FFmpeg
Loading...
Searching...
No Matches
h264dsp_init.c
Go to the documentation of this file.
1/*
2 * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
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 "libavutil/attributes.h"
22#include "libavutil/cpu.h"
23#include "libavutil/x86/cpu.h"
24#include "libavcodec/h264dsp.h"
25
26/***********************************/
27/* IDCT */
28#define IDCT_ADD_FUNC(NUM, DEPTH, OPT) \
29void ff_h264_idct ## NUM ## _add_ ## DEPTH ## _ ## OPT(uint8_t *dst, \
30 int16_t *block, \
31 ptrdiff_t stride);
32
33IDCT_ADD_FUNC(, 8, sse2)
34IDCT_ADD_FUNC(, 8, avx)
35IDCT_ADD_FUNC(, 10, sse2)
36IDCT_ADD_FUNC(_dc, 8, sse2)
37IDCT_ADD_FUNC(_dc, 8, avx)
38IDCT_ADD_FUNC(_dc, 10, sse2)
39IDCT_ADD_FUNC(8_dc, 8, sse2)
40IDCT_ADD_FUNC(8_dc, 10, sse2)
41IDCT_ADD_FUNC(8, 8, sse2)
42IDCT_ADD_FUNC(8, 10, sse2)
43IDCT_ADD_FUNC(, 10, avx)
44IDCT_ADD_FUNC(8_dc, 10, avx)
45IDCT_ADD_FUNC(8, 10, avx)
46
47
48#define IDCT_ADD_REP_FUNC(NUM, REP, DEPTH, OPT) \
49void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
50 (uint8_t *dst, const int *block_offset, \
51 int16_t *block, ptrdiff_t stride, const uint8_t nnzc[5 * 8]);
52
53IDCT_ADD_REP_FUNC(8, 4, 8, sse2)
54IDCT_ADD_REP_FUNC(8, 4, 10, sse2)
55IDCT_ADD_REP_FUNC(8, 4, 10, avx)
56IDCT_ADD_REP_FUNC(, 16, 8, sse2)
57IDCT_ADD_REP_FUNC(, 16, 10, sse2)
58IDCT_ADD_REP_FUNC(, 16intra, 8, sse2)
59IDCT_ADD_REP_FUNC(, 16intra, 10, sse2)
60IDCT_ADD_REP_FUNC(, 16, 10, avx)
61IDCT_ADD_REP_FUNC(, 16intra, 10, avx)
62
63
64#define IDCT_ADD_REP_FUNC2(NUM, REP, DEPTH, OPT) \
65void ff_h264_idct ## NUM ## _add ## REP ## _ ## DEPTH ## _ ## OPT \
66 (uint8_t **dst, const int *block_offset, \
67 int16_t *block, ptrdiff_t stride, const uint8_t nnzc[15 * 8]);
68
69IDCT_ADD_REP_FUNC2(, 8, 8, sse2)
70IDCT_ADD_REP_FUNC2(, 8, 10, sse2)
71IDCT_ADD_REP_FUNC2(, 8, 10, avx)
72
73IDCT_ADD_REP_FUNC2(, 8_422, 8, sse2)
74
75IDCT_ADD_REP_FUNC2(, 8_422, 10, sse2)
76IDCT_ADD_REP_FUNC2(, 8_422, 10, avx)
77
78void ff_h264_luma_dc_dequant_idct_sse2(int16_t *output, int16_t *input, int qmul);
79
80/***********************************/
81/* deblocking */
82
83void ff_h264_loop_filter_strength_sse2(int16_t bS[2][4][4], uint8_t nnz[40],
84 int8_t ref[2][40],
85 int16_t mv[2][40][2],
86 int bidir, int edges, int step,
87 int mask_mv0, int mask_mv1, int field);
88
89#define LF_FUNC(DIR, TYPE, DEPTH, OPT) \
90void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix, \
91 ptrdiff_t stride, \
92 int alpha, \
93 int beta, \
94 int8_t *tc0);
95#define LF_IFUNC(DIR, TYPE, DEPTH, OPT) \
96void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT(uint8_t *pix, \
97 ptrdiff_t stride, \
98 int alpha, \
99 int beta);
100
101#define LF_FUNCS(type, depth) \
102LF_FUNC(h, luma, depth, sse2) \
103LF_IFUNC(h, luma_intra, depth, sse2) \
104LF_FUNC(v, luma, depth, sse2) \
105LF_IFUNC(v, luma_intra, depth, sse2) \
106LF_FUNC(h, chroma, depth, sse2) \
107LF_IFUNC(h, chroma_intra, depth, sse2) \
108LF_FUNC(h, chroma422, depth, sse2) \
109LF_IFUNC(h, chroma422_intra, depth, sse2) \
110LF_FUNC(v, chroma, depth, sse2) \
111LF_IFUNC(v, chroma_intra, depth, sse2) \
112LF_FUNC(h, luma, depth, avx) \
113LF_IFUNC(h, luma_intra, depth, avx) \
114LF_FUNC(v, luma, depth, avx) \
115LF_IFUNC(v, luma_intra, depth, avx) \
116LF_FUNC(h, chroma, depth, avx) \
117LF_IFUNC(h, chroma_intra, depth, avx) \
