Version 2019.04.25 — The "OS/8 V3F and os8-run" release
The banner feature in this release is that Bill Cattey transformed our
mkos8tool into the
os8-runscript interpreter, giving us many new features and capabilities:
- The OS/8 V3D RK05 media build steps previously hard-coded in
mkos8are now in a series of scripts in the
os8-runlanguage. This abstracts the process, making it easier to understand and change. Non-Python programmers can examine these
os8-runinput scripts to learn how their media are built. It's easier to modify the
os8-runscripts to get custom results than to modify the prior release's
Replaced the hand-maintained
media/os8/os8.tu56OS/8 V3D TU56 tape image used by boot option IF=3 with an
os8-runscript that's run at build time to generate a series of four similar tape images from pristine, curated source media, just like we did for the RK05 media in the prior release.
You can ask to have the IF=3 boot option use the OS/8 V3F tape instead of the OS/8 V3D default. Both versions get built, always, but only one gets to be the boot tape for this option. This may be the first time that V3F was ever easily available to SIMH users. Previously, you had to assemble it from pieces found all over the Internet.
(Bootable V3F RK05 media are planned for a future release.)
You can also select whether the IF=3 boot option uses the boot tape configured for a PDP-8 with the TD8E tape controller instead of the default, which assumes a TC08. The prior release was hard-coded for the TD8E.
This new driver is more efficient and it will allow you to copy the medium image to an actual DECtape and boot it on a PDP-8 with the TC08 or compatible tape drive controller. That wasn't possible before because the TD8E and TC08 share the same IOT device ID, so you can have only one or the other installed into OS/8 at a time, and that tape didn't have
BUILDon it, so the manually maintained tape image we were running didn't allow you to easily switch it.
In the prior release, we offered the Python
simhAPI for scripting OS/8 and SIMH, plus the
teco-pi-demoscript to show off this API. The
os8-runcommand language is based on this same mechanism, offering a simpler method of achieving custom results. The
os8-runscript language is more like a command language at this time than a programming language, so it should be much easier for non-programmers to learn.
We've been working on this new mechanism for many months now, during which we did a lot of manual testing. But as with
mkos8before it, we've also got an automatic tester for
os8-run, which gives us tested quality assurance on several axes:
Complete: We've repeatedly tested all 32768 possible OS/8 RK05 configurations afforded by the configure script.
Repeatable: We've re-tested those builds on the original build systems, on multiple computers at a site, and across sites. This assures us that the builds are deterministic: given the same inputs, you always get the same outputs. That sounds like it should be obvious and easy, but it didn't come for free!
Platform-Independent: By copying these test exemplars between Pi and non-Pi systems and re-testing, we've convinced ourselves that the build is platform-independent, at least within the scope of the systems we've tried it on.
Reliable: Some of these configurations have been tested many times over, and all of them have been built at least four times. We've come to expect that this new mechanism will reliably build standard media on your Pi, too.
This is largely the work of Bill Cattey.
- The OS/8 V3D RK05 media build steps previously hard-coded in Python within
Added Bill Cattey's
os8-cpscript, which makes it nearly as easy as
cp(1)to get files into and out of a SIMH OS/8 media image.
Also added his
diff-os8program, which is not general-purpose, but it shows a useful application of
os8-cp: to compare two RK05 media images by copying out all of the files from OS/8 into the local filesystem and then comparing the files individually on the host. Imagine your own possibilities!
Major improvements to Ian Schofield’s CC8 compiler:
Bill Cattey retargeted the CC8 cross-compiler against SmallC-85, which greatly improves the capabilty of the compiler. It also allows this compiler to run without crashing on 64-bit hosts for the first time.
Bill merged in some of the updates Ian has made to CC8, which affects the native OS/8 compiler and its standard library.
Bill wrote an
os8-runscript to generate the
cc8.tu56DECtape image dynamically at build time, as needed, rather than use a static binary image shipped in the Fossil code repository. The primary practical upshot of this is that you can now change the native OS/8 CC8 source code on the host side and just say “
make” to get a new RK05 disk pack with the new code running in it. If you don’t get how cool this is, you don’t understand it properly. :)
Warren Young quadrupled the size of the CC8 user manual. It now answers many more questions, reveals many previously hidden details, fully documents LIBC’s interfaces and internal behaviors, and documents the CC8 memory model.
Warren and Ian collaborated on fixes to the native compiler and its LIBC to fix a bunch of bugs and improve its conformance to Standard C. It’s still miles from passing any ISO C conformance suite, but it should violate fewer expectations now. This work does change the API and ABI of CC8’s LIBC somewhat, so if you have existing code, you might want to read the new manual to work out what’s needed to port that code.
Notable improvements are that itoa() now takes a radix parameter to match its implementation in other C libraries; sprintf() returns an error code when the format string contains an unsupported format specifier; and the printf() family of functions now handle %x and %X properly. In the prior release, only %x was supported, and it gave uppercase output, not lowercase as the Standard requires.
Warren changed CC8 to use octal when generating constants in SABR output, that being SABR’s default radix. Since CC8 leaves SABR in its default octal mode, the primary user benefit of this is that inline assembly now behaves the same in CC8 as in OS/8 FORTRAN II, which is also built atop SABR. That is to say, your inline assembly code can safely assume that the assembler is in octal mode when it processes your code.
This does mean that if you had C programs built with CC8 that had inline assembly and that code had integer constants within it, it will have to be changed to work with the new compilers.
The default radix for C code remains 10, so if you were not using inline assembly, this change does not affect you.
Since the beginning of this project, we've called our modified version of the SIMH PDP-8 simulator
pidp8i-sim. With this release, we hard link that program to
pdp8, the simulator's name in the upstream distribution of SIMH. When called by that name, our simulator suppresses all of the PiDP-8/I extensions.
This not only means we don’t start the GPIO thread that blinks the LEDs and scans for switch changes, it means we revert behavioral changes like the one that affects how the HLT instruction ("halt") is processed: just as with the upstream version, this drops you down to the SIMH command prompt when you call the simulator as
pdp8, rather than halt the processor and wait for a front panel CONT or START keypress to get it out of STOP state.
