Add cubed sphere test cases #199

benegee · 2024-07-24T21:33:31Z

New test cases:

baroclinic instability
poor man's tracer advection with AMR (as shown at MOW24)

Requires: trixi-framework/Trixi.jl#1918

To test this in the meantime:

Build and install t8code
- has to be main

Build and install libtrixi

needs this branch: bg/baroclinic-instability

e.g.

cmake -DCMAKE_INSTALL_PREFIX=~/install/libtrixi-baro \
      -DCMAKE_BUILD_TYPE=DEBUG \
      -DT8CODE_ROOT=~/install/t8code \
      .. 
make
make install

Build julia project
- cd libtrixi-julia-baro
- e.g. ./libtrixi-baro/bin/libtrixi-init-julia --t8code-library ../t8code/lib/libt8.so ../libtrixi-baro
- JULIA_DEPOT_PATH=./julia-depot julia --project=.
- Package manager ]
- pkg> dev --local Trixi
- leave Julia, cd dev/Trixi, git switch feature-t8code-cubed-sphere-setup, cd -
- JULIA_DEPOT_PATH=./julia-depot julia --project=.
- Package manager ]
- update
- maybe downgrade XML2_jll: add [email protected]
- precompile

Run test case

cd `libtrixi-baro
LIBTRIXI_DEBUG=all ./bin/trixi_controller_baroclinic_c ../libtrixi-julia-baro  \
    ./share/libtrixi/LibTrixi.jl/examples/libelixir_t8code3d_euler_baroclinic_instability.jl

add parameter index and only get averaged of the variable at position index

proof of concept for creating data vectors in libelixirs which can be filled later by external applications

required in this PR, subject to change

demonstrates source terms via database

…ility.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]>

…i into more-data-access

codecov · 2024-12-20T09:16:28Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.62%. Comparing base (8e9d576) to head (b379e50).
Report is 2 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #199       +/-   ##
===========================================
+ Coverage   82.43%   95.62%   +13.19%     
===========================================
  Files          21       23        +2     
  Lines         905     1097      +192     
  Branches       52       63       +11     
===========================================
+ Hits          746     1049      +303     
+ Misses        155       44      -111     
  Partials        4        4

Flag	Coverage Δ
unittests	`95.62% <100.00%> (+13.19%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

benegee · 2024-12-20T09:16:58Z

Finally! 🎉

@jmark's cubed sphere fix was released with t8code v3.0.1. which made it into t8code_jll.jll v3.0.1, which made it into T8code v0.7.4, which made it into Trixi 0.9.12, and is finally here!

sloede

Very cool example case! Just a few minor notes...

sloede · 2024-12-22T06:36:45Z

.github/workflows/ci.yml

@@ -198,6 +196,7 @@ jobs:
          cd build
          cmake .. -DCMAKE_INSTALL_PREFIX=../install \
                   -DCMAKE_BUILD_TYPE=Debug \
+                   -DT8CODE_ROOT=$PWD/../t8code-local/prefix \


Good catch!
At some point I ran t8code elixirs also with PC, but not anymore.
Removed this line.

sloede · 2024-12-22T06:39:57Z

LibTrixi.jl/examples/libelixir_tree2d_warm_bubble.jl

Why delete this file?

I think it was temporary, for some demonstration I guess.
It does not have unique features. And it will at some point return when there will be moist air.

sloede · 2024-12-22T06:47:18Z

LibTrixi.jl/examples/libelixir_t8code3d_euler_baroclinic_instability.jl

+    # provide some data because calc_sources! will already be called during initialization
+    zero_data = zeros(Float64, nelements*nnodes)
+    registry[1] = zero_data
+    registry[2] = zero_data
+    registry[3] = zero_data
+    registry[4] = zero_data


The current implementation does not seem correct. All entries of registry are assigned the same vector zero_data, i.e., they point to a single data structure.

I see that they will be overwritten later again. However, my feeling is that it would be better to make sure that this is not accidentally considered a valid implementation:

Suggested change

# provide some data because calc_sources! will already be called during initialization

zero_data = zeros(Float64, nelements*nnodes)

registry[1] = zero_data

registry[2] = zero_data

registry[3] = zero_data

registry[4] = zero_data

# provide some data because calc_sources! will already be called during initialization

# Note: the data pointers in the registry will be overwritten before the first real use

zero_data = zeros(Float64, nelements*nnodes)

registry[1] = zero_data

registry[2] = zero_data

registry[3] = zero_data

registry[4] = zero_data

As a side note, I am starting to wonder if storage for source terms should maybe live on the Trixi.jl side instead of the C side. This goes against previous statements I made, and I wouldn't change it now before we have gathered more experience with it. However, it seems to me that it is overly convoluted this way, especially since it's effectively Trixi.jl that needs and controls the memory, not the C side

True! I now allocate separate arrays to avoid confusion.

