让 OpenAL 也支持 S16 Planar(辅以 FFmpeg)

正在制作某物品,现在做到音频部分了。

原本要采用 SDL2_mixer 的,不过实验结果表明其失真非常严重,还带有大量的电噪声。不知道是不是我打开的方式不对……

一气之下去看 OpenAL,结果吃了闭门羹(维护中,只有 mailing list 和 specification)。转投 FMOD,不过又考虑到其授权方式,还是放弃了。最终回到 OpenAL。使用的是 OpenAL-Soft。

OpenAL 呢,好的方面是开源+授权,坏的方面……呃,至少在刚刚的测试中,代码维护甚至没有 SDL 好。直接编译 .c 示例失败,耍小聪明改成 .cpp 拿去编译才成功。

在接下来的代码中,需要用到 OpenAL-Soft(1.15.1)和 FFmpeg。

看 OpenAL-Soft 自带的示例 alstream.c。为了方便起见,接下来的 C 源代码文件全部改成 C++ 源代码文件去……同时不要忘了在 FFmpeg 的头文件上下加 extern "C"!(为什么他们不考虑这一点?)

好,编译示例,运行。(注意,各种 dependencies 这里就不提了。)随便选择一个含有音频的、可以被 FFmpeg 解码的文件。

不对啊!很有可能出现以下错误信息:

Opened "OpenAL Soft"
AL_SOFT_buffer_samples supported!
Unsupported ffmpeg sample format: s16p
Error getting audio info for 01.mpg
Done.

这是……怎么回事?经过测试,SDL_mixer 可以播放同一个文件,不过正如之前所说的,失真&噪声。看其采样格式:S16P(Signed 16-bit, Planar←平面?)。再看源代码,S16(Signed 16-bit)是支持的。(当然,如果强制将那几个 if 修改一下的话,你会听到神奇的东西……)S16 和 S16P 的不同点是在于数据的排列方式,前者是相邻连续排列,后者是分离排列。但是现在有相当多的音频文件采用 planar 的方案,不仅是 S16,U8、S32、F32、F64 都有对应的 planar 方式。现在,目标就是:让这个示例支持 planar。

思路很简单。我的上一篇随笔中,有一个 AudioResampling() 函数,这里直接拿来用吧!(秉持拿来主义!鲁迅先生不谢。)

 

接下来就是好戏了。

又试验了一下,播放 U8/Mono 的时候出现崩溃,不知道原因。调试的时候内存是越界的。

 

先是添加对 libswresample 和 libavutil(要用到 opt_* 函数)的包含(别忘了添加对应的库):

1 #ifdef __cplusplus
2 extern "C" {
3 #endif
4 #include "libavutil/opt.h"
5 #include "libswresample/swresample.h"
6 #ifdef __cplusplus
7 }
8 #endif

然后是修改 MyStream 的定义:

 1 struct MyStream {
 2     AVCodecContext *CodecCtx;
 3     int StreamIdx;
 4 
 5     struct PacketList *Packets;
 6 
 7     AVFrame *Frame;
 8 
 9     // FrameData 没什么用了,不过为了保持代码结构,还是保留下来,其作用由 FrameBuffer 代替
10     const uint8_t *FrameData;
11     const uint8_t FrameBuffer[FRAME_BUFFER_SIZE];
12 
13     size_t FrameDataSize;
14 
15     FilePtr parent;
16 };

可以先定义一下 FRAME_BUFFER_SIZE:

1 // MP3 每一帧的大小是4608,所以如果设定成4096(一般音频可以播放)的话会造成溢出、崩溃
2 #define FRAME_BUFFER_SIZE            (4800)

直接插入 AudioResampling() 函数(如果对这错误的时态感到别扭,改一下就好了),添加重采样支持:

