CMake doesn't pass CUDA arguments? - cmake

Question

I am trying to compile a library with both, C++ and CUDA source files. I am using GNU make with CMake. My compiler of choice is clang, since CUDA only supports gcc up to version 5 and Debian 9 has gcc 6 as its oldest version and I have to use software that is provided by the Debian 9 or 10 repositories.
CMake version is 3.9.0
clang version is 3.8.1
cc and c++ in /usr/bin correctly link to clang and clang++ ẃhich also link to the correct files.
Unfortunately the initial checks of CMake for CUDA fail although everything, as far as I can see, seems to be set up correctly. It looks like the arguments aren't passed correctly to the CUDA compiler.
This is a part of my project's main CMake file:
cmake_minimum_required (VERSION 3.9.0 FATAL_ERROR)
project (dev)
find_package(CUDA REQUIRED)
set(CUDA_HOST_COMPILATION_CPP ON)
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -Wno-deprecated-gpu-targets -ccbin clang-3.8")
The library's CMake file looks like this:
cmake_minimum_required(VERSION 3.9.0 FATAL_ERROR)
project (libname LANGUAGES CXX CUDA)
file(GLOB SOURCES "*.cu" "*.cpp")
add_library(libname ${SOURCES})
set_target_properties(libname PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
set_target_properties(libname PROPERTIES POSITION_INDEPENDENT_CODE ON)
This is the output of CMake:
-- The C compiler identification is Clang 3.8.1
-- The CXX compiler identification is Clang 3.8.1
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - not found
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE
-- Found CUDA: /usr (found version "8.0")
-- The CUDA compiler identification is unknown
-- Check for working CUDA compiler: /usr/bin/nvcc
-- Check for working CUDA compiler: /usr/bin/nvcc -- broken
The CUDA/nvcc tests of CMake fails with the following errors:
Change Dir: /home/user/projects/hamonIC-linux-experimental/current_state/working_copy/code/build/CMakeFiles/CMakeTmp
Run Build Command:"/usr/bin/make" "cmTC_6d1a9/fast"
/usr/bin/make -f CMakeFiles/cmTC_6d1a9.dir/build.make
CMakeFiles/cmTC_6d1a9.dir/build
make[1]: Entering directory
'/home/user/projects/hamonIC-linux-experimental/current_state/working_copy/code/build/CMakeFiles/CMakeTmp'
Building CUDA object CMakeFiles/cmTC_6d1a9.dir/main.cu.o
/usr/bin/nvcc -x cu -c
/home/user/projects/hamonIC-linux-experimental/current_state/working_copy/code/build/CMakeFiles/CMakeTmp/main.cu
-o CMakeFiles/cmTC_6d1a9.dir/main.cu.o
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are
deprecated, and may be removed in a future release (Use
-Wno-deprecated-gpu-targets to suppress warning).
ERROR: No supported gcc/g++ host compiler found, but clang-3.8 is
available.
Use 'nvcc -ccbin clang-3.8' to use that instead.
CMakeFiles/cmTC_6d1a9.dir/build.make:65: recipe for target
'CMakeFiles/cmTC_6d1a9.dir/main.cu.o' failed
make[1]: *** [CMakeFiles/cmTC_6d1a9.dir/main.cu.o] Error 1
make[1]: Leaving directory
'/home/user/projects/hamonIC-linux-experimental/current_state/working_copy/code/build/CMakeFiles/CMakeTmp'
Makefile:126: recipe for target 'cmTC_6d1a9/fast' failed
make: *** [cmTC_6d1a9/fast] Error 2
CMake will not be able to correctly generate this project.
Call Stack (most recent call first):
hic_iplibrary/source/ciccommon/CMakeLists.txt:7 (project)

Answer 1

The CUDA NVCC flags in CMake have to be semicolon delimited and not whitespace delimited.
Change the flags in your CMakeLists.txt to use:
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS};-Wno-deprecated-gpu-targets;-ccbin=clang-3.8")
and that should allow CMake to pass your flags to NVCC.