is "preemptive kernel" available in FreeBSD?
Kenneth B. Harwood
devices.
while exploring the sound API i came across a project known (in the linux
world)
as the "preemptive kernel". basically it allows kernel processes to be
interrupted
to better support real-time applications (like audio). i realize that unix
and real-time
is sort of like oil & vinegar, but O/S's are a lot like salad dressings
anyhow.
this option is available on (redhat?) linux 2.4.6. please see:
http://kpreempt.sourceforge.net/
http://sourceforge.net/projects/kpreempt/
i'm handy with unix as a developer and end user but i've never poked around
the kernel.
1) is this option available on FreeBSD?
2) is there a simple kernel configuration file option for it?
3) can this option be turned-on via "module load" (.ko)?
4) does the linux/freebsd binary compatibility feature take care of these
issues?
thanks.
- ken harwood
Kris Kennaway
> 1) is this option available on FreeBSD?
FreeBSD 5.x is a preemptive kernel.
Kris
Kenneth B. Harwood
referring to kernel hacks which enable better performance in real-time
application conditions.
according to various unix books, the unix kernel runs at "highest" (most
~negative~?) priority. as i understand the preemptive kernel project, the
conventional approach can be detrimental when real-time application
performance is desired. for example, your application generates
time-critical voice samples needing encode/xmit. in the conventional
approach, while your data sits around waiting for CPU time, the CPU might be
busy working on some "higher" priority housekeeping task. while that task is
locked-in (scheduled to run to completion?), your important data is becoming
hopelessly latent.
the preemptive kernel is designed to overcome those sorts of bottlenecks by
somehow (it's kind of technical) allowing the application process to take-on
priorities even higher than the kernels. so iow, the kernel will wait while
the real-time data encode/xmit runs to completion.
but perhaps you're correct. when i read that "FreeBSD shares the same
binary/kernel/whatever with linux" that must mean that the development trees
share (at least at ~some~ point) a common ancestor. if FreeBSD's kernel has
been derived from linux's (which one?!) 2.5.4 kernel, then yes. FreeBSD
would already have this feature built-in.
sorry. i really should go back and read about the FreeBSD kernel and/or look
at the development branches. the ignorance is showing on my red face.
- ken h.
"Kris Kennaway" <kk...@rot13.obsecurity.org> wrote in message
news:e0vcgb.tm61.ln@xor...
Donn Miller
Maybe he means the low-latency option? Anyways, I tried the Linux
2.6.0-test2 kernel. It's got all these fancy new options, like an
anticipatory scheduler, interactivity estimator, and some sort of
scheduling elevator. Even without all those hacks, er, features,
FreeBSD 5.1 still seems to be smoother and more responsive.
-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 80,000 Newsgroups - 16 Different Servers! =-----
Patrick TJ McPhee
Kenneth B. Harwood <run10...@comcast.net> wrote:
% but perhaps you're correct. when i read that "FreeBSD shares the same
% binary/kernel/whatever with linux" that must mean that the development trees
% share (at least at ~some~ point) a common ancestor. if FreeBSD's kernel has
% been derived from linux's (which one?!) 2.5.4 kernel, then yes. FreeBSD
% would already have this feature built-in.
I don't where you read this, but it's absolutely not the case. There is
no common ancestry between the two kernels. In any case, the BSD kernel
predates the Linux one by a long time.
--
Patrick TJ McPhee
East York Canada
pt...@interlog.com
Bill Vermillion
>but perhaps you're correct. when i read that "FreeBSD shares
>the same binary/kernel/whatever with linux" that must mean that
>the development trees share (at least at ~some~ point) a common
>ancestor. if FreeBSD's kernel has been derived from linux's
>(which one?!) 2.5.4 kernel, then yes. FreeBSD would already have
>this feature built-in.
The BSD ancestry goes back to the version 7s of AT&T. Not related
to Linux at all.
One of the Unix-like RTOS that I recall is QNX. Been around for a
long time. www.qnx.com for more info.
The first question asked about audio and one of QNXs things [ or
related to them ] was the Q-sound of a few years back with
seemingly quad-synthesis in a two-channel environment. Pretty
impressive for it's work then - but there was just one sweet spot.
--
Bill Vermillion - bv @ wjv . com
Michael Sierchio
> The BSD ancestry goes back to the version 7s of AT&T. Not related
> to Linux at all.
Bill, you know that is a lie. System III version 7 and Linux
are both the work of Richard Stallman. 'fess up!
ken harwood
> Kris Kennaway wrote:
> > In article <LZCdnQQwXeK...@comcast.com>, Kenneth B. Harwood wrote:
> >
> >
> >>1) is this option available on FreeBSD?
> >
> >
> > FreeBSD 5.x is a preemptive kernel.
>
> Maybe he means the low-latency option? Anyway
s, I tried the Linux
> 2.6.0-test2 kernel. It's got all these fancy new options, like an
> anticipatory scheduler, interactivity estimator, and some sort of
> scheduling elevator. Even without all those hacks, er, features,
> FreeBSD 5.1 still seems to be smoother and more responsive.
