Messages in this thread | | | Date | Wed, 23 Dec 2009 10:01:26 +0100 | Subject | Re: workqueue thing | From | Stijn Devriendt <> |
| |
On Wed, Dec 23, 2009 at 9:43 AM, Peter Zijlstra <peterz@infradead.org> wrote: > On Wed, 2009-12-23 at 09:17 +0100, Stijn Devriendt wrote: >> On Tue, Dec 22, 2009 at 7:07 PM, Andi Kleen <andi@firstfloor.org> wrote: >> > One reason I liked a more dynamic frame work for this is that it >> > has the potential to be exposed to user space and allow automatic >> > work partitioning there based on available cores. User space >> > has a lot more CPU consumption than the kernel. >> > >> Basically, this is exactly what I was trying to solve with my >> sched_wait_block patch. It was broken in all ways, but the ultimate >> goal was to have concurrency managed workqueues (to nick the term) >> in userspace and have a way out when I/O hits the workqueue. > > Don't we have the problem of wakeup concurrency here? > > Forking on blocking is only half the problem (and imho the easy half). > >
The original design was to always have 1 spare thread handy that would wait until the worker-thread blocked. At that point it would wakeup and continue trying to keep the CPU busy. The current perf-event approach is to have threads poll based upon the concurrency as measured in the kernel by the perf-event. When too many threads are on the runqueue, the poll() blocks. When threads go to sleep/block another thread falls out of poll()to continue work.
Stijn -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |