.cu 源文件中默认包含哪些 headers？
Which headers are included by default in the .cu source file?
我想知道 .cu 编译单元中隐含了哪些 headers。例如，让我们举一些基本的例子：
#include <iostream>
__global__ void kernel(int a, int b, int *c)
{
*c = a * b;
}
int main()
{
int c, *dev_c;
cudaMalloc((void **) &dev_c, sizeof(*dev_c));
kernel<<<1,1>>>(5, 3, dev_c);
cudaMemcpy(&c, dev_c, sizeof(*dev_c), cudaMemcpyDeviceToHost);
cudaFree(dev_c);
std::cout << c << std::endl;
}
根据 C++ 标准，所有函数都必须在使用前声明（自 C99 以来在 C 中也是如此），所以我猜这些 CUDA headers 包含在 "behind the scence".
这个问题是相关的，因为我想在 Visual C++ 中为 CUDA API 调用获取 IntelliSense，但我没有找到任何关于此的 "oficial" 参考。
我 运行 对 Linux 上的空文件执行以下命令：
nvcc --generate-dependencies empty.cu
结果如下：
empty.o : empty.cu \
/usr/include/stdc-predef.h \
/opt/cuda/bin/..//include/cuda_runtime.h \
/opt/cuda/bin/..//include/host_config.h \
/usr/include/features.h \
/usr/include/x86_64-linux-gnu/sys/cdefs.h \
/usr/include/x86_64-linux-gnu/bits/wordsize.h \
/usr/include/x86_64-linux-gnu/gnu/stubs.h \
/usr/include/x86_64-linux-gnu/gnu/stubs-64.h \
/opt/cuda/bin/..//include/builtin_types.h \
/opt/cuda/bin/..//include/device_types.h \
/opt/cuda/bin/..//include/host_defines.h \
/opt/cuda/bin/..//include/driver_types.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/limits.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/syslimits.h \
/usr/include/limits.h \
/usr/include/x86_64-linux-gnu/bits/posix1_lim.h \
/usr/include/x86_64-linux-gnu/bits/local_lim.h \
/usr/include/linux/limits.h \
/usr/include/x86_64-linux-gnu/bits/posix2_lim.h \
/usr/include/x86_64-linux-gnu/bits/xopen_lim.h \
/usr/include/x86_64-linux-gnu/bits/stdio_lim.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h \
/opt/cuda/bin/..//include/surface_types.h \
/opt/cuda/bin/..//include/texture_types.h \
/opt/cuda/bin/..//include/vector_types.h \
/opt/cuda/bin/..//include/channel_descriptor.h \
/opt/cuda/bin/..//include/cuda_runtime_api.h \
/opt/cuda/bin/..//include/cuda_device_runtime_api.h \
/opt/cuda/bin/..//include/driver_functions.h \
/opt/cuda/bin/..//include/vector_functions.h \
/opt/cuda/bin/..//include/vector_functions.hpp \
/opt/cuda/bin/..//include/common_functions.h \
/usr/include/string.h \
/usr/include/xlocale.h \
/usr/include/time.h \
/usr/include/x86_64-linux-gnu/bits/time.h \
/usr/include/x86_64-linux-gnu/bits/types.h \
/usr/include/x86_64-linux-gnu/bits/typesizes.h \
/usr/include/x86_64-linux-gnu/bits/timex.h \
/usr/include/c++/4.8/new \
/usr/include/x86_64-linux-gnu/c++/4.8/bits/c++config.h \
/usr/include/x86_64-linux-gnu/c++/4.8/bits/os_defines.h \
/usr/include/x86_64-linux-gnu/c++/4.8/bits/cpu_defines.h \
/usr/include/c++/4.8/exception \
/usr/include/c++/4.8/bits/atomic_lockfree_defines.h \
/usr/include/stdio.h \
/usr/include/libio.h \
/usr/include/_G_config.h \
/usr/include/wchar.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include/stdarg.h \
/usr/include/x86_64-linux-gnu/bits/sys_errlist.h \
/usr/include/stdlib.h \
/usr/include/x86_64-linux-gnu/bits/waitflags.h \
/usr/include/x86_64-linux-gnu/bits/waitstatus.h \
/usr/include/endian.h \
/usr/include/x86_64-linux-gnu/bits/endian.h \
/usr/include/x86_64-linux-gnu/bits/byteswap.h \
/usr/include/x86_64-linux-gnu/bits/byteswap-16.h \
/usr/include/x86_64-linux-gnu/sys/types.h \
/usr/include/x86_64-linux-gnu/sys/select.h \
/usr/include/x86_64-linux-gnu/bits/select.h \
/usr/include/x86_64-linux-gnu/bits/sigset.h \
/usr/include/x86_64-linux-gnu/sys/sysmacros.h \
/usr/include/x86_64-linux-gnu/bits/pthreadtypes.h \
/usr/include/alloca.h \
/usr/include/x86_64-linux-gnu/bits/stdlib-float.h \