If you install this software on a non-Raspbian systemd-based Linux distribution, the new
pidp8i startcommand will notice that there is no PiDP-8/I front panel available and will run
pdp8instead of our extended
When the GPIO thread is not running, the simulator runs considerably faster. It’s not clear to me why that should be so on multi-core boxes, since the expensive parts of the PiDP-8/I extensions are running on a separate core, but the measurements are clear.
simhAPI now supports automatic transitions between OS/8 and SIMH context, largely removing the need to manage this manually as in the prior release. This is largely Bill Cattey's work.
pidp8iprogram now takes optional verb arguments that let you avoid giving verbose
servicecommands. Instead of
sudo systemctl restart pidp8ias in prior releases, you can now say
pidp8i restart, for example.
Gracefully stopping the simulator (e.g.
pidp8i stopor Ctrl-E, q) now turns off all front panel LEDs before shutting down.
We’ve made no attempt to do similar cleanups when the simulator is killed outright. The most common case is during a Pi reboot, where you’ll still see traces of the last state on the front panel briefly. We also make no attempt to cope with badness like
pidp8i-testprogram's scan-switch feature now debounces the switches in software so that each state change results in only one line printed on the console. This is not just cosmetic: the prior behavior could fool a builder into believing their correctly-working PiDP-8/I had a hardware problem.
Changed the SysV init script to a systemd unit file. This gets us several new features:
It fixes a problem introduced between Raspbian Jessie and Stretch that caused dangling
pidp8i-simprocesses when you stopped the simulator.
SysV init was pretty much limited to use by root, so we had to use
sudoin many cases even though we'd largely divorced the PiDP-8/I software from needing root privileges itself. systemd allows a service to be installed under the account of a normal user, which lets you start and stop the simulator without needing root privilege.
We get detailed service status for free via pidp8i status, a short alias for the full command, systemctl --user status pidp8i.
Gets us away from the SysV init backwards compatibility that may go away in a future release of Raspbian.
Applied a fix from Ian Schofield for a serious problem with the accuracy of the MB register lights in certain contexts, such as while in STOP mode. Bill Cattey verified this fix against a real PDP-8 at the Rhode Island Computer Museum.
Fixed some bugs in our version of the James L-W alt-serial mod feature relative to the mailing list patch. Bug reports and diagnosis by Dylan McNamee.
Fixed a bug going clear back to the epochal v20151215 release which can cause an OSR instruction to incorrectly set the Link bit if the next GPIO pin up from those used by the SR lines happens to be set when you issue that instruction.
tools/mkbootscriptprogram which translates palbart assembly listing files into SIMH boot scripts was only writing a SIMH "dep" command for the first word on a line of listing output. It’s legal in PAL format listings to have multiple words on a line, as is common with data arrays and such. Any program that makes use of that feature is affected, including the tty output from
While in there, made several other improvements to the script.
examples/hello.palprogram was badly broken in prior releases.
It was skipping the first character ("H") in its output message.
Set the 8th bit set on the ASCII output bytes in case you use a Teletype Model 33 ASR or similar, which requires mark parity.
It now uses the same optimized
pep001.pal, which we're also shipping as
Between these weaknesses and the
mkbootscriptbug fixed above, this example was entirely broken since being shipped. Our thanks for the tests and diagnosis of these problems go to Greg Christie and Bill Cattey.
This program no longer gets stuck in the TSF loop on startup if you run it with the terminal unready for output. The terminal would typically be ready when launching this program from within OS/8, but it could get stuck in other conditions, such as when running it under a freshly started simulator:
$ bin/pdp8 boot/pep001.script
Applied same high-bit improvement as for
Updated SIMH to commit ID 4e0450cff (2019-04-18):
Simulated PDP-8 devices now have description strings so that SIMH commands like
SHOW FEATURESgives more descriptive output.
SCP (the Ctrl-E command processor) has seen a lot of work since our last release, and the included PiDP-8/I boot scripts make use of two of these improvements:
ELSEdirectives in command scripts and
The core SIMH timing and throttling mechanism has seen a lot of work.
Lots of small improvements to the terminal muxer.
The software now builds and runs on FreeBSD. This just a step toward support for FreeBSD for the Raspberry Pi: we haven’t tried to make the GPIO stuff work at all yet. For now, it just lets this software be used on your FreeBSD desktop or server machine. It may allow building on other BSDs, but that is untested.
scanswitchnow returns 127 on “no GPIO” rather than 255.
16 months of maintenance and polishing: better documentation, build system improvements, etc.
Version 2017.12.22 — The "Languages and Custom OS/8 Disk Packs" release
All prior versions of the PiDP-8/I software distribution included
os8.rk05, a "Field Service Diagnostic" OS/8 disk pack image with uncertain provenance, configuration, and modification history. We have replaced that canned disk image with a script which is run at build time which programmatically assembles a set of clean OS/8 RK05 disk images from curated, pristine, tested sources based on the user's chosen configuration options.
This provides the following features and benefits as compared to the old
The PiDP-8/I software build process now builds up to three RK05 disk images:
os8v3d-bin.rk05is a bootable OS/8 V3D disk configured according to your chosen configuration options, which are described below and in
README.md. It is made from the pristine DECtape images shipped by DEC for OS/8 V3D plus several third-party tapes curated for and built by the project's maintainers. See the OS/8 media
README.mdfile for more details.
os8v3d-patched.rk05is a copy of the
bindisk with most of the patches DEC published over the years for OS/8 V3D applied. That set of patches was chosen and tested for applicability to our modern PiDP-8/I world and for mutual compatibility.
This is the boot disk used for the IF=0 and IF=7 cases unless you give
configurescript, in which case these boot options use the
os8v3d-src.rk05is a non-bootable disk containing the contents of the OS/8 V3D source code tapes plus the source code for the extensions to OS/8 V3D. The ten TU56 tape images used as input to this process are also included among the PiDP-8/I software — see
media/os8/al-*-sa-*.tu56— but we find it much more convenient to explore the sources on a single RK05 disk than to repeatedly attach and detach the TU56 tapes.
You can suppress building this with
Default versions of these disk images are also now published on the project's home page for the benefit of those not running our PiDP-8/I software. There are quite a few OS/8 RK05 disk images floating around on the Internet these days, and many of them have bugs and breakage in them that we've fixed. It would completely fail to break our hearts if these images were used by many people outside the PiDP-8/I project.
U/W FOCAL V4E is installed on SYS: by default. Start with our U/W FOCAL Manual Supplement for the PiDP-8/I, then follow links from there to further information.
The primary linked source is the U/W FOCAL Manual by Jim van Zee (creator of U/W FOCAL) converted from scanned and OCR'd PDF form to Markdown format, which Fossil renders nicely for us on the web.
