(NOTE: This is a work in progress.)
The goal of this project is to build an emulator for the hardware of a PDP-10 CPU called the KL10 model B. This was the ~1983 pinnacle of the PDP10 universe in that it could support up to 16MWords (36-bit) of memory, allowed huge I/O DMA capability for disk and tape drives, could host a DECnet or a TCP/IP networking interface, and supported up to (I think) about 256 TTY RS-232C serial terminals. These machines had a PDP-11/40 as their front-end processor for loading microcode, doing diagnostics, and interfacing the console and command and control for the system. Unlike most of the mainframes of their day, they had virtual no console switches or lights. All of the management of the system was done through the PDP-11/40's interface into the innards of the system.
This emulator runs (with no changes) the final TOPS-20 capable microcode release from DEC.
For now, the macro assembly listing file for the microcode is read by
a quick JavaScript utility called gencode.js to build source files
needed by the emulator. You can run this by
node gencode
This generates:
cram.jswhich is the CRAM (microcode) RAM contents.cram-lines.jsfor source level microcode debugging and single stepping.dram.jswhich is the DRAM (dispatch) RAM contents.fields-model.jswhich I don't currently use.
You can run the emulator by
node em
This starts the microcode debugger. This debugger will eventually handle two levels of debugging: the microcode level and the PDP-10 instruction level. For now you can start and stop the PDP-10 instruction level only. The rest of the features (some of which aren't yet implemented or are simply hacks to get past some bug I was working on) are for controlling microcode emulation.
I use yarn for package management. Some people use npm. I like
mine. You can use yarn run to select from the scripts to run. Or
just specify yarn run build and then yarn run run to run the
emulator.
The project has been going for a few months of spare time so far. I have most of the EBOX (execution box) emulated to at least the first order of detail now, and the microcode correctly fetches an instruction from the MBOX (memory) and begins to execute it.
Don't expect this to be stable or tagged for releases until I get a lot further down the path. For now, since it isn't stable anyway, I checkpoint my source tree to GitHub with impunity and little care. It could be in any state when I do that.
I have code from an earlier project that will eventually supply the emulation for disk and tape drives, and I'll write an Ethernet based emulator for the TCP/IP interface when the time comes as well.
I expect this emulation to run at only a fraction of the speed of the real DECSYSTEM-20 because I'm emulating at such a low level and I'm not being very concerned with performance at this point.
When this is working more completely, I want to take the understanding I have gained from creating this emulator and build one in C or C++ that is meant to go fast. I didn't do that first instead because it would have slowed me down considerably. The whole C/C++ edit, compile, link, debug loop is much slower than my JavaScript debugging process, and the JavaScript language is extremely expressive. It's just faster to get concepts from brain to working code.
Many many thanks to Al Kossow and his comrades at the Computer History Museum for having read magtapes, imaged documents, and generally supported the DEC historical community for projects like this one. Much of the information that I have used in various ways during this project came from his archive at
http://www.bitsavers.org/pdf/dec/pdp10/
of which there are fortunately many mirrors.