Add support for quadratic programming to the barrier method #500

chris-maes · 2025-10-14T18:58:33Z

Add support for quadratic programming to the barrier method.

Issue

Checklist

I am familiar with the Contributing Guidelines.
Testing
- New or existing tests cover these changes
- Added tests
- Created an issue to follow-up
- NA
Documentation
- The documentation is up to date with these changes
- Added new documentation
- NA

copy-pr-bot · 2025-10-14T18:58:37Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

greptile-apps

Greptile Overview

Greptile Summary

This PR extends cuOPT's barrier method solver to support quadratic programming (QP) problems alongside the existing linear programming (LP) functionality. The implementation adds storage and handling for a quadratic objective matrix Q throughout the optimization pipeline. The changes introduce three new vector fields (Q_offsets, Q_indices, Q_values) to store the quadratic matrix in compressed sparse format, following the same pattern used for constraint matrices. The quadratic objective function takes the form (1/2)x^T Q x + c^T x, enabling the solver to handle more complex optimization problems like portfolio optimization and machine learning applications.

The implementation modifies the barrier method's interior point algorithm to properly handle quadratic terms in the KKT system. Key architectural changes include updating the augmented system formation to include Q matrix entries, modifying dual residual computations to incorporate Q*x terms, and ensuring step size consistency for quadratic problems. The changes maintain backward compatibility by treating empty Q matrices as linear problems.

Important Files Changed

Changed Files

Filename	Score	Overview
`cpp/src/dual_simplex/barrier.cu`	4/5	Implements core quadratic programming support in barrier method with Q matrix handling, KKT system modifications, and residual computations
`cpp/include/cuopt/linear_programming/optimization_problem.hpp`	4/5	Adds API for quadratic objective matrix storage with setter and getter methods
`cpp/src/linear_programming/optimization_problem.cu`	4/5	Implements quadratic matrix storage methods and debug output functionality
`cpp/src/dual_simplex/presolve.cpp`	4/5	Adds quadratic matrix initialization and disables incompatible optimizations for QP problems
`cpp/src/mip/problem/problem.cu`	4/5	Integrates quadratic objective components into MIP problem constructors
`cpp/src/linear_programming/solve.cu`	4/5	Connects MPS quadratic data parsing to optimization problem structure
`cpp/src/mip/problem/problem.cuh`	4/5	Declares quadratic matrix storage vectors in problem class header
`cpp/src/dual_simplex/presolve.hpp`	4/5	Adds Q matrix field to lp_problem_t structure for quadratic programming
`cpp/src/dual_simplex/user_problem.hpp`	5/5	Adds quadratic matrix storage vectors to user problem structure
`cpp/src/linear_programming/translate.hpp`	4/5	Transfers quadratic matrix data between problem structures with debug output
`cpp/libmps_parser/src/mps_parser.cpp`	4/5	Adds debug output for quadratic objective matrix parsing
`cpp/src/dual_simplex/scaling.cpp`	3/5	Disables column scaling for quadratic problems but contains syntax error in logical operator

Confidence score: 4/5

This PR appears safe to merge with moderate risk due to the comprehensive nature of the changes and good architectural integration
Score reflects the extensive modifications across multiple solver components and the presence of debug code/syntax issues that should be addressed
Pay close attention to cpp/src/dual_simplex/scaling.cpp which has a logical operator syntax error, and cpp/src/dual_simplex/barrier.cu due to its complexity and central role in QP functionality

Sequence Diagram

sequenceDiagram
    participant User
    participant MPS_Parser as mps_parser_t
    participant Problem as optimization_problem_t
    participant Barrier as barrier_solver_t

    User->>MPS_Parser: Parse MPS file with QUADOBJ section
    MPS_Parser->>MPS_Parser: parse_quadobj(line)
    Note over MPS_Parser: Parse quadratic objective entries<br/>(upper triangular format)
    MPS_Parser->>MPS_Parser: Store quadobj_entries as tuples<br/>(var1_id, var2_id, value)
    
    MPS_Parser->>MPS_Parser: fill_problem(data_model)
    Note over MPS_Parser: Convert to CSR format using<br/>double transpose method
    MPS_Parser->>MPS_Parser: build_csr_via_transpose(quadobj_entries)
    Note over MPS_Parser: Expand upper triangular to<br/>full symmetric matrix
    MPS_Parser->>Problem: set_quadratic_objective_matrix(Q_values, Q_indices, Q_offsets)
    Problem->>Problem: Store Q_offsets_, Q_indices_, Q_values_
    
