We've created the first vectorized Quicksort
3 hours ago
- Open source code sorts arrays ~10x faster than C++ std::sort and outperforms architecture-specific algorithms.
- Leverages SIMD/vector instructions and compress-store (or permute emulation) for efficient partitioning.
- Implemented using Highway's portable SIMD functions, supporting 16-128 bit inputs across six instruction sets.
- Achieves record speeds: 1123 MB/s on Skylake AVX-512, 798 MB/s on AVX2, 499 MB/s on Apple M1.
- Available on GitHub under Apache2 license; paper provides detailed explanation.