Fixes #13, regards #12:

* Updated the compiler directives for suppressing relevant warnings to also properly work with NVCC.
* Test suite "replicated" for CUDA host-side compilation (by including the `.cpp` file from the `.cu` file)
* Test suite adapted for CUDA device-side testing - using some invocation wrappering machinery. It's a bit crude but it does the job.
* `CMakeLists.txt` for the tests reworked to support CUDA targets, architecture auto-detection, etc.

Caveat: At the moment, CUDA testing is only performed using single-block, single-thread grids.

test/CMakeLists.txt

@@ -1,12 +1,53 @@
-cmake_minimum_required(VERSION 3.9)
+cmake_minimum_required(VERSION 3.18)
 
 enable_language(CXX)
 
+option(BUILD_CUDA_TESTS "Include tests for use of the printf library with CUDA host-side and device-side code"  OFF)
+
 set(test_targets test_suite autotest)
 
 option(TEST_NONSTD_FORMATS "Include tests using non-standard-compliant format strings?" ON)
 # ... don't worry, we'll suppress the compiler warnings for those.
 
+set(cuda_tests)
+if(BUILD_CUDA_TESTS)
+	enable_language(CUDA)
+	list(APPEND CMAKE_CUDA_FLAGS "--extended-lambda --expt-relaxed-constexpr -Xcudafe --display_error_number")
+	if(CMAKE_VERSION VERSION_GREATER_EQUAL 3.18)
+	    cmake_policy(SET CMP0104 OLD)
+	endif()
+	include(FindCUDA/select_compute_arch)
+	list(APPEND cuda_tests cuda_test_suite_host cuda_test_suite_device)
+	if((NOT DEFINED CUDA_ARCH_FLAGS) OR ("${CUDA_ARCH_FLAGS}" STREQUAL ""))
+		cuda_select_nvcc_arch_flags(CUDA_ARCH_FLAGS_1 Auto)
+		set(CUDA_ARCH_FLAGS ${CUDA_ARCH_FLAGS} CACHE STRING "CUDA -gencode parameters")
+		string(REPLACE ";" " " CUDA_ARCH_FLAGS_STR "${CUDA_ARCH_FLAGS}")
+	else()
+		set(CUDA_ARCH_FLAGS_STR "${CUDA_ARCH_FLAGS}")
+	endif()
+	set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} ${CUDA_ARCH_FLAGS_STR}")
+endif()
+
+foreach(tgt ${cuda_tests})
+	string(REPLACE "cuda_" "cuda/" source_file_prefix ${tgt})
+	add_executable(${tgt} "${source_file_prefix}.cu")
+	set_target_properties(
+		${tgt}
+		PROPERTIES
+		CUDA_STANDARD 11
+		CUDA_STANDARD_REQUIRED YES
+		CUDA_EXTENSIONS NO
+	)
+endforeach()
+
+if(BUILD_CUDA_TESTS)
+	target_link_libraries(cuda_test_suite_device cuda_printf)
+	target_include_directories(cuda_test_suite_device PRIVATE "$<BUILD_INTERFACE:${GENERATED_INCLUDE_DIR}>")
+	set_target_properties(cuda_test_suite_device PROPERTIES CUDA_SEPARABLE_COMPILATION YES)
+endif()
+
+set(tests test_suite ${cuda_tests})
+
 foreach(tgt ${test_targets})
 	add_executable(${tgt} "${tgt}.cpp")
 	set_target_properties(