264_biweight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst, \
180 #define H264_BIWEIGHT_10_SSE(W, DEPTH) \
181 H264_WEIGHT_10(W, DEPTH, sse2) \
182 H264_WEIGHT_10(W, DEPTH, sse4) \
183 H264_BIWEIGHT_10(W, DEPTH, sse2) \
184 H264_BIWEIGHT_10(W, DEPTH, sse4)
186 H264_BIWEIGHT_10_SSE(16, 10)
187 H264_BIWEIGHT_10_SSE(8, 10)
188 H264_BIWEIGHT_10_SSE(4, 10)
190 av_cold
void ff_h264dsp_init_x86(H264DSPContext
*c
, const int bit_depth
,
191 const int chroma_format_idc
)
194 int cpu_flags
= av_get_cpu_flags();
196 if (EXTERNAL_MMXEXT(cpu_flags
) && chroma_format_idc
<= 1)
197 c
->h264_loop_filter_strength
= ff_h264_loop_filter_strength_mmxext
;
199 if (bit_depth
== 8) {
200 if (EXTERNAL_MMX(cpu_flags
)) {
201 if (chroma_format_idc
<= 1) {
203 c
->h264_idct_add8
= ff_h264_idct_add8_422_8_mmx
;
206 if (EXTERNAL_MMXEXT(cpu_flags
)) {
207 c
->h264_idct8_dc_add
= ff_h264_idct8_dc_add_8_mmxext
;
209 c
->weight_h264_pixels_tab
[2] = ff_h264_weight_4_mmxext
;
211 c
->biweight_h264_pixels_tab
[2] = ff_h264_biweight_4_mmxext
;
213 if (EXTERNAL_SSE2(cpu_flags
)) {
214 c
->h264_idct8_add
= ff_h264_idct8_add_8_sse2
;
216 c
->h264_idct_add16
= ff_h264_idct_add16_8_sse2
;
217 c
->h264_idct8_add4
= ff_h264_idct8_add4_8_sse2
;
218 if (chroma_format_idc
<= 1)
219 c
->h264_idct_add8
= ff_h264_idct_add8_8_sse2
;
220 c
->h264_idct_add16intra
= ff_h264_idct_add16intra_8_sse2
;
221 c
->h264_luma_dc_dequant_idct
= ff_h264_luma_dc_dequant_idct_sse2
;
223 c
->weight_h264_pixels_tab
[0] = ff_h264_weight_16_sse2
;
224 c
->weight_h264_pixels_tab
[1] = ff_h264_weight_8_sse2
;
226 c
->biweight_h264_pixels_tab
[0] = ff_h264_biweight_16_sse2
;
227 c
->biweight_h264_pixels_tab
[1] = ff_h264_biweight_8_sse2
;
229 c
->h264_v_loop_filter_luma
= ff_deblock_v_luma_8_sse2
;
230 c
->h264_h_loop_filter_luma
= ff_deblock_h_luma_8_sse2
;
231 c
->h264_v_loop_filter_luma_intra
= ff_deblock_v_luma_intra_8_sse2
;
232 c
->h264_h_loop_filter_luma_intra
= ff_deblock_h_luma_intra_8_sse2
;
235 c
->h264_h_loop_filter_luma_mbaff
= ff_deblock_h_luma_mbaff_8_sse2
;
238 c
->h264_v_loop_filter_chroma
= ff_deblock_v_chroma_8_sse2
;
239 c
->h264_v_loop_filter_chroma_intra
= ff_deblock_v_chroma_intra_8_sse2
;
240 if (chroma_format_idc
<= 1) {
241 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma_8_sse2
;
242 c
->h264_h_loop_filter_chroma_intra
= ff_deblock_h_chroma_intra_8_sse2
;
244 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma422_8_sse2
;
245 c
->h264_h_loop_filter_chroma_intra
= ff_deblock_h_chroma422_intra_8_sse2
;
