/*
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
#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 {
}
// Forward propagation
- const OutputType* propagate(
+ void propagate(
const InputType* input, OutputType* output) const {
#if defined(USE_AVX2)
}
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};
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;
}
};