  1 static int AudioResampling(AVCodecContext * audio_dec_ctx,
  2                     AVFrame * pAudioDecodeFrame,
  3                     int out_sample_fmt,
  4                     int out_channels,
  5                     int out_sample_rate,
  6                     uint8_t* out_buf)
  7 {
  8     SwrContext * swr_ctx = NULL;
  9     int data_size = 0;
 10     int ret = 0;
 11     int64_t src_ch_layout = audio_dec_ctx->channel_layout;
 12     int64_t dst_ch_layout = AV_CH_LAYOUT_STEREO;
 13     int dst_nb_channels = 0;
 14     int dst_linesize = 0;
 15     int src_nb_samples = 0;
 16     int dst_nb_samples = 0;
 17     int max_dst_nb_samples = 0;
 18     uint8_t **dst_data = NULL;
 19     int resampled_data_size = 0;
 20 
 21     swr_ctx = swr_alloc();
 22     if (!swr_ctx)
 23     {
 24         printf("swr_alloc error \n");
 25         return -1;
 26     }
 27 
 28     src_ch_layout = (audio_dec_ctx->channels ==
 29                      av_get_channel_layout_nb_channels(audio_dec_ctx->channel_layout)) ?
 30                      audio_dec_ctx->channel_layout :
 31                      av_get_default_channel_layout(audio_dec_ctx->channels);
 32 
 33     if (out_channels == 1)
 34     {
 35         dst_ch_layout = AV_CH_LAYOUT_MONO;
 36         //printf("dst_ch_layout: AV_CH_LAYOUT_MONO\n");
 37     }
 38     else if (out_channels == 2)
 39     {
 40         dst_ch_layout = AV_CH_LAYOUT_STEREO;
 41         //printf("dst_ch_layout: AV_CH_LAYOUT_STEREO\n");
 42     }
 43     else
 44     {
 45         dst_ch_layout = AV_CH_LAYOUT_SURROUND;
 46         //printf("dst_ch_layout: AV_CH_LAYOUT_SURROUND\n");
 47     }
 48 
 49     if (src_ch_layout <= 0)
 50     {
 51         printf("src_ch_layout error \n");
 52         return -1;
 53     }
 54 
 55     src_nb_samples = pAudioDecodeFrame->nb_samples;
 56     if (src_nb_samples <= 0)
 57     {
 58         printf("src_nb_samples error \n");
 59         return -1;
 60     }
 61 
 62     av_opt_set_int(swr_ctx, "in_channel_layout", src_ch_layout, 0);
 63     av_opt_set_int(swr_ctx, "in_sample_rate", audio_dec_ctx->sample_rate, 0);
 64     av_opt_set_sample_fmt(swr_ctx, "in_sample_fmt", audio_dec_ctx->sample_fmt, 0);
 65 
 66     av_opt_set_int(swr_ctx, "out_channel_layout", dst_ch_layout, 0);
 67     av_opt_set_int(swr_ctx, "out_sample_rate", out_sample_rate, 0);
 68     av_opt_set_sample_fmt(swr_ctx, "out_sample_fmt", (AVSampleFormat)out_sample_fmt, 0);
 69 
 70     if ((ret = swr_init(swr_ctx)) < 0) {
 71         printf("Failed to initialize the resampling context\n");
 72         return -1;
 73     }
 74 
 75     max_dst_nb_samples = dst_nb_samples = av_rescale_rnd(src_nb_samples,
 76                                                          out_sample_rate, audio_dec_ctx->sample_rate, AV_ROUND_UP);
 77     if (max_dst_nb_samples <= 0)
 78     {
 79         printf("av_rescale_rnd error \n");
 80         return -1;
 81     }
 82 
 83     dst_nb_channels = av_get_channel_layout_nb_channels(dst_ch_layout);
 84     ret = av_samples_alloc_array_and_samples(&dst_data, &dst_linesize, dst_nb_channels,
 85                                              dst_nb_samples, (AVSampleFormat)out_sample_fmt, 0);
 86     if (ret < 0)
 87     {
 88         printf("av_samples_alloc_array_and_samples error \n");
 89         return -1;
 90     }
 91 
 92 
 93     dst_nb_samples = av_rescale_rnd(swr_get_delay(swr_ctx, audio_dec_ctx->sample_rate) +
 94                                     src_nb_samples, out_sample_rate, audio_dec_ctx->sample_rate, AV_ROUND_UP);
 95     if (dst_nb_samples <= 0)
 96     {
 97         printf("av_rescale_rnd error \n");
 98         return -1;
 99     }
100     if (dst_nb_samples > max_dst_nb_samples)
101     {
102         av_free(dst_data[0]);
103         ret = av_samples_alloc(dst_data, &dst_linesize, dst_nb_channels,
104                                dst_nb_samples, (AVSampleFormat)out_sample_fmt, 1);
105         max_dst_nb_samples = dst_nb_samples;
106     }
107 
108     if (swr_ctx)
109     {
110         ret = swr_convert(swr_ctx, dst_data, dst_nb_samples,
111                           (const uint8_t **)pAudioDecodeFrame->data, pAudioDecodeFrame->nb_samples);
112         if (ret < 0)
113         {
114             printf("swr_convert error \n");
115             return -1;
116         }
117 
118         resampled_data_size = av_samples_get_buffer_size(&dst_linesize, dst_nb_channels,
119                                                          ret, (AVSampleFormat)out_sample_fmt, 1);
120         if (resampled_data_size < 0)
121         {
122             printf("av_samples_get_buffer_size error \n");
123             return -1;
124         }
125     }
126     else
127     {
128         printf("swr_ctx null error \n");
129         return -1;
130     }
131 
132     memcpy(out_buf, dst_data[0], resampled_data_size);
133 
134     if (dst_data)
135     {
136         av_freep(&dst_data[0]);
137     }
138     av_freep(&dst_data);
139     dst_data = NULL;
140 
141     if (swr_ctx)
142     {
143         swr_free(&swr_ctx);
144     }
145     return resampled_data_size;
146 }

