Really stumped with this one. Have installed toolkit and SDK v3.1 on four different machines. Three of which are XP and one is Windows 7. I am trying to use the CUFFT library and it worked fine with V2.2 of the CUDA toolkit/SDK, but I needed to upgrade to v3.1 so I can use the planMany feature to batch process lots of 2d ffts.
I have found that on all my machines with compute capability 1.2 or less, I consistently get the error message from Visual Studio 2008,
“Microsoft C++ expception: cudaError_enum at memory location xxxxxxxxx”
it comes out about 15 times during the plan call.
I dont get the error message on a machine that has a GPU of compute capability 1.3. This may or may not be related.
This error occurs on the Windows XP and Windows 7 machines, so its not OS related.
I have tried different device drivers but to no avail. I have also tried v3.2 of SDK/Toolkit and same problem.
If I ensure that I dont make any use of the CUFFT library, the error message goes away. I am using plenty other cuda features.
Even with the error the FFT does actually work though, but the string of errors is a bit unnerving.
Really stumped with this one. Have installed toolkit and SDK v3.1 on four different machines. Three of which are XP and one is Windows 7. I am trying to use the CUFFT library and it worked fine with V2.2 of the CUDA toolkit/SDK, but I needed to upgrade to v3.1 so I can use the planMany feature to batch process lots of 2d ffts.
I have found that on all my machines with compute capability 1.2 or less, I consistently get the error message from Visual Studio 2008,
“Microsoft C++ expception: cudaError_enum at memory location xxxxxxxxx”
it comes out about 15 times during the plan call.
I dont get the error message on a machine that has a GPU of compute capability 1.3. This may or may not be related.
This error occurs on the Windows XP and Windows 7 machines, so its not OS related.
I have tried different device drivers but to no avail. I have also tried v3.2 of SDK/Toolkit and same problem.
If I ensure that I dont make any use of the CUFFT library, the error message goes away. I am using plenty other cuda features.
Even with the error the FFT does actually work though, but the string of errors is a bit unnerving.