On Fri, 20 Jul 2018 at 17:51, Radim Krcmar <rkrcmar@xxxxxxxxxx> wrote: > > 2018-07-20 11:33+0800, Wanpeng Li: > > On Fri, 20 Jul 2018 at 00:28, Radim Krčmář <rkrcmar@xxxxxxxxxx> wrote: > > > 2018-07-03 14:21+0800, Wanpeng Li: > > > But because it is very similar to x2apic, I'd really need some real > > > performance data to see if this benefits a real workload. > > > > Thanks for your review, Radim! :) I will find another real benchmark > > instead of the micro one to evaluate the performance. > > Analyzing the cpu bitmap for every IPI request on a non-small guest (at > least 32 VCPUs, ideally >256) during various workloads could also > provide some insight regardless of workload/benchmark result -- we want > to know how many VM exits we would save. I will try ebizzy benchmark, just complete the patchset w/ __uint128 which Paolo just suggested. In addition, I remember aliyun posted the performance number for their online real workload "message oriented middleware" from 800 k/s vmexits w/ PV IPIs to 150 k/s vmexits w/ PV IPIs. Regards, Wanpeng Li