修改 getAVAudioData() 函数:

 1 uint8_t *getAVAudioData(StreamPtr stream, size_t *length)
 2 {
 3     int got_frame;
 4     int len;
 5 
 6     if(length) *length = 0;
 7 
 8     if(!stream || stream->CodecCtx->codec_type != AVMEDIA_TYPE_AUDIO)
 9         return NULL;
10 
11 next_packet:
12     if(!stream->Packets && !getNextPacket(stream->parent, stream->StreamIdx))
13         return NULL;
14 
15     /* Decode some data, and check for errors */
16     avcodec_get_frame_defaults(stream->Frame);
17     while((len=avcodec_decode_audio4(stream->CodecCtx, stream->Frame,
18                                      &got_frame, &stream->Packets->pkt)) < 0)
19     {
20         struct PacketList *self;
21 
22         /* Error? Drop it and try the next, I guess... */
23         self = stream->Packets;
24         stream->Packets = self->next;
25 
26         av_free_packet(&self->pkt);
27         av_free(self);
28 
29         if(!stream->Packets)
30             goto next_packet;
31     }
32 
33     if(len < stream->Packets->pkt.size)
34     {
35         /* Move the unread data to the front and clear the end bits */
36         int remaining = stream->Packets->pkt.size - len;
37         memmove(stream->Packets->pkt.data, &stream->Packets->pkt.data[len],
38                 remaining);
39         memset(&stream->Packets->pkt.data[remaining], 0,
40                stream->Packets->pkt.size - remaining);
41         stream->Packets->pkt.size -= len;
42     }
43     else
44     {
45         struct PacketList *self;
46 
47         self = stream->Packets;
48         stream->Packets = self->next;
49 
50         av_free_packet(&self->pkt);
51         av_free(self);
52     }
53 
54     if(!got_frame || stream->Frame->nb_samples == 0)
55         goto next_packet;
56 
57 
58     // 在这里插入重新采样代码
59 
60     *length = AudioResampling(stream->CodecCtx, stream->Frame, AV_SAMPLE_FMT_S16, stream->Frame->channels, stream->Frame->sample_rate, const_cast<uint8_t *>(stream->FrameBuffer));
61 
62     /* Set the output buffer size */
63     /*
64     *length = av_samples_get_buffer_size(NULL, stream->CodecCtx->channels,
65                                                stream->Frame->nb_samples,
66                                                stream->CodecCtx->sample_fmt, 1);
67 
68     return stream->Frame->data[0];
69     */
70 
71     return const_cast<uint8_t *>(stream->FrameBuffer);
72 }

