Half-float conversion algorithm in C?

Question

Possible Duplicate:
32-bit to 16-bit Floating Point Conversion

How do I convert between 32-bit floats and 16-bit half-precision floats in C?

What is this type with 16-bit half-precision float? – Kiril Kirov Aug 13 '12 at 11:05 — Kiril Kirov, Aug 13 '12 at 11:05

score 1 · Answer 1 · answered Aug 13 '12 at 11:09

1

I suggest you checking up this half_floats

You should also think about the way you will store these half-floats (there is no such type in C)

answered Aug 13 '12 at 11:09

Xentatt

score 1 · Accepted Answer · answered Aug 13 '12 at 11:12

Since C typically doesn't have a 16-bit floating point type built-in, I would expect you would either have to

Use some super-specific platform "magic", like a dedicated CPU instruction if available. This will of course not be a C solution, it will likely be inline assembly in the C source. Not portable.
Manually extract the required bits into perhaps an unsigned short. This will be portable, but might be a lot slower.

Both of these assume all you want to do is build the value's representation, you can't do arithmetic using C's operators of course.

2 Answers2