118LF_FUNC(h, chroma422, depth, avx) \
119LF_IFUNC(h, chroma422_intra, depth, avx) \
120LF_FUNC(v, chroma, depth, avx) \
121LF_IFUNC(v, chroma_intra, depth, avx)
122
123LF_FUNC(h, luma_mbaff, 8, sse2)
124LF_FUNC(h, luma_mbaff, 8, avx)
125
126LF_FUNCS(uint8_t, 8)
127LF_FUNCS(uint16_t, 10)
128
129/***********************************/
130/* weighted prediction */
131
132#define H264_WEIGHT(W, OPT) \
133void ff_h264_weight_ ## W ## _ ## OPT(uint8_t *dst, ptrdiff_t stride, \
134 int height, int log2_denom, \
135 int weight, int offset);
136
137#define H264_BIWEIGHT(W, OPT) \
138void ff_h264_biweight_ ## W ## _ ## OPT(uint8_t *dst, uint8_t *src, \
139 ptrdiff_t stride, int height, \
140 int log2_denom, int weightd, \
141 int weights, int offset);
142
143#define H264_BIWEIGHT_SSE(W) \
144 H264_WEIGHT(W, sse2) \
145 H264_BIWEIGHT(W, sse2) \
146 H264_BIWEIGHT(W, ssse3)
147
148H264_BIWEIGHT_SSE(16)
149H264_BIWEIGHT_SSE(8)
150H264_BIWEIGHT_SSE(4)
151
152#define H264_WEIGHT_10(W, DEPTH, OPT) \
153void ff_h264_weight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst, \
154 ptrdiff_t stride, \
155 int height, \
156 int log2_denom, \
157 int weight, \
158 int offset);
159
160#define H264_BIWEIGHT_10(W, DEPTH, OPT) \
161void ff_h264_biweight_ ## W ## _ ## DEPTH ## _ ## OPT(uint8_t *dst, \
162 uint8_t *src, \
163 ptrdiff_t stride, \
164 int height, \
165 int log2_denom, \
166 int weightd, \
167 int weights, \
168 int offset);
169
170#define H264_BIWEIGHT_10_SSE(W, DEPTH) \
171 H264_WEIGHT_10(W, DEPTH, sse2) \
172 H264_WEIGHT_10(W, DEPTH, sse4) \
173 H264_BIWEIGHT_10(W, DEPTH, sse2) \
174 H264_BIWEIGHT_10(W, DEPTH, sse4)
175
176H264_BIWEIGHT_10_SSE(16, 10)
177H264_BIWEIGHT_10_SSE(8, 10)
178H264_BIWEIGHT_10_SSE(4, 10)
179
180av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
181 const int chroma_format_idc)
182{
183 int cpu_flags = av_get_cpu_flags();
184
185 if (EXTERNAL_SSE2(cpu_flags) && chroma_format_idc <= 1)
186 c->loop_filter_strength = ff_h264_loop_filter_strength_sse2;
187
188 if (bit_depth == 8) {
189 if (EXTERNAL_SSE2(cpu_flags)) {
190 c->idct8_add = ff_h264_idct8_add_8_sse2;
191 c->idct8_dc_add = ff_h264_idct8_dc_add_8_sse2;
192 c->idct_add16 = ff_h264_idct_add16_8_sse2;
193 c->idct8_add4 = ff_h264_idct8_add4_8_sse2;
194 if (chroma_format_idc <= 1)
195 c->idct_add8 = ff_h264_idct_add8_8_sse2;
196 else
197 c->idct_add8 = ff_h264_idct_add8_422_8_sse2;
198 c->idct_add16intra = ff_h264_idct_add16intra_8_sse2;
199 c->luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_sse2;
200
201 c->weight_pixels_tab[0] = ff_h264_weight_16_sse2;
202 c->weight_pixels_tab[1] = ff_h264_weight_8_sse2;
203 c->weight_pixels_tab[2] = ff_h264_weight_4_sse2;
204
205 c->biweight_pixels_tab[0] = ff_h264_biweight_16_sse2;
206 c->biweight_pixels_tab[1] = ff_h264_biweight_8_sse2;
207 c->biweight_pixels_tab[2] = ff_h264_biweight_4_sse2;
208
209 c->v_loop_filter_luma = ff_deblock_v_luma_8_sse2;
210 c->h_loop_filter_luma = ff_deblock_h_luma_8_sse2;
211 c->v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_sse2;
212 c->h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_sse2;
213
214#if ARCH_X86_64
215 c->h_loop_filter_luma_mbaff = ff_deblock_h_luma_mbaff_8_sse2;
216#endif
217
218 c->v_loop_filter_chroma = ff_deblock_v_chroma_8_sse2;
219 c->v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_8_sse2;
