initial dfrc support #1684
initial dfrc support #1684
Conversation
src/chess/board.h
Outdated
| } | ||
|
|
||
| private: | ||
| // Position of "left" (queenside) rook in starting game position. | ||
| std::uint8_t queenside_rook_ : 3; | ||
| std::uint8_t our_queenside_rook_ : 3; |
There was a problem hiding this comment.
Actually, do we even care about board representation size anymore? I think we don't, we actually care about performanse more.
It probably makes sense to remove all bit fields and have just a byte per field, and split data_ into 4 bools (not sure about this part).
Also, we mix std::uint8_t and uint8_t a lot in this file, probably makes sense to remove std::.
There was a problem hiding this comment.
I left data_ as an int8_t for the time being.
|
|
||
| // Note: this is not a strict xfen compatible output. Without access to the | ||
| // board its not possible to know whether there is ambiguity so all cases | ||
| // with any non-standard rook positions are encoded in the x-fen format | ||
| std::string as_string() const { | ||
| if (data_ == 0) return "-"; | ||
| std::string result; | ||
| if (queenside_rook() == FILE_A && kingside_rook() == FILE_H) { | ||
| if (our_queenside_rook() == FILE_A && our_kingside_rook() == FILE_H && |
There was a problem hiding this comment.
It's tempting to create a helper struct, something like
struct RookPair {
uint8 kingside;
uint8 queenside;
bool can_castle_queenside;
bool can_castle_kingside;
bool is_standard_chess() const { return queenside == FILE_A && kindside == FILE_H; }
// and other similar functions.
};
Not sure if worth it though.
There was a problem hiding this comment.
Started doing it, but didn't see a lot of potential, so I gave up.
src/chess/board.cc
Outdated
| // Finding rightmost rook. | ||
| for (right_rook = FILE_H; right_rook > king_col; --right_rook) { | ||
| if (rooks.get(is_black ? RANK_8 : RANK_1, right_rook)) break; | ||
| for (our_right_rook = FILE_H; our_right_rook > king_col; |
There was a problem hiding this comment.
Feels that this code should be refactored, too duplicated..
Maybe have a function to find lowest and highest bit on the bit level (on "our rooks" bitmask)?
But I guess it let's decide whether to factor our struct RookPair first.
There was a problem hiding this comment.
Did some refactoring for now.
Naphthalin
added
the
testing required
|
Actually there is a ~1% speed gain on a single thread trivial backend test. |
Very lightly tested