tomoto 0.2.3 → 0.3.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +6 -0
- data/README.md +8 -10
- data/ext/tomoto/extconf.rb +6 -2
- data/ext/tomoto/{ext.cpp → tomoto.cpp} +1 -1
- data/lib/tomoto/version.rb +1 -1
- data/lib/tomoto.rb +5 -1
- data/vendor/EigenRand/EigenRand/Core.h +10 -10
- data/vendor/EigenRand/EigenRand/Dists/Basic.h +208 -9
- data/vendor/EigenRand/EigenRand/Dists/Discrete.h +52 -31
- data/vendor/EigenRand/EigenRand/Dists/GammaPoisson.h +9 -8
- data/vendor/EigenRand/EigenRand/Dists/NormalExp.h +28 -21
- data/vendor/EigenRand/EigenRand/EigenRand +11 -6
- data/vendor/EigenRand/EigenRand/Macro.h +13 -7
- data/vendor/EigenRand/EigenRand/MorePacketMath.h +348 -740
- data/vendor/EigenRand/EigenRand/MvDists/Multinomial.h +5 -3
- data/vendor/EigenRand/EigenRand/MvDists/MvNormal.h +9 -3
- data/vendor/EigenRand/EigenRand/PacketFilter.h +11 -253
- data/vendor/EigenRand/EigenRand/PacketRandomEngine.h +21 -47
- data/vendor/EigenRand/EigenRand/RandUtils.h +50 -344
- data/vendor/EigenRand/EigenRand/arch/AVX/MorePacketMath.h +619 -0
- data/vendor/EigenRand/EigenRand/arch/AVX/PacketFilter.h +149 -0
- data/vendor/EigenRand/EigenRand/arch/AVX/RandUtils.h +228 -0
- data/vendor/EigenRand/EigenRand/arch/NEON/MorePacketMath.h +473 -0
- data/vendor/EigenRand/EigenRand/arch/NEON/PacketFilter.h +142 -0
- data/vendor/EigenRand/EigenRand/arch/NEON/RandUtils.h +126 -0
- data/vendor/EigenRand/EigenRand/arch/SSE/MorePacketMath.h +501 -0
- data/vendor/EigenRand/EigenRand/arch/SSE/PacketFilter.h +133 -0
- data/vendor/EigenRand/EigenRand/arch/SSE/RandUtils.h +120 -0
- data/vendor/EigenRand/EigenRand/doc.h +24 -12
- data/vendor/EigenRand/README.md +57 -4
- data/vendor/eigen/COPYING.APACHE +203 -0
- data/vendor/eigen/COPYING.BSD +1 -1
- data/vendor/eigen/COPYING.MINPACK +51 -52
- data/vendor/eigen/Eigen/Cholesky +0 -1
- data/vendor/eigen/Eigen/Core +112 -265
- data/vendor/eigen/Eigen/Eigenvalues +2 -3
- data/vendor/eigen/Eigen/Geometry +5 -8
- data/vendor/eigen/Eigen/Householder +0 -1
- data/vendor/eigen/Eigen/Jacobi +0 -1
- data/vendor/eigen/Eigen/KLUSupport +41 -0
- data/vendor/eigen/Eigen/LU +2 -5
- data/vendor/eigen/Eigen/OrderingMethods +0 -3
- data/vendor/eigen/Eigen/PaStiXSupport +1 -0
- data/vendor/eigen/Eigen/PardisoSupport +0 -0
- data/vendor/eigen/Eigen/QR +2 -3
- data/vendor/eigen/Eigen/QtAlignedMalloc +0 -1
- data/vendor/eigen/Eigen/SVD +0 -1
- data/vendor/eigen/Eigen/Sparse +0 -2
- data/vendor/eigen/Eigen/SparseCholesky +0 -8
- data/vendor/eigen/Eigen/SparseLU +4 -0
- data/vendor/eigen/Eigen/SparseQR +0 -1
- data/vendor/eigen/Eigen/src/Cholesky/LDLT.h +42 -27
- data/vendor/eigen/Eigen/src/Cholesky/LLT.h +39 -23
- data/vendor/eigen/Eigen/src/CholmodSupport/CholmodSupport.h +90 -47
- data/vendor/eigen/Eigen/src/Core/ArithmeticSequence.h +413 -0
- data/vendor/eigen/Eigen/src/Core/Array.h +99 -11
- data/vendor/eigen/Eigen/src/Core/ArrayBase.h +3 -3
- data/vendor/eigen/Eigen/src/Core/ArrayWrapper.h +21 -21
- data/vendor/eigen/Eigen/src/Core/Assign.h +1 -1
- data/vendor/eigen/Eigen/src/Core/AssignEvaluator.h +125 -50
- data/vendor/eigen/Eigen/src/Core/Assign_MKL.h +10 -10
- data/vendor/eigen/Eigen/src/Core/BandMatrix.h +16 -16
- data/vendor/eigen/Eigen/src/Core/Block.h +56 -60
- data/vendor/eigen/Eigen/src/Core/BooleanRedux.h +29 -31
- data/vendor/eigen/Eigen/src/Core/CommaInitializer.h +7 -3
- data/vendor/eigen/Eigen/src/Core/CoreEvaluators.h +325 -272
- data/vendor/eigen/Eigen/src/Core/CoreIterators.h +5 -0
- data/vendor/eigen/Eigen/src/Core/CwiseBinaryOp.h +21 -22
- data/vendor/eigen/Eigen/src/Core/CwiseNullaryOp.h +153 -18
- data/vendor/eigen/Eigen/src/Core/CwiseUnaryOp.h +6 -6
- data/vendor/eigen/Eigen/src/Core/CwiseUnaryView.h +14 -10
- data/vendor/eigen/Eigen/src/Core/DenseBase.h +132 -42
- data/vendor/eigen/Eigen/src/Core/DenseCoeffsBase.h +25 -21
- data/vendor/eigen/Eigen/src/Core/DenseStorage.h +153 -71
- data/vendor/eigen/Eigen/src/Core/Diagonal.h +21 -23
- data/vendor/eigen/Eigen/src/Core/DiagonalMatrix.h +50 -2
- data/vendor/eigen/Eigen/src/Core/DiagonalProduct.h +1 -1
- data/vendor/eigen/Eigen/src/Core/Dot.h +10 -10
- data/vendor/eigen/Eigen/src/Core/EigenBase.h +10 -9
- data/vendor/eigen/Eigen/src/Core/ForceAlignedAccess.h +8 -4
- data/vendor/eigen/Eigen/src/Core/Fuzzy.h +3 -3
- data/vendor/eigen/Eigen/src/Core/GeneralProduct.h +20 -10
- data/vendor/eigen/Eigen/src/Core/GenericPacketMath.h +599 -152
- data/vendor/eigen/Eigen/src/Core/GlobalFunctions.h +40 -33
- data/vendor/eigen/Eigen/src/Core/IO.h +40 -7
- data/vendor/eigen/Eigen/src/Core/IndexedView.h +237 -0
- data/vendor/eigen/Eigen/src/Core/Inverse.h +9 -10
- data/vendor/eigen/Eigen/src/Core/Map.h +7 -7
- data/vendor/eigen/Eigen/src/Core/MapBase.h +10 -3
- data/vendor/eigen/Eigen/src/Core/MathFunctions.h +767 -125
- data/vendor/eigen/Eigen/src/Core/MathFunctionsImpl.h +118 -19
- data/vendor/eigen/Eigen/src/Core/Matrix.h +131 -25
- data/vendor/eigen/Eigen/src/Core/MatrixBase.h +21 -3
- data/vendor/eigen/Eigen/src/Core/NestByValue.h +25 -50
- data/vendor/eigen/Eigen/src/Core/NoAlias.h +4 -3
- data/vendor/eigen/Eigen/src/Core/NumTraits.h +107 -20
- data/vendor/eigen/Eigen/src/Core/PartialReduxEvaluator.h +232 -0
- data/vendor/eigen/Eigen/src/Core/PermutationMatrix.h +3 -31
- data/vendor/eigen/Eigen/src/Core/PlainObjectBase.h +152 -59
- data/vendor/eigen/Eigen/src/Core/Product.h +30 -25
- data/vendor/eigen/Eigen/src/Core/ProductEvaluators.h +192 -125
- data/vendor/eigen/Eigen/src/Core/Random.h +37 -1
- data/vendor/eigen/Eigen/src/Core/Redux.h +180 -170
- data/vendor/eigen/Eigen/src/Core/Ref.h +121 -23
- data/vendor/eigen/Eigen/src/Core/Replicate.h +8 -8
- data/vendor/eigen/Eigen/src/Core/Reshaped.h +454 -0
- data/vendor/eigen/Eigen/src/Core/ReturnByValue.h +7 -5
- data/vendor/eigen/Eigen/src/Core/Reverse.h +18 -12
- data/vendor/eigen/Eigen/src/Core/Select.h +8 -6
- data/vendor/eigen/Eigen/src/Core/SelfAdjointView.h +33 -20
- data/vendor/eigen/Eigen/src/Core/Solve.h +14 -14
- data/vendor/eigen/Eigen/src/Core/SolveTriangular.h +16 -16
- data/vendor/eigen/Eigen/src/Core/SolverBase.h +41 -3
- data/vendor/eigen/Eigen/src/Core/StableNorm.h +100 -70
- data/vendor/eigen/Eigen/src/Core/StlIterators.h +463 -0
- data/vendor/eigen/Eigen/src/Core/Stride.h +9 -4
- data/vendor/eigen/Eigen/src/Core/Swap.h +5 -4
- data/vendor/eigen/Eigen/src/Core/Transpose.h +88 -27
- data/vendor/eigen/Eigen/src/Core/Transpositions.h +26 -47
- data/vendor/eigen/Eigen/src/Core/TriangularMatrix.h +93 -75
- data/vendor/eigen/Eigen/src/Core/VectorBlock.h +5 -5
- data/vendor/eigen/Eigen/src/Core/VectorwiseOp.h +159 -70
- data/vendor/eigen/Eigen/src/Core/Visitor.h +137 -29
- data/vendor/eigen/Eigen/src/Core/arch/AVX/Complex.h +50 -129
- data/vendor/eigen/Eigen/src/Core/arch/AVX/MathFunctions.h +126 -337
- data/vendor/eigen/Eigen/src/Core/arch/AVX/PacketMath.h +1092 -155
- data/vendor/eigen/Eigen/src/Core/arch/AVX/TypeCasting.h +65 -1
- data/vendor/eigen/Eigen/src/Core/arch/AVX512/Complex.h +422 -0
- data/vendor/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h +207 -236
- data/vendor/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h +1482 -495
- data/vendor/eigen/Eigen/src/Core/arch/AVX512/TypeCasting.h +89 -0
- data/vendor/eigen/Eigen/src/Core/arch/AltiVec/Complex.h +152 -165
- data/vendor/eigen/Eigen/src/Core/arch/AltiVec/MathFunctions.h +19 -251
- data/vendor/eigen/Eigen/src/Core/arch/AltiVec/MatrixProduct.h +2937 -0
- data/vendor/eigen/Eigen/src/Core/arch/AltiVec/MatrixProductCommon.h +221 -0
- data/vendor/eigen/Eigen/src/Core/arch/AltiVec/MatrixProductMMA.h +629 -0
- data/vendor/eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h +2042 -392
- data/vendor/eigen/Eigen/src/Core/arch/CUDA/Complex.h +235 -80
- data/vendor/eigen/Eigen/src/Core/arch/Default/BFloat16.h +700 -0
- data/vendor/eigen/Eigen/src/Core/arch/Default/ConjHelper.h +102 -14
- data/vendor/eigen/Eigen/src/Core/arch/Default/GenericPacketMathFunctions.h +1649 -0
- data/vendor/eigen/Eigen/src/Core/arch/Default/GenericPacketMathFunctionsFwd.h +110 -0
- data/vendor/eigen/Eigen/src/Core/arch/Default/Half.h +942 -0
- data/vendor/eigen/Eigen/src/Core/arch/Default/Settings.h +1 -1
- data/vendor/eigen/Eigen/src/Core/arch/Default/TypeCasting.h +120 -0
- data/vendor/eigen/Eigen/src/Core/arch/{CUDA → GPU}/MathFunctions.h +16 -4
- data/vendor/eigen/Eigen/src/Core/arch/GPU/PacketMath.h +1685 -0
- data/vendor/eigen/Eigen/src/Core/arch/GPU/TypeCasting.h +80 -0
- data/vendor/eigen/Eigen/src/Core/arch/HIP/hcc/math_constants.h +23 -0
- data/vendor/eigen/Eigen/src/Core/arch/MSA/Complex.h +648 -0
- data/vendor/eigen/Eigen/src/Core/arch/MSA/MathFunctions.h +387 -0
- data/vendor/eigen/Eigen/src/Core/arch/MSA/PacketMath.h +1233 -0
- data/vendor/eigen/Eigen/src/Core/arch/NEON/Complex.h +313 -219
- data/vendor/eigen/Eigen/src/Core/arch/NEON/GeneralBlockPanelKernel.h +183 -0
- data/vendor/eigen/Eigen/src/Core/arch/NEON/MathFunctions.h +54 -70
- data/vendor/eigen/Eigen/src/Core/arch/NEON/PacketMath.h +4376 -549
- data/vendor/eigen/Eigen/src/Core/arch/NEON/TypeCasting.h +1419 -0
- data/vendor/eigen/Eigen/src/Core/arch/SSE/Complex.h +59 -179
- data/vendor/eigen/Eigen/src/Core/arch/SSE/MathFunctions.h +65 -428
- data/vendor/eigen/Eigen/src/Core/arch/SSE/PacketMath.h +893 -283
- data/vendor/eigen/Eigen/src/Core/arch/SSE/TypeCasting.h +65 -0
- data/vendor/eigen/Eigen/src/Core/arch/SVE/MathFunctions.h +44 -0
- data/vendor/eigen/Eigen/src/Core/arch/SVE/PacketMath.h +752 -0
- data/vendor/eigen/Eigen/src/Core/arch/SVE/TypeCasting.h +49 -0
- data/vendor/eigen/Eigen/src/Core/arch/SYCL/InteropHeaders.h +232 -0
- data/vendor/eigen/Eigen/src/Core/arch/SYCL/MathFunctions.h +301 -0
- data/vendor/eigen/Eigen/src/Core/arch/SYCL/PacketMath.h +670 -0
- data/vendor/eigen/Eigen/src/Core/arch/SYCL/SyclMemoryModel.h +694 -0
- data/vendor/eigen/Eigen/src/Core/arch/SYCL/TypeCasting.h +85 -0
- data/vendor/eigen/Eigen/src/Core/arch/ZVector/Complex.h +212 -183
- data/vendor/eigen/Eigen/src/Core/arch/ZVector/MathFunctions.h +101 -5
- data/vendor/eigen/Eigen/src/Core/arch/ZVector/PacketMath.h +510 -395
- data/vendor/eigen/Eigen/src/Core/functors/AssignmentFunctors.h +11 -2
- data/vendor/eigen/Eigen/src/Core/functors/BinaryFunctors.h +112 -46
- data/vendor/eigen/Eigen/src/Core/functors/NullaryFunctors.h +31 -30
- data/vendor/eigen/Eigen/src/Core/functors/StlFunctors.h +32 -2
- data/vendor/eigen/Eigen/src/Core/functors/UnaryFunctors.h +355 -16
- data/vendor/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h +1075 -586
- data/vendor/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h +49 -24
- data/vendor/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h +41 -35
- data/vendor/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h +6 -6
- data/vendor/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h +4 -2
- data/vendor/eigen/Eigen/src/Core/products/GeneralMatrixVector.h +382 -483
- data/vendor/eigen/Eigen/src/Core/products/Parallelizer.h +22 -5
- data/vendor/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h +53 -30
- data/vendor/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h +16 -8
- data/vendor/eigen/Eigen/src/Core/products/SelfadjointMatrixVector.h +8 -6
- data/vendor/eigen/Eigen/src/Core/products/SelfadjointProduct.h +4 -4
