180-Line Python Compiler Shows SIMD Vectorization Lowering Step

A 180-line Python compiler demonstrates how scalar loop kernels get lowered into explicit data-parallel IR, making SIMD/SIMT vectorization visible as an intermediate step. The toy compiler takes hand-written ASTs and emits vectorized IR with lane IDs and active masks, illustrating the core transformation that production compilers like LLVM or Triton perform internally.