_mm256_reduce_max_ph
Classification
AVX-512, Arithmetic, CPUID Test: AVX512_FP16
Header File
Synopsis
_mm256_reduce_max_ph(__m256h a);
Description
Reduce the packed half-precision (16-bit) floating-point elements in "a" by maximum. Returns the maximum of all elements in "a".
Operation
tmp := a
FOR i := 0 to 7
tmp.fp16[i] := (tmp.fp16[i] > tmp.fp16[i+8] ? tmp.fp16[i] : tmp.fp16[i+8])
ENDFOR
FOR i := 0 to 3
tmp.fp16[i] := (tmp.fp16[i] > tmp.fp16[i+4] ? tmp.fp16[i] : tmp.fp16[i+4])
ENDFOR
FOR i := 0 to 1
tmp.fp16[i] := (tmp.fp16[i] > tmp.fp16[i+2] ? tmp.fp16[i] : tmp.fp16[i+2])
ENDFOR
dst.fp16[0] := (tmp.fp16[0] > tmp.fp16[1] ? tmp.fp16[0] : tmp.fp16[1])