This is a text-only version of the following page on https://raymii.org: --- Title : Install NetBSD 9.2 on a DEC Alpha CPU in QEMU with X11 Author : Remy van Elst Date : 18-08-2021 Last update : 27-08-2021 URL : https://raymii.org/s/articles/NetBSD_on_QEMU_Alpha.html Format : Markdown/HTML --- This is a guide on installing and running NetBSD for the Alpha CPU architecture on QEMU, including a GUI (X11 via VNC). It requires you to patch and compile QEMU yourself. It was never possible, until now, to run an actual operating system *easily* with QEMU Alpha, so this is amazing! It is very cool that [Jason Thorpe][2] is putting in [so much effort on the QEMU side][29], as all but one patch is upstream already. Alpha emulation has always been a niche of a niche, so seeing this improve in QEMU is wonderful. OpenVMS does not boot yet since many more things are missing on the QEMU side, but who knows what the future might bring? Maybe even Windows NT for Alpha will boot on QEMU one day?

Recently I removed all Google Ads from this site due to their invasive tracking, as well as Google Analytics. Please, if you found this content useful, consider a small donation using any of the options below:

I'm developing an open source monitoring app called Leaf Node Monitoring, for windows, linux & android. Go check it out!

Consider sponsoring me on Github. It means the world to me if you show your appreciation and you'll help pay the server costs.

You can also sponsor me by getting a Digital Ocean VPS. With this referral link you'll get $200 credit for 60 days. Spend $25 after your credit expires and I'll get $25!

This article is based on [work][3] by [astr0baby][4]/[DoktorCranium][5] and his twitter screenshots. His guide is quite sparse, which I try to improve with this one. [astr0baby][4] does a lot of cool emulation stuff, his twitter is full of screenshots of his M1 Macbook running all kinds of legacy operating systems, from AIX to OpenVMS to MorphOS on whatever old CPU you can think of. Here is how the end result of this guide looks: ![openttd][31] > Modern OpenTTD and browsing on NetBSD on Alpha! If you're interested in Alpha emulation or running OpenVMS, check out [AXPbox][6] or read [this page][10] for more information on available emulators. AXPbox [can boot][12] the NetBSD 9.1 installer, but it doesn't install yet, due to disk emulation issues. If you get it working with 9.2 or the current NetBSD development release, please let me know! There were earlier attempts to use QEMU Alpha emulation, [here is a guide from 2014 on running Debian][27]. However, Debian [dropped the Alpha architecture] [28] after version 5.0 (lenny), so NetBSD is our only modern option. And maybe Gentoo, but I'm unsure of that. Tru64 will not run on QEMU currently. The Alpha machine QEMU emulates is based on the Alpha DP264 platform, this is probably best known as the basis of the Compaq AlphaServer ES40 and Compaq AlphaStation ES40 systems, although a number of other AlphaStations (e.g. XP900 and XP1000), AlphaServers (e.g. DS10 and DS20) and OEM boards (e.g. AlphaPC 264DP and API UP2000) were also based on it. ([via][30]) I can't even do a `cat /proc/cpuinfo`. The NetBSD specific `cpuctl identify 0` command reports `CPU identification not implemented for this architecture`. With `dmesg` you get some info, but not much: dmesg | grep MHz [ 1.000000] 6600 family, unknown model variation 0x1400, 250MHz, s/n QEMU Here's a screenshot of NeoFetch in the QEMU console: ![neofetch][41] This guide is also available for [AXPbox][40], the other open source DEC Alpha emulator. #### Article changelog * 2021-08-27: Added `openssl speed` and `sysbench` benchmarks * 2021-08-20: `pkgin` works again, repo issue fixed. Updated section in article accordingly. * 2021-08-19: Added a few responses from Jason's mailing list answer regarding kernel / emulation. ### Patching and building QEMU Due to not all patches [being upstream][2] you