I got to talk to Thomas Gleixner (kernel developer) when I was in China for the Real-Time Linux Workshop. Thomas works on real-time issues with the standard kernel, and is paid by TimeSys. Thomas gave a talk on high resolution timers that were added starting with
2.6.18. Now, using the standard system call nanosleep(), you will wake up +/- 80 microseconds or so of your programmed wake-up time (assuming you have the priority needed or the system is not too busy). And that was the data Thomas got using old hardware. Of course these things will benefit things like video playback as well, and this is why it was accepted into the main-line kernel.
See this article for more information on the enhancements to 2.6.18, and also more about Thomas Gleixner:
http://linuxdevices.com/news/NS9566944929.html
On 1/17/07, Mel Wilson <m("the-wire.com","mwilson");">> wrote:
Eddie Dawydiuk wrote:
> "We are a very old-school set of engineers that have been
> well served with the saying "if it ain't broke, don't fix it" and
> nothing seems obviously broken with 2.4. This, of course, can change
> if 1) we have a large customer legitimately require and/or fund 2.6
> development or 2) we can be shown real-life advantages of 2.6 that
> outweigh the costs and risks. (Big-Oh notation algorithm
> improvements, 200% performance improvements for 1% of apps, driver
> updates that can only be described as "better...", or other marketing
> dribble are not likely to impress)"
I've read that 2.6.18 (or is it .22?) makes changes that enhance
real-time performance. When I get into this further, I'll have more
to say (or less, if for instance I find out it's all on behalf of
symmetric multi-processing, then who cares?)
[ ... ]
> We'd really rather invest energy in making Linux more transparent in
> customer applications. What we'd like to do is create a more cohesive
> environment with better tools (perhaps Eclipse) and to reduce the need
> to understand Linux as much as possible for customer applications.
[ ... ]
This is a two-edged sword. People who are spared having to understand
Linux will have to understand Technologic's simplifications.
Nonetheless, my task right now is to come up with a "linuxrc-ourown"
that will let us combine the fast-booting mdt kernel and root tree
with our own software on the SD card. The released linuxrc- examples
are pretty clear and pretty transparent. With luck I'll have it
figured out today.
Cheers, Mel.
-- Dr. Don W. Carr J. G. Montenegro 2258 Guadalajara, Mexico +52-333-630-0704 +52-333-836-4500 ext 2930
__._,_.___
![]()
SPONSORED LINKS
__,_._,___
|