ianfab
commented
5 years ago Thanks for reporting it. I unfortunately do not have a setup for testing the compilation with MSVC at the moment, but there are a few issues I found in the code that are related to the issues you are seeing:
- Bitbases do not work and are not used for large boards at the moment, so the initialization should consequently also be skipped. This should fix the first issue, as you already mentioned.
- The bitboard operations popcount, lsb and msb are not implemented properly for MSVC in the large-board version, so it can not work at the moment. https://github.com/ianfab/Fairy-Stockfish/blob/0811ca24daad29b003c00b00a2bfc520229f4ac9/src/bitboard.h#L373-L377 https://github.com/ianfab/Fairy-Stockfish/blob/0811ca24daad29b003c00b00a2bfc520229f4ac9/src/bitboard.h#L415-L457 I am not entirely sure whether fixing these three functions would already completely solve the issues, but this should be the first step.
I hope I will find the time to address this soon. If you are interested, you can also try it yourself. The implementation of the bitboard operations can be analogous to how I generalized it from 64bit to 128bit for GCC: https://github.com/ianfab/Fairy-Stockfish/blob/0811ca24daad29b003c00b00a2bfc520229f4ac9/src/bitboard.h#L379-L383 https://github.com/ianfab/Fairy-Stockfish/blob/0811ca24daad29b003c00b00a2bfc520229f4ac9/src/bitboard.h#L391-L411
htomtom
When compiled with msvc using LARGEBOARDS, it seems running unlimitted time. When run into the Bitbases::init() or Threads.set(Options["Threads"]) in main(), the error occurs, throws "access conflict at 0x000000000085C070", stop at the line: repeat |= (db[idx] == UNKNOWN && db[idx].classify(db) != UNKNOWN); when comment line Bitbases::init() , the error throws within set_state( ) module，throws "access conflict at 0x000000009F71798C" stop at the line si->key ^= Zobrist::psq[pc][s];