v3.54.0

LinearSolve v3.54.0

Diff since v3.53.0

Merged pull requests:

• Refactor LinearVerbosity to use SciMLLogging's @verbosity_specifier macro (#859) (@ChrisRackauckas-Claude)

v3.53.0

LinearSolve v3.53.0

Diff since v3.52.0

Merged pull requests:

• Fix Tridiagonal cache mutation with parameterized DirectLdiv!{cache} (#856) (@ChrisRackauckas-Claude)

Closed issues:

• Calling solve! on cache built with tridiagonal matrix, mutates the matrix and ruins the cache. (#825)

v3.52.0

LinearSolve v3.52.0

Diff since v3.51.0

Merged pull requests:

• Use nonzeros(A) instead of A.nzval in pattern check (#854) (@j-fu)
• Fix type inference for ForwardDiff Dual problems (#855) (@ChrisRackauckas-Claude)

v3.51.0

LinearSolve v3.51.0

Diff since v3.50.3

Merged pull requests:

• Fix and speed up sparsity pattern check (#845) (@hersle)

Closed issues:

• Size mismatch in dual number handling (#841)

v3.50.3

LinearSolve v3.50.3

Diff since v3.50.2

Merged pull requests:

• fix another method overwrite (#853) (@j-fu)

v3.50.2

LinearSolve v3.50.2

Diff since v3.50.1

Merged pull requests:

• Fix precompilation (#852) (@ChrisRackauckas)

v3.50.1

LinearSolve v3.50.1

Diff since v3.50.0

Merged pull requests:

• Remove LUFactorization method for AbstractSciMLOperator (#849) (@ChrisRackauckas)

Closed issues:

• Precompile warnings for method redefinition in LinearSolveSparseArraysExt in v3.50 (#848)

v3.50.0

LinearSolve v3.50.0

Diff since v3.49.0

Merged pull requests:

• Setup use of has_concretization and test with SciMLOperators/WOperator (#847) (@ChrisRackauckas-Claude)

v3.49.0

LinearSolve v3.49.0

Diff since v3.48.1

Merged pull requests:

• fix default solver selection for CuSparseMatrix (#834) (@hexaeder)
• Fix verbosity log tests to use regex matching (#843) (@ChrisRackauckas-Claude)
• Remove UnPack (#846) (@devmotion)

Closed issues:

• Default algs for CUSPARSE failing (#827)
• dgetrf: Singular matrix when solving a boundary value problem (#830)
• Preconditioners do not work correctly (#839)

v3.48.1

LinearSolve v3.48.1

Diff since v3.48.0

Merged pull requests:

• Revert "Fix default algorithm for sparse CUDA matrices to LUFactorization" (#833) (@hexaeder)
• Update reltol and abstol handling for odd (dual) number types (#840) (@ChrisRackauckas)

Closed issues:

• Intel Fortran Runtime overwriting Interputs (#757)