X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Fnnue%2Fnnue_feature_transformer.h;h=4f6a174a486667ea23acab3e3d99c4f1fe13d438;hb=4766dfc3956f78d853c5e0c4636d6f90fd93df9a;hp=ab05f8842a5cebf6bbaa838029f4fe98bd02c43a;hpb=ce4c523ad3defce2d19342d2dd1b43858ec33819;p=stockfish

diff --git a/src/nnue/nnue_feature_transformer.h b/src/nnue/nnue_feature_transformer.h
index ab05f884..4f6a174a 100644
--- a/src/nnue/nnue_feature_transformer.h
+++ b/src/nnue/nnue_feature_transformer.h
@@ -336,10 +336,17 @@ namespace Stockfish::Eval::NNUE {
       {
           const IndexType offset = HalfDimensions * p;
           const auto out = reinterpret_cast<int8x8_t*>(&output[offset]);
-          for (IndexType j = 0; j < NumChunks; ++j)
+
+          constexpr IndexType UnrollFactor = 16;
+          static_assert(UnrollFactor % UnrollFactor == 0);
+          for (IndexType j = 0; j < NumChunks; j += UnrollFactor)
           {
-              int16x8_t sum = reinterpret_cast<const int16x8_t*>(accumulation[perspectives[p]])[j];
-              out[j] = vmax_s8(vqmovn_s16(sum), Zero);
+              int16x8_t sums[UnrollFactor];
+              for (IndexType i = 0; i < UnrollFactor; ++i)
+                sums[i] = reinterpret_cast<const int16x8_t*>(accumulation[perspectives[p]])[j+i];
+
+              for (IndexType i = 0; i < UnrollFactor; ++i)
+                out[j+i] = vmax_s8(vqmovn_s16(sums[i]), Zero);
           }
       }
       return psqt;
@@ -370,7 +377,6 @@ namespace Stockfish::Eval::NNUE {
       // That might depend on the feature set and generally relies on the
       // feature set's update cost calculation to be correct and never
       // allow updates with more added/removed features than MaxActiveDimensions.
-      using IndexList = ValueList<IndexType, FeatureSet::MaxActiveDimensions>;
 
   #ifdef VECTOR
       // Gcc-10.2 unnecessarily spills AVX2 registers if this array
@@ -383,7 +389,7 @@ namespace Stockfish::Eval::NNUE {
       // of the estimated gain in terms of features to be added/subtracted.
       StateInfo *st = pos.state(), *next = nullptr;
       int gain = FeatureSet::refresh_cost(pos);
-      while (st->accumulator.state[perspective] == EMPTY)
+      while (st->previous && !st->accumulator.computed[perspective])
       {
         // This governs when a full feature refresh is needed and how many
         // updates are better than just one full refresh.
@@ -394,7 +400,7 @@ namespace Stockfish::Eval::NNUE {
         st = st->previous;
       }
 
-      if (st->accumulator.state[perspective] == COMPUTED)
+      if (st->accumulator.computed[perspective])
       {
         if (next == nullptr)
           return;
@@ -404,16 +410,16 @@ namespace Stockfish::Eval::NNUE {
 
         // Gather all features to be updated.
         const Square ksq = pos.square<KING>(perspective);
-        IndexList removed[2], added[2];
+        FeatureSet::IndexList removed[2], added[2];
         FeatureSet::append_changed_indices(
-          ksq, next, perspective, removed[0], added[0]);
+          ksq, next->dirtyPiece, perspective, removed[0], added[0]);
         for (StateInfo *st2 = pos.state(); st2 != next; st2 = st2->previous)
           FeatureSet::append_changed_indices(
-            ksq, st2, perspective, removed[1], added[1]);
+            ksq, st2->dirtyPiece, perspective, removed[1], added[1]);
 
         // Mark the accumulators as computed.
-        next->accumulator.state[perspective] = COMPUTED;
-        pos.state()->accumulator.state[perspective] = COMPUTED;
+        next->accumulator.computed[perspective] = true;
+        pos.state()->accumulator.computed[perspective] = true;
 
         // Now update the accumulators listed in states_to_update[], where the last element is a sentinel.
         StateInfo *states_to_update[3] =
@@ -533,8 +539,8 @@ namespace Stockfish::Eval::NNUE {
       {
         // Refresh the accumulator
         auto& accumulator = pos.state()->accumulator;
-        accumulator.state[perspective] = COMPUTED;
-        IndexList active;
+        accumulator.computed[perspective] = true;
+        FeatureSet::IndexList active;
         FeatureSet::append_active_indices(pos, perspective, active);
 
   #ifdef VECTOR