Wire NNUE eval into real search path; fix critical perf regression

Load-and-call infra: export_nnue_weights.py-produced weight files can
now be loaded at runtime (nnueeval.c's NnueLoadWeights, "nnue load"
command, --nnueweights CLI flag for scripted match-play), and the FC
head (NnueForward) runs the real forward pass against the live
accumulator.

Real integration: Eval() now short-circuits to NnueForward when
g_fUseNnueEval && g_fNnueWeightsLoaded, so ./typhoon can actually play
on the NNUE eval -- ./eval_tune/test_vs_head.sh works unmodified to
compare it against the old HCE. ReInitializeSearcherContext (root.c)
gained a defensive accumulator recompute hook covering setboard's
downstream effect on the next go/ponder; it only fires when both
kings are validly placed, since testrecogn.c's synthetic positions
don't always have one at the moment this is called.

Critical fix: the NNUE accumulator maintenance added to move.c's 8
primitives (and MakeMove/UnmakeMove's king-move backup/restore) was
running unconditionally, on the mistaken assumption this was cheap.
It cost ~50% of engine throughput even with NNUE off (673k nps
measured vs the expected 1.4-1.5M baseline). Gated all of it behind
g_fUseNnueEval; classic HCE is back to ~1.52M nps, NNUE eval runs at
a fair ~409k nps (expected for an unquantized float32 forward pass).

This gating also exposed that testnnue.c's TestNnueAccumulator never
set g_fUseNnueEval, so its own direct NnueAccumRecompute calls raced
ahead of now-gated-off incremental maintenance; fixed by having the
test set/restore the flag around its run.

board.c: added en passant legality check to IsBoardLegal (rank vs
side-to-move, origin-square occupancy) -- a real, previously-latent
bug surfaced while building the NNUE test fixtures, unrelated to
NNUE itself.

Co-Authored-By: Claude Sonnet 5 <[email protected]>

1 files changed, 28 insertions, 2 deletions

diff --git a/src/chess.h b/src/chess.h
index 1f67df1..cfe1d4e 100755
--- a/src/chess.h
+++ b/src/chess.h
@@ -205,8 +205,7 @@ typedef struct _DLIST_ENTRY
 //
 // Program version number
 //
-#define VERSION                    "1.00"
-#define REVISION                   "$Id: chess.h 354 2008-06-30 05:10:08Z scott $\n"
+#define VERSION                    "2.00"
 
 // GIT_COMMIT is normally injected by GNUmakefile via -DGIT_COMMIT (short
 // hash, plus a -dirty suffix if the tree had uncommitted changes at build
@@ -1641,6 +1640,8 @@ MOVE_SCORING_SCHEME;
 // weights exist -- not something to conflate with debugging the
 // harder plumbing now.
 #define NNUE_ACCUMULATOR_SIZE (256)  // L1, nnue_model.py
+#define NNUE_L2 (32)                 // nnue_model.py's L2
+#define NNUE_L3 (32)                 // nnue_model.py's L3
 
 typedef struct _NNUE_ACCUMULATOR
 {
@@ -2325,6 +2326,31 @@ NnueAccumRemoveFeature(NNUE_ACCUMULATOR *pAccum, ULONG uPerspective,
 void
 NnueAccumRecompute(NNUE_ACCUMULATOR *pAccum, POSITION *pos, ULONG uPerspective);
 
+//
+// FC head weights (nnueeval.c) -- the dense 2*L1 -> L2 -> L3 -> 1
+// layers that sit on top of the two accumulator halves. Still a
+// placeholder (zero-initialized) until NnueLoadWeights successfully
+// reads a real exported checkpoint (annotate_pgns/
+// export_nnue_weights.py's output format). g_fNnueWeightsLoaded is
+// FALSE until that happens -- NnueForward does not check it itself
+// (callers should, before trusting its output for anything beyond
+// plumbing validation).
+//
+extern float g_rgfNnueFc1Weight[NNUE_L2][2 * NNUE_ACCUMULATOR_SIZE];
+extern float g_rgfNnueFc1Bias[NNUE_L2];
+extern float g_rgfNnueFc2Weight[NNUE_L3][NNUE_L2];
+extern float g_rgfNnueFc2Bias[NNUE_L3];
+extern float g_rgfNnueFc3Weight[NNUE_L3];
+extern float g_rgfNnueFc3Bias;
+extern FLAG g_fNnueWeightsLoaded;
+extern FLAG g_fUseNnueEval;
+
+FLAG
+NnueLoadWeights(CHAR *szFilename);
+
+SCORE
+NnueForward(SEARCHER_THREAD_CONTEXT *ctx);
+
 #ifdef PERF_COUNTERS
 #define INC(x)                     ((x) += 1)
 #else