Re: [fluid-dev] New Lua module; therefore various questions

[David Henningsson]

On 2014-08-20 01:13, Peter Billam wrote:
> Greetings,
>
> I've just got version 1.0 of a fluidsynth.lua module up on
>    https://rocks.moonscript.org/
>    https://rocks.moonscript.org/modules/peterbillam
> What there is (more or less) works (it was a wrestling match)
> but it's basically a place-holder (e.g. no Sequencer, yet;
> no test script; bad doc).  1.1 will be more serious...

Interesting stuff! I hope you manage to keep it updated over time - not 
that we're putting many releases out there right now :-)

> So, I've got some Fluidsynth API questions...
>
> After
>    new_fluid_synth(settings)
> has been called, the doc says:
>    "The settings parameter is used directly,
>     and should not be modified or freed independently."
> but is it still permissible to change those parameters not
> related to creating the synth, such as audio.file.name ?

You can change any parameter safely, but don't expect all of them to 
take effect.

The sentence means that doing this is wrong:

settings = new_fluid_settings();
synth = new_fluid_synth(settings);
delete_fluid_settings(settings);
fluid_synth_xxx(synth); /* Accessing any fluid_synth_* function might 
use the freed settings object */

> When
>    new_fluid_audio_driver(settings, synth)
> is called, does the settings have to be the same settings
> that was used to create the synth ? or can it be a new
> settings created in order to configure the audio_driver ?

Hmm, good question. I guess it would be okay to have a different object, 
otherwise there wouldn't be another settings pointer in the first place. 
But we should probably look up this better.

> If I want to output to the soundcard, and output to a wav file
> at the same time, do I just create two simultaneous audio_drivers
> connected to the same synth ?
> Or would I have to roll my own, using fluid_synth_write_s16(),
> (and then convert raw->wav) ?

In general, using two audio drivers in parallel is not supported, 
because they will request audio in different chunks, i e, even if 
they're both set to the same sample rate, the clocks will likely drift 
after some time.

So yes, you would have to roll your own.

> When I'm using it to convert from .mid to .wav it runs in real
> time (which I can understand if the "just create two simultaneous
> audio_drivers" trick might get used). But timidity, for example,
> if converting .mid to .wav runs the CPU flat out and doesn't
> bother to wait for anything, and finishes a lot faster.
> Is there a way of asking fluidsynth to do the same ?

You can look in src/fluidsynth.c to see how it's done. There is a 
fluid_file_renderer_t object you can use for this purpose.

// David