Pflow Baseline Performance on Neptune
From WaterWiki
Contents |
[edit] Plot comparing baseline and profiling cases
[edit] Comments
The profiling overhead is quite high currently even after applying throttling during TAU profiling. Throttling is the processing of eliminating small functions which are called frequently from profiling. I fixed a few bugs in PETSc to get it to work with TAU. Still the integration of PETSc with TAU is suspect. I believe the fault lies within PETSc and not with TAU.
[edit] Base Pflow
Number of Processors followed by the time taken in each case.
1
CPU Time: 4.1590E-02 [sec] 6.9317E-04 [min] 1.1553E-05 [hr] Wall Clock Time: 4.1754E+02 [sec] 6.9591E+00 [min] 1.1598E-01 [hr]
2
CPU Time: 2.3439E-02 [sec] 3.9065E-04 [min] 6.5108E-06 [hr] Wall Clock Time: 2.5033E+02 [sec] 4.1722E+00 [min] 6.9537E-02 [hr]
4
CPU Time: 1.1320E-02 [sec] 1.8867E-04 [min] 3.1444E-06 [hr] Wall Clock Time: 1.3574E+02 [sec] 2.2624E+00 [min] 3.7706E-02 [hr]
8
CPU Time: 4.2620E-03 [sec] 7.1033E-05 [min] 1.1839E-06 [hr] Wall Clock Time: 7.9621E+01 [sec] 1.3270E+00 [min] 2.2117E-02 [hr]
16
CPU Time: 2.5560E-03 [sec] 4.2600E-05 [min] 7.1000E-07 [hr] Wall Clock Time: 8.9761E+01 [sec] 1.4960E+00 [min] 2.4934E-02 [hr]
[edit] Profiling with TAU
1
CPU Time: 4.3472E-02 [sec] 7.2453E-04 [min] 1.2076E-05 [hr] Wall Clock Time: 5.2547E+02 [sec] 8.7579E+00 [min] 1.4596E-01 [hr]
2
CPU Time: 2.8822E-02 [sec] 4.8037E-04 [min] 8.0061E-06 [hr] Wall Clock Time: 4.5234E+02 [sec] 7.5389E+00 [min] 1.2565E-01 [hr]
4
CPU Time: 1.8149E-02 [sec] 3.0248E-04 [min] 5.0414E-06 [hr] Wall Clock Time: 3.8928E+02 [sec] 6.4879E+00 [min] 1.0813E-01 [hr]
8
CPU Time: 1.1909E-02 [sec] 1.9848E-04 [min] 3.3081E-06 [hr] Wall Clock Time: 4.5149E+02 [sec] 7.5249E+00 [min] 1.2541E-01 [hr]
16
CPU Time: 1.0732E-02 [sec] 1.7887E-04 [min] 2.9811E-06 [hr] Wall Clock Time: 7.3258E+02 [sec] 1.2210E+01 [min] 2.0350E-01 [hr]