This is a fascinating programming language, well worth studying!
Ian Schofield's CC8 OS/8 C compiler is installed on
SYS:by default, and its examples and other files are on
DSK:. We have also merged in his
cc8host-side cross-compiler. See the CC8
This is a considerably improved compiler relative to what was distributed on the mailing list in August 2017. Ian has been working within the PiDP-8/I project since that initial public release, which we are now distributing publicly for the first time. We thank him for trusting us to host and distribute his project.
The MACREL v2 macro assembler and its associated FUTIL V8B tool are installed by default. Not only is this new functionality relative to prior releases of the PiDP-8/I software, it is a considerable upgrade over to the original MACREL and FUTIL V7 that are more commonly found on the net.
DCP disassembler is installed by default.
John Comeau's CHECKMO-II chess program is installed by default.
By default, SIMH no longer folds lowercase input and output to uppercase. Instead, we apply patches to OS/8's command processor and its BASIC implementation to up-case input, since neither OS/8 nor BASIC can cope with lowercase input.
All other programs are left to fend for themselves, which often works out fine. U/W FOCAL, Adventure, and TECO all handle lowercase input to some extent, for example, and all three can emit lowercase text if given it. With the prior SIMH setting, you could not use lowercase in these programs at all.
This default can be overridden. See the documentation for the new
--lowercaseconfiguration option in
INIT.TXmessage displayed by default on OS/8 boot is now more informative than the old
FIELD SERVICE PDP-8 DIAGNOSTIC SYSTEMmessage. It also now uses lowercase unless you built the simulator to force uppercase with
Those that do not want any boot message can disable it at configuration time with the
The message can be modified by editing
make, which will rebuild the OS/8 media.
All of the above features can be disabled if not wanted, as can several features present on the old
os8.rk05disk: Adventure, FORTRAN IV, FORTRAN II, Kermit-12, and the BASIC game and demo programs.
You can disable each feature above with a
--disable-os8-*option to the
configurescript, or you can disable all of them collectively with the
--os8-minimaloption, which gets you a nearly bare-bones OS/8 installation with lots of spare disk space with which you can do what you want.
Replaced the mismatched FORTRAN compiler and runtime with matched versions from the distribution DECtapes, and ensured that Adventure runs under this version of the FORTRAN Run Time System (FRTS). At various points in the past, either FORTRAN or Adventure has been broken.
Repaired several broken BASIC programs on
RKB0:by going back to primary sources. Either the
os8.rk05disk image was corrupted at some point or it is an image of a real RK05 disk pack that was corrupted, causing several of these BASIC programs to not run properly.
*.MUand music player files are left off of
RKB0:by default, since they apparently do not cause sufficient RFI on the PiDP-8/I hardware to be picked up by normal AM radios. This saves space for things that do work.
No longer installing the
VTEDITmacros for TECO by default. Although some may enjoy this alternative way of running TECO, it was decided that we should offer stock TECO by default for its historical value. If you want VTEDIT back, it can be re-enabled with a
In the old
os8.rk05disk image, both
DSK:were set to
RKA0:, which meant that to address anything on
RKB0:, you had to specify the device path explicitly or manually change the default in OS/8 with an
ASSIGN RKB0 DSKcommand or similar.
In the new disk pack, programs run with the OS/8
Rcommand are installed on
RKA0:which is the
SYS:disk, and user data files, FOCAL and BASIC programs, etc. are on
RKB0:which is assigned as
DSK:. This means OS/8 and its programs are now far more likely to find files without an explicit device name, because files are installed where OS/8 looks for them by default. Example:
.R FRTS ⇠ loads FRTS from SYS: (RKA0:) *ADVENT ⇠ loads ADVENT.LD from DSK: (RKB0:) *[Esc] Welcome to Adventure!!
Notice that no device names had to be specified. OS/8 did the right thing by default here, even though the files involved are on two separate OS/8 devices.
To a very rough approximation,
SYS:on these new RK05 disk packs acts like the Unix
DSK:acts like your user's home directory.
The idea for this came from the
cc8.rk05disk image which Ian Schofield included with his original CC8 distribution on the PiDP-8/I mailing list.
(A version of
cc8.rk05is still buried in our code repository if you want to pull it out, but it is not part of the release because our
os8v3d-patched.rk05disk image is functionally a complete replacement.)
OS/8 has a limit on the number of devices it can support, and we made different choices than the creator of
Briefly, we replaced the second floppy (
RXA1:) with a third RK05 disk, that being deemed a more useful configuration for this hard disk based OS/8 configuration. A dual-floppy configuration implies that you are booting from floppy and need the second one for user files and such. In our RK05 based configuration, users should need floppy disk support rarely, and then primarily to get data on and off of the attached hard disk(s).
We chose to stick with the dual TU56 tape drive setup of the prior version as we found the ability to mount two tapes very helpful, particularly during the
The difference in the
RESORCoutput between the versions is:
Old: RKA0,RKB0,RKA1,RKB1, RXA0,RXA1,DTA0,DTA1,TTY,LPT,PTP,PTR New: RKA0,RKB0,RKA1,RKB1,RKA2,RKB2,RXA0, DTA0,DTA1,TTY,LPT,PTP,PTR
This automatic OS/8 media build feature was suggested by Jonathan Trites who wrote the initial version of the script that is now called
libexec/mkos8. That script was then extended and factored into its current form by Bill Cattey and Warren Young. The author of Autosetup, Steve Bennett, helped with the code which allows the
mkos8scripts to share a single set of option definitions.
Warren thinks Bill did most of the hard work in the items above.
The source media used by the
mkos8script comes from many sources and was curated for the PiDP-8/I project by Bill Cattey. See the OS/8 media README for more details.
See the the top-level
READMEfor information on modifying the default OS/8 configuration. Pretty much everything above can be disabled if it's enabled by default, and vice versa.
Added several new wiki articles covering the above:
More Project Euler Problem #1 solutions in:
Demos in BASIC, describing
OS/8 Console TTY Setup, describing how we have modified the stock behavior of OS/8 to behave appropriately with a glass terminal or SSH on its console, as opposed to its default behavior, which assumes a teletype.
OS/8 LCSYS.BI Disassembled, a symbolic disassembly of the
LCSYS.BIpatch we distribute with the system, which is widely available online elsewhere. That script is a raw binary patch, which makes its operation a mystery unless you happen to be able to read PDP-8 machine code.
