Why doesn't convertPcm16ToFloat do source[i] / 32768.0f?

[dturner]

This line of code (from https://github.com/google/oboe/blob/master/src/common/Utilities.cpp#L43)

destination[i] = source[i] * (1.0f / 32768.0f);

Why not just use:

source[i] / 32768.0f

[dturner]

Originally asked by mnaganov

[philburk]

A multiply is generally much faster than a divide.
Some compilers will pre-calculate the 1/x at compile time and then compile a multiply for the first line. But they will compile a divide for the second line.

It might look better to define a const float kScaleI16ToFloat = (1.0f/32768.0f)
and then dest = source * kScaleI16ToFloat;