Re: New 'GObject' as base for GtkObject?

[Tim Janik <timj gtk org>]

On Mon, 20 Dec 1999, David Benson wrote:

> Yes, that was really my point: that the fast equivalent is
> conceivable.
> 
> However I really like the safety the casts provide usually,
> and usually the cost is bearable, it is only in these
> *extremely* direct casts.  In some situations prepare/check
> may be called as many as 400 times per second for each
> source that is going to be mixed together.
> Only there does it add up to enough, whereas usually the cost of
> the safe-casts is negligible.  (And very helpful while
> debugging...  a sound library is very annoying to debug
> if it can never keep up when any debugging info is present)

you are misusing the main loop mechanism there, at the point where you have
400 sources, you shouldn't care about the time taken up by the object casts,
but by the time spent in the glib's main loop and hook list implementations,
doing wild list walks. for your interest, in BSE i use the safety-guarded
object casts a lot, even from within prepare and check without any downfalls.
with proper design for mixer administration and integration into the source/
main loop model, the time spend on some object casts or check/prepare functions
is just laughable compared to what goes on in the innermost mixing loops.
but that assumes that you actually check/prepare for the /dev/dsp devices only,
and collect data from the various audio sources within your own code.

> 
> - Dave
> 

---
ciaoTJ