Release 0.9.2

[michel2323]

I had to bump MadNLPGPU from 0.8.0 to 0.9.0 because of a CUDA dependency change 9 (-> weak deps).

Release notes draft:

MadNLP v0.9.2

Highlights: new SchurComplementKKTSystem for two-stage stochastic programs (CPU + GPU), an API cleanup so solver and wrapper internals are accessed through getter functions, and a MUMPS 5.8.x fix.

New features

• Add SchurComplementKKTSystem for exploiting the block-arrowhead structure of two-stage stochastic programs. Reduces each Newton step to a dense nd × nd design-variable system via per-scenario eliminations. CPU variant uses Umfpack for the per-scenario blocks by default; GPU variant lives in the MadNLPGPU CUDA extension and uses batched cuDSS + strided-batched CUBLAS. (Add SchurComplementKKTSystem for two-stage stochastic programs #598)

API changes

• Solver member access is now through getter functions instead of direct field access. Downstream code that reaches into MadNLPSolver fields will need to migrate. (Overhaul solver member access via functions #497)
• Wrapper models use getters for inner-model access. (Wrappers should use getters for nlp models meta #595)
• solve is now re-exported from SolverCore, aligning MadNLP with the broader JuliaSmoothOptimizers ecosystem. (Revert and use function solve from SolverCore #601)

Fixes

• Update MUMPS bindings for the 5.8.x ABI. ([Mumps] Fix the MUMPS structure for v5.8.x #613)

Documentation

• Fix mu_init in the warm-start tutorial. ([doc] Fix mu_init in warm-start tutorial #607)

---

MadNLPGPU v0.9.0

Highlights: CUDA/CUDSS and AMDGPU are now weak dependencies, and the new Schur-complement KKT system has a GPU backend.

• CUDA/CUDSS moved to weak deps, GPU backend code deduplicated. Installing MadNLPGPU no longer pulls CUDA on machines without it; the CUDA-specific code (LapackCUDASolver, CUDSSSolver, GPU Schur) lives in the MadNLPGPUCUDAExt extension and loads only when both CUDA.jl and CUDSS.jl are present. The AMDGPU backend follows the same pattern via MadNLPGPUAMDGPUExt. (Move CUDA/CUDSS to weak deps and deduplicate GPU backend code #596)
• GPU SchurComplementKKTSystem: batched per-scenario factorizations via cuDSS, Schur accumulation via CUBLAS strided-batched GEMM. (Add SchurComplementKKTSystem for two-stage stochastic programs #598)
• Requires MadNLP ≥ 0.9.2.

Known issue — V100 / Volta (sm_70):
CUDA Toolkit 13 dropped support for Volta GPUs. Starting at CUDA.jl v5.10 (March 2026), Pkg may resolve to a CUDA.jl that auto-selects the CUDA 13 runtime, which then fails on V100. MadNLPGPU's compat allows this. Until JuliaGPU/CUDA.jl#3134 ships in a tagged release, V100 users should pin to a CUDA 12.x toolkit by running once per environment:

using CUDA
CUDA.set_runtime_version!(v"12.9")    # any 12.x works

Alternatively, pin CUDA = "~5.9" in your project's Project.toml.

---

MadNLPTests v0.6.1

• Add Schur-complement two-stage stochastic test fixtures used by MadNLP and MadNLPGPU Schur tests. (Add SchurComplementKKTSystem for two-stage stochastic programs #598)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.63%. Comparing base (f99a4ad) to head (5298883).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #616   +/-   ##
=======================================
  Coverage   85.63%   85.63%           
=======================================
  Files          52       52           
  Lines        4880     4881    +1     
=======================================
+ Hits         4179     4180    +1     
  Misses        701      701           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[KSepetanc]

Can you check, something is wrong for MadNLPGPU. Installation fails on clean 1.12 env (simple add MadNLPGPU#master). Is it related to changes here?

Failed to precompile MadNLPGPU [d72a61cc-809d-412f-99be-fd81f4b8a598] to "C:\\Users\\karlo\\.julia\\compiled\\v1.12\\MadNLPGPU\\jl_AC1F.tmp".
WARNING: Imported binding MadNLP.populate_RR_nn! was undeclared at import time during import to MadNLPGPU.
WARNING: Imported binding MadNLP.count_lu_bounds was undeclared at import time during import to MadNLPGPU.
ERROR: LoadError: invalid method definition in MadNLPGPU: exported function MadNLP.count_lu_bounds does not exist

Edit: MadNLPGPU 0.9 was not in registry and 0.10 produces the above error.

[KSepetanc]

I see releases go in tandem thus above error.

Please put MadNLPGPU 0.9 in registry (bot did not do it).