/usr/include/assert.h \
/opt/cuda/bin/..//include/math_functions.h \
/usr/include/math.h \
/usr/include/x86_64-linux-gnu/bits/huge_val.h \
/usr/include/x86_64-linux-gnu/bits/huge_valf.h \
/usr/include/x86_64-linux-gnu/bits/huge_vall.h \
/usr/include/x86_64-linux-gnu/bits/inf.h \
/usr/include/x86_64-linux-gnu/bits/nan.h \
/usr/include/x86_64-linux-gnu/bits/mathdef.h \
/usr/include/x86_64-linux-gnu/bits/mathcalls.h \
/usr/include/c++/4.8/cmath \
/usr/include/c++/4.8/bits/cpp_type_traits.h \
/usr/include/c++/4.8/ext/type_traits.h \
/usr/include/c++/4.8/cstdlib \
/opt/cuda/bin/..//include/math_functions.hpp \
/opt/cuda/bin/..//include/math_functions_dbl_ptx3.h \
/opt/cuda/bin/..//include/math_functions_dbl_ptx3.hpp \
/opt/cuda/bin/..//include/cuda_surface_types.h \
/opt/cuda/bin/..//include/cuda_texture_types.h \
/opt/cuda/bin/..//include/device_functions.h \
/opt/cuda/bin/..//include/device_functions.hpp \
/opt/cuda/bin/..//include/device_atomic_functions.h \
/opt/cuda/bin/..//include/device_atomic_functions.hpp \
/opt/cuda/bin/..//include/device_double_functions.h \
/opt/cuda/bin/..//include/device_double_functions.hpp \
/opt/cuda/bin/..//include/sm_20_atomic_functions.h \
/opt/cuda/bin/..//include/sm_20_atomic_functions.hpp \
/opt/cuda/bin/..//include/sm_32_atomic_functions.h \
/opt/cuda/bin/..//include/sm_32_atomic_functions.hpp \
/opt/cuda/bin/..//include/sm_35_atomic_functions.h \
/opt/cuda/bin/..//include/sm_20_intrinsics.h \
/opt/cuda/bin/..//include/sm_20_intrinsics.hpp \
/opt/cuda/bin/..//include/sm_30_intrinsics.h \
/opt/cuda/bin/..//include/sm_30_intrinsics.hpp \
/opt/cuda/bin/..//include/sm_32_intrinsics.h \
/opt/cuda/bin/..//include/sm_32_intrinsics.hpp \
/opt/cuda/bin/..//include/sm_35_intrinsics.h \
/opt/cuda/bin/..//include/surface_functions.h \
/opt/cuda/bin/..//include/surface_functions.hpp \
/opt/cuda/bin/..//include/texture_fetch_functions.h \
/opt/cuda/bin/..//include/texture_fetch_functions.hpp \
/opt/cuda/bin/..//include/texture_indirect_functions.h \
/opt/cuda/bin/..//include/texture_indirect_functions.hpp \
/opt/cuda/bin/..//include/surface_indirect_functions.h \
/opt/cuda/bin/..//include/surface_indirect_functions.hpp \
/opt/cuda/bin/..//include/device_launch_parameters.h
所以默认情况下 nvcc 已经包含很多 CUDA headers（在我的例子中是 /opt/cuda/...）。
然后我创建了一个包含以下内容的文件 test.cpp 和 运行 g++ -H -D__CUDACC__（伪造 nvcc）：
#include "/opt/cuda/bin/..//include/cuda_runtime.h"
结果如下：
. /opt/cuda/bin/..//include/cuda_runtime.h
.. /opt/cuda/bin/..//include/host_config.h
... /usr/include/features.h
.... /usr/include/x86_64-linux-gnu/sys/cdefs.h
..... /usr/include/x86_64-linux-gnu/bits/wordsize.h
.... /usr/include/x86_64-linux-gnu/gnu/stubs.h
..... /usr/include/x86_64-linux-gnu/gnu/stubs-64.h
.. /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_types.h
.... /opt/cuda/bin/..//include/host_defines.h
... /opt/cuda/bin/..//include/driver_types.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/limits.h
..... /usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/syslimits.h
...... /usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/limits.h
....... /usr/include/limits.h
........ /usr/include/x86_64-linux-gnu/bits/posix1_lim.h
......... /usr/include/x86_64-linux-gnu/bits/local_lim.h
.......... /usr/include/linux/limits.h
........ /usr/include/x86_64-linux-gnu/bits/posix2_lim.h
........ /usr/include/x86_64-linux-gnu/bits/xopen_lim.h
......... /usr/include/x86_64-linux-gnu/bits/stdio_lim.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
... /opt/cuda/bin/..//include/surface_types.h
