Way to test for non-output GPU

Is there a good way to test if the GPU is not used for video output, as would be the case with a Tesla GPU? For example, you might want to not limit amount of work done in a single kernel call if there’s no need to limit workloads to keep the system responsive at the expense of performance. Should there be a property from clGetDeviceInfo, like CL_DEVICE_NON_OUTPUT or something to that effect?

Is there a good way to test if the GPU is not used for video output, as would be the case with a Tesla GPU? For example, you might want to not limit amount of work done in a single kernel call if there’s no need to limit workloads to keep the system responsive at the expense of performance. Should there be a property from clGetDeviceInfo, like CL_DEVICE_NON_OUTPUT or something to that effect?

I don’t think there is such a thing. If you don’t use the device for the output it simply execute only your calculations. In the case of non-video device the performance related to the output are only related to when and how you use functions like clEnqueueReadBuffer to retrieve data. Performance issues related to graphic arise if you want use another video capable device for image output, in that case I think you have to use synchronization and data passing.

I don’t think there is such a thing. If you don’t use the device for the output it simply execute only your calculations. In the case of non-video device the performance related to the output are only related to when and how you use functions like clEnqueueReadBuffer to retrieve data. Performance issues related to graphic arise if you want use another video capable device for image output, in that case I think you have to use synchronization and data passing.