I use [font=“Courier”]cudaMalloc3DArray[/font] followed by [font=“Courier”]cudaMemcpy3D[/font] to send data from the host to the device.
On Windows 2003 64-bit Server with a Tesla D870, the call to [font=“Courier”]cudaMemcpy3D[/font] returns [font=“Courier”]cudaErrorLaunchFailure[/font]. But on Vista 32-bit with GeForce 8800 Ultra, the call succeeds. The reference manual only mentions a return value of [font=“Courier”]cudaErrorSuccess[/font].
Please help me to figure out what is going wrong. Thanks.