... /opt/cuda/bin/..//include/texture_types.h
... /opt/cuda/bin/..//include/vector_types.h
.... /opt/cuda/bin/..//include/builtin_types.h
..... /opt/cuda/bin/..//include/vector_types.h
.. /opt/cuda/bin/..//include/channel_descriptor.h
... /opt/cuda/bin/..//include/cuda_runtime_api.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/cuda_device_runtime_api.h
.. /opt/cuda/bin/..//include/driver_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/vector_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/vector_functions.hpp
.... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/common_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
... /usr/include/string.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
.... /usr/include/xlocale.h
... /usr/include/time.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
.... /usr/include/x86_64-linux-gnu/bits/time.h
..... /usr/include/x86_64-linux-gnu/bits/types.h
...... /usr/include/x86_64-linux-gnu/bits/wordsize.h
...... /usr/include/x86_64-linux-gnu/bits/typesizes.h
..... /usr/include/x86_64-linux-gnu/bits/timex.h
... /opt/cuda/bin/..//include/math_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /usr/include/math.h
..... /usr/include/x86_64-linux-gnu/bits/huge_val.h
..... /usr/include/x86_64-linux-gnu/bits/huge_valf.h
..... /usr/include/x86_64-linux-gnu/bits/huge_vall.h
..... /usr/include/x86_64-linux-gnu/bits/inf.h
..... /usr/include/x86_64-linux-gnu/bits/nan.h
..... /usr/include/x86_64-linux-gnu/bits/mathdef.h
..... /usr/include/x86_64-linux-gnu/bits/mathcalls.h
..... /usr/include/x86_64-linux-gnu/bits/mathcalls.h
..... /usr/include/x86_64-linux-gnu/bits/mathcalls.h
.... /usr/include/stdlib.h
..... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
..... /usr/include/x86_64-linux-gnu/bits/waitflags.h
..... /usr/include/x86_64-linux-gnu/bits/waitstatus.h
...... /usr/include/endian.h
....... /usr/include/x86_64-linux-gnu/bits/endian.h
....... /usr/include/x86_64-linux-gnu/bits/byteswap.h
........ /usr/include/x86_64-linux-gnu/bits/wordsize.h
........ /usr/include/x86_64-linux-gnu/bits/byteswap-16.h
..... /usr/include/x86_64-linux-gnu/sys/types.h
...... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
...... /usr/include/x86_64-linux-gnu/sys/select.h
....... /usr/include/x86_64-linux-gnu/bits/select.h
........ /usr/include/x86_64-linux-gnu/bits/wordsize.h
....... /usr/include/x86_64-linux-gnu/bits/sigset.h
....... /usr/include/x86_64-linux-gnu/bits/time.h
...... /usr/include/x86_64-linux-gnu/sys/sysmacros.h
...... /usr/include/x86_64-linux-gnu/bits/pthreadtypes.h
....... /usr/include/x86_64-linux-gnu/bits/wordsize.h
..... /usr/include/alloca.h
...... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
..... /usr/include/x86_64-linux-gnu/bits/stdlib-float.h
.... /usr/include/c++/4.8/cmath
..... /usr/include/x86_64-linux-gnu/c++/4.8/bits/c++config.h
...... /usr/include/x86_64-linux-gnu/c++/4.8/bits/os_defines.h
...... /usr/include/x86_64-linux-gnu/c++/4.8/bits/cpu_defines.h
..... /usr/include/c++/4.8/bits/cpp_type_traits.h
..... /usr/include/c++/4.8/ext/type_traits.h
.... /usr/include/c++/4.8/cstdlib
.... /opt/cuda/bin/..//include/math_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/math_functions_dbl_ptx3.h
..... /opt/cuda/bin/..//include/math_functions_dbl_ptx3.hpp
.. /opt/cuda/bin/..//include/cuda_surface_types.h
... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/cuda_texture_types.h
... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/device_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_functions.hpp
.... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_atomic_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/device_atomic_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_double_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/device_double_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_20_atomic_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_20_atomic_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_32_atomic_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_32_atomic_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_35_atomic_functions.h
... /opt/cuda/bin/..//include/sm_20_intrinsics.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_20_intrinsics.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_30_intrinsics.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_30_intrinsics.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_32_intrinsics.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_32_intrinsics.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_35_intrinsics.h
... /opt/cuda/bin/..//include/surface_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/surface_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/texture_fetch_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/texture_fetch_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/texture_indirect_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/texture_indirect_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/surface_indirect_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/surface_indirect_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/device_launch_parameters.h
看起来 cuda_runtime.h 包括所有其他与 CUDA 相关的 headers。
我想知道 .cu 编译单元中隐含了哪些 headers。例如，让我们举一些基本的例子：
#include <iostream>
__global__ void kernel(int a, int b, int *c)
{
*c = a * b;
}
int main()
{
int c, *dev_c;
cudaMalloc((void **) &dev_c, sizeof(*dev_c));
kernel<<<1,1>>>(5, 3, dev_c);
cudaMemcpy(&c, dev_c, sizeof(*dev_c), cudaMemcpyDeviceToHost);
cudaFree(dev_c);
std::cout << c << std::endl;
}
根据 C++ 标准，所有函数都必须在使用前声明（自 C99 以来在 C 中也是如此），所以我猜这些 CUDA headers 包含在 "behind the scence".
这个问题是相关的，因为我想在 Visual C++ 中为 CUDA API 调用获取 IntelliSense，但我没有找到任何关于此的 "oficial" 参考。
我 运行 对 Linux 上的空文件执行以下命令：
nvcc --generate-dependencies empty.cu
结果如下：
empty.o : empty.cu \
/usr/include/stdc-predef.h \
/opt/cuda/bin/..//include/cuda_runtime.h \
/opt/cuda/bin/..//include/host_config.h \
/usr/include/features.h \
/usr/include/x86_64-linux-gnu/sys/cdefs.h \
/usr/include/x86_64-linux-gnu/bits/wordsize.h \
/usr/include/x86_64-linux-gnu/gnu/stubs.h \
/usr/include/x86_64-linux-gnu/gnu/stubs-64.h \
/opt/cuda/bin/..//include/builtin_types.h \
/opt/cuda/bin/..//include/device_types.h \
/opt/cuda/bin/..//include/host_defines.h \
/opt/cuda/bin/..//include/driver_types.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/limits.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/syslimits.h \
/usr/include/limits.h \
/usr/include/x86_64-linux-gnu/bits/posix1_lim.h \
/usr/include/x86_64-linux-gnu/bits/local_lim.h \
/usr/include/linux/limits.h \
/usr/include/x86_64-linux-gnu/bits/posix2_lim.h \
/usr/include/x86_64-linux-gnu/bits/xopen_lim.h \
/usr/include/x86_64-linux-gnu/bits/stdio_lim.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h \
/opt/cuda/bin/..//include/surface_types.h \
/opt/cuda/bin/..//include/texture_types.h \
/opt/cuda/bin/..//include/vector_types.h \
/opt/cuda/bin/..//include/channel_descriptor.h \
