Compare dict vs. bitboard for tensor conversion

[JamesCarr1]

Currently, conversion from Board object to a tensor (through as_tensor) is very slow. Compare two options:

• use of bit boards (board.pieces)
• use of a dictionary (board.piecemap) and cycle through, updating the position.
• get fen and cycle through (current method)

[JamesCarr1]

Running the conversion 10000 times each yields the following results:

fen_method took 0.3049468994140625 seconds to execute 10000 times.

dict_method took 0.14049696922302246 seconds to execute 10000 times.

my_bb_method took 0.09064698219299316 seconds to execute 10000 times.

ex_bitboard_method took 0.12563681602478027 seconds to execute 10000 times.

Clearly my_bb_method is the best.

[JamesCarr1]

Now further improving the efficiency of my_bb_method. Previously, was defining the lists pieces and colours INSIDE the function. Now comparing two options:

• using variables.py
• defining them outside the function and passing them in
• using a class

[JamesCarr1]

Ever so slightly slower passing them in rather than using variables.py:

• Passing variables in: my_bitboard_method took 0.08989787101745605 seconds to execute 10000 times.
• Using variables.py:my_bitboard_method took 0.08812713623046875 seconds to execute 10000 times.
• Using a class (self.colours etc.): my_bitboard_method took 0.09048104286193848 seconds to execute 10000 times.

Therefore, use variables.py