v1.3.0

ForwardDiff v1.3.0

Diff since v1.2.2

Merged pull requests:

• Add derivative rule for LinearAlgebra.givensAlgorithm (#783) (@devmotion)

Closed issues:

• Method ambiguities reported by Aqua (#681)

v1.2.2

ForwardDiff v1.2.2

Diff since v1.2.1

Merged pull requests:

• Backport #770 to ForwardDiff 0.10 (#773) (@devmotion)
• CI tests with NaN-safe mode (#776) (@devmotion)
• Fix NaN-safe mode (#777) (@devmotion)
• Avoid non-lazy string interpolation and replace @assert with ArgumentError (#779) (@devmotion)
• Bump version from 1.2.1 to 1.2.2 (#782) (@devmotion)

Closed issues:

• Allocation tests broken since Julia 1.9 (#689)
• NaN gradient occurring in SpecialFunctions.gamma, when using float arguments. (#705)
• Product of Inf terms leading to NaNs (#745)
• NaN-safe mode is not NaN-safe enough (#774)
• Zero code coverage for NaN-safe mode (#775)
• JET.@test_opt fails on Julia >= 1.12 due to AssertionErrors (#778)
• Type stability of gradient and jacobian (#781)

v0.10.39

ForwardDiff v0.10.39

Diff since v0.10.38

This release has been identified as a backport.
Automated changelogs for backports tend to be wildly incorrect.
Therefore, the list of issues and pull requests is hidden.

v1.2.1

ForwardDiff v1.2.1

Diff since v1.2.0

Merged pull requests:

• Fix gradient and Jacobian for functions with Dual output (#770) (@devmotion)

Closed issues:

• ForwardDiff.gradient and ForwardDiff.jacobian don't handle functions with ForwardDiff.Dual output correctly (#769)

v1.2.0

ForwardDiff v1.2.0

Diff since v1.1.0

Merged pull requests:

• improve performance of hessians with static arrays (#751) (@longemen3000)
• Fix tests by using tag types instead of instances (#768) (@devmotion)
• prevent adding methods to the functions > and >= (#771) (@nsajko)
• v1.2.0 (#772) (@nsajko)

Closed issues:

• Performance issue/allocations in DiffResults-hessian! with StaticArrays. (#720)

v1.1.0

ForwardDiff v1.1.0

Diff since v1.0.1

Merged pull requests:

• Update CI (#730) (@devmotion)
• Require Julia v1.10+ (#731) (@dlfivefifty)
• Fix seeding with undefined elements (#743) (@devmotion)
• Add SmallTag type for more compact Dual types (#748) (@topolarity)
• Add section in docs about limitation for functions composed of non-differentiable functions (#753) (@JoshuaLampert)
• Also test zero allocations on 1.10 (#754) (@andreasnoack)
• Fix formatting in docs (#755) (@JoshuaLampert)
• Update badges in README.md (#763) (@andreasnoack)
• Update ci.yml to test on windows and mac as well (#764) (@ViralBShah)
• Revert "Add SmallTag type for more compact Dual types" (#765) (@oscardssmith)
• Adjust version to 1.1.0 (#766) (@devmotion)

Closed issues:

• Incorrect computation of Hessian of squared euclidean norm (#303)
• Manually preventing perturbation confusion (#423)
• Jacobian of in-place functions in presence of undefined references (#436)
• Why is seeding ydual necessary? (#740)
• searchsortedfirst of dual returning wrong result (#742)

v1.0.1

ForwardDiff v1.0.1

Diff since v1.0.0

Merged pull requests:

• Update how_it_works.md (#672) (@vpuri3)
• Bump codecov/codecov-action from 4 to 5 (#723) (@dependabot[bot])
• Do not seed structural zeros (#739) (@devmotion)

Closed issues:

