float16 provides IEEE 754 half-precision (binary16) with IEEE 754 default rounding for conversions. All 4+ billion possible conversions between float16 and float32 are tested to be correct.
Geek Repo:Geek Repo
Github PK Tool:Github PK Tool