21–30 of 43 posts
that’s also how they managed to fit all the software of the Apollo guidance computer into it.
> “The AGC had a sophisticated software interpreter, developed by the MIT Instrumentation Laboratory, that implemented a virtual machine with more complex and capable pseudo-instructions than the native AGC.” (https://en.wikipedia.org/wiki/Apollo_Guidance_Computer#softw...)
Does anyone know of any good books which describe the design and implementation of small VMs/interpreters like these?
I have encountered brief discussions of the technique in a few places though. Jon Bentley wrote a column called Programming Pearls, collected in a couple of compilation books, and one of those was called Squeezing Space which talked about this a bit. I think that Peter Norvig had a chapter on it in Paradigms of Artificial Intelligence Programming but couldn't swear to it. I've also read that KDB+ uses this technique, and I seem to remember that one of the few columns that discussed implementation (it's proprietary technology, so there's not much on it) talked about stuff like this.
The general idea was also briefly discussed in Dynamic Languages Wizards Series - Panel on Runtime (from 2001), though they didn't go into methods, just trade-offs [1]
An instance of compiling the program to a bespoke bytecode and then interpreting the bytecode. This is one of my favourite programming tricks - high effort and high effectiveness, rarely used. Dwarf does this as one of the size optimisation tricks. I used to know another example but it isn't coming to mind.
As opposed to containing implementations of various interpreters, VMs etc, as described here (and functions that 100s of other programs implemented as well).
Or even better: a selection of languages, letting programmers pick or even mix & match for various parts of the program.
Even today "big ROM, small RAM" is common (eg. in uC's). With flash offering extra flexibility for the 'ROM' part.
But the limitations of the day made many (system) ROMs little more than a glorified boatloader + some low level I/O code. If you were lucky, with an easy to learn but slow language like (interpreted!) BASIC.
Does anyone know of any good books which describe the design and implementation of small VMs/interpreters like these?
An instance of compiling the program to a bespoke bytecode and then interpreting the bytecode. This is one of my favourite programming tricks - high effort and high effectiveness, rarely used. Dwarf does this as one of the size optimisation tricks. I used to know another example but it isn't coming to mind.
Too bad ROM was expensive those days. If a language with a nice balance between expressiveness and performance (say Lisp, Forth, Pascal, maybe Lua or even Java) had been embedded in the machine's ROM, then many programs could contain just their core functionality. While having all the machine's capabilities easily accessible (and maybe some standard library of often-used functions). As opposed to containing implement…
I've always been fascinated by Zork despite coming of age decades after its release. Reminded of it a few years back through The Digital Antiquarian, I wrote a C# implementation of the Z-Machine ( https://github.com/raegwolf/zmachine ). Seeing how so much could be achieved with so little memory was eye opening.
the answer is to encode your program using a higher-level machine code targeting a portable virtual machine. that’s also how they managed to fit all the software of the Apollo guidance computer into it. > “The AGC had a sophisticated software interpreter, developed by the MIT Instrumentation Laboratory, that implemented a virtual machine with more complex and capable pseudo-instructions than the native AGC.” ( https:…