Function _mm256_rcp_ph 

pub fn _mm256_rcp_ph(a: __m256h) -> __m256h

Available on x86 and target feature avx512fp16 and target feature avx512vl only.

Compute the approximate reciprocal of packed 16-bit floating-point elements in a and stores the results in dst. The maximum relative error for this approximation is less than 1.5*2^-12.

Intel’s documentation