Explaining my fast 6502 code generator

To learn how optimizing compilers are made, I built one targeting the 6502 architecture. In a bizarre twist, my compiler generates faster code than GCC, LLVM, and every other compiler I compared it to.

I reckon my compiler isn't doing more when it comes to high-level optimizations, so the gains must be from the code generation side. This makes sense, as most compilers are multi-target, with backends designed for modern RISC-like systems, not the ancient 6502. It doesn't matter how good GCC or LLVM's high-level optimizations are if they falter at the last leg of the race.

Still, my compiler also beats those designed for retro and embedded systems, like VBCC, SDCC, and KickC. For this reason, it seemed like a good idea to write about my technique.

on Hacker News. Here's the article, and here's the HN discussion.