devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

merge

Browse Source
This commit is contained in:
Gael Guennebaud
2016-09-22 22:32:55 +02:00
parent ba0f844d6b 9bcdc8b756
commit 8f2bdde373
27 changed files with 374 additions and 108 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Eigen/src/Core/CwiseNullaryOp.h
View File

@@ -220,7 +220,7 @@ DenseBase<Derived>::Constant(const Scalar& value)
*
* The function generates 'size' equally spaced values in the closed interval [low,high].
* This particular version of LinSpaced() uses sequential access, i.e. vector access is
* assumed to be a(0), a(1), ..., a(size). This assumption allows for better vectorization
* assumed to be a(0), a(1), ..., a(size-1). This assumption allows for better vectorization
* and yields faster code than the random access version.
*
* When size is set to 1, a vector of length 1 containing 'high' is returned.
@@ -389,7 +389,7 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(Index newSize, con
/**
* \brief Sets a linearly spaced vector.
*
* The function fill *this with equally spaced values in the closed interval [low,high].
* The function fills *this with equally spaced values in the closed interval [low,high].
* When size is set to 1, a vector of length 1 containing 'high' is returned.
*
* \only_for_vectors

18
Eigen/src/Core/GeneralProduct.h
View File

@@ -159,20 +159,20 @@ struct gemv_static_vector_if<Scalar,Size,Dynamic,true>
template<typename Scalar,int Size,int MaxSize>
struct gemv_static_vector_if<Scalar,Size,MaxSize,true>
{
#if EIGEN_MAX_STATIC_ALIGN_BYTES!=0
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize),0> m_data;
EIGEN_STRONG_INLINE Scalar* data() { return m_data.array; }
#else
// Some architectures cannot align on the stack,
// => let's manually enforce alignment by allocating more data and return the address of the first aligned element.
enum {
ForceAlignment = internal::packet_traits<Scalar>::Vectorizable,
PacketSize = internal::packet_traits<Scalar>::size
};
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize)+(ForceAlignment?PacketSize:0),0> m_data;
#if EIGEN_MAX_STATIC_ALIGN_BYTES!=0
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize),0,EIGEN_PLAIN_ENUM_MIN(AlignedMax,PacketSize)> m_data;
EIGEN_STRONG_INLINE Scalar* data() { return m_data.array; }
#else
// Some architectures cannot align on the stack,
// => let's manually enforce alignment by allocating more data and return the address of the first aligned element.
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize)+(ForceAlignment?EIGEN_MAX_ALIGN_BYTES:0),0> m_data;
EIGEN_STRONG_INLINE Scalar* data() {
return ForceAlignment
? reinterpret_cast<Scalar*>((internal::UIntPtr(m_data.array) & ~(size_t(EIGEN_MAX_ALIGN_BYTES-1))) + EIGEN_MAX_ALIGN_BYTES)
? reinterpret_cast<Scalar*>((internal::UIntPtr(m_data.array) & ~(std::size_t(EIGEN_MAX_ALIGN_BYTES-1))) + EIGEN_MAX_ALIGN_BYTES)
: m_data.array;
}
#endif
@@ -207,7 +207,7 @@ template<> struct gemv_dense_selector<OnTheRight,ColMajor,true>
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType;
typedef Map<Matrix<ResScalar,Dynamic,1>, Aligned> MappedDest;
typedef Map<Matrix<ResScalar,Dynamic,1>, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits<ResScalar>::size)> MappedDest;
ActualLhsType actualLhs = LhsBlasTraits::extract(lhs);
ActualRhsType actualRhs = RhsBlasTraits::extract(rhs);

8
Eigen/src/Core/MathFunctionsImpl.h
View File

@@ -29,8 +29,12 @@ T generic_fast_tanh_float(const T& a_x)
// this range is +/-1.0f in single-precision.
const T plus_9 = pset1<T>(9.f);
const T minus_9 = pset1<T>(-9.f);
const T x = pmax(minus_9, pmin(plus_9, a_x));
// NOTE GCC prior to 6.3 might improperly optimize this max/min
// step such that if a_x is nan, x will be either 9 or -9,
// and tanh will return 1 or -1 instead of nan.
// This is supposed to be fixed in gcc6.3,
// see: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72867
const T x = pmax(minus_9,pmin(plus_9,a_x));
// The monomial coefficients of the numerator polynomial (odd).
const T alpha_1 = pset1<T>(4.89352455891786e-03f);
const T alpha_3 = pset1<T>(6.37261928875436e-04f);

