Commit | Line | Data |
---|---|---|
2ba45a60 DM |
1 | /* |
2 | * Interface to libmp3lame for mp3 encoding | |
3 | * Copyright (c) 2002 Lennert Buytenhek <buytenh@gnu.org> | |
4 | * | |
5 | * This file is part of FFmpeg. | |
6 | * | |
7 | * FFmpeg 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 | * FFmpeg 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 FFmpeg; if not, write to the Free Software | |
19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
20 | */ | |
21 | ||
22 | /** | |
23 | * @file | |
24 | * Interface to libmp3lame for mp3 encoding. | |
25 | */ | |
26 | ||
27 | #include <lame/lame.h> | |
28 | ||
29 | #include "libavutil/channel_layout.h" | |
30 | #include "libavutil/common.h" | |
31 | #include "libavutil/float_dsp.h" | |
32 | #include "libavutil/intreadwrite.h" | |
33 | #include "libavutil/log.h" | |
34 | #include "libavutil/opt.h" | |
35 | #include "avcodec.h" | |
36 | #include "audio_frame_queue.h" | |
37 | #include "internal.h" | |
38 | #include "mpegaudio.h" | |
39 | #include "mpegaudiodecheader.h" | |
40 | ||
41 | #define BUFFER_SIZE (7200 + 2 * MPA_FRAME_SIZE + MPA_FRAME_SIZE / 4+1000) // FIXME: Buffer size to small? Adding 1000 to make up for it. | |
42 | ||
43 | typedef struct LAMEContext { | |
44 | AVClass *class; | |
45 | AVCodecContext *avctx; | |
46 | lame_global_flags *gfp; | |
47 | uint8_t *buffer; | |
48 | int buffer_index; | |
49 | int buffer_size; | |
50 | int reservoir; | |
51 | int joint_stereo; | |
52 | int abr; | |
53 | float *samples_flt[2]; | |
54 | AudioFrameQueue afq; | |
55 | AVFloatDSPContext fdsp; | |
56 | } LAMEContext; | |
57 | ||
58 | ||
59 | static int realloc_buffer(LAMEContext *s) | |
60 | { | |
61 | if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) { | |
62 | int new_size = s->buffer_index + 2 * BUFFER_SIZE, err; | |
63 | ||
64 | av_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size, | |
65 | new_size); | |
66 | if ((err = av_reallocp(&s->buffer, new_size)) < 0) { | |
67 | s->buffer_size = s->buffer_index = 0; | |
68 | return err; | |
69 | } | |
70 | s->buffer_size = new_size; | |
71 | } | |
72 | return 0; | |
73 | } | |
74 | ||
75 | static av_cold int mp3lame_encode_close(AVCodecContext *avctx) | |
76 | { | |
77 | LAMEContext *s = avctx->priv_data; | |
78 | ||
79 | av_freep(&s->samples_flt[0]); | |
80 | av_freep(&s->samples_flt[1]); | |
81 | av_freep(&s->buffer); | |
82 | ||
83 | ff_af_queue_close(&s->afq); | |
84 | ||
85 | lame_close(s->gfp); | |
86 | return 0; | |
87 | } | |
88 | ||
89 | static av_cold int mp3lame_encode_init(AVCodecContext *avctx) | |
90 | { | |
91 | LAMEContext *s = avctx->priv_data; | |
92 | int ret; | |
93 | ||
94 | s->avctx = avctx; | |
95 | ||
96 | /* initialize LAME and get defaults */ | |
97 | if (!(s->gfp = lame_init())) | |
98 | return AVERROR(ENOMEM); | |
99 | ||
100 | ||
101 | lame_set_num_channels(s->gfp, avctx->channels); | |
102 | lame_set_mode(s->gfp, avctx->channels > 1 ? s->joint_stereo ? JOINT_STEREO : STEREO : MONO); | |
103 | ||
104 | /* sample rate */ | |
105 | lame_set_in_samplerate (s->gfp, avctx->sample_rate); | |
106 | lame_set_out_samplerate(s->gfp, avctx->sample_rate); | |
107 | ||
108 | /* algorithmic quality */ | |
109 | if (avctx->compression_level == FF_COMPRESSION_DEFAULT) | |
110 | lame_set_quality(s->gfp, 5); | |
111 | else | |
112 | lame_set_quality(s->gfp, avctx->compression_level); | |
113 | ||
114 | /* rate control */ | |
115 | if (avctx->flags & CODEC_FLAG_QSCALE) { // VBR | |
116 | lame_set_VBR(s->gfp, vbr_default); | |
117 | lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA); | |
118 | } else { | |
119 | if (avctx->bit_rate) { | |
