diff options
Diffstat (limited to 'Eigen/src/Core/util/BlasUtil.h')
-rw-r--r-- | Eigen/src/Core/util/BlasUtil.h | 120 |
1 files changed, 62 insertions, 58 deletions
diff --git a/Eigen/src/Core/util/BlasUtil.h b/Eigen/src/Core/util/BlasUtil.h index 972814dc9..7c976a5bc 100644 --- a/Eigen/src/Core/util/BlasUtil.h +++ b/Eigen/src/Core/util/BlasUtil.h @@ -28,109 +28,111 @@ // This file contains many lightweight helper classes used to // implement and control fast level 2 and level 3 BLAS-like routines. +namespace internal { + // forward declarations template<typename LhsScalar, typename RhsScalar, typename Index, int mr, int nr, bool ConjugateLhs=false, bool ConjugateRhs=false> -struct ei_gebp_kernel; +struct gebp_kernel; template<typename Scalar, typename Index, int nr, int StorageOrder, bool Conjugate = false, bool PanelMode=false> -struct ei_gemm_pack_rhs; +struct gemm_pack_rhs; template<typename Scalar, typename Index, int Pack1, int Pack2, int StorageOrder, bool Conjugate = false, bool PanelMode = false> -struct ei_gemm_pack_lhs; +struct gemm_pack_lhs; template< typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, int ResStorageOrder> -struct ei_general_matrix_matrix_product; +struct general_matrix_matrix_product; template<typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, typename RhsScalar, bool ConjugateRhs> -struct ei_general_matrix_vector_product; +struct general_matrix_vector_product; -template<bool Conjugate> struct ei_conj_if; +template<bool Conjugate> struct conj_if; -template<> struct ei_conj_if<true> { +template<> struct conj_if<true> { template<typename T> - inline T operator()(const T& x) { return ei_conj(x); } + inline T operator()(const T& x) { return conj(x); } }; -template<> struct ei_conj_if<false> { +template<> struct conj_if<false> { template<typename T> inline const T& operator()(const T& x) { return x; } }; -template<typename Scalar> struct ei_conj_helper<Scalar,Scalar,false,false> +template<typename Scalar> struct conj_helper<Scalar,Scalar,false,false> { - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return ei_pmadd(x,y,c); } - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const { return ei_pmul(x,y); } + EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return internal::pmadd(x,y,c); } + EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const { return internal::pmul(x,y); } }; -template<typename RealScalar> struct ei_conj_helper<std::complex<RealScalar>, std::complex<RealScalar>, false,true> +template<typename RealScalar> struct conj_helper<std::complex<RealScalar>, std::complex<RealScalar>, false,true> { typedef std::complex<RealScalar> Scalar; EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return c + pmul(x,y); } EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(ei_real(x)*ei_real(y) + ei_imag(x)*ei_imag(y), ei_imag(x)*ei_real(y) - ei_real(x)*ei_imag(y)); } + { return Scalar(real(x)*real(y) + imag(x)*imag(y), imag(x)*real(y) - real(x)*imag(y)); } }; -template<typename RealScalar> struct ei_conj_helper<std::complex<RealScalar>, std::complex<RealScalar>, true,false> +template<typename RealScalar> struct conj_helper<std::complex<RealScalar>, std::complex<RealScalar>, true,false> { typedef std::complex<RealScalar> Scalar; EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return c + pmul(x,y); } EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(ei_real(x)*ei_real(y) + ei_imag(x)*ei_imag(y), ei_real(x)*ei_imag(y) - ei_imag(x)*ei_real(y)); } + { return Scalar(real(x)*real(y) + imag(x)*imag(y), real(x)*imag(y) - imag(x)*real(y)); } }; -template<typename RealScalar> struct ei_conj_helper<std::complex<RealScalar>, std::complex<RealScalar>, true,true> +template<typename RealScalar> struct conj_helper<std::complex<RealScalar>, std::complex<RealScalar>, true,true> { typedef std::complex<RealScalar> Scalar; EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return c + pmul(x,y); } EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(ei_real(x)*ei_real(y) - ei_imag(x)*ei_imag(y), - ei_real(x)*ei_imag(y) - ei_imag(x)*ei_real(y)); } + { return Scalar(real(x)*real(y) - imag(x)*imag(y), - real(x)*imag(y) - imag(x)*real(y)); } }; -template<typename RealScalar,bool Conj> struct