Why there is no Float16 type?

[FengJungle]

As the title, why there is no float16 type?

[AlexandreEichenberger]

We typically add functionality as we encounter it. I would encourage you to contribute this simple extension.

[FengJungle]

We typically add functionality as we encounter it. I would encourage you to contribute this simple extension.

How about use ±65504 as the max/min limit for float16 type?

[AlexandreEichenberger]

I am not an expert in the limit values for float16, but if the integer representation you listed above corresponds to min & max float16, feel free to add this case to the switch.