Profiling

[michaelneuder]

Adding open_mp timing for calc explicit calls and entire loop. Timing results show that 60% of runtime comes from calc_explicit call. With Nx=640 and Ny=510, we have the following performance.

Example of the output:

  - calc_explicit(1) timing:   0.25514945201575756      (s)
  - calc_explicit(2) timing:   0.25498794508166611      (s)
  - calc_explicit(3) timing:   0.25438324990682304      (s)
  - calc_explicit(4) timing:   0.25450293906033039      (s)
 overall timing:    1.7029833139386028      (s)

[michaelneuder]

The six loops of calc_explicit are timed below. Based on this, I have been focusing most of my effort on loop 3 because it consumes 2/3 of the total time. For reference this is with Nx=1600, Ny=1275.

  - - l1 timing:    1.4956430066376925E-002 (s)
  - - l2 timing:    5.5703870952129364E-002 (s)
  - - l3 timing:   0.92254984402097762      (s)
  - - l4 timing:    8.3095165900886059E-002 (s)
  - - l5 timing:   0.32055144500918686      (s)
  - - l6 timing:    3.2401083968579769E-002 (s)
  - calc_explicit(4) timing:    1.4898755690082908      (s)