About four timings per process may be read out. These timings can be combined to obtain the processing times, including or not including communication, and the total latency caused by each processor.