4
Eigen/src/Core/MatrixBase.h
View File

@@ -330,15 +330,11 @@ template<typename Derived> class MatrixBase
/////////// LU module ///////////
EIGEN_DEVICE_FUNC
inline const FullPivLU<PlainObject> fullPivLu() const;
EIGEN_DEVICE_FUNC
inline const PartialPivLU<PlainObject> partialPivLu() const;
EIGEN_DEVICE_FUNC
inline const PartialPivLU<PlainObject> lu() const;
EIGEN_DEVICE_FUNC
inline const Inverse<Derived> inverse() const;
template<typename ResultType>

88
Eigen/src/Core/arch/CUDA/Complex.h Normal file
View File

@@ -0,0 +1,88 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_COMPLEX_CUDA_H
#define EIGEN_COMPLEX_CUDA_H
// clang-format off
namespace Eigen {
namespace internal {
#if defined(__CUDACC__) && defined(EIGEN_USE_GPU)
// Many std::complex methods such as operator+, operator-, operator* and
// operator/ are not constexpr. Due to this, clang does not treat them as device
// functions and thus Eigen functors making use of these operators fail to
// compile. Here, we manually specialize these functors for complex types when
// building for CUDA to avoid non-constexpr methods.
template<typename T> struct scalar_sum_op<std::complex<T>> {
typedef typename std::complex<T> result_type;
EIGEN_EMPTY_STRUCT_CTOR(scalar_sum_op)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const std::complex<T> operator() (const std::complex<T>& a, const std::complex<T>& b) const {
return std::complex<T>(numext::real(a) + numext::real(b),
numext::imag(a) + numext::imag(b));
}
};
template<typename T> struct scalar_difference_op<std::complex<T>> {
typedef typename std::complex<T> result_type;
EIGEN_EMPTY_STRUCT_CTOR(scalar_difference_op)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const std::complex<T> operator() (const std::complex<T>& a, const std::complex<T>& b) const {
return std::complex<T>(numext::real(a) - numext::real(b),
numext::imag(a) - numext::imag(b));
}
};
template<typename T> struct scalar_product_op<std::complex<T>, std::complex<T>> {
enum {
Vectorizable = packet_traits<std::complex<T>>::HasMul
};
typedef typename std::complex<T> result_type;
EIGEN_EMPTY_STRUCT_CTOR(scalar_product_op)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const std::complex<T> operator() (const std::complex<T>& a, const std::complex<T>& b) const {
const T a_real = numext::real(a);
const T a_imag = numext::imag(a);
const T b_real = numext::real(b);
const T b_imag = numext::imag(b);
return std::complex<T>(a_real * b_real - a_imag * b_imag,
a_real * b_imag + a_imag * b_real);
}
};
template<typename T> struct scalar_quotient_op<std::complex<T>, std::complex<T>> {
enum {
Vectorizable = packet_traits<std::complex<T>>::HasDiv
};
typedef typename std::complex<T> result_type;
EIGEN_EMPTY_STRUCT_CTOR(scalar_quotient_op)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const std::complex<T> operator() (const std::complex<T>& a, const std::complex<T>& b) const {
const T a_real = numext::real(a);
const T a_imag = numext::imag(a);
const T b_real = numext::real(b);
const T b_imag = numext::imag(b);
const T norm = T(1) / (b_real * b_real + b_imag * b_imag);
return std::complex<T>((a_real * b_real + a_imag * b_imag) * norm,
(a_imag * b_real - a_real * b_imag) * norm);
}
};
#endif
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_COMPLEX_CUDA_H

2
Eigen/src/Core/products/SelfadjointMatrixVector.h
View File

@@ -179,7 +179,7 @@ struct selfadjoint_product_impl<Lhs,LhsMode,false,Rhs,0,true>
{
typedef typename Dest::Scalar ResScalar;
typedef typename Rhs::Scalar RhsScalar;
typedef Map<Matrix<ResScalar,Dynamic,1>, Aligned> MappedDest;
typedef Map<Matrix<ResScalar,Dynamic,1>, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits<ResScalar>::size)> MappedDest;
eigen_assert(dest.rows()==a_lhs.rows() && dest.cols()==a_rhs.cols());

2
Eigen/src/Core/products/TriangularMatrixVector.h
View File

@@ -216,7 +216,7 @@ template<int Mode> struct trmv_selector<Mode,ColMajor>
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType;
typedef Map<Matrix<ResScalar,Dynamic,1>, Aligned> MappedDest;
typedef Map<Matrix<ResScalar,Dynamic,1>, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits<ResScalar>::size)> MappedDest;
typename internal::add_const_on_value_type<ActualLhsType>::type actualLhs = LhsBlasTraits::extract(lhs);
typename internal::add_const_on_value_type<ActualRhsType>::type actualRhs = RhsBlasTraits::extract(rhs);

