chipbuster / genetic-discrepancy Goto Github PK

When running with a fixed RANDOM_SEED (usually zero), it is sometimes still possible to create different outputs.

This issue disappears when OMP_NUM_THREADS=1 is added to the environment, suggesting a race condition.

Initial bug tested under g++, unconfirmed with other compilers.

Not sure if this was intentional or a side effect of prototyping, but the types in this code can be a bit inconsistent.

E.g. in calcFitness, n and M are unsigned, but d is signed. Is there a specific reason for this? If not, I can change it around.

Also, is there any reason we wouldn't want to fix the bitwidths of these fields? e.g. if we fix on 32-bits, using uint32_t instead of unsigned.