though it would also be nice if folks who are using arch-specific simd intrinsics right now, instead used portable vector extensions supported in gcc/clang