Skip to content

Dual simplex performance improvements #1043

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
chris-maes wants to merge 7 commits into
base: main
Choose a base branch
from
Open

Dual simplex performance improvements #1043

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
652aeeb
Dual Simplex Performance improvements: no BTRAN for steepest edge ent…
chris-maes Apr 7, 2026
b4121cc
Merge remote-tracking branch 'cuopt-nvidia/release/26.04' into steepe…
chris-maes Apr 7, 2026
3c20852
Improve clean_up_infeasibilities. Style fixes
chris-maes Apr 8, 2026
f41e81b
Fix issue if multiple threads altered the same Arow
chris-maes Apr 14, 2026
db17f4d
Remove dead debug code
chris-maes Apr 14, 2026
46d8a31
Style fixes
chris-maes Apr 14, 2026
45ca97e
Merge remote-tracking branch 'cuopt-nvidia/main' into steepest_edge_e…
chris-maes Apr 14, 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
2 changes: 1 addition & 1 deletion cpp/src/branch_and_bound/branch_and_bound.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2506,7 +2506,7 @@ mip_status_t branch_and_bound_t<i_t, f_t>::solve(mip_solution_t<i_t, f_t>& solut
set_uninitialized_steepest_edge_norms(original_lp_, basic_list, edge_norms_);

pc_.resize(original_lp_.num_cols);
original_lp_.A.transpose(pc_.AT);
pc_.Arow = Arow_;
{
raft::common::nvtx::range scope_sb("BB::strong_branching");
strong_branching<i_t, f_t>(original_lp_,
Expand Down
34 changes: 17 additions & 17 deletions cpp/src/branch_and_bound/pseudo_costs.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,14 +66,14 @@ template <typename i_t, typename f_t>
objective_change_estimate_t<f_t> single_pivot_objective_change_estimate(
const lp_problem_t<i_t, f_t>& lp,
const simplex_solver_settings_t<i_t, f_t>& settings,
const csc_matrix_t<i_t, f_t>& A_transpose,
const csr_matrix_t<i_t, f_t>& Arow,
const std::vector<variable_status_t>& vstatus,
i_t variable_j,
i_t basic_j,
const lp_solution_t<i_t, f_t>& lp_solution,
const std::vector<i_t>& basic_list,
const std::vector<i_t>& nonbasic_list,
const std::vector<i_t>& nonbasic_mark,
const std::vector<i_t>& nonbasic_end,
basis_update_mpf_t<i_t, f_t>& basis_factors,
std::vector<i_t>& workspace,
std::vector<f_t>& delta_z,
Expand All @@ -82,7 +82,6 @@ objective_change_estimate_t<f_t> single_pivot_objective_change_estimate(
// Compute the objective estimate for the down and up branches of variable j
assert(variable_j >= 0);
assert(basic_j >= 0);

// Down branch
i_t direction = -1;
sparse_vector_t<i_t, f_t> e_k(lp.num_rows, 0);
Expand All @@ -104,11 +103,11 @@ objective_change_estimate_t<f_t> single_pivot_objective_change_estimate(
std::vector<i_t> delta_z_indices;
// delta_z starts out all zero
if (use_transpose) {
compute_delta_z(A_transpose,
compute_delta_z(Arow,
delta_y,
variable_j,
direction,
nonbasic_mark,
nonbasic_end,
workspace,
delta_z_indices,
delta_z,
Expand Down Expand Up @@ -234,10 +233,8 @@ void initialize_pseudo_costs_with_estimate(const lp_problem_t<i_t, f_t>& lp,
basic_map[basic_list[i]] = i;
}

std::vector<i_t> nonbasic_mark(n, -1);
for (i_t i = 0; i < n - m; i++) {
nonbasic_mark[nonbasic_list[i]] = i;
}
std::vector<i_t> nonbasic_end(lp.num_rows);
compute_initial_nonbasic_end(basic_map, pc.Arow, nonbasic_end);

for (i_t k = 0; k < fractional.size(); k++) {
const i_t j = fractional[k];
Expand All @@ -246,14 +243,14 @@ void initialize_pseudo_costs_with_estimate(const lp_problem_t<i_t, f_t>& lp,
objective_change_estimate_t<f_t> estimate =
single_pivot_objective_change_estimate(lp,
settings,
pc.AT,
pc.Arow,
vstatus,
j,
basic_map[j],
lp_solution,
basic_list,
nonbasic_list,
nonbasic_mark,
nonbasic_end,
basis_factors,
workspace,
delta_z,
Expand Down Expand Up @@ -1476,25 +1473,28 @@ i_t pseudo_costs_t<i_t, f_t>::reliable_variable_selection(
basic_map[worker->basic_list[i]] = i;
}

std::vector<i_t> nonbasic_mark(n, -1);
for (i_t i = 0; i < n - m; i++) {
nonbasic_mark[worker->nonbasic_list[i]] = i;
}
// Each thread will have a different basis
// So we need to make a copy of Arow before we permute the columns
// so that nonbasic variables appear first
csr_matrix_t<i_t, f_t> local_Arow = Arow;

std::vector<i_t> nonbasic_end(m);
compute_initial_nonbasic_end(basic_map, local_Arow, nonbasic_end);

for (auto& [score, j] : unreliable_list) {
if (pseudo_cost_num_down[j] == 0 || pseudo_cost_num_up[j] == 0) {
// Estimate the objective change by performing a single pivot of dual simplex.
objective_change_estimate_t<f_t> estimate =
single_pivot_objective_change_estimate(worker->leaf_problem,
settings,
AT,
local_Arow,
node_ptr->vstatus,
j,
basic_map[j],
leaf_solution,
worker->basic_list,
worker->nonbasic_list,
nonbasic_mark,
nonbasic_end,
Comment thread
chris-maes marked this conversation as resolved.
worker->basis_factors,
workspace,
delta_z,
Expand Down
4 changes: 2 additions & 2 deletions cpp/src/branch_and_bound/pseudo_costs.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -435,7 +435,7 @@ class pseudo_costs_t {
pseudo_cost_num_up(num_variables),
pseudo_cost_mutex_up(num_variables),
pseudo_cost_mutex_down(num_variables),
AT(1, 1, 1)
Arow(1, 1, 1)
{
}

Expand Down Expand Up @@ -526,7 +526,7 @@ class pseudo_costs_t {

reliability_branching_settings_t<i_t, f_t> reliability_branching_settings;

csc_matrix_t<i_t, f_t> AT; // Transpose of the constraint matrix A
csr_matrix_t<i_t, f_t> Arow;
Copy link
Copy Markdown
Contributor

@mlubin mlubin Apr 7, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it an invariant that all the nonbasic coefficients appear first, or only in some contexts? Consider documenting. I'm also wondering if there's scope for a helper struct to group Arow with nonbasic_end since nonbasic_end is a property of the stored representation of the matrix.

Copy link
Copy Markdown
Contributor Author

@chris-maes chris-maes Apr 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In order to use the compute_delta_z function the nonbasic coefficients must appear first. I can add some documentation and consider grouping with a struct.

std::vector<omp_atomic_t<f_t>> pseudo_cost_sum_up;
std::vector<omp_atomic_t<f_t>> pseudo_cost_sum_down;
std::vector<omp_atomic_t<i_t>> pseudo_cost_num_up;
Expand Down
Loading