- data/vendor/eigen/Eigen/src/Core/products/SelfadjointRank2Update.h +5 -4
- data/vendor/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h +33 -27
- data/vendor/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h +14 -12
- data/vendor/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h +36 -34
- data/vendor/eigen/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h +8 -4
- data/vendor/eigen/Eigen/src/Core/products/TriangularSolverVector.h +13 -10
- data/vendor/eigen/Eigen/src/Core/util/BlasUtil.h +304 -119
- data/vendor/eigen/Eigen/src/Core/util/ConfigureVectorization.h +512 -0
- data/vendor/eigen/Eigen/src/Core/util/Constants.h +25 -9
- data/vendor/eigen/Eigen/src/Core/util/DisableStupidWarnings.h +26 -3
- data/vendor/eigen/Eigen/src/Core/util/ForwardDeclarations.h +29 -9
- data/vendor/eigen/Eigen/src/Core/util/IndexedViewHelper.h +186 -0
- data/vendor/eigen/Eigen/src/Core/util/IntegralConstant.h +272 -0
- data/vendor/eigen/Eigen/src/Core/util/MKL_support.h +8 -1
- data/vendor/eigen/Eigen/src/Core/util/Macros.h +709 -246
- data/vendor/eigen/Eigen/src/Core/util/Memory.h +222 -52
- data/vendor/eigen/Eigen/src/Core/util/Meta.h +355 -77
- data/vendor/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h +5 -1
- data/vendor/eigen/Eigen/src/Core/util/ReshapedHelper.h +51 -0
- data/vendor/eigen/Eigen/src/Core/util/StaticAssert.h +8 -5
- data/vendor/eigen/Eigen/src/Core/util/SymbolicIndex.h +293 -0
- data/vendor/eigen/Eigen/src/Core/util/XprHelper.h +65 -30
- data/vendor/eigen/Eigen/src/Eigenvalues/ComplexEigenSolver.h +1 -1
- data/vendor/eigen/Eigen/src/Eigenvalues/ComplexSchur.h +7 -4
- data/vendor/eigen/Eigen/src/Eigenvalues/EigenSolver.h +2 -2
- data/vendor/eigen/Eigen/src/Eigenvalues/GeneralizedSelfAdjointEigenSolver.h +1 -1
- data/vendor/eigen/Eigen/src/Eigenvalues/HessenbergDecomposition.h +2 -2
- data/vendor/eigen/Eigen/src/Eigenvalues/MatrixBaseEigenvalues.h +2 -2
- data/vendor/eigen/Eigen/src/Eigenvalues/RealQZ.h +9 -6
- data/vendor/eigen/Eigen/src/Eigenvalues/RealSchur.h +21 -9
- data/vendor/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h +77 -43
- data/vendor/eigen/Eigen/src/Eigenvalues/Tridiagonalization.h +20 -15
- data/vendor/eigen/Eigen/src/Geometry/AlignedBox.h +99 -5
- data/vendor/eigen/Eigen/src/Geometry/AngleAxis.h +4 -4
- data/vendor/eigen/Eigen/src/Geometry/EulerAngles.h +3 -3
- data/vendor/eigen/Eigen/src/Geometry/Homogeneous.h +15 -11
- data/vendor/eigen/Eigen/src/Geometry/Hyperplane.h +1 -1
- data/vendor/eigen/Eigen/src/Geometry/OrthoMethods.h +3 -2
- data/vendor/eigen/Eigen/src/Geometry/ParametrizedLine.h +39 -2
- data/vendor/eigen/Eigen/src/Geometry/Quaternion.h +70 -14
- data/vendor/eigen/Eigen/src/Geometry/Rotation2D.h +3 -3
- data/vendor/eigen/Eigen/src/Geometry/Scaling.h +23 -5
- data/vendor/eigen/Eigen/src/Geometry/Transform.h +88 -67
- data/vendor/eigen/Eigen/src/Geometry/Translation.h +6 -12
- data/vendor/eigen/Eigen/src/Geometry/Umeyama.h +1 -1
- data/vendor/eigen/Eigen/src/Geometry/arch/Geometry_SIMD.h +168 -0
- data/vendor/eigen/Eigen/src/Householder/BlockHouseholder.h +9 -2
- data/vendor/eigen/Eigen/src/Householder/Householder.h +8 -4
- data/vendor/eigen/Eigen/src/Householder/HouseholderSequence.h +123 -48
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h +15 -15
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/BiCGSTAB.h +7 -23
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h +5 -22
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/IncompleteCholesky.h +41 -47
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/IncompleteLUT.h +51 -60
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/IterativeSolverBase.h +70 -20
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/LeastSquareConjugateGradient.h +2 -20
- data/vendor/eigen/Eigen/src/IterativeLinearSolvers/SolveWithGuess.h +11 -9
- data/vendor/eigen/Eigen/src/Jacobi/Jacobi.h +31 -10
- data/vendor/eigen/Eigen/src/KLUSupport/KLUSupport.h +358 -0
- data/vendor/eigen/Eigen/src/LU/Determinant.h +35 -19
- data/vendor/eigen/Eigen/src/LU/FullPivLU.h +29 -43
- data/vendor/eigen/Eigen/src/LU/InverseImpl.h +25 -8
- data/vendor/eigen/Eigen/src/LU/PartialPivLU.h +71 -58
- data/vendor/eigen/Eigen/src/LU/arch/InverseSize4.h +351 -0
- data/vendor/eigen/Eigen/src/OrderingMethods/Amd.h +7 -17
- data/vendor/eigen/Eigen/src/OrderingMethods/Eigen_Colamd.h +297 -277
- data/vendor/eigen/Eigen/src/OrderingMethods/Ordering.h +6 -10
- data/vendor/eigen/Eigen/src/PaStiXSupport/PaStiXSupport.h +1 -1
- data/vendor/eigen/Eigen/src/PardisoSupport/PardisoSupport.h +9 -7
- data/vendor/eigen/Eigen/src/QR/ColPivHouseholderQR.h +41 -20
- data/vendor/eigen/Eigen/src/QR/CompleteOrthogonalDecomposition.h +100 -27
- data/vendor/eigen/Eigen/src/QR/FullPivHouseholderQR.h +59 -22
- data/vendor/eigen/Eigen/src/QR/HouseholderQR.h +48 -23
- data/vendor/eigen/Eigen/src/SPQRSupport/SuiteSparseQRSupport.h +25 -3
- data/vendor/eigen/Eigen/src/SVD/BDCSVD.h +183 -63
- data/vendor/eigen/Eigen/src/SVD/JacobiSVD.h +22 -14
- data/vendor/eigen/Eigen/src/SVD/SVDBase.h +83 -22
- data/vendor/eigen/Eigen/src/SVD/UpperBidiagonalization.h +3 -3
- data/vendor/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h +17 -9
- data/vendor/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h +12 -37
- data/vendor/eigen/Eigen/src/SparseCore/AmbiVector.h +3 -2
- data/vendor/eigen/Eigen/src/SparseCore/CompressedStorage.h +16 -0
- data/vendor/eigen/Eigen/src/SparseCore/ConservativeSparseSparseProduct.h +6 -6
- data/vendor/eigen/Eigen/src/SparseCore/SparseAssign.h +81 -27
- data/vendor/eigen/Eigen/src/SparseCore/SparseBlock.h +25 -57
- data/vendor/eigen/Eigen/src/SparseCore/SparseCompressedBase.h +40 -11
- data/vendor/eigen/Eigen/src/SparseCore/SparseCwiseBinaryOp.h +11 -15
- data/vendor/eigen/Eigen/src/SparseCore/SparseCwiseUnaryOp.h +4 -2
- data/vendor/eigen/Eigen/src/SparseCore/SparseDenseProduct.h +30 -8
- data/vendor/eigen/Eigen/src/SparseCore/SparseMatrix.h +126 -11
- data/vendor/eigen/Eigen/src/SparseCore/SparseMatrixBase.h +5 -12
- data/vendor/eigen/Eigen/src/SparseCore/SparseProduct.h +13 -1
- data/vendor/eigen/Eigen/src/SparseCore/SparseRef.h +7 -7
- data/vendor/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h +5 -2
- data/vendor/eigen/Eigen/src/SparseCore/SparseUtil.h +8 -0
- data/vendor/eigen/Eigen/src/SparseCore/SparseVector.h +1 -1
- data/vendor/eigen/Eigen/src/SparseCore/SparseView.h +1 -0
- data/vendor/eigen/Eigen/src/SparseLU/SparseLU.h +162 -12
- data/vendor/eigen/Eigen/src/SparseLU/SparseLU_Memory.h +1 -1
- data/vendor/eigen/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h +76 -2
- data/vendor/eigen/Eigen/src/SparseLU/SparseLU_column_dfs.h +2 -2
- data/vendor/eigen/Eigen/src/SparseLU/SparseLU_gemm_kernel.h +1 -1
- data/vendor/eigen/Eigen/src/SparseLU/SparseLU_panel_bmod.h +1 -1
- data/vendor/eigen/Eigen/src/SparseQR/SparseQR.h +19 -6
- data/vendor/eigen/Eigen/src/StlSupport/StdDeque.h +2 -12
- data/vendor/eigen/Eigen/src/StlSupport/StdList.h +2 -2
- data/vendor/eigen/Eigen/src/StlSupport/StdVector.h +2 -2
- data/vendor/eigen/Eigen/src/SuperLUSupport/SuperLUSupport.h +6 -8
- data/vendor/eigen/Eigen/src/UmfPackSupport/UmfPackSupport.h +175 -39
- data/vendor/eigen/Eigen/src/misc/lapacke.h +5 -4
- data/vendor/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h +28 -2
- data/vendor/eigen/Eigen/src/plugins/ArrayCwiseUnaryOps.h +155 -11
- data/vendor/eigen/Eigen/src/plugins/BlockMethods.h +626 -242
- data/vendor/eigen/Eigen/src/plugins/CommonCwiseUnaryOps.h +14 -0
- data/vendor/eigen/Eigen/src/plugins/IndexedViewMethods.h +262 -0
- data/vendor/eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h +4 -4
- data/vendor/eigen/Eigen/src/plugins/MatrixCwiseUnaryOps.h +10 -0
- data/vendor/eigen/Eigen/src/plugins/ReshapedMethods.h +149 -0
- data/vendor/eigen/README.md +2 -0
- data/vendor/eigen/bench/btl/README +1 -1
- data/vendor/eigen/bench/tensors/README +6 -7
- data/vendor/eigen/ci/README.md +56 -0
- data/vendor/eigen/demos/mix_eigen_and_c/README +1 -1
- data/vendor/eigen/unsupported/Eigen/CXX11/src/Tensor/README.md +213 -158
- data/vendor/eigen/unsupported/README.txt +1 -1
- data/vendor/tomotopy/README.kr.rst +21 -0
- data/vendor/tomotopy/README.rst +20 -0
- data/vendor/tomotopy/src/Labeling/FoRelevance.cpp +2 -2
- data/vendor/tomotopy/src/Labeling/Phraser.hpp +1 -1
- data/vendor/tomotopy/src/TopicModel/CTModel.hpp +2 -1
- data/vendor/tomotopy/src/TopicModel/DMRModel.hpp +2 -1
- data/vendor/tomotopy/src/TopicModel/DTModel.hpp +1 -1
- data/vendor/tomotopy/src/TopicModel/GDMRModel.hpp +2 -2
- data/vendor/tomotopy/src/TopicModel/HDP.h +1 -0
- data/vendor/tomotopy/src/TopicModel/HDPModel.hpp +53 -2
- data/vendor/tomotopy/src/TopicModel/HLDAModel.hpp +1 -1
- data/vendor/tomotopy/src/TopicModel/HPAModel.hpp +1 -0
- data/vendor/tomotopy/src/TopicModel/LDACVB0Model.hpp +2 -2
- data/vendor/tomotopy/src/TopicModel/LDAModel.hpp +16 -5
- data/vendor/tomotopy/src/TopicModel/LLDAModel.hpp +1 -0
- data/vendor/tomotopy/src/TopicModel/MGLDAModel.hpp +1 -0
- data/vendor/tomotopy/src/TopicModel/PLDAModel.hpp +1 -0
- data/vendor/tomotopy/src/TopicModel/PT.h +3 -1
- data/vendor/tomotopy/src/TopicModel/PTModel.hpp +31 -1
- data/vendor/tomotopy/src/TopicModel/SLDAModel.hpp +2 -2
- data/vendor/tomotopy/src/TopicModel/TopicModel.hpp +7 -5
- data/vendor/tomotopy/src/Utils/EigenAddonOps.hpp +36 -1
- data/vendor/tomotopy/src/Utils/exception.h +6 -0
- data/vendor/tomotopy/src/Utils/sample.hpp +14 -12
- data/vendor/tomotopy/src/Utils/sse_gamma.h +0 -3
- metadata +60 -14
- data/vendor/eigen/Eigen/CMakeLists.txt +0 -19
- data/vendor/eigen/Eigen/src/Core/arch/CUDA/Half.h +0 -674
- data/vendor/eigen/Eigen/src/Core/arch/CUDA/PacketMath.h +0 -333
- data/vendor/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h +0 -1124
- data/vendor/eigen/Eigen/src/Core/arch/CUDA/TypeCasting.h +0 -212
- data/vendor/eigen/Eigen/src/Geometry/arch/Geometry_SSE.h +0 -161
- data/vendor/eigen/Eigen/src/LU/arch/Inverse_SSE.h +0 -338
@@ -11,19 +11,56 @@
|
|
11
11
|
#ifndef EIGEN_MACROS_H
|
12
12
|
#define EIGEN_MACROS_H
|
13
13
|
|
14
|
+
//------------------------------------------------------------------------------------------
|
15
|
+
// Eigen version and basic defaults
|
16
|
+
//------------------------------------------------------------------------------------------
|
17
|
+
|
14
18
|
#define EIGEN_WORLD_VERSION 3
|
15
|
-
#define EIGEN_MAJOR_VERSION
|
16
|
-
#define EIGEN_MINOR_VERSION
|
19
|
+
#define EIGEN_MAJOR_VERSION 4
|
20
|
+
#define EIGEN_MINOR_VERSION 0
|
17
21
|
|
18
22
|
#define EIGEN_VERSION_AT_LEAST(x,y,z) (EIGEN_WORLD_VERSION>x || (EIGEN_WORLD_VERSION>=x && \
|
19
23
|
(EIGEN_MAJOR_VERSION>y || (EIGEN_MAJOR_VERSION>=y && \
|
20
24
|
EIGEN_MINOR_VERSION>=z))))
|
21
25
|
|
26
|
+
#ifdef EIGEN_DEFAULT_TO_ROW_MAJOR
|
27
|
+
#define EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION Eigen::RowMajor
|
28
|
+
#else
|
29
|
+
#define EIGEN_DEFAULT_MATRIX_STORAGE_ORDER_OPTION Eigen::ColMajor
|
30
|
+
#endif
|
31
|
+
|
32
|
+
#ifndef EIGEN_DEFAULT_DENSE_INDEX_TYPE
|
33
|
+
#define EIGEN_DEFAULT_DENSE_INDEX_TYPE std::ptrdiff_t
|
34
|
+
#endif
|
35
|
+
|
36
|
+
// Upperbound on the C++ version to use.
