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

* add bench/benchVecAdd.cpp by Gael, fix crash (ei_pload on non-aligned)

Browse Source 
* introduce packet(int), make use of it in linear vectorized paths
  --> completely fixes the slowdown noticed in benchVecAdd.
* generalize coeff(int) to linear-access xprs
* clarify the access flag bits
* rework api dox in Coeffs.h and util/Constants.h
* improve certain expressions's flags, allowing more vectorization
* fix bug in Block: start(int) and end(int) returned dyn*dyn size
* fix bug in Block: just because the Eval type has packet access
  doesn't imply the block xpr should have it too.
This commit is contained in:
Benoit Jacob
2008-06-26 16:06:41 +00:00
parent 5b0da4b778
commit 25ba9f377c
23 changed files with 558 additions and 264 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
Eigen/src/Core/Matrix.h
View File

@@ -128,6 +128,11 @@ class Matrix : public MatrixBase<Matrix<_Scalar, _Rows, _Cols, _MaxRows, _MaxCol
return m_storage.data()[row + col * m_storage.rows()];
}
inline const Scalar& _coeff(int index) const
{
return m_storage.data()[index];
}
inline Scalar& _coeffRef(int row, int col)
{
if(Flags & RowMajorBit)
@@ -136,20 +141,33 @@ class Matrix : public MatrixBase<Matrix<_Scalar, _Rows, _Cols, _MaxRows, _MaxCol
return m_storage.data()[row + col * m_storage.rows()];
}
inline Scalar& _coeffRef(int index)
{
return m_storage.data()[index];
}
template<int LoadMode>
inline PacketScalar _packet(int row, int col) const
{
ei_internal_assert(Flags & PacketAccessBit);
if(Flags & RowMajorBit)
if (LoadMode==Aligned)
return ei_pload(&m_storage.data()[col + row * m_storage.cols()]);
return ei_pload(m_storage.data() + col + row * m_storage.cols());
else
return ei_ploadu(&m_storage.data()[col + row * m_storage.cols()]);
return ei_ploadu(m_storage.data() + col + row * m_storage.cols());
else
if (LoadMode==Aligned)
return ei_pload(&m_storage.data()[row + col * m_storage.rows()]);
return ei_pload(m_storage.data() + row + col * m_storage.rows());
else
return ei_ploadu(&m_storage.data()[row + col * m_storage.rows()]);
return ei_ploadu(m_storage.data() + row + col * m_storage.rows());
}
template<int LoadMode>
inline PacketScalar _packet(int index) const
{
if (LoadMode==Aligned)
return ei_pload(m_storage.data() + index);
else
return ei_ploadu(m_storage.data() + index);
}
template<int StoreMode>
@@ -158,14 +176,23 @@ class Matrix : public MatrixBase<Matrix<_Scalar, _Rows, _Cols, _MaxRows, _MaxCol
ei_internal_assert(Flags & PacketAccessBit);
if(Flags & RowMajorBit)
if (StoreMode==Aligned)
ei_pstore(&m_storage.data()[col + row * m_storage.cols()], x);
ei_pstore(m_storage.data() + col + row * m_storage.cols(), x);
else
ei_pstoreu(&m_storage.data()[col + row * m_storage.cols()], x);
ei_pstoreu(m_storage.data() + col + row * m_storage.cols(), x);
else
if (StoreMode==Aligned)
ei_pstore(&m_storage.data()[row + col * m_storage.rows()], x);
ei_pstore(m_storage.data() + row + col * m_storage.rows(), x);
else
ei_pstoreu(&m_storage.data()[row + col * m_storage.rows()], x);
ei_pstoreu(m_storage.data() + row + col * m_storage.rows(), x);
}
template<int StoreMode>
inline void _writePacket(int index, const PacketScalar& x)
{
if (StoreMode==Aligned)
ei_pstore(m_storage.data() + index, x);
else
ei_pstoreu(m_storage.data() + index, x);
}
public: