Re: [PATCH] 2.6.17-rt add clockevent to ixp4xx

[Milan Svoboda]

Kevin,

> On Thu, 2006-07-13 at 10:53 +0000, Milan Svoboda wrote:
> 
> > there are patches that enable clock event on ixp4xx platform. This 
should
> > enable high resolution timers... Option for hrtimers in menuconfig is
> > also enabled.
> 
> Milan,
> 
> I've also done a clockevent driver for ixp4xx and it looks pretty much
> like yours.  I've been waiting to submit as Thomas has recently reworked
> the clockevent layer a bit in his -hrt-dyntick patchset.
> 
> Here are a couple comments on your patchset
> 
> - since you've registered the clockevent with CAP_TICK, both the
> arch interrupt handler and the clockevent handler are handling the tick
> and calling do_timer().  While I don't think this will negatively affect
> timekeeping, it's unncessary overhead.

Yeah, I had a headache from this mistake, I found it a few minutes
after sendig mail ;-)

> - why the addition of clockevent_hz2mult()? since shift is 32, you
> could use existing div_sc32()

I didn't know about div_sc32... Simply, I have no problem with using
div_sc32 now when I know that is exists ;-)

> The attached patch is a combination of my patch and yours and addresses
> my comments above.  I simply removed the CLOCK_CAP_TICK and removed your
> clockevent_hz2mult() and used div_sc32().
> 

Good.

> Also, below are a few runs of the nanosleep_jitter test that comes with
> the sourceforge HRT test suite.  Something strange is that with the
> nanosleep_jitter test, I only see max sleeps of ~300-400 usec but with
> your test I see max sleeps up to 1.3 msec.
> 

Maybe my test program is not as good as nanosleep_jitter :-)

I did patch against -hrt-dyntick (see lkml, this thread) wich is almost 
the same
as this one against -rt (for ixp4xx part only) and I found that when timer 
is
loaded with IXP4XX_OST_ONE_SHOT the latency time suddenly drops to ~30usec instead
of ~2000usec! I'd like know what's the reason for this...

Best regards,
Milan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/