[aspect-devel] "Not converge" problem with variable viscosity.

[Scott King]

I think we had the same number of elements but the linear vs. quadratic elements is/was an issue that I'm not sure we factored in correctly.  The way Aspect refines grids is quite different conceptually from CitcomS, which uses 12 cubes to tesselate (sort of) the sphere.  Those cubes are not 'refined' as one does in Aspect but within each cube one specifies a number of elements for the cube, say 32x32x32 so one has 32x32x32x12 elements or 64x64x64x12 element for the whole sphere (this need not be NxNxN or a power of 2 but often is -- not relevant here).  Comparing the number of nodes in the output file from Aspect with the number I calculate from the CitcomS grid, I think that the global refinement 4 is about equivalent to the 32x32x32x12 CitcomS grid while global refinement 5 is equivalent to the 64x64x64x12 element grid.  

In fact I asked Shangxin to set up and run the low Ra l-3,m=2 problem not so much to look at time but to begin to look at accuracy.   But I have not seen his results yet.  The goal was to try to do this on various grids and see 1) how it compares with CitcomS and 2) how the various diagnostics compare on different levels of refinement.  

Yes indeed the S40RTS initial condition is not implemented efficiently and a lot of computing time is used unnecessarily.   We have not gone back to fix that.

On Sep 4, 2015, at 5:24 PM, Timo Heister <heister at clemson.edu> wrote:

> (yes, back to the mailing list)
> 
> summarizing some more information that Shangxin sent offline to me:
> 
> ./ depth dependent profile from a file needs to be formatted correctly
> (was only checked in debug mode, see
> https://github.com/geodynamics/aspect/pull/597 )
> 
> ./ Timing information for a typical run:
> Stokes Solver takes most of the time as below:
> | Solving Stokes system   | 1 | 1.1e+04 s | 70 %|
> | Initialization                    | 2  | 4.43e+03 | 28 % |
> 
> Most of the time is spent in the Stokes solver, but initialization of
> RTS40 initial condition is still way too expensive to be used in such
> a large computation.
> 
> ./ More statistics:
> # 4: Number of Stokes degrees of freedom 79'283'080
> # 7: Iterations for Stokes solver 576
> # 8: Velocity iterations in Stokes preconditioner 11'835
> # 9: Schur complement iterations in Stokes preconditioner 6848
> 
> The linear solver is certainly worth looking into. I will report back.
> 
> 
> -- 
> Timo Heister
> http://www.math.clemson.edu/~heister/
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel