Compression Bench
C++, SIMD, multithreading, CMake, benchmarking
Compression Bench is a systems project focused on implementation quality and measurement discipline. It benchmarks multiple lossless compression algorithms under a consistent harness and publishes results in terminal, CSV, and HTML formats.
What it includes
- Implementations of Huffman, LZ77, DEFLATE, RLE, and LZW from scratch
- Streaming APIs and a self-describing container format with CRC32 verification
- Multi-threaded DEFLATE with AVX2-accelerated byte matching
- Benchmark routines with warmup iterations and median-of-N reporting across a fixed corpus
Engineering focus
The project emphasizes repeatable performance measurement, explicit correctness checks, and enough tooling around the algorithms to make comparisons meaningful instead of anecdotal.