must build QEMU yourself. This process is rather simple, as you would expect it is nothing more than getting the source code, applying the patch and the regular `./configure` followed by `make`. To install all required build dependencies on Ubuntu, use the following command: apt-get build-dep qemu The patch file is rather small, which is why I can share it here. Place the below contents in a file named `alpha-mmclock-qemu.patch`: --- qemu/hw/rtc/mc146818rtc.c.orig 2020-10-01 00:56:55.574093880 +0000 +++ qemu/hw/rtc/mc146818rtc.c 2020-10-01 00:58:40.326479896 +0000 @@ -155,9 +155,15 @@ static uint32_t rtc_periodic_clock_ticks { int period_code; +#if 0 + /* + * Real hardware sets the PF bit rergardless if it actually + * raises an interrupt. + */ if (!(s->cmos_data[RTC_REG_B] & REG_B_PIE)) { return 0; } +#endif period_code = s->cmos_data[RTC_REG_A] & 0x0f; @@ -944,6 +950,7 @@ static void rtc_realizefn(DeviceState *d } s->periodic_timer = timer_new_ns(rtc_clock, rtc_periodic_timer, s); + periodic_timer_update(s, qemu_clock_get_ns(rtc_clock), 0, true); s->update_timer = timer_new_ns(rtc_clock, rtc_update_timer, s); check_update_timer(s); The manual process to build and install QEMU from the latest release is on [the official site][7]. You can keep an eye on [the source file mx146818rtc.c git log][1] to see if the patch process is still required. As of writing this article, it is, the latest commit there is `2021-05-13 Philippe Mathieu... hw/rtc/mc146818rtc: Convert to 3-phase reset (Resettabl...) (fae5a0420754453beca3407250899254fb6745a1)`. I'm building git commit hash `bd44d64a3879bb6b0ca19bff3be16e0093502fac`, but if you notice that the git log of the above file is different, skip the patch/checkout step and let me know if I need to update this article. Clone the QEMU source repository: git clone https://gitlab.com/qemu-project/qemu.git Navigate into the source folder: cd qemu Checkout the specific commit I used when writing this article: git checkout bd44d64a3879bb6b0ca19bff3be16e0093502fac Apply the patch: patch -p1 < alpha-mmclock-qemu.patch Output should look like: patching file hw/rtc/mc146818rtc.c Hunk #2 succeeded at 934 (offset -16 lines). Initialize all git submodules: git submodule init git submodule update --recursive Start the actual build process, first the configure step, for speed of compilation specifically for the Alpha architecture only: ./configure --target-list=alpha-softmmu Followed by make to start compilation. Replace `-j8` with your amount of CPU cores: make -j8 No `make install`, since I suspect you already have qemu installed and just want to test out stuff. This build, due to being configured with a specific target, will only work with that, not "regular" QEMU as provided by your distro. Make will take some time, when it is finished compiling, you should have a binary named `qemu-system-alpha` inside the `build` folder: file build/qemu-system-alpha Output: build/qemu-system-alpha: ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=bc41c3dfc1d574362343dbe4f8d06ad1f72806a6, for GNU/Linux 4.4.0, with debug_info, not stripped If you run the binary, you should see a black window with a green text greeting you (`Hello`): ![screenshot qemu][11] I had some configure errors (for example `ERROR: Program 'keycodemapdb/tools/keymap-gen' not found`) that were the result of the `git submodule update` command failing. I had to reset all submodules and try again with the following commands: git clean -xfd git submodule foreach --recursive git clean -xfd git reset --hard git submodule foreach --recursive git reset --hard git submodule update --init --recursive ### Installing NetBSD The installation process runs as usual once it's booted up. You use an interactive installer to install NetBSD, but to boot, we must provide the kernel to QEMU manually. The kernel file for installation is different from the one for "regular" use. #### Downloading the ISO and kernels Check [the following directory][8] for the current version of the latest `current` ISO, in my case `NetBSD-9.99.88-alpha.iso`. Download the ISO: wget https://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/images/NetBSD-9.99.88-alpha.iso Note: If the version number changed, the above command will return a 404 error and you must specify the new URL yourself. More information on how the `current` (development/daily) builds work can be found [here][9] on the NetBSD site. I asked on [the mailing list][35] why the kernels should be provided to QEMU outside of the installed disk, this is what Jason had to say: > The Qemu "bios" (palcode-clipper) for its Alpha emulation does not contain any of the I/O callbacks implemented by real DEC SRM firmware. Nor does it contain any of the CLI implemented by real DEC SRM firmware. As such, there is no way to use a boot loader program to load the kernel from the file system in the emulated machines disks. Download the two kernel files, start with the one for installation: wget http://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/alpha/installation/instkernel/netbsd.gz Uncompress it and rename so we know which one is which: gunzip netbsd.gz mv netbsd inst-netbsd My file is for 9.99.88, yours may differ: file inst-netbsd Output: inst-netbsd: ELF 64-bit LSB executable, Alpha (unofficial), version 1 (SYSV), statically linked, for NetBSD 9.99.88, stripped Same for the one for regular use: wget http://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/alpha/binary/kernel/netbsd-GENERIC.QEMU.gz Extract it: gunzip netbsd-GENERIC.QEMU.gz File information: netbsd-GENERIC.QEMU: ELF 64-bit LSB executable, Alpha(unofficial), version 1 (SYSV), statically linked, for NetBSD 9.99.88, not stripped I asked what the difference is between the installation kernel and the regular QEMU kernel file on the mailing list: > GENERIC.QEMU is exactly the same as GENERIC except that it has the kernel symbol table injected into it for use by the in-kernel debugger (and, eventually, the in-kernel linker for modules, once I finish implementing all of the necessary relocations for Alpha). #### NetBSD Installation Create a disk file which the installation will use: qemu-img create -f qcow2 disk.img 20G Start QEMU with the following command to start the installation: ./build/qemu-system-alpha \ -m 1024 \ -kernel inst-netbsd \ -drive file=disk.img,media=disk,format=qcow2,index=0 \ -drive file=NetBSD-9.99.88-alpha.iso,media=cdrom,format=raw,index=1 \ -net nic \ -net user,hostfwd=tcp::2222-:22 Change the ISO filename if the version number has changed upstream (`NetBSD-9.99.88-alpha.iso`). If you haven't applied the patch, the install will boot but hang with an error regarding the clock: ![screenshot rtc error][13] Otherwise, you will be asked what type of terminal you're using and the installation will start as you're used to: ![screenshot installer][14] Continue on with the interactive installer. I'm not covering the entire installation process since there are no special things you need to do (like the qemu patch above). ![screenshot installer][15] I choose a full installation (not minimal or without X11), installed to disk `wd0` using the CDROM as source: ![screenshot installer][16] I also setup networking during the installation, as well as enable SSH and mdns: ![screenshot installer][17] Here's the summary: ![screenshot installer][18] Once you're done installing, shut down the VM with the following command: shutdown -p now ### Running NetBSD Once NetBSD is installed, the command to start qemu is different, since we're using a different kernel file and we're telling NetBSD which drive to boot from (`wd0`). This is the new command: ./build/qemu-system-alpha \ -m 1024 \ -kernel netbsd-GENERIC.QEMU \ -drive file=disk.img,media=disk,format=qcow2,index=0 \ -drive file=NetBSD-9.99.88-alpha.iso,media=cdrom,format=raw,index=1 \ -append "rootdev=/dev/wd0" \ -net nic \ -net