使用FindCUDA.cmake和带推力装置的推力系统OMP时出现编译错误

我最近发现，除了经典的cuda功能外，推力还能够处理自动OMP和TBB并行化

虽然我能够在一个简单的示例中使用这个非常简单的功能，但我的cmake配置生成了编译错误，可能是我使用FindCUDA.cmake的方式不对，或者这个模块不能以这种方式使用

这是我的Test.cu文件：

#include <thrust/device_vector.h>
#include <cstdio>

struct cuda_hello
{
  __host__ __device__
  void operator()(int x)
  {
    printf("Hello, world from Cuda!\n");
  }
};

int main()
{
    thrust::device_vector<int> cuda_vec(1, 0);
    thrust::for_each(cuda_vec.begin(),cuda_vec.end(),cuda_hello());
}

现在，当使用推力\设备\系统\ OMP（为CC 3.0设备编译）时，无法生成在linux下正确编译的Makefiles的cmakefile：

编译错误的类型为：

/usr/local/cuda/include/thrust/system/omp/detail/for_each.inl(53): error: incomplete type is not allowed
          detected during:
            instantiation of "RandomAccessIterator thrust::system::omp::detail::for_each_n(thrust::system::omp::detail::execution_policy<DerivedPolicy> &, RandomAccessIterator, Size, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, RandomAccessIterator=thrust::device_ptr<int>, Size=unsigned long, UnaryFunction=thrust::detail::host_generate_functor<thrust::detail::fill_functor<int>>]" 
/usr/local/cuda/include/thrust/detail/for_each.inl(69): here
            instantiation of "InputIterator thrust::for_each_n(const thrust::detail::execution_policy_base<DerivedPolicy> &, InputIterator, Size, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, InputIterator=thrust::device_ptr<int>, Size=unsigned long, UnaryFunction=thrust::detail::host_generate_functor<thrust::detail::fill_functor<int>>]" 
/usr/local/cuda/include/thrust/system/detail/generic/generate.inl(52): here
            instantiation of "OutputIterator thrust::system::detail::generic::generate_n(thrust::execution_policy<ExecutionPolicy> &, OutputIterator, Size, Generator) [with ExecutionPolicy=thrust::system::omp::detail::tag, OutputIterator=thrust::device_ptr<int>, Size=unsigned long, Generator=thrust::detail::fill_functor<int>]" 
/usr/local/cuda/include/thrust/detail/generate.inl(56): here
            instantiation of "OutputIterator thrust::generate_n(const thrust::detail::execution_policy_base<DerivedPolicy> &, OutputIterator, Size, Generator) [with DerivedPolicy=thrust::system::omp::detail::tag, OutputIterator=thrust::device_ptr<int>, Size=unsigned long, Generator=thrust::detail::fill_functor<int>]" 
/usr/local/cuda/include/thrust/system/detail/generic/fill.h(45): here
            instantiation of "OutputIterator thrust::system::detail::generic::fill_n(thrust::execution_policy<DerivedPolicy> &, OutputIterator, Size, const T &) [with DerivedPolicy=thrust::system::omp::detail::tag, OutputIterator=thrust::device_ptr<int>, Size=unsigned long, T=int]" 
/usr/local/cuda/include/thrust/detail/fill.inl(50): here
            [ 6 instantiation contexts not shown ]
            instantiation of "void thrust::detail::contiguous_storage<T, Alloc>::uninitialized_fill_n(thrust::detail::contiguous_storage<T, Alloc>::iterator, thrust::detail::contiguous_storage<T, Alloc>::size_type, const thrust::detail::contiguous_storage<T, Alloc>::value_type &) [with T=int, Alloc=thrust::device_malloc_allocator<int>]" 
/usr/local/cuda/include/thrust/detail/vector_base.inl(164): here
            instantiation of "void thrust::detail::vector_base<T, Alloc>::fill_init(thrust::detail::vector_base<T, Alloc>::size_type, const T &) [with T=int, Alloc=thrust::device_malloc_allocator<int>]" 
/usr/local/cuda/include/thrust/detail/vector_base.inl(139): here
            instantiation of "void thrust::detail::vector_base<T, Alloc>::init_dispatch(IteratorOrIntegralType, IteratorOrIntegralType, thrust::detail::true_type) [with T=int, Alloc=thrust::device_malloc_allocator<int>, IteratorOrIntegralType=int]" 
/usr/local/cuda/include/thrust/detail/vector_base.inl(224): here
            instantiation of "thrust::detail::vector_base<T, Alloc>::vector_base(InputIterator, InputIterator) [with T=int, Alloc=thrust::device_malloc_allocator<int>, InputIterator=int]" 
/usr/local/cuda/include/thrust/device_vector.h(148): here
            instantiation of "thrust::device_vector<T, Alloc>::device_vector(InputIterator, InputIterator) [with T=int, Alloc=thrust::device_malloc_allocator<int>, InputIterator=int]" 