220 if (chroma_format_idc <= 1) {
221 c->h_loop_filter_chroma = ff_deblock_h_chroma_8_sse2;
222 c->h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_sse2;
223 } else {
224 c->h_loop_filter_chroma = ff_deblock_h_chroma422_8_sse2;
225 c->h_loop_filter_chroma_intra = ff_deblock_h_chroma422_intra_8_sse2;
226 }
227
228 c->idct_add = ff_h264_idct_add_8_sse2;
229 c->idct_dc_add = ff_h264_idct_dc_add_8_sse2;
230 }
231 if (EXTERNAL_SSSE3(cpu_flags)) {
232 c->biweight_pixels_tab[0] = ff_h264_biweight_16_ssse3;
233 c->biweight_pixels_tab[1] = ff_h264_biweight_8_ssse3;
234 c->biweight_pixels_tab[2] = ff_h264_biweight_4_ssse3;
235 }
236 if (EXTERNAL_AVX(cpu_flags)) {
237 c->v_loop_filter_luma = ff_deblock_v_luma_8_avx;
238 c->h_loop_filter_luma = ff_deblock_h_luma_8_avx;
239 c->v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_avx;
240 c->h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_avx;
241#if ARCH_X86_64
242 c->h_loop_filter_luma_mbaff = ff_deblock_h_luma_mbaff_8_avx;
243#endif
244
245 c->v_loop_filter_chroma = ff_deblock_v_chroma_8_avx;
246 c->v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_8_avx;
247 if (chroma_format_idc <= 1) {
248 c->h_loop_filter_chroma = ff_deblock_h_chroma_8_avx;
249 c->h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_avx;
250 } else {
251 c->h_loop_filter_chroma = ff_deblock_h_chroma422_8_avx;
252 c->h_loop_filter_chroma_intra = ff_deblock_h_chroma422_intra_8_avx;
253 }
254
255 c->idct_add = ff_h264_idct_add_8_avx;
256 c->idct_dc_add = ff_h264_idct_dc_add_8_avx;
257 }
258 } else if (bit_depth == 10) {
259 if (EXTERNAL_SSE2(cpu_flags)) {
260 c->idct_dc_add = ff_h264_idct_dc_add_10_sse2;
261 c->idct_add = ff_h264_idct_add_10_sse2;
262 c->idct8_dc_add = ff_h264_idct8_dc_add_10_sse2;
263
264 c->idct_add16 = ff_h264_idct_add16_10_sse2;
265 if (chroma_format_idc <= 1) {
266 c->idct_add8 = ff_h264_idct_add8_10_sse2;
267 } else {
268 c->idct_add8 = ff_h264_idct_add8_422_10_sse2;
269 }
270 c->idct_add16intra = ff_h264_idct_add16intra_10_sse2;
271#if HAVE_ALIGNED_STACK
272 c->idct8_add = ff_h264_idct8_add_10_sse2;
273 c->idct8_add4 = ff_h264_idct8_add4_10_sse2;
274#endif /* HAVE_ALIGNED_STACK */
275
276 c->weight_pixels_tab[0] = ff_h264_weight_16_10_sse2;
277 c->weight_pixels_tab[1] = ff_h264_weight_8_10_sse2;
278 c->weight_pixels_tab[2] = ff_h264_weight_4_10_sse2;
279
280 c->biweight_pixels_tab[0] = ff_h264_biweight_16_10_sse2;
281 c->biweight_pixels_tab[1] = ff_h264_biweight_8_10_sse2;
282 c->biweight_pixels_tab[2] = ff_h264_biweight_4_10_sse2;
283
284 c->v_loop_filter_chroma = ff_deblock_v_chroma_10_sse2;
285 c->v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_sse2;
286 if (chroma_format_idc <= 1) {
287 c->h_loop_filter_chroma = ff_deblock_h_chroma_10_sse2;
288 } else {
289 c->h_loop_filter_chroma = ff_deblock_h_chroma422_10_sse2;
290 }
291 c->v_loop_filter_luma = ff_deblock_v_luma_10_sse2;
292 c->h_loop_filter_luma = ff_deblock_h_luma_10_sse2;
293 c->v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_sse2;
294 c->h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_sse2;
295 }
296 if (EXTERNAL_SSE4(cpu_flags)) {
297 c->weight_pixels_tab[0] = ff_h264_weight_16_10_sse4;
298 c->weight_pixels_tab[1] = ff_h264_weight_8_10_sse4;
299 c->weight_pixels_tab[2] = ff_h264_weight_4_10_sse4;
300
