Reputation: 21
My application is doing image processing in an openCL kernel and writing output to an openGL texture for display.
I am creating the context, command queue, compiling the program and creating the kernel, sampler and 2D image objects without error. Everything is running fine with no reported errors until I try to queue the kernel for execution:
errNum = clEnqueueNDRangeKernel(commandQueue, kernel, 2, NULL, globalWorkSize, localWorkSize, 0, NULL, NULL);
this call returns a value of -1000.
according to the online manpage for clEnqueueNDRangeKernel, there are several values errNum could be set to (defined in the cl.h header file). none of these values match -1000.
Any ideas about this error?
I'm running on an Nvidia NVS 4200M gpu if that's relevant.
Upvotes: 2
Views: 1928
Reputation: 1
I assume you are using cl gl interop? see cl_gl.h:
#define CL_INVALID_GL_SHAREGROUP_REFERENCE_KHR -1000
Upvotes: 0
Reputation: 10896
I'm not entirely sure, but it is possible that you are seeing a (negative?) CUDA_ERROR_UNKNOWN (used to be 999, but may have been bumped up to 1000) from the underlying CUDA libraries that nvidia's OpenCL implementation piggy-backs on top of. I have seen a CUDA_ERROR_INVALID_VALUE error pop up (look here at the first value in my ErrorCode enum) before, which is why I suspect this.
Upvotes: 0