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

Adds missing EIGEN_STRONG_INLINE to support MSVC properly inlining small vector calculations

Browse Source 
When working with MSVC often small vector operations are not properly inlined. This behaviour is observed even on the most recent compiler versions.
This commit is contained in:
Basil Fierz
2017-10-26 22:44:28 +02:00
parent 746a6b7b81
commit 624df50945
4 changed files with 26 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
Eigen/src/Core/Dot.h
View File

@@ -31,7 +31,8 @@ struct dot_nocheck
typedef scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> conj_prod;
typedef typename conj_prod::result_type ResScalar;
EIGEN_DEVICE_FUNC
static inline ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
EIGEN_STRONG_INLINE
static ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
{
return a.template binaryExpr<conj_prod>(b).sum();
}
@@ -43,7 +44,8 @@ struct dot_nocheck<T, U, true>
typedef scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> conj_prod;
typedef typename conj_prod::result_type ResScalar;
EIGEN_DEVICE_FUNC
static inline ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
EIGEN_STRONG_INLINE
static ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
{
return a.transpose().template binaryExpr<conj_prod>(b).sum();
}
@@ -65,6 +67,7 @@ struct dot_nocheck<T, U, true>
template<typename Derived>
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE
typename ScalarBinaryOpTraits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
MatrixBase<Derived>::dot(const MatrixBase<OtherDerived>& other) const
{
@@ -102,7 +105,7 @@ EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename NumTraits<typename internal::trai
* \sa lpNorm(), dot(), squaredNorm()
*/
template<typename Derived>
EIGEN_DEVICE_FUNC inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
{
return numext::sqrt(squaredNorm());
}
@@ -117,7 +120,7 @@ EIGEN_DEVICE_FUNC inline typename NumTraits<typename internal::traits<Derived>::
* \sa norm(), normalize()
*/
template<typename Derived>
EIGEN_DEVICE_FUNC inline const typename MatrixBase<Derived>::PlainObject
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
MatrixBase<Derived>::normalized() const
{
typedef typename internal::nested_eval<Derived,2>::type _Nested;
@@ -139,7 +142,7 @@ MatrixBase<Derived>::normalized() const
* \sa norm(), normalized()
*/
template<typename Derived>
EIGEN_DEVICE_FUNC inline void MatrixBase<Derived>::normalize()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void MatrixBase<Derived>::normalize()
{
RealScalar z = squaredNorm();
// NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU
@@ -160,7 +163,7 @@ EIGEN_DEVICE_FUNC inline void MatrixBase<Derived>::normalize()
* \sa stableNorm(), stableNormalize(), normalized()
*/
template<typename Derived>
EIGEN_DEVICE_FUNC inline const typename MatrixBase<Derived>::PlainObject
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
MatrixBase<Derived>::stableNormalized() const
{
typedef typename internal::nested_eval<Derived,3>::type _Nested;
@@ -185,7 +188,7 @@ MatrixBase<Derived>::stableNormalized() const
* \sa stableNorm(), stableNormalized(), normalize()
*/
template<typename Derived>
EIGEN_DEVICE_FUNC inline void MatrixBase<Derived>::stableNormalize()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void MatrixBase<Derived>::stableNormalize()
{
RealScalar w = cwiseAbs().maxCoeff();
RealScalar z = (derived()/w).squaredNorm();