VRCP28PS - ReCiProcal 28 Packed Single

VRCP28PS zmm1{k1}{z}, zmm2/m512/m32bcst{sae}    (V5+ER
__m512 _mm512_rcp28_round_ps(__m512 a, int sae)
__m512 _mm512_mask_rcp28_round_ps(__m512 s, __mmask16 k, __m512 a, int sae)
__m512 _mm512_maskz_rcp28_round_ps(__mmask16 k, __m512 a, int sae)

For each float, calculate approximate reciprocal of (1) and set the result to (2). (relative error < 2-28)

x86/x64 SIMD Instruction List  Feedback