|| ||Gregory Haskins <email@example.com>|
|| ||[PATCH 0/9] RT: balance rt tasks v6|
|| ||Wed, 24 Oct 2007 09:52:31 -0400|
Gregory Haskins <firstname.lastname@example.org>,
Steven Rostedt <email@example.com>,
Dmitry Adamushko <firstname.lastname@example.org>,
Peter Zijlstra <email@example.com>,
Ingo Molnar <firstname.lastname@example.org>, Darren Hart <email@example.com>|
Applies to 23-rt1
Changes since v5:
*) Folded some of the smaller patches together to address feedback
*) Fixed several minor bugs related to PI re-factoring and wakeup paths
I still have yet to make Ingo Oeser's suggestion to the last patch.
At this point, things are looking really really good. So this is a request to
provide a formal review (if one has not yet been done) and to consider for
inclusion in -rt.
In my testing, cyclictest look great (ran overnight on a 4-way with the max
outlier at 32us), the preempt-test passes with flying colors, and we see a
300% improvement in latency spikes on large SMP systems (16-way is the largest
we have in our small lab).
I will rebase to -rt2 when its available.
You can find the preempt-test here: