Re: [PATCH v5] KVM: halt-polling: poll for the upcoming fire timers
From: yunhong jiang
Date: Fri May 27 2016 - 20:23:36 EST
On Fri, 27 May 2016 11:49:26 +0200
Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
> On 26/05/2016 22:33, yunhong jiang wrote:
> > On Thu, 26 May 2016 12:26:27 +0200
> > Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
> >> On 25/05/2016 04:47, Wanpeng Li wrote:
> >>> From: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
> >>> If an emulated lapic timer will fire soon(in the scope of 10us the
> >>> base of dynamic halt-polling, lower-end of message passing
> >>> workload latency TCP_RR's poll time < 10us) we can treat it as a
> >>> short halt, and poll to wait it fire, the fire callback
> >>> apic_timer_fn() will set KVM_REQ_PENDING_TIMER, and this flag
> >>> will be check during busy poll. This can avoid context switch
> >>> overhead and the latency which we wake up vCPU.
> >> As discussed on IRC, I would like to understand why the adaptive
> > Glad to know the IRC channel. Is #kvm channel on freenode the IRC
> > you are talking about?
> No, it's #qemu on irc.oftc.net
Thanks for the information.