user,hostfwd=tcp::2222-:22 If you do get questions asked during boot, just press ENTER to accept the defaults. If all went well, you should be able to login to your fresh new shiny Alpha NetBSD: ![screenshot netbsd][19] Networking works out of the box as well: ![screenshot netbsd][20] Awesome, well done. For the final part, we're going to install and setup TigerVNC so we can play around with a GUI. #### Connecting via SSH to NetBSD The QEMU VM is started with the option `-net user,hostfwd=tcp::2222-:22`. This means that on you local machine, port `2222` is forwarded to port `22` on the VM. To SSH into the NetBSD machine (assuming networking is setup and working), you can connect to that port on localhost: ssh root@127.0.0.1 -p 2222 I had to allow root login (`vi /etc/ssh/sshd_config`, `PermitRootLogin yes`, `service sshd restart`), which in this case, for this local test machine not directly exposed to the internet, is not a security risk. ![SSH][25] If you're unable to install packages (for whatever reason), you can use the built-in features to have a GUI via SSH and Xnest. See [my other article] [38] how to do that. ### Setup X11 and VNC (GUI) X is installed by default, but crashes with a segfault if [you run it][33] on the QEMU VGA console (`X -configure`, then `X -config /root/xorg.conf.new`). Therefore, we're going to use VNC, which does run and is quite fast and usable. The `tigervnc` server is not included in the installation, which is why we start with setting up the package manager and installing a few packages. A screenshot of the segfault: ![x11 crash][23] #### Installing binary packages (pkgsrc) Assuming networking works, we can use [the usual way][32] to install binary packages. How amazing it is that NetBSD provides binary packages for Alpha, compiling them would otherwise take very long. Type the following commands in the shell to install the binary package manager: PATH="/usr/pkg/sbin:/usr/pkg/bin:$PATH" PKG_PATH="https://cdn.NetBSD.org/pub/pkgsrc/packages" PKG_PATH="$PKG_PATH/NetBSD/alpha/9.2/All/" export PATH PKG_PATH pkg_add pkgin Update the package sources: pkgin update After that, you can install tigervnc: pkgin install tigervnc **When writing this article, there was an issue with the repository, the index file was corrupt. The below section covers that, but you can probably skip it if you're not interested in NetBSD's packaging system. You can continue with the section `TigerVNC Configuration`. After sending a message on the `port-alpha` mailing list, [John fixed the repository][36].** In my case, `pkgin update` gave an error, `pkgin: Short read of pkg_summary: truncated bzip2 input`. This seems to be a [corrupt file][21], which I cannot fix from here. I did however saw on the directory index, that there also was a gzipped version (not bzip2), `pkg_summary.gz`, but I could not find a way to get that file instead of the bzip2 file. `pkgin` is a handy wrapper that mimicks `apt/yum`, but underneath it uses `pkg_*`, as far as I know, which we can also do manually. The repository URL is setup after we've exported the variables earlier, so you should be able to just do: pkg_add fetch You can also specify the entire URL to a package: pkg_add ftp://ftp.netbsd.org/pub/pkgsrc/packages/NetBSD/alpha/9.2/All/fetch-1.9nb1.tgz For the exact filename you can [browse the web index of Alpha packages][22]. You can also use `ftp` to download the package, then install it locally: ftp -i ftp://ftp.netbsd.org/pub/pkgsrc/packages/NetBSD/alpha/9.2/All/fetch-1.9nb1.tgz pkg_add ./fetch-1.9nb1.tgz That will however still connect to the network to fetch dependencies, so if you're doing this offline (lets say via an NFS share), you might run into issues there. Install the `tigervnc` package: pkg_add tigervnc It will fetch some dependencies and should install correctly: ![vnc install][24] ### TigerVNC configuration Configuring tigerVNC is simple since most configuration is already done via the