rust port

[Disservin]

How to test this branch, check it out via git and run
make this should give you a ./fastchess executable in the same dir.
This requires cargo/rust
https://rust-lang.org/tools/install/

• curious about things which dont work yet

changes:

• has experimental shogi support with usi
• logging, uses a separate lightweight thread to allow for very fast time controls with realtime engine logging enabled
• ctrl c uses a separate lightweight thread too
• massive cleanup on game assignments and color usage
• easier tests
• crc32fast dependency which uses simd
• better future variant support
• faster engine shutdown time

breaking changes:

• non intended

issues (known):

• none known

fixed issues (done):

• not exiting when all respond with Engine sf-dev2 did not respond to uciok in time.
  This wasn't actually an issue but rather slow termination, similar to
  refactor: make termination of engines happen in parallel #984
• drop illegal enpassant squares when parsing fen
• missing compliance test

perf:

./fastchess -engine name=sf-dev1 cmd=./stockfish \
    -engine name=sf-dev2 cmd=./stockfish \
    -rounds 50 -concurrency -1 -openings file=UHO_Lichess_4852_v1.epd format=epd \
    -each tc=1+0.01 dir=/home/me/Github/Stockfish/src/ -check-mate-pvs;

Language	Memory	Time
Rust	216M	18s
C++	217M	20s

[Disservin]

todo: drop illegal enpassant squares when parsing fen

[KazApps]

works on my end too - no time losses
even with 192 concurrency at 1.00+0.01
huge thanks!

[Disservin]

cool very nice

[KazApps]

fastchess/app/src/variants/shogi.rs

Line 933 in 891d92a

assert!(!startpos.is_in_check());

is this assert really necessary?
the book I use for stoat tests includes some positions where the side to move is in check, so fastchess crashes because of it

[Disservin]

hm that came straight from @87flowers https://github.com/87flowers/shogitest/blob/f620e5b69911ff4d37807f11bcf4da93820e7fd7/src/shogi.rs#L1002 maybe they can clarify the intention?

I know very little about shogi

[87flowers]

I hadn't thought too hard about the last_not_in_check_ply logic when the opening is in check, and had put that in so I could think about it at a later date.
Removing the assert is fine.

I suggest the following change:

---

diff --git a/app/src/variants/shogi.rs b/app/src/variants/shogi.rs
index cc874d8..0cd6769 100644
--- a/app/src/variants/shogi.rs
+++ b/app/src/variants/shogi.rs
@@ -930,12 +930,15 @@ pub(crate) struct Game {
 
 impl Game {
     pub fn new(startpos: Position) -> Game {
-        assert!(!startpos.is_in_check());
+        let mut last_not_in_check_ply = [-1; 2];
+        if !startpos.is_in_check() {
+            last_not_in_check_ply[startpos.stm.to_index()] = 0;
+        }
         Game {
             current_position: startpos,
             moves: vec![],
             history: vec![startpos],
-            last_not_in_check_ply: [-1, -1],
+            last_not_in_check_ply,
         }
     }

[KazApps]

here you're assigning ply - 1:

fastchess/app/src/variants/shogi.rs

Line 984 in 891d92a

self.last_not_in_check_ply[stm.to_index()] = (self.history.len() - 1) as isize

but here you're assigning the ply:
last_not_in_check_ply[startpos.stm.to_index()] = 0;

won't that cause any issues?

[87flowers]

             history: vec![startpos],

history starts with history.len() == 1

[KazApps]

ohhh, you're right

[Disservin]

okay thanks you two :) pushed a patch

[Disservin]

if there are more shogi things to support you can list them, though I'll (myself) will probably only implement the sane ones and not delve into weird cutechess things.. unless someone else takes that up

[KazApps]

@Disservin

fastchess/app/src/variants/shogi.rs

Lines 1416 to 1420 in 2de63ca

	fn is_threefold_repetition(&self) -> bool {
	// Shogi uses fourfold repetition (sennichite), not threefold
	// For now, return false as we don't track position history
	false
	}

could you update this?

[KazApps]

fastchess/app/src/variants/shogi.rs

Lines 1353 to 1357 in 2de63ca

	fn status(&self) -> GameStatus {
	// Check for game over conditions
	// For now, we return ongoing - the actual outcome is tracked by do_move
	GameStatus::ONGOING
	}

this part as well

[Disservin]

wdym

[KazApps]

I mean I think it should be something like this:

--- a/app/src/variants/shogi.rs
+++ b/app/src/variants/shogi.rs
@@ -1211,7 +1211,7 @@
 
 // ── Game Trait Implementation ────────────────────────────────────────────────
 
-use crate::game::{GameStatus, Side as GameColor};
+use crate::game::{GameOverReason, GameStatus, Side as GameColor};
 use crate::types::VariantType;
 use crate::variants::GameMove;
 
@@ -1262,6 +1262,7 @@
 pub struct ShogiGame {
     game: Game,
     ply_count: u32,
+    last_outcome: GameOutcome,
 }
 
 impl ShogiGame {
@@ -1270,6 +1271,7 @@
         Self {
             game: Game::new(Position::default()),
             ply_count: 0,
+            last_outcome: GameOutcome::Undetermined,
         }
     }
 
@@ -1279,6 +1281,7 @@
         Some(Self {
             game: Game::new(position),
             ply_count: 0,
+            last_outcome: GameOutcome::Undetermined,
         })
     }
 
@@ -1290,6 +1293,7 @@
     /// Makes a move and returns true if successful.
     fn make_shogi_move(&mut self, mv: &ShogiMove) -> bool {
         let outcome = self.game.do_move(mv.inner());
+        self.last_outcome = outcome;
         if outcome != GameOutcome::LossByIllegal(self.game.stm()) {
             self.ply_count += 1;
             true
@@ -1351,9 +1355,12 @@
     }
 
     fn status(&self) -> GameStatus {
-        // Check for game over conditions
-        // For now, we return ongoing - the actual outcome is tracked by do_move
-        GameStatus::ONGOING
+        let reason = match self.last_outcome {
+            GameOutcome::Checkmated(_) => GameOverReason::Checkmate,
+            GameOutcome::DrawBySennichite => GameOverReason::Repetition,
+            _ => GameOverReason::None,
+        };
+        GameStatus { reason }
     }
 
     fn parse_move(&self, notation: &str) -> Option<Box<dyn GameMove>> {
@@ -1414,8 +1421,6 @@
     }
 
     fn is_threefold_repetition(&self) -> bool {
-        // Shogi uses fourfold repetition (sennichite), not threefold
-        // For now, return false as we don't track position history
-        false
+        self.last_outcome == GameOutcome::DrawBySennichite
     }
 }