[ros-kernel] Process scheduler / timer resolution

[Mike Nordell]

welch wrote:

> Quoting Mike Nordell [e-mail snipped]:
[please don't quote e-mail adresses]

> > That is never the case. Ever. Performance is always reduced
> > with increased interrupt rates. Always.
> >
> Batch performance is reduced; interactive performance might improve if
> there are cpu bound tasks running in the background.

Any task (given CPU-time) have impact of the overall system performance,
obviously. However, that's not the point. The point is that the more
interrupts you are handling, the less time is left for threads to do their
work. That's not some kind of "soft" truth - it's an absolute. Not only can
I prove it empirically, I can also prove it mathematically.

Interactive _percieved performance_ might change, but that has nothing to do
with real performance but instead latency, and we then are stepping into
areas like "mouse lag in X under VMWare sucks" to display that an
input+display-update rate less than 10 FPS is annoying.


/Mike