Added Bill Cattey's
txt2ptpprogram which converts plain ASCII text files files to the paper tape format used by SIMH, which eases transfer of text data into the simulator. That program is also linked to
ptp2txt, which makes it perform the inverse function: given a SIMH paper tape image file, produce an ASCII text file on the host machine with its contents.
This program was originally written to ease the movement of FOCAL program text between SIMH and its host OS, but it is now a key part of the OS/8 RK05 disk building process, used whenever we need to inject the contents of a text file from the host into the simulated PDP-8 running OS/8.
These filters should prove broadly useful.
Integrated Robert Krten's
d8tapePDP-8 host-side disassembler. This is distinct from the OS/8 DCP disassembler, which runs inside the simulator. It is intended as a companion to
palbart, which we integrated last year.
Added a new "blinkenlights" demo program called
bin/teco-pi-demowhich drives SIMH from the outside, running a TECO macro under OS/8 to calculate pi to 248 digits at a very slow rate, producing a display that manages to land somewhere between the random default display of Deeper Thought and the clear, boring patterns of our preexisting IF=5 demo script.
Why 248 digits? Because at 249, TECO8 runs out of memory, aborting the demo early. At the default execution rate, it would take over 17 years to complete this demo, making it a good choice to run on PiDP-8/I units primarily being used as objets d'art. The demo has a finite run time, but your Raspberry Pi is likely to die before it completes.
This script is also included as a demonstration of how the end user can reuse the technology that we developed to automatically build the custom OS/8 disk images described above to achieve different ends. Perhaps you have some other program you'd like to run within SIMH in an automated fashion? This shows one way how, and demonstrates a pre-built and tested set of tools for achieving it.
We have also written a tutorial explaining how
bin/teco-pi-demoworks and how to reuse the components it is built atop for your own ends.
This demo also has a benchmark mode (command line option
-b) which has two purposes:
- It lets you see how much faster your host system runs PDP-8 code than a Raspberry Pi Model B+ running the PiDP-8/I simulator.
- Given that information, the benchmark overrides a hardcoded
timing value in the source code as distributed which prevents
teco-pi-demofrom spamming the OS/8 terminal input handler. The default is for the slowest host we support this software on, that same Model B+ referred to above, but if we know you're running on a faster host, we can shorten this delay and remain reliable.
If you run the demo in benchmark mode twice, you'll notice that the TECO script is input nearly instantaneously the second time, whereas you can see the demo "type" the script in very quickly the first time. (Remove
make reconfigand run the demo again to see the difference.)
SING_STEPfeature for automatically attaching binary media images to the simulator from files on USB sticks now looks at all directories under
/media, not just
usb7so that it works with several other common Linux USB automounting schemes, such as the one Raspbian Desktop uses.
Fixed the order of initialization in the GPIO setup code for the James L-W serial mod case. Fix by Dylan McNamee.
The helper program that selects which boot script to run when the PiDP-8/I boots based on the IF switch settings broke at some point in the past, apparently because it was using its own idiosyncratic GPIO handling code, and thus did not track our evolving GPIO handling methods. Now it shares the same code used by
pidp8i-test, so it works properly again.
The SysV init script that starts
pidp8i-simunder GNU Screen on the PiDP-8/I now sets the working directory to
$prefix/share/mediaon start, so relative paths given to SIMH commands (e.g.
ATTACH) are more likely to do what you want. In prior releases, you generally had to give absolute paths to attach media and such because CWD would be set somewhere unhelpful.
The Fetch LED is no longer lit when in STOP or single-step mode. In real hardware, it can be either on or off in this mode, depending on various conditions, but it is most often off, so while it is not perfectly correct now, it is less wrong. Most of the investigation into this issue is by Bill Cattey, with the current partial fix by me. A more precise fix may come later. (See ticket [347ae45403] if you care to know the details.)
The Pause LED state was over-counted in the LED sub-sampling scheme so that it would tend to be brighter than it should have been. Problem noticed by Ian Schofield.
The MB row's state was not showing the right thing. The problem was noticed in comparison to real PDP-8/I hardware by Vincent Slyngstad and verified by William Cattey. Ian Schofield suggested the current fix.
Updated SIMH to upstream checkin ID 27f9fc3c3, December 11, 2017.
There have been no substantial changes to the PDP-8 simulator since the last update, 8 months ago, but there have been a lot of bug fixes to the SCP, that being the common core of all of the SIMH simulators.
One upstream change had to be backed out to work around a bug they introduced, which was not fixed by release time.
(See GitHub issue #508.)
Updated for Raspbian Stretch, released in September 2017. (Our binary OS images are built against the subsequent 2017-11-29 release, with updates as of 2017-12-22 applied.)
The only significant difference found is that the old, abandoned
usbmountpackage no longer works, apparently due to some change in
systemd. We've replaced that with a set of scripts based on those by Mike Blackwell.
It should still run on Raspbian Jessie, however.
The binary OS images produced with this version automatically expand the root partition to fill your SD card. (Prior versions required that you manually do this with
Assorted portability, build system, and documentation improvements.
Removed the PDP-8 CPU idle detection feature. Oscar Vermeulen reports that it also interfered with NLS LED driving mode in his last version, and we have no reason to believe our NLS code is sufficiently different to avoid the problem.
This does not affect ILS users, since enabling ILS mode disables this change.
NLS system upgrades wouldn't normally be affected because the changed files are not normally overwritten on installation. If you're affected, you know it, and how to fix it.
Replaced the earlier attempts at finding an ideal IPS rate for the simulator when run on single-core hosts with usage of SIMH's percentage style throttle rates. We now explicitly set the throttle rate to "50%" which not only achieves an even higher throttle rate than in earlier releases, it's reliable in the face of varying background CPU usage. See the single-core section of
Version 2017.04.01 — The "I May Be a Fool, but I am Your Fool" release
configure --alt-serial-modoption to change the GPIO code to work with James L-W's alternative serial mod.
Increased the stock CPU throttle from 0.67 MIPS to 0.85 MIPS on most Pi 1 class devices, except for the Pi Zero which is a bit faster and so is able to run at 1.25 MIPS.
(Testing shows that increasing it further causes SIMH to complain that it doesn't have enough CPU power to run that fast, despite the fact that top(1) shows it taking only about half the available CPU power. It's just as well: we don't want to hog all the CPU power on a single-core Pi, anyway.)
When built in NLS mode, most of the PDP-8 simulator configuration files we generate now have CPU idle detection enabled, allowing host CPU usage to drop when the simulated CPU is basically idle, such as when waiting for keyboard input.