120 | if (s->abr) { // ABR | |
121 | lame_set_VBR(s->gfp, vbr_abr); | |
122 | lame_set_VBR_mean_bitrate_kbps(s->gfp, avctx->bit_rate / 1000); | |
123 | } else // CBR | |
124 | lame_set_brate(s->gfp, avctx->bit_rate / 1000); | |
125 | } | |
126 | } | |
127 | ||
128 | /* do not get a Xing VBR header frame from LAME */ | |
129 | lame_set_bWriteVbrTag(s->gfp,0); | |
130 | ||
131 | /* bit reservoir usage */ | |
132 | lame_set_disable_reservoir(s->gfp, !s->reservoir); | |
133 | ||
134 | /* set specified parameters */ | |
135 | if (lame_init_params(s->gfp) < 0) { | |
136 | ret = -1; | |
137 | goto error; | |
138 | } | |
139 | ||
140 | /* get encoder delay */ | |
141 | avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1; | |
142 | ff_af_queue_init(avctx, &s->afq); | |
143 | ||
144 | avctx->frame_size = lame_get_framesize(s->gfp); | |
145 | ||
146 | /* allocate float sample buffers */ | |
147 | if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) { | |
148 | int ch; | |
149 | for (ch = 0; ch < avctx->channels; ch++) { | |
150 | s->samples_flt[ch] = av_malloc(avctx->frame_size * | |
151 | sizeof(*s->samples_flt[ch])); | |
152 | if (!s->samples_flt[ch]) { | |
153 | ret = AVERROR(ENOMEM); | |
154 | goto error; | |
155 | } | |
156 | } | |
157 | } | |
158 | ||
159 | ret = realloc_buffer(s); | |
160 | if (ret < 0) | |
161 | goto error; | |
162 | ||
163 | avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT); | |
164 | ||
165 | return 0; | |
166 | error: | |
167 | mp3lame_encode_close(avctx); | |
168 | return ret; | |
169 | } | |
170 | ||
171 | #define ENCODE_BUFFER(func, buf_type, buf_name) do { \ | |
172 | lame_result = func(s->gfp, \ | |
173 | (const buf_type *)buf_name[0], \ | |
174 | (const buf_type *)buf_name[1], frame->nb_samples, \ | |
175 | s->buffer + s->buffer_index, \ | |
176 | s->buffer_size - s->buffer_index); \ | |
177 | } while (0) | |
178 | ||
179 | static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, | |
180 | const AVFrame *frame, int *got_packet_ptr) | |
181 | { | |
182 | LAMEContext *s = avctx->priv_data; | |
183 | MPADecodeHeader hdr; | |
184 | int len, ret, ch; | |
185 | int lame_result; | |
186 | uint32_t h; | |
187 | ||
188 | if (frame) { | |
189 | switch (avctx->sample_fmt) { | |
190 | case AV_SAMPLE_FMT_S16P: | |
191 | ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data); | |
192 | break; | |
193 | case AV_SAMPLE_FMT_S32P: | |
194 | ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data); | |
195 | break; | |
196 | case AV_SAMPLE_FMT_FLTP: | |
197 | if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) { | |
198 | av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n"); | |
199 | return AVERROR(EINVAL); | |
200 | } | |
201 | for (ch = 0; ch < avctx->channels; ch++) { | |
202 | s->fdsp.vector_fmul_scalar(s->samples_flt[ch], | |
203 | (const float *)frame->data[ch], | |
204 | 32768.0f, | |
205 | FFALIGN(frame->nb_samples, 8)); | |
206 | } | |
207 | ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt); | |
208 | break; | |
209 | default: | |
210 | return AVERROR_BUG; | |
211 | } | |
212 | } else if (!s->afq.frame_alloc) { | |
213 | lame_result = 0; | |
214 | } else { | |
215 | lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index, | |
216 | s->buffer_size - s->buffer_index); | |
217 | } | |
218 | if (lame_result < 0) { | |
219 | if (lame_result == -1) { | |
220 | av_log(avctx, AV_LOG_ERROR, | |
221 | "lame: output buffer too small (buffer index: %d, free bytes: %d)\n", | |
222 | s->buffer_index, s->buffer_size - s->buffer_index); | |
223 | } | |
224 | return -1; | |
225 | } | |
