1 files changed, 55 insertions, 2 deletions
diff --git a/unsupported/test/cxx11_tensor_random_cuda.cu b/unsupported/test/cxx11_tensor_random_cuda.cu
index 5d091de15..b3be199e1 100644
--- a/unsupported/test/cxx11_tensor_random_cuda.cu
+++ b/unsupported/test/cxx11_tensor_random_cuda.cu
@@ -13,10 +13,61 @@
 #define EIGEN_DEFAULT_DENSE_INDEX_TYPE int
 #define EIGEN_USE_GPU
 
+#if defined __CUDACC_VER__ && __CUDACC_VER__ >= 70500
+#include <cuda_fp16.h>
+#endif
 #include "main.h"
 #include <Eigen/CXX11/Tensor>
 
-static void test_default()
+
+void test_cuda_random_uniform()
+{
+  Tensor<float, 2> out(72,97);
+  out.setZero();
+
+  std::size_t out_bytes = out.size() * sizeof(float);
+
+  float* d_out;
+  cudaMalloc((void**)(&d_out), out_bytes);
+
+  Eigen::CudaStreamDevice stream;
+  Eigen::GpuDevice gpu_device(&stream);
+
+  Eigen::TensorMap<Eigen::Tensor<float, 2> > gpu_out(d_out, 72,97);
+
+  gpu_out.device(gpu_device) = gpu_out.random();
+
+  assert(cudaMemcpyAsync(out.data(), d_out, out_bytes, cudaMemcpyDeviceToHost, gpu_device.stream()) == cudaSuccess);
+  assert(cudaStreamSynchronize(gpu_device.stream()) == cudaSuccess);
+
+  // For now we just check thes code doesn't crash.
+  // TODO: come up with a valid test of randomness
+}
+
+
+void test_cuda_random_normal()
+{
+  Tensor<float, 2> out(72,97);
+  out.setZero();
+
+  std::size_t out_bytes = out.size() * sizeof(float);
+
+  float* d_out;
+  cudaMalloc((void**)(&d_out), out_bytes);
+
+  Eigen::CudaStreamDevice stream;
+  Eigen::GpuDevice gpu_device(&stream);
+
+  Eigen::TensorMap<Eigen::Tensor<float, 2> > gpu_out(d_out, 72,97);
+
+  Eigen::internal::NormalRandomGenerator<float> gen(true);
+  gpu_out.device(gpu_device) = gpu_out.random(gen);
+
+  assert(cudaMemcpyAsync(out.data(), d_out, out_bytes, cudaMemcpyDeviceToHost, gpu_device.stream()) == cudaSuccess);
+  assert(cudaStreamSynchronize(gpu_device.stream()) == cudaSuccess);
+}
+
+static void test_complex()
 {
   Tensor<std::complex<float>, 1> vec(6);
   vec.setRandom();
@@ -31,5 +82,7 @@ static void test_default()
 
 void test_cxx11_tensor_random_cuda()
 {
-  CALL_SUBTEST(test_default());
+  CALL_SUBTEST(test_cuda_random_uniform());
+  CALL_SUBTEST(test_cuda_random_normal());
+  CALL_SUBTEST(test_complex());
 }