v0.9.7

Krylov v0.9.7

Diff since v0.9.6

Merged pull requests:

• Add cgls-shift algorithm (#853) (@tmigot)
• Support preconditioners in BiLQ and QMR (#862) (@amontoison)
• Add tests with LinearOperators.jl (#863) (@amontoison)
• Don't use the macro @. (#864) (@amontoison)
• Fix copy_triangle for Julia 1.11 (#866) (@amontoison)
• Fix preconditioning in cg when optional radius argument is positive (#868) (@mpf)
• Fix download count badge (#869) (@pricklypointer)
• Use GHA to test on Apple Silicon (#871) (@amontoison)
• Update test_allocations.jl (#872) (@amontoison)
• Update Aqua.yml (#873) (@amontoison)
• Bump actions/download-artifact from 3 to 4.1.7 in /.github/workflows (#874) (@dependabot[bot])
• Validate CITATION.cff (#875) (@tmigot)
• [Buildkite] Update pipeline.yml (#877) (@amontoison)
• Test block-GMRES on GPUs (#878) (@amontoison)
• Update the documentation (#879) (@amontoison)
• [documentation] Add an example with RandomizedPreconditioners.jl (#880) (@amontoison)
• [documentation] Update preconditioners.md (#881) (@amontoison)
• [documentation] Add a matrix-free example with FFTW.jl (#883) (@amontoison)
• Add a reference for USYMQR (#885) (@amontoison)
• Use metaprogramming to generate out-of-place methods (#887) (@amontoison)
• [documentation] Replace IncompleteLU.jl by KrylovPreconditioners.jl (#889) (@amontoison)
• Improve type inference of extract_parameters (#890) (@amontoison)
• Define cg_lanczos_shift with metaprogramming (#892) (@amontoison)
• Use metaprogramming to generate more out-of-place methods (#893) (@amontoison)
• Generate all out-of-place methods in krylov_solve.jl (#894) (@amontoison)
• [documentation] Improve the matrix-free example with FFT (#895) (@amontoison)
• Add a macro kfill! (#897) (@amontoison)
• Swap the arguments in the macro kcopy! (#898) (@amontoison)
• Use the macro kcopy! instead of broadcast (#899) (@amontoison)
• Improve the docstrings of block-Krylov processes (#900) (@amontoison)
• Specify that memory is an optional argument for some KrylovSolvers (#901) (@amontoison)
• [documentation] Improve the structure of matrix_free.md (#903) (@amontoison)

Closed issues:

• Use 5-argument mul! (#353)
• Add documentation about memory parameter (#521)
• convert function for solvers (#585)
• Use Real as FC instead of AbstractFloat ? (#646)
• ReshapedArrays support (#770)
• Download badge is broken (#851)
• block_gmres fails with LinearOperator (#854)
• Add tests with LinearOperators.jl (#861)
• Possible support for Dual types for Automatic Differentiation (#870)
• Use metaprogramming to generate all out-of-place methods (#888)

v0.9.6

Krylov v0.9.6

Diff since v0.9.5

Merged pull requests:

• Add comments about using block-GMRES on GPUs (#844) (@amontoison)
• [documentation] KrylovOperator -- optimized sparse products on NVIDIA and AMD GPUs (#848) (@amontoison)
• Add copyto! for KrylovStats (#849) (@amontoison)
• Update CITATION.cff (#852) (@tmigot)
• Update ci.yml (#857) (@amontoison)
• Support ComponentArrays.jl in Krylov.jl (#859) (@amontoison)

Closed issues:

• [documentation] Add an optimized linear operator for sparse matrix-vector product on Nvidia GPUs (#783)
• Implement copyto! for KrylovStats (#846)

v0.9.5

Krylov v0.9.5

Diff since v0.9.4

Merged pull requests:

