m = log2(n) where n is the number of data points in the input.
log2(n) = the logarithm, base 2, of n. Essentially the number of
bits in the value of n.
Since the size of data[] is 128, the n is 7 because 2**7 == 128.
The number of input points/samples for FFT's should be a nice
power of 2 (8,16,32,etc). If not, zero-filled at the end to fill that
data array to the next highest power of two before an FFT is
performed.
Hope that helps.