Fast Reciprocal Square Root with Labview

[Intaris]

The problem with the performance is precisely the

evil floating point bit level hacking 

The original code uses pointer redirection to do a no-op re-interpretation of a SGL pointer to I32 (or the other waay around I can't remember).

It's a core principle for the speed, but bear in mind that modern processors have functionalities to do this way faster than this method. This comes from the era where math co-processors were a thing and were not automatically part of the CPU.

[altenbach]

Yes, it's more an academic curiosity today. It is still interesting to implement it in LabVIEW as an exercise. 😄 

As discussed elsewhere, if we just remove the FOR loop and operate on larger arrays, the speed per element can be 40-50x faster with the exact same result. (of course for small arrays, the FOR loop version is faster)