Skip to content

Speed up sparsity-pattern sort with packed integer keys #301

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

Merged
lxmota merged 1 commit into from
May 21, 2026
Merged

Speed up sparsity-pattern sort with packed integer keys #301

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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: 10 additions & 5 deletions src/assemblers/SparsityPatterns.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,8 +93,11 @@ function SparseMatrixPattern(dof::DofManager)
cscrowval = Vector{Int64}(undef, 0)
cscnzval = Vector{Float64}(undef, 0)

# set permutation
ks = map((i, j) -> (i, j), Is, Js)
# set permutation — pack (row, col) into one Int64 key so sortperm takes the
# integer radix-sort path instead of an O(N log N) tuple comparison sort.
# DOF indices are << 2^32, so the packed key preserves (i, j) lexicographic
# order and (sortperm being stable) yields an identical permutation.
ks = map((i, j) -> (Int64(i) << 32) | Int64(j), Is, Js)
permutation = sortperm(ks)

pattern = SparseMatrixPattern(
Expand Down Expand Up @@ -246,9 +249,11 @@ function _update_dofs!(pattern::SparseMatrixPattern, dof, dirichlet_dofs)
resize!(pattern.csrcolval, length(pattern.Is))
resize!(pattern.csrnzval, length(pattern.Is))

# TODO below is half of the runtime in this method
# it can likely be sped up
ks = map((i, j) -> (i, j), pattern.Is, pattern.Js)
# Pack (row, col) into one Int64 key so sortperm takes the integer radix-sort
# path instead of an O(N log N) tuple comparison sort. DOF indices are
# << 2^32, so the packed key preserves (i, j) lexicographic order and
# (sortperm being stable) yields an identical permutation.
ks = map((i, j) -> (Int64(i) << 32) | Int64(j), pattern.Is, pattern.Js)
permutation = sortperm(ks)
resize!(pattern.permutation, length(permutation))
pattern.permutation .= permutation
Expand Down
Loading