On 09/09/2002 23:20:24, Emmanuel Dreyfus <manu@netbsd.org> said:

> One last issue before converting the keyboard handling to this: If I run
> top with a null delay (In top, hit s then request 0 seconds), I can see
> the time spent in interrput raise to about 80% when I touch the screen.
> I wonder if j720tp_disable really does its job.

I checked this. The tp interrupt is not called anymore once I disable
it, so j720tp_disable really works. The interrupt swarm seems only
related to callout activity.

And reading more code, this is not surprising, since there a 5ms delay
when talking to the ssp in j720ssp_readwrite. If I request a 25Hz poll,
I get 125ms of CPU time wasted per second. It seems a bit too much.

Similarly we loose at least 10ms waiting when we read a keyboard key.

Don't we have a better way of handling this? 10ms is a time slice for
the scheduler, it seems enormous to me.=20

--=20
Emmanuel Dreyfus.
Windows 95, Windows 98, Windows 2000, Windows Millenium
D=E9sormais, les OS portent la date de peremption dans leurs noms.
manu@netbsd.org