More flexibility & consistency (across dimensions) to `compute_coefficients!` for discontinuous ICs #2433

DanielDoehring · 2025-06-10T07:57:50Z

See the item in the Trixi Meeting

Major issue: The extensively used initial_condition_weak_blast_wave - if this is changed, we have lots of CI values to update.

github-actions · 2025-06-10T07:58:05Z

codecov · 2025-06-10T08:26:38Z

Codecov Report

Attention: Patch coverage is 55.37634% with 83 lines in your changes missing coverage. Please review.

Project coverage is 96.56%. Comparing base (da9e72e) to head (1b82431).

Files with missing lines	Patch %	Lines
src/solvers/dg.jl	56.20%	53 Missing ⚠️
...dgsem/elixir_euler_riemannproblem_quadrants_amr.jl	0.00%	15 Missing ⚠️
...dgsem/elixir_euler_riemannproblem_quadrants_amr.jl	0.00%	15 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2433      +/-   ##
==========================================
- Coverage   96.77%   96.56%   -0.20%     
==========================================
  Files         495      497       +2     
  Lines       41055    41214     +159     
==========================================
+ Hits        39727    39798      +71     
- Misses       1328     1416      +88

Flag	Coverage Δ
unittests	`96.56% <55.38%> (-0.20%)`	⬇️

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.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

ranocha

Interesting proposal. I am looking forward to the discussion in our next meeting.

ranocha · 2025-06-11T15:24:54Z

examples/structured_1d_dgsem/elixir_advection_shockcapturing.jl

-
-initial_condition = initial_condition_composite
+# Note calling the constructor of the struct: `InitialConditionComposite()` instead of `initial_condition_composite` !
+const initial_condition = InitialConditionComposite()


Does setting this to const conflict with including different elixirs in a single REPL session?

Yes, that will cause errors. I took inspiration from

Trixi.jl/src/equations/numerical_fluxes.jl

Line 244 in f35d749

const flux_lax_friedrichs = FluxLaxFriedrichs()

It is required there in our code but not here since we just pass the IC to the semidiscretization. So it should be fine without const here

ranocha · 2025-06-11T15:30:43Z

src/solvers/dg.jl

+            if i == 1
+                x_node = SVector(nextfloat(x_node[1]), x_node[2])
+            elseif i == nnodes(dg)
+                x_node = SVector(prevfloat(x_node[1]), x_node[2])
+            end
+            if j == 1
+                x_node = SVector(x_node[1], nextfloat(x_node[2]))
+            elseif j == nnodes(dg)
+                x_node = SVector(x_node[1], prevfloat(x_node[2]))
+            end


This assumes that the element lies in the coordinate direction, which may not be the case for unstructured and curved meshes in general, is it?

True, we would need to check the direction of the face.

As discussed in the meeting, it would also be fine to implement this only for the TreeMesh when a hard error is thrown for the other mesh types in 2D and 3D.

ranocha · 2025-06-11T15:31:03Z

src/solvers/dg.jl

+            if i == 1
+                x_node = SVector(nextfloat(x_node[1]), x_node[2], x_node[3])
+            elseif i == nnodes(dg)
+                x_node = SVector(prevfloat(x_node[1]), x_node[2], x_node[3])
+            end


DanielDoehring · 2025-06-13T07:27:29Z

src/solvers/dg.jl

+                # NOTE: This still assumes that the first node index somewhat resembles the first coordinate direction!
+                if x_node[1] < x_node_n[1]
+                    x_node = SVector(nextfloat(x_node[1]), x_node[2])
+                else
+                    x1_min_at_node1 = false
+                    x_node = SVector(prevfloat(x_node[1]), x_node[2])
+                end


This is probably the most convenient way of dragging the nodes inwards. As mentioned in the comment, this is probably not really fail-safe if a mesh is supplied/constructed where the i index does not really expand the element in 1 direction but 2 direction. I would assume that this is possible depending on the exact order of node numbering in a supplied mesh file. @andrewwinters5000 is that right?

