moonshot
Source
Source
// Definition of a program that is invoked by
// typing VERB 0 and NOUN 0 on the DSKY
prog {
.verb = 00;
.noun = 00;
.entry = main;
}
// The entry point of a program. State is kept
// and restored between invocations of the program
// (i.e. between DSKY commands or system reboots)
state main () [
a: WORD = 0;
] {
a = add(a: a, b: 1);
}
// Basic subroutines
sub add(a: WORD, b: WORD) {
return a + b;
}
You know the Apollo Guidance Computer (AGC)? It’s a flight computer that was used onboard both the Command Module (the bit that orbited the moon) and Lunar Module (the bit that landed on the moon) during the Apollo missions. This was a really impressive piece of technology for the time, and its architecture is crazy. Seriously, read the Wikipedia article for this thing.
Recently, the Luminary source code was released into the public domain, and my complete inability to read the assembly language made me wonder if there was a way to write programs for the AGC in a higher-level language – could I write software to perform a simulated moon landing in my own language compiled for the AGC? So, down the rabbit hole I went.
The Virtual AGC is an amazing project that provides a complete simulation environment for the AGC, including the DSKY (the goofy display and keyboard that the astronauts used to interact with the computer) as well as an assembler for the AGC assembly language. So all the tools I needed to run my programs were already available. I just needed to write a compiler that would take my code and translate it into the AGC assembly language.
The compiler has gone through a few iterations. The first version was way overengineered and I scrapped it pretty quickly. The current design is much less complex. It supports a single 15-bit integer type (yes, really), arithmetic operations, and conditional statements. The syntax is tightly bound to a task-based runtime model that would provide scheduled code execution, similar to the way the actual AGC software worked (but this runtime is not implemented yet). Additionally, the goal is to provide a simple standard library that provides a few basic functions for interacting with the DSKY and other components of the AGC.
As I mentioned, this whole thing is only partially implemented. I do have integration tests stood up that can actually compile and run simple programs on the Virtual AGC, which I expect to rely on heavily as I continue to develop the compiler. I come back to this project every few months because of how fascinating the AGC is, and I hope to eventually work my way towards a complete compiler that can run more complex programs.