|
37
|
+
// Expected values are 03, 11, 14, 17, etc.
|
38
|
+
// By default, let's use an arbitrarily large C++ version.
|
39
|
+
#ifndef EIGEN_MAX_CPP_VER
|
40
|
+
#define EIGEN_MAX_CPP_VER 99
|
41
|
+
#endif
|
42
|
+
|
43
|
+
/** Allows to disable some optimizations which might affect the accuracy of the result.
|
44
|
+
* Such optimization are enabled by default, and set EIGEN_FAST_MATH to 0 to disable them.
|
45
|
+
* They currently include:
|
46
|
+
* - single precision ArrayBase::sin() and ArrayBase::cos() for SSE and AVX vectorization.
|
47
|
+
*/
|
48
|
+
#ifndef EIGEN_FAST_MATH
|
49
|
+
#define EIGEN_FAST_MATH 1
|
50
|
+
#endif
|
51
|
+
|
52
|
+
#ifndef EIGEN_STACK_ALLOCATION_LIMIT
|
53
|
+
// 131072 == 128 KB
|
54
|
+
#define EIGEN_STACK_ALLOCATION_LIMIT 131072
|
55
|
+
#endif
|
56
|
+
|
57
|
+
//------------------------------------------------------------------------------------------
|
22
58
|
// Compiler identification, EIGEN_COMP_*
|
59
|
+
//------------------------------------------------------------------------------------------
|
23
60
|
|
24
61
|
/// \internal EIGEN_COMP_GNUC set to 1 for all compilers compatible with GCC
|
25
62
|
#ifdef __GNUC__
|
26
|
-
#define EIGEN_COMP_GNUC
|
63
|
+
#define EIGEN_COMP_GNUC (__GNUC__*10+__GNUC_MINOR__)
|
27
64
|
#else
|
28
65
|
#define EIGEN_COMP_GNUC 0
|
29
66
|
#endif
|
@@ -35,6 +72,12 @@
|
|
35
72
|
#define EIGEN_COMP_CLANG 0
|
36
73
|
#endif
|
37
74
|
|
75
|
+
/// \internal EIGEN_COMP_CASTXML set to 1 if being preprocessed by CastXML
|
76
|
+
#if defined(__castxml__)
|
77
|
+
#define EIGEN_COMP_CASTXML 1
|
78
|
+
#else
|
79
|
+
#define EIGEN_COMP_CASTXML 0
|
80
|
+
#endif
|
38
81
|
|
39
82
|
/// \internal EIGEN_COMP_LLVM set to 1 if the compiler backend is llvm
|
40
83
|
#if defined(__llvm__)
|
@@ -71,14 +114,44 @@
|
|
71
114
|
#define EIGEN_COMP_MSVC 0
|
72
115
|
#endif
|
73
116
|
|
117
|
+
#if defined(__NVCC__)
|
118
|
+
#if defined(__CUDACC_VER_MAJOR__) && (__CUDACC_VER_MAJOR__ >= 9)
|
119
|
+
#define EIGEN_COMP_NVCC ((__CUDACC_VER_MAJOR__ * 10000) + (__CUDACC_VER_MINOR__ * 100))
|
120
|
+
#elif defined(__CUDACC_VER__)
|
121
|
+
#define EIGEN_COMP_NVCC __CUDACC_VER__
|
122
|
+
#else
|
123
|
+
#error "NVCC did not define compiler version."
|
124
|
+
#endif
|
125
|
+
#else
|
126
|
+
#define EIGEN_COMP_NVCC 0
|
127
|
+
#endif
|
128
|
+
|
74
129
|
// For the record, here is a table summarizing the possible values for EIGEN_COMP_MSVC:
|
75
|
-
// name
|
76
|
-
// 2008
|
77
|
-
// 2010
|
78
|
-
// 2012
|
79
|
-
// 2013
|
80
|
-
// 2015
|
81
|
-
// "15"
|
130
|
+
// name ver MSC_VER
|
131
|
+
// 2008 9 1500
|
132
|
+
// 2010 10 1600
|
133
|
+
// 2012 11 1700
|
134
|
+
// 2013 12 1800
|
135
|
+
// 2015 14 1900
|
136
|
+
// "15" 15 1900
|
137
|
+
// 2017-14.1 15.0 1910
|
138
|
+
// 2017-14.11 15.3 1911
|
139
|
+
// 2017-14.12 15.5 1912
|
140
|
+
// 2017-14.13 15.6 1913
|
141
|
+
// 2017-14.14 15.7 1914
|
142
|
+
|
143
|
+
/// \internal EIGEN_COMP_MSVC_LANG set to _MSVC_LANG if the compiler is Microsoft Visual C++, 0 otherwise.
|
144
|
+
#if defined(_MSVC_LANG)
|
145
|
+
#define EIGEN_COMP_MSVC_LANG _MSVC_LANG
|
146
|
+
#else
|
147
|
+
#define EIGEN_COMP_MSVC_LANG 0
|
148
|
+
#endif
|
149
|
+
|
150
|
+
// For the record, here is a table summarizing the possible values for EIGEN_COMP_MSVC_LANG:
|
151
|
+
// MSVC option Standard MSVC_LANG
|
152
|
+
// /std:c++14 (default as of VS 2019) C++14 201402L
|
153
|
+
// /std:c++17 C++17 201703L
|
154
|
+
// /std:c++latest >C++17 >201703L
|
82
155
|
|
83
156
|
/// \internal EIGEN_COMP_MSVC_STRICT set to 1 if the compiler is really Microsoft Visual C++ and not ,e.g., ICC or clang-cl
|
84
157
|
#if EIGEN_COMP_MSVC && !(EIGEN_COMP_ICC || EIGEN_COMP_LLVM || EIGEN_COMP_CLANG)
|
@@ -87,16 +160,21 @@
|
|
87
160
|
#define EIGEN_COMP_MSVC_STRICT 0
|
88
161
|
#endif
|
89
162
|
|
90
|
-
/// \internal EIGEN_COMP_IBM set to
|
91
|
-
|
92
|
-
|
163
|
+
/// \internal EIGEN_COMP_IBM set to xlc version if the compiler is IBM XL C++
|
164
|
+
// XLC version
|
165
|
+
// 3.1 0x0301
|
166
|
+
// 4.5 0x0405
|
167
|
+
// 5.0 0x0500
|
168
|
+
// 12.1 0x0C01
|
169
|
+
#if defined(__IBMCPP__) || defined(__xlc__) || defined(__ibmxl__)
|
170
|
+
#define EIGEN_COMP_IBM __xlC__
|
93
171
|
#else
|
94
172
|
#define EIGEN_COMP_IBM 0
|
95
173
|
#endif
|
96
174
|
|
97
|
-
/// \internal EIGEN_COMP_PGI set to
|
175
|
+
/// \internal EIGEN_COMP_PGI set to PGI version if the compiler is Portland Group Compiler
|
98
176
|
#if defined(__PGI)
|
99
|
-
#define EIGEN_COMP_PGI
|
177
|
+
#define EIGEN_COMP_PGI (__PGIC__*100+__PGIC_MINOR__)
|
100
178
|
#else
|
101
179
|
#define EIGEN_COMP_PGI 0
|
102
180
|
#endif
|
@@ -108,7 +186,7 @@
|
|
108
186
|
#define EIGEN_COMP_ARM 0
|
109
187
|
#endif
|
110
188
|
|
111
|
-
/// \internal
|
189
|
+
/// \internal EIGEN_COMP_EMSCRIPTEN set to 1 if the compiler is Emscripten Compiler
|
112
190
|
#if defined(__EMSCRIPTEN__)
|
113
191
|
#define EIGEN_COMP_EMSCRIPTEN 1
|
114
192
|
#else
|
@@ -142,9 +220,13 @@
|
|
142
220
|
#endif
|
143
221
|
|
144
222
|
|
223
|
+
|
224
|
+
//------------------------------------------------------------------------------------------
|
145
225
|
// Architecture identification, EIGEN_ARCH_*
|
226
|
+
//------------------------------------------------------------------------------------------
|
227
|
+
|
146
228
|
|
147
|
-
#if defined(__x86_64__) || defined(_M_X64) || defined(__amd64)
|
229
|
+
#if defined(__x86_64__) || (defined(_M_X64) && !defined(_M_ARM64EC)) || defined(__amd64)
|
148
230
|
#define EIGEN_ARCH_x86_64 1
|
149
231
|
#else
|
150
232
|
#define EIGEN_ARCH_x86_64 0
|
@@ -170,18 +252,61 @@
|
|
170
252
|
#endif
|
171
253
|
|
172
254
|
/// \internal EIGEN_ARCH_ARM64 set to 1 if the architecture is ARM64
|
173
|
-
#if defined(__aarch64__)
|
255
|
+
#if defined(__aarch64__) || defined(_M_ARM64) || defined(_M_ARM64EC)
|
174
256
|
#define EIGEN_ARCH_ARM64 1
|
175
257
|
#else
|
176
258
|
#define EIGEN_ARCH_ARM64 0
|
177
259
|
#endif
|
178
260
|
|
261
|
+
/// \internal EIGEN_ARCH_ARM_OR_ARM64 set to 1 if the architecture is ARM or ARM64
|
179
262
|
#if EIGEN_ARCH_ARM || EIGEN_ARCH_ARM64
|
180
263
|
#define EIGEN_ARCH_ARM_OR_ARM64 1
|
181
264
|
#else
|
182
265
|
#define EIGEN_ARCH_ARM_OR_ARM64 0
|
183
266
|
#endif
|
184
267
|
|
268
|
+
/// \internal EIGEN_ARCH_ARMV8 set to 1 if the architecture is armv8 or greater.