ei_conj_helper<std::complex<RealScalar>, RealScalar, Conj,false> +template<typename RealScalar,bool Conj> struct conj_helper<std::complex<RealScalar>, RealScalar, Conj,false> { typedef std::complex<RealScalar> Scalar; EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const RealScalar& y, const Scalar& c) const - { return ei_padd(c, pmul(x,y)); } + { return padd(c, pmul(x,y)); } EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const RealScalar& y) const - { return ei_conj_if<Conj>()(x)*y; } + { return conj_if<Conj>()(x)*y; } }; -template<typename RealScalar,bool Conj> struct ei_conj_helper<RealScalar, std::complex<RealScalar>, false,Conj> +template<typename RealScalar,bool Conj> struct conj_helper<RealScalar, std::complex<RealScalar>, false,Conj> { typedef std::complex<RealScalar> Scalar; EIGEN_STRONG_INLINE Scalar pmadd(const RealScalar& x, const Scalar& y, const Scalar& c) const - { return ei_padd(c, pmul(x,y)); } + { return padd(c, pmul(x,y)); } EIGEN_STRONG_INLINE Scalar pmul(const RealScalar& x, const Scalar& y) const - { return x*ei_conj_if<Conj>()(y); } + { return x*conj_if<Conj>()(y); } }; -template<typename From,typename To> struct ei_get_factor { +template<typename From,typename To> struct get_factor { EIGEN_STRONG_INLINE static To run(const From& x) { return x; } }; -template<typename Scalar> struct ei_get_factor<Scalar,typename NumTraits<Scalar>::Real> { - EIGEN_STRONG_INLINE static typename NumTraits<Scalar>::Real run(const Scalar& x) { return ei_real(x); } +template<typename Scalar> struct get_factor<Scalar,typename NumTraits<Scalar>::Real> { + EIGEN_STRONG_INLINE static typename NumTraits<Scalar>::Real run(const Scalar& x) { return real(x); } }; // Lightweight helper class to access matrix coefficients. // Yes, this is somehow redundant with Map<>, but this version is much much lighter, // and so I hope better compilation performance (time and code quality). template<typename Scalar, typename Index, int StorageOrder> -class ei_blas_data_mapper +class blas_data_mapper { public: - ei_blas_data_mapper(Scalar* data, Index stride) : m_data(data), m_stride(stride) {} + blas_data_mapper(Scalar* data, Index stride) : m_data(data), m_stride(stride) {} EIGEN_STRONG_INLINE Scalar& operator()(Index i, Index j) { return m_data[StorageOrder==RowMajor ? j + i*m_stride : i + j*m_stride]; } protected: @@ -140,10 +142,10 @@ class ei_blas_data_mapper // lightweight helper class to access matrix coefficients (const version) template<typename Scalar, typename Index, int StorageOrder> -class ei_const_blas_data_mapper +class const_blas_data_mapper { public: - ei_const_blas_data_mapper(const Scalar* data, Index stride) : m_data(data), m_stride(stride) {} + const_blas_data_mapper(const Scalar* data, Index stride) : m_data(data), m_stride(stride) {} EIGEN_STRONG_INLINE const Scalar& operator()(Index i, Index j) const { return m_data[StorageOrder==RowMajor ? j + i*m_stride : i + j*m_stride]; } protected: @@ -155,9 +157,9 @@ class ei_const_blas_data_mapper /* Helper class to analyze the factors of a Product expression. * In particular it allows to pop out operator-, scalar multiples, * and conjugate */ -template<typename XprType> struct ei_blas_traits +template<typename XprType> struct blas_traits { - typedef typename ei_traits<XprType>::Scalar Scalar; + typedef typename traits<XprType>::Scalar Scalar; typedef const XprType& ExtractType; typedef XprType _ExtractType; enum { @@ -168,10 +170,10 @@ template<typename XprType> struct ei_blas_traits && ( /* Uncomment this when the low-level matrix-vector product functions support strided vectors bool(XprType::IsVectorAtCompileTime) || */ - int(ei_inner_stride_at_compile_time<XprType>::ret) == 1) + int(inner_stride_at_compile_time<XprType>::ret) == 1) ) ? 