Debian Code Search: Fast TurboPFor with Go SIMD
Summary
This technical blog post describes implementing a fast TurboPFor integer compressor/decoder in Debian Code Search using Go SIMD, eliminating C dependencies, and achieving near-C performance through SIMD and bit-width specialization. It covers API design, setup, benchmarking, and optimizations including PGO, memory allocation reductions, and positional popcount, with a focus on practical Go performance engineering.