| 21 | |
| 22 | === Results for 8 processors === |
| 23 | |
| 24 | * Standard IOIPSL : |
| 25 | |
| 26 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/PERF_STD_8PROCS.png,70%)]] |
| 27 | |
| 28 | * IOIPSL with patch : |
| 29 | |
| 30 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/PERF_PATCH_8PROCS.png,70%)]] |
| 31 | |
| 32 | === Results for 16 processors === |
| 33 | |
| 34 | * Standard IOIPSL : |
| 35 | |
| 36 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/STD_16PROCS.png,70%)]] |
| 37 | |
| 38 | |
| 39 | * IOIPSL with patch : |
| 40 | |
| 41 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/PATCH_PERF_16PROCS.png,70%)]] |
| 42 | |
| 43 | * Standard IOIPSL, processor 16 only : |
| 44 | |
| 45 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/STD_PROC_16.png,70%)]] |
| 46 | |
| 47 | |
| 48 | * Modified IOIPSL, processor 16 only : |
| 49 | |
| 50 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/PATCH_PROC_16.png,70%)]] |
| 51 | |
| 52 | |
| 53 | |
| 87 | [[Image(http://dods.ipsl.jussieu.fr/orchidee/WIKI/VAMPIR_RESULTS_PATCH/STD_96_PROCS.png,70%)]] |
| 88 | |
| 89 | |
| 90 | == PARTIAL CONCLUSION (September 2012) == |
| 91 | |
| 92 | * The modified IOIPSL helps to improve the performances of the model. On 16 processors, we divide by 10 the computational time taken by the subroutine moycum. It seems that the last processor always takes much more time that the others. Should we revise the algorithm for the LoadBalance.dat ? |
| 93 | * An other big problem appears with the subroutine flinopen_work : on 96 processors, this subroutine takes 90% of the time computing! This problem neutralizes the patch. |
| 94 | * Possible explanations : bug in IOIPSL, bug in the parallelization, bad use of Vampir? |
| 95 | * To investigate : why flinopen takes so much time? Test with previous ORCHIDEE tag and IOIPSL tag. Study the influence of th file LoadBalance.dat. |