This code is usually running on an embedded CPU without FPU. In
addtition to its inefficience, the algorithm is also much harder to
understand.

Replace the logarithm formula with a simple loop.

(cherry picked from commit f2e0f7e3)