Did you know tokenizing and parsing don’t need to be so serial? Why iterate over 1 byte at a time when you could iterate over 64 at a time? Why write a loop when a single instruction would do?
In this talk I’ll explain the most important operations for exploiting data-level parallelism in tokenizing and parsing Zig code. Come ready to learn about AVX-512 SIMD, bit-manipulation, and branchless programming for a compiler front-end!