AOMedia AV1 Codec
twopass_encoder
1/*
2 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
12// Two Pass Encoder
13// ================
14//
15// This is an example of a two pass encoder loop. It takes an input file in
16// YV12 format, passes it through the encoder twice, and writes the compressed
17// frames to disk in IVF format. It builds upon the simple_encoder example.
18//
19// Twopass Variables
20// -----------------
21// Twopass mode needs to track the current pass number and the buffer of
22// statistics packets.
23//
24// Updating The Configuration
25// ---------------------------------
26// In two pass mode, the configuration has to be updated on each pass. The
27// statistics buffer is passed on the last pass.
28//
29// Encoding A Frame
30// ----------------
31// Encoding a frame in two pass mode is identical to the simple encoder
32// example.
33//
34// Processing Statistics Packets
35// -----------------------------
36// Each packet of type `AOM_CODEC_CX_FRAME_PKT` contains the encoded data
37// for this frame. We write a IVF frame header, followed by the raw data.
38//
39//
40// Pass Progress Reporting
41// -----------------------------
42// It's sometimes helpful to see when each pass completes.
43//
44//
45// Clean-up
46// -----------------------------
47// Destruction of the encoder instance must be done on each pass. The
48// raw image should be destroyed at the end as usual.
49
50#include <stdio.h>
51#include <stdlib.h>
52#include <string.h>
53
54#include "aom/aom_encoder.h"
55#include "aom/aomcx.h"
56#include "common/tools_common.h"
57#include "common/video_writer.h"
58
59static const char *exec_name;
60
61void usage_exit(void) {
62 fprintf(stderr,
63 "Usage: %s <codec> <width> <height> <infile> <outfile> "
64 "<limit(optional)>\n",
65 exec_name);
66 exit(EXIT_FAILURE);
67}
68
69static int get_frame_stats(aom_codec_ctx_t *ctx, const aom_image_t *img,
70 aom_codec_pts_t pts, unsigned int duration,
72 aom_fixed_buf_t *stats) {
73 int got_pkts = 0;
74 aom_codec_iter_t iter = NULL;
75 const aom_codec_cx_pkt_t *pkt = NULL;
76 const aom_codec_err_t res = aom_codec_encode(ctx, img, pts, duration, flags);
77 if (res != AOM_CODEC_OK) die_codec(ctx, "Failed to get frame stats.");
78
79 while ((pkt = aom_codec_get_cx_data(ctx, &iter)) != NULL) {
80 got_pkts = 1;
81
82 if (pkt->kind == AOM_CODEC_STATS_PKT) {
83 const uint8_t *const pkt_buf = pkt->data.twopass_stats.buf;
84 const size_t pkt_size = pkt->data.twopass_stats.sz;
85 stats->buf = realloc(stats->buf, stats->sz + pkt_size);
86 if (!stats->buf) die("Failed to allocate frame stats buffer.");
87 memcpy((uint8_t *)stats->buf + stats->sz, pkt_buf, pkt_size);
88 stats->sz += pkt_size;
89 }
90 }
91
92 return got_pkts;
93}
94
95static int encode_frame(aom_codec_ctx_t *ctx, const aom_image_t *img,
96 aom_codec_pts_t pts, unsigned int duration,
97 aom_enc_frame_flags_t flags, AvxVideoWriter *writer) {
98 int got_pkts = 0;
99 aom_codec_iter_t iter = NULL;
100 const aom_codec_cx_pkt_t *pkt = NULL;
101 const aom_codec_err_t res = aom_codec_encode(ctx, img, pts, duration, flags);
102 if (res != AOM_CODEC_OK) die_codec(ctx, "Failed to encode frame.");
103
104 while ((pkt = aom_codec_get_cx_data(ctx, &iter)) != NULL) {
105 got_pkts = 1;
106 if (pkt->kind == AOM_CODEC_CX_FRAME_PKT) {
107 const int keyframe = (pkt->data.frame.flags & AOM_FRAME_IS_KEY) != 0;
108
109 if (!aom_video_writer_write_frame(writer, pkt->data.frame.buf,
110 pkt->data.frame.sz,
111 pkt->data.frame.pts))
112 die_codec(ctx, "Failed to write compressed frame.");
113 printf(keyframe ? "K" : ".");
114 fflush(stdout);
115 }
116 }
117
118 return got_pkts;
119}
120
121static aom_fixed_buf_t pass0(aom_image_t *raw, FILE *infile,
122 aom_codec_iface_t *encoder,
123 const aom_codec_enc_cfg_t *cfg, int limit) {
124 aom_codec_ctx_t codec;
125 int frame_count = 0;
126 aom_fixed_buf_t stats = { NULL, 0 };
127
128 if (aom_codec_enc_init(&codec, encoder, cfg, 0))
129 die("Failed to initialize encoder");
130
131 // Calculate frame statistics.
