v26.02.00

Release Notes 26.02

New Features (26.02)

• New parallel reliability branching inside MIP solver
• Mixed Integer Gomory, Mixed Integer Rounding, Knapsack and Strong Chvatal Gomory cuts are now added at root node
• Added an option to use batch PDLP when running strong branching at the root. Based on Batched First-Order Methods for Parallel LP Solving in MIP (Nicolas Blin, Stefano Gualandi, Christopher Maes, Andrea Lodi, Bartolomeo Stellato)
• Quadratic programming (QP) solver is now generally available (previously beta)
• New infeasibility detection option for PDLP's default solver mode Stable3
• Solutions callbacks added to C API. Users can now retrieve the dual bound and pass in user data
• Multiple new diving techniques added for finding integer feasible solutions
• The PSLP presolver is enabled by default for LP problems. Use the presolve option to select Papilo or disable
• Added a batch solve for routing to boost throughput for many similar instances
• Added experimental support for determinism in the parallel branch-and-bound solver. GPU heuristics are not supported yet in this mode

Breaking Changes (26.02)

• The signatures of the solution callbacks have changed for the Python API
• To use PDLP warm start, presolve must now be explicitly disabled by setting CUOPT_PRESOLVE=0. Previously, presolve was disabled automatically

Improvements (26.02)

• Improved primal/dual warm start for PDLP's default solver mode Stable3
• Quadratic objectives can now be constructed via a matrix in Python API
• QP barrier now updates and solves augmented system on the GPU
• Improved performance for LP folding
• Probing implications and better variable ordering to strengthen presolve and branching
• Replace deprecated cuDF Column/Buffer APIs with pylibcudf and public cuDF interfaces
• Modernize dependency pinnings; make CUDA runtime linkage static for portability
• Build/tooling: add --split-compile, --jobserver, Clang host build, ThreadSanitizer, improved container scripts, and branch/commit metadata in images
• Use explicit cudaStream_t with cub::DeviceTransform and non-blocking streams for GPU control
• Enable barrier LP tests, add regression testing, and add SonarQube static analysis
• Added parameter for specifying the random seed used by the solver

Bug Fixes (26.02)

• Fixed an issue with incorrect signs of dual variables and reduced costs on maximization problems
• Fix out-of-bounds in dense-column detection in barrier
• Correct infeasible-list handling to avoid incorrect infeasibility reports in dual simplex
• Fix race conditions found via Clang host build + ThreadSanitizer
• Resolve CUDA–Numba version mismatches with cuDF
• Fix device code to include required trailing return types
• Fix issue in crossover after dualization in barrier
• Repair container build and test failures
• Miscellaneous additional fixes and stability improvements

Documentation (26.02)

• Update README and top-level docs for current build and usage
• Document new repository branching strategies and release-cycle details in README and CONTRIBUTING
• Add best practices for batch solving

[NIGHTLY] v25.12.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• Disable flaky barrier tests (#520) @rgsl888prabhu
• Fix TSP when order locations are set (#503) @hlinsen
• Fix nightly container testing suite (#495) @rgsl888prabhu

📖 Documentation

• Add release notes to main repository for quick view (#512) @rgsl888prabhu

🛠️ Improvements

• Refactor build scripts to use RAPIDS_BRANCH and remove RAPIDS_VERSION (#517) @bdice
• Add code-rabbit config file to the repo (#516) @rgsl888prabhu
• Reduce package sizes (#507) @rgsl888prabhu
• Branch 25.12 merge branch 25.10 (#505) @rgsl888prabhu
• Branch 25.12 merge branch 25.10 (#490) @rgsl888prabhu
• Forward-merge branch-25.10 into branch-25.12 (#476) @rapids-bot[bot]

