Re: [LMMS-devel] Oscillator rewrite

[Tobias D. <tob...@gm...>]

Am Montag, 5. Juni 2006 22:34 schrieb Javier Serrano Polo:
> Hi all.
>
> I've recently committed several changes to CVS.
>
> The most noticeable is the triple oscillator's behaviour. Organic plugin
> has been modified too.
> When you alter a knob, waveform, modulation type... changes will apply
> within the next buffer period (every 512 frames, latency 11.6ms).
this is indeed nice, the only thing is that the wave-update-code is *much*=
=20
slower than the previous one, as already discussed in an earlier mail... (y=
ou=20
removed the macros etc.) I don't know what to do now. The thing is that now=
=20
we have branches inside the most-inner loops, which the compiler probably=20
cannot move outside the loops... maybe function-pointers are faster, as the=
re=20
are at least no branches. but all this gives really bad code. May be we=20
should make some benchmarks and decide afterwards about what we're doing.

> Also FM has changed, now it really is frequency modulation (the former
> behaviour was phase modulation). Of course, this affects presets that
> use FM. Help descriptions should be changed: osc1+osc2 actually means
> osc1+output-of(osc2+osc3); oscillator 1 is the one who is modulated with
> output-of(osc2+osc3). I'd like to hear whether the current behaviour is
> the expected one.
this breaks all existing presets and songs! please revert this and add it a=
s a=20
further modulation-method! we *have* to keep an eye on compatibility!

> For developers, oscillator's control has changed. Every parameter is a
> pointer and the plugin is who converts user-friendly values to
> oscillator values. Triple oscillator and organic are already adapted.
that's ok;-)

toby

=2D-=20
Die wahre Beredsamkeit besteht darin, das zu sagen, was zur Sache
geh=F6rt, und eben nur das.
		-- Fran=E7ois de La Rochefoucault