Scheduling for new computing platforms with GPUs