[NIGHTLY] v25.10.00

🔗 Links

• Development Branch
• Compare with main branch

🚨 Breaking Changes

• cuPDLPx (#388) @Kh4ster

🐛 Bug Fixes

• Hot fix 25.10 (#552) @hlinsen
• Fix handling zero sized array in case of presolve (#491) @rgsl888prabhu
• Fix issue with barrier incorrectly returning numerical in concurrent (#484) @chris-maes
• Gracefully handle OOM in Barrier (#483) @hlinsen
• Don't access free variables in the original problem on the folded problem (#477) @chris-maes
• [BUG] Fixed starting variable bounds for diving (#474) @nguidotti
• Fix an UMA on ns1952667 due to ignoring rank in factorize_basis when no singletons present (#470) @chris-maes
• Further CPUFJ assert and numerical fixes (#466) @aliceb-nv
• Disable TBB because of a bug in CliqueMerging presolver (#463) @rg20
• Barrier post-solve, find_dense_columns fix, folding fix (#462) @chris-maes
• Don't eliminate free variables in dual simplex (#457) @chris-maes
• More numerically robust CPUFJ feasibility checking (#454) @aliceb-nv
• Explicitely create new stream for Barrier concurrent (#453) @hlinsen
• Fixed delayed termination after timeout (#451) @nguidotti
• Fix compile and log (#450) @akifcorduk
• Short run fine tuning (#444) @aliceb-nv
• GF2 presolve infeasibility fix + CI occasional failure fix (#437) @aliceb-nv
• Fix variables out of bounds caused by CPUFJ LP scratch thread (#425) @aliceb-nv
• Fixed the maybe-uninitialized compilation error (#424) @nguidotti
• Fix linking errors in the test suite when disabling C adaptor (#420) @nguidotti
• Compute relative gap with respect to user objectives (#418) @rg20
• add http timeout values for general, send, and receive to client (#404) @tmckayus
• Fix bug in fixed_problem_computation (#403) @akifcorduk
• Remove limiting_resource_adaptor leftover (#398) @aliceb-nv
• Add support for cuda13 container and fix cuda13 lib issues in wheel (#379) @rgsl888prabhu
• Return Infeasible if the user problem contains crossing bounds (#361) @aliceb-nv
• Fix out-of-bound access in clean_up_infeasibilities. (#346) @legrosbuffle
• Skip downloading netlib data (#335) @rgsl888prabhu
• Don't remove empty columns with infinite bounds (#287) @chris-maes

📖 Documentation

• Update introduction to use engine instead of library (#536) @rgsl888prabhu
• Add default accuracy for each method in LP (#514) @rgsl888prabhu
• Release notes for v25.10 (#432) @cwilkinson76
• Add tutorial video links to docs (#416) @rgsl888prabhu
• Add warmstart, model update, update docs (#411) @Iroy30
• add docs on CI workflow inputs (#407) @jameslamb
• Add name to drop down for video link (#396) @rgsl888prabhu
• Add video link to the docs and Readme (#393) @rgsl888prabhu
• Add documentation on nightly installation commands (#367) @rgsl888prabhu
• Fix version in version tab, change log and fix typos (#267) @rgsl888prabhu

🚀 New Features

• Add optional timestamps enabled by env CUOPT_EXTRA_TIMESTAMPS (#392) @tmckayus
• cuPDLPx (#388) @Kh4ster
• QPS extension for MPS (#352) @Franc-Z
• Barrier method for LPs with cuDSS sparse Cholesky / LDLT on the GPU (#272) @chris-maes
• Add root node presolve using Papilo (#234) @hlinsen

