Bin sizes follow floating point (exponent + mantissa) distribution (piecewise linear log approx) This ensures that for each size class, the average overhead percentage stays the same
source not stored for this graph (policy: none)
no test coverage detected