Replaced a simplistic 2-second delay in the startup sequence of the simulator,
pidp8i-test, and "Incandescent Thought" with a smarter interlocked startup sequencing mechanism that largely eliminates the delay.
Fixed a problem introduced in v20170204 which causes the
DEPOSITswitch handlers to generate incorrect core addresses when the SIMH PDP-8 CPU core sets bits beyond the lower 12 in the PC register. We were assuming this register is always 12-bit clean, but it isn't.
Merged in upstream SIMH improvements. Changes relevant to the PiDP-8/I include:
- The PDP-8 CPU reset mechanism now does more of what our
STARTswitch handler did, so we now delegate to that upstream mechanism, reducing behavior differences between the cases.
- Improved keyboard polling behavior in terminal handler.
- Fixed many bugs identified by Coverity Scan in many different subsystems of the simulator. Normally I wouldn't note such a low-level change in this user-centric document, but it is possible that some of these improvements fix user-visible bugs.
- The PDP-8 CPU reset mechanism now does more of what our preexisting
SIMH's default PDP-8 configuration enables the DF32 disk device with the short name "DF", but since the SIMH
DEPOSITcommand works on both devices and registers, a command like
d df 0is ambiguous, causing the default configuration of SIMH to give a "Too few arguments" error for this command, even though it's obvious that you mean the CPU DF register here. (Surely you didn't mean to overwrite the first word of your disk image instead?) Since upstream refuses to fix it, I have disabled the DF32 device in all of the default
Since these scripts aren't overwritten on installation, this will only affect new installs unless you say
make mediainstall, in which case your binary OS media is also overwritten. Do this at your own risk!
Many improvements to the
DFUSB auto-mounting and SIMH attaching feature:
- Prior versions apparently could only mount paper tape images correctly. This release includes a fix that allows RX type floppy images and TU type DECtape images to autoattach.
- Changed the meaning of
DF=7from attaching an RL type removable hard disk cartridge image to the first RL01 drive to attaching an older and more period-correct RK type image to the second RK05 drive. The second half of the change lets you use this feature along with the stock OS/8 image to attach a second hard disk at runtime.
- The file name matching code used by this feature was only using
the first character of the file name extension, and it wasn't
pinning the match to the end of the file name string. Thus, if
DF=0, it would look for the first file with
.panywhere in the file name, not
.ptat the end, as expected.
- Improved error messages given when this feature doesn't work.
pidp8i-testprogram now accepts a
-vflag to make it give the version and configuration string and exit.
pidp8i-testnow exits more cleanly, shutting down ncurses, turning off the front panel LEDs, and writing a diagnostic message on signal exits.
The version number part of the configuration string written by
pidp8i-test -vand as the first line of output from the simulator now uses the same 10-digit Fossil checkin ID format as the Fossil timeline, making it easier to match that version to a particular Fossil checkin.
The Raspberry Pi model detection code behind the Pi model tag in the configuration string was only doing the right thing for the more common Pi models. It now reports the correct Pi model for the various flavors of Compute Module and the Pi Zero.
Improved error handling in the process that inserts the version info into the configuration string emitted when the simulator and test programs start up.
We now build and link in the upstream
sim_videomodule, which allows access to a video display via SDL. We do not currently use this in the project core, but I recall hearing about a third-party project that uses this for a local graphical X-Y vector display implementation for playing Spacewar! When built on a system without SDL or even a local bitmap display, this code becomes nearly a no-op, affecting build time very little.
SIMH changes to a different delay mechanism at CPU throttle rates below 1000 IPS, which prevents our incandescent lamp simulator from running correctly. Therefore, when you give a
./configure --throttleflag value that would use this throttle mode, we disable the ILS even when building on multi-core Raspberry Pis.
(This fix doesn't disable the ILS at run time if you manually set a throttle value under 1000 IPS via a SIMH command. We'll try to help you avoid accidentally shooting yourself in the foot, but if you're going to aim, you're on your own.)
Several internal refactorings to improve code style, reduce the upstream SIMH patch footprint, and fix corner case bugs.
Largely rewrote the incandescent lamp simulator (ILS) feature. The core of Ian Schofield's original contribution is still hiding in there if you go spelunking, but everything surrounding it is different.
The changes and design decisions surrounding this are complicated and subtle, but the end result is that the ILS is now free of judders, blips, shudders, and chugs. (Those are nuanced technical terms for "badness.") The ILS is now buttery smooth from 1 kIPS to the many-MIPS rate you get running un-throttled on a Pi 3.
Although most of the ILS work does not directly apply to the "no lamp simulator" (NLS) case, the sample rate dithering reduces display update artifacts seen in this case as well.
Slowed the ILS brightness rates down a bit: more lampy, less snappy. Whether this is accurate or not is something we'll have to determine through research separately in progress.
The ILS display is a bit brighter: the delay values used in prior versions put a cap on max brightness that was well under the full LED brightness achievable.
For the first time, it is possible to build Deeper Thought (any version) against the ILS, with minor adjustments. Prior versions of the ILS had too different an external interface to allow this. Full details are in a new wiki article.
In normal free-running mode, the simulator lights the Fetch and Execute LEDs at 50%, whereas before there was an imbalance that purely had to do with the much lower complexity of fetching an instruction inside the simulator vs executing it.
(In real hardware, the complexities were different: fetch involved a core memory retrieval, very much non-instantaneous, whereas the execution of the fetched instruction kind of happened all at once in complicated electron flows, rather than the sequential C code of the SIMH PDP-8 simulator. Thus, it was reasonable for DEC to talk about PDP-8/I fetch-and-execute cycles as if the two steps were of equal time complexity.)
I haven't compared the resulting LED appearance to a real PDP-8/I.
Several other tweaks to LED state handling to better match real hardware.
pidp8i-testprogram to allow manual stepping forwards and backwards in addition to the previous auto-advancing behavior.
As soon as you press one of the arrow keys, the test program moves to the next or previous action in the sequence and stops auto-advancing. This mode is useful when testing the hardware with a multimeter or similar, and you need a certain row or column to stay lit up indefinitely.
You can also press R to resume auto-advancing behavior, or either Ctrl-C or X to exit the program gracefully.
This requires that you have
libncurses-devinstalled on your Pi.
