well, in addition to setting stathz to 0, you'd also have to
change the splhigh()s in kern_synch.c and sys_generic.c
to splstatclock() for there to be any benefit to the zs driver,
and that's a lot more intrusive.

so is it just me, or does MI code calling splhigh() seem
pretty bogus in general?  I don't see why MI code should be
saying "it is currently unsafe to process ANY interrupts",
which is how I interpret splhigh().  I'd guess this is just
an overly conservative bit left from long ago, and
really should be made more aggressive at this point.

-Chuck

In message <199710252029.QAA00951@bach.mc.com>, "Gordon W. Ross" writes:
> > >lower processor IPL during processing of level 14 interrupts ...
> 
> I too found the idea interesting, though I suspect we could accomplish
> pretty much the same thing by just leaving stathz=0 on the sparc, so
> that statclock() will be called from the normal clock interrupt on
> level 10.  This means there would be no profiling interrupts in any
> of the interrupt handlers for fd(11), zs(12), or audio(13)] but I
> don't think we really WANT profiling interrupts during any of those.
> 
> This fix would be simple enough so it could go into this release.
> Would anything important be lost with this approach?
> 
> Gordon