From 0ac1fc52ddc4042e1a4e07061207db5a2223affa Mon Sep 17 00:00:00 2001
From: Antonio Sanchez <cantonios@google.com>
Date: Thu, 13 Mar 2025 20:47:31 -0700
Subject: [PATCH] Fix CUDA clang again with new C++11 usages

---
 Eigen/src/Core/arch/CUDA/Complex.h                        | 4 ++--
 test/CMakeLists.txt                                       | 6 +-----
 .../Eigen/CXX11/src/Tensor/TensorContractionCuda.h        | 2 +-
 unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h    | 8 +++-----
 unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h           | 2 +-
 unsupported/test/CMakeLists.txt                           | 6 +-----
 6 files changed, 9 insertions(+), 19 deletions(-)

diff --git a/Eigen/src/Core/arch/CUDA/Complex.h b/Eigen/src/Core/arch/CUDA/Complex.h
index 526e59fd0..57d1201f4 100644
--- a/Eigen/src/Core/arch/CUDA/Complex.h
+++ b/Eigen/src/Core/arch/CUDA/Complex.h
@@ -55,7 +55,7 @@ template<typename T> struct scalar_difference_op<std::complex<T>, std::complex<T
 // Product
 template<typename T> struct scalar_product_op<const std::complex<T>, const std::complex<T> >  : binary_op_base<const std::complex<T>, const std::complex<T> > {
   enum {
-    Vectorizable = packet_traits<std::complex<T>>::HasMul
+    Vectorizable = packet_traits<std::complex<T> >::HasMul
   };
   typedef typename std::complex<T> result_type;
 
@@ -76,7 +76,7 @@ template<typename T> struct scalar_product_op<std::complex<T>, std::complex<T> >
 // Quotient
 template<typename T> struct scalar_quotient_op<const std::complex<T>, const std::complex<T> > : binary_op_base<const std::complex<T>, const std::complex<T> > {
   enum {
-    Vectorizable = packet_traits<std::complex<T>>::HasDiv
+    Vectorizable = packet_traits<std::complex<T> >::HasDiv
   };
   typedef typename std::complex<T> result_type;
 
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index 014e6a432..a60d3912b 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -371,11 +371,7 @@ if (EIGEN_TEST_CUDA)
     string(REPLACE "-Wnon-virtual-dtor" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
     string(REPLACE "-fno-check-new" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
     if( (NOT EIGEN_TEST_CXX11) OR (CMAKE_VERSION VERSION_LESS 3.3))
-      if (MSVC)
-        string(APPEND EIGEN_CUDA_CXX11_FLAGS " /std:c++11")
-      else()
-        string(APPEND EIGEN_CUDA_CXX11_FLAGS " -std=c++11")
-      endif(MSVC)
+      string(APPEND EIGEN_CUDA_CXX_FLAGS " -std=c++11")
     endif()
 
     if(EIGEN_TEST_CUDA_CLANG)
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorContractionCuda.h b/unsupported/Eigen/CXX11/src/Tensor/TensorContractionCuda.h
index b78467b2a..678423b44 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorContractionCuda.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorContractionCuda.h
@@ -1220,7 +1220,7 @@ struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgT
   typedef typename PacketType<CoeffReturnType, GpuDevice>::type PacketReturnType;
 
   enum {
-    Layout = TensorEvaluator<LeftArgType, Device>::Layout,
+    Layout = TensorEvaluator<LeftArgType, Device>::Layout
   };
 
   // Most of the code is assuming that both input tensors are ColMajor. If the
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h b/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
index 8b0e9bf3b..620f523b6 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorConvolution.h
@@ -606,7 +606,7 @@ __global__ void EigenConvolutionKernel1D(
     }
     __syncthreads();
   }
-};
+}
 
 template <typename InputEvaluator, typename Index, typename InputDims,
           int StaticKernelSizeX, int StaticKernelSizeY>
@@ -674,7 +674,7 @@ __global__ void EigenConvolutionKernel2D(
 
     __syncthreads();
   }
-};
+}
 
 template <typename InputEvaluator, typename Index, typename InputDims>
 __global__ void EigenConvolutionKernel3D(
@@ -740,9 +740,7 @@ __global__ void EigenConvolutionKernel3D(
     }
     __syncthreads();
   }
-};
-
-
+}
 
 template<typename Indices, typename InputArgType, typename KernelArgType>
 struct TensorEvaluator<const TensorConvolutionOp<Indices, InputArgType, KernelArgType>, GpuDevice>
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h b/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h
index 615559d44..bf8196ae1 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorMeta.h
@@ -78,7 +78,7 @@ struct PacketType<half, GpuDevice> {
     HasLog    = 1,
     HasLog1p  = 0,
     HasLog10  = 0,
-    HasPow    = 1,
+    HasPow    = 1
   };
 };
 #endif
diff --git a/unsupported/test/CMakeLists.txt b/unsupported/test/CMakeLists.txt
index a896b42e7..6d32e60d9 100644
--- a/unsupported/test/CMakeLists.txt
+++ b/unsupported/test/CMakeLists.txt
@@ -218,11 +218,7 @@ if(CUDA_FOUND AND EIGEN_TEST_CUDA)
   message(STATUS "Flags used to compile cuda code: " ${CMAKE_CXX_FLAGS})
 
   if( (NOT EIGEN_TEST_CXX11) OR (CMAKE_VERSION VERSION_LESS 3.3))
-    if (MSVC)
-      string(APPEND EIGEN_CUDA_CXX11_FLAGS " /std:c++11")
-    else()
-      string(APPEND EIGEN_CUDA_CXX11_FLAGS " -std=c++11")
-    endif(MSVC)
+    string(APPEND EIGEN_CUDA_CXX_FLAGS " -std=c++11")
   endif()
   if(EIGEN_TEST_CUDA_CLANG)
     string(APPEND CMAKE_CXX_FLAGS " --cuda-path=${CUDA_TOOLKIT_ROOT_DIR}")