| |||||||||||||
The realtime preemption mini-summitThe realtime preemption mini-summitPosted Oct 4, 2009 13:56 UTC (Sun) by dvhart (guest, #19636)In reply to: The realtime preemption mini-summit by simlo Parent article: The realtime preemption mini-summit
WRT rwlocks. We actually cap reader count to 1 in PREEMPT_RT for that very reason. This is unfortunate, and one of the causes for performance degradation on -rt for certain workloads. There was some discussion during the rt-summit in Dresden about making kernel rwlocks non-pi-aware for this reason. Some more investigation is needed before we make a decision there.
(Log in to post comments)
The realtime preemption mini-summit Posted Oct 6, 2009 15:37 UTC (Tue) by simlo (subscriber, #10866) [Link]
As I said: You could leave half-PI-aware : Let readers boost the writer, but not the other way around. This will most likely work in many cases. It means a RT task can't write-lock a rwlock must defer the operation to another task. Config options are needed....
|
|||||||||||||