/usr/local/cuda/include/thrust/system/omp/detail/for_each.inl(53): error: incomplete type is not allowed
          detected during:
            instantiation of "RandomAccessIterator thrust::system::omp::detail::for_each_n(thrust::system::omp::detail::execution_policy<DerivedPolicy> &, RandomAccessIterator, Size, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, RandomAccessIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, Size=long, UnaryFunction=cuda_hello]" 
(89): here
            instantiation of "RandomAccessIterator thrust::system::omp::detail::for_each(thrust::system::omp::detail::execution_policy<DerivedPolicy> &, RandomAccessIterator, RandomAccessIterator, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, RandomAccessIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, UnaryFunction=cuda_hello]" 
/usr/local/cuda/include/thrust/detail/for_each.inl(43): here
            instantiation of "InputIterator thrust::for_each(const thrust::detail::execution_policy_base<DerivedPolicy> &, InputIterator, InputIterator, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, InputIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, UnaryFunction=cuda_hello]" 
/usr/local/cuda/include/thrust/detail/for_each.inl(57): here
            instantiation of "InputIterator thrust::for_each(InputIterator, InputIterator, UnaryFunction) [with InputIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, UnaryFunction=cuda_hello]" 


2 errors detected in the compilation of "/tmp/tmpxft_00002d3a_00000000-6_Test.cpp1.ii".
CMake Error at ExecutableCmake_generated_Test.cu.o.cmake:264 (message):
  Error generating file



make[2]: *** [CMakeFiles/ExecutableCmake.dir/./ExecutableCmake_generated_Test.cu.o] Erreur 1
make[1]: *** [CMakeFiles/ExecutableCmake.dir/all] Erreur 2
make: *** [all] Erreur 2