The SIMH PDP-8 simulator's internal SR register now matches the hardware switches when you say Ctrl-E then
ex sr. Prior versions only loaded the hardware switch register values into the internal register when it executed an
Copied the KiCad design files into the source tree since they are now formally released by Oscar Vermeulen under a Creative Commons license. Also included the PDF version of the schematic produced by Tony Hill. (This is all in the
Lowered the default simulator throttle value for single-core Pi boards from 1332 kIPS to 666 kIPS after doing some testing with the current code on a Raspberry Pi 1 Model B+. This value was chosen since it is approximately twice the speed of a PDP-8/I. This leaves a fair bit of CPU left over for background tasks, including interactive shell use.
This value may be a bit low for Pi Zero users, but it is easily adjustable.
Merged in the relevant SIMH updates. This is all internal stuff that doesn't affect current PiDP-8/I behavior.
Many build system and documentation improvements.
When any program that talks to the PiDP-8/I front panel starts up, it now prints out a banner modeled on the Erlang configuration line. For example, when I run the software in the development tree on my PiDP-8/I, I get the following:
PiDP-8/I trunk:i49cd065c [pi3b] [ils] [serpcb] [gpio]
It tells me that:
- I'm running code built from Fossil checkin ID 49cd065c on the
trunk branch, as opposed to a release version, which would be
release:v20170123or similar. (The
ihere is a tag standing for "ID", as in Fossil checkin ID. Contrast
vused to tag release version numbers.)
- I'm running it on a Raspberry Pi 3 Model B with Ian Schofield's incandescent lamp simulator (ILS) feature enabled.
- The software is built to expect that the PiDP-8/I PCB and the Pi board attached to it have had the serial mods made to them.
- The GPIO module found the GPIO hardware and was able to attach to it.
- I'm running code built from Fossil checkin ID 49cd065c on the trunk branch, as opposed to a release version, which would be marked
I get a very different result when running it on my desktop machine:
PiDP-8/I trunk:id8536d91 [cake] [nls] [nopcb] [rt]
This tells me:
- I'm running a different version of the development branch (i.e. the "trunk") of the code than what's running on the Pi.
- It's not running on a Pi at all. (Cake ≠ pi.)
- I've disabled the ILS feature, so it's running with the "no lamp simulator" (NLS) GPIO module.
- Which is all to the good, because there's no point burning CPU power running the ILS code on a host machine that doesn't have a PiDP-8/I PCB attached.
- The GPIO thread is running with real-time privileges.
The ILS feature can now be disabled at
configuretime via the new
--no-lamp-simulatorflag. This option is automatically set when building on a single-core Raspberry Pi. (The flag is there only to allow someone building the software on a multi-core host machine to disable the ILS.)
Tweaked the ILS decay constants to be asymmetric, better mimicking the way real incandescent lamps work: they heat up to full brightness faster than they fade to perceptively "off."
The LED values used by the GPIO thread were being recalculated way too often.
In the ILS case, it was updating the values approximately at the same rate as the ILS's PWM core frequency, roughly 7,500 times per second, which is far higher than the human persistence of vision limit. While the PWM rate does need to be that fast to do its job, the underlying LED state values do not need to change nearly that often to fool the human into seeing instantaneous updates.
The NLS case was actually worse, recalculating the LED values on every instruction executed by the PDP-8 CPU simulator, which even on a Pi 1 is likely to be a few MHz.
In both the ILS and NLS cases, we now update the LED values about 100 times a second, maintaining that rate dynamically based on the current execution speed of the simulator.
In prior versions, the ILS was only updating at its intended rate when the PDP-8 simulator was running flat-out on a current multi-core Raspberry Pi. If you throttled the SIMH simulator to a slower execution rate, the display quality would start to degrade noticeably below about 1 MIPS.
With the prior fix, we now ship 5.script (i.e. the handler for starting the simulator with IF=5, or restarting it with IF=5 +
SING_STEP) set to a throttle value of 30 kIPS, which allows the human to see each AC/MQ modification. The built-in delay loops are still there, else we'd have to drop this to well under 1 kIPS.
SING_INSTswitch now immediately puts the processor into single instruction mode, not requiring a separate press of the
STOPkey, as in prior versions. This is the correct behavior according to the 1967-1968 edition of DEC's Small Computer Handbook for the PDP-8/I.
Greatly simplified the way single-instruction mode, processor stoppage, and the
CONTkey work. The prior implementation was error-prone and difficult to understand. This fixes a particularly bad interaction between the way
CONTkey presses were handled, causing the processor to not resume correctly from
Consolidated and cleaned up the bulk of the PiDP-8/I switch handling code so that it is not so intimately tied into the guts of the PDP-8 CPU emulator. This will greatly increase the chance that future updates to the upstream SIMH code will apply cleanly to our version.
Fixed a bug in
examples/bit-rotate.palwhich caused it to skip the actual bit rotation step. We were trying to microcode two instructions into one that the PDP-8 won't accept together, and we didn't catch it until now because the HLT bug masked it, and the
palbartassembler we ship isn't smart enough to notice the bug.
Fully generalized the mechanism for generating
examples/*.pal. You can now drop any PAL assembly language program into the
examplesdirectory and type
maketo build these various output forms automatically using the shipped version of
palbart. This effectively turns this PiDP-8/I software distribution into a PDP-8 assembly language development environment: rapidly build, test, and debug your PAL programs on your PC before you deploy them to real hardware. Or, write PAL programs to debug the hardware or simulator, as we did with
Fixed a sorting bug in the tool that generates
obj/*.lst, resulting in
depinstructions that weren't sorted by core address. This didn't cause any real problem, but it made tracing the execution of a PAL assembly program difficult if you were trying to refer to the
boot/*.scriptfile to check that the PiDP-8/I's front panel is showing the correct register values.
Updated SIMH to the latest upstream version and shipping a subset of the SIMH docs as unversioned files from tangentsoft.com.
configurescript now aborts the build if it sees that you're trying to build the software as root, since that means it generates the init script and the pidp8i script expecting to run the installed software as root, not as your normal user. The most common way this happens is that you have a configured source tree, then change one of the
*.infiles and say
sudo make install, thinking to build and install the change in one step. This fixes that.
Several improvements to the build system.
Prior releases did not include proper licensing for many of the included files. This project was, therefore, not a proper Open Source Software project. This problem has been fixed.
In this release, many files that were previously technically only under standard copyright due to having no grant of license now have an explicit license, typically the same as SIMH itself. (Thank you to all of the authors who allowed me to apply this license to their contributions!)
