v0.8.5

MadNLP v0.8.5

Diff since v0.8.4

Merged pull requests:

• [MOI] update signatures of callbacks in MOI wrapper (#351) (@frapac)
• [KKT] Add K2.5 formulation for augmented KKT system (#352) (@frapac)
• Remove benchmark scripts (#359) (@frapac)
• Make library relocatable (#362) (@tmmsartor)
• [MOI] Do not overwrite MadNLP's options in wrapper (#367) (@frapac)
• [FIX] fix missing check in get_status_output (#369) (@frapac)
• CI path update (#370) (@sshin23)
• [DOC] Add tutorial on multi-precision (#371) (@frapac)
• [DOC] Add tutorial on warm-start (#372) (@frapac)
• MadNLPGraph migration (#374) (@sshin23)
• fix bug in MOI impl when quadratic term has param (#379) (@guimarqu)
• Update the link for libHSL (#383) (@amontoison)
• [documentation] Replace the old link of Julia-HSL (#384) (@amontoison)
• CompatHelper: bump compat for CUDSS to 0.4 for package MadNLPGPU, (keep existing compat) (#385) (@github-actions[bot])
• [MadNLPGPU] Release 0.7.4 (#386) (@amontoison)
• Remove some deps in the Project.toml of MadNLPGPU (#391) (@amontoison)
• Fix warning related to memory in MadNLPGPU (#392) (@amontoison)
• Iterative refinement tolerance tuning (#393) (@sshin23)
• Fix MadNLPGPU on CUDA.jl 5.6.0 (#395) (@frapac)
• [CI] Test MadNLPGPU on Julia 1.11 (#396) (@amontoison)
• [MadNLP] Release 0.8.5 (#397) (@amontoison)

Closed issues:

• Problems with relocatability (#250)
• Project.toml: No such file (#375)
• The function _set_con_scale_sparse! is not GPU-friendly (#377)
• No method matching _eval_lag_hess_wrapper! (#378)
• Why one can get right result on A6000, but problem can't converge on H100 in FP64? (#382)
• Why FP32 is slower than FP64 ? (#387)
• Error with CUDA.jl v5.6.0 (#394)

v0.8.4

MadNLP v0.8.4

Diff since v0.8.3

Merged pull requests:

• [MadNLPGPU] Upgrade CUDSS -- support iterative refinement and hybrid mode (#329) (@amontoison)
• CompatHelper: bump compat for NLPModels to 0.21, (keep existing compat) (#331) (@github-actions[bot])
• [MadNLPGPU] Refactoring - Part I (#340) (@frapac)
• [KKT] Fix symmetric K3 formulation (#345) (@amontoison)
• [KKT] add two folders Sparse and Dense for KKT formulations (#347) (@frapac)
• Update kkt.md (#354) (@amontoison)
• bump MadNLP version to v0.8.4 (#358) (@frapac)

v0.8.3

MadNLP v0.8.3

Diff since v0.8.2

Merged pull requests:

• [hotfix] LDLFactorizations.jl added as dependency (#343) (@sshin23)
• bump MadNLP version to v0.8.3 (#344) (@sshin23)

Closed issues:

• Move LDLSolver to weakdeps (#342)

v0.8.2

MadNLP v0.8.2

Diff since v0.8.1

Merged pull requests:

• Add an ordering for cuDSS (#317) (@amontoison)
• [LinearSolvers] Add support for LDL factorization in CHOLMOD (#321) (@frapac)
• [MOI] Add support for nonlinear problems without Hessian (#322) (@frapac)
• Reexport MadNLP from MadNLP/libs (#325) (@sshin23)
• [MadNLPGPU] Bug fix for empty Hessian (#326) (@sshin23)
• MadNLPMumps StaticArray dependency removed (#328) (@sshin23)
• [MOI] fix eval_constraint_jacobian_product (#337) (@frapac)
• Use GH Action Julia cache (#339) (@michel2323)
• bump MadNLP version to 0.8.2 (#341) (@sshin23)

Closed issues:

• JuMP interface with register: AssertionError: d.want_hess error (#318)
• Running into a weird issue, with the getting started example (#323)
• CUDSS internal error (invalid argument: inertia correction) (#333)
• CuCholeskySolver GPU memory leak (#334)
• User-defined operator JuMP: UndefVarError: ∇f not defined (#336)

v0.8.1

MadNLP v0.8.1

Diff since v0.8.0

Merged pull requests:

• Update TagBot.yml (#308) (@sshin23)
• Update README.md (#309) (@sshin23)
• Add logos to documentation (#310) (@sshin23)
• CuDSS synchronize added (#314) (@sshin23)
• Introduce linear solvers with version info (#315) (@sshin23)
• Fix optional arguments in get_index_constraints (#316) (@frapac)
• Bump MadNLP version to v0.8.1 (#319) (@sshin23)

Closed issues:

• Display the version number of the linear solver in the logs (#313)

v0.7.0

MadNLP v0.7.0

Diff since v0.6.0

Closed issues:

• LBFGS approximation of the inverse Hessian (#39)
• Unable to install HSL extension in Windows (#217)

Merged pull requests:

