vp9: Add bsf to merge superframes
[libav.git] / libavcodec / vp9_superframe_bsf.c
1 /*
2 * VP9 invisible (alt-ref) frame to superframe merge bitstream filter
3 * Copyright (c) 2016 Ronald S. Bultje <rsbultje@gmail.com>
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include "libavutil/avassert.h"
23 #include "avcodec.h"
24 #include "bsf.h"
25 #include "get_bits.h"
26
27 #define MAX_CACHE 8
28 typedef struct VP9BSFContext {
29 int n_cache;
30 struct CachedBuf {
31 uint8_t *data;
32 int size;
33 } cache[MAX_CACHE];
34 } VP9BSFContext;
35
36 static void stats(const struct CachedBuf *in, int n_in,
37 unsigned *_max, unsigned *_sum)
38 {
39 int n;
40 unsigned max = 0, sum = 0;
41
42 for (n = 0; n < n_in; n++) {
43 unsigned sz = in[n].size;
44
45 if (sz > max)
46 max = sz;
47 sum += sz;
48 }
49
50 *_max = max;
51 *_sum = sum;
52 }
53
54 static int merge_superframe(const struct CachedBuf *in, int n_in, AVPacket *out)
55 {
56 unsigned max, sum, mag, marker, n, sz;
57 uint8_t *ptr;
58 int res;
59
60 stats(in, n_in, &max, &sum);
61 mag = av_log2(max) >> 3;
62 marker = 0xC0 + (mag << 3) + (n_in - 1);
63 sz = sum + 2 + (mag + 1) * n_in;
64 res = av_new_packet(out, sz);
65 if (res < 0)
66 return res;
67 ptr = out->data;
68 for (n = 0; n < n_in; n++) {
69 memcpy(ptr, in[n].data, in[n].size);
70 ptr += in[n].size;
71 }
72
73 #define wloop(mag, wr) do { \
74 for (n = 0; n < n_in; n++) { \
75 wr; \
76 ptr += mag + 1; \
77 } \
78 } while (0)
79
80 // write superframe with marker 110[mag:2][nframes:3]
81 *ptr++ = marker;
82 switch (mag) {
83 case 0:
84 wloop(mag, *ptr = in[n].size);
85 break;
86 case 1:
87 wloop(mag, AV_WL16(ptr, in[n].size));
88 break;
89 case 2:
90 wloop(mag, AV_WL24(ptr, in[n].size));
91 break;
92 case 3:
93 wloop(mag, AV_WL32(ptr, in[n].size));
94 break;
95 }
96 *ptr++ = marker;
97 av_assert0(ptr == &out->data[out->size]);
98
99 return 0;
100 }
101
102 static int vp9_superframe_filter(AVBSFContext *ctx, AVPacket *out)
103 {
104 GetBitContext gb;
105 VP9BSFContext *s = ctx->priv_data;
106 AVPacket *in;
107 int res, invisible, profile, marker, uses_superframe_syntax = 0, n;
108
109 res = ff_bsf_get_packet(ctx, &in);
110 if (res < 0)
111 return res;
112
113 marker = in->data[in->size - 1];
114 if ((marker & 0xe0) == 0xc0) {
115 int nbytes = 1 + ((marker >> 3) & 0x3);
116 int n_frames = 1 + (marker & 0x7), idx_sz = 2 + n_frames * nbytes;
117
118 uses_superframe_syntax = in->size >= idx_sz && in->data[in->size - idx_sz] == marker;
119 }
120
121 if ((res = init_get_bits8(&gb, in->data, in->size)) < 0)
122 goto done;
123
124 get_bits(&gb, 2); // frame marker
125 profile = get_bits1(&gb);
126 profile |= get_bits1(&gb) << 1;
127 if (profile == 3) profile += get_bits1(&gb);
128
129 if (get_bits1(&gb)) {
130 invisible = 0;
131 } else {
132 get_bits1(&gb); // keyframe
133 invisible = !get_bits1(&gb);
134 }
135
136 if (uses_superframe_syntax && s->n_cache > 0) {
137 av_log(ctx, AV_LOG_ERROR,
138 "Mixing of superframe syntax and naked VP9 frames not supported");
139 res = AVERROR(ENOSYS);
140 goto done;
141 } else if ((!invisible || uses_superframe_syntax) && !s->n_cache) {
142 // passthrough
143 av_packet_move_ref(out, in);
144 goto done;
145 } else if (s->n_cache + 1 >= MAX_CACHE) {
146 av_log(ctx, AV_LOG_ERROR,
147 "Too many invisible frames");
148 res = AVERROR_INVALIDDATA;
149 goto done;
150 }
151
152 if (invisible) {
153 s->cache[s->n_cache].data = av_malloc(in->size);
154 if (!s->cache[s->n_cache].data) {
155 res = AVERROR(ENOMEM);
156 goto done;
157 }
158 memcpy(s->cache[s->n_cache].data, in->data, in->size);
159 s->cache[s->n_cache++].size = in->size;
160 res = AVERROR(EAGAIN);
161 goto done;
162 }
163 av_assert0(s->n_cache > 0);
164
165 s->cache[s->n_cache].data = in->data;
166 s->cache[s->n_cache].size = in->size;
167
168 // build superframe
169 if ((res = merge_superframe(s->cache, s->n_cache + 1, out)) < 0)
170 goto done;
171
172 for (n = 0; n < s->n_cache; n++)
173 av_freep(&s->cache[n].data);
174 s->n_cache = 0;
175
176 res = av_packet_copy_props(out, in);
177 if (res < 0)
178 goto done;
179
180 done:
181 if (res < 0)
182 av_packet_unref(out);
183 av_packet_free(&in);
184 return res;
185 }
186
187 static void vp9_superframe_close(AVBSFContext *ctx)
188 {
189 VP9BSFContext *s = ctx->priv_data;
190 int n;
191
192 // free cached data
193 for (n = 0; n < s->n_cache; n++)
194 av_freep(&s->cache[n].data);
195 }
196
197 static const enum AVCodecID codec_ids[] = {
198 AV_CODEC_ID_VP9, AV_CODEC_ID_NONE,
199 };
200
201 const AVBitStreamFilter ff_vp9_superframe_bsf = {
202 .name = "vp9_superframe",
203 .priv_data_size = sizeof(VP9BSFContext),
204 .filter = vp9_superframe_filter,
205 .close = vp9_superframe_close,
206 .codec_ids = codec_ids,
207 };