>I have a query about scheduling in multi core systems. The user level
>threads are mapped to LWP in Solaris and those LWPs are scheduled by the
>OS.
>
>Is there any flexibility for the user to write some module and plug into
>the kernel so that he can schedule all the LWPs (to which his threads
>are attached) as per his requirement?
>
> 
>
>The requirement we have is to improve the performance of our multi
>threaded software by making/forcing the threads run across multiple
>cores.


You can write your own scheduling class but the first question I have is:
why does the user expect it to perform better than the standard scheduling
classes?  (Currently all the user can do is pick a different standard 
scheduling class or bind threads to specific CPUs (i.e., processor thread
of execution)

Casper

_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to