Clarity Around Inefficient Datasource Usage Mentioned in Docs

[cdigit]

In documentation, in the Datasources section we can find the following:

Do not use ma_decoder_seek_to_pcm_frame() as a means to reuse a data source to play multiple instances of the same sound simultaneously. Instead, initialize multiple data sources for each instance. This can be extremely inefficient depending on the data source and can result in glitching due to subtle changes to the state of internal filters.

The question is: What exactly is inefficient? Is it using ma_decoder_seek_to_pcm_frame() as a means to reuse a data source to play multiple sound simultaneously or is it initializing multiple data sources for each instance. I would think the former is inefficient, but it's unclear (to me) what the subject of 'this' is?

Also does it make sense to do:

ma_data_source ds = ma_sound_get_data_source(pOrig_sound);
ma_sound_init_from_data_source(&g_engine, ds,  0, NULL, new_sound);

And use new_sound as a copy of pOrig_sound or should one initialize new_sound from the file behind pOrig_sound, for example.

[mackron]

The documentation is saying that the seeking is the inefficient part. Each decoding backend is different. MP3, for example, does not have a built-in mechanism for accelerated seeking which means it needs to brute force it by seeking to the start of the sound and the reading and discarding every sample until the seek target which is extremely inefficient.

It does not make any sense to use the code you suggested. You need to get the idea of reusing data sources out of your head. It's not a thing in miniaudio. It's one data source instance to one sound instance in all cases, no exceptions.

[cdigit]

Thanks for the clarification.

You need to get the idea of reusing data sources out of your head. It's not a thing in miniaudio. It's one data source instance to one sound instance in all cases, no exceptions.

Ok. I think I got the idea from seeing ma_sound_init_from_datasource and thinking I could reuse existing datasources to create copies of sounds. But I get it now!