• [documentation] AMDGPU.jl supports rocSPARSE (#781) (@amontoison)
• [MINARES] Fix warm-start (#794) (@amontoison)
• Fix a typo in test/gpu/amd.jl (#796) (@amontoison)
• [documentation] Update the section about AMD GPUs (#798) (@amontoison)
• Improve the function hermitian_lanczos (#799) (@amontoison)
• Block-Krylov processes (#800) (@amontoison)
• Add an option reorthogonalization for two processes (#803) (@amontoison)
• Update the pseudocode of the Krylov processes (#805) (@amontoison)
• [documentation] Explain the transpose-free variant of the non-Hermitian Lanczos process (#807) (@amontoison)
• Update paper.md (#809) (@danielskatz)
• Fix the issue with the adjoint in JOSS (#810) (@amontoison)
• [documentation] Update Documenter.jl (#811) (@amontoison)
• Update Krylov processes to return more arguments (#813) (@amontoison)
• [JOSS] Update CITATION.cff and DOI badge (#815) (@amontoison)
• [JOSS] Add a BibTeX entry (#817) (@amontoison)
• Update CITATION.cff (#818) (@amontoison)
• Add a DOI link for CAR and MINARES (#820) (@amontoison)
• Test view of AMD GPUs (#823) (@amontoison)
• Update pipeline.yml (#829) (@amontoison)
• [documentation] Add a reference for CRAIG (#834) (@amontoison)
• Add compat entries for Test and Random (#835) (@amontoison)
• Add an implementation of BLOCK-GMRES (#840) (@amontoison)
• Release 0.9.5 (#841) (@amontoison)

Closed issues:

• [documentation] Update GPU support (#627)
• Solve matrix-matrix problem (#759)
• Add an option reorthogonalization for the processes (#801)
• Question for GPU computation: lots of time on vector products (#822)
• Automatically promote atol and rtol to eltype(b) (#828)
• Move Block-GMRES in Krylov.jl (#831)
• Add a reference for CRAIG (#833)

v0.9.4

Krylov v0.9.4

Diff since v0.9.3

Closed issues:

• GPU Example is broken (#754)
• GMRES Convergence Without Meeting atol/rtol (#774)
• Add an example and a test for the CuSparseMatrixCOO format (#775)

Merged pull requests:

• [documentation] Update tips.md (#777) (@amontoison)
• [documentation] Improve preconditioners.md (#778) (@amontoison)
• [documentation] Add an example with CuSparseMatrixCOO (#779) (@amontoison)
• Update paper.md (#785) (@amontoison)
• An implementation of CAR (#786) (@amontoison)
• An implementation of MINARES (#787) (@amontoison)
• Remove .zenodo.json (#788) (@amontoison)
• Add an example for MINARES (#789) (@amontoison)
• Release 0.9.4 (#790) (@amontoison)
• CITATION.bib -> CITATION.cff (#791) (@amontoison)

v0.9.3

Krylov v0.9.3

Diff since v0.9.2

Closed issues:

• GPU scalar indexing when using LsmrSolver (ERROR: LoadError: Scalar indexing is disallowed.) (#764)

Merged pull requests:

• [JOSS] Review comments of LeilaGhaffari (#765) (@amontoison)
• StaticArrays support (#768) (@amontoison)
• Release 0.9.3 (#772) (@amontoison)

v0.9.2

Krylov v0.9.2

Diff since v0.9.1

Closed issues:

• ktypeof(::AbstractVector) is breaking between 0.9.0 and 0.9.1 (#749)
• Optimize the function solve! (#751)

Merged pull requests:

• Improve the reliability of solver.stats.timer (#748) (@amontoison)
• Optimize the generic function solve! (#752) (@amontoison)
• [CI] Remove the build with Linux ARM (#753) (@amontoison)
• Use metaprogramming to reduce the code (#755) (@amontoison)
• Fix tests with Apple GPUs (#756) (@amontoison)
• Update .builkite/pipeline.yml (#757) (@amontoison)
• Release 0.9.2 (#758) (@amontoison)

v0.9.1

Krylov v0.9.1

Diff since v0.9.0

Closed issues:

• Add tests for verbose mode (#474)
• Add stopping criterion on time limit (#661)
• [documentation] Add keyword arguments in the docstrings (#663)
• [CI] Update Cirrus script (#694)
• [documentation] Update the comment about oneMKL for oneAPI.jl (#709)
• [documentation] Update examples with Nvidia preconditioners (#710)
• PCG produces different output on GPU (#724)
• [JOSS] Add a sentence about Apple Accelerate for LBT (#725)
• Unnecessary eltype restriciton on A (#736)

