Skip to content

Add parallel intitialization for TLSPH arrays to ensure NUMA awareness [Alternative] #1294

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
efaulhaber wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Add parallel intitialization for TLSPH arrays to ensure NUMA awareness [Alternative] #1294

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
010fecb
Add parallel intitialization for TLSPH arrays to ensure NUMA awareness
efaulhaber Aug 17, 2026
fa834b3
Fix
efaulhaber Aug 17, 2026
2a94489
Update comment
efaulhaber Aug 17, 2026
f9ca795
Add NEWS.md entry
efaulhaber Aug 17, 2026
104fe9a
Fix bounds checks
efaulhaber Aug 17, 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
22 changes: 18 additions & 4 deletions src/schemes/structure/total_lagrangian_sph/system.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,10 +145,24 @@ function TotalLagrangianSPHSystem(initial_condition; smoothing_kernel, smoothing
poisson_ratio_sorted = poisson_ratio
end

initial_coordinates = copy(initial_condition_sorted.coordinates)
current_coordinates = copy(initial_condition_sorted.coordinates)
mass = copy(initial_condition_sorted.mass)
material_density = copy(initial_condition_sorted.density)
initial_coordinates = similar(initial_condition_sorted.coordinates)
current_coordinates = similar(initial_condition_sorted.coordinates)
mass = similar(initial_condition_sorted.mass)
material_density = similar(initial_condition_sorted.density)

# Initialize the runtime arrays in parallel so that first-touch allocation places
Comment thread
efaulhaber marked this conversation as resolved.
Outdated
# their memory close to the threads that will access it during the simulation.
# The semidiscretization backend is not available in the constructor yet, and using
# Polyester here is harmless for serial and GPU simulations.
parallelization_backend = PolyesterBackend()
Comment thread
efaulhaber marked this conversation as resolved.
Comment thread
efaulhaber marked this conversation as resolved.
copyto!(ThreadedBroadcastArray(initial_coordinates; parallelization_backend),
initial_condition_sorted.coordinates)
copyto!(ThreadedBroadcastArray(current_coordinates; parallelization_backend),
initial_condition_sorted.coordinates)
copyto!(ThreadedBroadcastArray(mass; parallelization_backend),
initial_condition_sorted.mass)
copyto!(ThreadedBroadcastArray(material_density; parallelization_backend),
initial_condition_sorted.density)
correction_matrix = Array{ELTYPE, 3}(undef, NDIMS, NDIMS, n_particles)
pk1_rho2 = Array{ELTYPE, 3}(undef, NDIMS, NDIMS, n_particles)
deformation_grad = Array{ELTYPE, 3}(undef, NDIMS, NDIMS, n_particles)
Expand Down
Loading