最后是 getAVAudioInfo() 函数,我们要让它允许 planar 音频输入:

 1 int getAVAudioInfo(StreamPtr stream, ALuint *rate, ALenum *channels, ALenum *type)
 2 {
 3     if(!stream || stream->CodecCtx->codec_type != AVMEDIA_TYPE_AUDIO)
 4         return 1;
 5 
 6     /* Get the sample type for OpenAL given the format detected by ffmpeg. */
 7     if(stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_U8 || stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_U8P)
 8         *type = AL_UNSIGNED_BYTE_SOFT;
 9     else if (stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_S16 || stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_S16P)
10         *type = AL_SHORT_SOFT;
11     else if(stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_S32 || stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_S32P)
12         *type = AL_INT_SOFT;
13     else if(stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_FLT || stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_FLTP)
14         *type = AL_FLOAT_SOFT;
15     else if(stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_DBL || stream->CodecCtx->sample_fmt == AV_SAMPLE_FMT_DBLP)
16         *type = AL_DOUBLE_SOFT;
17     else
18     {
19         fprintf(stderr, "Unsupported ffmpeg sample format: %s\n",
20                 av_get_sample_fmt_name(stream->CodecCtx->sample_fmt));
21         return 1;
22     }
23 
24     /* Get the OpenAL channel configuration using the channel layout detected
25      * by ffmpeg. NOTE: some file types may not specify a channel layout. In
26      * that case, one must be guessed based on the channel count. */
27     if(stream->CodecCtx->channel_layout == AV_CH_LAYOUT_MONO)
28         *channels = AL_MONO_SOFT;
29     else if(stream->CodecCtx->channel_layout == AV_CH_LAYOUT_STEREO)
30         *channels = AL_STEREO_SOFT;
31     else if(stream->CodecCtx->channel_layout == AV_CH_LAYOUT_QUAD)
32         *channels = AL_QUAD_SOFT;
33     else if(stream->CodecCtx->channel_layout == AV_CH_LAYOUT_5POINT1_BACK)
34         *channels = AL_5POINT1_SOFT;
35     else if(stream->CodecCtx->channel_layout == AV_CH_LAYOUT_7POINT1)
36         *channels = AL_7POINT1_SOFT;
37     else if(stream->CodecCtx->channel_layout == 0)
38     {
39         /* Unknown channel layout. Try to guess. */
40         if(stream->CodecCtx->channels == 1)
41             *channels = AL_MONO_SOFT;
42         else if(stream->CodecCtx->channels == 2)
43             *channels = AL_STEREO_SOFT;
44         else
45         {
46             fprintf(stderr, "Unsupported ffmpeg raw channel count: %d\n",
47                     stream->CodecCtx->channels);
48             return 1;
49         }
50     }
51     else
52     {
53         char str[1024];
54         av_get_channel_layout_string(str, sizeof(str), stream->CodecCtx->channels,
55                                      stream->CodecCtx->channel_layout);
56         fprintf(stderr, "Unsupported ffmpeg channel layout: %s\n", str);
57         return 1;
58     }
59 
60     *rate = stream->CodecCtx->sample_rate;
61 
62     return 0;
63 }

 

嗯,基本上就可以了。现在播放的话,对应的 planar 是不会显示出来的,因为显示调用的是 alhelpers.cpp 的 GetFormat(),而它是按照 OpenAL 的格式输出的。

不过这不影响播放嘛。

posted @ 2014-04-05 17:01  Grid Science  阅读(1015)  评论(0编辑  收藏  举报