Trying to Make a Loop Auto-Vectorize
Summary
A technical blog post by jsgroth exploring auto-vectorization in Rust. It compares scalar dot-product implementations with SSE/AVX code, demonstrates how compilers emit 128-bit and 256-bit vectors, and discusses the floating-point ordering caveats of IEEE 754 and the impact of compiler flags and algebraic math features. The piece also experiments with manual chunking and intrinsics versus auto-vectorization to measure performance.