package. Start by setting up a VNC password: vncpasswd Enter the password twice: Password: Verify: Answer no to a view only password: Would you like to enter a view-only password (y/n)? n Start the VNC server: vncserver Output: New 'netbsd.alpha.raymii.org:1 (root)' desktop is netbsd.alpha.raymii.org:1 Creating default startup script /root/.vnc/xstartup Creating default config /root/.vnc/config Starting applications specified in /root/.vnc/xstartup Log file is /root/.vnc/netbsd.alpha.raymii.org:1.log To connect to your shiny new awesome desktop, you must use SSH to forward a port. Disconnect your SSH session and reconnect with a new flag for a local port forward: ssh root@127.0.0.1 -p 2222 -L 5901:127.0.0.1:5901 This forwards port 5901 (VNC port for display `:1`) on your local machine to port `5901` on the virtual machine. After logging in, you should be able to use any VNC client to connect right away: ![vnc yay][26] How cool is that? A modern NetBSD with a graphical desktop on a CPU architecture (Alpha) running in QEMU, with networking. After rebooting, you need to enter the `vncserver` command again to start VNC. You can also setup a service to autostart at boot, but I'm not going to cover how to do that in this guide. You've gotten this far on a weird CPU emulation article, I bet you that you're very capable of figuring autostarting VNC out yourself. For some more fun, you can install a browser (`netsurf` or `dillo`, firefox or chrome are not available as packages) or a game (my favorite, `openttd` is available as a package including the free graphics `openttd-data`). OpenTTD is quite playable. Not very fast, but a small map seems to run okay. It even seems `xfce4` packages are available in the repository, so you could deck out a full featured desktop. I've tried installing it, but after 18 hours or so, it's still running. Netsurf took just as long, and it seems to be the installation of GTK that takes so much time. As Jason says on the mailing list, it might be related to disk IO: > Yikes. I'll need pkg install experts to weigh in on this. It's entirely possible that it's bottlenecked on the disk emulation. I'm working on getting VirtIO to work on alpha so hopefully we can speed disk and network access up a little. If there is any software you do want to run but not available as a binary package, compiling it yourself (or letting `pkgsrc` do it for you) is always an option. ![openttd][31] I managed to compile `python2` via `pkgsrc`, the `cvs checkout` of `pkgsrc` took a day or so to complete: ![python2][37] The setup of `pkgsrc` for source compilation is documented on the NetBSD website but this is the process I used, if you want to try it out: cd /usr && cvs -q -z2 -d anoncvs@anoncvs.NetBSD.org:/cvsroot checkout -r pkgsrc-2021Q1 -P pkgsrc cd /usr/pkgsrc/lang/python27 make fetch-list | sh; # fetch dependencies make NO_CHECKSUM=yes make install If you get any segfaults or `internal compiler errors` during building, my experience is that those go away if you retry the build. (Do a `make clean` first). Below is an example error message, which did not occur when I retried the compilation: mv -f .deps/loadapi.Tpo .deps/loadapi.Po gcc -DLOCALEDIR=\"/usr/pkg/share/locale\" -DLIBDIR=\"/usr/pkg/lib\" -DINCLUDEDIR=\"/usr/pkg/include\" -DHAVE_CONFIG_H -I. -I./glob -I/usr/include -I/usr/pkg/include -O2 -mieee -D_FORTIFY_SOURCE=2 -I/usr/include -MT main.o -MD -MP -MF .deps/main.Tpo -c -o main.o main.c during GIMPLE pass: cddce main.c: In function 'main': main.c:1066:1: internal compiler error: Segmentation fault 1066 | main (int argc, char **argv, char **envp) | ^~~~ no stack trace because unwind library not available Please submit a full bug report, with preprocessed source if appropriate. See for instructions. *** Error code 1 Stop. make[2]: stopped in /usr/pkgsrc/devel/gmake/work/make-4.2.1 *** Error code 1 Astrobaby even got DOOM running: ![doom][34] ### OpenSSL Speed Benchmark [This