/usr/local/cuda/include/推力/system/omp/detail/for_each.inl（53）：错误：不允许类型不完整
在以下期间检测到：
实例化“RandomAccess迭代器推力：：系统：：omp：：细节：：for_each_n（推力：：系统：：omp：：细节：：执行_policy&，RandomAccess迭代器，大小，UnaryFunction）[使用DerivedPolicy=推力：：系统：：omp：：细节：：标记，RandomAccess迭代器=推力：：设备_ptr，大小=无符号长，UnaryFunction=推力：：细节：：主机生成_函子]”
/usr/local/cuda/include/推力/细节/for_each.inl（69）：这里
实例化“InputIterator推力：：for_each_n（const推力：：detail：：execution_policy_base&，InputIterator，Size，UnaryFunction）[使用DerivedPolicy=推力：：系统：：omp：：detail：：标记，InputIterator=推力：：设备_ptr，Size=unsigned long，UnaryFunction=推力：：detail：：主机生成_函子]”
/usr/local/cuda/include/推力/system/detail/generic/generate.inl（52）：这里
实例化“输出迭代器推力：：系统：：细节：：通用：：生成\n（推力：：执行\u策略&，输出迭代器，大小，生成器）[使用ExecutionPolicy=推力：：系统：：omp：：细节：：标记，输出迭代器=推力：：设备\u ptr，大小=无符号长，生成器=推力：：细节：：填充\u函子]”
/usr/local/cuda/include/推力/detail/generate.inl（56）：这里
实例化“输出迭代器推力：：生成\n（常量推力：：细节：：执行\u策略\u基和，输出迭代器，大小，生成器）[使用DerivedPolicy=推力：：系统：：omp：：细节：：标记，输出迭代器=推力：：设备\u ptr，大小=无符号长，生成器=推力：：细节：：填充\u函子]”
/usr/local/cuda/include/推力/system/detail/generic/fill.h（45）：这里
实例化“输出迭代器推力：：系统：：细节：：通用：：填充(推力：：执行)策略&，输出迭代器，大小，常数T&）[带有DerivedPolicy=推力：：系统：：omp：：细节：：标记，输出迭代器=推力：：设备(ptr，大小=无符号长，T=int]”
/usr/local/cuda/include/推力/细节/fill.inl（50）：这里
[6未显示实例化上下文]
实例化“void-推力：：细节：：连续存储：：未初始化的\u填充\n（推力：：细节：：连续存储：：迭代器，推力：：细节：：连续存储：：大小\u类型，常量推力：：细节：：连续存储：：值\u类型&）[带T=int，Alloc=推力：：设备\u malloc \u分配器]”
/usr/local/cuda/include/推力/detail/vector_base.inl（164）：这里
实例化“void-推力：：细节：：向量\基：：填充\初始（推力：：细节：：向量\基：：大小\类型，常数T&）[使用T=int，Alloc=推力：：设备\ malloc \分配器]”
/usr/local/cuda/include/推力/detail/vector_base.inl（139）：这里
实例化“无效推力：：细节：：向量基：：初始调度（迭代器类型，迭代器类型，推力：：细节：：真类型）[使用T=int，Alloc=推力：：设备分配程序，迭代器类型=int]”
/usr/local/cuda/include/推力/detail/vector_base.inl（224）：这里
实例化“推力：：细节：：矢量\基：：矢量\基（输入编译器，输入编译器）[使用T=int，Alloc=推力：：设备\分配程序，输入编译器=int]”
/usr/local/cuda/include/推力/device_vector.h（148）：这里
实例化“推力：：设备\向量：：设备\向量（InputIterator，InputIterator）[带T=int，Alloc=推力：：设备\ malloc \ u分配器，InputIterator=int]”
/usr/local/cuda/include/THEST/system/omp/detail/for_each.inl（53）：错误：不允许类型不完整
在以下期间检测到：
实例化“RandomAccessIterator推力：：系统：：omp：：细节：：for_each_n（推力：：系统：：omp：：细节：：执行_策略&，RandomAccessIterator，大小，UnaryFunction）[使用DerivedPolicy=推力：：系统：：omp：：细节：：标记，RandomAccessIterator=推力：：细节：：普通_迭代器，大小=长，UnaryFunction=cuda_你好]”
（89）：这里
实例化“RandomAccess迭代器推力：：系统：：omp：：细节：：for_each（推力：：系统：：omp：：细节：：执行_策略&，RandomAccess迭代器，RandomAccess迭代器，UnaryFunction）[使用DerivedPolicy=推力：：系统：：omp：：细节：：标记，RandomAccess迭代器=推力：：细节：：普通_迭代器，UnaryFunction=cuda_hello]”
/usr/local/cuda/include/推力/细节/for_each.inl（43）：这里
实例化“InputIterator推力：：for_each（const推力：：detail：：execution_policy_base&，InputIterator，InputIterator，UnaryFunction）[使用DerivedPolicy=推力：：系统：：omp：：detail：：标记，InputIterator=推力：：detail：：normal_迭代器，UnaryFunction=cuda_hello]”
/usr/local/cuda/include/推力/细节/for_each.inl（57）：这里
实例化“InputIterator推力：：for_each（InputIterator，InputIterator，UnaryFunction）[使用InputIterator=推力：：细节：：普通迭代器，UnaryFunction=cuda\u hello]”
在编译“/tmp/tmpxft_00002d3a_00000000-6_Test.cpp1.ii”时检测到2个错误。
在ExecutableCmake生成的测试中出现CMake错误。cu.o.CMake:264（消息）：
生成文件时出错
make[2]：***[CMakeFiles/ExecutableCmake.dir//ExecutableCmake\u生成的\u Test.cu.o]错误1
make[1]：***[CMakeFiles/ExecutableCmake.dir/all]Erreur 2
制造：**[全部]错误2

