devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add support for replicate in CUDA

Browse Source
This commit is contained in:
Gael Guennebaud
2015-07-20 10:53:03 +02:00
parent 6799c26cd6
commit 88e352adac
4 changed files with 29 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
test/cuda_basic.cu
View File

@@ -47,6 +47,23 @@ struct coeff_wise {
}
};
template<typename T>
struct replicate {
EIGEN_DEVICE_FUNC
void operator()(int i, const typename T::Scalar* in, typename T::Scalar* out) const
{
using namespace Eigen;
T x1(in+i);
int step = x1.size() * 4;
int stride = 3 * step;
typedef Map<Array<typename T::Scalar,Dynamic,Dynamic> > MapType;
MapType(out+i*stride+0*step, x1.rows()*2, x1.cols()*2) = x1.replicate(2,2);
MapType(out+i*stride+1*step, x1.rows()*3, x1.cols()) = in[i] * x1.colwise().replicate(3);
MapType(out+i*stride+2*step, x1.rows(), x1.cols()*3) = in[i] * x1.rowwise().replicate(3);
}
};
template<typename T>
struct redux {
EIGEN_DEVICE_FUNC
@@ -117,7 +134,7 @@ void test_cuda_basic()
Eigen::VectorXf in, out;
#ifndef __CUDA_ARCH__
int data_size = nthreads * 16;
int data_size = nthreads * 512;
in.setRandom(data_size);
out.setRandom(data_size);
#endif
@@ -125,6 +142,9 @@ void test_cuda_basic()
CALL_SUBTEST( run_and_compare_to_cuda(coeff_wise<Vector3f>(), nthreads, in, out) );
CALL_SUBTEST( run_and_compare_to_cuda(coeff_wise<Array44f>(), nthreads, in, out) );
CALL_SUBTEST( run_and_compare_to_cuda(replicate<Array4f>(), nthreads, in, out) );
CALL_SUBTEST( run_and_compare_to_cuda(replicate<Array33f>(), nthreads, in, out) );
CALL_SUBTEST( run_and_compare_to_cuda(redux<Array4f>(), nthreads, in, out) );
CALL_SUBTEST( run_and_compare_to_cuda(redux<Matrix3f>(), nthreads, in, out) );