Changes between Initial Version and Version 1 of Ticket #3781, comment 18


Ignore:
Timestamp:
2017-07-21T22:18:59Z (8 years ago)
Author:
Francesco Casella

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #3781, comment 18

    initial v1  
    88Particularly, the largest figures went down from 1800+ seconds to about 300. This also reduced the total time to run the whole library testing from 7250 s to 4588 s, which is remarkable. In fact, if you comparer the [https://test.openmodelica.org/hudson/job/LibraryExperimental/ running times of the Hudson task] it is apparent that the improvement took place on 21 july betweeen 10:09 and 21:01.
    99
    10 I am not sure where these huge savings (about 1500 s) come from, unfortunately the detailed log is only available for the latest results, not for the history.
     10I am not sure where these huge savings (about 1500 s) come from, unfortunately the detailed log is only available for the latest results, not for the historical ones.
    1111
    12 However, the scaling of detectJacobianSparsePattern is still not satisfactory. If you check {{{DistributionSystemLinearIndividual_N_XX_M_XX}}}, you can see
     12However, the scaling of detectJacobianSparsePattern is still not satisfactory. If you check {{{DistributionSystemLinearIndividual_N_XX_M_XX}}}, on [https://test.openmodelica.org/libraries/ScalableTestSuite_Experimental/BuildModelRecursive.html the latest run] you can see
    1313|| N || M || Time ||
    1414|| 28 || 28 || 7.5||