Merge pull request #153 from kaipartmann/threads_static

Change multithreading backend to Base.Threads

CITATION.cff

@@ -4,11 +4,13 @@ authors:
 - family-names: Partmann
   given-names: Kai
   orcid: https://orcid.org/0000-0002-5238-4355
+- family-names: Dienst
+  given-names: Manuel
+  orcid: https://orcid.org/0009-0004-9195-0112
 - family-names: Weinberg
   given-names: Kerstin
   orcid: https://orcid.org/0000-0002-2213-8401
 title: "Peridynamics.jl"
-version: 0.2.0
-doi: 10.5281/zenodo.8187478
+doi: 10.5281/zenodo.8187477
 date-released: 2023-07-26
 url: "https://github.com/kaipartmann/Peridynamics.jl"

Project.toml

@@ -13,7 +13,6 @@ LightXML = "9c8b4983-aa76-5018-a973-4c85ecc9e179"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 MPI = "da04e1cc-30fd-572f-bb4f-1f8673147195"
 PointNeighbors = "1c4d5385-0a27-49de-8e2c-43b175c8985c"
-Polyester = "f517fe37-dbe3-4b94-8317-1923a5111588"
 PrecompileTools = "aea7be01-6a6a-4083-8856-8a6e6704d82a"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 ProgressMeter = "92933f4c-e287-5a05-a399-4b506db050ca"
@@ -31,8 +30,7 @@ LibGit2 = "1.8"
 LightXML = "0.9"
 LinearAlgebra = "1.8"
 MPI = "0.20.0"
-PointNeighbors = "0.4"
-Polyester = "0.7.5"
+PointNeighbors = "0.4.5"
 PrecompileTools = "1.0"
 Printf = "1.8"
 ProgressMeter = "1.0"

src/Peridynamics.jl

@@ -7,7 +7,6 @@ using Base.Threads, Printf, LinearAlgebra, StaticArrays, PointNeighbors, Progres
 end
 
 import LibGit2, Dates
-import Polyester: @batch
 
 # Material models
 export BBMaterial, OSBMaterial, NOSBMaterial, CKIMaterial

src/VtkReader/VtkReader.jl

@@ -1,6 +1,5 @@
 module VtkReader
 
-using Polyester: @batch
 using Base64: base64decode
 using CodecZlib: ZlibDecompressor
 using LightXML: LightXML, XMLDocument, XMLElement, parse_string, attribute, child_elements,
@@ -89,7 +88,7 @@ end
 function get_raw_xml_and_data(files::Vector{String})
     vec_raw_xml_str = Vector{String}(undef, length(files))
     vec_raw_data = Vector{Vector{UInt8}}(undef, length(files))
-    @batch for i in eachindex(files)
+    Threads.@threads for i in eachindex(files)
         raw_xml_str, raw_data = get_raw_xml_and_data(files[i])
         vec_raw_xml_str[i] = raw_xml_str
         vec_raw_data[i] = raw_data

src/auxiliary/process_each_export.jl

@@ -101,7 +101,7 @@ function process_each_export_threads(f::F, vtk_files::Vector{String}) where {F<:
     ref_result = read_vtk(first(vtk_files))
     p = Progress(length(vtk_files); dt=1, color=:normal, barlen=20,
                  enabled=progress_bars())
-    @batch for file_id in eachindex(vtk_files)
+    @threads :static for file_id in eachindex(vtk_files)
         process_step(f, ref_result, vtk_files[file_id], file_id)
         next!(p)
     end

src/core/threads_body_data_handler.jl

@@ -25,7 +25,7 @@ function _chop_body_threads(::Type{ChunkType}, body::AbstractBody,
                             solver::AbstractTimeSolver, point_decomp::PointDecomposition,
                             param_spec::AbstractParamSpec) where {ChunkType}
     chunks = Vector{ChunkType}(undef, point_decomp.n_chunks)
-    @batch for chunk_id in eachindex(point_decomp.decomp)
+    @threads :static for chunk_id in eachindex(point_decomp.decomp)
         chunk = BodyChunk(body, solver, point_decomp, chunk_id, param_spec)
         apply_precracks!(chunk, body)
         apply_initial_conditions!(chunk, body)
@@ -38,7 +38,7 @@ end
 function find_halo_exchanges(chunks::Vector{B}) where {B<:AbstractBodyChunk}
     lth_exs = Vector{Vector{HaloExchange}}(undef, length(chunks))
     htl_exs = Vector{Vector{HaloExchange}}(undef, length(chunks))
-    @batch for chunk_id in eachindex(chunks)
+    @threads :static for chunk_id in eachindex(chunks)
         _lth_exs, _htl_exs = find_exs(chunks, chunk_id)
         lth_exs[chunk_id] = _lth_exs
         htl_exs[chunk_id] = _htl_exs
@@ -63,7 +63,7 @@ end
 
 function reorder_htl_exs!(htl_exs::Vector{Vector{HaloExchange}})
     all_htl_exs = reduce(vcat, htl_exs)
-    @batch for chunk_id in eachindex(htl_exs)
+    @threads :static for chunk_id in eachindex(htl_exs)
         htl_exs[chunk_id] = filter(x -> x.dest_chunk_id == chunk_id, all_htl_exs)
     end
     return nothing
@@ -166,7 +166,7 @@ end
 
 function export_reference_results(dh::ThreadsBodyDataHandler, options::AbstractJobOptions)
     options.export_allowed || return nothing
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         _export_results(options, dh.chunks[chunk_id], chunk_id, dh.n_chunks, 0, 0.0)
     end
     return nothing

