hyperbolic Serre-Green-Naghdi equations #139
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #139 +/- ##
==========================================
+ Coverage 97.57% 97.75% +0.18%
==========================================
Files 20 22 +2
Lines 1569 1741 +172
==========================================
+ Hits 1531 1702 +171
- Misses 38 39 +1 ☔ View full report in Codecov by Sentry. |
| # TODO: There is another name clash. For the SerreGreenNaghdiEquations1D, | ||
| # the corresponding function is called initial_condition_convergence_test | ||
| # However, we cannot use that name since it's not an analytical solution. | ||
| # How shall we handle this? | ||
| """ | ||
| initial_condition_soliton(x, t, equations::HyperbolicSerreGreenNaghdiEquations1D, mesh) | ||
|
|
||
| A soliton solution of the [`SerreGreenNaghdiEquations1D`](@ref) | ||
| used for convergence tests in a periodic domain. | ||
|
|
||
| See also [`initial_condition_convergence_test`](@ref). | ||
| """ |
There was a problem hiding this comment.
Another name clash
There was a problem hiding this comment.
Good question. I would be fine keeping it like it is now (maybe with a bit more explanations in the docstring that this function is the same as initial_condition_convergence_test for SerreGreenNaghdiEquations1D, but is not an analytical solution of HyperbolicSerreGreenNaghdiEquations1D). I think it's not too bad that these two initial conditions are not called the same.
There was a problem hiding this comment.
Well, they represent the same physics, so it's a bit annoying but I don't know a better solution right now 🤷
There was a problem hiding this comment.
Yeah, it' not perfect, but I can live with in (at least for now). The name initial_condition_convergence_test doesn't describe the physics, but rather the fact of it being an analytical solution of some equations, so it's not that confusing, I think.
There was a problem hiding this comment.
Thanks! A preliminary review regarding your comments/questions.
I saw some non-ASCII code characters in executable code. Could you replace them by ASCII code characters, please?
How to initialize variables? Ideally, I would like to use w = -(D1 * v) instead of having to specify a value manually in the initial condition.
I also thought about that. I agree that this would be really nice. I think it would be ideal if both would be possible, i.e., defining an initial condition consisting of three variables with defaults for compute_coefficients by adding some dispatching on HyperbolicSerreGreenNaghdiEquations1D and some branching whether the output of initial_condition is three or five elements long?
| # TODO: There is another name clash. For the SerreGreenNaghdiEquations1D, | ||
| # the corresponding function is called initial_condition_convergence_test | ||
| # However, we cannot use that name since it's not an analytical solution. | ||
| # How shall we handle this? | ||
| """ | ||
| initial_condition_soliton(x, t, equations::HyperbolicSerreGreenNaghdiEquations1D, mesh) | ||
|
|
||
| A soliton solution of the [`SerreGreenNaghdiEquations1D`](@ref) | ||
| used for convergence tests in a periodic domain. | ||
|
|
||
| See also [`initial_condition_convergence_test`](@ref). | ||
| """ |
There was a problem hiding this comment.
Good question. I would be fine keeping it like it is now (maybe with a bit more explanations in the docstring that this function is the same as initial_condition_convergence_test for SerreGreenNaghdiEquations1D, but is not an analytical solution of HyperbolicSerreGreenNaghdiEquations1D). I think it's not too bad that these two initial conditions are not called the same.
Co-authored-by: Joshua Lampert <[email protected]>
I used them on purpose to make it easier to search and replace stuff after deciding how to call the new variables |
julia> convergence_test("examples/hyperbolic_serre_green_naghdi_1d/hyperbolic_serre_green_naghdi_manufactured.jl", 4);
[...]
