2012-11-01

New "mini-ranks" hybrid programming model.



Table 1: Comparison of MPI ranks with mini-ranks on the Colosse
super-computer at Laval University.
+-------+---------------------------------------------------+
| Cores | Average round-trip latency (us)                   |
+-------+-----------------------+---------------------------+
|       | MPI ranks             | mini-ranks                |
|       | (pure MPI)            | (MPI + pthread)           |
+-------+-----------------------+---------------------------+
| 8     | 11.25 +/- 0           | 24.1429 +/- 0             |
| 16    | 35.875 +/- 6.92369    | 43.0179 +/- 8.76275       |
| 32    | 66.3125 +/- 6.76387   | 41.7143 +/- 1.23924       |
| 64    | 90 +/- 16.5265        | 37.75 +/- 6.41984         |
| 128   | 126.562 +/- 25.0116   | 43.0179 +/- 8.76275       |
| 256   | 203.637 +/- 67.4579   | 44.6429 +/- 6.11862       |
| 512   |                       |                           |
+-------+-----------------------+---------------------------+

No comments:

There was an error in this gadget