half precision to double

[rickford66]

Anyone come up with a way to do a fast conversion of half precision float to double precision float?  I haven't found anything on it and it seems that NI might have already addressed this so I don't have to reinvent the wheel.

Thanks.

[crossrulz]

How are you recieving the half-precision value?  String?  U16?  U8s?  I'm looking for a place to start.  I'm not familiar with any built-in functions, but it looks like it should be simple enough to make.

[altenbach]

---

@rickford66 wrote:

Anyone come up with a way to do a fast conversion of half precision float to double precision float?

---

Do you already have a slow version and are trying to speed it up?

I think all reasonable solution (i.e. not invoving boolean arrays in intermediary steps) will be very fast. I would start here, mask out the sign, exponent, and mantissa, shift if needed, and reassemble into a DBL.

If speed is really critical, you could make a lookup table, it's only 2^16 possible outcomes! 😄

[paul_a_cardinale]

.

[rickford66]

I have coded up a conversion using tables for the exponent and fraction, but it didn't occur to me to make a table for the entire range.  That would be by far the fastest.  Thanks.