6/29/2026 at 6:44:40 PM
I've always really enjoyed Andrew Kelley's article about trying to statically recompile NES code from 2013 [1]. Basically he makes a ton of progress but gets hung up not just on the realities of the handwritten assembler of the era just not being all that great at mapping to higher level LLVM IR. In the conclusion he specifically calls out a JIT-type methodology as probably being the way to go, where you live-recompile the hot paths when you have the runtime data required to actually understand them, and don't worry about the parts you can't.Very cool to see something like that in action.
by mikepurvis