Function copy_ns_per_element_of 

pub const fn copy_ns_per_element_of(class: HostClass) -> f64

Serial throughput of a copy in nanoseconds per element, of a host class.

• x86_64-avx512: 0.171, the median of seq time per element of the strided copy probe at 16384, 32768 and 65536 elements (L2 resident; 0.171, 0.131, 0.201) on an AMD Ryzen AI 9 HX 470 at 2.0 GHz (boost off), faer-benchmark data/probes/20261026-awake-team-linux/tables-confirm-exec3.md.
• aarch64-apple: 0.090, Apple M5 Max (thresholds.md section 3.2).
• every other class: the generic row, the fastest known value (the minimum over the measured classes), which keeps work serial: 0.090.

A row’s rates and the dispatch costs of the same class in threadteam come from the same machine in the same clock state (boost on or off moves both by about 1.66x and keeps their ratio within a few percent).