v0.10.20

ForwardDiff v0.10.20

Diff since v0.10.19

Closed issues:

• Addition of LogExpFunctions generates lots of warnings. (#543)
• Segfault from collecting a range of duals (#548)

Merged pull requests:

• define ArithmeticStyle for Dual (#549) (@KristofferC)

v0.10.19

ForwardDiff v0.10.19

Diff since v0.10.18

Closed issues:

• ForwardDiff.derivative(float, 1) returns Float16(1.0) (#362)
• float(::Dual) disagrees with Base (#492)
• Support normalize (#521)
• DiffRules [1.1,1.2] breaks precompilation (#529)
• method conj overwritten in dual.jl (#531)

Merged pull requests:

• Int(::Dual) and Integer(::Dual) (#508) (@dlfivefifty)
• Error message contains "expexts"; changed it to expects (#522) (@jwscook)
• Overload floatmin, floatmax (#525) (@dlfivefifty)
• Ignore rules with more than two arguments (#530) (@mcabbott)
• Remove rule for conj (#532) (@mcabbott)
• fix float(::Dual) and add float(::Type{<:Dual}) (#535) (@stevengj)

v0.10.18

ForwardDiff v0.10.18

Diff since v0.10.17

Closed issues:

• Broadcast max is not supported (#512)

Merged pull requests:

• Fix "FowardDiff.can_dual" typo (#500) (@mcabbott)
• allow printing using at-printf (#511) (@ranocha)
• eigen and eigvals for Symmetric, Hermitian, and SymTridiagonal (#513) (@dlfivefifty)

v0.10.17

ForwardDiff v0.10.17

Diff since v0.10.16

Closed issues:

• ForwardDiff 10x slower than FiniteDiff equivalent (#502)
• Support structured (banded) Jacobians? (#507)

Merged pull requests:

• Update README.md (#505) (@ViralBShah)
• Remove pure function annotation (#509) (@mohamed82008)

v0.10.16

ForwardDiff v0.10.16

Diff since v0.10.15

Closed issues:

• ForwardDiff.jl does not throw an error even when function outputting an array is used to calculate the Hessian (#501)

Merged pull requests:

• Overload div for Dual (#503) (@sethaxen)

v0.10.15

ForwardDiff v0.10.15

Diff since v0.10.14

Closed issues:

• Having a jvp function for jacobian vector product (#487)
• AD Hessian is inaccurate for only one entry (#488)
• inaccurate gradients for matrix logarithm (#493)
• Derivative through sincospi gives StackOverflowError (#497)

Merged pull requests:

• Friendly errors on array vs. scalar mistakes (#489) (@mcabbott)
• Fix for namespace bug in binary_dual_definition (#491) (@torfjelde)
• Fix for derivative through sincospi gives StackOverflowError (#499) (@mchitre)

v0.10.14

ForwardDiff v0.10.14

Diff since v0.10.13

Merged pull requests:

• Make seed! and extract_jacobian! gpu-friendly (#472) (@charleskawczynski)

v0.10.13

ForwardDiff v0.10.13

Diff since v0.10.12

Closed issues:

• Support for complex numbers (#157)
• Vote: should we add Complex number support as a feature of ForwardDiff. (#456)
• Missing website (#464)
• extract_jacobian calls generic function in generated function (#468)

Merged pull requests:

• Document multidimensional arrays (#449) (@thisrod)
• Add prevfloat and nextfloat for Dual (#453) (@johnmyleswhite)
• README: properly format citation (#465) (@zlatanvasovic)
• Avoid calling highly overloaded function in generated extraction functions (#469) (@andreasnoack)
• Fix link to upgrade guide in README.md (#470) (@jeffreyesun)
• Update advanced.md with NANSAFE file location. (#473) (@Se7enroll)
• CompatHelper: bump compat for "SpecialFunctions" to "1.0" (#474) (@github-actions[bot])
• Relax print tests (#476) (@oxinabox)
• Complete switch to GitHub Actions based CI (#483) (@andreasnoack)
• CompatHelper: bump compat for "StaticArrays" to "1.0" (#484) (@github-actions[bot])

v0.10.12

ForwardDiff v0.10.12

Diff since v0.10.11

Closed issues:

• Package author for citations? (#460)
• Question: does ForwardDiff work with Python functions? (#461)

Merged pull requests:

• Address CommonSubexpressions 0.3 deprecation (#462) (@omus)

v0.10.11

ForwardDiff v0.10.11

Diff since v0.10.10

Closed issues:

• AD of DifferentialEquations.jl ODE solver clashes with AD in Turing.jl (#446)
• Simple code fails (#457)
• Sparse Hessian (#458)

Merged pull requests:

• CompatHelper: bump compat for "CommonSubexpressions" to "0.3" (#459) (@github-actions[bot])