Karl W. MacMillan wrote:
(Benno wrote:)
> > Although better than standard kernels we are still about 20times worse
> > than Ingo's 2.2.10/2.2.15 patches, and speaking in other OSes terms
> > still 5-7 times worse than Windows/MacOs.
> > So far for the 2.4 kernels.
> >
>
> I am curious where the worst case numbers for Windows/Mac come from. My
> experiences with real world apps running on these platforms is that the
> latency is quite a bit worse than linux (these are apps that don't use
> special techniques to acheive lower latency - i.e. gigasampler). Also,
> the only tests I have seen on Windows indicated that its average
> latency was in the 150ms range (this was an ICMC paper by Roger Dannenberg
> - I may be remembering that number incorrectly). Just wondering.
We saw worst-case latencies on that order for Windows 95 and NT 4.
Windows 98 RC0 was markedly better. NT 5 beta1 was not, but this was
a long time before the release of Win2K. The paper is up at
http://www.cs.cmu.edu/~eli/papers/eub-rbd-icmc98.pdf
(My understanding, though I'm not a Mac person, is that MacOS (pre-X)
should not be conjoined-with-a-slash to Windows (post-3.1) because
one's not preemptive and the other is.)
--
Eli Brandt | [EMAIL PROTECTED] | http://www.cs.cmu.edu/~eli/