article][30] has some benchmarks on Debian 5.0 on QEMU Alpha, including `OpenSSL speed`. [EmuVM][39] has a few benchmarks as well, both real Alpha machines as their commercial emulator. Most of the other benchmarks do not run on NetBSD (Only OpenVMS or Tru64), so reproducing that is hard. OpenSSL however, is included on NetBSD by default. Newer version, but hopefully comparable results. `OpenSSL speed` is a performance benchmark for each of the available algorithms. Because this is an old CPU architecture, and because [the other article][30] does so (as well as [EmuVM][39]), I'm trying just the `md5` and `rsa` benchmarks. **The OpenSSL benchmark fails**. Results overflow on the RSA side and MD5 doesn't give any numbers. Not sure if this is an emulation issue, or a NetBSD issue. Might also be in OpenSSL, it's quite a new version(1.1.1) and I'm not sure if the OpenSSL project still tests on Alpha machines. I'm including the results, for future reference it might be interesting. For comparison however, they are useless. Results of `rsa` calculations: netbsd# openssl speed rsa Doing 512 bits private rsa's for 10s: Doing 512 bits public rsa's for 10s: 100761 512 bits public RSA's in 0.00s Doing 1024 bits private rsa's for 10s: 2142 1024 bits private RSA's in 0.00s Doing 1024 bits public rsa's for 10s: 113632 1024 bits public RSA's in 0.00s Doing 2048 bits private rsa's for 10s: 1138 2048 bits private RSA's in 0.00s Doing 2048 bits public rsa's for 10s: 40512 2048 bits public RSA's in 0.00s Doing 3072 bits private rsa's for 10s: 385 3072 bits private RSA's in 0.00s Doing 3072 bits public rsa's for 10s: 20155 3072 bits public RSA's in 0.00s Doing 4096 bits private rsa's for 10s: 174 4096 bits private RSA's in 0.00s Doing 4096 bits public rsa's for 10s: 11464 4096 bits public RSA's in 0.00s Doing 7680 bits private rsa's for 10s: 29 7680 bits private RSA's in 0.00s Doing 7680 bits public rsa's for 10s: 3540 7680 bits public RSA's in 0.00s Doing 15360 bits private rsa's for 10s: 4 15360 bits private RSA's in 0.00s Doing 15360 bits public rsa's for 10s: 913 15360 bits public RSA's in 0.00s OpenSSL 1.1.1k 25 Mar 2021 NetBSD 9.99.88 options:bn(64,64) rc4(int) des(int) aes(partial) idea(int) blowfish(ptr) gcc version 10.3.0 (NetBSD nb1 20210411) sign verify sign/s verify/s rsa 512 bits infs 0.000000s 0.0 923036700778489577472.0 rsa 1024 bits 0.000000s 0.000000s 20953569512310820864.0 1054135706452407615488.0 rsa 2048 bits 0.000000s 0.000000s 11156638280728354816.0 374733567055141208064.0 rsa 3072 bits 0.000000s 0.000000s 3761703789465102336.0 186142468490230857728.0 rsa 4096 bits 0.000000s 0.000000s 1695153782552369408.0 105969518684333555712.0 rsa 7680 bits 0.000000s 0.000000s 278990477569769280.0 32383696070560972800.0 rsa 15360 bits 0.000000s 0.000000s 37610866572920040.0 8378888909113309184.0 Results of `md5` calculations: netbsd# openssl speed md5 Doing md5 for 3s on 16 size blocks: Doing md5 for 3s on 64 size blocks: Doing md5 for 3s on 256 size blocks: Doing md5 for 3s on 1024 size blocks: Doing md5 for 3s on 8192 size blocks: Doing md5 for 3s on 16384 size blocks: OpenSSL 1.1.1k 25 Mar 2021 NetBSD 9.99.88 options:bn(64,64) rc4(int) des(int) aes(partial) idea(int) blowfish(ptr) gcc version 10.3.0 (NetBSD nb1 20210411) The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes md5 0.00 0.00 0.00 0.00 0.00 0.00 I did try to play with some options (longer runtime, `-elapsed` flag (`Use wall-clock time instead of CPU user time as divisor`) but none helped. The weird thing is that these two commands work fine on [AXPbox with NetBSD][40], which makes me suspect that there is an issue with QEMU emulation as opposed to an issue in OpenSSL. The other article has OpenSSL results from Debian 5.0, so it might be a combination of this OpenSSL version with QEMU, but I'm not sure. ### SysBench benchmark The [NetBSD Alpha][22] package repository does have a `sysbench` package. Sysbench is quite a populair cross