🛠️ Improvements

• Switch to using papilo develop branch (#493) @rg20
• Fix return on size mismatch in Barrier (#489) @hlinsen
• Print scaling information about the user model (#488) @chris-maes
• Enable TBB and use fixed commit version (#481) @akifcorduk
• Barrier log fixes (#478) @chris-maes
• Remove raft-dask dependency (#475) @rgsl888prabhu
• Add setting to change the dual initial point used by barrier (#468) @chris-maes
• Add docs and example tests to barrier (#449) @rgsl888prabhu
• Enable green context in Barrier (#448) @hlinsen
• Remove left over kernels (#430) @akifcorduk
• Fix issue stemming because Rocky 8.10 is EOL and replaced by 8.11 (#427) @rgsl888prabhu
• Combine conda install for local and channel based artifacts into one env create call (#422) @rgsl888prabhu
• Integrate CliqueMerging presolver and tune presolver settings (#415) @rg20
• Enable miplib datasets which was disabled and also update README (#414) @rgsl888prabhu
• cuopt-server: update dependencies (drop httpx, add psutil) (#413) @jameslamb
• Parallel Branch-and-Bound (#412) @nguidotti
• build pure-Python wheels without waiting for dependencies (#409) @jameslamb
• avoid triggering nightly tests until builds are complete (#408) @jameslamb
• GF(2) presolve reduction for enlight_hard.mps (#406) @aliceb-nv
• CPUFJ for local search + simple rounding (#405) @aliceb-nv
• Add jump testing to nightly testing (#394) @rgsl888prabhu
• [FIX] Do not run compression tests when compression is disabled. (#389) @legrosbuffle
• Add sanitizer build option (#385) @akifcorduk
• Doc update for container version update and add nvidia-cuda-runtime as dependency (#384) @rgsl888prabhu
• Heuristic Improvements: balance between generation and improvement heuristics (#382) @akifcorduk
• Loosen presolve tolerance and update timers to report cumulative presolve/solve time (#381) @hlinsen
• CUDA 13 support: follow-ups (#377) @jameslamb
• Warn in case a dependent library is not found in libcuopt load (#375) @rgsl888prabhu
• Combined variable bounds (#372) @kaatish
• Enable parallelism for root node presolve (#371) @hlinsen
• Print cuOpt version / machine info before solving (#370) @aliceb-nv
• Implement node presolve (#368) @rg20
• Build and test with CUDA 13.0.0 (#366) @jameslamb
• Add Commit Sha to container for reference (#362) @rgsl888prabhu
• use GCC 14, consolidate dependency groups, update pre-commit hooks (#358) @jameslamb
• Decompression for .mps.gz and .mps.bz2 files (#357) @ahehn-nv
• Enable singleton stuffing and use papilo default params (#354) @hlinsen
• [FIX] Fix high GPU memory usage (#351) @aliceb-nv
• Adding support nightly cuopt-examples notebook testing (#342) @rgsl888prabhu
• Reduce hard-coded version usage in repo (#337) @rgsl888prabhu
• Make infeasibility checks consistent between the main solver and presolver (#326) @rg20
• Add read/write MPS and relaxation to python API (#323) @Iroy30
• Add maximization support for root node presolve (#322) @hlinsen
• Container to work on all different users including root (#321) @rgsl888prabhu
• Changes to download lp and milp dataset and also disable cvxpy testing for 3.10 (#317) @rgsl888prabhu
• Faster engine compile time (#316) @Kh4ster
• Performance tweak for dual_simplex/right_looking_lu (#315) @ahehn-nv
• Fix pre-commit for trailing whitespace and end of file (#306) @rgsl888prabhu
• Simple diving for Branch-and-Bound (#305) @nguidotti
• Bump libcuopt size to 775MiB (#303) @rgsl888prabhu
• Merge update version and fix version format bugs (#286) @rgsl888prabhu
• Support cuda 12.9 (#269) @rgsl888prabhu
• Adding actions to automate auto issue assignment and reminder on inactivity (#268) @rgsl888prabhu
• Sub-MIP recombiner and B&B global variable changes (#259) @akifcorduk
• FP as a local search (#258) @akifcorduk
• Update rapids-build-backend to 0.4.1 (#253) @KyleFromNVIDIA
• Branch 25.10 merge branch 25.08 (#251) @rgsl888prabhu

[NIGHTLY] v25.08.00

🔗 Links

• Development Branch
• Compare with main branch

🚨 Breaking Changes

• Fix versioning for nightly and release package (#175) @rgsl888prabhu
• Bump cuOpt and Rapids version (#55) @rgsl888prabhu