    User->>Problem: solve_lp()
    Problem->>Problem: cuopt_problem_to_simplex_problem()
    Note over Problem: Convert problem structure<br/>for dual simplex format
    Problem->>Problem: Copy Q data to user_problem
    Note over Problem: user_problem.Q_offsets = Q_offsets<br/>user_problem.Q_indices = Q_indices<br/>user_problem.Q_values = Q_values
    
    Problem->>Barrier: solve_linear_program_with_barrier()
    Barrier->>Barrier: Create lp_problem_t with Q matrix
    Note over Barrier: Q.m = lp.num_cols<br/>Q.n = lp.num_cols<br/>Convert row format to column format
    Barrier->>Barrier: form_augmented(Q matrix)
    Note over Barrier: Build augmented system with<br/>quadratic terms: -Q - diag - dual_perturb
    Barrier->>Barrier: Initialize barrier method
    Note over Barrier: Handle quadratic programming<br/>with interior point method
    Barrier-->>Problem: Return solution
    Problem-->>User: Return optimized solution

_{12 files reviewed, 17 comments}

_{Edit Code Review Agent Settings | Greptile}

cpp/src/dual_simplex/scaling.cpp

greptile-apps · 2025-10-14T18:59:52Z

cpp/src/linear_programming/translate.hpp

+  printf("user_problem.Q_offsets: %zu\n", user_problem.Q_offsets.size());
+  printf("user_problem.Q_indices: %zu\n", user_problem.Q_indices.size());
+  printf("user_problem.Q_values: %zu\n", user_problem.Q_values.size());


style: Debug printf statements should be removed before merging - use logging instead

Suggested change

printf("user_problem.Q_offsets: %zu\n", user_problem.Q_offsets.size());

printf("user_problem.Q_indices: %zu\n", user_problem.Q_indices.size());

printf("user_problem.Q_values: %zu\n", user_problem.Q_values.size());

CUOPT_LOG_DEBUG("user_problem.Q_offsets: {}", user_problem.Q_offsets.size());

CUOPT_LOG_DEBUG("user_problem.Q_indices: {}", user_problem.Q_indices.size());

CUOPT_LOG_DEBUG("user_problem.Q_values: {}", user_problem.Q_values.size());

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/src/linear_programming/translate.hpp Line: 112:114 Comment: **style:** Debug printf statements should be removed before merging - use logging instead ```suggestion CUOPT_LOG_DEBUG("user_problem.Q_offsets: {}", user_problem.Q_offsets.size()); CUOPT_LOG_DEBUG("user_problem.Q_indices: {}", user_problem.Q_indices.size()); CUOPT_LOG_DEBUG("user_problem.Q_values: {}", user_problem.Q_values.size()); ``` How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2025-10-14T18:59:53Z

cpp/libmps_parser/src/mps_parser.cpp


  // Process QUADOBJ data if present (upper triangular format)
  if (!quadobj_entries.empty()) {
+    printf("Quadratic objective matrix: %zu nonzeros\n", quadobj_entries.size());


style: Consider using proper logging instead of printf for consistency with the rest of the codebase

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/libmps_parser/src/mps_parser.cpp Line: 509:509 Comment: **style:** Consider using proper logging instead of `printf` for consistency with the rest of the codebase How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2025-10-14T18:59:54Z

cpp/include/cuopt/linear_programming/optimization_problem.hpp

+  /** Quadratic objective matrix in CSR format (for (1/2) * x^T * Q * x term) */
+  std::vector<i_t> Q_offsets_;
+  std::vector<i_t> Q_indices_;
+  std::vector<f_t> Q_values_;


style: Quadratic matrix stored as host std::vectors while other matrices use rmm::device_uvector. This creates inconsistency in memory management - consider if device storage is needed for GPU operations.

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/include/cuopt/linear_programming/optimization_problem.hpp Line: 435:438 Comment: **style:** Quadratic matrix stored as host std::vectors while other matrices use rmm::device_uvector. This creates inconsistency in memory management - consider if device storage is needed for GPU operations. How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2025-10-14T18:59:54Z

cpp/src/linear_programming/optimization_problem.cu

+                                                                      const f_t* Q_values,
+                                                                      i_t size_values)
+{
+  printf("optimization_problem_t Quadratic objective matrix: %d nonzeros\n", size_values);


style: Debug printf should be removed or replaced with proper logging using CUOPT_LOG_DEBUG or similar logging infrastructure

