Executing multiple independent neural network operations concurrently across different processing units.