You mean developers, right? Compiler auto-vectorization hasn't really advanced that much. You still need to manually use SIMD intrinsics (or inline asm) to get any kind of worthwhile performance increase.