|
269
|
+
#if EIGEN_ARCH_ARM_OR_ARM64 && defined(__ARM_ARCH) && __ARM_ARCH >= 8
|
270
|
+
#define EIGEN_ARCH_ARMV8 1
|
271
|
+
#else
|
272
|
+
#define EIGEN_ARCH_ARMV8 0
|
273
|
+
#endif
|
274
|
+
|
275
|
+
|
276
|
+
/// \internal EIGEN_HAS_ARM64_FP16 set to 1 if the architecture provides an IEEE
|
277
|
+
/// compliant Arm fp16 type
|
278
|
+
#if EIGEN_ARCH_ARM64
|
279
|
+
#ifndef EIGEN_HAS_ARM64_FP16
|
280
|
+
#if defined(__ARM_FP16_FORMAT_IEEE)
|
281
|
+
#define EIGEN_HAS_ARM64_FP16 1
|
282
|
+
#else
|
283
|
+
#define EIGEN_HAS_ARM64_FP16 0
|
284
|
+
#endif
|
285
|
+
#endif
|
286
|
+
#endif
|
287
|
+
|
288
|
+
/// \internal EIGEN_HAS_ARM64_FP16_VECTOR_ARITHMETIC set to 1 if the architecture
|
289
|
+
/// supports Neon vector intrinsics for fp16.
|
290
|
+
#if EIGEN_ARCH_ARM64
|
291
|
+
#ifndef EIGEN_HAS_ARM64_FP16_VECTOR_ARITHMETIC
|
292
|
+
#if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC)
|
293
|
+
#define EIGEN_HAS_ARM64_FP16_VECTOR_ARITHMETIC 1
|
294
|
+
#else
|
295
|
+
#define EIGEN_HAS_ARM64_FP16_VECTOR_ARITHMETIC 0
|
296
|
+
#endif
|
297
|
+
#endif
|
298
|
+
#endif
|
299
|
+
|
300
|
+
/// \internal EIGEN_HAS_ARM64_FP16_SCALAR_ARITHMETIC set to 1 if the architecture
|
301
|
+
/// supports Neon scalar intrinsics for fp16.
|
302
|
+
#if EIGEN_ARCH_ARM64
|
303
|
+
#ifndef EIGEN_HAS_ARM64_FP16_SCALAR_ARITHMETIC
|
304
|
+
#if defined(__ARM_FEATURE_FP16_SCALAR_ARITHMETIC)
|
305
|
+
#define EIGEN_HAS_ARM64_FP16_SCALAR_ARITHMETIC 1
|
306
|
+
#endif
|
307
|
+
#endif
|
308
|
+
#endif
|
309
|
+
|
185
310
|
/// \internal EIGEN_ARCH_MIPS set to 1 if the architecture is MIPS
|
186
311
|
#if defined(__mips__) || defined(__mips)
|
187
312
|
#define EIGEN_ARCH_MIPS 1
|
@@ -212,7 +337,9 @@
|
|
212
337
|
|
213
338
|
|
214
339
|
|
340
|
+
//------------------------------------------------------------------------------------------
|
215
341
|
// Operating system identification, EIGEN_OS_*
|
342
|
+
//------------------------------------------------------------------------------------------
|
216
343
|
|
217
344
|
/// \internal EIGEN_OS_UNIX set to 1 if the OS is a unix variant
|
218
345
|
#if defined(__unix__) || defined(__unix)
|
@@ -299,9 +426,17 @@
|
|
299
426
|
#define EIGEN_OS_WIN_STRICT 0
|
300
427
|
#endif
|
301
428
|
|
302
|
-
/// \internal EIGEN_OS_SUN set to
|
429
|
+
/// \internal EIGEN_OS_SUN set to __SUNPRO_C if the OS is SUN
|
430
|
+
// compiler solaris __SUNPRO_C
|
431
|
+
// version studio
|
432
|
+
// 5.7 10 0x570
|
433
|
+
// 5.8 11 0x580
|
434
|
+
// 5.9 12 0x590
|
435
|
+
// 5.10 12.1 0x5100
|
436
|
+
// 5.11 12.2 0x5110
|
437
|
+
// 5.12 12.3 0x5120
|
303
438
|
#if (defined(sun) || defined(__sun)) && !(defined(__SVR4) || defined(__svr4__))
|
304
|
-
#define EIGEN_OS_SUN
|
439
|
+
#define EIGEN_OS_SUN __SUNPRO_C
|
305
440
|
#else
|
306
441
|
#define EIGEN_OS_SUN 0
|
307
442
|
#endif
|
@@ -314,26 +449,137 @@
|
|
314
449
|
#endif
|
315
450
|
|
316
451
|
|
452
|
+
//------------------------------------------------------------------------------------------
|
453
|
+
// Detect GPU compilers and architectures
|
454
|
+
//------------------------------------------------------------------------------------------
|
317
455
|
|
318
|
-
|
319
|
-
|
320
|
-
|
321
|
-
#
|
322
|
-
#define EIGEN_SAFE_TO_USE_STANDARD_ASSERT_MACRO 1
|
456
|
+
// NVCC is not supported as the target platform for HIPCC
|
457
|
+
// Note that this also makes EIGEN_CUDACC and EIGEN_HIPCC mutually exclusive
|
458
|
+
#if defined(__NVCC__) && defined(__HIPCC__)
|
459
|
+
#error "NVCC as the target platform for HIPCC is currently not supported."
|
323
460
|
#endif
|
324
461
|
|
325
|
-
|
326
|
-
//
|
327
|
-
#define
|
462
|
+
#if defined(__CUDACC__) && !defined(EIGEN_NO_CUDA)
|
463
|
+
// Means the compiler is either nvcc or clang with CUDA enabled
|
464
|
+
#define EIGEN_CUDACC __CUDACC__
|
465
|
+
#endif
|
328
466
|
|
329
|
-
#
|
330
|
-
|
467
|
+
#if defined(__CUDA_ARCH__) && !defined(EIGEN_NO_CUDA)
|
468
|
+
// Means we are generating code for the device
|
469
|
+
#define EIGEN_CUDA_ARCH __CUDA_ARCH__
|
470
|
+
#endif
|
471
|
+
|
472
|
+
#if defined(EIGEN_CUDACC)
|
473
|
+
#include <cuda.h>
|
474
|
+
#define EIGEN_CUDA_SDK_VER (CUDA_VERSION * 10)
|
331
475
|
#else
|
332
|
-
#define
|
476
|
+
#define EIGEN_CUDA_SDK_VER 0
|
333
477
|
#endif
|
334
478
|
|
335
|
-
#
|
336
|
-
|
479
|
+
#if defined(__HIPCC__) && !defined(EIGEN_NO_HIP)
|
480
|
+
// Means the compiler is HIPCC (analogous to EIGEN_CUDACC, but for HIP)
|
481
|
+
#define EIGEN_HIPCC __HIPCC__
|
482
|
+
|
483
|
+
// We need to include hip_runtime.h here because it pulls in
|
484
|
+
// ++ hip_common.h which contains the define for __HIP_DEVICE_COMPILE__
|
485
|
+
// ++ host_defines.h which contains the defines for the __host__ and __device__ macros
|
486
|
+
#include <hip/hip_runtime.h>
|
487
|
+
|
488
|
+
#if defined(__HIP_DEVICE_COMPILE__)
|
489
|
+
// analogous to EIGEN_CUDA_ARCH, but for HIP
|
490
|
+
#define EIGEN_HIP_DEVICE_COMPILE __HIP_DEVICE_COMPILE__
|
491
|
+
#endif
|
492
|
+
|
493
|
+
// For HIP (ROCm 3.5 and higher), we need to explicitly set the launch_bounds attribute
|
494
|
+
// value to 1024. The compiler assigns a default value of 256 when the attribute is not
|
495
|
+
// specified. This results in failures on the HIP platform, for cases when a GPU kernel
|
496
|
+
// without an explicit launch_bounds attribute is called with a threads_per_block value
|
497
|
+
// greater than 256.
|
498
|
+
//
|
499
|
+
// This is a regression in functioanlity and is expected to be fixed within the next
|
500
|
+
// couple of ROCm releases (compiler will go back to using 1024 value as the default)
|
501
|
+
//
|
502
|
+
// In the meantime, we will use a "only enabled for HIP" macro to set the launch_bounds
|
503
|
+
// attribute.
|
504
|
+
|
505
|
+
#define EIGEN_HIP_LAUNCH_BOUNDS_1024 __launch_bounds__(1024)
|
506
|
+
|
507
|
+
#endif
|
508
|
+
|
509
|
+
#if !defined(EIGEN_HIP_LAUNCH_BOUNDS_1024)
|
510
|
+
#define EIGEN_HIP_LAUNCH_BOUNDS_1024
|
511
|
+
#endif // !defined(EIGEN_HIP_LAUNCH_BOUNDS_1024)
|
512
|
+
|
513
|
+
// Unify CUDA/HIPCC
|
514
|
+
|
515
|
+
#if defined(EIGEN_CUDACC) || defined(EIGEN_HIPCC)
|
516
|
+
//
|
517
|
+
// If either EIGEN_CUDACC or EIGEN_HIPCC is defined, then define EIGEN_GPUCC
|
518
|
+
//
|
519
|
+
#define EIGEN_GPUCC
|
520
|
+
//
|
521
|
+
// EIGEN_HIPCC implies the HIP compiler and is used to tweak Eigen code for use in HIP kernels
|
522
|
+
// EIGEN_CUDACC implies the CUDA compiler and is used to tweak Eigen code for use in CUDA kernels
|
523
|
+
//
|
524
|
+
// In most cases the same tweaks are required to the Eigen code to enable in both the HIP and CUDA kernels.
|
525
|
+
// For those cases, the corresponding code should be guarded with
|
526
|
+
// #if defined(EIGEN_GPUCC)
|
527
|
+
// instead of
|
528
|
+
// #if defined(EIGEN_CUDACC) || defined(EIGEN_HIPCC)
|
529
|
+
//
|
530
|
+
// For cases where the tweak is specific to HIP, the code should be guarded with
|
531
|
+
// #if defined(EIGEN_HIPCC)
|
532
|
+
//
|
533
|
+
// For cases where the tweak is specific to CUDA, the code should be guarded with
|
534
|
+
// #if defined(EIGEN_CUDACC)
|
535
|
+
//
|
536
|
+
#endif
|
537
|
+
|
538
|
+
#if defined(EIGEN_CUDA_ARCH) || defined(EIGEN_HIP_DEVICE_COMPILE)
|
539
|
+
//
|
540
|
+
// If either EIGEN_CUDA_ARCH or EIGEN_HIP_DEVICE_COMPILE is defined, then define EIGEN_GPU_COMPILE_PHASE
|
541
|
+
//
|
542
|
+
#define EIGEN_GPU_COMPILE_PHASE
|
543
|
+
//
|
544
|
+
// GPU compilers (HIPCC, NVCC) typically do two passes over the source code,
|
545
|
+
// + one to compile the source for the "host" (ie CPU)
|
546
|
+
// + another to compile the source for the "device" (ie. GPU)
|
547
|
+
//
|
548
|
+
// Code that needs to enabled only during the either the "host" or "device" compilation phase
|
549
|
+
// needs to be guarded with a macro that indicates the current compilation phase
|
550
|
+
//
|
551
|
+
// EIGEN_HIP_DEVICE_COMPILE implies the device compilation phase in HIP
|
552
|
+
// EIGEN_CUDA_ARCH implies the device compilation phase in CUDA
|
553
|
+
//
|
554
|
+
// In most cases, the "host" / "device" specific code is the same for both HIP and CUDA
|
555
|
+
// For those cases, the code should be guarded with
|
556
|
+
// #if defined(EIGEN_GPU_COMPILE_PHASE)
|
557
|
+
// instead of
|
558
|
+
// #if defined(EIGEN_CUDA_ARCH) || defined(EIGEN_HIP_DEVICE_COMPILE)
|
559
|
+
//
|
560
|
+
// For cases where the tweak is specific to HIP, the code should be guarded with
|
561
|
+
// #if defined(EIGEN_HIP_DEVICE_COMPILE)
|
562
|
+
//
|
563
|
+
// For cases where the tweak is specific to CUDA, the code should be guarded with
|
564
|
+
// #if defined(EIGEN_CUDA_ARCH)
|
565
|
+
//
|
566
|
+
#endif
|
567
|
+
|
568
|
+
#if defined(EIGEN_USE_SYCL) && defined(__SYCL_DEVICE_ONLY__)
|
569
|
+
// EIGEN_USE_SYCL is a user-defined macro while __SYCL_DEVICE_ONLY__ is a compiler-defined macro.
|
570
|
+
// In most cases we want to check if both macros are defined which can be done using the define below.
|
571
|
+
#define SYCL_DEVICE_ONLY
|
572
|
+
#endif
|
573
|
+
|
574
|
+
//------------------------------------------------------------------------------------------
|
575
|
+
// Detect Compiler/Architecture/OS specific features
|
576
|
+
//------------------------------------------------------------------------------------------
|
577
|
+
|
578
|
+
#if EIGEN_GNUC_AT_MOST(4,3) && !EIGEN_COMP_CLANG
|
579
|
+
// see bug 89
|
580
|
+
#define EIGEN_SAFE_TO_USE_STANDARD_ASSERT_MACRO 0
|
581
|
+
#else
|
582
|
+
#define EIGEN_SAFE_TO_USE_STANDARD_ASSERT_MACRO 1
|
337
583
|
#endif
|
338
584
|
|
339
585
|
// Cross compiler wrapper around LLVM's __has_builtin
|
@@ -349,26 +595,79 @@
|
|
349
595
|
# define __has_feature(x) 0
|
350
596
|
#endif
|
351
597
|
|
352
|
-
//
|
353
|
-
//
|
354
|
-
|
355
|
-
|
356
|
-
|
598
|
+
// Some old compilers do not support template specializations like:
|
599
|
+
// template<typename T,int N> void foo(const T x[N]);
|
600
|
+
#if !( EIGEN_COMP_CLANG && ( (EIGEN_COMP_CLANG<309) \
|
601
|
+
|| (defined(__apple_build_version__) && (__apple_build_version__ < 9000000))) \
|
602
|
+
|| EIGEN_COMP_GNUC_STRICT && EIGEN_COMP_GNUC<49)
|
603
|
+
#define EIGEN_HAS_STATIC_ARRAY_TEMPLATE 1
|
604
|
+
#else
|
605
|
+
#define EIGEN_HAS_STATIC_ARRAY_TEMPLATE 0
|
606
|
+
#endif
|
607
|
+
|
608
|
+
// The macro EIGEN_CPLUSPLUS is a replacement for __cplusplus/_MSVC_LANG that
|
609
|
+
// works for both platforms, indicating the C++ standard version number.
