Yes, I know. My point was that CPU-only deep learning is possible on the Pi if you don't need real-time inference. What I wasn't sure of is whether that specifi

Yes, I know. My point was that CPU-only deep learning is possible on the Pi if you don't need real-time inference. What I wasn't sure of is whether that specific port does anything on the GPU at all, or if it's only using NEON intrinsics.