Weaviate 1.39 extends Rotational Quantization with 4-bit support, achieving 45% heap reduction with less than 1% recall loss compared to TurboQuant. The release improves encoding performance through SIMD-accelerated Fast Walsh-Hadamard Transforms, optimizes distance kernels with SIMD operations, and enhances memory access patterns via CPU prefetching and vector centering techniques.
Researchers have developed a vectorized Quicksort implementation that sorts arrays roughly ten times faster than C++ std::sort by leveraging SIMD instructions across multiple CPU architectures. The approach uses compress-store and permute instructions for efficient partitioning, achieving portable performance across AVX2, AVX-512, and Arm NEON while supporting 16-128 bit inputs.