users

users@lists.genode.org

1 participants
2310 discussions

Re: Re: Benchmark Armv8 VMM (Stefan Kalkowski)
by Chris Hofer 07 May '20

07 May '20

Hi Stefan, Hi Chris, On Wed, May 06, 2020 at 01:23:59PM +0200, Chris Hofer wrote: > Hi, > > to measure the performance overhead of the Armv8 Genode VMM, I ran some > benchmarks (cyclictest and lmbench). I compared it against a native Linux > installation on the Imx8-evk board (same kernel version). For some > measurements (for example fork + executing /bin/sh from lmbench), I observe > counter-intuitive results in the sense that the virtualized version runs > *faster* than the native one (Vm 1,1ms vs. native 2,5ms). > > My understanding is, that the virtual timer offset (cntvoff_el2) is not used > and thus, the guest sees wall clock time. I would expect that the > virtualization incurs an overhead and this overhead manifests itself in > longer (wall clock) time measurements. > >> Correct, but you assume that native Linux and virtualized version run >> with the same frequencies (cpu + timer) - don't you? >> >> Actually, I think the timer frequency should be constant in both >> scenarios, whereby at least the cpu will be controlled dynamically in >> the native Linux case. >> >> When comparing such complex scenarios to avoid mistakes I would >> strongly encourage you to not measure time on-target. It often leads >> to wrong assumptions. It is better to measure long-term scenarios from >> another host. If you need to measure micro-benchmarks, you might loop >> the operation quite often and measure the whole time-span outside, and >> divide the result. >> >> Alternatively, if you insist to measure on-target, you can instrument >> Linux to use a constant CPU frequency (google for "linux cpu >> governor"). Moreover, you might check whether the timer frequency is >> the same in both settings (CNTFRQ_EL0) - well it should be. Thank you for this advice; it makes perfect sense to capture the time off-target. I finally found a big mistake in my setup to measure the duration of forking a process and calling /bin/sh. The native installation uses a dynamically linked shell and the guest has a statically linked binary. After using statically linked binaries on both systems I get reasonable results. Nevertheless, I will take advice and verify the measurements off-target. Best, Chris

1 0

Benchmark Armv8 VMM
by Chris Hofer 07 May '20

07 May '20

Hi, to measure the performance overhead of the Armv8 Genode VMM, I ran some benchmarks (cyclictest and lmbench). I compared it against a native Linux installation on the Imx8-evk board (same kernel version). For some measurements (for example fork + executing /bin/sh from lmbench), I observe counter-intuitive results in the sense that the virtualized version runs *faster* than the native one (Vm 1,1ms vs. native 2,5ms). My understanding is, that the virtual timer offset (cntvoff_el2) is not used and thus, the guest sees wall clock time. I would expect that the virtualization incurs an overhead and this overhead manifests itself in longer (wall clock) time measurements. Do you have any explanation for this or give me a hint on how to confirm correctness of measurements? Best, Chris

2 1

Gui issue.
by Shlomo Pongratz 23 Apr '20

23 Apr '20

Hi, While installing vbox5-nova-scupt there is a need to configure ROM(usb_device). The option list is truncated in the middle of "dynamic_depot_rom" (in the middle of the line), and there is no way to see if there are other options below. I guess the best thing is to change the window size however changing the fb_drv hight in the config didn't do the trick. Best regard. S.P.

4 4

Running on QEMU.
by Shlomo Pongratz 22 Apr '20

22 Apr '20

Hi, I'm running the scultp-20.02 on QEMU and I'm trying tor run vbox5-nova-sculpt. And I get these errors: page fault, pd='init -> runtime -> vbox5-nova-sculpt -> vbox' thread='EMT-1' cpu=0 ip=0xaeb70 address=0x30 stack pointer=0xa09fec48 qualifiers=0x4 irUwp reason=1 no RM attachment (READ pf_addr=0x30 pf_ip=0xaeb70 from pager_object: pd='init -> runtime -> vbox5-nova-sculpt -> vbox' thread='VMPwrUp') page fault, pd='init -> runtime -> vbox5-nova-sculpt -> vbox' thread='VMPwrUp' cpu=0 ip=0xaeb70 address=0x30 stack pointer=0xa06fe848 qualifiers=0x4 irUwp reason=1 no RM attachment (READ pf_addr=0x30 pf_ip=0xaeb70 from pager_object: pd='init -> runtime -> vbox5-nova-sculpt -> vbox' thread='EMT-0') page fault, pd='init -> runtime -> vbox5-nova-sculpt -> vbox' thread='EMT-0' cpu=0 ip=0xaeb70 address=0x30 stack pointer=0xa08fead8 qualifiers=0x4 irUwp reason=1 Does this implies that this can't be done on QEMU but only on real HW. Best regards, S.P.

