Q. How to correctly compute the solver scale up for parallel runs?

A. When computing solver scale up for parallel runs, the correct 'Clock Time' to use is the "CFD Solver wall clock seconds" printed right after the last iteration in the Output file. Example:

======================================================================
OUTER LOOP ITERATION = 784 ( 200) CPU SECONDS = 3.593E+03 (1.068E+03)
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution |
+----------------------+------+---------+---------+------------------+
| U-Mom | 0.97 | 1.5E-05 | 3.3E-04 | 6.9E-03 OK|
| V-Mom | 0.98 | 3.1E-06 | 9.4E-05 | 1.6E-02 OK|
| W-Mom | 1.05 | 4.1E-06 | 3.6E-05 | 6.2E-03 OK|
| P-Mass | 0.82 | 2.4E-06 | 3.3E-05 | 8.5 2.9E-02 OK|
+----------------------+------+---------+---------+------------------+
| H-Energy | 0.90 | 4.0E-05 | 8.7E-04 | 5.6 6.1E-03 OK|
+----------------------+------+---------+---------+------------------+
| K-TurbKE | 0.87 | 1.1E-05 | 4.2E-04 | 5.6 1.1E-02 OK|
| E-Diss.K | 0.71 | 1.5E-05 | 4.5E-04 | 7.1 4.1E-03 OK|
+----------------------+------+---------+---------+------------------+
| Water Vapour at 100 | 0.94 | 3.4E-05 | 7.2E-04 | 5.6 1.7E-02 OK|
+----------------------+------+---------+---------+------------------+

CFD Solver finished: Thu Jun 9 13:51:42 2005
CFD Solver wall clock seconds: 1.1380E+03

All other CPU times and Wall clocks times reported will include disk I/O overheads and partitioning overheads and hence will not give a true representation of the the solver scale up as you add partitions.


Q. How to correctly compute the solver scale up for parallel runs?

A. When computing solver scale up for parallel runs, the correct 'Clock Time' to use is the "CFD Solver wall clock seconds" printed right after the last iteration in the Output file. Example:

======================================================================
OUTER LOOP ITERATION = 784 ( 200) CPU SECONDS = 3.593E+03 (1.068E+03)
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution |
+----------------------+------+---------+---------+------------------+
| U-Mom | 0.97 | 1.5E-05 | 3.3E-04 | 6.9E-03 OK|
| V-Mom | 0.98 | 3.1E-06 | 9.4E-05 | 1.6E-02 OK|
| W-Mom | 1.05 | 4.1E-06 | 3.6E-05 | 6.2E-03 OK|
| P-Mass | 0.82 | 2.4E-06 | 3.3E-05 | 8.5 2.9E-02 OK|
+----------------------+------+---------+---------+------------------+
| H-Energy | 0.90 | 4.0E-05 | 8.7E-04 | 5.6 6.1E-03 OK|
+----------------------+------+---------+---------+------------------+
| K-TurbKE | 0.87 | 1.1E-05 | 4.2E-04 | 5.6 1.1E-02 OK|
| E-Diss.K | 0.71 | 1.5E-05 | 4.5E-04 | 7.1 4.1E-03 OK|
+----------------------+------+---------+---------+------------------+
| Water Vapour at 100 | 0.94 | 3.4E-05 | 7.2E-04 | 5.6 1.7E-02 OK|
+----------------------+------+---------+---------+------------------+

CFD Solver finished: Thu Jun 9 13:51:42 2005
CFD Solver wall clock seconds: 1.1380E+03

All other CPU times and Wall clocks times reported will include disk I/O overheads and partitioning overheads and hence will not give a true representation of the the solver scale up as you add partitions.





Show Form
No comments yet. Be the first to add a comment!