3 questions about ma_audio_buffer and ma_sound

[lukasz-szczur]

Hello. I use the code shown below and have three questions.

1) Do I have to call free(decodedOggFile); when stb_vorbis_decode_memory(...) returns -1? | No
2) I know that instances of ma_audio_buffer can share the same memory buffer (eg. decodedOggFile below). Is it required to create ma_audio_buffer for each new ma_sound to play multiple sounds at once? Can I pass one ma_audio_buffer to multiple ma_sound variables?
3) I load and decode a sound beforehand. Does the MA_SOUND_FLAG_DECODE flag passed to ma_sound_init_from_data_source make any difference?

        ma_engine soundEngine;
        ma_audio_buffer audioBuffer;
        ma_sound bgSound;

        std::vector<unsigned char> soundBytes;
        assets.load("publish/bg.ogg", soundBytes);

        short* decodedOggFile;
        int channels = 0, sampleRate = 0;
        int numSamples = stb_vorbis_decode_memory(soundBytes.data(), soundBytes.size(), &channels, &sampleRate, &decodedOggFile);
        if (numSamples == -1)
            return pik::GLLoaderJobState::INTERRUPTED;
        PIK_LOG("numSamples: %d | channels: %d | sampleRate: %d | ", numSamples, channels, sampleRate);
        
        if (ma_engine_init(NULL, &soundEngine) != MA_SUCCESS)
            return pik::GLLoaderJobState::INTERRUPTED;

        ma_audio_buffer_config config = ma_audio_buffer_config_init(ma_format_s16, channels, numSamples, decodedOggFile, nullptr);
        if (ma_audio_buffer_init(&config, &audioBuffer) != MA_SUCCESS)
            return pik::GLLoaderJobState::INTERRUPTED;

       if (ma_sound_init_from_data_source(&soundEngine, &audioBuffer, MA_SOUND_FLAG_DECODE , nullptr, &bgSound) != MA_SUCCESS) 
            return pik::GLLoaderJobState::INTERRUPTED;

        ma_sound_start(&bgSound);

[mackron]

2. Yes, you need a separate ma_audio_buffer for each ma_sound. It's a one-to-one relationship.
3. No, MA_SOUND_FLAG_DECODE is only used when loading from a file path, not when initializing a sound from a data source. You can exclude it.

[lukasz-szczur]

Thank you.
A data source (eg. ma_audio_buffer, ma_decoder, etc.) has to be duplicated for each new ma_sound, but I should share the buffer in memory between data sources. Is that right?

I have one more question. I test my code on an old device with Android 4.4.2 (32-bit, SDK 19) and CPU Qualcomm MSM8926 Snapdragon 400 (32-bit, 2 cores).

Watch the video: https://www.youtube.com/shorts/BhElZVgcw60

On first device the sound is OK. On second device it is distorted. Why? I use the code I have shown here. Do I miss to configure something? I remember that the sound was good with ma_decoder. Do I have to decode a sound for each device? I use the following ogg file: https://drive.google.com/file/d/1EDL04etramopwFm_E8NcSDQNBNXRTi4G/view?usp=sharing

[mackron]

Could be a sample rate inconsistency. When I first wrote ma_audio_buffer it didn't have the notion of a sample rate. I added support for it later, but I couldn't add it as a parameter to ma_audio_buffer_config_init() because it would have broken the API (I'll be adding it in version 0.12). You can still set the sample rate, however. Try adding this line after your call to ma_audio_buffer_config_init():

config.sampleRate = sampleRate;

I'm not 100% sure if that will fix the problem, but it's something worth trying.

And yes, you should share the underlying data buffer between multiple ma_audio_buffer objects. But you need a separate ma_audio_buffer for each ma_sound.

[lukasz-szczur]

config.sampleRate = sampleRate;

The above line fixed my problem. Thank you very much!

Do the sharing principles also work for ma_decoder? A separate ma_decoder for each ma_sound. Memory buffer shared between decoders.

[mackron]

Yes it applies for all data sources, and ma_decoder is a data source.