cvs commit: src/sys/isa psm.c
scottl at freebsd.org
Thu Dec 11 14:29:16 PST 2003
On Thu, 11 Dec 2003, Brian F. Feldman wrote:
> Alfred Perlstein <alfred at FreeBSD.org> wrote:
> > * Eivind Eklund <eivind at FreeBSD.org> [031211 04:08] wrote:
> > > On Thu, Dec 11, 2003 at 03:28:11AM -0800, Alfred Perlstein wrote:
> > > > Log:
> > > > Significantly reduce the "jitter" that is typical for PS/2 mice
> > > > when using a KVM.
> > > [...]
> > > >
> > > > The actual solution that appears to offer the best clamping of
> > > > jitter is to buffer the mouse packets if we've not seen mouse
> > > > activity for more than .5 seconds. Then waiting to flush that data
> > > > for 1/20th of a second. If within that 20th of a second we get any
> > > > packets that do fail the weak test we drop the entire queue and
> > > > back off accepting data from the mouse for 2 seconds and then repeat
> > > > the whole deal.
> > >
> > > Have you tested this with enough high speed interactive games? 50ms
> > > (1/20s) is three to four frames, and quite a lot of delaying input.
> > > There are a number of types of games that require faster response than
> > > this. I'm not entirely sure how it influences most mouse-run games - I
> > > know that at least some 2D fighting games with joystick absolutely
> > > required 1-frame (20ms) response on 50Hz displays.
> > There is only a delay if the mouse has been idle for .5 seconds,
> > otherwise the reaction is immediate. I guess we could tune that
> > up to 2 seconds, it's also tunable via sysctls. Feedback, testing
> > and tweaking are encouraged, this is -current afterall. :)
> I can't imagine that most people that are playing these sorts of games use a
> KVM switch, and as such probably have no need for a PS/2 mouse and would
> want to use a USB mouse instead.
This is a bold and mostly useless statement. I don't think that it's at
all unreasonable to have a deskto/game machine share a switch with development
boxes or servers. I'm in this position myself right now.
> Still, the default should probably be
> several seconds at least; when I used a KVM I don't think I ever managed to
> switch from and back to one of the computers in less than a few seconds :)
Again, this is lacks broad perspective. Good switches (the non-Belkin
kind) can switch rather quickly.
I think that Alfred's approach is going to require a lot of feedback and
tuning to get right, though I'm not saying that it is the wrong approach.
As always, we might want to look at Linux to see what solutions have been
More information about the cvs-all