asynchronus multiprocessing

asynchronus multiprocessing

[Martin A. Brooks]

[-- Attachment #1: Type: text/plain, Size: 231 bytes --]

Has there been any talk of (or work on) AMP support in the kernel?

-- 
Martin A. Brooks   Systems Administrator
Jtrix Ltd          t: +44 7395 4990
57-59 Neal Street  f: +44 7395 4991
London, WC2H 9PP   e: martin@jtrix.com

[-- Attachment #2: Type: application/pgp-signature, Size: 240 bytes --]

Re: asynchronus multiprocessing

[Martin A. Brooks]

[-- Attachment #1: Type: text/plain, Size: 322 bytes --]

On Wed, 2001-12-19 at 15:40, Martin A. Brooks wrote:
> Has there been any talk of (or work on) AMP support in the kernel?

I meant /asymmetric/ MP. Sorry.

-- 
Martin A. Brooks   Systems Administrator
Jtrix Ltd          t: +44 7395 4990
57-59 Neal Street  f: +44 7395 4991
London, WC2H 9PP   e: martin@jtrix.com

[-- Attachment #2: Type: application/pgp-signature, Size: 240 bytes --]

Re: asynchronus multiprocessing

[John Clemens]

I looked into this a little while ago... never got a patch working
100% correctly...there was some discussions about it on the SGI Linux
Scalability list (as far as i can tell, it was the -only- discussion on
the SGI list, as the IBM list started about the same time.. ahhh, orphaned
mailing lists..).. It was a quick-n-dirty hack, and only the beginnings of
one at that...

i put a quick page up about it at
http://www.deater.net/john/processorgroups.html

Conclusion: Could be done, not worth it for miniscule speed gains, could
be much more of a benefit on NUMA machines, but the linux scalable
scheduler's out there are probably a much better approach to doing the
same thing...

john.c

On 19 Dec 2001, Martin A. Brooks wrote:

> On Wed, 2001-12-19 at 15:40, Martin A. Brooks wrote:
> > Has there been any talk of (or work on) AMP support in the kernel?
>
> I meant /asymmetric/ MP. Sorry.
>
>

-- 
John Clemens          http://www.deater.net/john
john@deater.net     ICQ: 7175925, IM: PianoManO8
      "I Hate Quotes" -- Samuel L. Clemens

Re: asynchronus multiprocessing

[george anzinger]

"Martin A. Brooks" wrote:
> 
> On Wed, 2001-12-19 at 15:40, Martin A. Brooks wrote:
> > Has there been any talk of (or work on) AMP support in the kernel?
> 
> I meant /asymmetric/ MP. Sorry.
> 
I have heard of some work.  As I understand it they are making an API
for cpu affinity.  It is real time, so they are also interested in the
schedule routines around cpu affinity as well.

What did you have in mind? 
-- 
George           george@mvista.com
High-res-timers: http://sourceforge.net/projects/high-res-timers/
Real time sched: http://sourceforge.net/projects/rtsched/

Re: asynchronus multiprocessing

[Martin A. Brooks]

[-- Attachment #1: Type: text/plain, Size: 655 bytes --]

On Wed, 2001-12-19 at 18:18, george anzinger wrote:

> I have heard of some work.  As I understand it they are making an API
> for cpu affinity.  It is real time, so they are also interested in the
> schedule routines around cpu affinity as well.
> 
> What did you have in mind? 

Nothing too major, I just intend to buy myself a dual Athlon. I would
rather recycle my perfectly good 1.3ghz chip and use it in conjuction
with a 1.5ghz athlon rather than buying two new processors.

-- 
Martin A. Brooks   Systems Administrator
Jtrix Ltd          t: +44 7395 4990
57-59 Neal Street  f: +44 7395 4991
London, WC2H 9PP   e: martin@jtrix.com

[-- Attachment #2: Type: application/pgp-signature, Size: 240 bytes --]

Re: asynchronus multiprocessing

[george anzinger]

"Martin A. Brooks" wrote:
> 
> On Wed, 2001-12-19 at 18:18, george anzinger wrote:
> 
> > I have heard of some work.  As I understand it they are making an API
> > for cpu affinity.  It is real time, so they are also interested in the
> > schedule routines around cpu affinity as well.
> >
> > What did you have in mind?
> 
> Nothing too major, I just intend to buy myself a dual Athlon. I would
> rather recycle my perfectly good 1.3ghz chip and use it in conjuction
> with a 1.5ghz athlon rather than buying two new processors.
> 
Apart from the clock differences (which if actually presented to the two
cpus would cause the TSC to drift apart cause who knows what mayhem)
what is the real difference?  I think you would have to make sure you
configure for the subset of chip capabilities so you don't find the
system trying to use some capability that is only present on one cpu and
then switching the ap to the other.  As I recall, someone has already
explored this turf to some extent.  Check the archives.

But really, I was wondering what you thought needed to be done in
software?
-- 
George           george@mvista.com
High-res-timers: http://sourceforge.net/projects/high-res-timers/
Real time sched: http://sourceforge.net/projects/rtsched/