The evaluation functions takes up a significant amount of time in the whole search process (second only to probably the legal move generation). The current implementation takes about 0.050 ms, which I believe could be cut down to 0.025 ms if we would use C to evaluate the board position.
It appears as though numpy also has a C API which could be used to use numpy in C. This would also make it a lot easier and faster to call the transposition table (which is also implemented in C) and given that these two mainly need to communicate with each other, this might be a significant performance boost.
The evaluation functions takes up a significant amount of time in the whole search process (second only to probably the legal move generation). The current implementation takes about
0.050 ms, which I believe could be cut down to0.025 msif we would use C to evaluate the board position.It appears as though numpy also has a C API which could be used to use numpy in C. This would also make it a lot easier and faster to call the transposition table (which is also implemented in C) and given that these two mainly need to communicate with each other, this might be a significant performance boost.