248 c
->h264_idct_add
= ff_h264_idct_add_8_sse2
;
249 c
->h264_idct_dc_add
= ff_h264_idct_dc_add_8_sse2
;
251 if (EXTERNAL_SSSE3(cpu_flags
)) {
252 c
->biweight_h264_pixels_tab
[0] = ff_h264_biweight_16_ssse3
;
253 c
->biweight_h264_pixels_tab
[1] = ff_h264_biweight_8_ssse3
;
255 if (EXTERNAL_AVX(cpu_flags
)) {
256 c
->h264_v_loop_filter_luma
= ff_deblock_v_luma_8_avx
;
257 c
->h264_h_loop_filter_luma
= ff_deblock_h_luma_8_avx
;
258 c
->h264_v_loop_filter_luma_intra
= ff_deblock_v_luma_intra_8_avx
;
259 c
->h264_h_loop_filter_luma_intra
= ff_deblock_h_luma_intra_8_avx
;
261 c
->h264_h_loop_filter_luma_mbaff
= ff_deblock_h_luma_mbaff_8_avx
;
264 c
->h264_v_loop_filter_chroma
= ff_deblock_v_chroma_8_avx
;
265 c
->h264_v_loop_filter_chroma_intra
= ff_deblock_v_chroma_intra_8_avx
;
266 if (chroma_format_idc
<= 1) {
267 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma_8_avx
;
268 c
->h264_h_loop_filter_chroma_intra
= ff_deblock_h_chroma_intra_8_avx
;
270 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma422_8_avx
;
271 c
->h264_h_loop_filter_chroma_intra
= ff_deblock_h_chroma422_intra_8_avx
;
274 c
->h264_idct_add
= ff_h264_idct_add_8_avx
;
275 c
->h264_idct_dc_add
= ff_h264_idct_dc_add_8_avx
;
277 } else if (bit_depth
== 10) {
278 if (EXTERNAL_MMXEXT(cpu_flags
)) {
279 #if ARCH_X86_32 && !HAVE_ALIGNED_STACK
280 c
->h264_v_loop_filter_luma
= ff_deblock_v_luma_10_mmxext
;
281 c
->h264_h_loop_filter_luma
= ff_deblock_h_luma_10_mmxext
;
282 c
->h264_v_loop_filter_luma_intra
= ff_deblock_v_luma_intra_10_mmxext
;
283 c
->h264_h_loop_filter_luma_intra
= ff_deblock_h_luma_intra_10_mmxext
;
284 #endif /* ARCH_X86_32 && !HAVE_ALIGNED_STACK */
285 c
->h264_idct_dc_add
= ff_h264_idct_dc_add_10_mmxext
;
287 if (EXTERNAL_SSE2(cpu_flags
)) {
288 c
->h264_idct_add
= ff_h264_idct_add_10_sse2
;
289 c
->h264_idct8_dc_add
= ff_h264_idct8_dc_add_10_sse2
;
291 c
->h264_idct_add16
= ff_h264_idct_add16_10_sse2
;
292 if (chroma_format_idc
<= 1) {
293 c
->h264_idct_add8
= ff_h264_idct_add8_10_sse2
;
295 c
->h264_idct_add8
= ff_h264_idct_add8_422_10_sse2
;
297 c
->h264_idct_add16intra
= ff_h264_idct_add16intra_10_sse2
;
298 #if HAVE_ALIGNED_STACK
299 c
->h264_idct8_add
= ff_h264_idct8_add_10_sse2
;
300 c
->h264_idct8_add4
= ff_h264_idct8_add4_10_sse2
;
301 #endif /* HAVE_ALIGNED_STACK */
303 c
->weight_h264_pixels_tab
[0] = ff_h264_weight_16_10_sse2
;
304 c
->weight_h264_pixels_tab
[1] = ff_h264_weight_8_10_sse2
;
305 c
->weight_h264_pixels_tab
[2] = ff_h264_weight_4_10_sse2
;
307 c
