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

* implement slice vectorization. Because it uses unaligned

Browse Source 
packet access, it is not certain that it will bring a performance
  improvement: benchmarking needed.
* improve logic choosing slice vectorization.
* fix typo in SSE packet math, causing crash in unaligned case.
* fix bug in Product, causing crash in unaligned case.
* add TEST_SSE3 CMake option.
This commit is contained in:
Benoit Jacob
2008-06-22 15:02:05 +00:00
parent 8cef541b5a
commit 8a967fb17c
5 changed files with 64 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
Eigen/src/Core/Assign.h
View File

@@ -52,6 +52,9 @@ private:
InnerSize = int(Derived::Flags)&RowMajorBit
? Derived::ColsAtCompileTime
: Derived::RowsAtCompileTime,
InnerMaxSize = int(Derived::Flags)&RowMajorBit
? Derived::MaxColsAtCompileTime
: Derived::MaxRowsAtCompileTime,
PacketSize = ei_packet_traits<typename Derived::Scalar>::size
};
@@ -60,7 +63,9 @@ private:
&& ((int(Derived::Flags)&RowMajorBit)==(int(OtherDerived::Flags)&RowMajorBit)),
MayInnerVectorize = MightVectorize && InnerSize!=Dynamic && int(InnerSize)%int(PacketSize)==0,
MayLinearVectorize = MightVectorize && (int(Derived::Flags) & int(OtherDerived::Flags) & LinearAccessBit),
MaySliceVectorize = MightVectorize && InnerSize==Dynamic
MaySliceVectorize = MightVectorize && InnerMaxSize==Dynamic /* slice vectorization can be slow, so we only
want it if the slices are big, which is indicated by InnerMaxSize rather than InnerSize, think of the case
of a dynamic block in a fixed-size matrix */
};
public:
@@ -349,7 +354,7 @@ struct ei_assign_impl<Derived1, Derived2, LinearVectorization, NoUnrolling>
template<typename Derived1, typename Derived2>
struct ei_assign_impl<Derived1, Derived2, LinearVectorization, CompleteUnrolling>
{
inline static void run(Derived1 &dst, const Derived2 &src)
static void run(Derived1 &dst, const Derived2 &src)
{
const int size = Derived1::SizeAtCompileTime;
const int packetSize = ei_packet_traits<typename Derived1::Scalar>::size;
@@ -383,8 +388,30 @@ struct ei_assign_impl<Derived1, Derived2, SliceVectorization, NoUnrolling>
{
static void run(Derived1 &dst, const Derived2 &src)
{
//FIXME unimplemented, so for now we fall back to non-vectorized path
ei_assign_impl<Derived1, Derived2, NoVectorization, NoUnrolling>::run(dst, src);
const int packetSize = ei_packet_traits<typename Derived1::Scalar>::size;
const bool rowMajor = Derived1::Flags&RowMajorBit;
const int innerSize = rowMajor ? dst.cols() : dst.rows();
const int outerSize = rowMajor ? dst.rows() : dst.cols();
const int alignedInnerSize = (innerSize/packetSize)*packetSize;
for(int i = 0; i < outerSize; i++)
{
// do the vectorizable part of the assignment
for (int index = 0; index<alignedInnerSize ; index+=packetSize)
{
const int row = rowMajor ? i : index;
const int col = rowMajor ? index : i;
dst.template writePacket<UnAligned>(row, col, src.template packet<UnAligned>(row, col));
}
// do the non-vectorizable part of the assignment
for (int index = alignedInnerSize; index<innerSize ; index++)
{
const int row = rowMajor ? i : index;
const int col = rowMajor ? index : i;
dst.coeffRef(row, col) = src.coeff(row, col);
}
}
}
};