[Leaplist] need some jiffies

Richard F. Ostrow Jr. kshots at warfaresdl.com
Thu Oct 11 10:45:22 EDT 2007


Right... I'm sure you already know this, but you cannot use an x86 PC as a
real-time machine under linux. It may usually get close to such, but you
cannot absolutely gaurantee that it will at all times. For that you need a
RTOS or a microcontroller.

I think you had posted earlier about using some variant of an RT system...
I'd assumed before that you meant it wasn't necessary to *always* be
exactly 100 Hz or whatever you're aiming at, but just verifying :).
-- 
Life without passion is death in disguise

On Thu, October 11, 2007 10:11 am, Andrew wrote:
>
> On Oct 9, 2007, at 8:01 PM, John Simpson wrote:
>
>> On 2007-10-09, at 1052, Richard F. Ostrow Jr. wrote:
>>>
>>> Hmm... the last time I'd heard of the word "jiffies" was back when
>>> I was
>>> coding on the C64 when I was 12 years old. I believe the
>>> definition back
>>> then was more linked to machine cycles than any measurement of
>>> time... so
>>> a jiffy on one machine will be different than a jiffy on another
>>> machine(hence a "jiffy" is not equal to x microseconds).
>>
>> that's exactly what it is under linux. the kernel programs the
>> timer chip to send interrupts 100 times per second. (for old-school
>> DOS hackers, this is the same timer chip that DOS programs for 18.2
>> times per second. it uses hardware interrupt line 0, which maps to
>> the processor's INT 0x08.)
>
> Something to be aware of is that you cannot absolutely count on it
> being 100 times per second.  On the Alpha, it has always been 1000
> times per second, and the IA-64 port followed that model.  IIRC, Red
> Hat increased the timer resolution in their kernels to 1000 per
> second as well, for better performance characteristics.  Also, I have
> not been closely following the "tickless kernel" work, so there may
> be some additional caveats that come into play there.
> _______________________________________________
> Leaplist mailing list
> Leaplist at leap-cf.org
> http://lists.leap-cf.org/mailman/listinfo/leaplist
>




More information about the Leaplist mailing list