For several other files, I was able to trace down some prior license and include its text here for the first time.
There remain a few "gray" items: the TSS/8 and ETOS disk images. See the
COPYING.mdfile for more on the current status of these OS images. If the legal status of these files clarifies in the future, this software distribution will react accordingly, even if that means removing these files from the distribution if we learn that these files are not freely-redistributable, as we currently believe them to be today.
The Step Counter LEDs on the front panel weren't being lit properly when EAE instructions were being used. Thanks for this patch go to Henk Gooijen and Paul R. Bernard.
boot/5.scriptfiles are no longer simply opaque lists of octal addresses and machine code. They are generated from PAL assembly files provided in the
examplesdirectory, so that you can now modify the assembly code and type
maketo rebuild these boot scripts.
The mechanism behind the prior item is fully general-purpose, not something that only works with
examples/*.palfile found at
maketime is transformed into a SIMH boot script named after the PAL file and placed in the
bootdirectory. This gives you an easier way to run PDP-8 assembly code inside the simulator. After saying
*.scriptfiles, you can run the program with
bin/pidp8i-sim boot/my-program.scriptto poke your program's octal values into core and run it. This round-trip edit-and-run process is far faster than any of the options given in the examples'
Disassembled both versions of the RIM loader to commented, labeled PAL assembly language files. If you ever wanted to know what those 16 mysterious instructions printed on the front panel of your PiDP-8/I did, you can now read my pidgin interpretation of these programs in
examples/*-rim.loader.paland become just as confused as I am now. :)
The two RIM loader implementations now start with a nonstandard
HLTinstruction so that when you fire up the simulator with IF=1 to start the high-speed RIM loader, it automatically halts for you, so you don't have to remember to STOP the processor manually.
There is currently a bug in the way the simulator handles
HLTinstructions which prevents you from simply pressing START or CONT to enter the RIM loader after you've attached your paper tape, so you still have to manually toggle in the 7756 starting address and press START to load the tape into core. (I hope to fix this before the next release, but no promises.)
configure --throttlefeature for making the simulator run at a different speed than it normally does. See
The build system now reacts differently when building the PiDP-8/I software on a single-core Raspberry Pi:
- If you're building the trunk or release branch, you'll get a configure error because it knows you can't run the current implementation of the incandescent lamp simulator on a single-core Pi. (Not enough spare CPU power, even with heavy amounts of throttling.)
- If you're building the no-lamp-simulator branch, it inserts a
throttle value into the generated
boot/*.scriptfiles that do not already contain a throttle value so that the simulator doesn't hog 100% of the lone core, leaving some spare cycles for background tasks. The above
--throttlefeature overrides this.
These features effectively replace the manual instructions in the old
README-single-core.mdfile, which is no longer included with this software distribution, starting with this release.
Lowered the real-time priority of the GPIO thread from 98 to 4. This should not result in a user-visible change in behavior, but it is called out here in case it does. (In our testing, such high values simply aren't necessary to get the necessary performance, even on the trunk branch with the incandescent lamp simulator.)
Since v20161128, when you
make installon a system with an existing PiDP-8/I software installation, the binary OS media images were not being overwritten, on purpose, since you may have modified them locally, so the installer chose not to overwrite your versions.
With this release, the same principle applies to the SIMH boot scripts (e.g.
$prefix/share/boot/0.script) since those are also things the user might want to modify.
This release and prior ones do have important changes to some of these files, so if you do not wish to overwrite your local changes with a
make mediainstallcommand, you might want to diff the two versions and decide which changes to copy over or merge into your local files.
Automated the process for merging in new SIMH updates. From within the PiDP-8/I software build directory, simply say
make simh-updateand it will do its best to merge in the latest upstream changes.
This process is more for the PiDP-8/I software maintainers than for the end users of that software, but if you wish to update your SIMH software without waiting for a new release of this software, you now have a nice automated system for doing that.
Updated SIMH using that new process. The changes relevant to the PiDP-8/I since the prior update in release v20161226 are:
- Many more improvements to the simulator's internal timer system. This should make deliberate underclocking more accurate.
- It is now possible to get hex debug logs for the simulator console port by cranking up the simulator's debug level.
The simulator now reports the upstream Git commit ID it is based on in its version string, so that if you report bugs upstream to the SIMH project, you can give them a version number that will be meaningful to them. (They don't care about our vYYYYMMDD release numbers or our Fossil checkin IDs.)
Version 2016.12.26 — The Boxing Day release
Tony Hill updated SIMH to the latest upstream version.
This change represents about 15 months worth of work in the upstream project — plus a fair bit of work by Tony to merge it all — so I will only summarize the improvements affecting the PDP-8 simulator here:
Many improvements to the internal handling of timers.
The most user-visible improvement is that you can now clock your emulated PDP-8 down to well below the performance of a real PDP-8 via
SET THROTTLE, which can be useful for making blinkenlights demos run at human speeds without adding huge delay loops to the PDP-8 code implementing that demo.
Increased the number of supported terminals from four to either twelve or sixteen, depending on how you look at it. Eight of the additional supported terminal devices are conflict-free, while the final four variously conflict with one or more of the other features of the simulated PDP-8. If you want to use all 16, you will be unable to use the FPP, CT, MT and TSC features of the system.
This limitation reflects the way the PDP-8 worked. It is not an arbitrary limitation of SIMH.
Added support for the LS8E printer interface option used by the WPS8 word processing system.
The simulator's command console now shows the FPP register descriptions when using it as a PDP-8 debugger.
SHOW TTIX/TTOX DEVNOSIMH command to display the device numbers used for TTIX and TTOX.
SHOW TTIX SUMMARYSIMH command is now case-insensitive.
Upstream improvements to host OS/compiler compatibility. This increases the chances that this software will build out of the box on random non-Raspbian systems such as your development laptop running some uncommon operating system.
make install, we now disable Deeper Thought 2 and the legacy
pidp8service if we find them, since they conflict with our
Added the install user to the
gpiogroup if you
make installif that group is present at install time. This is useful when building and installing the software on an existing Raspbian SD card while logged in as a user other than
The entire software stack now runs without explicit root privileges. It now runs under the user and group of the one who built the software.
For the few instances where it does need elevated privileges, a limited-scope set of sudo rules are installed that permit the simulator to run the necessary helper programs.
The power down and reboot front panel switch combinations are no longer sensitive to the order you flip the switches.
