Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 11/10/2011 10:05 PM, Avi Kivity wrote: On 11/10/2011 03:28 PM, Xiao Guangrong wrote: I have tested RHEL.6.1 setup/boot/reboot/shutdown and the complete output of scan_results.py is attached. The result shows the performance is improved: before:After: 570529 555538 552531 546528 553559 553527 550523 553533 547538 550526 How do you think about it? :) Well, either I was sloppy in my measurements, or maybe RHEL 6 is very different from F9 (unlikely). I'll measure it again and see. Thanks for your time. :) btw, this is with ept=0, yes? Yeah. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 11/10/2011 03:28 PM, Xiao Guangrong wrote: > > I have tested RHEL.6.1 setup/boot/reboot/shutdown and the complete > output of scan_results.py is attached. > > The result shows the performance is improved: > before:After: > 570529 > 555538 > 552531 > 546528 > 553559 > 553527 > 550523 > 553533 > 547538 > 550526 > > How do you think about it? :) Well, either I was sloppy in my measurements, or maybe RHEL 6 is very different from F9 (unlikely). I'll measure it again and see. btw, this is with ept=0, yes? -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 11/06/2011 11:35 PM, Avi Kivity wrote: On 11/04/2011 11:16 AM, Xiao Guangrong wrote: I have done kernbench tests several times on my desktop, but it shows very well: before patchset: real 212.27 real 213.47 real 204.99 real 200.58 real 199.99 real 199.94 real 201.51 real 199.83 real 198.19 real 205.13 after patchset: real 199.90 real 201.89 real 194.54 real 188.71 real 185.75 real 187.70 real 188.99 real 188.53 real 186.29 real 188.25 I will test it on our server using kvm-autotest, could you share me your config file please? # Copy this file to tests.cfg and edit it. # # This file contains the test set definitions. Define your test sets here. Thanks Avi! I have tested RHEL.6.1 setup/boot/reboot/shutdown and the complete output of scan_results.py is attached. The result shows the performance is improved: before: After: 570 529 555 538 552 531 546 528 553 559 553 527 550 523 553 533 547 538 550 526 How do you think about it? :) result.tar.bz2 Description: application/bzip
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 11/04/2011 11:16 AM, Xiao Guangrong wrote: > > I have done kernbench tests several times on my desktop, but it shows > very well: > > before patchset: > real 212.27 > real 213.47 > real 204.99 > real 200.58 > real 199.99 > real 199.94 > real 201.51 > real 199.83 > real 198.19 > real 205.13 > > after patchset: > real 199.90 > real 201.89 > real 194.54 > real 188.71 > real 185.75 > real 187.70 > real 188.99 > real 188.53 > real 186.29 > real 188.25 > > I will test it on our server using kvm-autotest, could you share me > your config file please? # Copy this file to tests.cfg and edit it. # # This file contains the test set definitions. Define your test sets here. include base.cfg include subtests.cfg include guest-os.cfg include cdkeys.cfg extra_params += ' -enable-kvm' image_name(_.*)? ?<= /images/autotest/ #cdrom(_.*)? ?<= isos/ # Modify/comment the following lines if you wish to modify # the paths of the image files, ISO files, step files or qemu binaries. # # As for the defaults: # * qemu and qemu-img are expected to be found under /usr/bin/qemu-kvm and # /usr/bin/qemu-img respectively. # * All image files are expected under /tmp/kvm_autotest_root/images/ # * All iso files are expected under /tmp/kvm_autotest_root/isos/ # * All step files are expected under /tmp/kvm_autotest_root/steps/ #image_name.* ?<= images/ #cdrom.* ?<= isos/ drive_cache = writeback # -no-kvm-irqchip #timeout_multiplier = 8 #iterations = 30 vga = cirrus # Here are the test sets variants. The variant 'qemu_kvm_windows_quick' is # fully commented, the following ones have comments only on noteworthy points variants: - @avi: only no_pci_assignable only qcow2 only ide #only default only smp2 #only up only Fedora.9.32 Fedora.9.64 WinVista.64sp1 WinXP only install setup boot reboot migrate shutdown only rtl8139 only smallpages #only default_host no migrate.exec # Uncomment the following lines to enable abort-on-error mode: # abort_on_error = yes kill_vm.* ?= no kill_unresponsive_vms.* ?