GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-12 19:59:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

Implemented protate() for CUDA

Browse Source
This commit is contained in:
Benoit Steiner 2016-02-19 15:16:54 +00:00
parent cd042dbbfd
commit dc26459b99
1 changed files with 29 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
Eigen/src/Core/arch/CUDA/PacketMath.h
View File

@ -272,6 +272,35 @@ template<> EIGEN_DEVICE_FUNC inline double predux_mul<double2>(const double2& a)
return a.x * a.y; return a.x * a.y;
} }
template<size_t offset>
struct protate_impl<offset, float4>
{
static float4 run(const float4& a) {
if (offset == 0) {
return make_float4(a.x, a.y, a.z, a.w);
}
if (offset == 1) {
return make_float4(a.w, a.x, a.y, a.z);
}
if (offset == 2) {
return make_float4(a.z, a.w, a.x, a.y);
}
return make_float4(a.y, a.z, a.w, a.x);
}
};
template<size_t offset>
struct protate_impl<offset, double2>
{
static double2 run(const double2& a) {
if (offset == 0) {
return make_double2(a.x, a.y);
}
return make_double2(a.y, a.x);
}
};
template<> EIGEN_DEVICE_FUNC inline float4 pabs<float4>(const float4& a) { template<> EIGEN_DEVICE_FUNC inline float4 pabs<float4>(const float4& a) {
return make_float4(fabsf(a.x), fabsf(a.y), fabsf(a.z), fabsf(a.w)); return make_float4(fabsf(a.x), fabsf(a.y), fabsf(a.z), fabsf(a.w));
} }