[aspect-devel] temperature assembly timing

[Thieulot, C. (Cedric)]

Hi all,

I have installed aspect's latest version and I have been running the convection-box cookbook to satisfaction.
Since i wish to run case 1c of Blankenbach et al 1989, I have set g= 10^16 so that Ra=10^6.
I have set the resolution to a constant 512x512 (set Initial global refinement = 9).
So far so good, until I looked at the timings after a substantial number of time steps:
Why does the temperature assembly system take so much time ? (10 times more than Stokes assembly, 20times more than its solve )

Thanks,

Cedric.

 784 +---------------------------------------------+------------+------------+
 785 | Total wallclock time elapsed since start    |  3.25e+05s |            |
 786 |                                             |            |            |
 787 | Section                         | no. calls |  wall time | % of total |
 788 +---------------------------------+-----------+------------+------------+
 789 | Assemble Stokes system          |     11101 |     2e+04s |       6.1% |
 790 | Assemble temperature system     |     11101 |  2.36e+05s |        73% |
 791 | Build Stokes preconditioner     |         1 |      15.2s |    0.0047% |
 792 | Build temperature preconditioner|     11101 |  3.65e+03s |       1.1% |
 793 | Solve Stokes system             |     11101 |  4.11e+04s |        13% |
 794 | Solve temperature system        |     11101 |  1.06e+04s |       3.3% |
 795 | Initialization                  |         2 |     0.917s |   0.00028% |
 796 | Postprocessing                  |     11101 |  8.65e+03s |       2.7% |
 797 | Setup dof systems               |         1 |      11.7s |    0.0036% |
 798 +---------------------------------+-----------+------------+------------+
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://geodynamics.org/pipermail/aspect-devel/attachments/20130918/476df48a/attachment.html>