= no WinXP.64: no shutdown no reboot Win2003.64: no shutdown no reboot # Choose your test list from the testsets defined only avi pci_assignable = no serial_console = no -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 10/11/2011 04:36 PM, Xiao Guangrong wrote: On 10/09/2011 09:37 PM, Avi Kivity wrote: On 10/09/2011 02:24 PM, Avi Kivity wrote: On 10/08/2011 06:06 AM, Xiao Guangrong wrote: On 10/07/2011 01:53 AM, Marcelo Tosatti wrote: On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: This patchset is against https://github.com/avikivity/kvm.git next branch. In this version, some changes come from Avi's comments: - fix instruction retried for nested guest - skip write-flooding for the sp whose level is 1 - rename some functions Please rebase. Marcelo, These patches can be applied without any conflict and it also works well, the current code was pulled from https://github.com/avikivity/kvm.git next branch. What problem did you meet when you applied these? :( I guess it was a user error - it applies cleanly here too (and pushed to next, thanks). However, it seems to reduce performance. Ouch, will look into it soon. Hi Avi, I have done kernbench tests several times on my desktop, but it shows very well: before patchset: real 212.27 real 213.47 real 204.99 real 200.58 real 199.99 real 199.94 real 201.51 real 199.83 real 198.19 real 205.13 after patchset: real 199.90 real 201.89 real 194.54 real 188.71 real 185.75 real 187.70 real 188.99 real 188.53 real 186.29 real 188.25 I will test it on our server using kvm-autotest, could you share me your config file please? -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 10/09/2011 09:37 PM, Avi Kivity wrote: > On 10/09/2011 02:24 PM, Avi Kivity wrote: >> On 10/08/2011 06:06 AM, Xiao Guangrong wrote: >>> On 10/07/2011 01:53 AM, Marcelo Tosatti wrote: >>> > On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: >>> >> This patchset is against https://github.com/avikivity/kvm.git next >>> >> branch. >>> >> >>> >> In this version, some changes come from Avi's comments: >>> >> - fix instruction retried for nested guest >>> >> - skip write-flooding for the sp whose level is 1 >>> >> - rename some functions >>> > >>> > Please rebase. >>> > >>> > >>> >>> Marcelo, >>> >>> These patches can be applied without any conflict and it also works well, >>> the current code was pulled from https://github.com/avikivity/kvm.git next >>> branch. >>> >>> What problem did you meet when you applied these? :( >> >> I guess it was a user error - it applies cleanly here too (and pushed to >> next, thanks). >> > > However, it seems to reduce performance. > Ouch, will look into it soon. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 10/09/2011 02:24 PM, Avi Kivity wrote: On 10/08/2011 06:06 AM, Xiao Guangrong wrote: On 10/07/2011 01:53 AM, Marcelo Tosatti wrote: > On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: >> This patchset is against https://github.com/avikivity/kvm.git next branch. >> >> In this version, some changes come from Avi's comments: >> - fix instruction retried for nested guest >> - skip write-flooding for the sp whose level is 1 >> - rename some functions > > Please rebase. > > Marcelo, These patches can be applied without any conflict and it also works well, the current code was pulled from https://github.com/avikivity/kvm.git next branch. What problem did you meet when you applied these? :( I guess it was a user error - it applies cleanly here too (and pushed to next, thanks). However, it seems to reduce performance. Autotest results before: TestStatus Seconds Info -- --- (Result file: ../../results/default/status) unittestGOOD 147 completed successfully Fedora.9.32.install.smp2GOOD 865 completed successfully Fedora.9.32.boot.smp2 GOOD 46 completed successfully Fedora.9.32.reboot.smp2 GOOD 49 completed successfully Fedora.9.32.shutdown.smp2 GOOD 15 completed successfully Fedora.9.64.install.smp2GOOD 943 completed successfully Fedora.9.64.boot.smp2 GOOD 47 completed successfully Fedora.9.64.reboot.smp2 GOOD 48 completed successfully Fedora.9.64.shutdown.smp2 GOOD 14 completed successfully WinXP.32.install.smp2 GOOD 772 completed successfully WinXP.32.setup.smp2 GOOD 53 completed successfully WinXP.32.boot.smp2 GOOD 57 completed successfully WinXP.32.reboot.smp2GOOD 34 completed successfully WinXP.32.shutdown.smp2 GOOD 5 completed successfully WinXP.64.install.smp2 GOOD 636 completed successfully After: unittestGOOD 150 completed successfully Fedora.9.32.install.smp2GOOD 879 completed successfully Fedora.9.32.boot.smp2 GOOD 50 completed successfully Fedora.9.32.reboot.smp2 GOOD 48 completed successfully Fedora.9.32.shutdown.smp2 