2 1

Warning messages
by Shlomo Pongratz 21 Apr '20

21 Apr '20

Hi, I'm running the downloaded sculpt-20.02.img according to https://genode.org/download/sculpt qemu-system-x86_64 -display sdl -cpu core2duo -machine q35 \ -m 1024 -drive format=raw,file=sculpt-20.02.img (and also a built scultp according to the document sculpt-20-02.pdf) and I always see the following messages. [init -> config_fs] tar archive 'launcher.tar' local at 0x184000, size is 12288 [init -> config_fs_rom] /fonts ROM file is missing [init -> config_fs_rom] /fonts ROM file is missing [init -> config_fs_rom] /nic_router ROM file is missing [init -> config_fs_rom] /nic_router ROM file is missing [init -> config_fs_rom] /wifi ROM file is missing [init -> config_fs_rom] /wifi ROM file is missing [init -> report_fs_rom] /runtime/launcher_query/listing ROM file is missing [init -> report_fs_rom] /runtime/launcher_query/listing ROM file is missing [init -> config_fs_rom] /managed/deploy ROM file is missing [init -> config_fs_rom] /managed/deploy ROM file is missing [init -> config_fs_rom] /installation ROM file is missing [init -> config_fs_rom] /installation ROM file is missing [init -> report_fs_rom] /runtime/editor/saved ROM file is missing [init -> report_fs_rom] /runtime/editor/saved ROM file is missing [init -> config_fs_rom] /runtime ROM file is missing [init -> config_fs_rom] /runtime ROM file is missing [init -> report_fs_rom] /runtime/depot_query/scan ROM file is missing [init -> report_fs_rom] /runtime/depot_query/scan ROM file is missing [init -> report_fs_rom] /runtime/depot_query/index ROM file is missing [init -> report_fs_rom] /runtime/depot_query/index ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing [init -> report_fs_rom] /runtime/ahci-0.part_block/partitions ROM file is missing Is this normal? If not, what should I do? Best regards, S.P.

4 3

I/O Terminal for VM on Imx8q-evk
by Chris Hofer 20 Apr '20

20 Apr '20

Hey, I'm continuously exploring for my thesis the Genode virtualization on the Imx8q-evk board. Currently the vmm_arm-runscript only allows to specify a static command, which is repeatedly sent to the VM after it booted. In order to run some benchmarks on the VM, I would like to interact with the VM by sending dynamically input. Is there a way to interactively send input to the TTY of the VM via e.g. the serial interface of the Imx8q-evk? If there is none, can you think of an easy way to build it? Best, Chris Hofer

2 1

Error while building boot image.
by Shlomo Pongratz 19 Apr '20

19 Apr '20

Hi, I'm trying to build sculpt according to document https://genode.org/documentation/sculpt-20-02.pdf. In section 5.2 Building the boot image stage 7 one finds: Create the Sculpt boot image (defined by the run script at repos/gems/run/sculpt.run) make -C build/x86_64 run/sculpt KERNEL=nova BOARD=pc The file repos/gems/run/sculpt.run exists. $ file repos/gems/run/sculpt.run repos/gems/run/sculpt.run: exported SGML document, ASCII text However there is an error: $ make -C build/x86_64 run/sculpt KERNEL=nova BOARD=pc make: Entering directory '/home/shlomo/development/all_git/genode.git/genode/build/x86_64' Error: No run script for sculpt make: *** [Makefile:323: run/sculpt] Error 255 make: Leaving directory '/home/shlomo/development/all_git/genode.git/genode/build/x86_64' Best regards, S.P.

2 2

Warning and Error while running on QEMU.
by Shlomo Pongratz 19 Apr '20

19 Apr '20

Hi, While running sculpt-20.02 on QEMU according to the instructions from https://genode.org/download/sculpt e.g. qemu-system-x86_64 -display sdl -cpu core2duo -machine q35 \ -m 1024 -drive format=raw,file=sculpt-20.02.img I've noticed the following warning and error: [init] Warning: runtime assigned RAM (32G) exceeds available RAM (659914929) [init] Error: RAM preservation exceeds available memory The maximum amount I can give the QEMU is 8G, is there anyway one can lower the memory requirement for a prebuild image? Best regards, S.P.

1 0

Shutdown
by Shlomo Pongratz 17 Apr '20

17 Apr '20

Hi, I'm running sculpt-20.02.img on QEMU and I wonder if the only way to shutdown sculpt is to kill QEMU? Is it the same for real HW (e.g. to push the reset button)? I'm sure it is somewhere in the documentation but I've failed to find it. Best regards, S.P.

4 5

Creating files with the new file manager
by Star Lord 14 Apr '20

14 Apr '20

Hello friends! Can I create files using the new built in file manager? I can view and edit existing files and directories, but can't find an option to create a new file. nooby

2 1

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

users