X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Fnnue%2Flayers%2Fclipped_relu.h;h=48cd6c6934547a06852a248b6ad936733f8a0faa;hb=8a912951de6d4bff78d3ff5258213a0c7e6f494e;hp=ffd2e3b76a94523eec1beb01d6b67bec122a89d2;hpb=cb9c2594fcedc881ae8f8bfbfdf130cf89840e4c;p=stockfish

diff --git a/src/nnue/layers/clipped_relu.h b/src/nnue/layers/clipped_relu.h
index ffd2e3b7..48cd6c69 100644
--- a/src/nnue/layers/clipped_relu.h
+++ b/src/nnue/layers/clipped_relu.h
@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -21,6 +21,10 @@
 #ifndef NNUE_LAYERS_CLIPPED_RELU_H_INCLUDED
 #define NNUE_LAYERS_CLIPPED_RELU_H_INCLUDED
 
+#include <algorithm>
+#include <cstdint>
+#include <iosfwd>
+
 #include "../nnue_common.h"
 
 namespace Stockfish::Eval::NNUE::Layers {
@@ -59,7 +63,7 @@ namespace Stockfish::Eval::NNUE::Layers {
     }
 
     // Forward propagation
-    const OutputType* propagate(
+    void propagate(
         const InputType* input, OutputType* output) const {
 
   #if defined(USE_AVX2)
@@ -131,24 +135,6 @@ namespace Stockfish::Eval::NNUE::Layers {
       }
       constexpr IndexType Start = NumChunks * SimdWidth;
 
-  #elif defined(USE_MMX)
-      constexpr IndexType NumChunks = InputDimensions / SimdWidth;
-      const __m64 k0x80s = _mm_set1_pi8(-128);
-      const auto in = reinterpret_cast<const __m64*>(input);
-      const auto out = reinterpret_cast<__m64*>(output);
-      for (IndexType i = 0; i < NumChunks; ++i) {
-        const __m64 words0 = _mm_srai_pi16(
-            _mm_packs_pi32(in[i * 4 + 0], in[i * 4 + 1]),
-            WeightScaleBits);
-        const __m64 words1 = _mm_srai_pi16(
-            _mm_packs_pi32(in[i * 4 + 2], in[i * 4 + 3]),
-            WeightScaleBits);
-        const __m64 packedbytes = _mm_packs_pi16(words0, words1);
-        out[i] = _mm_subs_pi8(_mm_adds_pi8(packedbytes, k0x80s), k0x80s);
-      }
-      _mm_empty();
-      constexpr IndexType Start = NumChunks * SimdWidth;
-
   #elif defined(USE_NEON)
       constexpr IndexType NumChunks = InputDimensions / (SimdWidth / 2);
       const int8x8_t Zero = {0};
@@ -168,18 +154,8 @@ namespace Stockfish::Eval::NNUE::Layers {
 
       for (IndexType i = Start; i < InputDimensions; ++i) {
         output[i] = static_cast<OutputType>(
-            std::max(0, std::min(127, input[i] >> WeightScaleBits)));
-      }
-
-      // Affine transform layers expect that there is at least
-      // ceil_to_multiple(OutputDimensions, 32) initialized values.
-      // We cannot do this in the affine transform because it requires
-      // preallocating space here.
-      for (IndexType i = OutputDimensions; i < PaddedOutputDimensions; ++i) {
-        output[i] = 0;
+            std::clamp(input[i] >> WeightScaleBits, 0, 127));
       }
-
-      return output;
     }
   };