my next "spare time" project....

[Mark L. Chang]

On Wed, 9 May 2001, Marc Swanson wrote:

> Agreed.  If the engine is running at 6000 rpms, that's one revolution of the
> crank every 1/100th of a second.  The severity of the CPU running kernel
> interrupts would depend on the duration of the interrupt.  As far as I know
> on a bare bones kernel that should only account for a few thousandths of a
> second at the worst.. if this is the case then there isn't too much to worry
> about.  As you say though.. probably best to run a real-time OS just to be
> "safe...er".

A little looking around and it seems that some tests have shown kernel
blocking times for interrupts (ie. when the kernel can't service any
interrupts) on the order of 150us if things are done smartly. This goes
down to around 100us or so for RAM-based filesystems.

Find more information at:
http://www.mvista.com/realtime/latency and http://www.mvista.com/realtime

So, if you are doing pulsed injection, you need a resolution on the order
of 10-100 ms, and you'll get +/- 200us or so. Guess it isn't so bad.

> > LinuxBIOS helps, as does
> > Bob's "burn to ROM" suggestion. Okay. I'll concede this one.
>
> Cool, I'm wearing down on you ;-)

LinuxBIOS reports from power on to on the network and at a prompt in 3
seconds, which is nifty:
http://www.acl.lanl.gov/linuxbios/news/index.html

> yeah, that's the reason I even considered it... Even getting a
> simulation to work would be nifty.  As far as I know nobody has tried
> anything like this... which doesn't speak too well for the strength of
> the idea....

As do some of the comments on the list :>... now what was it that Huw was
saying about waxing rotors? :)

-- 
http://www.mchang.org/
http://decss.zoy.org/