I am working on Hikey970 board with Mali G72-MP12 GPU. I am running multiple neural networks at once on CPU Little, CPU Big and GPU. The OpenCL version is opencl-1.2-stubs. Could you guys help me understand when running multiple networks for GPU, is it time shared or space shared? Are networks assigned with different time slots?
Old question, but networks on a single Arm GPU are time-shared.