Merged pull requests:

• Add tests for verbose mode (#475) (@amontoison)
• final pass over paper (#690) (@amontoison)
• Update CI for aarch64 Mac -- Apple M1 (#691) (@amontoison)
• Add an example with BasicLU.jl in preconditioners.md (#692) (@amontoison)
• Fix #694 (#695) (@amontoison)
• Switch Metal.jl CI to the juliaecosystem pipeline. (#696) (@maleadt)
• Use JET to fix bugs and type instabilities (#697) (@amontoison)
• Add new tests for GPUs (#699) (@amontoison)
• Add more details about the number of Krylov methods (#703) (@amontoison)
• Allow more generic storage type S (#705) (@amontoison)
• Check that the KrylovSolvers are compatible with the dimension of the linear problems (#707) (@amontoison)
• [buildkite] Remove the build with AMDGPU.jl] (#708) (@amontoison)
• [documentation] Update the page related to the GPU support (#711) (@amontoison)
• Update LICENSE.md (#712) (@amontoison)
• Improve the JOSS paper with the comments of prj- (#713) (@amontoison)
• [documentation] Update the GPU section (#714) (@amontoison)
• Update the reference of the GPMR paper (#715) (@amontoison)
• Fix the documentation for CUDA.jl 4.1 (#718) (@amontoison)
• Add a threshold for codecov (#719) (@amontoison)
• Adress the remaining comments of prj- (#720) (@amontoison)
• Fix some typos (#721) (@goggle)
• [documentation] Use CUDA.zeros instead of similar (#722) (@amontoison)
• Update examples of the JOSS paper (#723) (@amontoison)
• Add Apple Accelerate for BLAS backends available with LBT (#726) (@amontoison)
• Add a reference for CR (#728) (@amontoison)
• Add the keyword argument timemax in all Krylov methods (#730) (@amontoison)
• Test Krylov.jl with CUDA 12.1.1 (#731) (@amontoison)
• Improve test_mp.jl (#733) (@amontoison)
• Use metaprogramming to define the kwargs of all Krylov methods (#735) (@amontoison)
• Update Breakage.yml for LinearSolve.jl (#737) (@amontoison)
• Update the kwargs of the in-place methods (#738) (@amontoison)
• Display a warning when the operator and the right-hand side don't have the same type (#739) (@amontoison)
• Update codecov.yml (#740) (@amontoison)
• Combine the allocation of the KrylovSolver with the warm-start (#742) (@amontoison)
• Update timemax when a KrylovSolver is allocated (#743) (@amontoison)
• Update the verbose mode to display a timer (#744) (@amontoison)
• Add a field timer in all KrylovStats (#745) (@amontoison)
• Release 0.9.1 (#747) (@amontoison)

v0.9.0

Krylov v0.9.0

Diff since v0.8.4

Closed issues:

• Document callbacks (#587)
• Add GPU support for MINRES-QLP (#632)
• Suggestion: Overhauled minres implementation? (#652)
• [documentation] Change the category of USYMLQ and USYMQR (#665)
• Add ktypeof support for view arrays (#667)

Merged pull requests:

