Event redirection adventures

[Deron Johnson]

Keith Packard wrote On 01/12/06 17:45,:

>  Pick is a whole lot
> faster when accelerated by the 3D hardware anyway.

Just as an aside, Java3D uses software for picking. Some graphics
hardware devices do provide some rudimentary picking support but the
Java3D folks found that it is generally faster and simpler to do the
necessary intersections in the CPU.

>>I think it is always worth to think about ideas that can reduce the
>>number of round trips.

I've been talking with Keith about the round-trip issues for a long time
now and initially I was extremely concerned about the overhead of the
round trip. At one point I was thinking of abandonning the event
round-trip and sucking the entire X grabbing code into the LG Display
Server. But Keith convinced me to stick with the event round trips and I
have been working with such an implementation for a year now and I
haven't observed any adverse performance effects from the round trip.
The cost of the additional overhead is only a few milliseconds (worst
case). My target event latency time is 50ms, so the additional overhead
of the event round trip is negligible.