This is a compiler that lowers a C-like language into x86 assembly, built (almost) entirely from scratch in Rust. I previously implemented this in C++ for the Compiler Construction course at Northwestern University, but I rewrote it in Rust just for fun.
The IR is split into several layers which get progressively lowered by their respective compilers.
Bridge between the higher-level IRs and x86-64 assembly (AT&T syntax). The L1 compiler does a nearly a 1-1 mapping to ASM.
Introduces variables. The L2 compiler performs register allocation using the Iterated Register Coalescing algorithm (George and Appel, 1996).
Simpler, C-style syntax that eliminates registers and calling convention. The L3 compiler implements instruction selection with the tiling method on trees, using dynamic programming to cover each tree optimally.
Explicit control flow to enable easy analysis and transformation. The IR compiler constructs SSA (Cytron et al., 1991) and optimizes the code.
Special thanks to Prof. Campanoni for inspiring my interest in compilers.