• CompatHelper: bump compat for NLPModels to 0.20, (keep existing compat) (#252) (@github-actions[bot])
• CompatHelper: bump compat for NLPModels to 0.20 for package MadNLPTests, (keep existing compat) (#253) (@github-actions[bot])
• fix version of MadNLPGPU in Project.toml (#255) (@frapac)
• [MadNLPGPU] Migrate to KernelAbstractions 0.9 (#258) (@frapac)
• [MadNLPTests] Remove ADNLPModels in deps (#259) (@frapac)
• Update README.md for JuMP documentation (#260) (@odow)
• Update for julia v1.9 (#261) (@sshin23)
• Using HSL_jll instead of custom compile (#263) (@sshin23)
• bump MadNLP version to 0.7.0 (#266) (@sshin23)

v0.6.0

MadNLP v0.6.0

Diff since v0.5.2

Merged pull requests:

• Add dense BFGS and compact LBFGS algorithms (#221) (@frapac)
• CompatHelper: bump compat for ADNLPModels to 0.5 for package MadNLPTests, (keep existing compat) (#240) (@github-actions[bot])
• bump CUDA.jl to 4.0 (#248) (@frapac)
• bump MadNLP version to 0.6.0 (#251) (@frapac)

v0.5.2

MadNLP v0.5.2

Diff since v0.5.1

Closed issues:

• Invalid number in NLP Hessian Lagrangian detrected / error in MadNLP (#239)
• MadNLP stucks at an iteration (#242)

Merged pull requests:

• CompatHelper: bump compat for SolverCore to 0.3, (keep existing compat) (#222) (@github-actions[bot])
• Remove allocations in callbacks (#230) (@frapac)
• [TTFS] Replace broadcast operators by explicit for loops in IPM kernels (#231) (@frapac)
• Ma27 bug fix (#235) (@sshin23)
• Minor README link fix (#236) (@michel2323)
• Fix invalid return code in MOI.TerminationStatus (#241) (@odow)
• Check whether initial KKT is valid before initiating multipliers (#243) (@frapac)
• Fix infinite loop in feasibility restoration (#244) (@frapac)
• bump version to 0.5.2 (#246) (@frapac)

v0.5.1

MadNLP v0.5.1

Diff since v0.5.0

Closed issues:

• Number of upper and lower bounded vars are incorrectly reported (#209)
• Possibly wrong init of dual variables (#214)
• Test for solve! function (#218)
• Better error message when invalid number detected (#225)

Merged pull requests:

• Update quickstart.md (#208) (@baggepinnen)
• [doc] fix quickstart (#210) (@frapac)
• [hotfix] Fix detection of upper/lower bounds (#211) (@frapac)
• Fix solve! with y provided (#215) (@baggepinnen)
• fix changed names of meta fields (#216) (@baggepinnen)
• add proper tests for solve! function (#219) (@frapac)
• Make nlp a concretely typed field in MadNLPSolver (#220) (@baggepinnen)
• update MOI wrapper (#224) (@frapac)
• Improve error messages when invalid number is detected (#226) (@frapac)
• [TTFS] Improve type inference in MadNLP (#227) (@frapac)
• [TTFS] clean interior-point kernels to improve inference (#228) (@frapac)
• Add new constructor for custom KKT type (#232) (@frapac)
• Improve MOI interface (#233) (@sshin23)
• Bump version to 0.5.1 (#234) (@frapac)

v0.5.0

MadNLP v0.5.0

Diff since v0.4.2

Closed issues:

• Change MadNLP behavior to not delete the entries inside the options' Dict (#178)
• Implement a proper interface for MadNLPGPU (#179)
• tril_to_full! function in QR and LU algorithms error on GPU (#181)
• Add option for specifying the matrix sparsity structure (#183)
• Change function names (#192)

Merged pull requests:

• [MadNLPHSL] build improvment (#170) (@sshin23)
• Remove StrideOneVector alias (#174) (@frapac)
• Define linear solver as a struct instead of as a module (#175) (@frapac)
• add HS15Model in MadNLPTests (#184) (@frapac)
• add CuInteriorPointSolver constructor (#185) (@frapac)
• Add support for Float32 (#187) (@sshin23)
• Add support for NLSModels (#190) (@sshin23)
• [Breaking change] Refactor the way we pass options to MadNLP (#191) (@frapac)
• Improve code coverage (#193) (@frapac)
• [Breaking change] Change names to follow JSO conventions (#194) (@sshin23)
• HSL build hotfix (#196) (@sshin23)
• CompatHelper: bump compat for ADNLPModels to 0.4 for package MadNLPTests, (keep existing compat) (#197) (@github-actions[bot])
• [hotfix] Symmetric mul! for dense/condensed KKT systems (#198) (@sshin23)
• fix: do not overwrite number of threads in BLAS (#199) (@frapac)
• add functions to decompose the time spent in callbacks/linear solver (#200) (@frapac)
• [hotfix] Symmetric mul! for GPUs (#201) (@sshin23)
• fix unscaling of the constraints during post-processing (#202) (@frapac)
• update README (#203) (@frapac)
• update documentation (#204) (@frapac)
• prep for release 0.5.0 (#206) (@frapac)