A friend shared Sylvain Kerkour's post SIMD programming in pure Rust which covers AVX-512 on AMD Zen 5. That got me curious about ARM's side of the story—specifically how std::simd compares to hand-written NEON intrinsics. My main development machine is a MacBook Pro with Apple M4, so I ran the benchmarks there.

I ran 9 benchmarks comparing three...