|
610
|
+
//
|
611
|
+
// With MSVC, without defining /Zc:__cplusplus, the __cplusplus macro will
|
612
|
+
// report 199711L regardless of the language standard specified via /std.
|
613
|
+
// We need to rely on _MSVC_LANG instead, which is only available after
|
614
|
+
// VS2015.3.
|
615
|
+
#if EIGEN_COMP_MSVC_LANG > 0
|
616
|
+
#define EIGEN_CPLUSPLUS EIGEN_COMP_MSVC_LANG
|
617
|
+
#elif EIGEN_COMP_MSVC >= 1900
|
618
|
+
#define EIGEN_CPLUSPLUS 201103L
|
619
|
+
#elif defined(__cplusplus)
|
620
|
+
#define EIGEN_CPLUSPLUS __cplusplus
|
621
|
+
#else
|
622
|
+
#define EIGEN_CPLUSPLUS 0
|
623
|
+
#endif
|
624
|
+
|
625
|
+
// The macro EIGEN_COMP_CXXVER defines the c++ verson expected by the compiler.
|
626
|
+
// For instance, if compiling with gcc and -std=c++17, then EIGEN_COMP_CXXVER
|
627
|
+
// is defined to 17.
|
628
|
+
#if EIGEN_CPLUSPLUS > 201703L
|
629
|
+
#define EIGEN_COMP_CXXVER 20
|
630
|
+
#elif EIGEN_CPLUSPLUS > 201402L
|
631
|
+
#define EIGEN_COMP_CXXVER 17
|
632
|
+
#elif EIGEN_CPLUSPLUS > 201103L
|
633
|
+
#define EIGEN_COMP_CXXVER 14
|
634
|
+
#elif EIGEN_CPLUSPLUS >= 201103L
|
635
|
+
#define EIGEN_COMP_CXXVER 11
|
636
|
+
#else
|
637
|
+
#define EIGEN_COMP_CXXVER 03
|
638
|
+
#endif
|
639
|
+
|
640
|
+
#ifndef EIGEN_HAS_CXX14_VARIABLE_TEMPLATES
|
641
|
+
#if defined(__cpp_variable_templates) && __cpp_variable_templates >= 201304 && EIGEN_MAX_CPP_VER>=14
|
642
|
+
#define EIGEN_HAS_CXX14_VARIABLE_TEMPLATES 1
|
643
|
+
#else
|
644
|
+
#define EIGEN_HAS_CXX14_VARIABLE_TEMPLATES 0
|
645
|
+
#endif
|
357
646
|
#endif
|
358
647
|
|
359
|
-
|
648
|
+
|
649
|
+
// The macros EIGEN_HAS_CXX?? defines a rough estimate of available c++ features
|
650
|
+
// but in practice we should not rely on them but rather on the availabilty of
|
651
|
+
// individual features as defined later.
|
652
|
+
// This is why there is no EIGEN_HAS_CXX17.
|
653
|
+
// FIXME: get rid of EIGEN_HAS_CXX14 and maybe even EIGEN_HAS_CXX11.
|
654
|
+
#if EIGEN_MAX_CPP_VER>=11 && EIGEN_COMP_CXXVER>=11
|
360
655
|
#define EIGEN_HAS_CXX11 1
|
361
656
|
#else
|
362
657
|
#define EIGEN_HAS_CXX11 0
|
363
658
|
#endif
|
364
659
|
|
660
|
+
#if EIGEN_MAX_CPP_VER>=14 && EIGEN_COMP_CXXVER>=14
|
661
|
+
#define EIGEN_HAS_CXX14 1
|
662
|
+
#else
|
663
|
+
#define EIGEN_HAS_CXX14 0
|
664
|
+
#endif
|
365
665
|
|
366
666
|
// Do we support r-value references?
|
367
667
|
#ifndef EIGEN_HAS_RVALUE_REFERENCES
|
368
668
|
#if EIGEN_MAX_CPP_VER>=11 && \
|
369
669
|
(__has_feature(cxx_rvalue_references) || \
|
370
|
-
|
371
|
-
(EIGEN_COMP_MSVC >= 1600))
|
670
|
+
(EIGEN_COMP_CXXVER >= 11) || (EIGEN_COMP_MSVC >= 1600))
|
372
671
|
#define EIGEN_HAS_RVALUE_REFERENCES 1
|
373
672
|
#else
|
374
673
|
#define EIGEN_HAS_RVALUE_REFERENCES 0
|
@@ -376,11 +675,14 @@
|
|
376
675
|
#endif
|
377
676
|
|
378
677
|
// Does the compiler support C99?
|
678
|
+
// Need to include <cmath> to make sure _GLIBCXX_USE_C99 gets defined
|
679
|
+
#include <cmath>
|
379
680
|
#ifndef EIGEN_HAS_C99_MATH
|
380
681
|
#if EIGEN_MAX_CPP_VER>=11 && \
|
381
682
|
((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901)) \
|
382
683
|
|| (defined(__GNUC__) && defined(_GLIBCXX_USE_C99)) \
|
383
|
-
|| (defined(_LIBCPP_VERSION) && !defined(_MSC_VER))
|
684
|
+
|| (defined(_LIBCPP_VERSION) && !defined(_MSC_VER)) \
|
685
|
+
|| (EIGEN_COMP_MSVC >= 1900) || defined(SYCL_DEVICE_ONLY))
|
384
686
|
#define EIGEN_HAS_C99_MATH 1
|
385
687
|
#else
|
386
688
|
#define EIGEN_HAS_C99_MATH 0
|
@@ -388,21 +690,73 @@
|
|
388
690
|
#endif
|
389
691
|
|
390
692
|
// Does the compiler support result_of?
|
693
|
+
// result_of was deprecated in c++17 and removed in c++ 20
|
391
694
|
#ifndef EIGEN_HAS_STD_RESULT_OF
|
392
|
-
#if
|
695
|
+
#if EIGEN_HAS_CXX11 && EIGEN_COMP_CXXVER < 17
|
393
696
|
#define EIGEN_HAS_STD_RESULT_OF 1
|
394
697
|
#else
|
395
698
|
#define EIGEN_HAS_STD_RESULT_OF 0
|
396
699
|
#endif
|
397
700
|
#endif
|
398
701
|
|
702
|
+
// Does the compiler support std::hash?
|
703
|
+
#ifndef EIGEN_HAS_STD_HASH
|
704
|
+
// The std::hash struct is defined in C++11 but is not labelled as a __device__
|
705
|
+
// function and is not constexpr, so cannot be used on device.
|
706
|
+
#if EIGEN_HAS_CXX11 && !defined(EIGEN_GPU_COMPILE_PHASE)
|
707
|
+
#define EIGEN_HAS_STD_HASH 1
|
708
|
+
#else
|
709
|
+
#define EIGEN_HAS_STD_HASH 0
|
710
|
+
#endif
|
711
|
+
#endif // EIGEN_HAS_STD_HASH
|
712
|
+
|
713
|
+
#ifndef EIGEN_HAS_STD_INVOKE_RESULT
|
714
|
+
#if EIGEN_MAX_CPP_VER >= 17 && EIGEN_COMP_CXXVER >= 17
|
715
|
+
#define EIGEN_HAS_STD_INVOKE_RESULT 1
|
716
|
+
#else
|
717
|
+
#define EIGEN_HAS_STD_INVOKE_RESULT 0
|
718
|
+
#endif
|
719
|
+
#endif
|
720
|
+
|
721
|
+
#ifndef EIGEN_HAS_ALIGNAS
|
722
|
+
#if EIGEN_MAX_CPP_VER>=11 && EIGEN_HAS_CXX11 && \
|
723
|
+
( __has_feature(cxx_alignas) \
|
724
|
+
|| EIGEN_HAS_CXX14 \
|
725
|
+
|| (EIGEN_COMP_MSVC >= 1800) \
|
726
|
+
|| (EIGEN_GNUC_AT_LEAST(4,8)) \
|
727
|
+
|| (EIGEN_COMP_CLANG>=305) \
|
728
|
+
|| (EIGEN_COMP_ICC>=1500) \
|
729
|
+
|| (EIGEN_COMP_PGI>=1500) \
|
730
|
+
|| (EIGEN_COMP_SUNCC>=0x5130))
|
731
|
+
#define EIGEN_HAS_ALIGNAS 1
|
732
|
+
#else
|
733
|
+
#define EIGEN_HAS_ALIGNAS 0
|
734
|
+
#endif
|
735
|
+
#endif
|
736
|
+
|
737
|
+
// Does the compiler support type_traits?
|
738
|
+
// - full support of type traits was added only to GCC 5.1.0.
|
739
|
+
// - 20150626 corresponds to the last release of 4.x libstdc++
|
740
|
+
#ifndef EIGEN_HAS_TYPE_TRAITS
|
741
|
+
#if EIGEN_MAX_CPP_VER>=11 && (EIGEN_HAS_CXX11 || EIGEN_COMP_MSVC >= 1700) \
|
742
|
+
&& ((!EIGEN_COMP_GNUC_STRICT) || EIGEN_GNUC_AT_LEAST(5, 1)) \
|
743
|
+
&& ((!defined(__GLIBCXX__)) || __GLIBCXX__ > 20150626)
|
744
|
+
#define EIGEN_HAS_TYPE_TRAITS 1
|
745
|
+
#define EIGEN_INCLUDE_TYPE_TRAITS
|
746
|
+
#else
|
747
|
+
#define EIGEN_HAS_TYPE_TRAITS 0
|
748
|
+
#endif
|
749
|
+
#endif
|
750
|
+
|
399
751
|
// Does the compiler support variadic templates?
|
400
752
|
#ifndef EIGEN_HAS_VARIADIC_TEMPLATES
|
401
|
-
#if EIGEN_MAX_CPP_VER>=11 && (
|
402
|
-
&& (!defined(__NVCC__) || !EIGEN_ARCH_ARM_OR_ARM64 || (
|
753
|
+
#if EIGEN_MAX_CPP_VER>=11 && (EIGEN_COMP_CXXVER >= 11) \
|
754
|
+
&& (!defined(__NVCC__) || !EIGEN_ARCH_ARM_OR_ARM64 || (EIGEN_COMP_NVCC >= 80000) )
|
403
755
|
// ^^ Disable the use of variadic templates when compiling with versions of nvcc older than 8.0 on ARM devices:
|
404
756
|
// this prevents nvcc from crashing when compiling Eigen on Tegra X1
|
405
757
|
#define EIGEN_HAS_VARIADIC_TEMPLATES 1
|
758
|
+
#elif EIGEN_MAX_CPP_VER>=11 && (EIGEN_COMP_CXXVER >= 11) && defined(SYCL_DEVICE_ONLY)
|
759
|
+
#define EIGEN_HAS_VARIADIC_TEMPLATES 1
|
406
760
|
#else
|
407
761
|
#define EIGEN_HAS_VARIADIC_TEMPLATES 0
|
408
762
|
#endif
|
@@ -410,27 +764,33 @@
|
|
410
764
|
|
411
765
|
// Does the compiler fully support const expressions? (as in c++14)
|
412
766
|
#ifndef EIGEN_HAS_CONSTEXPR
|
767
|
+
#if defined(EIGEN_CUDACC)
|
768
|
+
// Const expressions are supported provided that c++11 is enabled and we're using either clang or nvcc 7.5 or above
|
769
|
+
#if EIGEN_MAX_CPP_VER>=14 && (EIGEN_COMP_CXXVER >= 11 && (EIGEN_COMP_CLANG || EIGEN_COMP_NVCC >= 70500))
|
770
|
+
#define EIGEN_HAS_CONSTEXPR 1
|
771
|
+
#endif
|
772
|
+
#elif EIGEN_MAX_CPP_VER>=14 && (__has_feature(cxx_relaxed_constexpr) || (EIGEN_COMP_CXXVER >= 14) || \
|
773
|
+
(EIGEN_GNUC_AT_LEAST(4,8) && (EIGEN_COMP_CXXVER >= 11)) || \
|
774
|
+
(EIGEN_COMP_CLANG >= 306 && (EIGEN_COMP_CXXVER >= 11)))
|
775
|
+
#define EIGEN_HAS_CONSTEXPR 1
|
776
|
+
#endif
|
413
777
|
|
414
|
-
#
|
415
|
-
|
416
|
-
#
|
417
|
-
#define EIGEN_HAS_CONSTEXPR 1
|
418
|
-
#endif
|
419
|
-
#elif EIGEN_MAX_CPP_VER>=14 && (__has_feature(cxx_relaxed_constexpr) || (defined(__cplusplus) && __cplusplus >= 201402L) || \
|
420
|
-
(EIGEN_GNUC_AT_LEAST(4,8) && (__cplusplus > 199711L)))
|
421
|
-
#define EIGEN_HAS_CONSTEXPR 1
|
422
|
-
#endif
|
778
|
+
#ifndef EIGEN_HAS_CONSTEXPR
|
779
|
+
#define EIGEN_HAS_CONSTEXPR 0
|
780
|
+
#endif
|
423
781
|
|
424
|
-
#
|
425
|
-
#define EIGEN_HAS_CONSTEXPR 0
|
426
|
-
#endif
|
782
|
+
#endif // EIGEN_HAS_CONSTEXPR
|
427
783
|
|
784
|
+
#if EIGEN_HAS_CONSTEXPR
|
785
|
+
#define EIGEN_CONSTEXPR constexpr
|
786
|
+
#else
|
787
|
+
#define EIGEN_CONSTEXPR
|
428
788
|
#endif
|
429
789
|
|
430
790
|
// Does the compiler support C++11 math?