🐛 Bug Fixes

• Fix a segfault on bnatt500 due to small mu leading to inf/nan (#254) @chris-maes
• Fix a bug in basis repair. Recover from numerical issues in primal update. (#249) @chris-maes
• Unset NDEBUG in cmake in assert mode (#248) @hlinsen
• Manual cuda graph creation in load balanced bounds presolve (#242) @kaatish
• Fix assert failure on MIP problem reducing to LP after presolve (#235) @aliceb-nv
• align service api with SDK (compat with deprecations) (#227) @tmckayus
• Fix bug on initial solution size in the check and cuda set device order (#226) @akifcorduk
• No cuda graph in batched pdlp (#225) @Kh4ster
• Disable feasibility jump test for one of the dataset (#215) @rgsl888prabhu
• Fix logging levels (#201) @akifcorduk
• Fix bug in scaling of dual slacks and sign of dual variables for >= constraints (#191) @chris-maes
• Fix inversion crossover bug with pdp and prize collection (#179) @hlinsen
• Fix a bug in extract_best_per_route kernel (#156) @rg20
• Fix several bugs appeared in unit testing of JuMP interface (#149) @rg20
• Fix incorrect reported solving time (#131) @aliceb-nv
• Fix max offset (#113) @Kh4ster
• Fix batch graph capture issue caused by pinned memory allocator (#110) @Kh4ster
• Fix bug in optimization_problem_solution_t::copy_from (#109) @rg20
• Fix issue when problem has an empty problem in pdlp (#107) @Kh4ster
• Fix crash on models with variables but no constraints (#105) @aliceb-nv
• Fix inversion of constraint bounds in conditional bounds presolve (#75) @kaatish
• Fix data initialization in create depot node for max travel time feature (#74) @hlinsen

📖 Documentation

• Add link checker for doc build and test (#229) @rgsl888prabhu
• Adding few more pre-commit checks to ensure coding standards (#213) @rgsl888prabhu
• Mention GAMS and GAMSPy in third-party modeling languages page in documentation (#206) @0x17
• Enable doc build workflow and build script for PR and Nightly (#203) @rgsl888prabhu
• Fix the link to Python docs in README (#118) @Abinashbunty
• Update PR template to be concise with checklist (#73) @rgsl888prabhu

🚀 New Features

• Add python API (#223) @Iroy30
• add a simple smoketest for cvxpy integration (#199) @tmckayus

🛠️ Improvements

• Bump actions/checkout in nightly.yaml to v4 (#230) @ScottBrenner
• Add helm chart for cuopt service (#224) @rgsl888prabhu
• Remove cuda 11 specific changes from repo (#222) @rgsl888prabhu
• Heuristic Improvements with solution hash, MAB and simplex root solution (#216) @akifcorduk
• Add license headers to missing files (#210) @rgsl888prabhu
• Fix: Prevent Github language statistics from mistaking MPS files as JetBrains related (#202) @aliceb-nv
• Fix: various typos in comments and strings, note on result dir (#200) @MohdAatifSiddi
• Performance improvements to dual simplex: BFRT, hypersparsity, basis updates, primal infeasible list (#192) @chris-maes
• Feat: rework cmakelists.txt to facilitate deb file creation. (#190) @mostroot
• Add nightly container support (#180) @rgsl888prabhu
• New heuristic improvements (#178) @akifcorduk
• Fix versioning for nightly and release package (#175) @rgsl888prabhu
• Split very large tests into smaller individual test cases (#152) @vitor1001
• Fix compile error when using clang with C++20. (#145) @legrosbuffle
• Small fixes for corner cases: (#130) @vitor1001
• Small improvements on how paths are handled in tests (#129) @vitor1001
• Update cxxopts to v3.3.1 (#128) @vitor1001
• refactor(rattler): remove cuda11 options and general cleanup (#127) @gforsyth
• use 25.08 versions of RAPIDS CI workflows (#126) @jameslamb
• Relax pinnings on several dependencies, remove nvidia channel (#125) @bdice
• Several fixes needed to compile cuOpt with LLVM (#121) @vitor1001
• Use cusparsespmv_preprocess() now that Raft implements it. (#120) @vitor1001
• Fix compile error when building with clang. (#119) @legrosbuffle
• cuopt service add healthcheck for / (#114) @tmckayus
• refactor(shellcheck): fix all remaining shellcheck errors/warnings (#99) @gforsyth
• Branch 25.08 merge branch 25.05 (#95) @rgsl888prabhu
• Create a bash script to run mps files in parallel (#87) @rg20
• Branch 25.08 merge 25.05 (#79) @rgsl888prabhu
• Branch 25.08 merge 25.05 (#60) @rgsl888prabhu
• Add CTK 12.9 fatbin flags to maintain existing binary sizes (#58) @robertmaynard
• expand RAPIDS build-infra CODEOWNERS assignments (#57) @jameslamb
• Bump cuOpt and Rapids version (#55) @rgsl888prabhu
• stop uploading to downloads.rapids.ai, use rapids-init-pip in wheel CI (#50) @jameslamb
• Branch 25.08 merge 25.05 (#43) @rgsl888prabhu
• Branch 25.08 merge 25.05 (#35) @rgsl888prabhu

