Robust options for Enzyme by michel2323 · Pull Request #378 · JuliaSmoothOptimizers/ADNLPModels.jl

michel2323 · 2026-02-17T16:55:23Z

Alexis, you had issues with OptimalControl.jl and Enzyme. I've set here the most robust options for Enzyme. Maybe it works now. The compile time might be slower now.

I couldn't run your example. It gives an unrelated error. Maybe I use it wrong.

╰─$ julia --project enzyme.jl                                                                                                                                           1 ↵
ERROR: LoadError: MethodError: no method matching iterate(::CTDirect.DOCP{CTDirect.Midpoint, Model{CTModels.Autonomous, CTModels.TimesModel{CTModels.FixedTimeModel{Int64}, CTModels.FixedTimeModel{Float64}}, CTModels.StateModel, CTModels.ControlModel, CTModels.EmptyVariableModel, var"#fun##296#16", CTModels.LagrangeObjectiveModel{var"#fun##301#17"}, CTModels.ConstraintsModel{Tuple{Vector{Float64}, CTModels.var"#path_cons_nl!#build##1"{Tuple{}, Vector{Int64}, Int64}, Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, var"#fun##289#15", Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, Vector{Int64}, Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, Vector{Int64}, Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, Vector{Int64}, Vector{Float64}, Vector{Symbol}}}, var"#18#19"}, CTDirect.ADNLPBackend})
The function `iterate` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  iterate(::Pkg.Resolve.NodePerm, Any...)
   @ Pkg ~/.julia/juliaup/julia-1.12.5+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/Resolve/maxsum.jl:242
  iterate(::Pipe, ::Int64)
   @ Base stream.jl:1297
  iterate(::Pipe)
   @ Base stream.jl:1296
  ...

Stacktrace:
 [1] indexed_iterate(I::CTDirect.DOCP{CTDirect.Midpoint, Model{CTModels.Autonomous, CTModels.TimesModel{CTModels.FixedTimeModel{Int64}, CTModels.FixedTimeModel{Float64}}, CTModels.StateModel, CTModels.ControlModel, CTModels.EmptyVariableModel, var"#fun##296#16", CTModels.LagrangeObjectiveModel{var"#fun##301#17"}, CTModels.ConstraintsModel{Tuple{Vector{Float64}, CTModels.var"#path_cons_nl!#build##1"{Tuple{}, Vector{Int64}, Int64}, Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, var"#fun##289#15", Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, Vector{Int64}, Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, Vector{Int64}, Vector{Float64}, Vector{Symbol}}, Tuple{Vector{Float64}, Vector{Int64}, Vector{Float64}, Vector{Symbol}}}, var"#18#19"}, CTDirect.ADNLPBackend}, i::Int64)
   @ Base ./tuple.jl:165
 [2] top-level scope
   @ ~/project/enzyme.jl:24

Copilot

Pull request overview

This PR adjusts the Enzyme-based AD extension to use more “robust” Enzyme configuration and runtime-activity mode selection, aiming to reduce Enzyme failures on complex structs/typing patterns (e.g., OptimalControl.jl use cases), at the cost of potentially slower compilation.

Changes:

Set Enzyme global robustness options (strictAliasing!, looseTypeAnalysis!) when the extension loads.
Switch multiple Enzyme entrypoints (autodiff, jacobian) to use Enzyme.set_runtime_activity(...) for Forward/Reverse modes.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-17T16:58:48Z

ext/ADNLPModelsEnzymeExt.jl

+# Configure Enzyme for robustness
+Enzyme.API.strictAliasing!(false)   # handle Union types in Printf/@sprintf
+Enzyme.API.looseTypeAnalysis!(true) # handle unresolved types in complex structs


These Enzyme configuration calls are executed at module top-level. In Julia, top-level side effects may run during precompilation and won’t reliably re-run when loading the precompiled extension, so the settings may not actually be applied in user sessions. Consider moving this configuration into __init__() (and optionally guarding with isdefined/try to avoid hard load failures) so it is applied at runtime when the extension is loaded. Also note this mutates global Enzyme state for the whole Julia process; if possible, scope it (or make it opt-in) to avoid impacting other packages using Enzyme.

