As implemented, this is a multiplication of the branch probabilities (extracted from LLVM) along the path. However, in this way, longer paths are penalized, when instead longer paths should be chosen as error accumulates more along a longer path. Possible fix could be addition of branch probabilities along the path.
domainexpert
commented
7 years ago
As implemented, this is a multiplication of the branch probabilities (extracted from LLVM) along the path. However, in this way, longer paths are penalized, when instead longer paths should be chosen as error accumulates more along a longer path. Possible fix could be addition of branch probabilities along the path.