The argument for memory management on C side was that some large software frameworks have their own storage pool and would like to use it.

LibTrixi.jl/examples/libelixir_t8code3d_euler_baroclinic_instability.jl

benegee and others added 30 commits February 22, 2024 17:49

add getter for number of dofs

d5ecc49

change trixi_load_cell_average

72709e5

add parameter index and only get averaged of the variable at position index

fix: ensure Int32

71ba4bf

adapt reference value

9c83b05

add getter for all dofs values

5b3960b

adapt next value

3e31f4a

add trixi_store_in_database

2d45c62

proof of concept for creating data vectors in libelixirs which can be filled later by external applications

add trixi_ndofs_element

92d817a

add tests

530f5cc

add missing parts in tests

c7d38d4

fix tests

ba00c6f

deallocate first

cdf6e5e

add trixi_load_prim to Fortran API

aa6d8c7

reference value

cf43795

get doxygen right

dda491a

Merge branch 'main' into more-data-access

35513ac

Merge branch 'main' into bg/store-in-database

bf0b732

add trixi_get_time, trixi_load_node_coordinates

b864850

required in this PR, subject to change

implement source terms controller

9430f2d

demonstrates source terms via database

libelixir with source terms via database

5956a78

change Array to Vector

ccd23fc

spelling

1be73d9

libelixir for baroclinic instability

584a674

remove method only required by steady state correction

7c82c90

Update LibTrixi.jl/examples/libelixir_p4est3d_euler_baroclinic_instab…

d1850a7

…ility.jl Co-authored-by: Michael Schlottke-Lakemper <[email protected]>

Merge branch 'main' into bg/baroclinic-instability

fd103d9

Merge branch 'more-data-access' into bg/store-in-database

6170610

Merge branch 'more-data-access' of github.com:trixi-framework/libtrix…

94cc1c7

…i into more-data-access

make everything more consistent!

7dbb2c6

use const double *

7906080

benegee added 2 commits November 20, 2024 15:17

fixes

c6ec30a

adapt index.md to README.md

b866b8b

benegee changed the title ~~Add t8code baroclinic instability test case~~ Add cubed sphere test cases Nov 20, 2024

benegee added 12 commits November 20, 2024 16:29

change to non-adaptive time integrator

ac1b653

missing controlers in ci checks

58444f3

Merge branch 'main' into bg/baroclinic-instability

5f02f44

use zeros in initial data

218cd91

missing renamings

47a2e4c

time not needed anymore

8bd1b23

tuned simulation parameters

85f882c

itree and ielement need to be 0-based

49fa815

one more to free

5cf1aec

ouch...

8ccad7b

less output

5599568

typo in docstring

142e6a2

benegee mentioned this pull request Dec 2, 2024

RFC: raw data access #219

Open

benegee added 4 commits December 17, 2024 09:20

adopt new t8code ForestWrapper

8794de7

Merge branch 'main' into bg/baroclinic-instability

b7bd4c4

switch to newest t8code release

5f5fe48

cubed sphere stuff only works with latest Trixi release

f341561

benegee marked this pull request as ready for review December 20, 2024 07:46

benegee added 2 commits December 20, 2024 09:46

missed renaming

889f990

enable t8code for package compiler as well

63ece7a

benegee requested a review from sloede December 20, 2024 09:17

sloede requested changes Dec 22, 2024

View reviewed changes

benegee added 2 commits January 7, 2025 12:27

review

90312a1

filter out baroclinic example

b379e50

benegee requested a review from sloede January 7, 2025 16:45

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add cubed sphere test cases #199

Add cubed sphere test cases #199

benegee commented Jul 24, 2024 •

edited by sloede

Loading

codecov bot commented Dec 20, 2024 •

edited

Loading

benegee commented Dec 20, 2024

sloede left a comment

sloede Dec 22, 2024

benegee Jan 7, 2025

sloede Dec 22, 2024

benegee Jan 7, 2025

sloede Dec 22, 2024

sloede Dec 22, 2024

benegee Jan 7, 2025

benegee Jan 7, 2025

Add cubed sphere test cases #199

Are you sure you want to change the base?

Add cubed sphere test cases #199

Conversation

benegee commented Jul 24, 2024 • edited by sloede Loading

codecov bot commented Dec 20, 2024 • edited Loading

Codecov Report

benegee commented Dec 20, 2024

sloede left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

benegee commented Jul 24, 2024 •

edited by sloede

Loading

codecov bot commented Dec 20, 2024 •

edited

Loading