src/core/threads_multibody_data_handler.jl

@@ -37,7 +37,7 @@ function update_caches!(dh::ThreadsMultibodyDataHandler)
     for body_idx in each_body_idx(dh)
         body_dh = get_body_dh(dh, body_idx)
         position_cache = dh.position_caches[body_idx]
-        @batch for chunk in body_dh.chunks
+        @threads :static for chunk in body_dh.chunks
             @inbounds for (li, i) in enumerate(chunk.ch.loc_points)
                 position_cache[1, i] = chunk.storage.position[1, li]
                 position_cache[2, i] = chunk.storage.position[2, li]

src/discretization/bond_system_corrections.jl

@@ -54,10 +54,10 @@ end
 
 function initialize!(dh::AbstractThreadsBodyDataHandler{BondSystem{EnergySurfaceCorrection}},
                      ::AbstractTimeSolver)
-    @batch for chunk in dh.chunks
+    @threads :static for chunk in dh.chunks
         calc_mfactor!(chunk)
     end
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         exchange_loc_to_halo!(get_mfactor, dh, chunk_id)
         calc_scfactor!(dh.chunks[chunk_id])
     end

src/physics/short_range_force_contact.jl

@@ -113,12 +113,12 @@ function calc_short_range_force_contacts!(dh::AbstractThreadsMultibodyDataHandle
         # update neighborhoodsearches
         posc_a = dh.position_caches[body_a_idx]
         posc_b = dh.position_caches[body_b_idx]
-        update_grid!(contact.nhs, posc_b)
+        update_grid!(contact.nhs, posc_b; parallelization_backend=ThreadsStaticBackend())
 
         # calc contact
         body_dh_a = get_body_dh(dh, body_a_idx)
         volc_b = dh.volume_caches[body_b_idx]
-        @batch for chunk_a in body_dh_a.chunks
+        @threads :static for chunk_a in body_dh_a.chunks
             calc_contact_force_density!(chunk_a, contact, posc_a, posc_b, volc_b)
         end
     end

src/time_solvers/dynamic_relaxation.jl

@@ -107,7 +107,7 @@ function solve!(dh::AbstractDataHandler, dr::DynamicRelaxation,
 end
 
 function init_density_matrix!(dh::AbstractThreadsBodyDataHandler, dr::DynamicRelaxation)
-    @batch for chunk in dh.chunks
+    @threads :static for chunk in dh.chunks
         _init_density_matrix!(chunk, dr, chunk.paramsetup)
     end
     return nothing
@@ -168,16 +168,16 @@ function relaxation_timestep!(dh::AbstractThreadsBodyDataHandler,
                               options::AbstractJobOptions,
                               Δt::Float64, n::Int)
     t = n * Δt
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         chunk = dh.chunks[chunk_id]
         apply_boundary_conditions!(chunk, t)
         update_disp_and_pos!(chunk, Δt)
     end
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         exchange_loc_to_halo!(dh, chunk_id)
         calc_force_density!(dh.chunks[chunk_id])
     end
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         exchange_halo_to_loc!(dh, chunk_id)
         chunk = dh.chunks[chunk_id]
         calc_damage!(chunk)

src/time_solvers/velocity_verlet.jl

@@ -135,7 +135,7 @@ end
 
 function calc_stable_timestep(dh::ThreadsBodyDataHandler, safety_factor::Float64)
     Δt = zeros(length(dh.chunks))
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         Δt[chunk_id] = calc_timestep(dh.chunks[chunk_id])
     end
     return minimum(Δt) * safety_factor
@@ -181,17 +181,17 @@ end
 function verlet_timestep!(dh::AbstractThreadsBodyDataHandler, options::AbstractJobOptions,
                           Δt::Float64, Δt½::Float64, n::Int)
     t = n * Δt
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         chunk = dh.chunks[chunk_id]
         update_vel_half!(chunk, Δt½)
         apply_boundary_conditions!(chunk, t)
         update_disp_and_pos!(chunk, Δt)
     end
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         exchange_loc_to_halo!(dh, chunk_id)
         calc_force_density!(dh.chunks[chunk_id])
     end
-    @batch for chunk_id in eachindex(dh.chunks)
+    @threads :static for chunk_id in eachindex(dh.chunks)
         exchange_halo_to_loc!(dh, chunk_id)
         chunk = dh.chunks[chunk_id]
         calc_damage!(chunk)
@@ -206,13 +206,13 @@ function verlet_timestep!(dh::AbstractThreadsMultibodyDataHandler,
     t = n * Δt
     for body_idx in each_body_idx(dh)
         body_dh = get_body_dh(dh, body_idx)
-        @batch for chunk_id in eachindex(body_dh.chunks)
+        @threads :static for chunk_id in eachindex(body_dh.chunks)
             chunk = body_dh.chunks[chunk_id]
             update_vel_half!(chunk, Δt½)
             apply_boundary_conditions!(chunk, t)
             update_disp_and_pos!(chunk, Δt)
         end
-        @batch for chunk_id in eachindex(body_dh.chunks)
+        @threads :static for chunk_id in eachindex(body_dh.chunks)
             exchange_loc_to_halo!(body_dh, chunk_id)
             calc_force_density!(body_dh.chunks[chunk_id])
         end
@@ -222,7 +222,7 @@ function verlet_timestep!(dh::AbstractThreadsMultibodyDataHandler,
     for body_idx in each_body_idx(dh)
         body_dh = get_body_dh(dh, body_idx)
         body_name = get_body_name(dh, body_idx)
-        @batch for chunk_id in eachindex(body_dh.chunks)
+        @threads :static for chunk_id in eachindex(body_dh.chunks)
             exchange_halo_to_loc!(body_dh, chunk_id)
             chunk = body_dh.chunks[chunk_id]
             calc_damage!(chunk)