####################################################################################################
l2
η v D w H
N error EOC N error EOC N error EOC N error EOC N error EOC
50 5.65e-02 - 50 3.46e-02 - 50 0.00e+00 - 50 1.17e+00 - 50 5.85e-02 -
100 3.61e-03 3.97 100 2.25e-03 3.94 100 0.00e+00 NaN 100 7.51e-02 3.96 100 3.73e-03 3.97
200 2.26e-04 4.00 200 1.42e-04 3.99 200 0.00e+00 NaN 200 4.71e-03 3.99 200 2.34e-04 4.00
400 1.42e-05 4.00 400 1.02e-05 3.79 400 0.00e+00 NaN 400 2.95e-04 4.00 400 1.46e-05 4.00
mean 3.99 mean 3.91 mean NaN mean 3.99 mean 3.99
----------------------------------------------------------------------------------------------------
linf
η v D w H
N error EOC N error EOC N error EOC N error EOC N error EOC
50 1.21e-01 - 50 5.21e-02 - 50 0.00e+00 - 50 2.19e+00 - 50 1.27e-01 -
100 7.81e-03 3.95 100 3.36e-03 3.95 100 0.00e+00 NaN 100 1.41e-01 3.96 100 8.59e-03 3.88
200 4.91e-04 3.99 200 2.16e-04 3.96 200 0.00e+00 NaN 200 8.83e-03 3.99 200 5.38e-04 4.00
400 3.15e-05 3.96 400 2.33e-05 3.21 400 0.00e+00 NaN 400 5.51e-04 4.00 400 3.37e-05 4.00
mean 3.97 mean 3.71 mean NaN mean 3.98 mean 3.96
----------------------------------------------------------------------------------------------------
|
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
|
Including some optimizations for flat bathymetry: julia> q = copy(ode.u0); dq = similar(q); @benchmark $(ode.f)($dq, $q, $(ode.p), $0.0)
BenchmarkTools.Trial: 10000 samples with 4 evaluations.
Range (min … max): 7.864 μs … 13.969 μs ┊ GC (min … max): 0.00% … 0.00%
Time (median): 7.927 μs ┊ GC (median): 0.00%
Time (mean ± σ): 8.165 μs ± 647.504 ns ┊ GC (mean ± σ): 0.00% ± 0.00%
█▇▆▂▂▃▃ ▂▁▂▁▂ ▁ ▂
█████████▅▅▄▅▁▄▃█████████▇▆▅▅▆▃▆▆▆▇██▇▇▇▇▆▇▇███▇▅▅▄▆▆▅▅▃▅▆▆ █
7.86 μs Histogram: log(frequency) by time 11 μs <
Memory estimate: 304 bytes, allocs estimate: 5.
julia> redirect_stdio(stdout = devnull, stderr = devnull) do
trixi_include("examples/hyperbolic_serre_green_naghdi_1d/hyperbolic_serre_green_naghdi_soliton.jl", interval = 1000, bathymetry_type = bathymetry_mild_slope);
end;
julia> q = copy(ode.u0); dq = similar(q); @benchmark $(ode.f)($dq, $q, $(ode.p), $0.0)
BenchmarkTools.Trial: 10000 samples with 3 evaluations.
Range (min … max): 8.597 μs … 17.208 μs ┊ GC (min … max): 0.00% … 0.00%
Time (median): 8.917 μs ┊ GC (median): 0.00%
Time (mean ± σ): 9.159 μs ± 774.910 ns ┊ GC (mean ± σ): 0.00% ± 0.00%
▇▇▂ █▇▇▇▂ ▁▁▁▁▁▁ ▁ ▁ ▂
███▆██████▇███▇▇█▇█▇██████████████▇█▇▇▇▆██▇███▇▇▇▆▆▆▅▆▅▆▅▅▅ █
8.6 μs Histogram: log(frequency) by time 12.2 μs <
Memory estimate: 16 bytes, allocs estimate: 2. |
There was a problem hiding this comment.
This looks already really nice! I think we are not far from merging this. I especially like your trait system to allow all or only a reduced set of variables in the initial condition. I knew you would come up with a very nice idea for implementing that what was very fuzzy in my head 🙂
|
The convergence test and benchmark look nice. Thanks! |
Co-authored-by: Joshua Lampert <[email protected]>
examples/hyperbolic_serre_green_naghdi_1d/hyperbolic_serre_green_naghdi_manufactured.jl
Show resolved
Hide resolved
Co-authored-by: Joshua Lampert <[email protected]>
Co-authored-by: Joshua Lampert <[email protected]>
Next step of #129
TODO
w = -(D1 * v)instead of having to specify a value manually in the initial condition.