I copy pasted the code and ran a unit test, which it failed to pass. Inputs are a positive floats and the outputs are compared to the results of log2f, then the average error is calculated, which was way off.
Did you try to compare results of input values between 0.5 and 1.5 to the library log2f function?
Did you try to compare results of input values between 0.5 and 1.5 to the library log2f function?