Perf: Fix type instabilities in DifferentiationInterface closures #1132
+237
−206
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This PR fixes a significant performance issue in the gradient, hessian, and other derivative closures generated by `instantiate_function` when using DifferentiationInterface (DI) AD backends. ## Problem The closures were capturing `prep_grad`, `prep_hess`, etc. variables with `Core.Box` types due to conditional assignments inside if-blocks. This caused: - Type instability in the hot path functions - Extra allocations on every gradient/hessian call - ~2x slower performance than direct DI calls ## Solution Use `let` blocks to capture preparation objects with concrete types when creating closures. This ensures the closures are type-stable and allocation-free. ## Performance Improvement Gradient evaluation (ForwardDiff, 2D Rosenbrock): - Before: ~80 ns, 1 allocation (16 bytes), 124% overhead vs DI - After: ~36 ns, 0 allocations, <1% overhead vs DI This is a **2.2x speedup** and **eliminates all allocations** in the gradient evaluation hot path. ## Additional Fix Fixed incorrect use of `value_derivative_and_second_derivative!` (for scalar functions) in fgh! closure. Replaced with correct `value_gradient_and_hessian!` (for multivariate optimization functions). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Core.Boxcapturesvalue_derivative_and_second_derivative!withvalue_gradient_and_hessian!Problem
The closures generated by
instantiate_functionwere capturingprep_grad,prep_hess, etc. withCore.Boxtypes due to conditional assignments inside if-blocks. This caused:@code_warntype)Solution
Use
letblocks to capture preparation objects with concrete types when creating closures. This is a well-known Julia performance pattern to avoid Core.Box boxing.Benchmark Results
Gradient evaluation (ForwardDiff, 2D Rosenbrock):
Additional Fix
Fixed incorrect use of
value_derivative_and_second_derivative!(for scalar functions) in thefgh!closure. This function is for scalar→scalar differentiation. For optimization (vector→scalar), the correct function isvalue_gradient_and_hessian!.Test Plan
@code_warntypeshows no type instabilitiescc @ChrisRackauckas
🤖 Generated with Claude Code