CUDA linking error - Visual Express 2008 - nvcc fatal due to (null) configuration file

I've been searching extensively for a possible solution to my error for the past 2 weeks. I have successfully installed the Cuda 64-bit compiler (tools) and SDK as well as the 64-bit version of Visual Studio Express 2008 and Windows 7 SDK with Framework 3.5. I'm using windows XP 64-bit. I have confirmed that VSE is able to compile in 64-bit as I have all of the 64-bit options available to me using the steps on the following website: (since Visual Express does not inherently include the 64-bit packages)

http://jenshuebel.wordpress.com/2009/02/12/visual-c-2008-express-edition-and-64-bit-targets/

The registry updates for 64-bit installation are found in a user comment on the same page as the above link.

I have confirmed the 64-bit compile ability since the "x64" is available from the pull-down menu under "Tools->Options->VC++ Directories" and compiling in 64-bit does not result in the entire project being "skipped". I have included all the needed directories for 64-bit cuda tools, 64 SDK and Visual Express (\VC\bin\amd64).

Here's the error message I receive when trying to compile in 64-bit:

1>------ Build started: Project: New, Configuration: Release x64 ------
1>Compiling with CUDA Build Rule...
1>"C:\CUDA\bin64\nvcc.exe"    -arch sm_10 -ccbin "C:\Program Files (x86)\Microsoft    Visual Studio 9.0\VC\bin"    -Xcompiler "/EHsc /W3 /nologo /O2 /Zi   /MT  "  -maxrregcount=32  --compile -o "x64\Release\template.cu.obj" "c:\Documents and Settings\All Users\Application Data\NVIDIA Corporation\NVIDIA GPU Computing SDK\C\src\CUDA_Walkthrough_DeviceKernels\template.cu" 
1>nvcc fatal   : Visual Studio configuration file '(null)' could not be found for installation at 'C:/Program Files (x86)/Microsoft Visual Studio 9.0/VC/bin/../..'
1>Linking...
1>LINK : fatal error LNK1181: cannot open input file '.\x64\Release\template.cu.obj'
1>Build log was saved at "file://c:\Documents and Settings\Administrator\My Documents\Visual Studio 2008\Projects\New\New\x64\Release\BuildLog.htm"
1>New - 1 error(s), 0 warning(s)
========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========

Here's the simple code I'm trying to compile/run in 64-bit:

#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>

#include <cuda.h>

void mypause () 
{ 
  printf ( "Press [Enter] to continue . . ." );
  fflush ( stdout );
  getchar();
} 

__global__ void VecAdd1_Kernel(float* A, float* B, float* C, int N)
{
 int i = blockDim.x*blockIdx.x+threadIdx.x;
 if (i<N)
  C[i] = A[i] + B[i]; //result should be a 16x1 array of 250s
} 

__global__ void VecAdd2_Kernel(float* B, float* C, int N)
{
 int i = blockDim.x*blockIdx.x+threadIdx.x;
 if (i<N)
  C[i] = C[i] + B[i]; //result should be a 16x1 array o开发者_如何学运维f 400s
}

int main()
{
 int N = 16;
 float A[16];float B[16];
 size_t size = N*sizeof(float);

 for(int i=0; i<N; i++) 
 {
  A[i] = 100.0;
  B[i] = 150.0;
 }

 // Allocate input vectors h_A and h_B in host memory
 float* h_A = (float*)malloc(size);
        float* h_B = (float*)malloc(size);
        float* h_C = (float*)malloc(size);

 //Initialize Input Vectors
 memset(h_A,0,size);memset(h_B,0,size);
 h_A = A;h_B = B;

 printf("SUM = %f\n",A[1]+B[1]); //simple check for initialization

 //Allocate vectors in device memory
 float* d_A;
 cudaMalloc((void**)&d_A,size);
 float* d_B;
 cudaMalloc((void**)&d_B,size);
 float* d_C;
 cudaMalloc((void**)&d_C,size);

 //Copy vectors from host memory to device memory
 cudaMemcpy(d_A,h_A,size,cudaMemcpyHostToDevice);
 cudaMemcpy(d_B,h_B,size,cudaMemcpyHostToDevice);

 //Invoke kernel
 int threadsPerBlock = 256;
 int blocksPerGrid = (N+threadsPerBlock-1)/threadsPerBlock;
 VecAdd1(blocksPerGrid, threadsPerBlock,d_A,d_B,d_C,N);
 VecAdd2(blocksPerGrid, threadsPerBlock,d_B,d_C,N);

 //Copy results from device memory to host memory
 //h_C contains the result in host memory
 cudaMemcpy(h_C,d_C,size,cudaMemcpyDeviceToHost);

 for(int i=0; i<N; i++) //output result from the kernel "VecAdd"
 {
  printf("%f ", h_C[i] );
  printf("\n");
 }
 printf("\n");

 cudaFree(d_A); 
 cudaFree(d_B); 
 cudaFree(d_C);
 free(h_A);
 free(h_B);
 free(h_C);

 mypause();
 return 0;
}

---

I solved the problem by

1. installing Windows SDK (don't forget to choose all x64 options for 64 bit OS)
2. include "c:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin\amd64" in PATH
3. create file vcvars64.bat inside directory "c:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin\amd64" with following content: call "C:\Program Files\Microsoft SDKs\Windows\v7.1\Bin\SetEnv.cmd" /x64

Note: I did this because:

1. I am using VC++ Express 2010
2. I dont have "vcvars64.bat" in any directory ??

---

UPDATE 06/04/2010:

Ok, I found the solution to the problem. The code is fine. After following the steps from the original link above and adding in the needed registry key make sure that Windows SDK is targeting the correct version (7.0) by launching the Windows SDK Configuration Tool from start menu, choose the right version (v7.0), and click "Make Current".

Make sure to include the following directories for x64 compiling (under "Tools->Options->Projects And Solutions->VC++ Directories): C:\CUDA\bin64 C:\CUDA\lib64 C:\CUDA\include C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC\bin\amd64

Now there's one other thing to change for a 64-bit system. Apparently the cuda compiler has a "hard-coded" directory for 64-bit compilers on Visual Express 2008. To make the fix, copy the needed file "vcvars64.bat" and rename it to "vcvarsamd64.bat" as follows:

C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC\bin\vcvars64.bat

to

C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC\bin\amd64\vcvarsamd64.bat

Upon the change, the program compiled and ran successfully.

I've read 'other' postings all over the web for other solutions to the error: "nvcc fatal: Visual Studio configuration file '(null)'", but very few have specified the above as a requirement to get nvcc to find the necessary visual studio configuration file.

Disclaimer: I installed to a clean machine and OS.

Hopefully this helps others with similar problems.

---

Same error message, but different solution yet again. I was trying to compile 32 bits CUDA libs on a 64bits machine. I had to add --machine 32 argument to nvcc call, as suggested by Imperian in a comment, here : Error compiling CUDA from Command Prompt

in hope this helps someone

---

I found that I also had to change the CUDA_PATH and CUDA_LIB_PATH environment variables, because they were pointing to the x86 toolchain (which I had installed after the x64 tools).

After some faffing around with some linker errors, I managed to build a x64 CUDA app!