Skip to content

Add optional CUDA neighbor-loop and RunSimulationCUDA #139

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
AhmedSalih3d wants to merge 2 commits into
base: codex/optimize-neighborlist-code-performance
Choose a base branch
from
Open

Add optional CUDA neighbor-loop and RunSimulationCUDA #139

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
61deef3
Add CUDA neighbor loop path
AhmedSalih3d Jan 15, 2026
45f9242
Remove no_escape usage in CUDA loop
AhmedSalih3d Jan 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,7 @@ src/
├── PreProcess.jl # Load inputs and allocate arrays
├── ProduceHDFVTK.jl # Write simulation data in HDF5/VTK format
├── SPHCellList.jl # Custom neighbour search and time stepping
├── SPHCUDA.jl # Optional CUDA interaction path
├── SPHDensityDiffusionModels.jl # Density diffusion implementations
├── SPHExample.jl # Glue module re-exporting all functions
├── SPHKernels.jl # SPH kernel definitions
Expand Down Expand Up @@ -91,6 +92,20 @@ one plus the particles in its neighbor stencil.
Neighbor rebuilds use a per-cell ordering buffer without reordering particle
storage.

### CUDA Acceleration (Optional)

The solver can offload the particle interaction loops to a CUDA GPU when
`CUDA.jl` is available in your Julia environment. The helper `CUDAAvailable()`
reports whether CUDA is functional, and the `RunSimulationCUDA` entry point
mirrors `RunSimulation` while delegating the interaction loops to the GPU.
Currently, the CUDA path supports the `NoShifting` and `NoKernelOutput`
configurations with `LinearDensityDiffusion` or `ZeroDensityDiffusion` plus
`ArtificialViscosity` or `ZeroViscosity`.

To try the GPU workflow, install CUDA.jl in your environment and run
`example/StillWedgeMDBC.jl`, which will automatically pick `RunSimulationCUDA`
when CUDA is available.

## Help

Questions or issues can be posted on the GitHub issue tracker. Response times may vary but all feedback is welcome.
Expand Down
39 changes: 26 additions & 13 deletions example/StillWedgeMDBC.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ let
SimConstantsWedge = SimulationConstants{FloatType}(dx=0.02,c₀=42.48576250492629, δᵩ = 0.1, CFL=0.5)
# SimConstantsWedge = SimulationConstants{FloatType}(dx=0.01,c₀=43.4, δᵩ = 0.1, CFL=0.2)
#
SimMetaDataWedge = SimulationMetaData{Dimensions,FloatType,NoShifting,NoKernelOutput,NoMDBC,StoreLog}(
SimMetaDataWedge = SimulationMetaData{Dimensions,FloatType,NoShifting,NoKernelOutput,SimpleMDBC,StoreLog}(
SimulationName="StillWedge",
SaveLocation="W:/Simulations/StillWedge2D_MDBC",
SimulationTime=4.0,
Expand Down Expand Up @@ -49,17 +49,31 @@ let

SimKernel = SPHKernelInstance{Dimensions, FloatType}(WendlandC2(); dx = SimConstantsWedge.dx)

@profview RunSimulation(
SimGeometry = SimulationGeometry,
SimMetaData = SimMetaDataWedge,
SimConstants = SimConstantsWedge,
SimKernel = SimKernel,
SimLogger = SimLogger,
SimParticles = SimParticles,
SimViscosity = ArtificialViscosity(),
SimDensityDiffusion = LinearDensityDiffusion(),
ParticleNormalsPath = "./input/still_wedge_mdbc/StillWedge_Dp$(SimConstantsWedge.dx)_GhostNodes_Correct.csv"
)
if CUDAAvailable()
@profview RunSimulationCUDA(
SimGeometry = SimulationGeometry,
SimMetaData = SimMetaDataWedge,
SimConstants = SimConstantsWedge,
SimKernel = SimKernel,
SimLogger = SimLogger,
SimParticles = SimParticles,
SimViscosity = ArtificialViscosity(),
SimDensityDiffusion = LinearDensityDiffusion(),
ParticleNormalsPath = "./input/still_wedge_mdbc/StillWedge_Dp$(SimConstantsWedge.dx)_GhostNodes_Correct.csv"
)
else
@profview RunSimulation(
SimGeometry = SimulationGeometry,
SimMetaData = SimMetaDataWedge,
SimConstants = SimConstantsWedge,
SimKernel = SimKernel,
SimLogger = SimLogger,
SimParticles = SimParticles,
SimViscosity = ArtificialViscosity(),
SimDensityDiffusion = LinearDensityDiffusion(),
ParticleNormalsPath = "./input/still_wedge_mdbc/StillWedge_Dp$(SimConstantsWedge.dx)_GhostNodes_Correct.csv"
)
end

# This can be used to plot pressure profile results after simulation
# using Plots
Expand Down Expand Up @@ -97,4 +111,3 @@ let
# display(plt)
end


Loading