163 | | |
| 163 | See ticket #40. |
| 164 | |
| 165 | |
| 166 | === Methodology (22/11/2012) === |
| 167 | |
| 168 | In order to study both the influence of the IO patches and the Load balance file, I make a survey using the following setup : |
| 169 | * NCC forcing file : 360*180, 15238 land points |
| 170 | * Loop 10 times over the same year to study the influence of the Load balance file. |
| 171 | * sechiba_hist_level = 4 |
| 172 | * stomate_hist_level = 5 |
| 173 | * Monthly outputs |
| 174 | * Tests done on Curie |
| 175 | I calculate the average computation time over the 5 last runs. So we are sure that the load balance file is "stabilized". |
| 176 | || Nb processors || Average time without patches (seconds) || Average time with patches (seconds) || Time gain (ratio) || |
| 177 | || 8 || 920.69 || 790.15 || ~15% || |
| 178 | || 16 || 495.84 || 406.85 || ~18% || |
| 179 | || 32 || 573.00 || 234.4 || ~59% || |
| 180 | || 48 || 510.61 || 202.55 || ~61% || |
| 181 | || 64 || 433.53 || 191.90 || ~55% || |
| 182 | Notice that the patch is significant for a high number of processors (>16). For 32 and 48 processors, the gain is about 60% (it seems that it is the optimal for NCC). |
| 183 | After 48, the gain diminished. [[BR]] |
| 184 | For a half degree meteorological data, we may expect to use until 192 processors ! |