Application: Gyro
Platform: Teragrid at NCSA. | User Guide NCSA's TeraGrid cluster consists of 887 IBM cluster nodes: 256 nodes with dual 1.3 GHz Intel® Itanium® 2 processors (Phase 1) and 631 nodes with dual 1.5 GHz Intel® Itanium® 2 processors (Phase 2).
Simulation: B3-gtc
UnInstrumented version results:
Timing.out files: 64 | 128 | 192 | 256 | 320 | 384 | 448 | 512
Batch output files: 64 | 128 | 192 | 256 | 320 | 384 | 448 | 512
Graphs from the data extracted from the above results:
The consistent format being used for all the timing results:
Alternative representations:
The data that was used to generate the above graphs:(This is from the timing results for the first 50 timesteps; As can be seen from the scalability graph,there is no significant difference between the cumulative times for the first 50 and the next 50 timesteps)
| 64 | 128 | 192 | 256 | 320 | 384 | 448 | 512 | |
| NL | 6.34E+02 | 2.34E+02 | 1.59E+02 | 1.67E+02 | 9.79E+01 | 1.05E+02 | 9.34E+01 | 6.14E+01 |
| NL_tr | 4.66E+02 | 3.12E+02 | 2.59E+02 | 8.47E+01 | 9.35E+01 | 9.55E+01 | 6.97E+01 | 5.42E+01 |
| Coll | 0.00E+00 | 0.00E+00 | 0.00E+00 | 0.00E+00 | 0.00E+00 | 0.00E+00 | 0.00E+00 | 0.00E+00 |
| Coll_tr | 5.45E-01 | 2.49E-01 | 1.67E-01 | 8.67E-02 | 6.88E-02 | 3.49E-02 | 1.74E-02 | 1.54E-02 |
| lin_RHS | 4.67E+01 | 2.31E+01 | 1.60E+01 | 1.16E+01 | 9.37E+00 | 7.94E+00 | 7.21E+00 | 5.83E+00 |
| field | 5.53E+01 | 4.15E+01 | 2.24E+01 | 3.49E+01 | 2.14E+01 | 1.54E+01 | 1.99E+01 | 3.80E+01 |
| extras | 1.07E+01 | 4.67E+00 | 3.37E+00 | 2.35E+00 | 2.33E+00 | 6.06E+00 | 1.28E+00 | 1.27E+00 |
| I/O | 1.75E+00 | 1.12E+00 | 1.02E+00 | 8.50E-01 | 7.67E-01 | 7.52E-01 | 6.57E-01 | 7.41E-01 |
Processors |
Timesteps/sec |
Total Time |
64 |
2.44E+03 |
0.04100041 |
128 |
1.23E+03 |
0.081234768 |
192 |
9.22E+02 |
0.108412836 |
256 |
6.03E+02 |
0.165727544 |
320 |
4.58E+02 |
0.218483723 |
384 |
4.48E+02 |
0.223463687 |
448 |
3.87E+02 |
0.258331181 |
512 |
3.23E+02 |
0.309693404 |