tomoto 0.2.2 → 0.3.0

data/vendor/EigenRand/EigenRand/MvDists/Multinomial.h

@@ -2,10 +2,10 @@
  * @file Multinomial.h
  * @author bab2min (bab2min@gmail.com)
  * @brief
- * @version 0.3.0
- * @date 2020-10-07
+ * @version 0.3.3
+ * @date 2021-03-31
  *
- * @copyright Copyright (c) 2020
+ * @copyright Copyright (c) 2020-2021
  *
  */
 
@@ -147,6 +147,8 @@ namespace Eigen
 		template<typename _Scalar, Index Dim = -1>
 		class DirichletGen : public MvVecGenBase<DirichletGen<_Scalar, Dim>, _Scalar, Dim>
 		{
+			static_assert(std::is_floating_point<_Scalar>::value, "`DirichletGen` needs floating point types.");
+
 			Matrix<_Scalar, Dim, 1> alpha;
 			std::vector<GammaGen<_Scalar>> gammas;
 		public:

data/vendor/EigenRand/EigenRand/MvDists/MvNormal.h

@@ -2,10 +2,10 @@
  * @file MvNormal.h
  * @author bab2min (bab2min@gmail.com)
  * @brief
- * @version 0.3.0
- * @date 2020-10-07
+ * @version 0.3.3
+ * @date 2021-03-31
  *
- * @copyright Copyright (c) 2020
+ * @copyright Copyright (c) 2020-2021
  *
  */
 
@@ -56,6 +56,8 @@ namespace Eigen
 		template<typename _Scalar, Index Dim = -1>
 		class MvNormalGen : public MvVecGenBase<MvNormalGen<_Scalar, Dim>, _Scalar, Dim>
 		{
+			static_assert(std::is_floating_point<_Scalar>::value, "`MvNormalGen` needs floating point types.");
+
 			Matrix<_Scalar, Dim, 1> mean;
 			Matrix<_Scalar, Dim, Dim> lt;
 			StdNormalGen<_Scalar> stdnorm;
@@ -170,6 +172,8 @@ namespace Eigen
 		template<typename _Scalar, Index Dim>
 		class WishartGen : public MvMatGenBase<WishartGen<_Scalar, Dim>, _Scalar, Dim>
 		{
+			static_assert(std::is_floating_point<_Scalar>::value, "`WishartGen` needs floating point types.");
+
 			Index df;
 			Matrix<_Scalar, Dim, Dim> chol;
 			StdNormalGen<_Scalar> stdnorm;
@@ -330,6 +334,8 @@ namespace Eigen
 		template<typename _Scalar, Index Dim>
 		class InvWishartGen : public MvMatGenBase<InvWishartGen<_Scalar, Dim>, _Scalar, Dim>
 		{
+			static_assert(std::is_floating_point<_Scalar>::value, "`InvWishartGen` needs floating point types.");
+
 			Index df;
 			Matrix<_Scalar, Dim, Dim> chol;
 			StdNormalGen<_Scalar> stdnorm;

data/vendor/EigenRand/EigenRand/PacketFilter.h

@@ -2,10 +2,10 @@
  * @file PacketFilter.h
  * @author bab2min (bab2min@gmail.com)
  * @brief 
- * @version 0.3.0
- * @date 2020-10-07
+ * @version 0.3.3
+ * @date 2021-03-31
  * 
- * @copyright Copyright (c) 2020
+ * @copyright Copyright (c) 2020-2021
  * 
  */
 
@@ -13,7 +13,7 @@
 #define EIGENRAND_PACKET_FILTER_H
 
 #include <array>
-#include <EigenRand/MorePacketMath.h>
+#include "MorePacketMath.h"
 
 namespace Eigen
 {
@@ -26,261 +26,19 @@ namespace Eigen
 		}
 	}
 }
-#ifdef EIGEN_VECTORIZE_AVX
-#include <immintrin.h>
-namespace Eigen
-{
-	namespace Rand
-	{
-		namespace detail
-		{
-			template<>
-			class CompressMask<32>
-			{
-				std::array<std::array<internal::Packet8i, 256>, 15> idx;
-				std::array<internal::Packet8f, 8> selector;
-				std::array<uint8_t, 256> cnt;
-
-				static internal::Packet8i make_compress(int mask, int offset = 0)
-				{
-					int32_t ret[8] = { 0, };
-					int n = offset;
-					for (int i = 0; i < 8; ++i)
-					{
-						int l = mask & 1;
-						mask >>= 1;
-						if (l)
-						{
-							if (n >= 0) ret[n] = i;
-							if (++n >= 8) break;
-						}
-					}
-					return _mm256_loadu_si256((internal::Packet8i*)ret);
-				}
-
-				static uint8_t count(int mask)
-				{
-					uint8_t ret = 0;
-					for (int i = 0; i < 8; ++i)
-					{
-						ret += mask & 1;
-						mask >>= 1;
-					}
-					return ret;
-				}
 