DanielDoehring · 2025-06-13T07:28:43Z

src/solvers/dgsem_structured/dg_2d.jl

@@ -437,7 +437,7 @@ end
    left_direction = right_direction - 1

    for i in eachnode(dg)
-        if orientation == 1
+        if orientation == 1 # (x-direction)


See

Trixi.jl/src/solvers/dgsem_structured/dg_2d.jl

Lines 404 to 411 in da9e72e

# Interfaces in x-direction (`orientation` = 1)

calc_interface_flux!(elements.surface_flux_values,

elements.left_neighbors[1, element],

element, 1, u, mesh,

nonconservative_terms, equations,

surface_integral, dg, cache)

# Interfaces in y-direction (`orientation` = 2)

DanielDoehring · 2025-06-13T07:29:08Z

src/solvers/dgsem_structured/dg_3d.jl

@@ -552,7 +552,7 @@ end
    left_direction = right_direction - 1

    for j in eachnode(dg), i in eachnode(dg)
-        if orientation == 1
+        if orientation == 1 # (x-direction)


see

Trixi.jl/src/solvers/dgsem_structured/dg_3d.jl

Lines 513 to 527 in da9e72e

# Interfaces in x-direction (`orientation` = 1)

calc_interface_flux!(elements.surface_flux_values,

elements.left_neighbors[1, element],

element, 1, u, mesh,

nonconservative_terms, equations,

surface_integral, dg, cache)

# Interfaces in y-direction (`orientation` = 2)

calc_interface_flux!(elements.surface_flux_values,

elements.left_neighbors[2, element],

element, 2, u, mesh,

nonconservative_terms, equations,

surface_integral, dg, cache)

# Interfaces in z-direction (`orientation` = 3)

DanielDoehring · 2025-06-13T09:52:22Z

Plot of current implementation for 4 Quadrant Riemann problem

Plot of new implementation for 4 Quadrant Riemann problem

Flexibility compute_coefficients for discont. ICs

fae1b81

DanielDoehring added discussion consistency Make Michael happy example Adding examples (elixirs) labels Jun 10, 2025

Merge branch 'main' into Example2D_RP

389ee32

shorten

2ce0f60

DanielDoehring added the breaking label Jun 10, 2025

non-weak-blast wave discontinuous ICs

1d9f567

DanielDoehring added the high-priority label Jun 10, 2025

DanielDoehring added 4 commits June 10, 2025 11:58

use correct ICs

1e3b39b

fix errors cont. test cases

bec8402

update

cfd114f

revert

105497b

ranocha reviewed Jun 11, 2025

View reviewed changes

DanielDoehring added 5 commits June 12, 2025 15:15

no const initial cond

d5a9dc3

rename

a066eb7

examples

78a2c73

comment

cd90f4f

3d mesh general

7c89473

DanielDoehring commented Jun 13, 2025

View reviewed changes

DanielDoehring added 5 commits June 13, 2025 10:10

fix

560bb78

Merge branch 'main' into Example2D_RP

cf52b09

typo

4d6bafb

test vals

5e3a019

test vals

1b82431

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

More flexibility & consistency (across dimensions) to `compute_coefficients!` for discontinuous ICs #2433

More flexibility & consistency (across dimensions) to `compute_coefficients!` for discontinuous ICs #2433

Uh oh!

DanielDoehring commented Jun 10, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Jun 10, 2025

Uh oh!

codecov bot commented Jun 10, 2025 •

edited

Loading

Uh oh!

ranocha left a comment

Uh oh!

ranocha Jun 11, 2025

Uh oh!

DanielDoehring Jun 11, 2025

Uh oh!

ranocha Jun 11, 2025

Uh oh!

ranocha Jun 11, 2025

Uh oh!

DanielDoehring Jun 11, 2025

Uh oh!

ranocha Jun 12, 2025

Uh oh!

ranocha Jun 11, 2025

Uh oh!

DanielDoehring Jun 13, 2025

Uh oh!

DanielDoehring Jun 13, 2025

Uh oh!

DanielDoehring Jun 13, 2025

Uh oh!

DanielDoehring commented Jun 13, 2025

Uh oh!

Uh oh!

	# Interfaces in x-direction (`orientation` = 1)
	calc_interface_flux!(elements.surface_flux_values,
	elements.left_neighbors[1, element],
	element, 1, u, mesh,
	nonconservative_terms, equations,
	surface_integral, dg, cache)

	# Interfaces in y-direction (`orientation` = 2)

More flexibility & consistency (across dimensions) to compute_coefficients! for discontinuous ICs #2433

Are you sure you want to change the base?

More flexibility & consistency (across dimensions) to compute_coefficients! for discontinuous ICs #2433

Uh oh!

Conversation

DanielDoehring commented Jun 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jun 10, 2025

Review checklist

Purpose and scope

Code quality

Documentation

Testing

Performance

Verification

Uh oh!

codecov bot commented Jun 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

ranocha left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

DanielDoehring commented Jun 13, 2025

Uh oh!

Uh oh!

More flexibility & consistency (across dimensions) to `compute_coefficients!` for discontinuous ICs #2433

More flexibility & consistency (across dimensions) to `compute_coefficients!` for discontinuous ICs #2433

DanielDoehring commented Jun 10, 2025 •

edited

Loading

codecov bot commented Jun 10, 2025 •

edited

Loading