mirror of
https://gitlab.com/libeigen/eigen.git
synced 2026-04-10 11:34:33 +08:00
Implemented the SSE version of the gather and scatter packet primitives.
This commit is contained in:
@@ -111,6 +111,24 @@ template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<
|
||||
template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&numext::real_ref(*to), from.v); }
|
||||
template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&numext::real_ref(*to), from.v); }
|
||||
|
||||
|
||||
template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather<std::complex<float>, Packet2cf>(const std::complex<float>* from, int stride)
|
||||
{
|
||||
return Packet2cf(_mm_set_ps(std::imag(from[1*stride]), std::real(from[1*stride]),
|
||||
std::imag(from[0*stride]), std::real(from[0*stride])));
|
||||
}
|
||||
|
||||
template<> EIGEN_DEVICE_FUNC inline void pscatter<std::complex<float>, Packet2cf>(std::complex<float>* to, const Packet2cf& from, int stride)
|
||||
{
|
||||
/* for (int i = 0; i < 2; i+=2) {
|
||||
to[stride*i] = std::complex<float>(from.v[i], from.v[i+1]);
|
||||
}*/
|
||||
to[stride*0] = std::complex<float>(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 0)),
|
||||
_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 1)));
|
||||
to[stride*1] = std::complex<float>(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 2)),
|
||||
_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 3)));
|
||||
}
|
||||
|
||||
template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> * addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); }
|
||||
|
||||
template<> EIGEN_STRONG_INLINE std::complex<float> pfirst<Packet2cf>(const Packet2cf& a)
|
||||
|
||||
Reference in New Issue
Block a user