|
431
791
|
// Let's be conservative and enable the default C++11 implementation only if we are sure it exists
|
432
792
|
#ifndef EIGEN_HAS_CXX11_MATH
|
433
|
-
#if EIGEN_MAX_CPP_VER>=11 && ((
|
793
|
+
#if EIGEN_MAX_CPP_VER>=11 && ((EIGEN_COMP_CXXVER > 11) || (EIGEN_COMP_CXXVER == 11) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_MSVC || EIGEN_COMP_ICC) \
|
434
794
|
&& (EIGEN_ARCH_i386_OR_x86_64) && (EIGEN_OS_GNULINUX || EIGEN_OS_WIN_STRICT || EIGEN_OS_MAC))
|
435
795
|
#define EIGEN_HAS_CXX11_MATH 1
|
436
796
|
#else
|
@@ -441,9 +801,8 @@
|
|
441
801
|
// Does the compiler support proper C++11 containers?
|
442
802
|
#ifndef EIGEN_HAS_CXX11_CONTAINERS
|
443
803
|
#if EIGEN_MAX_CPP_VER>=11 && \
|
444
|
-
((
|
445
|
-
|| ((
|
446
|
-
|| EIGEN_COMP_MSVC >= 1900)
|
804
|
+
((EIGEN_COMP_CXXVER > 11) \
|
805
|
+
|| ((EIGEN_COMP_CXXVER == 11) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_MSVC || EIGEN_COMP_ICC>=1400)))
|
447
806
|
#define EIGEN_HAS_CXX11_CONTAINERS 1
|
448
807
|
#else
|
449
808
|
#define EIGEN_HAS_CXX11_CONTAINERS 0
|
@@ -454,24 +813,88 @@
|
|
454
813
|
#ifndef EIGEN_HAS_CXX11_NOEXCEPT
|
455
814
|
#if EIGEN_MAX_CPP_VER>=11 && \
|
456
815
|
(__has_feature(cxx_noexcept) \
|
457
|
-
|| (
|
458
|
-
|| ((
|
459
|
-
|| EIGEN_COMP_MSVC >= 1900)
|
816
|
+
|| (EIGEN_COMP_CXXVER > 11) \
|
817
|
+
|| ((EIGEN_COMP_CXXVER == 11) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_MSVC || EIGEN_COMP_ICC>=1400)))
|
460
818
|
#define EIGEN_HAS_CXX11_NOEXCEPT 1
|
461
819
|
#else
|
462
820
|
#define EIGEN_HAS_CXX11_NOEXCEPT 0
|
463
821
|
#endif
|
464
822
|
#endif
|
465
823
|
|
466
|
-
|
467
|
-
|
468
|
-
|
469
|
-
|
470
|
-
|
471
|
-
#
|
472
|
-
#
|
824
|
+
#ifndef EIGEN_HAS_CXX11_ATOMIC
|
825
|
+
#if EIGEN_MAX_CPP_VER>=11 && \
|
826
|
+
(__has_feature(cxx_atomic) \
|
827
|
+
|| (EIGEN_COMP_CXXVER > 11) \
|
828
|
+
|| ((EIGEN_COMP_CXXVER == 11) && (EIGEN_COMP_MSVC==0 || EIGEN_COMP_MSVC >= 1700)))
|
829
|
+
#define EIGEN_HAS_CXX11_ATOMIC 1
|
830
|
+
#else
|
831
|
+
#define EIGEN_HAS_CXX11_ATOMIC 0
|
832
|
+
#endif
|
833
|
+
#endif
|
834
|
+
|
835
|
+
#ifndef EIGEN_HAS_CXX11_OVERRIDE_FINAL
|
836
|
+
#if EIGEN_MAX_CPP_VER>=11 && \
|
837
|
+
(EIGEN_COMP_CXXVER >= 11 || EIGEN_COMP_MSVC >= 1700)
|
838
|
+
#define EIGEN_HAS_CXX11_OVERRIDE_FINAL 1
|
839
|
+
#else
|
840
|
+
#define EIGEN_HAS_CXX11_OVERRIDE_FINAL 0
|
841
|
+
#endif
|
842
|
+
#endif
|
843
|
+
|
844
|
+
// NOTE: the required Apple's clang version is very conservative
|
845
|
+
// and it could be that XCode 9 works just fine.
|
846
|
+
// NOTE: the MSVC version is based on https://en.cppreference.com/w/cpp/compiler_support
|
847
|
+
// and not tested.
|
848
|
+
#ifndef EIGEN_HAS_CXX17_OVERALIGN
|
849
|
+
#if EIGEN_MAX_CPP_VER>=17 && EIGEN_COMP_CXXVER>=17 && ( \
|
850
|
+
(EIGEN_COMP_MSVC >= 1912) \
|
851
|
+
|| (EIGEN_GNUC_AT_LEAST(7,0)) \
|
852
|
+
|| ((!defined(__apple_build_version__)) && (EIGEN_COMP_CLANG>=500)) \
|
853
|
+
|| (( defined(__apple_build_version__)) && (__apple_build_version__>=10000000)) \
|
854
|
+
)
|
855
|
+
#define EIGEN_HAS_CXX17_OVERALIGN 1
|
856
|
+
#else
|
857
|
+
#define EIGEN_HAS_CXX17_OVERALIGN 0
|
858
|
+
#endif
|
859
|
+
#endif
|
860
|
+
|
861
|
+
#if defined(EIGEN_CUDACC) && EIGEN_HAS_CONSTEXPR
|
862
|
+
// While available already with c++11, this is useful mostly starting with c++14 and relaxed constexpr rules
|
863
|
+
#if defined(__NVCC__)
|
864
|
+
// nvcc considers constexpr functions as __host__ __device__ with the option --expt-relaxed-constexpr
|
865
|
+
#ifdef __CUDACC_RELAXED_CONSTEXPR__
|
866
|
+
#define EIGEN_CONSTEXPR_ARE_DEVICE_FUNC
|
867
|
+
#endif
|
868
|
+
#elif defined(__clang__) && defined(__CUDA__) && __has_feature(cxx_relaxed_constexpr)
|
869
|
+
// clang++ always considers constexpr functions as implicitly __host__ __device__
|
870
|
+
#define EIGEN_CONSTEXPR_ARE_DEVICE_FUNC
|
871
|
+
#endif
|
473
872
|
#endif
|
474
873
|
|
874
|
+
// Does the compiler support the __int128 and __uint128_t extensions for 128-bit
|
875
|
+
// integer arithmetic?
|
876
|
+
//
|
877
|
+
// Clang and GCC define __SIZEOF_INT128__ when these extensions are supported,
|
878
|
+
// but we avoid using them in certain cases:
|
879
|
+
//
|
880
|
+
// * Building using Clang for Windows, where the Clang runtime library has
|
881
|
+
// 128-bit support only on LP64 architectures, but Windows is LLP64.
|
882
|
+
#ifndef EIGEN_HAS_BUILTIN_INT128
|
883
|
+
#if defined(__SIZEOF_INT128__) && !(EIGEN_OS_WIN && EIGEN_COMP_CLANG)
|
884
|
+
#define EIGEN_HAS_BUILTIN_INT128 1
|
885
|
+
#else
|
886
|
+
#define EIGEN_HAS_BUILTIN_INT128 0
|
887
|
+
#endif
|
888
|
+
#endif
|
889
|
+
|
890
|
+
//------------------------------------------------------------------------------------------
|
891
|
+
// Preprocessor programming helpers
|
892
|
+
//------------------------------------------------------------------------------------------
|
893
|
+
|
894
|
+
// This macro can be used to prevent from macro expansion, e.g.:
|
895
|
+
// std::max EIGEN_NOT_A_MACRO(a,b)
|
896
|
+
#define EIGEN_NOT_A_MACRO
|
897
|
+
|
475
898
|
#define EIGEN_DEBUG_VAR(x) std::cerr << #x << " = " << x << std::endl;
|
476
899
|
|
477
900
|
// concatenate two tokens
|
@@ -488,7 +911,7 @@
|
|
488
911
|
// but it still doesn't use GCC's always_inline. This is useful in (common) situations where MSVC needs forceinline
|
489
912
|
// but GCC is still doing fine with just inline.
|
490
913
|
#ifndef EIGEN_STRONG_INLINE
|
491
|
-
#if EIGEN_COMP_MSVC || EIGEN_COMP_ICC
|
914
|
+
#if (EIGEN_COMP_MSVC || EIGEN_COMP_ICC) && !defined(EIGEN_GPUCC)
|
492
915
|
#define EIGEN_STRONG_INLINE __forceinline
|
493
916
|
#else
|
494
917
|
#define EIGEN_STRONG_INLINE inline
|
@@ -503,7 +926,7 @@
|
|
503
926
|
// Eval.h:91: sorry, unimplemented: inlining failed in call to 'const Eigen::Eval<Derived> Eigen::MatrixBase<Scalar, Derived>::eval() const'
|
504
927
|
// : function body not available
|
505
928
|
// See also bug 1367
|
506
|
-
#if EIGEN_GNUC_AT_LEAST(4,2)
|
929
|
+
#if EIGEN_GNUC_AT_LEAST(4,2) && !defined(SYCL_DEVICE_ONLY)
|
507
930
|
#define EIGEN_ALWAYS_INLINE __attribute__((always_inline)) inline
|
508
931
|
#else
|
509
932
|
#define EIGEN_ALWAYS_INLINE EIGEN_STRONG_INLINE
|
@@ -523,12 +946,43 @@
|
|
523
946
|
#define EIGEN_PERMISSIVE_EXPR
|
524
947
|
#endif
|
525
948
|
|
949
|
+
// GPU stuff
|
950
|
+
|
951
|
+
// Disable some features when compiling with GPU compilers (NVCC/clang-cuda/SYCL/HIPCC)
|
952
|
+
#if defined(EIGEN_CUDACC) || defined(SYCL_DEVICE_ONLY) || defined(EIGEN_HIPCC)
|
953
|
+
// Do not try asserts on device code
|
954
|
+
#ifndef EIGEN_NO_DEBUG
|
955
|
+
#define EIGEN_NO_DEBUG
|
956
|
+
#endif
|
957
|
+
|
958
|
+
#ifdef EIGEN_INTERNAL_DEBUGGING
|
959
|
+
#undef EIGEN_INTERNAL_DEBUGGING
|
960
|
+
#endif
|
961
|
+
|
962
|
+
#ifdef EIGEN_EXCEPTIONS
|
963
|
+
#undef EIGEN_EXCEPTIONS
|
964
|
+
#endif
|
965
|
+
#endif
|
966
|
+
|
967
|
+
#if defined(SYCL_DEVICE_ONLY)
|
968
|
+
#ifndef EIGEN_DONT_VECTORIZE
|
969
|
+
#define EIGEN_DONT_VECTORIZE
|
970
|
+
#endif
|
971
|
+
#define EIGEN_DEVICE_FUNC __attribute__((flatten)) __attribute__((always_inline))
|
972
|
+
// All functions callable from CUDA/HIP code must be qualified with __device__
|
973
|
+
#elif defined(EIGEN_GPUCC)
|
974
|
+
#define EIGEN_DEVICE_FUNC __host__ __device__
|
975
|
+
#else
|
976
|
+
#define EIGEN_DEVICE_FUNC
|
977
|
+
#endif
|
978
|
+
|
979
|
+
|
526
980
|
// this macro allows to get rid of linking errors about multiply defined functions.
|
527
981
|
// - static is not very good because it prevents definitions from different object files to be merged.
|
528
982
|
// So static causes the resulting linked executable to be bloated with multiple copies of the same function.
|
529
983
|
// - inline is not perfect either as it unwantedly hints the compiler toward inlining the function.
|
530
|
-
#define EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
|
531
|
-
#define EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS inline
|
984
|
+
#define EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_DEVICE_FUNC
|
985
|
+
#define EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_DEVICE_FUNC inline
|
532
986
|
|
533
987
|
#ifdef NDEBUG
|
534
988
|
# ifndef EIGEN_NO_DEBUG
|
@@ -538,7 +992,11 @@
|
|
538
992
|
|
539
993
|
// eigen_plain_assert is where we implement the workaround for the assert() bug in GCC <= 4.3, see bug 89
|
540
994
|
#ifdef EIGEN_NO_DEBUG
|
541
|
-
#
|
995
|
+
#ifdef SYCL_DEVICE_ONLY // used to silence the warning on SYCL device
|
996
|
+
#define eigen_plain_assert(x) EIGEN_UNUSED_VARIABLE(x)
|
997
|
+
#else
|
998
|
+
#define eigen_plain_assert(x)
|
999
|
+
#endif
|
542
1000
|
#else
|
543
1001
|
#if EIGEN_SAFE_TO_USE_STANDARD_ASSERT_MACRO
|
544
1002
|
namespace Eigen {
|
@@ -612,7 +1070,7 @@
|
|
612
1070
|
// Suppresses 'unused variable' warnings.
|
613
1071
|
namespace Eigen {
|
614
1072
|
namespace internal {
|
615
|
-
template<typename T> EIGEN_DEVICE_FUNC void ignore_unused_variable(const T&) {}
|
1073
|
+
template<typename T> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void ignore_unused_variable(const T&) {}
|
616
1074
|
}
|
617
1075
|
}
|
618
1076
|
#define EIGEN_UNUSED_VARIABLE(var) Eigen::internal::ignore_unused_variable(var);
|
@@ -626,169 +1084,75 @@ namespace Eigen {
|
|
626
1084
|
#endif
|
627
1085
|
|
628
1086
|
|
629
|
-
|
630
|
-
//
|
1087
|
+
// Acts as a barrier preventing operations involving `X` from crossing. This
|
1088
|
+
// occurs, for example, in the fast rounding trick where a magic constant is
|
1089
|
+
// added then subtracted, which is otherwise compiled away with -ffast-math.
