Fix nvcc CUDA test performance: propagate host compiler flags

libeigen/eigen!2434

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index e50f24f..5a7cdae 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -456,7 +456,6 @@
     endforeach()
     string(APPEND CMAKE_CXX_FLAGS " ${EIGEN_CUDA_CXX_FLAGS}")
   else()
-    set(CUDA_PROPAGATE_HOST_FLAGS OFF)
     set(NVCC_ARCH_FLAGS)
     # Define an -arch=sm_<arch>, otherwise if GPU does not exactly match one of
     # those in the arch list for -gencode, the kernels will fail to run with
diff --git a/unsupported/test/CMakeLists.txt b/unsupported/test/CMakeLists.txt
index e90ca8c..ec61f4c 100644
--- a/unsupported/test/CMakeLists.txt
+++ b/unsupported/test/CMakeLists.txt
@@ -254,7 +254,6 @@
     endforeach()
     string(APPEND CMAKE_CXX_FLAGS " ${EIGEN_CUDA_CXX_FLAGS}")
   else()
-    set(CUDA_PROPAGATE_HOST_FLAGS OFF)
     set(NVCC_ARCH_FLAGS)
     # Define an -arch=sm_<arch>, otherwise if GPU does not exactly match one of
     # those in the arch list for -gencode, the kernels will fail to run with