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/src/linear_programming/optimization_problem.cu Line: 154:154 Comment: **style:** Debug printf should be removed or replaced with proper logging using CUOPT_LOG_DEBUG or similar logging infrastructure How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2025-10-14T19:00:00Z

cpp/src/dual_simplex/barrier.cu

    i_t factorization_size   = n + m;
    const f_t dual_perturb   = 0.0;
-    const f_t primal_perturb = 1e-6;
+    const f_t primal_perturb = 1e-12;


style: Consider using a more robust perturbation value than 1e-12 for the primal perturbation, especially for ill-conditioned problems

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/src/dual_simplex/barrier.cu Line: 366:366 Comment: **style:** Consider using a more robust perturbation value than 1e-12 for the primal perturbation, especially for ill-conditioned problems How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2025-10-14T19:00:01Z

cpp/src/dual_simplex/barrier.cu

+        settings_.log.printf("augmented nnz %d predicted %d\n", q, 2 * nnzA + n + m + off_diag_Qnz);
+        exit(1);
+      }
+      if (A.col_start[n] != AT.col_start[m]) {
+        settings_.log.printf("A nz %d AT nz %d\n", A.col_start[n], AT.col_start[m]);
+        exit(1);


style: Using exit(1) for assertion failures can be problematic in library code - consider returning error codes or throwing exceptions instead

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/src/dual_simplex/barrier.cu Line: 417:422 Comment: **style:** Using `exit(1)` for assertion failures can be problematic in library code - consider returning error codes or throwing exceptions instead How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2025-10-14T19:00:02Z

cpp/src/dual_simplex/barrier.cu

+    if (data.Q.n > 0) {
+      matrix_vector_multiply(data.Q, -1.0, data.x, 1.0, c);
+    }


style: The modification of vector c here creates a temporary that includes Q*x. Consider documenting this side effect or using a separate variable name for clarity

Prompt To Fix With AI

This is a comment left during a code review. Path: cpp/src/dual_simplex/barrier.cu Line: 1708:1710 Comment: **style:** The modification of vector `c` here creates a temporary that includes Q*x. Consider documenting this side effect or using a separate variable name for clarity How can I resolve this? If you propose a fix, please make it concise.

cpp/src/dual_simplex/barrier.cu

…columns

coderabbitai · 2025-10-16T14:22:40Z

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Note

Other AI code review bot(s) detected

CodeRabbit has detected other AI code review bot(s) in this pull request and will avoid duplicating their findings in the review comments. This may lead to a less comprehensive review.

✨ Finishing touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

First stab at QP

886f0a6

chris-maes requested a review from a team as a code owner October 14, 2025 18:58

chris-maes requested review from akifcorduk and kaatish October 14, 2025 18:58

greptile-apps bot reviewed Oct 14, 2025

View reviewed changes

rg20 marked this pull request as draft October 14, 2025 19:00

chris-maes added feature request New feature or request non-breaking Introduces a non-breaking change labels Oct 14, 2025

chris-maes added this to the 25.12 milestone Oct 14, 2025

chris-maes added 3 commits October 14, 2025 12:03

Add space in scaling

a624e23

Style fixes

b34300a

Merge remote-tracking branch 'cuopt-nvidia/branch-25.12' into qp

e3ef96b

chris-maes changed the title ~~[Skip-CI] Add support for quadratic programming to the barrier method~~ Add support for quadratic programming to the barrier method Oct 14, 2025

chris-maes changed the base branch from branch-25.10 to branch-25.12 October 14, 2025 21:33

rgsl888prabhu and others added 5 commits October 14, 2025 15:03

add header

84eedbf

Fix crashes due to mismatch between Q and A when presolve eliminated …

29509f9

…columns

Style fixes

3f08031

Handle SIF files in the script

9f04972

Fix a bug in computing quadratic objective

5a0f39f

rg20 added 8 commits October 16, 2025 07:28

handle presolve flag correctly

ef5b72a

Resize Q matrix to include slack variables

d7069c7

Remove stale cout

781b4e4

Fix indexing issue

d1824bc

Fix warning

4176cf6

Fix out of bounds indexing

203212b

Add debug function for cudss factorization

62b20c3

Merge remote-tracking branch 'upstream/main' into qp

f8631ac

Uh oh!

Add support for quadratic programming to the barrier method #500

Are you sure you want to change the base?

Add support for quadratic programming to the barrier method #500

Conversation

chris-maes commented Oct 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Issue

Checklist

Uh oh!

copy-pr-bot bot commented Oct 14, 2025

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Greptile Overview

Greptile Summary

Important Files Changed

Confidence score: 4/5

Sequence Diagram

Uh oh!

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Oct 14, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

coderabbitai bot commented Oct 16, 2025

Review skipped

Other AI code review bot(s) detected

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

chris-maes commented Oct 14, 2025 •

edited

Loading