Make DEM run on GPUs and with Float32

[efaulhaber]

The rectangular tank example now runs on Apple GPUs (and should therefore also run on Nvidia and AMD):

julia> using TrixiParticles, Metal

julia> trixi_include_changeprecision(Float32, "examples/dem/rectangular_tank_2d.jl", sol=nothing);
[ Info: You just called `trixi_include`. Julia may now compile the code, please be patient.

julia> min_corner = minimum(tank.boundary.coordinates, dims=2)
2×1 Matrix{Float32}:
 -0.15
 -0.15

julia> max_corner = maximum(tank.boundary.coordinates, dims=2)
2×1 Matrix{Float32}:
 2.15
 4.15

julia> cell_list = FullGridCellList(; min_corner, max_corner);

julia> neighborhood_search = GridNeighborhoodSearch{2}(; cell_list, update_strategy=ParallelUpdate());

julia> trixi_include_changeprecision(Float32, "examples/dem/rectangular_tank_2d.jl", neighborhood_search=neighborhood_search, parallelization_backend=MetalBackend());

[efaulhaber]

/run-gpu-tests

[efaulhaber]

/run-gpu-tests

[Copilot]

Pull Request Overview

This pull request enables the Discrete Element Method (DEM) to run on GPUs with Float32 precision support, primarily for Apple Metal GPUs but compatible with NVIDIA and AMD as well. The changes focus on making the code GPU-compatible through Adapt.jl integration and ensuring proper Float32 handling throughout the DEM implementation.

Key changes:

• Refactored DEM and boundary system constructors to work with Adapt.jl for GPU compatibility
• Replaced hardcoded floating-point literals with type-safe conversions to support Float32
• Added GPU test coverage for the DEM rectangular tank example

Reviewed Changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
test/examples/gpu.jl	Added comprehensive GPU test for DEM rectangular tank example with Float32 precision
src/schemes/structure/discrete_element_method/system.jl	Refactored constructor for Adapt.jl compatibility and fixed Float32 support in radius calculations
src/schemes/structure/discrete_element_method/rhs.jl	Replaced hardcoded doubles with type-safe conversions and updated comments
src/schemes/structure/discrete_element_method/contact_models.jl	Reordered arithmetic to avoid hardcoded division
src/schemes/boundary/dem_boundary/system.jl	Refactored constructor for Adapt.jl compatibility with proper Float32 handling
src/general/gpu.jl	Registered DEMSystem and BoundaryDEMSystem with Adapt.jl
examples/dem/rectangular_tank_2d.jl	Updated perturbation logic and added default parallelization backend

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[efaulhaber]

/run-gpu-tests

[LasNikas]

/run-gpu-tests

[svchb]

/run-gpu-tests

[codecov]

Codecov Report

❌ Patch coverage is 75.86207% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.63%. Comparing base (7f6ec8a) to head (583ee09).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...c/schemes/structure/discrete_element_method/rhs.jl	0.00%	6 Missing ⚠️
...chemes/structure/discrete_element_method/system.jl	92.85%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #979       +/-   ##
===========================================
+ Coverage   64.87%   89.63%   +24.76%     
===========================================
  Files         120      120               
  Lines        8560     8587       +27     
===========================================
+ Hits         5553     7697     +2144     
+ Misses       3007      890     -2117     

Flag	Coverage Δ
total	89.63% <75.86%> (?)
unit	64.84% <34.48%> (-0.03%)	⬇️

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.