This is just an observation. When calling clBuildProgram, if I set my device argument list to NULL and the number of devices, it compiles the code. However, if, as outlined in the specification, I try to specify a device, then the application crashes (no message given - it just stops running without even a chance to retrieve the returned value).
I am building on a GT230M, with driver 195.62, under Windows 7, 64 bit. Has anyone seen a similar behaviour? Are there any constraints on the use of clBuildProgram for NVIDIA GPUs since OpenCL 1.0 is supposed to allow devices and device counts to be specified.
I can also confirm, having retrieved the device information from the clGetBuildInfo and clGetDeviceInfo calls, that the device being used is a GPU device, so it seems like I should be able to conform with the standard…it’s just not happening.