Reflections on writing a Game Boy emulator

Thoughts #

I’ve recently finished¹ my Game Boy emulator, based on a C++ emulator I started a couple of years ago. I felt compelled to rewrite it in Rust, partly as a way to learn the language, but also to finish what I started (my previous emulator was incomplete and broken). This post won’t be technical (there are plenty of resources for that²), but more reflective.

The project took me a bit over 2 years, and in retrospect, I would have taken a less perfectionist way of writing my emulator. Instead of spending a large amount of time to pass Blargg’s CPU tests, it would have been better to try to get Tetris running and displaying on the screen. To illustrate this, after writing a disassembler, I could have checked what opcodes the game used, then developed a CPU emulator that only emulated those opcodes. Then, I would have written a very inaccurate graphics renderer that looped over the Game Boy display and displayed the background tile data. I could mock the joypad inputs to alternate start to get into gameplay. From here, I could rewrite my graphics renderer to be scanline-based and then go back to implement window/object rendering and proper joypad support. I could move on to other games from here to further improve the emulator. This sounds really hacky and bad, but I also would have finished this project in 6 months instead of 2 years (obviously working on the project on-and-off).

I don’t regret my time spent, since it is the largest personal project I’ve worked on and also pushed my knowledge of emulators much more than the Chip-8 did. Still, I likely would have tackled similarly interesting problems contributing to pre-existing emulators or trying to take a crack at a non-emulated system. You don’t need to work your way up to newer systems.

On an unrelated note, when I was first developing the dot-matrix-c++, I tried not to reference other emulators to avoid “cheating”, but you learn a lot referencing other code, especially since I switched languages. Rust didn’t change my emulator architecture much; I used Rc<RefCell> for many components to match how I wrote dot-matrix in C++, though there is a performance penalty, which is why you’ll see most other emulators take different approaches.

Closing #

Despite the negative tone of this post, I did enjoy developing the emulator, and I was super excited finally getting my joypad working on it to be able to interact with these games; It’s impressive how complex even this piece of 1989 hardware can be.