in reply to Code for Constraint Satisfaction Problems
If you really want to do be able to do it computationally, this seems exceedingly difficult. One possible simplifying idea that I can suggest is to only look for linear codes. This will add some much-needed structure to the search space to exploit. You "only" need to search for a good parity-check matrix for the code. However, this might still be pretty computationally intensive. And since the numbers in that table are mostly powers of two, it is at least plausible that a linear code (whose number of codewords is always a power of two) can achieve the optimal parameters.