226 | s->buffer_index += lame_result; | |
227 | ret = realloc_buffer(s); | |
228 | if (ret < 0) { | |
229 | av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n"); | |
230 | return ret; | |
231 | } | |
232 | ||
233 | /* add current frame to the queue */ | |
234 | if (frame) { | |
235 | if ((ret = ff_af_queue_add(&s->afq, frame)) < 0) | |
236 | return ret; | |
237 | } | |
238 | ||
239 | /* Move 1 frame from the LAME buffer to the output packet, if available. | |
240 | We have to parse the first frame header in the output buffer to | |
241 | determine the frame size. */ | |
242 | if (s->buffer_index < 4) | |
243 | return 0; | |
244 | h = AV_RB32(s->buffer); | |
245 | if (ff_mpa_check_header(h) < 0) { | |
246 | av_log(avctx, AV_LOG_ERROR, "Invalid mp3 header at start of buffer\n"); | |
247 | return AVERROR_BUG; | |
248 | } | |
249 | if (avpriv_mpegaudio_decode_header(&hdr, h)) { | |
250 | av_log(avctx, AV_LOG_ERROR, "free format output not supported\n"); | |
251 | return -1; | |
252 | } | |
253 | len = hdr.frame_size; | |
254 | av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len, | |
255 | s->buffer_index); | |
256 | if (len <= s->buffer_index) { | |
257 | if ((ret = ff_alloc_packet2(avctx, avpkt, len)) < 0) | |
258 | return ret; | |
259 | memcpy(avpkt->data, s->buffer, len); | |
260 | s->buffer_index -= len; | |
261 | memmove(s->buffer, s->buffer + len, s->buffer_index); | |
262 | ||
263 | /* Get the next frame pts/duration */ | |
264 | ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts, | |
265 | &avpkt->duration); | |
266 | ||
267 | avpkt->size = len; | |
268 | *got_packet_ptr = 1; | |
269 | } | |
270 | return 0; | |
271 | } | |
272 | ||
273 | #define OFFSET(x) offsetof(LAMEContext, x) | |
274 | #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM | |
275 | static const AVOption options[] = { | |
276 | { "reservoir", "use bit reservoir", OFFSET(reservoir), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE }, | |
277 | { "joint_stereo", "use joint stereo", OFFSET(joint_stereo), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE }, | |
278 | { "abr", "use ABR", OFFSET(abr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE }, | |
279 | { NULL }, | |
280 | }; | |
281 | ||
282 | static const AVClass libmp3lame_class = { | |
283 | .class_name = "libmp3lame encoder", | |
284 | .item_name = av_default_item_name, | |
285 | .option = options, | |
286 | .version = LIBAVUTIL_VERSION_INT, | |
287 | }; | |
288 | ||
289 | static const AVCodecDefault libmp3lame_defaults[] = { | |
290 | { "b", "0" }, | |
291 | { NULL }, | |
292 | }; | |
293 | ||
294 | static const int libmp3lame_sample_rates[] = { | |
295 | 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0 | |
296 | }; | |
297 | ||
298 | AVCodec ff_libmp3lame_encoder = { | |
299 | .name = "libmp3lame", | |
300 | .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"), | |
301 | .type = AVMEDIA_TYPE_AUDIO, | |
302 | .id = AV_CODEC_ID_MP3, | |
303 | .priv_data_size = sizeof(LAMEContext), | |
304 | .init = mp3lame_encode_init, | |
305 | .encode2 = mp3lame_encode_frame, | |
306 | .close = mp3lame_encode_close, | |
307 | .capabilities = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME, | |
308 | .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P, | |
309 | AV_SAMPLE_FMT_FLTP, | |
310 | AV_SAMPLE_FMT_S16P, | |
311 | AV_SAMPLE_FMT_NONE }, | |
312 | .supported_samplerates = libmp3lame_sample_rates, | |
313 | .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO, | |
314 | AV_CH_LAYOUT_STEREO, | |
315 | 0 }, | |
316 | .priv_class = &libmp3lame_class, | |
317 | .defaults = libmp3lame_defaults, | |
318 | }; |