Fix compile flags for debug mode

cmake/cuda.cmake

@@ -15,22 +15,23 @@ if(ENABLE_CUDA)
 
   # Pass options to NVCC ( -ccbin /path  --compiler-options -lfftw3f --compiler-options -lm --verbose)
   list(APPEND CUDA_NVCC_FLAGS -DENABLE_CUDA --std c++${CMAKE_CXX_STANDARD} -Wno-deprecated-gpu-targets --ptxas-options=-v)
-  list(APPEND CUDA_NVCC_FLAGS_DEBUG --debug; --device-debug; --generate-line-info -Xcompiler "-Wextra" -Xcompiler "-Werror")
+  list(APPEND CUDA_NVCC_FLAGS_DEBUG --debug --generate-line-info  -Xcompiler "-Wextra")  # Do not use Xcompiler -Werror here as it prevents some kernels from execution
+  #list(APPEND CUDA_NVCC_FLAGS_DEBUG --debug --device-debug -Xcompiler "-Wextra" -Xcompiler "-Werror")
   list(APPEND CUDA_NVCC_FLAGS_PROFILE --generate-line-info)
-  #list(APPEND CUDA_NVCC_FLAGS -arch compute_35) # minumum compute level (Sps restriction)
   string(TOUPPER "${CMAKE_BUILD_TYPE}" uppercase_CMAKE_BUILD_TYPE)
   if(NOT uppercase_CMAKE_BUILD_TYPE MATCHES "DEBUG")
-        message("Enabling device specific compilation as not in DEBUG mode")
+  message("Enabling device specific compilation as not in DEBUG mode")
         list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_61,code=sm_61) # GTX1080Ti, Titan Xp
         list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_61,code=compute_61) # Compatibility
         #list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_52,code=sm_52) # TitanX
         #list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_50,code=sm_50) # Maxwell
         #list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_37,code=sm_37) # K80
+
+    if(CUDA_VERSION VERSION_GREATER 10.0)
+        list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_75,code=sm_75) # RTX2080, CUDA10
+    endif()
   endif(NOT uppercase_CMAKE_BUILD_TYPE MATCHES "DEBUG")
 
-  if(CUDA_VERSION VERSION_GREATER 10.0)
-    list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_75,code=sm_75) # RTX2080, CUDA10
-  endif()
 
   list(APPEND CUDA_NVCC_FLAGS -O3 -use_fast_math -restrict)