threadlock patch
Julian Elischer
julian at elischer.org
Thu May 10 21:29:50 UTC 2007
Marko,
We are just reorganising our effort on this stuff
to put it in a bit more of an organised manner.
for a start.. there is a perforce branch that covers this work..
http://perforce.freebsd.org/depotTreeBrowser.cgi?FSPC=/depot/user/attilio/attilio%5fschedlock&HIDEDEL=NO
(hopefully that didn't wrap)
also we are using the mailing list smp at freebsd.org
to discuss this. I suggest that you and any of the guys there who are intersted in
pushing this sign up to catch the fun. I've CC'd this to the list..
I suggest we use that list to communicate so that there is a record.
also makes is easier for others to follow and join when needed.
Julian
Jeff Roberson wrote:
> I just returned home from a long trip so I don't have time to review
> your turnstile and rwlock changes in detail. I do have a few comments
> about the patch that may be helpful as well as some questions.
>
> The threadlock patch is really just a first starting point to get us to
> per-cpu run queue locks. By itself it will only minorly reduce
> contention for sched lock since in most cases the per-cpu thread lock
> still points to this single scheduler lock. There is another bit of
> work that must be done in ULE to permit the use of per-cpu locks. I do
> not believe it is practical to do this on 4BSD. Once this is done we
> should see some impressive performance improvements on larger machines,
> especially with context-switch heavy workloads.
>
> I would like to ask what your environment is. How many cpus do you have
> per machine and what kind of workload are they running? What problems
> are you running into that lead you to this patch? I do not know in what
> capacity nokia uses FreeBSD. I'm very interested in finishing up this
> work and having a good test case and industry involvement would help
> that along.
>
> Thanks,
> Jeff
More information about the freebsd-smp
mailing list