-				CompressMask()
-				{
-					for (int i = 0; i < 256; ++i)
-					{
-						for (int o = 0; o < 15; ++o)
-						{
-							idx[o][i] = make_compress(i, o < 8 ? o : o - 15);
-						}
 
-						cnt[i] = count(i);
-					}
-
-					selector[0] = _mm256_castsi256_ps(_mm256_setr_epi32(0, 0, 0, 0, 0, 0, 0, 0));
-					selector[1] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, 0, 0, 0, 0, 0, 0, 0));
-					selector[2] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, -1, 0, 0, 0, 0, 0, 0));
-					selector[3] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, -1, -1, 0, 0, 0, 0, 0));
-					selector[4] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, -1, -1, -1, 0, 0, 0, 0));
-					selector[5] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, -1, -1, -1, -1, 0, 0, 0));
-					selector[6] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, -1, -1, -1, -1, -1, 0, 0));
-					selector[7] = _mm256_castsi256_ps(_mm256_setr_epi32(-1, -1, -1, -1, -1, -1, -1, 0));
-				}
-
-				static EIGEN_STRONG_INLINE internal::Packet8f permute(const internal::Packet8f& p, const internal::Packet8i& i)
-				{
-#ifdef EIGEN_VECTORIZE_AVX2
-					return _mm256_permutevar8x32_ps(p, i);
-#else
-					auto l = _mm256_permutevar_ps(p, i);
-					auto h = _mm256_permutevar_ps(_mm256_permute2f128_ps(p, p, 0x01), i);
-					internal::Packet4i i1, i2;
-					internal::split_two(i, i1, i2);
-					i1 = _mm_slli_epi32(i1, 29);
-					i2 = _mm_slli_epi32(i2, 29);
-					auto c = _mm256_castsi256_ps(
-						internal::combine_two(
-							_mm_cmplt_epi32(i1, internal::pset1<internal::Packet4i>(0)),
-							_mm_cmplt_epi32(internal::pset1<internal::Packet4i>(-1), i2)
-						)
-					);
-					return internal::pblendv(c, h, l);
-#endif
-				}
-
-			public:
-				enum { full_size = 8 };
-				static const CompressMask& get_inst()
-				{
-					static CompressMask cm;
-					return cm;
-				}
-
-				template<typename Packet>
-				EIGEN_STRONG_INLINE int compress_append(Packet& _value, const Packet& _mask,
-					Packet& _rest, int rest_cnt, bool& full) const
-				{
-					auto& value = reinterpret_cast<internal::Packet8f&>(_value);
-					auto& mask = reinterpret_cast<const internal::Packet8f&>(_mask);
-					auto& rest = reinterpret_cast<internal::Packet8f&>(_rest);
-
-					int m = _mm256_movemask_ps(mask);
-					if (cnt[m] == full_size)
-					{
-						full = true;
-						return rest_cnt;
-					}
-
-					auto p1 = permute(value, idx[rest_cnt][m]);
-					p1 = internal::pblendv(selector[rest_cnt], rest, p1);
-
-					auto new_cnt = rest_cnt + cnt[m];
-					if (new_cnt >= full_size)
-					{
-						if (new_cnt > full_size)
-						{
-							rest = permute(value, idx[new_cnt - cnt[m] + full_size - 1][m]);
-						}
-						value = p1;
-						full = true;
-						return new_cnt - full_size;
-					}
-					else
-					{
-						rest = p1;
-						full = false;
-						return new_cnt;
-					}
-				}
-			};
-		}
-	}
-}
+#ifdef EIGEN_VECTORIZE_AVX
+#include "arch/AVX/PacketFilter.h"
 #endif
 
