Added initial experimental support for explicit vectorization.

Currently only the following platform/operations are supported:
 - SSE2 compatible architecture
 - compiler compatible with intel's SSE2 intrinsics
 - float, double and int data types
 - fixed size matrices with a storage major dimension multiple of 4 (or 2 for double)
 - scalar-matrix product, component wise: +,-,*,min,max
 - matrix-matrix product only if the left matrix is vectorizable and column major
   or the right matrix is vectorizable and row major, e.g.:
   a.transpose() * b is not vectorized with the default column major storage.
To use it you must define EIGEN_VECTORIZE and EIGEN_INTEL_PLATFORM.
This commit is contained in:
Gael Guennebaud
2008-04-09 12:31:55 +00:00
parent 4920f2011e
commit 1985fb0551
25 changed files with 436 additions and 93 deletions

View File

@@ -60,7 +60,9 @@ struct ei_traits<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >
ColsAtCompileTime = Lhs::ColsAtCompileTime,
MaxRowsAtCompileTime = Lhs::MaxRowsAtCompileTime,
MaxColsAtCompileTime = Lhs::MaxColsAtCompileTime,
Flags = Lhs::Flags | Rhs::Flags,
Flags = ((Lhs::Flags | Rhs::Flags) & ~VectorizableBit)
| (ei_functor_traits<BinaryOp>::IsVectorizable && ((Lhs::Flags&RowMajorBit)==(Rhs::Flags&RowMajorBit))
? (Lhs::Flags & Rhs::Flags & VectorizableBit) : 0),
CoeffReadCost = Lhs::CoeffReadCost + Rhs::CoeffReadCost + ei_functor_traits<BinaryOp>::Cost
};
};
@@ -89,6 +91,11 @@ class CwiseBinaryOp : ei_no_assignment_operator,
return m_functor(m_lhs.coeff(row, col), m_rhs.coeff(row, col));
}
PacketScalar _packetCoeff(int row, int col) const
{
return m_functor.packetOp(m_lhs.packetCoeff(row, col), m_rhs.packetCoeff(row, col));
}
protected:
const typename Lhs::XprCopy m_lhs;
const typename Rhs::XprCopy m_rhs;