Suggested change

# Configure Enzyme for robustness

Enzyme.API.strictAliasing!(false) # handle Union types in Printf/@sprintf

Enzyme.API.looseTypeAnalysis!(true) # handle unresolved types in complex structs

function __init__()

# Configure Enzyme for robustness at runtime (not during precompilation)

try

if isdefined(Enzyme, :API)

Enzyme.API.strictAliasing!(false) # handle Union types in Printf/@sprintf

Enzyme.API.looseTypeAnalysis!(true) # handle unresolved types in complex structs

end

catch err

# Avoid hard failures if Enzyme's API or availability changes

@debug "Failed to configure Enzyme in ADNLPModelsEnzymeExt.__init__" exception = (err, catch_backtrace())

end

end

github-actions · 2026-02-17T17:06:45Z

Package name	latest	stable
JSOSuite
PartiallySeparableNLPModels
PartiallySeparableSolvers
SolverTest

amontoison · 2026-02-17T17:54:09Z

@michel2323
The MWE is:

using OptimalControl
using NLPModelsIpopt
using Enzyme

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

solve(ocp; adnlp_backend = :enzyme)

michel2323 · 2026-02-18T18:14:37Z

The MWE should work now 🙂.

amontoison · 2026-02-18T19:44:22Z

The unit tests for Enzyme are failing now :(

michel2323 · 2026-02-19T16:29:27Z

I'll take a look...

amontoison · 2026-02-24T17:51:44Z

@jbcaillau

amontoison · 2026-02-25T02:37:32Z

test/sparse_hessian.jl

+    # In short: this is not a numerical precision issue, but a limitation of
+    # Enzyme when differentiating LLVM bit-manipulation code generated for
+    # packed Dual numbers
+    (backend == ADNLPModels.SparseEnzymeADHessian) && (T == Float32) && continue


@jbcaillau @michel2323
The issue with single precision is very deep...

jbcaillau · 2026-02-25T12:49:30Z

thanks @michel2323 for the fix. @amontoison et al: looking forward the new release (cc: @PierreMartinon)

Robust options for Enzyme

303d293

Copilot AI review requested due to automatic review settings February 17, 2026 16:55

Copilot started reviewing on behalf of michel2323 February 17, 2026 16:56 View session

Copilot AI reviewed Feb 17, 2026

View reviewed changes

michel2323 added 2 commits February 18, 2026 11:52

Fixes

c0a15ac

Use ForwardDiff over Enzyme Reverse for second-order

d2b8dd0

jbcaillau mentioned this pull request Feb 24, 2026

[Bug] Reverse over forward AD issues with ADNLP control-toolbox/OptimalControl.jl#481

Open

Update sparse_hessian.jl

5836266

amontoison force-pushed the ms/robust_enzyme branch from a7e38dd to 5836266 Compare February 25, 2026 02:36

amontoison reviewed Feb 25, 2026

View reviewed changes

amontoison approved these changes Feb 25, 2026

View reviewed changes

amontoison merged commit e5fdf7e into main Feb 25, 2026
28 of 30 checks passed

amontoison deleted the ms/robust_enzyme branch February 25, 2026 02:45

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Robust options for Enzyme#378

Robust options for Enzyme#378
amontoison merged 4 commits intomainfrom
ms/robust_enzyme

michel2323 commented Feb 17, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 17, 2026

Uh oh!

github-actions bot commented Feb 17, 2026 •

edited

Loading

Uh oh!

amontoison commented Feb 17, 2026 •

edited

Loading

Uh oh!

michel2323 commented Feb 18, 2026 •

edited

Loading

Uh oh!

amontoison commented Feb 18, 2026

Uh oh!

michel2323 commented Feb 19, 2026

Uh oh!

amontoison commented Feb 24, 2026

Uh oh!

amontoison Feb 25, 2026

Uh oh!

Uh oh!

jbcaillau commented Feb 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

-# Configure Enzyme for robustness
-Enzyme.API.strictAliasing!(false)   # handle Union types in Printf/@sprintf
-Enzyme.API.looseTypeAnalysis!(true) # handle unresolved types in complex structs
+function __init__()
+  # Configure Enzyme for robustness at runtime (not during precompilation)
+  try
+    if isdefined(Enzyme, :API)
+      Enzyme.API.strictAliasing!(false)   # handle Union types in Printf/@sprintf
+      Enzyme.API.looseTypeAnalysis!(true) # handle unresolved types in complex structs
+    end
+  catch err
+    # Avoid hard failures if Enzyme's API or availability changes
+    @debug "Failed to configure Enzyme in ADNLPModelsEnzymeExt.__init__" exception = (err, catch_backtrace())
+  end
+end

Conversation

michel2323 commented Feb 17, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Feb 17, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Feb 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

amontoison commented Feb 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

michel2323 commented Feb 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

amontoison commented Feb 18, 2026

Uh oh!

michel2323 commented Feb 19, 2026

Uh oh!

amontoison commented Feb 24, 2026

Uh oh!

amontoison Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jbcaillau commented Feb 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

github-actions bot commented Feb 17, 2026 •

edited

Loading

amontoison commented Feb 17, 2026 •

edited

Loading

michel2323 commented Feb 18, 2026 •

edited

Loading