->biweight_h264_pixels_tab
[0] = ff_h264_biweight_16_10_sse2
;
308 c
->biweight_h264_pixels_tab
[1] = ff_h264_biweight_8_10_sse2
;
309 c
->biweight_h264_pixels_tab
[2] = ff_h264_biweight_4_10_sse2
;
311 c
->h264_v_loop_filter_chroma
= ff_deblock_v_chroma_10_sse2
;
312 c
->h264_v_loop_filter_chroma_intra
= ff_deblock_v_chroma_intra_10_sse2
;
313 if (chroma_format_idc
<= 1) {
314 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma_10_sse2
;
316 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma422_10_sse2
;
318 #if HAVE_ALIGNED_STACK
319 c
->h264_v_loop_filter_luma
= ff_deblock_v_luma_10_sse2
;
320 c
->h264_h_loop_filter_luma
= ff_deblock_h_luma_10_sse2
;
321 c
->h264_v_loop_filter_luma_intra
= ff_deblock_v_luma_intra_10_sse2
;
322 c
->h264_h_loop_filter_luma_intra
= ff_deblock_h_luma_intra_10_sse2
;
323 #endif /* HAVE_ALIGNED_STACK */
325 if (EXTERNAL_SSE4(cpu_flags
)) {
326 c
->weight_h264_pixels_tab
[0] = ff_h264_weight_16_10_sse4
;
327 c
->weight_h264_pixels_tab
[1] = ff_h264_weight_8_10_sse4
;
328 c
->weight_h264_pixels_tab
[2] = ff_h264_weight_4_10_sse4
;
330 c
->biweight_h264_pixels_tab
[0] = ff_h264_biweight_16_10_sse4
;
331 c
->biweight_h264_pixels_tab
[1] = ff_h264_biweight_8_10_sse4
;
332 c
->biweight_h264_pixels_tab
[2] = ff_h264_biweight_4_10_sse4
;
334 if (EXTERNAL_AVX(cpu_flags
)) {
335 c
->h264_idct_dc_add
=
336 c
->h264_idct_add
= ff_h264_idct_add_10_avx
;
337 c
->h264_idct8_dc_add
= ff_h264_idct8_dc_add_10_avx
;
339 c
->h264_idct_add16
= ff_h264_idct_add16_10_avx
;
340 if (chroma_format_idc
<= 1) {
341 c
->h264_idct_add8
= ff_h264_idct_add8_10_avx
;
343 c
->h264_idct_add8
= ff_h264_idct_add8_422_10_avx
;
345 c
->h264_idct_add16intra
= ff_h264_idct_add16intra_10_avx
;
346 #if HAVE_ALIGNED_STACK
347 c
->h264_idct8_add
= ff_h264_idct8_add_10_avx
;
348 c
->h264_idct8_add4
= ff_h264_idct8_add4_10_avx
;
349 #endif /* HAVE_ALIGNED_STACK */
351 c
->h264_v_loop_filter_chroma
= ff_deblock_v_chroma_10_avx
;
352 c
->h264_v_loop_filter_chroma_intra
= ff_deblock_v_chroma_intra_10_avx
;
353 if (chroma_format_idc
<= 1) {
354 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma_10_avx
;
356 c
->h264_h_loop_filter_chroma
= ff_deblock_h_chroma422_10_avx
;
358 #if HAVE_ALIGNED_STACK
359 c
->h264_v_loop_filter_luma
= ff_deblock_v_luma_10_avx
;
360 c
->h264_h_loop_filter_luma
= ff_deblock_h_luma_10_avx
;
361 c
->h264_v_loop_filter_luma_intra
= ff_deblock_v_luma_intra_10_avx
;
362 c
->h264_h_loop_filter_luma_intra
= ff_deblock_h_luma_intra_10_avx
;
363 #endif /* HAVE_ALIGNED_STACK */