|
631
1090
|
//
|
632
|
-
//
|
633
|
-
|
634
|
-
|
635
|
-
//
|
636
|
-
//
|
637
|
-
//
|
638
|
-
//
|
639
|
-
|
640
|
-
|
641
|
-
|
642
|
-
|
643
|
-
|
644
|
-
|
645
|
-
|
646
|
-
|
647
|
-
|
648
|
-
|
649
|
-
|
650
|
-
|
651
|
-
|
652
|
-
|
653
|
-
|
654
|
-
|
655
|
-
|
656
|
-
|
657
|
-
|
658
|
-
|
659
|
-
|
660
|
-
|
661
|
-
|
662
|
-
//
|
663
|
-
|
664
|
-
|
665
|
-
|
666
|
-
|
667
|
-
|
668
|
-
|
669
|
-
|
670
|
-
|
671
|
-
|
672
|
-
|
673
|
-
#
|
674
|
-
|
675
|
-
|
676
|
-
|
677
|
-
|
678
|
-
|
679
|
-
|
680
|
-
//
|
681
|
-
|
682
|
-
|
683
|
-
#if (defined(EIGEN_DONT_ALIGN_STATICALLY) || defined(EIGEN_DONT_ALIGN)) && defined(EIGEN_MAX_STATIC_ALIGN_BYTES) && EIGEN_MAX_STATIC_ALIGN_BYTES>0
|
684
|
-
#error EIGEN_MAX_STATIC_ALIGN_BYTES and EIGEN_DONT_ALIGN[_STATICALLY] are both defined with EIGEN_MAX_STATIC_ALIGN_BYTES!=0. Use EIGEN_MAX_STATIC_ALIGN_BYTES=0 as a synonym of EIGEN_DONT_ALIGN_STATICALLY.
|
685
|
-
#endif
|
686
|
-
|
687
|
-
// EIGEN_DONT_ALIGN_STATICALLY and EIGEN_DONT_ALIGN are deprectated
|
688
|
-
// They imply EIGEN_MAX_STATIC_ALIGN_BYTES=0
|
689
|
-
#if defined(EIGEN_DONT_ALIGN_STATICALLY) || defined(EIGEN_DONT_ALIGN)
|
690
|
-
#ifdef EIGEN_MAX_STATIC_ALIGN_BYTES
|
691
|
-
#undef EIGEN_MAX_STATIC_ALIGN_BYTES
|
692
|
-
#endif
|
693
|
-
#define EIGEN_MAX_STATIC_ALIGN_BYTES 0
|
694
|
-
#endif
|
695
|
-
|
696
|
-
#ifndef EIGEN_MAX_STATIC_ALIGN_BYTES
|
697
|
-
|
698
|
-
// Try to automatically guess what is the best default value for EIGEN_MAX_STATIC_ALIGN_BYTES
|
699
|
-
|
700
|
-
// 16 byte alignment is only useful for vectorization. Since it affects the ABI, we need to enable
|
701
|
-
// 16 byte alignment on all platforms where vectorization might be enabled. In theory we could always
|
702
|
-
// enable alignment, but it can be a cause of problems on some platforms, so we just disable it in
|
703
|
-
// certain common platform (compiler+architecture combinations) to avoid these problems.
|
704
|
-
// Only static alignment is really problematic (relies on nonstandard compiler extensions),
|
705
|
-
// try to keep heap alignment even when we have to disable static alignment.
|
706
|
-
#if EIGEN_COMP_GNUC && !(EIGEN_ARCH_i386_OR_x86_64 || EIGEN_ARCH_ARM_OR_ARM64 || EIGEN_ARCH_PPC || EIGEN_ARCH_IA64)
|
707
|
-
#define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 1
|
708
|
-
#elif EIGEN_ARCH_ARM_OR_ARM64 && EIGEN_COMP_GNUC_STRICT && EIGEN_GNUC_AT_MOST(4, 6)
|
709
|
-
// Old versions of GCC on ARM, at least 4.4, were once seen to have buggy static alignment support.
|
710
|
-
// Not sure which version fixed it, hopefully it doesn't affect 4.7, which is still somewhat in use.
|
711
|
-
// 4.8 and newer seem definitely unaffected.
|
712
|
-
#define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 1
|
713
|
-
#else
|
714
|
-
#define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 0
|
715
|
-
#endif
|
716
|
-
|
717
|
-
// static alignment is completely disabled with GCC 3, Sun Studio, and QCC/QNX
|
718
|
-
#if !EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT \
|
719
|
-
&& !EIGEN_GCC3_OR_OLDER \
|
720
|
-
&& !EIGEN_COMP_SUNCC \
|
721
|
-
&& !EIGEN_OS_QNX
|
722
|
-
#define EIGEN_ARCH_WANTS_STACK_ALIGNMENT 1
|
723
|
-
#else
|
724
|
-
#define EIGEN_ARCH_WANTS_STACK_ALIGNMENT 0
|
725
|
-
#endif
|
726
|
-
|
727
|
-
#if EIGEN_ARCH_WANTS_STACK_ALIGNMENT
|
728
|
-
#define EIGEN_MAX_STATIC_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES
|
1091
|
+
// See bug 1674
|
1092
|
+
#if !defined(EIGEN_OPTIMIZATION_BARRIER)
|
1093
|
+
#if EIGEN_COMP_GNUC
|
1094
|
+
// According to https://gcc.gnu.org/onlinedocs/gcc/Constraints.html:
|
1095
|
+
// X: Any operand whatsoever.
|
1096
|
+
// r: A register operand is allowed provided that it is in a general
|
1097
|
+
// register.
|
1098
|
+
// g: Any register, memory or immediate integer operand is allowed, except
|
1099
|
+
// for registers that are not general registers.
|
1100
|
+
// w: (AArch32/AArch64) Floating point register, Advanced SIMD vector
|
1101
|
+
// register or SVE vector register.
|
1102
|
+
// x: (SSE) Any SSE register.
|
1103
|
+
// (AArch64) Like w, but restricted to registers 0 to 15 inclusive.
|
1104
|
+
// v: (PowerPC) An Altivec vector register.
|
1105
|
+
// wa:(PowerPC) A VSX register.
|
1106
|
+
//
|
1107
|
+
// "X" (uppercase) should work for all cases, though this seems to fail for
|
1108
|
+
// some versions of GCC for arm/aarch64 with
|
1109
|
+
// "error: inconsistent operand constraints in an 'asm'"
|
1110
|
+
// Clang x86_64/arm/aarch64 seems to require "g" to support both scalars and
|
1111
|
+
// vectors, otherwise
|
1112
|
+
// "error: non-trivial scalar-to-vector conversion, possible invalid
|
1113
|
+
// constraint for vector type"
|
1114
|
+
//
|
1115
|
+
// GCC for ppc64le generates an internal compiler error with x/X/g.
|
1116
|
+
// GCC for AVX generates an internal compiler error with X.
|
1117
|
+
//
|
1118
|
+
// Tested on icc/gcc/clang for sse, avx, avx2, avx512dq
|
1119
|
+
// gcc for arm, aarch64,
|
1120
|
+
// gcc for ppc64le,
|
1121
|
+
// both vectors and scalars.
|
1122
|
+
//
|
1123
|
+
// Note that this is restricted to plain types - this will not work
|
1124
|
+
// directly for std::complex<T>, Eigen::half, Eigen::bfloat16. For these,
|
1125
|
+
// you will need to apply to the underlying POD type.
|
1126
|
+
#if EIGEN_ARCH_PPC && EIGEN_COMP_GNUC_STRICT
|
1127
|
+
// This seems to be broken on clang. Packet4f is loaded into a single
|
1128
|
+
// register rather than a vector, zeroing out some entries. Integer
|
1129
|
+
// types also generate a compile error.
|
1130
|
+
// General, Altivec, VSX.
|
1131
|
+
#define EIGEN_OPTIMIZATION_BARRIER(X) __asm__ ("" : "+r,v,wa" (X));
|
1132
|
+
#elif EIGEN_ARCH_ARM_OR_ARM64
|
1133
|
+
// General, NEON.
|
1134
|
+
#define EIGEN_OPTIMIZATION_BARRIER(X) __asm__ ("" : "+g,w" (X));
|
1135
|
+
#elif EIGEN_ARCH_i386_OR_x86_64
|
1136
|
+
// General, SSE.
|
1137
|
+
#define EIGEN_OPTIMIZATION_BARRIER(X) __asm__ ("" : "+g,x" (X));
|
1138
|
+
#else
|
1139
|
+
// Not implemented for other architectures.
|
1140
|
+
#define EIGEN_OPTIMIZATION_BARRIER(X)
|
1141
|
+
#endif
|
729
1142
|
#else
|
730
|
-
|
731
|
-
|
732
|
-
|
733
|
-
#endif
|
734
|
-
|
735
|
-
// If EIGEN_MAX_ALIGN_BYTES is defined, then it is considered as an upper bound for EIGEN_MAX_ALIGN_BYTES
|
736
|
-
#if defined(EIGEN_MAX_ALIGN_BYTES) && EIGEN_MAX_ALIGN_BYTES<EIGEN_MAX_STATIC_ALIGN_BYTES
|
737
|
-
#undef EIGEN_MAX_STATIC_ALIGN_BYTES
|
738
|
-
#define EIGEN_MAX_STATIC_ALIGN_BYTES EIGEN_MAX_ALIGN_BYTES
|
739
|
-
#endif
|
740
|
-
|
741
|
-
#if EIGEN_MAX_STATIC_ALIGN_BYTES==0 && !defined(EIGEN_DISABLE_UNALIGNED_ARRAY_ASSERT)
|
742
|
-
#define EIGEN_DISABLE_UNALIGNED_ARRAY_ASSERT
|
743
|
-
#endif
|
744
|
-
|
745
|
-
// At this stage, EIGEN_MAX_STATIC_ALIGN_BYTES>0 is the true test whether we want to align arrays on the stack or not.
|
746
|
-
// It takes into account both the user choice to explicitly enable/disable alignment (by settting EIGEN_MAX_STATIC_ALIGN_BYTES)
|
747
|
-
// and the architecture config (EIGEN_ARCH_WANTS_STACK_ALIGNMENT).
|
748
|
-
// Henceforth, only EIGEN_MAX_STATIC_ALIGN_BYTES should be used.
|
749
|
-
|
750
|
-
|
751
|
-
// Shortcuts to EIGEN_ALIGN_TO_BOUNDARY
|
752
|
-
#define EIGEN_ALIGN8 EIGEN_ALIGN_TO_BOUNDARY(8)
|
753
|
-
#define EIGEN_ALIGN16 EIGEN_ALIGN_TO_BOUNDARY(16)
|
754
|
-
#define EIGEN_ALIGN32 EIGEN_ALIGN_TO_BOUNDARY(32)
|
755
|
-
#define EIGEN_ALIGN64 EIGEN_ALIGN_TO_BOUNDARY(64)
|
756
|
-
#if EIGEN_MAX_STATIC_ALIGN_BYTES>0
|
757
|
-
#define EIGEN_ALIGN_MAX EIGEN_ALIGN_TO_BOUNDARY(EIGEN_MAX_STATIC_ALIGN_BYTES)
|
758
|
-
#else
|
759
|
-
#define EIGEN_ALIGN_MAX
|
760
|
-
#endif
|
761
|
-
|
762
|
-
|
763
|
-
// Dynamic alignment control
|
764
|
-
|
765
|
-
#if defined(EIGEN_DONT_ALIGN) && defined(EIGEN_MAX_ALIGN_BYTES) && EIGEN_MAX_ALIGN_BYTES>0
|
766
|
-
#error EIGEN_MAX_ALIGN_BYTES and EIGEN_DONT_ALIGN are both defined with EIGEN_MAX_ALIGN_BYTES!=0. Use EIGEN_MAX_ALIGN_BYTES=0 as a synonym of EIGEN_DONT_ALIGN.
|
767
|
-
#endif
|
768
|
-
|
769
|
-
#ifdef EIGEN_DONT_ALIGN
|
770
|
-
#ifdef EIGEN_MAX_ALIGN_BYTES
|
771
|
-
#undef EIGEN_MAX_ALIGN_BYTES
|
1143
|
+
// Not implemented for other compilers.
|
1144
|
+
#define EIGEN_OPTIMIZATION_BARRIER(X)
|
772
1145
|
#endif
|
773
|
-
#define EIGEN_MAX_ALIGN_BYTES 0
|
774
|
-
#elif !defined(EIGEN_MAX_ALIGN_BYTES)
|
775
|
-
#define EIGEN_MAX_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES
|
776
1146
|
#endif
|
777
1147
|
|
778
|
-
#if
|
779
|
-
|
1148
|
+
#if EIGEN_COMP_MSVC
|
1149
|
+
// NOTE MSVC often gives C4127 warnings with compiletime if statements. See bug 1362.
|
1150
|
+
// This workaround is ugly, but it does the job.
