MUSIC-DSP digest 321

Eli Brandt eli at gs160.sp.cs.cmu.edu
Thu Sep 10 11:23:34 EDT 1998


Martin Fay wrote:
> That isn't strictly true...and idling 95 box with no win16 apps 
> running never seems to show jitter this bad.

Dunno about "never".  I saw 396 ms on an idle freshly-booted 95 box
with the usual tricks (no network, no CD-ROM polling, that type of
thing).  And I bet you'll see higher if you want to wait long enough.
(Should have mentioned, though, that a W95 DSound app can quite well
do 20-ms latency by using a 16-bit timer.)

> This relates back to the 
> jitter being caused by the mutex used to control access to 16bit OS 
> services. A simple demonstration of this is moving the window of a 
> win16 app around...in fact that can block timer callbacks 
> indefinately.

I think the UI-update mutex can cause similar problems even in the
absence of 16-bit apps.  (Those slide-out menus in W98 will hammer the
timer for 200 ms, I can tell you.)  Last I heard, nobody knows _what_
all the jitter in NT4 is caused by.

-- 
     Eli Brandt  |  eli+ at cs.cmu.edu  |  http://www.cs.cmu.edu/~eli/






More information about the music-dsp mailing list