std :: ml::float16
#include <std/ml/float16.h> source
#include <std/ml/float16.h>
#define HALF_CPU_BF16_CONVERT 4
float32 -> bf16 instructions (AVX512-BF16 / FEAT_BF16)