 #ifdef EIGEN_VECTORIZE_SSE2
-#include <xmmintrin.h>
-
-namespace Eigen
-{
-	namespace Rand
-	{
-		namespace detail
-		{
-			template<>
-			class CompressMask<16>
-			{
-				std::array<std::array<uint8_t, 16>, 7> idx;
-				std::array<internal::Packet4f, 4> selector;
-				std::array<uint8_t, 64> cnt;
-
-				static uint8_t make_compress(int mask, int offset = 0)
-				{
-					uint8_t ret = 0;
-					int n = offset;
-					for (int i = 0; i < 4; ++i)
-					{
-						int l = mask & 1;
-						mask >>= 1;
-						if (l)
-						{
-							if (n >= 0) ret |= (i & 3) << (2 * n);
-							if (++n >= 4) break;
-						}
-					}
-					return ret;
-				}
-
-				static uint8_t count(int mask)
-				{
-					uint8_t ret = 0;
-					for (int i = 0; i < 4; ++i)
-					{
-						ret += mask & 1;
-						mask >>= 1;
-					}
-					return ret;
-				}
-
-				CompressMask()
-				{
-					for (int i = 0; i < 16; ++i)
-					{
-						for (int o = 0; o < 7; ++o)
-						{
-							idx[o][i] = make_compress(i, o < 4 ? o : o - 7);
-						}
-
-						cnt[i] = count(i);
-					}
-
-					selector[0] = _mm_castsi128_ps(_mm_setr_epi32(0, 0, 0, 0));
-					selector[1] = _mm_castsi128_ps(_mm_setr_epi32(-1, 0, 0, 0));
-					selector[2] = _mm_castsi128_ps(_mm_setr_epi32(-1, -1, 0, 0));
-					selector[3] = _mm_castsi128_ps(_mm_setr_epi32(-1, -1, -1, 0));
-				}
-
-				static EIGEN_STRONG_INLINE internal::Packet4f permute(const internal::Packet4f& p, uint8_t i)
-				{
-					float u[4];
-					_mm_storeu_ps(u, p);
-					return _mm_setr_ps(u[i & 3], u[(i >> 2) & 3], u[(i >> 4) & 3], u[(i >> 6) & 3]);
-				}
-
-			public:
-
-				enum { full_size = 4 };
-
-				static const CompressMask& get_inst()
-				{
-					static CompressMask cm;
-					return cm;
-				}
-
-				template<typename Packet>
-				EIGEN_STRONG_INLINE int compress_append(Packet& _value, const Packet& _mask,
-					Packet& _rest, int rest_cnt, bool& full) const
-				{
-					auto& value = reinterpret_cast<internal::Packet4f&>(_value);
-					auto& mask = reinterpret_cast<const internal::Packet4f&>(_mask);
-					auto& rest = reinterpret_cast<internal::Packet4f&>(_rest);
-
-					int m = _mm_movemask_ps(mask);
-					if (cnt[m] == full_size)
-					{
-						full = true;
-						return rest_cnt;
-					}
-
-					auto p1 = permute(value, idx[rest_cnt][m]);
-					p1 = internal::pblendv(selector[rest_cnt], rest, p1);
+#include "arch/SSE/PacketFilter.h"
+#endif
 
-					auto new_cnt = rest_cnt + cnt[m];
-					if (new_cnt >= full_size)
-					{
-						if (new_cnt > full_size)
-						{
-							rest = permute(value, idx[new_cnt - cnt[m] + full_size - 1][m]);
-						}
-						value = p1;
-						full = true;
-						return new_cnt - full_size;
-					}
-					else
-					{
-						rest = p1;
-						full = false;
-						return new_cnt;
-					}
-				}
-			};
-		}
-	}
-}
+#ifdef EIGEN_VECTORIZE_NEON
+#include "arch/NEON/PacketFilter.h"
 #endif
 
+
 #endif

data/vendor/EigenRand/EigenRand/PacketRandomEngine.h

@@ -2,10 +2,10 @@
  * @file PacketRandomEngine.h
  * @author bab2min (bab2min@gmail.com)
  * @brief
- * @version 0.3.0
- * @date 2020-10-07
+ * @version 0.3.3
+ * @date 2021-03-31
  *
- * @copyright Copyright (c) 2020
+ * @copyright Copyright (c) 2020-2021
  *
  */
 
@@ -15,41 +15,11 @@
 #include <array>
 #include <random>
 #include <type_traits>
-#include <EigenRand/MorePacketMath.h>
+#include "MorePacketMath.h"
 #include <fstream>
 
 namespace Eigen
 {
-	namespace internal
-	{
-		template<typename Ty>
-		struct IsIntPacket : std::false_type {};
-
-		template<typename Ty>
-		struct HalfPacket;
-
-#ifdef EIGEN_VECTORIZE_AVX2
-		template<>
-		struct IsIntPacket<Packet8i> : std::true_type {};
-
-		template<>
-		struct HalfPacket<Packet8i>
-		{
-			using type = Packet4i;
-		};
-#endif
-#ifdef EIGEN_VECTORIZE_SSE2
-		template<>
-		struct IsIntPacket<Packet4i> : std::true_type {};
-
-		template<>
-		struct HalfPacket<Packet4i>
-		{
-			using type = uint64_t;
-		};
-#endif
-	}
-
 	namespace Rand
 	{
 		namespace detail
@@ -277,7 +247,7 @@ namespace Eigen
 				Packet prev = state[0] = x0;
 				for (int i = 1; i < _Nx; ++i)
 				{
-					prev = state[i] = pmuluadd64(pxor(prev, psrl64(prev, word_size - 2)), _Fx, i);
+					prev = state[i] = pmuluadd64(pxor(prev, psrl64<word_size - 2>(prev)), _Fx, i);
 				}
 				stateIdx = _Nx;
 			}
@@ -320,10 +290,10 @@ namespace Eigen
 				using namespace Eigen::internal;
 
