Added support for CUDA 9.0.

This commit is contained in:
Benoit Steiner
2017-08-31 02:49:39 +00:00
parent 304ef29571
commit a4089991eb
17 changed files with 175 additions and 63 deletions

View File

@@ -15,6 +15,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
template <int Layout>

View File

@@ -16,6 +16,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
void test_cuda_conversion() {

View File

@@ -14,6 +14,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
void test_cuda_nullary() {

View File

@@ -14,6 +14,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
template<typename T>

View File

@@ -17,6 +17,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
typedef Tensor<float, 1>::DimensionPair DimPair;

View File

@@ -15,6 +15,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
void test_cuda_nullary() {

View File

@@ -16,6 +16,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
using Eigen::RowMajor;

View File

@@ -16,6 +16,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
template<typename>

View File

@@ -16,6 +16,12 @@
#include "main.h"
#include <Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
void test_cuda_random_uniform()
{

View File

@@ -15,6 +15,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
template<typename Type, int DataLayout>
static void test_full_reductions() {

View File

@@ -16,6 +16,12 @@
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
// The EIGEN_CUDACC_VER macro is provided by
// unsupported/Eigen/CXX11/Tensor included above
#if defined EIGEN_CUDACC_VER && EIGEN_CUDACC_VER >= 70500
#include <cuda_fp16.h>
#endif
using Eigen::Tensor;
typedef Tensor<float, 1>::DimensionPair DimPair;