Messages in this thread | | | Date | Tue, 01 May 2012 18:33:08 +0300 | From | Avi Kivity <> | Subject | Re: [RFC PATCH v1 3/5] KVM: Add paravirt kvm_flush_tlb_others |
| |
On 05/01/2012 06:11 PM, Peter Zijlstra wrote: > On Tue, 2012-05-01 at 15:12 +0300, Avi Kivity wrote: > > We're now moving the freeing of kvm shadow page tables from using rcu to > > using an irq-protected scheme like gup_fast(), because of the > > performance differences. We didn't track it down but I expect the cause > > is less reuse of cache-hot pages. > > It would be good to understand these things.. just doing things because > tends to come back and bite you :-)
Agree. Anyway that was only part of the motivation, the other part was establishing a bound on the number of pages under deferred freeing.
> If its cache behaviour you should be able to clearly see that using perf > stat.
It's hard to see it clearly because the changes are in the order of 1-2% in a very noisy benchmark.
-- error compiling committee.c: too many arguments to function
| |