AVX Advanced Vector Extensions
Intel's SIMD instruction set with 256-bit registers. Introduced in Sandy Bridge 2011. AVX2 (Haswell, 2013) added integer ops. AVX-512 (Skylake-X, 2017) went to 512 bits.
Widely used in high-performance compute, data-format libraries (SIMD JSON parsing), ML code, codec encode/decode. AVX2 is the safe default — every modern CPU has it. AVX-512 is politically messy: Intel pulled it from consumer CPUs in 2020, AMD added it in Zen 4 (2022). Power: AVX-512 use historically lowered clock frequency (less so with Sapphire Rapids 2023). Auto-vectorization in modern compilers (LLVM, GCC) generates AVX when possible.