Missing EIGEN_DEVICE_FUNCs to get gpu_basic passing with CUDA 9.

CUDA 9 seems to require labelling defaulted constructors as
`EIGEN_DEVICE_FUNC`, despite giving warnings that such labels are
ignored.  Without these labels, the `gpu_basic` test fails to
compile, with errors about calling `__host__` functions from
`__host__ __device__` functions.


(cherry picked from commit 998bab4b04)
This commit is contained in:
Antonio Sanchez
2021-08-31 08:32:21 -07:00
committed by Antonio Sánchez
parent b8cf1ed753
commit f03d3e7072
2 changed files with 9 additions and 9 deletions

View File

@@ -260,19 +260,19 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
}
template<int LoadMode>
inline PacketScalar packet(Index rowId, Index colId) const
EIGEN_DEVICE_FUNC inline PacketScalar packet(Index rowId, Index colId) const
{
return m_xpr.template packet<Unaligned>(rowId + m_startRow.value(), colId + m_startCol.value());
}
template<int LoadMode>
inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
EIGEN_DEVICE_FUNC inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
{
m_xpr.template writePacket<Unaligned>(rowId + m_startRow.value(), colId + m_startCol.value(), val);
}
template<int LoadMode>
inline PacketScalar packet(Index index) const
EIGEN_DEVICE_FUNC inline PacketScalar packet(Index index) const
{
return m_xpr.template packet<Unaligned>
(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
@@ -280,7 +280,7 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
}
template<int LoadMode>
inline void writePacket(Index index, const PacketScalar& val)
EIGEN_DEVICE_FUNC inline void writePacket(Index index, const PacketScalar& val)
{
m_xpr.template writePacket<Unaligned>
(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),