platform benchmarking tool and it even gave reasonable results, as opposed to OpenSSL. It's an older version, currently we're at sysbench 1.0.20 and NetBSD has 0.4.12. CPU: sysbench --test=cpu run --max-time=30 sysbench 0.4.12: multi-threaded system evaluation benchmark Running the test with following options: Number of threads: 1 Doing CPU performance benchmark Threads started! Time limit exceeded, exiting... Done. Maximum prime number checked in CPU test: 10000 Test execution summary: total time: 30.0547s total number of events: 532 total time taken by event execution: 30.0458 per-request statistics: min: 53.15ms avg: 56.48ms max: 132.93ms approx. 95 percentile: infms Threads fairness: events (avg/stddev): 532.0000/0.00 execution time (avg/stddev): 30.0458/0.00 Memory: netbsd# sysbench --test=memory run --max-time=60 sysbench 0.4.12: multi-threaded system evaluation benchmark Running the test with following options: Number of threads: 1 Doing memory operations speed test Memory block size: 1K Memory transfer size: 102400M Memory operations type: write Memory scope type: global Threads started! Time limit exceeded, exiting... Done. Operations performed: 1257653 (20960.15 ops/sec) 1228.18 MB transferred (20.47 MB/sec) Test execution summary: total time: 60.0021s total number of events: 1257653 total time taken by event execution: 56.4853 per-request statistics: min: 0.02ms avg: 0.04ms max: 202.10ms approx. 95 percentile: infms Threads fairness: events (avg/stddev): 1257653.0000/0.00 execution time (avg/stddev): 56.4853/0.00 A 30 second timeout failed the test, which is why the max time for memory is 60. File I/O: netbsd# sysbench --test=fileio --file-test-mode=seqwr --file-num=8 run --max-time=60 sysbench 0.4.12: multi-threaded system evaluation benchmark Running the test with following options: Number of threads: 1 Extra file open flags: 0 8 files, 256Mb each 2Gb total file size Block size 16Kb Periodic FSYNC enabled, calling fsync() each 100 requests. Calling fsync() at the end of test, Enabled. Using synchronous I/O mode Doing sequential write (creation) test Threads started! Done. Operations performed: 0 Read, 131072 Write, 8 Other = 131080 Total Read 0b Written 2Gb Total transferred 2Gb (34.332Mb/sec) 2197.25 Requests/sec executed Test execution summary: total time: 59.6529s total number of events: 131072 total time taken by event execution: 57.6667 per-request statistics: min: 0.11ms avg: 0.44ms max: 13918.59ms approx. 95 percentile: infms Threads fairness: events (avg/stddev): 131072.0000/0.00 execution time (avg/stddev): 57.6667/0.00 The default number of files (128) failed with an error message, too many open files, which is why I tried with 8. Comparing the results to AXPbox is interesting. Summarized below are the AXPbox results: CPU: total number of events: 19 Memory: 447.07 MB transferred (7.45 MB/sec) Disk: Operations performed: 0 Read, 37353 Write, 0 Other = 37353 Total Read 0b Written 583.64Mb Total transferred 583.64Mb (9.7263Mb/sec) 622.48 Requests/sec executed These are the QEMU results, running on the same host machine and the same NetBSD version: CPU: total number of events: 532 Memory: 1228.18 MB transferred (20.47 MB/sec) Disk: Operations performed: 0 Read, 131072 Write, 8 Other = 131080 Total Read 0b Written 2Gb Total transferred 2Gb (34.332Mb/sec) 2197.25 Requests/sec executed The emulation provided by QEMU seems to be faster by a large amount, on all three tests. This is a basic benchmark and we have to keep in mind that both emulators emulate different hardware and that the core code of AXPbox is over 10 years old. A comparison with the machine that runs the QEMU VM (x86, Intel i7-8750H): $ sysbench --test=cpu run --max-time=30 WARNING: the --test option is deprecated. You can pass a script name or path on the command line without any options. WARNING: --max-time is deprecated, use --time instead sysbench 1.0.20 (using system LuaJIT 2.0.5) Running the