Use a package extension for Enzyme.jl

[amontoison]

12/02/2026: I also remove all the code for Zygote, it is not anymore relevant in 2026...
I prefer to spend time on the support of DI.jl.

[tmigot]

Connected to #231 ?
If you are going for this, could ou please first drop julia 1.6 for 1.10 from the Project.toml

ADNLPModels.jl/Project.toml

Line 24 in e73cb3e

julia = "^1.6"

(this is breaking change)

[github-actions]

Package name	latest	stable
JSOSuite
PartiallySeparableNLPModels
PartiallySeparableSolvers
SolverTest

[amontoison]

@tmigot it's a wip for #303 but yes I plan to do a last release for 0.8.8 before.
I will do the breaking changes after.

[amontoison]

@michel2323
I try to use OptimalControl.jl with Enzyme as backend for ADNLPModels.jl but I have an error:

using OptimalControl
using NLPModelsIpopt

x0 = [ 0.
       1. ]
A = [  0. 1.
      -1. 0. ]
B = [ 0.
      1. ]
Q = [ 1. 0.
      0. 1. ]
R = 1.
tf = 3.

ocp = @def begin
    t ∈ [0, tf], time
    x ∈ R², state
    u ∈ R, control
    x(0) == x0
    ẋ(t) == A * x(t) + B * u(t)
    0.5∫( x(t)' * Q * x(t) + u(t)' * R * u(t) ) → min
end

docp, nlp = direct_transcription(ocp)

solve(ocp) # error "Cannot determine ordering of Dual tags ForwardDiff.Tag{ReverseDiff.var..."
solve(ocp; adnlp_backend = :enzyme)

The error is:

ERROR: EnzymeRuntimeActivityError: Detected potential need for runtime activity.

Constant memory is stored (or returned) to a differentiable variable and correctness cannot be guaranteed with static activity analysis.
This might be due to the use of a constant variable as temporary storage for active memory (https://enzyme.mit.edu/julia/stable/faq/#faq-runtime-activity).
If Enzyme should be able to prove this use non-differentable, open an issue!

To work around this issue, either:
   a) rewrite this variable to not be conditionally active (fastest performance, slower to setup), or
   b) set the Enzyme mode to turn on runtime activity (e.g. autodiff(set_runtime_activity(Reverse), ...) ). This will maintain correctness, but may slightly reduce performance.

Failure within method:

DOCP_objective(::Vector{Float64}, ::CTDirect.DOCP{CTDirect.Midpoint, Model{…}, CTDirect.ADNLPBackend})
     @ CTDirect ~/.julia/packages/CTDirect/DgVuH/src/docp.jl:614

Hint: catch this exception as `err` and call `code_typed(err)` to inspect the surrounding code.

Mismatched activity for:   %value_phi = phi ptr addrspace(10) [ %6, %L12 ], [ %5, %L9 ] const val:   %5 = load atomic ptr addrspace(10), ptr addrspace(11) %4 unordered, align 8, !dbg !92, !tbaa !94, !alias.scope !98, !noalias !99, !enzyme_type !100, !enzymejl_source_type_Vector\7BFloat64\7D !0, !enzymejl_byref_MUT_REF !0
 Julia value causing error:  Unknown object of type Vector{Float64}
 LLVM view of erring value:     %5 = load atomic ptr addrspace(10), ptr addrspace(11) %4 unordered, align 8, !dbg !92, !tbaa !94, !alias.scope !98, !noalias !99, !enzyme_type !100, !enzymejl_source_type_Vector\7BFloat64\7D !0, !enzymejl_byref_MUT_REF !0

Stacktrace:
 [1] getproperty
   @ ./Base_compiler.jl:54
 [2] get_OCP_variable
   @ ~/.julia/packages/CTDirect/DgVuH/src/disc/common.jl:12
 [3] DOCP_objective
   @ ~/.julia/packages/CTDirect/DgVuH/src/docp.jl:622

Stacktrace:
  [1] getproperty
    @ ./Base_compiler.jl:54 [inlined]
  [2] DOCP_objective
    @ ~/.julia/packages/CTDirect/DgVuH/src/docp.jl:620
  [3] #4
    @ ~/.julia/packages/CTDirect/DgVuH/ext/CTDirectExtADNLP.jl:30 [inlined]
  [4] augmented_julia__4_40863_inner_36wrap
    @ ~/.julia/packages/CTDirect/DgVuH/ext/CTDirectExtADNLP.jl:0
  [5] macro expansion
    @ ~/.julia/packages/Enzyme/S3nC6/src/compiler.jl:6672 [inlined]
  [6] enzyme_call
    @ ~/.julia/packages/Enzyme/S3nC6/src/compiler.jl:6151 [inlined]
  [7] AugmentedForwardThunk
    @ ~/.julia/packages/Enzyme/S3nC6/src/compiler.jl:6099 [inlined]
  [8] autodiff(mode::ReverseMode{…}, f::Const{…}, ::Type{…}, args::Duplicated{…})
    @ Enzyme ~/.julia/packages/Enzyme/S3nC6/src/Enzyme.jl:419
  [9] gradient!(::ADNLPModels.EnzymeReverseADGradient, g::Vector{Float64}, f::Function, x::Vector{Float64})
    @ ADNLPModelsEnzymeExt ~/Bureau/git/ADNLPModels.jl/ext/ADNLPModelsEnzymeExt.jl:73
 [10] grad!(nlp::ADNLPModels.ADNLPModel{Float64, Vector{Float64}, Vector{Int64}}, x::Vector{Float64}, g::Vector{Float64})

The related code is defined here.
I think the culprit is the closure.

cc @ocots @jbcaillau