[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: test (was: [f-cpu] No latches, please !)
On Sat, 16 Feb 2002, Yann Guidon wrote:
> hi !
>
> Juergen Goeritz wrote:
> >
> > Hi guys,
> >
> > very, very intersting discussion.
> my feeling, too :-)
>
> > Let me add another thing:
> >
> > The test strategy of a device is very important. But also
> > very important is WHEN you are able to start a test. The
> > computer world is moving to cluster designs. It may be a
> > very good idea to implement a test strategy that can be
> > activated anytime! I.e. my suggestion is to make selftest
> > a part of the normal operation of the device that can be
> > activated by software!
>
> so this boils down to be able to trigger a "hard reset" with SW.
> that's pretty easy to do, no ?
Why do you want to do a hard reset? Just because Intel & Co.
didn't think of other ways of doing a system test? My opinion
is that it's not the state of the art any more. Just imagine
you have a dual processor on a single SoC and you want this
system to keep running if one of its components is tested.
The hard reset wouldn't do it - you need a new test strategy.
Other example: imagine a 500 processor cluster in an orbit
around Jupiter at high rad doses (I love these examples :)
1/2 of your cluster sleeps to recover from radiation, the
other half is operating. Before sleep you test, after sleep
you test and during operation you test at regular intervals.
JG
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu in the body. http://f-cpu.seul.org/