Apologies everyone... didn't send this to the correct address. Alex. ----- Original Message ----- From: "Christoph Reichenbach" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Sunday, September 23, 2001 8:23 PM Subject: [freesci-develop] Re: process_sound_events() (fwd) > Hi, > > this appears not to have been received by Listar... > > > ---------- Forwarded message ---------- > Date: Sat, 22 Sep 2001 18:25:40 +0200 (MET DST) > From: Christoph Reichenbach <[EMAIL PROTECTED]> > To: Alexander R Angas <[EMAIL PROTECTED]> > Subject: Re: process_sound_events() > > Hi, > > > > I was just looking at process_sound_events() which receives signals back > > from the sound server. It appears to be called in a fairly arbitrary > > fashion. Am I correct on this? > > Yes. It's called in a number of kernel functions that are known to be used > regularly (particularly Wait, DoSound, and Animate). > > > If so, is there any way it can be improved? > > We can't do asynchronous stuff portably here, since heap reads/writes are > not atomic on some machines and mutexes would be overkill. A possible > improvement would be to use a counter that calls process_sound_events() > every 250 instructions or so, in addition to the current calls (DoSound > tends to trigger sound events that can be processed immediately after it > has sent its commands, whereas Wait and Animate may take some time and are > therefore good places for checks. > > llap, > Christoph > > > >