1 : 0 }; - typedef typename ei_meta_if<bool(HasUsableDirectAccess), + typedef typename meta_if<bool(HasUsableDirectAccess), ExtractType, typename _ExtractType::PlainObject >::ret DirectLinearAccessType; @@ -181,11 +183,11 @@ template<typename XprType> struct ei_blas_traits // pop conjugate template<typename Scalar, typename NestedXpr> -struct ei_blas_traits<CwiseUnaryOp<ei_scalar_conjugate_op<Scalar>, NestedXpr> > - : ei_blas_traits<NestedXpr> +struct blas_traits<CwiseUnaryOp<scalar_conjugate_op<Scalar>, NestedXpr> > + : blas_traits<NestedXpr> { - typedef ei_blas_traits<NestedXpr> Base; - typedef CwiseUnaryOp<ei_scalar_conjugate_op<Scalar>, NestedXpr> XprType; + typedef blas_traits<NestedXpr> Base; + typedef CwiseUnaryOp<scalar_conjugate_op<Scalar>, NestedXpr> XprType; typedef typename Base::ExtractType ExtractType; enum { @@ -193,16 +195,16 @@ struct ei_blas_traits<CwiseUnaryOp<ei_scalar_conjugate_op<Scalar>, NestedXpr> > NeedToConjugate = Base::NeedToConjugate ? 0 : IsComplex }; static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } - static inline Scalar extractScalarFactor(const XprType& x) { return ei_conj(Base::extractScalarFactor(x.nestedExpression())); } + static inline Scalar extractScalarFactor(const XprType& x) { return conj(Base::extractScalarFactor(x.nestedExpression())); } }; // pop scalar multiple template<typename Scalar, typename NestedXpr> -struct ei_blas_traits<CwiseUnaryOp<ei_scalar_multiple_op<Scalar>, NestedXpr> > - : ei_blas_traits<NestedXpr> +struct blas_traits<CwiseUnaryOp<scalar_multiple_op<Scalar>, NestedXpr> > + : blas_traits<NestedXpr> { - typedef ei_blas_traits<NestedXpr> Base; - typedef CwiseUnaryOp<ei_scalar_multiple_op<Scalar>, NestedXpr> XprType; + typedef blas_traits<NestedXpr> Base; + typedef CwiseUnaryOp<scalar_multiple_op<Scalar>, NestedXpr> XprType; typedef typename Base::ExtractType ExtractType; static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } static inline Scalar extractScalarFactor(const XprType& x) @@ -211,11 +213,11 @@ struct ei_blas_traits<CwiseUnaryOp<ei_scalar_multiple_op<Scalar>, NestedXpr> > // pop opposite template<typename Scalar, typename NestedXpr> -struct ei_blas_traits<CwiseUnaryOp<ei_scalar_opposite_op<Scalar>, NestedXpr> > - : ei_blas_traits<NestedXpr> +struct blas_traits<CwiseUnaryOp<scalar_opposite_op<Scalar>, NestedXpr> > + : blas_traits<NestedXpr> { - typedef ei_blas_traits<NestedXpr> Base; - typedef CwiseUnaryOp<ei_scalar_opposite_op<Scalar>, NestedXpr> XprType; + typedef blas_traits<NestedXpr> Base; + typedef CwiseUnaryOp<scalar_opposite_op<Scalar>, NestedXpr> XprType; typedef typename Base::ExtractType ExtractType; static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } static inline Scalar extractScalarFactor(const XprType& x) @@ -224,15 +226,15 @@ struct ei_blas_traits<CwiseUnaryOp<ei_scalar_opposite_op<Scalar>, NestedXpr> > // pop/push transpose template<typename NestedXpr> -struct ei_blas_traits<Transpose<NestedXpr> > - : ei_blas_traits<NestedXpr> +struct blas_traits<Transpose<NestedXpr> > + : blas_traits<NestedXpr> { typedef typename NestedXpr::Scalar Scalar; - typedef ei_blas_traits<NestedXpr> Base; + typedef blas_traits<NestedXpr> Base; typedef Transpose<NestedXpr> XprType; typedef Transpose<typename Base::_ExtractType> ExtractType; typedef Transpose<typename Base::_ExtractType> _ExtractType; - typedef typename ei_meta_if<bool(Base::HasUsableDirectAccess), + typedef typename meta_if<bool(Base::HasUsableDirectAccess), ExtractType, typename ExtractType::PlainObject >::ret DirectLinearAccessType; @@ -243,22 +245,24 @@ struct ei_blas_traits<Transpose<NestedXpr> > static inline Scalar extractScalarFactor(const XprType& x) { return Base::extractScalarFactor(x.nestedExpression()); } }; -template<typename T, bool HasUsableDirectAccess=ei_blas_traits<T>::HasUsableDirectAccess> -struct ei_extract_data_selector { +template<typename T, bool HasUsableDirectAccess=blas_traits<T>::HasUsableDirectAccess> +struct extract_data_selector { static const typename T::Scalar* run(const T& m) { - return &ei_blas_traits<T>::extract(m).const_cast_derived().coeffRef(0,0); // FIXME this should be .data() + return &blas_traits<T>::extract(m).const_cast_derived().coeffRef(0,0); // FIXME this should be .data() } }; template<typename T> -struct ei_extract_data_selector<T,false> { +struct extract_data_selector<T,false> { static typename T::Scalar* run(const T&) { return 0; } }; -template<typename T> const typename T::Scalar* ei_extract_data(const T& m) +template<typename T> const typename T::Scalar* extract_data(const T& m) { - return ei_extract_data_selector<T>::run(m); + return extract_data_selector<T>::run(m); } +} // end namespace internal + #endif // EIGEN_BLASUTIL_H |