• Add six Krylov processes (#618) (@amontoison)
• [documentation] update callbacks.md (#629) (@amontoison)
• An implementation of FGMRES (#636) (@amontoison)
• Add vector_to_matrix function (#637) (@amontoison)
• Update multiple docstrings and comments (#638) (@amontoison)
• Reduce the storage required by DIOM and DQGMRES (#639) (@amontoison)
• Remove allocations in roots_quadratic (#640) (@amontoison)
• Add a comment about complex matrices in gpu.md (#641) (@amontoison)
• Add FGMRES in processes.md (#642) (@amontoison)
• Add discussion in readme (#643) (@tmigot)
• Add GPU support for to_boundary function (#644) (@amontoison)
• add imput and output arguments in all docstrings (#647) (@amontoison)
• Add the workflow Invalidations.yml (#649) (@amontoison)
• Use Aqua.jl in the tests (#650) (@amontoison)
• Add the support of CuSparseMatrixCOO in gpu.md (#651) (@amontoison)
• Add m and n in all Krylov solvers (#653) (@amontoison)
• Better display of Krylov solvers (#654) (@amontoison)
• [KrylovSolvers] Improve GPU support (#655) (@amontoison)
• Add optional argument(s) in the docstrings (#656) (@amontoison)
• Create a Aqua workflow (#657) (@amontoison)
• Fix Aqua workflow (#658) (@amontoison)
• Fix again the Aqua workflow (#659) (@amontoison)
• Improve allocation tests (#660) (@amontoison)
• [LNLQ] Use utolx and utoly instead of etolx and etoly (#662) (@amontoison)
• [documentation] Replace symmetric by Hermitian (#664) (@amontoison)
• [documentation] Add a section storage requirements (#666) (@amontoison)
• Change the category of USYMLQ and USYMQR (#668) (@amontoison)
• Support a view of a matrix as a right-hand side b (#669) (@amontoison)
• Improve Krylov processes (#671) (@amontoison)
• Update test/test_utils.jl (#672) (@amontoison)
• Fix a typo in processes.md (#674) (@amontoison)
• Comment processes (#675) (@amontoison)
• Use the same tolerance for all allocation tests (#676) (@amontoison)
• Add a keyword argument iostream for all Krylov methods (#677) (@amontoison)
• Update the show function for Krylov solvers (#678) (@amontoison)
• pass over paper (#681) (@dpo)
• Add documentation for all keyword arguments (#682) (@amontoison)
• [documentation] Remove the subsection restarted methods (#684) (@amontoison)
• Add notes about relations between processes (#685) (@amontoison)
• Check all docstrings (#688) (@amontoison)
• Release 0.9.0 (#689) (@amontoison)

v0.8.4

Krylov v0.8.4

Diff since v0.8.3

Closed issues:

• Add documentation about preconditioners (#222)
• Tests failures on macOS ARM M1 (#560)
• Failing CR test (#598)
• fill out residual in output stats (#612)

Merged pull requests:

• comment out failing tests (#561) (@dpo)
• Setup JOSS paper (#603) (@amontoison)
• use strict=true in docs (#604) (@dpo)
• Update ktypeof function to support BlockArrays and FillArrays (#606) (@amontoison)
• Use the multiple dispatch for ktypeof function (#607) (@amontoison)
• [documentation] Add a section about preconditioners (#609) (@amontoison)
• Replace mul! by mulorldiv! for preconditioners (#610) (@amontoison)
• [documentation] Refactor examples (#611) (@amontoison)
• [documentation] Add a section for each Krylov method (#613) (@amontoison)
• Improve rNorm estimate in CR (#614) (@amontoison)
• Use a testset for all allocation tests (#615) (@amontoison)
• Update .cirrus.yml for FreeBSD 13.0 (#616) (@amontoison)
• Remove two tests for USYMQR (#619) (@amontoison)
• Update CommentPR.yml (#620) (@JarLob)
• add FPS in breakage tests (#621) (@tmigot)
• Use Aᴴ instead of Aᵀ (#622) (@amontoison)
• Add a reference to BiCG paper (#623) (@amontoison)
• [KrylovSolvers] inner constructors -> outer constructors (#624) (@amontoison)
• Use buildkite to test the GPU support (#625) (@amontoison)
• Test Krylov macros (#626) (@amontoison)
• Update workflows (#628) (@amontoison)
• [documentation] update gpu.md (#630) (@amontoison)
• [GPU] Remove random tests (#631) (@amontoison)
• Add more examples with the preconditioners (#633) (@amontoison)
• Release 0.8.4 (#634) (@amontoison)

v0.8.3

Krylov v0.8.3

Diff since v0.8.2

Closed issues:

• Add an option restart for FOM and GMRES (#412)
• Add documentation about warm start (#422)
• Add stopping tests based on finite-precision system used (#589)
• Issue when solving a sparse linear system on the GPU (#594)

Merged pull requests:

• [GMRES] Callback utility function to get solver.x (#599) (@geoffroyleconte)
• use push_preview = true in deploydocs (#600) (@dpo)
• New LSMR stats (#601) (@AntoninKns)
• Bumping version to 0.8.3 (#602) (@AntoninKns)