When you have a batch of work and several GPUs sitting idle, proto-tools can spread the batch across all of them automatically. The pieces run in parallel, and the results come back in the original order, so the multi-GPU execution is invisible to your code.
1 0 5