GOOD 15 completed successfully Fedora.9.64.install.smp2GOOD 997 completed successfully Fedora.9.64.boot.smp2 GOOD 47 completed successfully Fedora.9.64.reboot.smp2 GOOD 48 completed successfully Fedora.9.64.shutdown.smp2 GOOD 14 completed successfully WinXP.32.install.smp2 GOOD 764 completed successfully WinXP.32.setup.smp2 GOOD 51 completed successfully WinXP.32.boot.smp2 GOOD 40 completed successfully WinXP.32.reboot.smp2GOOD 34 completed successfully WinXP.32.shutdown.smp2 GOOD 5 completed successfully WinXP.64.install.smp2 GOOD 666 completed successfully -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 10/08/2011 06:06 AM, Xiao Guangrong wrote: On 10/07/2011 01:53 AM, Marcelo Tosatti wrote: > On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: >> This patchset is against https://github.com/avikivity/kvm.git next branch. >> >> In this version, some changes come from Avi's comments: >> - fix instruction retried for nested guest >> - skip write-flooding for the sp whose level is 1 >> - rename some functions > > Please rebase. > > Marcelo, These patches can be applied without any conflict and it also works well, the current code was pulled from https://github.com/avikivity/kvm.git next branch. What problem did you meet when you applied these? :( I guess it was a user error - it applies cleanly here too (and pushed to next, thanks). -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 10/07/2011 01:53 AM, Marcelo Tosatti wrote: > On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: >> This patchset is against https://github.com/avikivity/kvm.git next branch. >> >> In this version, some changes come from Avi's comments: >> - fix instruction retried for nested guest >> - skip write-flooding for the sp whose level is 1 >> - rename some functions > > Please rebase. > > Marcelo, These patches can be applied without any conflict and it also works well, the current code was pulled from https://github.com/avikivity/kvm.git next branch. What problem did you meet when you applied these? :( -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: > This patchset is against https://github.com/avikivity/kvm.git next branch. > > In this version, some changes come from Avi's comments: > - fix instruction retried for nested guest > - skip write-flooding for the sp whose level is 1 > - rename some functions Please rebase. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: > This patchset is against https://github.com/avikivity/kvm.git next branch. > > In this version, some changes come from Avi's comments: > - fix instruction retried for nested guest > - skip write-flooding for the sp whose level is 1 > - rename some functions Please rebase. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 09/23/2011 02:51 PM, Marcelo Tosatti wrote: On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: > This patchset is against https://github.com/avikivity/kvm.git next branch. > > In this version, some changes come from Avi's comments: > - fix instruction retried for nested guest > - skip write-flooding for the sp whose level is 1 > - rename some functions Looks good to me. To me as well. -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On 09/23/2011 07:51 PM, Marcelo Tosatti wrote: > On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: >> This patchset is against https://github.com/avikivity/kvm.git next branch. >> >> In this version, some changes come from Avi's comments: >> - fix instruction retried for nested guest >> - skip write-flooding for the sp whose level is 1 >> - rename some functions > > Looks good to me. > Ping.. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v4 00/11] KVM: x86: optimize for writing guest page
On Thu, Sep 22, 2011 at 04:52:40PM +0800, Xiao Guangrong wrote: > This patchset is against https://github.com/avikivity/kvm.git next branch. > > In this version, some changes come from Avi's comments: > - fix instruction retried for nested guest > - skip write-flooding for the sp whose level is 1 > - rename some functions Looks good to me. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH v4 00/11] KVM: x86: optimize for writing guest page
This patchset is against https://github.com/avikivity/kvm.git next branch. In this version, some changes come from Avi's comments: - fix instruction retried for nested guest - skip write-flooding for the sp whose level is 1 - rename some functions -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html