test with following options: Number of threads: 1 Initializing random number generator from current time Prime numbers limit: 10000 Initializing worker threads... Threads started! CPU speed: events per second: 1369.69 General statistics: total time: 30.0010s total number of events: 41094 Latency (ms): min: 0.71 avg: 0.73 max: 3.78 95th percentile: 0.75 sum: 29991.75 Threads fairness: events (avg/stddev): 41094.0000/0.00 execution time (avg/stddev): 29.9918/0.00 [1]: https://git.qemu.org/?p=qemu.git;a=history;f=hw/rtc/mc146818rtc.c [2]: https://archive.md/7LXZK [3]: http://web.archive.org/web/20210817034654/http://45.76.81.249/AXP/NetBSD/ [4]: https://mobile.twitter.com/astr0baby/ [5]: https://github.com/DoktorCranium [6]: /s/blog/AXPBox-version-1.0.0-released.html [7]: https://www.qemu.org/download/ [8]: https://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/images/ [9]: https://www.netbsd.org/releases/current.html [10]: http://www.avanthar.com/healyzh/decemulation/Alpha.html [11]: /s/inc/img/qemu-system-alpha-1.png [12]: https://github.com/lenticularis39/axpbox/wiki/Guest-support [13]: /s/inc/img/qemu-system-alpha-2.png [14]: /s/inc/img/qemu-system-alpha-3.png [15]: /s/inc/img/qemu-system-alpha-4.png [16]: /s/inc/img/qemu-system-alpha-5.png [17]: /s/inc/img/qemu-system-alpha-6.png [18]: /s/inc/img/qemu-system-alpha-7.png [19]: /s/inc/img/qemu-system-alpha-8.png [20]: /s/inc/img/qemu-system-alpha-9.png [21]: http://mail-index.netbsd.org/port-sparc64/2020/06/24/msg002966.html [22]: https://cdn.netbsd.org/pub/pkgsrc/packages/NetBSD/alpha/9.2/All/ [23]: /s/inc/img/qemu-system-alpha-11.png [24]: /s/inc/img/qemu-system-alpha-10.png [25]: /s/inc/img/qemu-system-alpha-12.png [26]: /s/inc/img/qemu-system-alpha-13.png [27]: http://web.archive.org/web/20160509075529/http://systems.firstworks.com/2014/02/debian-5010-alpha.html [28]: https://lists.debian.org/debian-devel-announce/2009/10/msg00000.html [29]: https://web.archive.org/web/20210306121313/http://mail-index.netbsd.org/port-alpha/2020/02/26/msg001040.html [30]: https://boxes-of-tat.blogspot.com/2021/06/linux-for-alpha-on-qemu-system.html [31]: /s/inc/img/qemu-system-alpha-14.png [32]: https://www.netbsd.org/docs/pkgsrc/using.html [33]: https://www.netbsd.org/docs/guide/en/chap-x.html [34]: /s/inc/img/qemu-system-alpha-15.png [35]: http://mail-index.netbsd.org/port-alpha/2021/08/19/msg001212.html [36]: http://mail-index.netbsd.org/port-alpha/2021/08/20/msg001216.html [37]: /s/inc/img/qemu-system-alpha-16.png [38]: /s/articles/Remote_desktop_on_NetBSD_with_Xnest_no_VNC.html [39]: http://web.archive.org/web/20210827055720/https://emuvm.com/support/benchmarks/ [40]: /s/articles/NetBSD_Alpha_AXPbox.html [41]: /s/inc/img/qemu-system-alpha-17.png --- License: All the text on this website is free as in freedom unless stated otherwise. This means you can use it in any way you want, you can copy it, change it the way you like and republish it, as long as you release the (modified) content under the same license to give others the same freedoms you've got and place my name and a link to this site with the article as source. This site uses Google Analytics for statistics and Google Adwords for advertisements. You are tracked and Google knows everything about you. Use an adblocker like ublock-origin if you don't want it. All the code on this website is licensed under the GNU GPL v3 license unless already licensed under a license which does not allows this form of licensing or if another license is stated on that page / in that software: This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . Just to be clear, the information on this website is for meant for educational purposes and you use it at your own risk. I do not take responsibility if you screw something up. Use common sense, do not 'rm -rf /' as root for example. If you have any questions then do not hesitate to contact me. See https://raymii.org/s/static/About.html for details.