I am taking my first faltering steps with CUDA Toolkit 5.0 RC using VS2010.

Separate compilation has me confused.

I tried to set up a project as a Static Library (.lib), but when I try to build it, it does not create a device-link.obj and I don't understand why.

For instance, there are 2 files:

A caller function that uses a function f

#include "thrust\host_vector.h"
#include "thrust\device_vector.h"
using namespace thrust::placeholders;

extern __device__ double f(double x);

struct f_func 
{
__device__ double operator()(const double& x) const
{
    return f(x);
}
};

void test(const int len, double * data, double * res)
{
thrust::device_vector<double> d_data(data, data + len);
thrust::transform(d_data.begin(), d_data.end(), d_data.begin(), f_func());
thrust::copy(d_data.begin(),d_data.end(), res);
}

And a library file that defines f

__device__ double f(double x)
{
return x+2.0;
}

If I set the option generate relocatable device code to No, the first file will not compile due to unresolved extern function f.

If I set it to -rdc, it will compile, but does not produce a device-link.obj file and so the linker fails.

If I put the definition of f into the first file and delete the second it builds successfully, but now it isn't separate compilation anymore.

How can I build a static library like this with separate source files?

[Updated here]

I called the first caller file "caller.cu" and the second "libfn.cu". The compiler lines that VS2010 outputs (which I don't fully understand) are (for caller):

nvcc.exe 
-ccbin "C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin"  
-I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0\include" 
-I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0\include"  
-G   
--keep-dir "Debug" 
-maxrregcount=0  
--machine 32 
--compile  
-g   
-D_MBCS 
-Xcompiler "/EHsc /W3 /nologo /Od /Zi /RTC1 /MDd  " 
-o "Debug\caller.cu.obj" "G:\Test_Linking\caller.cu" 
-clean

and the same for libfn, then:

nvcc.exe 
-gencode=arch=compute_20,code=\"sm_20,compute_20\" 
--use-local-env 
--cl-version 2010 
-ccbin "C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin" 
-rdc=true 
-I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0\include" 
-I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0\include"  
-G   
--keep-dir "Debug" 
-maxrregcount=0  
--machine 32 
--compile  
-g   
-D_MBCS 
-Xcompiler "/EHsc /W3 /nologo /Od /Zi /RTC1 /MDd  " 
-o "Debug\caller.cu.obj" "G:\Test_Linking\caller.cu"

and again for libfn.