I’m trying to compute all third partial derivatives of a function using ForwardDiff, but cannot get it to be type stable and I don’t understand why. Here is a minimal example: using ForwardDiff using StaticArrays function tf( x ) return ( exp( 0.1*x[1] + 0.2*x[2] ) ) end function get_derivati…

yes, i found the same problem, but when needing the primal, first and second derivative. for what i understand, when using StaticArrays, the chunk is always the length of the SArray

related: [image] Allocation on ForwardDiff + DiffResults + StaticArrays General Usage Cf https://github.com/JuliaDiff/ForwardDiff.jl/pull/315 Something similar for Hessian may help resolve this.

Type stability for higher derivatives in ForwardDiff

General Usage

longemen3000 August 19, 2021, 10:07pm 3

Topic		Replies	Views
Type Stabled Jacobian of a Hessian Performance forwarddiff , type-stability , autodiff	3	168	November 28, 2025
Jacobian of StaticArray function becoming type-unstable Performance	3	709	December 28, 2018
Gradient to jacobian in type stable way with ForwardDiff General Usage forwarddiff , type-stability	0	190	December 1, 2023
Type instability in ForwardDiff.gradient Numerics performance , forwarddiff , type-stability	8	1162	October 20, 2021
Allocations with nested Jacobians with StaticArrays and ForwardDiff General Usage	6	221	May 23, 2026

Type stability for higher derivatives in ForwardDiff

Related topics