Joerg 'Herkules' Plewe wrote:
> > > correct?). Should I implement my own waiting cycles (i.e. don't use
> > > sleep()) to be sure of accurate timing?
> > This one is easy to answer : do not expect more than 1/10s in
> > System.currentMillis.
> > Be happy, it's worst in 95/98 ;-).
>
> Checked that out:
>
> 60ms on Win98
> 16ms on NT
> 1-2ms on Linux
>
> But why the hell....
Here's an Interesting extra.
I'm am testing on a top-of-the-line NT and a HP-UX box (unix) both using
identical graphics hardware from HP.
In my test, where I request a sleep in the postSwap() (really needs to be
preSwap but the closets is postRender)
I get this.....
On NT my frames per second work of to be something like XX.234365 - XX.509237
On the HP-UX a clean XX.0
for target rates from 10htz to 30htz on empty scenes as well as scenes that
can be rendered in the alloted time slice.
We are trying to determine our expectations for a multi-PC system and how well
we can sync frames across machines. Unfortunately, it looks as though the
regular thread.sleep() calls are not going to work on NT. :-(
We are planning to code in our own "sleep" loop and see how accurate it can
get. If you have done this please share with the group!
--
TTFN
__________________________________________________________
Shawn Kendall Full Sail Real World Education
Course Instructor 3300 University BLVD
Virtual Reality Winter Park FL 32792
[EMAIL PROTECTED] http://www.fullsail.com
__________________________________________________________
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