>
the L-LO. that got me thinking ... ouch. that hurts.
according to the "Roadmap for 5-Stable":
http://www.freebsd.org/doc/en_US.ISO8859-1/articles/5-roadmap/index.html
--> The foundations for fine-grained locking and preemption in the
kernel exist, but much more work is left to be done. Work on Kernel
Schedulable Entities (KSE), similar to Scheduler Activations, has been
ongoing but needs a push to realize its benefit. Performance compared
to FreeBSD 4.X has declined and must be restored and surpassed. <--
er, that's what i'm getting at. from what i've read in the past 12
hours, different O/S's - and indeed even different unices - have
different locking characteristics. apparently some variant of fine
grain locking is a crucial to support SMP. now as i understand it,
kernel preemption requires implementation of some pretty fine-grain
locking as well.
so the question becomes: how far along are those "foundations for fine
grain locking and preemption in the kernel"?
- ken h.
Richard Tobin
Michael Sierchio <ku...@tenebras.com> wrote:
>System III version 7
No such thing ever existed.
-- Richard
--
Spam filter: to mail me from a .com/.net site, put my surname in the headers.
FreeBSD rules!
Eric Masson
Richard> No such thing ever existed.
I think that Michael was somewhat ironic ;)
Eric MAsson
--
J'aurai aimé savoir si en Norvège il y avait effectivement des panneaux
de signalisation sur les routes indiquant la présence éventuelle de
fantômes? Merci.
-+- DM in :GNU- Il y a quelque chose de pouri au royaume du neuneu -+-
Michael Sierchio
> In article <euqdnZ4lAM8...@speakeasy.net>,
> Michael Sierchio <ku...@tenebras.com> wrote:
>
>
>>System III version 7
>
First of all, you missed my humor.
And, are you the sort who, upon seeing a giraffe for the first time,
declares that there is no such animal? Are you perhaps a literal-minded
prat? Or am I missing your subtle irony?
I left a comma out, I see, see below.
I used "UNIX System III, Version 7" (aka Version 7) in its incarnation
as Microsoft XENIX in 1981. While XENIX was based on 7th Edition
UNIX, which appeared in 1979, it also incorporated lotsa changes that
went into AT&T's commercial release of System III (ca. 1982).
You'd have to ask Ballmer, Gates and Simonyi, who were there at the
time, but I think I have the timeline right. All of Microsoft was
in a couple of floors of a bank building in Bellevue. The Apple ][
was the number 1 CP/M machine in the world, thanks to the Z80
card by Microsoft.
@my_last_name.org.uk Richard Caley
ms> Bill, you know that is a lie. System III version 7 and Linux
ms> are both the work of Richard Stallman. 'fess up!
But Stallman is a robot built by Alan Kay.
--
Mail me as MYFIR...@MYLASTNAME.org.uk _O_
|<
Michael Sierchio
> In article <euqdnZ4lAM8...@speakeasy.net>, Michael Sierchio (ms) writes:
>
> ms> Bill, you know that is a lie. System III version 7 and Linux
> ms> are both the work of Richard Stallman. 'fess up!
>
> But Stallman is a robot built by Alan Kay.
You are too fucking brilliant for words, I am in awe,
practically in tears. Honor is due.
"I invented Alan Kay. I invented computing. Why do you think
they're called algorithms?"
- Al Gore
Kris Kennaway
> --> The foundations for fine-grained locking and preemption in the
> kernel exist, but much more work is left to be done. Work on Kernel
> Schedulable Entities (KSE), similar to Scheduler Activations, has been
> ongoing but needs a push to realize its benefit. Performance compared
> to FreeBSD 4.X has declined and must be restored and surpassed. <--
>
> er, that's what i'm getting at. from what i've read in the past 12
> hours, different O/S's - and indeed even different unices - have
> different locking characteristics. apparently some variant of fine
> grain locking is a crucial to support SMP. now as i understand it,
> kernel preemption requires implementation of some pretty fine-grain
> locking as well.
Pre-emption doesn't, but avoiding large interrupt latencies does.
> so the question becomes: how far along are those "foundations for fine
> grain locking and preemption in the kernel"?
Work is ongoing, and it's improving steadily.
Kris
Bill Vermillion
No. Stallman stole the code from Dan Quales mother, who invented
Perl.
Steve O'Hara-Smith
b...@wjv.comREMOVE (Bill Vermillion) wrote:
BV> No. Stallman stole the code from Dan Quales mother, who invented
BV> Perl.
One nit, one perl ?
--
C:>WIN | Directable Mirrors
The computer obeys and wins. |A Better Way To Focus The Sun
You lose and Bill collects. | licenses available - see:
| http://www.sohara.org/
Bill Vermillion
Steve O'Hara-Smith <ste...@eircom.net> wrote:
>On Sat, 02 Aug 2003 01:54:32 GMT
>b...@wjv.comREMOVE (Bill Vermillion) wrote:
>
>BV> No. Stallman stole the code from Dan Quales mother, who invented
>BV> Perl.
> One nit, one perl ?
Thanks for the great laugh. I hereby grant you permission to
use this smiley.
))))
))))
:::: ))))
:::: ))))
---- ))))
---- ))))
:::: ))))
:::: ))))
))))
))))
Bill