Adding support for using Eigen in HIP kernels.

This commit enables the use of Eigen on HIP kernels / AMD GPUs. Support has been added along the same lines as what already exists for using Eigen in CUDA kernels / NVidia GPUs.

Application code needs to explicitly define EIGEN_USE_HIP when using Eigen in HIP kernels. This is because some of the CUDA headers get picked up by default during Eigen compile (irrespective of whether or not the underlying compiler is CUDACC/NVCC, for e.g. Eigen/src/Core/arch/CUDA/Half.h). In order to maintain this behavior, the EIGEN_USE_HIP macro is used to switch to using the HIP version of those header files (see Eigen/Core and unsupported/Eigen/CXX11/Tensor)


Use the "-DEIGEN_TEST_HIP" cmake option to enable the HIP specific unit tests.

1 files changed, 12 insertions, 0 deletions

diff --git a/unsupported/Eigen/CXX11/src/util/CXX11Meta.h b/unsupported/Eigen/CXX11/src/util/CXX11Meta.h
index 49d315a66..bb584e3f9 100644
--- a/unsupported/Eigen/CXX11/src/util/CXX11Meta.h
+++ b/unsupported/Eigen/CXX11/src/util/CXX11Meta.h
@@ -268,6 +268,9 @@ template<
   typename Reducer
 > struct reduce<Reducer>
 {
+  #if defined(EIGEN_HIPCC)
+  EIGEN_DEVICE_FUNC
+  #endif
   constexpr static inline int run() { return Reducer::Identity; }
 };
 
@@ -276,6 +279,9 @@ template<
   typename A
 > struct reduce<Reducer, A>
 {
+  #if defined(EIGEN_HIPCC)
+  EIGEN_DEVICE_FUNC
+  #endif
   constexpr static inline A run(A a) { return a; }
 };
 
@@ -285,6 +291,9 @@ template<
   typename... Ts
 > struct reduce<Reducer, A, Ts...>
 {
+  #if defined(EIGEN_HIPCC)
+  EIGEN_DEVICE_FUNC
+  #endif
   constexpr static inline auto run(A a, Ts... ts) -> decltype(Reducer::run(a, reduce<Reducer, Ts...>::run(ts...))) {
     return Reducer::run(a, reduce<Reducer, Ts...>::run(ts...));
   }
@@ -324,6 +333,9 @@ struct greater_equal_zero_op { template<typename A> constexpr static inline auto
 // together in front... (13.0 doesn't work with array_prod/array_reduce/... anyway, but 13.1
 // does...
 template<typename... Ts>
+#if defined(EIGEN_HIPCC)
+EIGEN_DEVICE_FUNC
+#endif
 constexpr inline decltype(reduce<product_op, Ts...>::run((*((Ts*)0))...)) arg_prod(Ts... ts)
 {
   return reduce<product_op, Ts...>::run(ts...);