这些错误看起来与我将cuda代码放入非.cu文件时遇到的错误完全相同，但我对cmake了解不够，无法理解为什么会出现这个问题

提前谢谢你

PROJECT(ExecutableCmake)
set (OUTPUT_NAME ExecutableCmake)

cmake_minimum_required (VERSION 2.8)

#test variable
#set(THRUST_DEVICE_SYSTEM THRUST_DEVICE_SYSTEM_CUDA)
set(THRUST_DEVICE_SYSTEM THRUST_DEVICE_SYSTEM_OMP)
#set(THRUST_DEVICE_SYSTEM THRUST_DEVICE_SYSTEM_TBB)

########################################
####            Cuda Part           ####
########################################
find_package(CUDA REQUIRED)

list( APPEND CUDA_NVCC_FLAGS -gencode arch=compute_30,code=sm_30 -DTHRUST_DEVICE_SYSTEM=${THRUST_DEVICE_SYSTEM} )

set (sources_gpu_cuda
    Test.cu
)

########################################
####         /Cuda Part             ####
########################################

########################################
####           OMP Part             ####
########################################
set(omp_deps gomp)
########################################
####           /OMP Part            ####
########################################

set (sources
#cuda source files
${sources_gpu_cuda}
)

cuda_add_executable(${OUTPUT_NAME} ${sources} ${headers})
target_link_libraries (${OUTPUT_NAME} ${omp_deps})