/opt/cuda/bin/..//include/cuda_runtime_api.h \
/opt/cuda/bin/..//include/cuda_device_runtime_api.h \
/opt/cuda/bin/..//include/driver_functions.h \
/opt/cuda/bin/..//include/vector_functions.h \
/opt/cuda/bin/..//include/vector_functions.hpp \
/opt/cuda/bin/..//include/common_functions.h \
/usr/include/string.h \
/usr/include/xlocale.h \
/usr/include/time.h \
/usr/include/x86_64-linux-gnu/bits/time.h \
/usr/include/x86_64-linux-gnu/bits/types.h \
/usr/include/x86_64-linux-gnu/bits/typesizes.h \
/usr/include/x86_64-linux-gnu/bits/timex.h \
/usr/include/c++/4.8/new \
/usr/include/x86_64-linux-gnu/c++/4.8/bits/c++config.h \
/usr/include/x86_64-linux-gnu/c++/4.8/bits/os_defines.h \
/usr/include/x86_64-linux-gnu/c++/4.8/bits/cpu_defines.h \
/usr/include/c++/4.8/exception \
/usr/include/c++/4.8/bits/atomic_lockfree_defines.h \
/usr/include/stdio.h \
/usr/include/libio.h \
/usr/include/_G_config.h \
/usr/include/wchar.h \
/usr/lib/gcc/x86_64-linux-gnu/4.8/include/stdarg.h \
/usr/include/x86_64-linux-gnu/bits/sys_errlist.h \
/usr/include/stdlib.h \
/usr/include/x86_64-linux-gnu/bits/waitflags.h \
/usr/include/x86_64-linux-gnu/bits/waitstatus.h \
/usr/include/endian.h \
/usr/include/x86_64-linux-gnu/bits/endian.h \
/usr/include/x86_64-linux-gnu/bits/byteswap.h \
/usr/include/x86_64-linux-gnu/bits/byteswap-16.h \
/usr/include/x86_64-linux-gnu/sys/types.h \
/usr/include/x86_64-linux-gnu/sys/select.h \
/usr/include/x86_64-linux-gnu/bits/select.h \
/usr/include/x86_64-linux-gnu/bits/sigset.h \
/usr/include/x86_64-linux-gnu/sys/sysmacros.h \
/usr/include/x86_64-linux-gnu/bits/pthreadtypes.h \
/usr/include/alloca.h \
/usr/include/x86_64-linux-gnu/bits/stdlib-float.h \
/usr/include/assert.h \
/opt/cuda/bin/..//include/math_functions.h \
/usr/include/math.h \
/usr/include/x86_64-linux-gnu/bits/huge_val.h \
/usr/include/x86_64-linux-gnu/bits/huge_valf.h \
/usr/include/x86_64-linux-gnu/bits/huge_vall.h \
/usr/include/x86_64-linux-gnu/bits/inf.h \
/usr/include/x86_64-linux-gnu/bits/nan.h \
/usr/include/x86_64-linux-gnu/bits/mathdef.h \
/usr/include/x86_64-linux-gnu/bits/mathcalls.h \
/usr/include/c++/4.8/cmath \
/usr/include/c++/4.8/bits/cpp_type_traits.h \
/usr/include/c++/4.8/ext/type_traits.h \
/usr/include/c++/4.8/cstdlib \
/opt/cuda/bin/..//include/math_functions.hpp \
/opt/cuda/bin/..//include/math_functions_dbl_ptx3.h \
/opt/cuda/bin/..//include/math_functions_dbl_ptx3.hpp \
/opt/cuda/bin/..//include/cuda_surface_types.h \
/opt/cuda/bin/..//include/cuda_texture_types.h \
/opt/cuda/bin/..//include/device_functions.h \
/opt/cuda/bin/..//include/device_functions.hpp \
/opt/cuda/bin/..//include/device_atomic_functions.h \
/opt/cuda/bin/..//include/device_atomic_functions.hpp \
/opt/cuda/bin/..//include/device_double_functions.h \
/opt/cuda/bin/..//include/device_double_functions.hpp \
/opt/cuda/bin/..//include/sm_20_atomic_functions.h \
/opt/cuda/bin/..//include/sm_20_atomic_functions.hpp \
/opt/cuda/bin/..//include/sm_32_atomic_functions.h \
/opt/cuda/bin/..//include/sm_32_atomic_functions.hpp \
/opt/cuda/bin/..//include/sm_35_atomic_functions.h \
/opt/cuda/bin/..//include/sm_20_intrinsics.h \
/opt/cuda/bin/..//include/sm_20_intrinsics.hpp \
/opt/cuda/bin/..//include/sm_30_intrinsics.h \
/opt/cuda/bin/..//include/sm_30_intrinsics.hpp \
/opt/cuda/bin/..//include/sm_32_intrinsics.h \
/opt/cuda/bin/..//include/sm_32_intrinsics.hpp \
/opt/cuda/bin/..//include/sm_35_intrinsics.h \
/opt/cuda/bin/..//include/surface_functions.h \
/opt/cuda/bin/..//include/surface_functions.hpp \
/opt/cuda/bin/..//include/texture_fetch_functions.h \
