| | Date | Sat, 23 Feb 2008 00:03:45 -0800 | | From | Andrew Morton <> | | Subject | Re: [PATCH [RT] 00/14] RFC - adaptive real-time locks |
On Thu, 21 Feb 2008 22:24:20 +0100 Ingo Molnar <mingo@elte.hu> wrote:
> regarding the concept: adaptive mutexes have been talked about in the
> past, but their advantage is not at all clear, that's why we havent done
> them. It's definitely not an unambigiously win-win concept.
When ext3 was converted from sleeping locks to spinlocks, dbench-on-numaq
throughput went up by a factor of ten. I'd expect that what RT has done
was a truly awful change for lots of workloads on lots of machines.
Yeah, there's the dont-enable-it-if-you're-doing-that option, but that adds
the we-just-doubled-the-number-of-kernels-distros-need-to-ship problem.
Does -rt also make bit_spin_lock preemptible? If not, I'd have thought the
change was of little benefit to ext3/jbd and it might as well go back to
spinning locks.
|