132 while (aom_img_read(raw, infile) && frame_count < limit) {
133 ++frame_count;
134 get_frame_stats(&codec, raw, frame_count, 1, 0, &stats);
135 }
136
137 // Flush encoder.
138 while (get_frame_stats(&codec, NULL, frame_count, 1, 0, &stats)) {
139 }
140
141 printf("Pass 0 complete. Processed %d frames.\n", frame_count);
142 if (aom_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec.");
143
144 return stats;
145}
146
147static void pass1(aom_image_t *raw, FILE *infile, const char *outfile_name,
148 aom_codec_iface_t *encoder, const aom_codec_enc_cfg_t *cfg,
149 int limit) {
150 AvxVideoInfo info = { get_fourcc_by_aom_encoder(encoder),
151 cfg->g_w,
152 cfg->g_h,
153 { cfg->g_timebase.num, cfg->g_timebase.den },
154 0 };
155 AvxVideoWriter *writer = NULL;
156 aom_codec_ctx_t codec;
157 int frame_count = 0;
158
159 writer = aom_video_writer_open(outfile_name, kContainerIVF, &info);
160 if (!writer) die("Failed to open %s for writing", outfile_name);
161
162 if (aom_codec_enc_init(&codec, encoder, cfg, 0))
163 die("Failed to initialize encoder");
164
165 if (aom_codec_control(&codec, AOME_SET_CPUUSED, 2))
166 die_codec(&codec, "Failed to set cpu-used");
167
168 // Encode frames.
169 while (aom_img_read(raw, infile) && frame_count < limit) {
170 ++frame_count;
171 encode_frame(&codec, raw, frame_count, 1, 0, writer);
172 }
173
174 // Flush encoder.
175 while (encode_frame(&codec, NULL, -1, 1, 0, writer)) {
176 }
177
178 printf("\n");
179
180 if (aom_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec.");
181
182 aom_video_writer_close(writer);
183
184 printf("Pass 1 complete. Processed %d frames.\n", frame_count);
185}
186
187int main(int argc, char **argv) {
188 FILE *infile = NULL;
189 int w, h;
190 aom_codec_ctx_t codec;
192 aom_image_t raw;
193 aom_codec_err_t res;
194 aom_fixed_buf_t stats;
195
196 const int fps = 30; // TODO(dkovalev) add command line argument
197 const int bitrate = 200; // kbit/s TODO(dkovalev) add command line argument
198 const char *const codec_arg = argv[1];
199 const char *const width_arg = argv[2];
200 const char *const height_arg = argv[3];
201 const char *const infile_arg = argv[4];
202 const char *const outfile_arg = argv[5];
203 int limit = 0;
204 exec_name = argv[0];
205
206 if (argc < 6) die("Invalid number of arguments");
207
208 if (argc > 6) limit = (int)strtol(argv[6], NULL, 0);
209
210 if (limit == 0) limit = 100;
211
212 aom_codec_iface_t *encoder = get_aom_encoder_by_short_name(codec_arg);
213 if (!encoder) die("Unsupported codec.");
214
215 w = (int)strtol(width_arg, NULL, 0);
216 h = (int)strtol(height_arg, NULL, 0);
217
218 if (w <= 0 || h <= 0 || (w % 2) != 0 || (h % 2) != 0)
219 die("Invalid frame size: %dx%d", w, h);
220
221 if (!aom_img_alloc(&raw, AOM_IMG_FMT_I420, w, h, 1))
222 die("Failed to allocate image (%dx%d)", w, h);
223
224 printf("Using %s\n", aom_codec_iface_name(encoder));
225
226 // Configuration
227 res = aom_codec_enc_config_default(encoder, &cfg, 0);
228 if (res) die_codec(&codec, "Failed to get default codec config.");
229
230 cfg.g_w = w;
231 cfg.g_h = h;
232 cfg.g_timebase.num = 1;
233 cfg.g_timebase.den = fps;
234 cfg.rc_target_bitrate = bitrate;
235
236 if (!(infile = fopen(infile_arg, "rb")))
237 die("Failed to open %s for reading", infile_arg);
238
239 // Pass 0
241 stats = pass0(&raw, infile, encoder, &cfg, limit);
242
243 // Pass 1
244 rewind(infile);
246 cfg.rc_twopass_stats_in = stats;
247 pass1(&raw, infile, outfile_arg, encoder, &cfg, limit);
248 free(stats.buf);
249
250 aom_img_free(&raw);
251 fclose(infile);
252
253 return EXIT_SUCCESS;
254}
Describes the encoder algorithm interface to applications.
aom_image_t * aom_img_alloc(aom_image_t *img, aom_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align)
Open a descriptor, allocating storage for the underlying image.
@ AOM_IMG_FMT_I420
Definition aom_image.h:45
void aom_img_free(aom_image_t *img)
Close an image descriptor.
Provides definitions for using AOM or AV1 encoder algorithm within the aom Codec Interface.
@ AOME_SET_CPUUSED
Codec control function to set encoder internal speed settings, int parameter.
Definition aomcx.h:220
const char * aom_codec_iface_name(aom_codec_iface_t *iface)
Return the name for a given interface.
aom_codec_err_t aom_codec_control(aom_codec_ctx_t *ctx, int ctrl_id,...)
Algorithm Control.
const struct aom_codec_iface aom_codec_iface_t
Codec interface structure.
Definition aom_codec.h:254
int64_t aom_codec_pts_t
Time Stamp Type.
Definition aom_codec.h:235
aom_codec_err_t aom_codec_destroy(aom_codec_ctx_t *ctx)
Destroy a codec instance.
aom_codec_err_t
Algorithm return codes.
Definition aom_codec.h:155
const void * aom_codec_iter_t
Iterator.
Definition aom_codec.h:288
#define AOM_FRAME_IS_KEY
Definition aom_codec.h:271
@ AOM_CODEC_OK
Operation completed without error.
Definition aom_codec.h:157
const aom_codec_cx_pkt_t * aom_codec_get_cx_data(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter)
Encoded data iterator.
aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img, aom_codec_pts_t pts, unsigned long duration, aom_enc_frame_flags_t flags)
Encode a frame.
#define aom_codec_enc_init(ctx, iface, cfg, flags)
Convenience macro for aom_codec_enc_init_ver()
Definition aom_encoder.h:938
aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface, aom_codec_enc_cfg_t *cfg, unsigned int usage)
Get the default configuration for a usage.
long aom_enc_frame_flags_t
Encoded Frame Flags.
Definition aom_encoder.h:375
@ AOM_RC_LAST_PASS
Definition aom_encoder.h:179
@ AOM_RC_FIRST_PASS
Definition aom_encoder.h:176
@ AOM_CODEC_CX_FRAME_PKT
Definition aom_encoder.h:108
@ AOM_CODEC_STATS_PKT
Definition aom_encoder.h:109
Codec context structure.
Definition aom_codec.h:298
Encoder output packet.
Definition aom_encoder.h:120
size_t sz
Definition aom_encoder.h:125
enum aom_codec_cx_pkt_kind kind
Definition aom_encoder.h:121
aom_fixed_buf_t twopass_stats
Definition aom_encoder.h:138
union aom_codec_cx_pkt::@1 data
aom_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition aom_encoder.h:127
struct aom_codec_cx_pkt::@1::@2 frame
aom_codec_frame_flags_t flags
Definition aom_encoder.h:130
void * buf
Definition aom_encoder.h:124
Encoder configuration structure.
Definition aom_encoder.h:385
struct aom_rational g_timebase
Stream timebase units.
Definition aom_encoder.h:487
unsigned int g_h
Height of the frame.
Definition aom_encoder.h:433
unsigned int g_w
Width of the frame.
Definition aom_encoder.h:424
enum aom_enc_pass g_pass
Multi-pass Encoding Mode.
Definition aom_encoder.h:502
unsigned int rc_target_bitrate
Target data rate.
Definition aom_encoder.h:641
aom_fixed_buf_t rc_twopass_stats_in
Two-pass stats buffer.
Definition aom_encoder.h:628
Generic fixed size buffer structure.
Definition aom_encoder.h:86
size_t sz
Definition aom_encoder.h:88
void * buf
Definition aom_encoder.h:87
Image Descriptor.
Definition aom_image.h:182
int num
Definition aom_encoder.h:163
int den
Definition aom_encoder.h:164