/opt/cuda/bin/..//include/texture_fetch_functions.hpp \
/opt/cuda/bin/..//include/texture_indirect_functions.h \
/opt/cuda/bin/..//include/texture_indirect_functions.hpp \
/opt/cuda/bin/..//include/surface_indirect_functions.h \
/opt/cuda/bin/..//include/surface_indirect_functions.hpp \
/opt/cuda/bin/..//include/device_launch_parameters.h
所以默认情况下 nvcc 已经包含很多 CUDA headers（在我的例子中是 /opt/cuda/...）。
然后我创建了一个包含以下内容的文件 test.cpp 和 运行 g++ -H -D__CUDACC__（伪造 nvcc）：
#include "/opt/cuda/bin/..//include/cuda_runtime.h"
结果如下：
. /opt/cuda/bin/..//include/cuda_runtime.h
.. /opt/cuda/bin/..//include/host_config.h
... /usr/include/features.h
.... /usr/include/x86_64-linux-gnu/sys/cdefs.h
..... /usr/include/x86_64-linux-gnu/bits/wordsize.h
.... /usr/include/x86_64-linux-gnu/gnu/stubs.h
..... /usr/include/x86_64-linux-gnu/gnu/stubs-64.h
.. /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_types.h
.... /opt/cuda/bin/..//include/host_defines.h
... /opt/cuda/bin/..//include/driver_types.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/limits.h
..... /usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/syslimits.h
...... /usr/lib/gcc/x86_64-linux-gnu/4.8/include-fixed/limits.h
....... /usr/include/limits.h
........ /usr/include/x86_64-linux-gnu/bits/posix1_lim.h
......... /usr/include/x86_64-linux-gnu/bits/local_lim.h
.......... /usr/include/linux/limits.h
........ /usr/include/x86_64-linux-gnu/bits/posix2_lim.h
........ /usr/include/x86_64-linux-gnu/bits/xopen_lim.h
......... /usr/include/x86_64-linux-gnu/bits/stdio_lim.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
... /opt/cuda/bin/..//include/surface_types.h
... /opt/cuda/bin/..//include/texture_types.h
... /opt/cuda/bin/..//include/vector_types.h
.... /opt/cuda/bin/..//include/builtin_types.h
..... /opt/cuda/bin/..//include/vector_types.h
.. /opt/cuda/bin/..//include/channel_descriptor.h
... /opt/cuda/bin/..//include/cuda_runtime_api.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/cuda_device_runtime_api.h
.. /opt/cuda/bin/..//include/driver_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/vector_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/vector_functions.hpp
.... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/common_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
... /usr/include/string.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
.... /usr/include/xlocale.h
... /usr/include/time.h
.... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
.... /usr/include/x86_64-linux-gnu/bits/time.h
..... /usr/include/x86_64-linux-gnu/bits/types.h
...... /usr/include/x86_64-linux-gnu/bits/wordsize.h
...... /usr/include/x86_64-linux-gnu/bits/typesizes.h
..... /usr/include/x86_64-linux-gnu/bits/timex.h
... /opt/cuda/bin/..//include/math_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /usr/include/math.h
..... /usr/include/x86_64-linux-gnu/bits/huge_val.h
..... /usr/include/x86_64-linux-gnu/bits/huge_valf.h
..... /usr/include/x86_64-linux-gnu/bits/huge_vall.h
..... /usr/include/x86_64-linux-gnu/bits/inf.h
..... /usr/include/x86_64-linux-gnu/bits/nan.h
..... /usr/include/x86_64-linux-gnu/bits/mathdef.h
..... /usr/include/x86_64-linux-gnu/bits/mathcalls.h
..... /usr/include/x86_64-linux-gnu/bits/mathcalls.h
..... /usr/include/x86_64-linux-gnu/bits/mathcalls.h
.... /usr/include/stdlib.h
..... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
..... /usr/include/x86_64-linux-gnu/bits/waitflags.h
..... /usr/include/x86_64-linux-gnu/bits/waitstatus.h