8
Eigen/src/Core/util/XprHelper.h
View File

@@ -671,6 +671,14 @@ struct scalar_div_cost {
enum { value = 8*NumTraits<T>::MulCost };
};
template<typename T,bool Vectorized>
struct scalar_div_cost<std::complex<T>, Vectorized> {
enum { value = 2*scalar_div_cost<T>::value
+ 6*NumTraits<T>::MulCost
+ 3*NumTraits<T>::AddCost
};
};
template<bool Vectorized>
struct scalar_div_cost<signed long,Vectorized,typename conditional<sizeof(long)==8,void,false_type>::type> { enum { value = 24 }; };

14
Eigen/src/Geometry/EulerAngles.h
View File

@@ -55,7 +55,12 @@ MatrixBase<Derived>::eulerAngles(Index a0, Index a1, Index a2) const
res[0] = atan2(coeff(j,i), coeff(k,i));
if((odd && res[0]<Scalar(0)) || ((!odd) && res[0]>Scalar(0)))
{
res[0] = (res[0] > Scalar(0)) ? res[0] - Scalar(EIGEN_PI) : res[0] + Scalar(EIGEN_PI);
if(res[0] > Scalar(0)) {
res[0] -= Scalar(EIGEN_PI);
}
else {
res[0] += Scalar(EIGEN_PI);
}
Scalar s2 = Vector2(coeff(j,i), coeff(k,i)).norm();
res[1] = -atan2(s2, coeff(i,i));
}
@@ -84,7 +89,12 @@ MatrixBase<Derived>::eulerAngles(Index a0, Index a1, Index a2) const
res[0] = atan2(coeff(j,k), coeff(k,k));
Scalar c2 = Vector2(coeff(i,i), coeff(i,j)).norm();
if((odd && res[0]<Scalar(0)) || ((!odd) && res[0]>Scalar(0))) {
res[0] = (res[0] > Scalar(0)) ? res[0] - Scalar(EIGEN_PI) : res[0] + Scalar(EIGEN_PI);
if(res[0] > Scalar(0)) {
res[0] -= Scalar(EIGEN_PI);
}
else {
res[0] += Scalar(EIGEN_PI);
}
res[1] = atan2(-coeff(i,k), -c2);
}
else

4
Eigen/src/Householder/Householder.h
View File

@@ -119,7 +119,7 @@ void MatrixBase<Derived>::applyHouseholderOnTheLeft(
{
*this *= Scalar(1)-tau;
}
else
else if(tau!=Scalar(0))
{
Map<typename internal::plain_row_type<PlainObject>::type> tmp(workspace,cols());
Block<Derived, EssentialPart::SizeAtCompileTime, Derived::ColsAtCompileTime> bottom(derived(), 1, 0, rows()-1, cols());
@@ -156,7 +156,7 @@ void MatrixBase<Derived>::applyHouseholderOnTheRight(
{
*this *= Scalar(1)-tau;
}
else
else if(tau!=Scalar(0))
{
Map<typename internal::plain_col_type<PlainObject>::type> tmp(workspace,rows());
Block<Derived, Derived::RowsAtCompileTime, EssentialPart::SizeAtCompileTime> right(derived(), 0, 1, rows(), cols()-1);

2
Eigen/src/LU/FullPivLU.h
View File

@@ -879,7 +879,7 @@ struct Assignment<DstXprType, Inverse<FullPivLU<MatrixType> >, internal::assign_
*
* \sa class FullPivLU
*/
template<typename Derived> EIGEN_DEVICE_FUNC
template<typename Derived>
inline const FullPivLU<typename MatrixBase<Derived>::PlainObject>
MatrixBase<Derived>::fullPivLu() const
{

2
Eigen/src/LU/InverseImpl.h
View File

@@ -327,7 +327,7 @@ struct Assignment<DstXprType, Inverse<XprType>, internal::assign_op<typename Dst
*
* \sa computeInverseAndDetWithCheck()
*/
template<typename Derived> EIGEN_DEVICE_FUNC
template<typename Derived>
inline const Inverse<Derived> MatrixBase<Derived>::inverse() const
{
EIGEN_STATIC_ASSERT(!NumTraits<Scalar>::IsInteger,THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES)

4
Eigen/src/LU/PartialPivLU.h
View File

@@ -584,7 +584,7 @@ struct Assignment<DstXprType, Inverse<PartialPivLU<MatrixType> >, internal::assi
*
* \sa class PartialPivLU
*/
template<typename Derived> EIGEN_DEVICE_FUNC
template<typename Derived>
inline const PartialPivLU<typename MatrixBase<Derived>::PlainObject>
MatrixBase<Derived>::partialPivLu() const
{
@@ -599,7 +599,7 @@ MatrixBase<Derived>::partialPivLu() const
*
* \sa class PartialPivLU
*/
template<typename Derived> EIGEN_DEVICE_FUNC
template<typename Derived>
inline const PartialPivLU<typename MatrixBase<Derived>::PlainObject>
MatrixBase<Derived>::lu() const
{