Parallel performance indices
| Performance index | Explanation |
|---|---|
| Average wall clock time (per iteration) (s) | Actual time needed to complete one iteration |
| Message count (per iteration) | Number of messages sent between all processes per iteration; important for communication latency |
| Data transfer (per iteration) (MB) | Amount of data communicated between processors per iteration. Increasing with problem size, number of partitions and complexity of underlying physics |
| Total wall clock time (s) | Actual total time for the simulation |
These indices are used to calculate speed-up, efficiency and cost.