mirror of
https://gitlab.com/libeigen/eigen.git
synced 2025-08-10 10:49:04 +08:00
Enable vectorized version on GPUs. The underlying bug has been fixed.
This commit is contained in:
parent
e289f44c56
commit
2b07018140
@ -218,7 +218,6 @@ struct InnerMostDimReducer<Self, Op, false, true> {
|
|||||||
}
|
}
|
||||||
};
|
};
|
||||||
|
|
||||||
#if !defined(EIGEN_USE_GPU) || !defined(__CUDACC__)
|
|
||||||
template <typename Self, typename Op>
|
template <typename Self, typename Op>
|
||||||
struct InnerMostDimReducer<Self, Op, true, true> {
|
struct InnerMostDimReducer<Self, Op, true, true> {
|
||||||
static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename Self::CoeffReturnType
|
static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename Self::CoeffReturnType
|
||||||
@ -258,7 +257,6 @@ struct InnerMostDimReducer<Self, Op, true, true> {
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
};
|
};
|
||||||
#endif
|
|
||||||
|
|
||||||
template <int DimIndex, typename Self, typename Op, bool vectorizable = (Self::InputPacketAccess && Self::ReducerTraits::PacketAccess)>
|
template <int DimIndex, typename Self, typename Op, bool vectorizable = (Self::InputPacketAccess && Self::ReducerTraits::PacketAccess)>
|
||||||
struct InnerMostDimPreserver {
|
struct InnerMostDimPreserver {
|
||||||
|
Loading…
x
Reference in New Issue
Block a user