 				Packet res = state[stateIdx++];
-				res = pxor(res, pand(psrl64(res, _Ux), pseti64<Packet>(_Dx)));
-				res = pxor(res, pand(psll64(res, _Sx), pseti64<Packet>(_Bx)));
-				res = pxor(res, pand(psll64(res, _Tx), pseti64<Packet>(_Cx)));
-				res = pxor(res, psrl64(res, _Lx));
+				res = pxor(res, pand(psrl64<_Ux>(res), pseti64<Packet>(_Dx)));
+				res = pxor(res, pand(psll64<_Sx>(res), pseti64<Packet>(_Bx)));
+				res = pxor(res, pand(psll64<_Tx>(res), pseti64<Packet>(_Cx)));
+				res = pxor(res, psrl64<_Lx>(res));
 				return res;
 			}
 
@@ -371,7 +341,7 @@ namespace Eigen
 						pand(state[i + _Nx + 1], lmask));
 
 					state[i] = pxor(pxor(
-						psrl64(tmp, 1),
+						psrl64<1>(tmp),
 						pand(pcmpeq64(pand(tmp, one), one), px)),
 						state[i + _Nx + _Mx]
 					);
@@ -383,7 +353,7 @@ namespace Eigen
 						pand(state[i + _Nx + 1], lmask));
 
 					state[i] = pxor(pxor(
-						psrl64(tmp, 1),
+						psrl64<1>(tmp),
 						pand(pcmpeq64(pand(tmp, one), one), px)),
 						state[i - _Nx + _Mx]
 					);
@@ -392,7 +362,7 @@ namespace Eigen
 				Packet tmp = por(pand(state[i + _Nx], hmask),
 					pand(state[0], lmask));
 				state[i] = pxor(pxor(
-					psrl64(tmp, 1),
+					psrl64<1>(tmp),
 					pand(pcmpeq64(pand(tmp, one), one), px)),
 					state[_Mx - 1]
 				);
@@ -414,7 +384,7 @@ namespace Eigen
 						pand(state[i - _Nx + 1], lmask));
 
 					state[i] = pxor(pxor(
-						psrl64(tmp, 1),
+						psrl64<1>(tmp),
 						pand(pcmpeq64(pand(tmp, one), one), px)),
 						state[i - _Nx + _Mx]
 					);
@@ -567,7 +537,7 @@ namespace Eigen
 
 			float uniform_real()
 			{
-				internal::bit_scalar<float> bs;
+				internal::BitScalar<float> bs;
 				return bs.to_ur(this->operator()());
 			}
 		};
@@ -599,7 +569,7 @@ namespace Eigen
 
 #ifdef EIGEN_VECTORIZE_AVX2
 		using Vmt19937_64 = Pmt19937_64<internal::Packet8i>;
-#elif defined(EIGEN_VECTORIZE_AVX) || defined(EIGEN_VECTORIZE_SSE2)
+#elif defined(EIGEN_VECTORIZE_AVX) || defined(EIGEN_VECTORIZE_SSE2) || defined(EIGEN_VECTORIZE_NEON)
 		using Vmt19937_64 = Pmt19937_64<internal::Packet4i>;
 #else
 		/**
@@ -612,12 +582,16 @@ namespace Eigen
 		 */
 		using Vmt19937_64 = std::mt19937_64;
 #endif
+		template<typename UIntType = uint64_t>
+		using P8_mt19937 = ParallelRandomEngineAdaptor<UIntType, Vmt19937_64, 8>;
+
 		/**
 		 * @brief a vectorized mt19937_64 which generates 8 integers of 64bit simultaneously.
 		 * It always yields the same value regardless of SIMD ISA.
 		 */
-		template<typename UIntType = uint64_t>
-		using P8_mt19937_64 = ParallelRandomEngineAdaptor<UIntType, Vmt19937_64, 8>;
+		using P8_mt19937_64 = P8_mt19937<uint64_t>;
+
+		using P8_mt19937_64_32 = P8_mt19937<uint32_t>;
 	}
 }