Skip to contentSkip to conversation

Where to find good doc ?

DISCUSSION Philippe Martinole 2 participants 2 messages Dec 15, 2004 – Dec 16, 2004

lore
Share & tools
lore
#0 lore raw

Hello everyone,

I've been looking for a doc on SMP since a while but I can't find what I want. Do you know some good doc to read ?
To be more precese, I want to put all processes on only one cpu except nly one real time process on the second cpu.

Thanks,

Philippe

#1 lore raw
Quoted 6 lines · Philippe Martinole

Hello everyone,

I've been looking for a doc on SMP since a while but I can't find what I want. Do you know some good doc to read ?
To be more precese, I want to put all processes on only one cpu except nly one real time process on the second cpu.

What you want to do is Asymetric multiprocessing using SMP (Symmetric Multi-Processor) so SMP doc won't help you so much :))

You may have a look to this rthowto:
http://people.mech.kuleuven.ac.be/~bruyninc/rthowto/
which contains valuable information RT and linux including SMP aspects.

Regarding linux SMP handling and [soft] realtime you must know that each process may have a processor affinity which may be set with the sched_setaffinity syscall. (beware of glibc moving interface regarding this syscal)

Then each interrupt has its smp_affinity which may be set using /proc/irq/<num>/smp_affinity

As you can see there is NO standard SMP linux feature to isolate/reserve one (or several) CPU from being used by processes.

This kind of behavior may be obtained using either sub-kernel approach (RTLinux or RTAI
http://www.aero.polimi.it/~rtai/) or using proprietary extension such as CPU shielding
http://www.linuxdevices.com/articles/AT8610061752.html
http://linuxdevices.com/files/article030/shielded-cpu.pdf

Hope this helps.

Quoted 9 lines

Thanks,

Philippe

-
To unsubscribe from this list: send the line "unsubscribe linux-smp" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 2 lines
Eric NOULARD
E-mail: [email protected]