[NIGHTLY] v25.05.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• Fix issue with reported solving time on time-limit termination (#93) @aliceb-nv
• Fix to decouple thin client requirements from cuopt main module (#92) @rgsl888prabhu
• Fix occasional incorrect solution bound on maximization problems (#83) @rgsl888prabhu
• Fix inversion of constraint bounds in conditional bounds presolve (#82) @kaatish
• Fix nightly build which is failing on jq request (#78) @rgsl888prabhu
• Pdlp fix batch cuda graph (#68) @Kh4ster
• allow long client version in service version check' (#61) @tmckayus
• Fix obj constant on max. Fix undefined memory access at root (#52) @chris-maes
• include Python constraints for pure-Python conda packages (#51) @jameslamb
• Disable failing FJ test on ARM CI (#39) @aliceb-nv
• Fix bug on probing cache usage of lb bounds prop and probing cache bounds lookup with wrong variable id (#36) @akifcorduk
• Fix graph bug seen when using batch (#34) @Kh4ster
• Fix bug: dual simplex does not fill in additional info (#33) @chris-maes
• Switch to using v3.2 of argparser for wheels (#31) @rg20
• Bug fix: solution incorrect when simplex presolve removes empty columns (#29) @chris-maes
• Bug fixes on preprocess, line segment returning worse solution, temporary fix add B&B solution directly (#27) @akifcorduk
• Fix non-trivial presolve not reporting Optimality (#26) @aliceb-nv
• Fix bug in C API for bool parameters. Fix bug on irish-electricty (#25) @chris-maes
• Fix bug on consecutive runs caused by static lp_state_t object (#24) @aliceb-nv
• Fix bug in crossover where numerical was returned instead of time limit (#23) @chris-maes
• Handle out of memory and out of time limit issues gracefully (#22) @rg20
• PDLP solution file generation and binary fix (#21) @Kh4ster
• Fix objective value when LP is unbounded (#19) @chris-maes
• Fix bugs on empty problem and pre_process assignment (#18) @akifcorduk
• Fix LP termination status 9 errors in batch LP solves (#17) @chris-maes
• Fix branch build and nightly (#12) @rgsl888prabhu

📖 Documentation

• Fix test commands in CONTRIBUTING.md (#88) @rgsl888prabhu
• Update README.md (#85) @bbozkaya
• Documentation update to add third party modeling languages and nightly documentation (#77) @rgsl888prabhu
• Update Source build instructions (#62) @rgsl888prabhu
• Update README and Docs for version and broken links (#56) @rgsl888prabhu
• Fix documentation for missing links and typos (#48) @rgsl888prabhu
• Address cuOpt doc reviews (#41) @rgsl888prabhu
• Fix documentation structure and fine tune explanations (#16) @rgsl888prabhu

🛠️ Improvements

• Migrate to modern logger interface (#80) @emmanuel-ferdman
• Enable nightly package publishing and fix nightly testing workflow (#72) @rgsl888prabhu
• Write out mps file containing user problem. Useful for debugging (#66) @chris-maes
• Update README.md (#49) @chris-maes
• Fix build.yaml to build for python versions apart from 3.13 (#45) @rgsl888prabhu
• Fix branch and nightly build for libcuopt wheel to build on only python3.12 (#38) @rgsl888prabhu
• Deprecate and remove cuda 11 support (#32) @rgsl888prabhu
• Remove unused and broken build flags (#15) @rg20
• Internal PR sync (#14) @rgsl888prabhu
• Fixing FJ bugs, Termination status and MIP GAP parameters (#6) @rgsl888prabhu
• Update Code owners file (#5) @rgsl888prabhu
• Adding C API support for LP and MILP (#3) @rgsl888prabhu