GitLab

Fixed issue with UINT_MAX not being found when executed using the compiled program by adding the header.

Added sqrt and if. If can be found easily, sqrt does not produce any good result ever.

Added fitness-penalty to programs above the maximum size. Multiplies the fitness by the percentage above the maximum size.
Also, fixed bug causing the best-fitness to be displayed at 0.

Added it as redistribution in the crossover, bigger blocks being taken from bigger programs if above the size.
However, if both programs are too big, it does not help much.

What I thought was a bug was actually intended behaviour. Removed from todo and adapted the testcase.

Changed the import-program to use a hashset for sorting out multiple occurances of programs. Had to add a hash to lines, programs and the Pair.

Fixed bug that did not group fitness-classes after the first in the little program that checks what's the best.
Adapted the scripts for evaluating to be easier to use and include that^.

Created little program that takes exported programs.json as input, and extracts the best fitness and the best programs from this set.

In the opencl-evaluation changed "found solution" to exclude programs that hit the runtime-limit.

Fixed running non-heuristic on island-model (issue was that copying function-pointer when copying object into vector had dangling reference)

TODO: Fix at island-model.

TODO: Fix at island-model.

Added methods to only generate new programs from scratch, and not use the operators to adapt programs, to be able to make the comparisons.

TODO: Add that to gui, add that to island-model.

Fixed segmentation-fault on accessing/adding elements on the maps, at running the compiled programs.

still TODO: Race-conditions on increasing DSS-difficulty on compiled programs.

STILL NEEDS SYNCHRONIZATION TO AVOID RACE-CONDITIONS!

Still takes FAR too long to evaluate, and the results are faulty.

Still need to use OpenMP for faster execution, setting the run-limits on the gui, "caching" outputs.

Uses dynamic_bitset for the calculation of the unique ID now. Unique-ID has delimiter between operands + sizes now, and should really be unique now.
Fixed a bug in unique-id-calculation, now it's a lot faster.

Now starts proper subprocess for running binary program, waits for amount seconds, then sends signal to stop. Gets final outputs on completion, and current values on kill.

Issue: Running it gives me some error. Maybe really use boost/subprocess?

Issue: Running it gives me some error. Maybe really use boost/subprocess?