diff options
Diffstat (limited to 'Eigen/src/Core/functors/AssignmentFunctors.h')
| -rw-r--r-- | Eigen/src/Core/functors/AssignmentFunctors.h | 13 |
1 files changed, 11 insertions, 2 deletions
diff --git a/Eigen/src/Core/functors/AssignmentFunctors.h b/Eigen/src/Core/functors/AssignmentFunctors.h index 4153b877c..bf64ef4ed 100644 --- a/Eigen/src/Core/functors/AssignmentFunctors.h +++ b/Eigen/src/Core/functors/AssignmentFunctors.h @@ -144,7 +144,7 @@ template<typename Scalar> struct swap_assign_op { EIGEN_EMPTY_STRUCT_CTOR(swap_assign_op) EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Scalar& a, const Scalar& b) const { -#ifdef __CUDACC__ +#ifdef EIGEN_GPUCC // FIXME is there some kind of cuda::swap? Scalar t=b; const_cast<Scalar&>(b)=a; a=t; #else @@ -157,7 +157,16 @@ template<typename Scalar> struct functor_traits<swap_assign_op<Scalar> > { enum { Cost = 3 * NumTraits<Scalar>::ReadCost, - PacketAccess = packet_traits<Scalar>::Vectorizable + PacketAccess = + #if defined(EIGEN_VECTORIZE_AVX) && EIGEN_COMP_CLANG && (EIGEN_COMP_CLANG<800 || defined(__apple_build_version__)) + // This is a partial workaround for a bug in clang generating bad code + // when mixing 256/512 bits loads and 128 bits moves. + // See http://eigen.tuxfamily.org/bz/show_bug.cgi?id=1684 + // https://bugs.llvm.org/show_bug.cgi?id=40815 + 0 + #else + packet_traits<Scalar>::Vectorizable + #endif }; }; |