applying EIGEN_DECLARE_TEST to *gpu* tests

Also, a few minor fixes for GPU tests running in HIP mode.

1. Adding an include for hip/hip_runtime.h in the Macros.h file
   For HIP __host__ and __device__ are macros which are defined in hip headers.
   Their definitions need to be included before their use in the file.

2. Fixing the compile failure in TensorContractionGpu introduced by the commit to
   "Fuse computations into the Tensor contractions using output kernel"

3. Fixing a HIP/clang specific compile error by making the struct-member assignment explicit

2 files changed, 6 insertions, 7 deletions

diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h b/unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h
index 0fbffa34c..c7c443a59 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h
@@ -449,8 +449,7 @@ struct TensorContractionEvaluatorBase
     // tensor dimensions (i, j) into the original tensor dimensions.
     // TODO(ezhulenev): Add parameters required to infer output tensor index for
     // more complex contractions than 2x2 on internal dimension.
-    m_tensor_contraction_params = {
-        /**swapped_arguments=*/static_cast<int>(Layout) == RowMajor};
+    m_tensor_contraction_params.swapped_arguments = static_cast<int>(Layout) == RowMajor;
   }
 
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return m_dimensions; }
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorContractionGpu.h b/unsupported/Eigen/CXX11/src/Tensor/TensorContractionGpu.h
index a4f92ee44..b9956cd43 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorContractionGpu.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorContractionGpu.h
@@ -1215,16 +1215,16 @@ EigenFloatContractionKernel16x16(const LhsMapper lhs, const RhsMapper rhs,
 }
 
 
-template<typename Indices, typename LeftArgType, typename RightArgType>
-struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgType>, GpuDevice> :
-    public TensorContractionEvaluatorBase<TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgType>, GpuDevice> > {
+template<typename Indices, typename LeftArgType, typename RightArgType, typename OutputKernelType>
+struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgType, OutputKernelType>, GpuDevice> :
+    public TensorContractionEvaluatorBase<TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgType, OutputKernelType>, GpuDevice> > {
 
   typedef GpuDevice Device;
 
-  typedef TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgType>, Device> Self;
+  typedef TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgType, OutputKernelType>, Device> Self;
   typedef TensorContractionEvaluatorBase<Self> Base;
 
-  typedef TensorContractionOp<Indices, LeftArgType, RightArgType> XprType;
+  typedef TensorContractionOp<Indices, LeftArgType, RightArgType, OutputKernelType> XprType;
   typedef typename internal::remove_const<typename XprType::Scalar>::type Scalar;
   typedef typename XprType::Index Index;
   typedef typename XprType::CoeffReturnType CoeffReturnType;