|
1151
|
+
# define EIGEN_CONST_CONDITIONAL(cond) (void)0, cond
|
780
1152
|
#else
|
781
|
-
#define
|
782
|
-
#endif
|
783
|
-
|
784
|
-
|
785
|
-
#ifndef EIGEN_UNALIGNED_VECTORIZE
|
786
|
-
#define EIGEN_UNALIGNED_VECTORIZE 1
|
1153
|
+
# define EIGEN_CONST_CONDITIONAL(cond) cond
|
787
1154
|
#endif
|
788
1155
|
|
789
|
-
//----------------------------------------------------------------------
|
790
|
-
|
791
|
-
|
792
1156
|
#ifdef EIGEN_DONT_USE_RESTRICT_KEYWORD
|
793
1157
|
#define EIGEN_RESTRICT
|
794
1158
|
#endif
|
@@ -796,10 +1160,6 @@ namespace Eigen {
|
|
796
1160
|
#define EIGEN_RESTRICT __restrict
|
797
1161
|
#endif
|
798
1162
|
|
799
|
-
#ifndef EIGEN_STACK_ALLOCATION_LIMIT
|
800
|
-
// 131072 == 128 KB
|
801
|
-
#define EIGEN_STACK_ALLOCATION_LIMIT 131072
|
802
|
-
#endif
|
803
1163
|
|
804
1164
|
#ifndef EIGEN_DEFAULT_IO_FORMAT
|
805
1165
|
#ifdef EIGEN_MAKING_DOCS
|
@@ -814,8 +1174,23 @@ namespace Eigen {
|
|
814
1174
|
// just an empty macro !
|
815
1175
|
#define EIGEN_EMPTY
|
816
1176
|
|
817
|
-
|
818
|
-
|
1177
|
+
|
1178
|
+
// When compiling CUDA/HIP device code with NVCC or HIPCC
|
1179
|
+
// pull in math functions from the global namespace.
|
1180
|
+
// In host mode, and when device code is compiled with clang,
|
1181
|
+
// use the std versions.
|
1182
|
+
#if (defined(EIGEN_CUDA_ARCH) && defined(__NVCC__)) || defined(EIGEN_HIP_DEVICE_COMPILE)
|
1183
|
+
#define EIGEN_USING_STD(FUNC) using ::FUNC;
|
1184
|
+
#else
|
1185
|
+
#define EIGEN_USING_STD(FUNC) using std::FUNC;
|
1186
|
+
#endif
|
1187
|
+
|
1188
|
+
#if EIGEN_COMP_MSVC_STRICT && (EIGEN_COMP_MSVC < 1900 || (EIGEN_COMP_MSVC == 1900 && EIGEN_COMP_NVCC))
|
1189
|
+
// For older MSVC versions, as well as 1900 && CUDA 8, using the base operator is necessary,
|
1190
|
+
// otherwise we get duplicate definition errors
|
1191
|
+
// For later MSVC versions, we require explicit operator= definition, otherwise we get
|
1192
|
+
// use of implicitly deleted operator errors.
|
1193
|
+
// (cf Bugs 920, 1000, 1324, 2291)
|
819
1194
|
#define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \
|
820
1195
|
using Base::operator =;
|
821
1196
|
#elif EIGEN_COMP_CLANG // workaround clang bug (see http://forum.kde.org/viewtopic.php?f=74&t=102653)
|
@@ -835,11 +1210,48 @@ namespace Eigen {
|
|
835
1210
|
#endif
|
836
1211
|
|
837
1212
|
|
1213
|
+
/**
|
1214
|
+
* \internal
|
1215
|
+
* \brief Macro to explicitly define the default copy constructor.
|
1216
|
+
* This is necessary, because the implicit definition is deprecated if the copy-assignment is overridden.
|
1217
|
+
*/
|
1218
|
+
#if EIGEN_HAS_CXX11
|
1219
|
+
#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) CLASS(const CLASS&) = default;
|
1220
|
+
#else
|
1221
|
+
#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS)
|
1222
|
+
#endif
|
1223
|
+
|
1224
|
+
|
1225
|
+
|
838
1226
|
/** \internal
|
839
1227
|
* \brief Macro to manually inherit assignment operators.
|
840
1228
|
* This is necessary, because the implicitly defined assignment operator gets deleted when a custom operator= is defined.
|
1229
|
+
* With C++11 or later this also default-implements the copy-constructor
|
841
1230
|
*/
|
842
|
-
#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived)
|
1231
|
+
#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) \
|
1232
|
+
EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \
|
1233
|
+
EIGEN_DEFAULT_COPY_CONSTRUCTOR(Derived)
|
1234
|
+
|
1235
|
+
/** \internal
|
1236
|
+
* \brief Macro to manually define default constructors and destructors.
|
1237
|
+
* This is necessary when the copy constructor is re-defined.
|
1238
|
+
* For empty helper classes this should usually be protected, to avoid accidentally creating empty objects.
|
1239
|
+
*
|
1240
|
+
* Hiding the default destructor lead to problems in C++03 mode together with boost::multiprecision
|
1241
|
+
*/
|
1242
|
+
#if EIGEN_HAS_CXX11
|
1243
|
+
#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \
|
1244
|
+
Derived() = default; \
|
1245
|
+
~Derived() = default;
|
1246
|
+
#else
|
1247
|
+
#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \
|
1248
|
+
Derived() {}; \
|
1249
|
+
/* ~Derived() {}; */
|
1250
|
+
#endif
|
1251
|
+
|
1252
|
+
|
1253
|
+
|
1254
|
+
|
843
1255
|
|
844
1256
|
/**
|
845
1257
|
* Just a side note. Commenting within defines works only by documenting
|
@@ -856,7 +1268,8 @@ namespace Eigen {
|
|
856
1268
|
typedef typename Eigen::internal::ref_selector<Derived>::type Nested; \
|
857
1269
|
typedef typename Eigen::internal::traits<Derived>::StorageKind StorageKind; \
|
858
1270
|
typedef typename Eigen::internal::traits<Derived>::StorageIndex StorageIndex; \
|
859
|
-
enum
|
1271
|
+
enum CompileTimeTraits \
|
1272
|
+
{ RowsAtCompileTime = Eigen::internal::traits<Derived>::RowsAtCompileTime, \
|
860
1273
|
ColsAtCompileTime = Eigen::internal::traits<Derived>::ColsAtCompileTime, \
|
861
1274
|
Flags = Eigen::internal::traits<Derived>::Flags, \
|
862
1275
|
SizeAtCompileTime = Base::SizeAtCompileTime, \
|
@@ -901,6 +1314,14 @@ namespace Eigen {
|
|
901
1314
|
|
902
1315
|
#define EIGEN_IMPLIES(a,b) (!(a) || (b))
|
903
1316
|
|
1317
|
+
#if EIGEN_HAS_BUILTIN(__builtin_expect) || EIGEN_COMP_GNUC
|
1318
|
+
#define EIGEN_PREDICT_FALSE(x) (__builtin_expect(x, false))
|
1319
|
+
#define EIGEN_PREDICT_TRUE(x) (__builtin_expect(false || (x), true))
|
1320
|
+
#else
|
1321
|
+
#define EIGEN_PREDICT_FALSE(x) (x)
|
1322
|
+
#define EIGEN_PREDICT_TRUE(x) (x)
|
1323
|
+
#endif
|
1324
|
+
|
904
1325
|
// the expression type of a standard coefficient wise binary operation
|
905
1326
|
#define EIGEN_CWISE_BINARY_RETURN_TYPE(LHS,RHS,OPNAME) \
|
906
1327
|
CwiseBinaryOp< \
|
@@ -932,14 +1353,14 @@ namespace Eigen {
|
|
932
1353
|
const typename internal::plain_constant_type<EXPR,SCALAR>::type, const EXPR>
|
933
1354
|
|
934
1355
|
// Workaround for MSVC 2010 (see ML thread "patch with compile for for MSVC 2010")
|
935
|
-
#if EIGEN_COMP_MSVC_STRICT<=1600
|
1356
|
+
#if EIGEN_COMP_MSVC_STRICT && (EIGEN_COMP_MSVC_STRICT<=1600)
|
936
1357
|
#define EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(X) typename internal::enable_if<true,X>::type
|
937
1358
|
#else
|
938
1359
|
#define EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(X) X
|
939
1360
|
#endif
|
940
1361
|
|
941
1362
|
#define EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(METHOD,OPNAME) \
|
942
|
-
template <typename T> EIGEN_DEVICE_FUNC
|
1363
|
+
template <typename T> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE \
|
943
1364
|
EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,typename internal::promote_scalar_arg<Scalar EIGEN_COMMA T EIGEN_COMMA EIGEN_SCALAR_BINARY_SUPPORTED(OPNAME,Scalar,T)>::type,OPNAME))\
|
944
1365
|
(METHOD)(const T& scalar) const { \
|
945
1366
|
typedef typename internal::promote_scalar_arg<Scalar,T,EIGEN_SCALAR_BINARY_SUPPORTED(OPNAME,Scalar,T)>::type PromotedT; \
|
@@ -948,7 +1369,7 @@ namespace Eigen {
|
|
948
1369
|
}
|
949
1370
|
|
950
1371
|
#define EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(METHOD,OPNAME) \
|
951
|
-
template <typename T> EIGEN_DEVICE_FUNC
|
1372
|
+
template <typename T> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE friend \
|
952
1373
|
EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename internal::promote_scalar_arg<Scalar EIGEN_COMMA T EIGEN_COMMA EIGEN_SCALAR_BINARY_SUPPORTED(OPNAME,T,Scalar)>::type,Derived,OPNAME)) \
|
953
1374
|
(METHOD)(const T& scalar, const StorageBaseType& matrix) { \
|
954
1375
|
typedef typename internal::promote_scalar_arg<Scalar,T,EIGEN_SCALAR_BINARY_SUPPORTED(OPNAME,T,Scalar)>::type PromotedT; \
|
@@ -961,15 +1382,23 @@ namespace Eigen {
|
|
961
1382
|
EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(METHOD,OPNAME)
|
962
1383
|
|
963
1384
|
|
1385
|
+
#if (defined(_CPPUNWIND) || defined(__EXCEPTIONS)) && !defined(EIGEN_CUDA_ARCH) && !defined(EIGEN_EXCEPTIONS) && !defined(EIGEN_USE_SYCL) && !defined(EIGEN_HIP_DEVICE_COMPILE)
|
1386
|
+
#define EIGEN_EXCEPTIONS
|
1387
|
+
#endif
|
1388
|
+
|
1389
|
+
|
964
1390
|
#ifdef EIGEN_EXCEPTIONS
|
965
1391
|
# define EIGEN_THROW_X(X) throw X
|
966
1392
|
# define EIGEN_THROW throw
|
967
1393
|
# define EIGEN_TRY try
|
968
1394
|
# define EIGEN_CATCH(X) catch (X)
|
969
1395
|
#else
|
970
|
-
#
|
1396
|
+
# if defined(EIGEN_CUDA_ARCH)
|
971
1397
|
# define EIGEN_THROW_X(X) asm("trap;")
|
972
1398
|
# define EIGEN_THROW asm("trap;")
|
1399
|
+
# elif defined(EIGEN_HIP_DEVICE_COMPILE)
|
1400
|
+
# define EIGEN_THROW_X(X) asm("s_trap 0")
|
1401
|
+
# define EIGEN_THROW asm("s_trap 0")
|
973
1402
|
# else
|
974
1403
|
# define EIGEN_THROW_X(X) std::abort()
|
975
1404
|
# define EIGEN_THROW std::abort()
|
@@ -989,13 +1418,47 @@ namespace Eigen {
|
|
989
1418
|
# define EIGEN_NOEXCEPT
|
990
1419
|
# define EIGEN_NOEXCEPT_IF(x)
|
991
1420
|
# define EIGEN_NO_THROW throw()
|
992
|
-
# if EIGEN_COMP_MSVC
|
1421
|
+
# if EIGEN_COMP_MSVC || EIGEN_COMP_CXXVER>=17
|
993
1422
|
// MSVC does not support exception specifications (warning C4290),
|
994
|
-
// and they are deprecated in c++11 anyway.
|
1423
|
+
// and they are deprecated in c++11 anyway. This is even an error in c++17.
|
995
1424
|
# define EIGEN_EXCEPTION_SPEC(X) throw()
|
996
1425
|
# else
|
997
1426
|
# define EIGEN_EXCEPTION_SPEC(X) throw(X)
|
998
1427
|
# endif
|
999
1428
|
#endif
|
1000
1429
|
|
1430
|
+
#if EIGEN_HAS_VARIADIC_TEMPLATES
|
1431
|
+
// The all function is used to enable a variadic version of eigen_assert which can take a parameter pack as its input.
|
1432
|
+
namespace Eigen {
|
1433
|
+
namespace internal {
|
1434
|
+
|
1435
|
+
inline bool all(){ return true; }
|
1436
|
+
|
1437
|
+
template<typename T, typename ...Ts>
|
1438
|
+
bool all(T t, Ts ... ts){ return t && all(ts...); }
|
1439
|
+
|
1440
|
+
}
|
1441
|
+
}
|
1442
|
+
#endif
|
1443
|
+
|
1444
|
+
#if EIGEN_HAS_CXX11_OVERRIDE_FINAL
|
1445
|
+
// provide override and final specifiers if they are available:
|
1446
|
+
# define EIGEN_OVERRIDE override
|
1447
|
+
# define EIGEN_FINAL final
|
1448
|
+
#else
|
1449
|
+
# define EIGEN_OVERRIDE
|
1450
|
+
# define EIGEN_FINAL
|
1451
|
+
#endif
|
1452
|
+
|
1453
|
+
// Wrapping #pragma unroll in a macro since it is required for SYCL
|
1454
|
+
#if defined(SYCL_DEVICE_ONLY)
|
1455
|
+
#if defined(_MSC_VER)
|
1456
|
+
#define EIGEN_UNROLL_LOOP __pragma(unroll)
|
1457
|
+
#else
|
1458
|
+
#define EIGEN_UNROLL_LOOP _Pragma("unroll")
|
1459
|
+
#endif
|
1460
|
+
#else
|
1461
|
+
#define EIGEN_UNROLL_LOOP
|
1462
|
+
#endif
|
1463
|
+
|
1001
1464
|
#endif // EIGEN_MACROS_H
|