301 c->biweight_pixels_tab[0] = ff_h264_biweight_16_10_sse4;
302 c->biweight_pixels_tab[1] = ff_h264_biweight_8_10_sse4;
303 c->biweight_pixels_tab[2] = ff_h264_biweight_4_10_sse4;
304 }
305 if (EXTERNAL_AVX(cpu_flags)) {
306 c->idct_dc_add =
307 c->idct_add = ff_h264_idct_add_10_avx;
308 c->idct8_dc_add = ff_h264_idct8_dc_add_10_avx;
309
310 c->idct_add16 = ff_h264_idct_add16_10_avx;
311 if (chroma_format_idc <= 1) {
312 c->idct_add8 = ff_h264_idct_add8_10_avx;
313 } else {
314 c->idct_add8 = ff_h264_idct_add8_422_10_avx;
315 }
316 c->idct_add16intra = ff_h264_idct_add16intra_10_avx;
317#if HAVE_ALIGNED_STACK
318 c->idct8_add = ff_h264_idct8_add_10_avx;
319 c->idct8_add4 = ff_h264_idct8_add4_10_avx;
320#endif /* HAVE_ALIGNED_STACK */
321
322 c->v_loop_filter_chroma = ff_deblock_v_chroma_10_avx;
323 c->v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_avx;
324 if (chroma_format_idc <= 1) {
325 c->h_loop_filter_chroma = ff_deblock_h_chroma_10_avx;
326 } else {
327 c->h_loop_filter_chroma = ff_deblock_h_chroma422_10_avx;
328 }
329 c->v_loop_filter_luma = ff_deblock_v_luma_10_avx;
330 c->h_loop_filter_luma = ff_deblock_h_luma_10_avx;
331 c->v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_avx;
332 c->h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_avx;
333 }
334 }
335}
bit_depth
static void bit_depth(AudioStatsContext *s, const uint64_t *const mask, uint8_t *depth)
Definition af_astats.c:246
h264dsp.h
H.264 DSP functions.
ff_h264dsp_init_x86
void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
Definition h264dsp_init.c:180
mv
static const int8_t mv[256][2]
Definition 4xm.c:81
attributes.h
Macro definitions for various function/variable attributes.
av_cold
#define av_cold
Definition attributes.h:117
cpu_flags
static atomic_int cpu_flags
Definition cpu.c:56
av_get_cpu_flags
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition cpu.c:109
EXTERNAL_AVX
#define EXTERNAL_AVX(flags)
Definition cpu.h:65
EXTERNAL_SSSE3
#define EXTERNAL_SSSE3(flags)
Definition cpu.h:60
EXTERNAL_SSE4
#define EXTERNAL_SSE4(flags)
Definition cpu.h:63
EXTERNAL_SSE2
#define EXTERNAL_SSE2(flags)
Definition cpu.h:54
H264DSPContext
Context for storing H.264 DSP functions.
Definition h264dsp.h:42
ref
static int ref[MAX_W *MAX_W]
Definition jpeg2000dwt.c:117
h
h
Definition vp9dsp_template.c:2070
c
static double c[64]
Definition vsrc_mptestsrc.c:89
IDCT_ADD_REP_FUNC2
#define IDCT_ADD_REP_FUNC2(NUM, REP, DEPTH, OPT)
Definition h264dsp_init.c:64
ff_h264_loop_filter_strength_sse2
void ff_h264_loop_filter_strength_sse2(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2], int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field)
ff_h264_luma_dc_dequant_idct_sse2
void ff_h264_luma_dc_dequant_idct_sse2(int16_t *output, int16_t *input, int qmul)
H264_BIWEIGHT_SSE
#define H264_BIWEIGHT_SSE(W)
Definition h264dsp_init.c:143
LF_FUNCS
#define LF_FUNCS(type, depth)
Definition h264dsp_init.c:101
IDCT_ADD_FUNC
#define IDCT_ADD_FUNC(NUM, DEPTH, OPT)
Definition h264dsp_init.c:28
H264_BIWEIGHT_10_SSE
#define H264_BIWEIGHT_10_SSE(W, DEPTH)
Definition h264dsp_init.c:170
LF_FUNC
#define LF_FUNC(DIR, TYPE, DEPTH, OPT)
Definition h264dsp_init.c:89
IDCT_ADD_REP_FUNC
#define IDCT_ADD_REP_FUNC(NUM, REP, DEPTH, OPT)
Definition h264dsp_init.c:48
Generated on Sat Nov 14 2026 19:20:56 for FFmpeg by doxygen 1.13.2