...... /usr/include/endian.h
....... /usr/include/x86_64-linux-gnu/bits/endian.h
....... /usr/include/x86_64-linux-gnu/bits/byteswap.h
........ /usr/include/x86_64-linux-gnu/bits/wordsize.h
........ /usr/include/x86_64-linux-gnu/bits/byteswap-16.h
..... /usr/include/x86_64-linux-gnu/sys/types.h
...... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
...... /usr/include/x86_64-linux-gnu/sys/select.h
....... /usr/include/x86_64-linux-gnu/bits/select.h
........ /usr/include/x86_64-linux-gnu/bits/wordsize.h
....... /usr/include/x86_64-linux-gnu/bits/sigset.h
....... /usr/include/x86_64-linux-gnu/bits/time.h
...... /usr/include/x86_64-linux-gnu/sys/sysmacros.h
...... /usr/include/x86_64-linux-gnu/bits/pthreadtypes.h
....... /usr/include/x86_64-linux-gnu/bits/wordsize.h
..... /usr/include/alloca.h
...... /usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h
..... /usr/include/x86_64-linux-gnu/bits/stdlib-float.h
.... /usr/include/c++/4.8/cmath
..... /usr/include/x86_64-linux-gnu/c++/4.8/bits/c++config.h
...... /usr/include/x86_64-linux-gnu/c++/4.8/bits/os_defines.h
...... /usr/include/x86_64-linux-gnu/c++/4.8/bits/cpu_defines.h
..... /usr/include/c++/4.8/bits/cpp_type_traits.h
..... /usr/include/c++/4.8/ext/type_traits.h
.... /usr/include/c++/4.8/cstdlib
.... /opt/cuda/bin/..//include/math_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/math_functions_dbl_ptx3.h
..... /opt/cuda/bin/..//include/math_functions_dbl_ptx3.hpp
.. /opt/cuda/bin/..//include/cuda_surface_types.h
... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/cuda_texture_types.h
... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/device_functions.h
... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_functions.hpp
.... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_atomic_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/device_atomic_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/device_double_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/device_double_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_20_atomic_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_20_atomic_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_32_atomic_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_32_atomic_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_35_atomic_functions.h
... /opt/cuda/bin/..//include/sm_20_intrinsics.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_20_intrinsics.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_30_intrinsics.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_30_intrinsics.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_32_intrinsics.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/sm_32_intrinsics.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/sm_35_intrinsics.h
... /opt/cuda/bin/..//include/surface_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/surface_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/texture_fetch_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/texture_fetch_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/texture_indirect_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/texture_indirect_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
... /opt/cuda/bin/..//include/surface_indirect_functions.h
.... /opt/cuda/bin/..//include/builtin_types.h
.... /opt/cuda/bin/..//include/surface_indirect_functions.hpp
..... /opt/cuda/bin/..//include/builtin_types.h
.. /opt/cuda/bin/..//include/device_launch_parameters.h
看起来 cuda_runtime.h 包括所有其他与 CUDA 相关的 headers。