Short Repeated Sounds Current Best Practice?

[Cornfritter]

Thanks for the great project!

After reading the docs and past discussion, and trying out some things, I'm wondering if this is the current recommended way to handle multiple repeated (overlapping) short sounds -- the method with the least latency. This is what I'm doing:

• Init 1 ma_engine.
• Init multiple ma_sound objects with ma_sound_init_from_file and ma_sound_init_copy, with MA_SOUND_FLAG_DECODE.
• Set volumes with ma_sound_set_volume.
• Cycle through the ma_sound objects when playing.
• Use ma_sound_uninit and re-init when changing sounds (if all not loaded initially).

Is this right generally? I realize some details might depend on other factors.

Also, can you clarify how to adjust the internal buffer size with this method?

Thanks a lot!

[mackron]

Yes this is the correct way to do it. The most important thing is using MA_SOUND_FLAG_DECODE and to initialize a ma_sound object at load time to keep it in memory, and then use ma_sound_init_copy() off that pre-loaded sound.

Also, can you clarify how to adjust the internal buffer size with this method?

Not quite sure what you mean by this. Sounds don't have the notion of an internal buffer, except for streams which is not configurable. Do you perhaps mean the buffer that controls overall latency of the engine as a whole? periodSizeInMilliseconds in the engine config might help you there.

[Cornfritter]

Great, many thanks!

Yes, that's what I meant. I'll check out periodSizeInMilliseconds.

[jankrassnigg]

I tried to do this, but in ma_sound_init_copy there is this comment:
/* Cloning only works for data buffers (not streams) that are loaded from the resource manager. */

And that's where my code fails now, because I load the data myself and set up the decoders myself.

So what would be the steps to do this, if I don't use the resource manager?

[mackron]

You won't be able to use init_copy() in that case. You'll have to use init_with_data_source(). If your sounds are short and repeating a lot, pre-decode them with ma_decoder, then for each sound initialize a separate ma_audio_buffer instance and keep it alive side-by-side with your sound instance. ma_audio_buffer will not make a copy of the audio data (unless you ask it to) so there will not be a huge amount of overhead.