On Mon, Feb 19, 2018 at 02:20:26PM -0500, nitesh narayan lal wrote: > Hi Michael, > > > On 25 January 2018 at 12:28, Michael S. Tsirkin <mst@xxxxxxxxxx> wrote: > > On Fri, Dec 01, 2017 at 12:31:29PM -0500, nilal@xxxxxxxxxx wrote: > > Changelog in v6: > > -Moved CONFIG_KVM_FREE_PAGE_HINTING from virt/kvm/Kconfig to arch/ > x86/kvm/Kconfig in order to resolve conflicts with other architectures: > > *s390 has its own page hinting functionality > > *powerpc has its own arch_free_page implementation > > -Fixed the typo in virtio_balloon structure which caused > compilation issues when guest page hinting was disabled > > -Separated init_vqs implementation with and without hinting_vq > enablement based on the CONFIG_KVM_FREE_PAGE_HINTING ifdef > > > > Virtio interface changes are picked up from Wei's patch-set for > Virtio-balloon enhancement[3]. "Wei, How would you like me to credit you in > the final patch?") > > So the main point missing in all this is any kind of performance > numbers: what kind of gain do you see with these patches and for which > workloads? What kind of overhead and for which workloads? > > -- > MST > > > > Sorry for the delay in my response. > I have shared the performance numbers earlier[1] but the load was mostly in the > userspace. > I did run hackbench and netperf for the same but unfortunately I don't have the > numbers with me. > I am planning to re-run the test and share the results once I return in May. > > [1] https://www.spinics.net/lists/kvm/msg158054.html > > Regards > Nitesh Sorry I wasn't clear. The numbers you posted seem to show no slowdown, which is good. However, there's little point in merging the patches unless we also see some kind of gain, isn't there?