/usr/local/cuda/include/thrust/system/omp/detail/for_each.inl(53): error: incomplete type is not allowed
          detected during:
            instantiation of "RandomAccessIterator thrust::system::omp::detail::for_each_n(thrust::system::omp::detail::execution_policy<DerivedPolicy> &, RandomAccessIterator, Size, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, RandomAccessIterator=thrust::device_ptr<int>, Size=unsigned long, UnaryFunction=thrust::detail::host_generate_functor<thrust::detail::fill_functor<int>>]" 
/usr/local/cuda/include/thrust/detail/for_each.inl(69): here
            instantiation of "InputIterator thrust::for_each_n(const thrust::detail::execution_policy_base<DerivedPolicy> &, InputIterator, Size, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, InputIterator=thrust::device_ptr<int>, Size=unsigned long, UnaryFunction=thrust::detail::host_generate_functor<thrust::detail::fill_functor<int>>]" 
/usr/local/cuda/include/thrust/system/detail/generic/generate.inl(52): here
            instantiation of "OutputIterator thrust::system::detail::generic::generate_n(thrust::execution_policy<ExecutionPolicy> &, OutputIterator, Size, Generator) [with ExecutionPolicy=thrust::system::omp::detail::tag, OutputIterator=thrust::device_ptr<int>, Size=unsigned long, Generator=thrust::detail::fill_functor<int>]" 
/usr/local/cuda/include/thrust/detail/generate.inl(56): here
            instantiation of "OutputIterator thrust::generate_n(const thrust::detail::execution_policy_base<DerivedPolicy> &, OutputIterator, Size, Generator) [with DerivedPolicy=thrust::system::omp::detail::tag, OutputIterator=thrust::device_ptr<int>, Size=unsigned long, Generator=thrust::detail::fill_functor<int>]" 
/usr/local/cuda/include/thrust/system/detail/generic/fill.h(45): here
            instantiation of "OutputIterator thrust::system::detail::generic::fill_n(thrust::execution_policy<DerivedPolicy> &, OutputIterator, Size, const T &) [with DerivedPolicy=thrust::system::omp::detail::tag, OutputIterator=thrust::device_ptr<int>, Size=unsigned long, T=int]" 
/usr/local/cuda/include/thrust/detail/fill.inl(50): here
            [ 6 instantiation contexts not shown ]
            instantiation of "void thrust::detail::contiguous_storage<T, Alloc>::uninitialized_fill_n(thrust::detail::contiguous_storage<T, Alloc>::iterator, thrust::detail::contiguous_storage<T, Alloc>::size_type, const thrust::detail::contiguous_storage<T, Alloc>::value_type &) [with T=int, Alloc=thrust::device_malloc_allocator<int>]" 
/usr/local/cuda/include/thrust/detail/vector_base.inl(164): here
            instantiation of "void thrust::detail::vector_base<T, Alloc>::fill_init(thrust::detail::vector_base<T, Alloc>::size_type, const T &) [with T=int, Alloc=thrust::device_malloc_allocator<int>]" 
/usr/local/cuda/include/thrust/detail/vector_base.inl(139): here
            instantiation of "void thrust::detail::vector_base<T, Alloc>::init_dispatch(IteratorOrIntegralType, IteratorOrIntegralType, thrust::detail::true_type) [with T=int, Alloc=thrust::device_malloc_allocator<int>, IteratorOrIntegralType=int]" 
/usr/local/cuda/include/thrust/detail/vector_base.inl(224): here
            instantiation of "thrust::detail::vector_base<T, Alloc>::vector_base(InputIterator, InputIterator) [with T=int, Alloc=thrust::device_malloc_allocator<int>, InputIterator=int]" 
/usr/local/cuda/include/thrust/device_vector.h(148): here
            instantiation of "thrust::device_vector<T, Alloc>::device_vector(InputIterator, InputIterator) [with T=int, Alloc=thrust::device_malloc_allocator<int>, InputIterator=int]" 


/usr/local/cuda/include/thrust/system/omp/detail/for_each.inl(53): error: incomplete type is not allowed
          detected during:
            instantiation of "RandomAccessIterator thrust::system::omp::detail::for_each_n(thrust::system::omp::detail::execution_policy<DerivedPolicy> &, RandomAccessIterator, Size, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, RandomAccessIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, Size=long, UnaryFunction=cuda_hello]" 
(89): here
            instantiation of "RandomAccessIterator thrust::system::omp::detail::for_each(thrust::system::omp::detail::execution_policy<DerivedPolicy> &, RandomAccessIterator, RandomAccessIterator, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, RandomAccessIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, UnaryFunction=cuda_hello]" 
/usr/local/cuda/include/thrust/detail/for_each.inl(43): here
            instantiation of "InputIterator thrust::for_each(const thrust::detail::execution_policy_base<DerivedPolicy> &, InputIterator, InputIterator, UnaryFunction) [with DerivedPolicy=thrust::system::omp::detail::tag, InputIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, UnaryFunction=cuda_hello]" 
/usr/local/cuda/include/thrust/detail/for_each.inl(57): here
            instantiation of "InputIterator thrust::for_each(InputIterator, InputIterator, UnaryFunction) [with InputIterator=thrust::detail::normal_iterator<thrust::device_ptr<int>>, UnaryFunction=cuda_hello]" 


2 errors detected in the compilation of "/tmp/tmpxft_00002d3a_00000000-6_Test.cpp1.ii".
CMake Error at ExecutableCmake_generated_Test.cu.o.cmake:264 (message):
  Error generating file



make[2]: *** [CMakeFiles/ExecutableCmake.dir/./ExecutableCmake_generated_Test.cu.o] Erreur 1
make[1]: *** [CMakeFiles/ExecutableCmake.dir/all] Erreur 2
make: *** [all] Erreur 2

list(APPEND CUDA_NVCC_FLAGS -Xcompiler -fopenmp)