Changed the powerdown front panel switch combination to the more mnemonically sensible
Its prior switch combo —
Start— is now the reboot sequence, with the mnemomic "restart."
Removed the USB stick mount/unmount front panel switch combos. The automount feature precludes a need for a manual mount command, and unmount isn't necessary for paper tape images on FAT sticks.
The simulator now runs correctly on systems where the GPIO setup process fails. (Basically, anything that isn't a Raspberry Pi.) Prior to this, this failure was just blithely ignored, causing subsequent code to behave as though all switches were being pressed at the same time, causing utter havoc.
The practical benefit of this is that you can now work with the software on your non-Pi desktop machine, losing out only on the front panel LEDs and switches. Everything else works just as on the Pi. You no longer need a separate vanilla SimH setup.
Added a locking mechanism that prevents
pidp8i-simfrom fighting over the front panel LEDs. While one of the two is running, the other refuses to run.
examples/ac-mq-blinker.pal, the PAL8 assembly code for the
Fixed two unrelated problems with OS/8's FORTRAN IV implementation which prevented it from a) building new software; and b) running already-built binaries. Thanks go to Rick Murphy for providing the working OS/8 images from which the files needed to fix these two problems were extracted.
Added the VT100-patched
VTEDITTECO macro from Rick Murphy's OS/8 images, and made it automatically run when you run TECO from the OS/8 disk pack. Also added documentation for it in
VTEDIT.DCon the disk pack as well as in the wiki.
The default user name on the binary OS images is now
pi, its password has changed to
edsonDeCastro1968, and it demands a password change on first login. I realize it's a hassle, but I decided I didn't want to contribute to the plague of open-to-the-world IoT boxes.
Many build system and documentation improvements.
pidp8i-testprogram's LED test routines did not work correctly when built against the incandescent lamp simulator version of the GPIO module. Reworked the build so that this test program builds against the no-lamp-simulator version instead so that you don't have to choose between having the lamp simulator or having a working
More improvements to
PRINTSroutine from that example as
This release marks the first binary SD card image released under my maintainership of the software. As such, the major user-visible features in this release of the Fossil tree simply support that:
pidp8i-initscript now understands that the OS's SSH host keys may be missing, and re-generates them. Without this security measure, anyone who downloads that binary OS image could impersonate the SSH server on your PiDP-8/I.
- Added a
RELEASE-PROCESS.mddocument. This is primarily for my own benefit, to ensure that I don't miss a step, particularly given the complexity of producing the binary OS image. However, you may care to look into it to see what goes on over here on the other side of the Internet. :)
Added an OS/8 BASIC solution to Project Euler Problem #1, so you can see how much simpler it is compared to the PAL8 assembly language version added in the prior release.
Updated the PAL8 assembly version with several clever optimizations by Rick Murphy, the primary effect of which is that it now fits into a single page of PDP-8 core memory.
Debounced the switches. See the mailing list post announcing this fix for details.
Added a solution to Project Euler Problem #1 in PAL8 assembly language and wrote the saga of my battle with this problem into the wiki. This also adds a couple of useful PAL8 routines in
Integrated David Gesswein's latest
palbartprogram (v2.13) into the source distribution so that we don't have to:
- ship pre-built RIM format paper tapes for the examples; and
- put up with the old versions that OS package repos tend to have (Ubuntu is still shipping v2.4, from 6 years ago!)
Fixed a bug in the
make installscript that caused it to skip installing
usbmountfrom the OS's package repo when they are found to be missing.
Fixed a related bug that prevented it from disabling the serial console if you configure the software without
--serial-modand then install it, causing the serial console and the GPIO code in the PiDP-8/I simulator to fight over GPIO pins 14 and 15.
Removed the last of the duplicate binary media entries. This makes the zip files for this version well under half the size of those for the 2015.12.15 upstream release despite having more features.
Added an intelligent, powerful build system, replacing the bare-bones
Makefilebased build system in the upstream version. See
README.mdfor more info on this.
The installation is now completely relocatable via
./configure --prefix=/some/other/place. The upstream version would not work if installed somewhere other than
/opt/pidp8due to many hard-coded absolute paths. (This is enabled by the new build system, but fixing it was not simply a matter of swapping out the build system.)
Changed all of the various "PDP," "PDP-8", and "PiDP-8" strings to variants on "PiDP-8/I", partly for consistency and partly because it seems unlikely that this software will ever be used with anything other than the PiDP-8/I project.
Part of this renaming means that the default installation location is now
/opt/pidp8i, which has the nice side benefit that installing this version of the software will not overwrite an existing installation of the upstream version in
Another user-visible aspect of this change is that the upstream version's
pdp.shscript to [re]enter the simulator is now called
The scripts that control the startup sequence of the PiDP-8/I simulator now include helpful header comments and emit useful status messages to the console. Stare no more at opaque lists of SimH commands, wondering what each script does!
scanswitchinto the top-level
srcdirectory, since the only thing keeping it in a separate directory was the redundant
gpio.hfile. There were minor differences between the two
gpio.hfiles, but their differences do not matter.
Installing multiple times no longer overwrites the binary OS/program media, since the disk images in particular may contain local changes. If you want your media images overwritten, you can insist on it via
The installation tree follows the Linux Filesystem Hierarchy Standard, so that files are in locations an experienced Linux user would expect to find them. The biggest changes are that the content of the upstream
bootscriptstree is now installed into
$prefix/share/boot, and the OS/program media images which used to be in
imagefilesare now in
Added a bunch of ancillary material: wiki articles, USB stick label artwork, a PAL8 assembly example program for you to toggle in, etc. Also filed a bunch of tickets detailing feature proposals, known bugs and weaknesses, etc. If you were looking for ways to contribute to the development effort, these new resources provide a bunch of ideas.
Made some efforts toward portability.
While this project will always center around Raspbian and the PiDP-8/I add-on board, the intent is that you should be able to unpack the project on any other Unix type system and at least get the simulator up and running with the normal SimH manual control over execution instead of the nice front panel controls provided by the PiDP-8/I board.
In particular, the software now builds under Mac OS X, though it does not yet run properly. (The modified SimH PDP-8 simulator currently misbehaves if the PiDP-8/I panel is not present. Fixing this is on the radar.)
Fixed a bunch of bugs!
The last release of the software made by Oscar Vermeulen, the creator of the PiDP-8/I project. This version of the software derives from it, but differs in many ways. (See above.)
Since May of 2017, this version of the software is now considered the current stable version.