• Duals and ranges (#275)
• Compiletime for tensor very slow (#278)
• gradients of hessians with divisions (#285)
• printf formatting causes stack overflow (#289)
• FieldVectors (#305)
• Problem with abs (#377)
• Complex valued pow gives stackoverflow error (#486)
• Complex derivative of power fails (#514)
• Replace values of Duals with other numbers (#526)
• Thread Safety of ForwardDiff.GradientConfig (#573)
• use StaticArraysCore.jl (#591)
• Violating assumptions about Dual comparisons (#609)
• ERROR: MethodError: no method matching Float32(::ForwardDiff.Dual{ForwardDiff.Tag{var"#11#12", Float32}, Float32, 12}) (#611)
• Derivative is wrong for this inverse quadratic form (#633)
• Different results on Intel vs. AMD with ForwardDiff (#636)
• Type instability in jacobian! filling JacobianResult when using StaticArrays (#639)
• incorrect 2nd derivative of complex exponential (#653)
• Derivative of a function of derivatives (#673)
• ERROR: Package ForwardDiff errored during testing (#700)
• UpperTriangular / LowerTriangular broken on 1.0.0 (#738)

v1.0.0

ForwardDiff v1.0.0

Diff since v0.10.38

Breaking changes

• Equality (==) on Dual numbers now requires both the real and dual part to match (#481). This removes a large number of bugs where the "structure" of e.g, non-zero values in an array was inspected, leading to erroneous derivatives. This might cause slightly different behavior in programs but should in general be more correct than previously.

Merged pull requests:

• Specialize on functions in StaticArrays extension (#721) (@devmotion)
• Remove CI check for invalidations (#722) (@devmotion)
• Bump version to 1.0.0 (#729) (@devmotion)
• Fix typo in DerivativeConfig docs (#733) (@gdalle)
• use Base.Fix1 instead of closures in ForwardDiffStaticArraysExt.jl (#735) (@jondeuce)

Closed issues:

• Can you take derivative of complicated function whose symbolic form is not explicit or not known? (#654)
• Tag v1.0 (#719)
• Modification of equality checking between two Duals not synchronized between master and last release (#726)
• Throwing errors after 1000 time steps simulation (#736)
• Matrix inverse is broken (#737)

v0.10.38

ForwardDiff v0.10.38

Diff since v0.10.37

Merged pull requests:

• Fix exponentiation for NaNMath.pow (#717) (@jClugstor)
• Backport #717 to 0.10 (#718) (@devmotion)

Closed issues:

• Implementation of Dual for NaNMath.pow does not use NaNMath.log (#716)

v0.10.37

ForwardDiff v0.10.37

Diff since v0.10.36

Merged pull requests:

• Update how_it_works.md (#660) (@thomvet)
• Bump actions/checkout from 3 to 4 (#663) (@dependabot[bot])
• Update for Documenter.jl v1 (#683) (@hyrodium)
• Broaden promote_rule for Dual to include AbstractIrrational (#687) (@ElOceanografo)
• Restrict StaticArrays to compatible versions (#688) (@andreasnoack)
• Bump actions/cache from 3 to 4 (#692) (@dependabot[bot])
• Bump codecov/codecov-action from 3 to 4 (#693) (@dependabot[bot])
• Make isless, <, >, <=, and >= consistent with == and isequal (#695) (@devmotion)
• Bump julia-actions/setup-julia from 1 to 2 (#699) (@dependabot[bot])
• Add support for StaticArrays >= 1.7 (#703) (@devmotion)
• Remove redundant DEFAULT_CHUNK_THRESHOLD (#704) (@sjkelly)
• Stop precomputing chunk sizes (#708) (@KristofferC)
• Apply #656 to master (#711) (@devmotion)
• Backport bug and test fixes to release-0.10 to prepare for new 0.10 release (#712) (@devmotion)
• Fix method ambiguity for Base.TwicePrecision (#715) (@devmotion)

Closed issues:

• MethodError: no method matching getrf!(::CuArray{ForwardDiff.Dual{ForwardDiff.Tag{ODEFunction (#612)
• Broken external link (#649)
• Correctly forming nested dual numbers. (#671)
• ForwardDiff fails to compute correct derivative (#676)
• Error requiring Symbolics from Optimization (#685)
• promote_rule ambiguity with AbstractIrrational and ForwardDiff.Dual (#686)
• Working with anonymous functions (#694)