[GSoC 2026] Proposal Draft: Enhancing Clad's C++ Interoperability & Standard Library Coverage

[ayush4874]

Hi @vgvassilev,

I've been working on the array segfaults (PR #1691) and the Elliptic Integrals (PR #1674) recently. Digging into those issues made me realize there's still a pretty big gap in how Clad handles standard C++ types and modern math functions.

I’ve put together a GSoC proposal to tackle this properly.

The Core Idea

I want to focus on "Interoperability." Right now, passing a basic C-array crashes Clad, and we're missing most of the C++17 <cmath> headers. My plan is to fix the plumbing so users can just pass their standard data structures and get derivatives without fighting the tool.

The Plan

1. Static Arrays: Implement proper compile-time unpacking for T[N] and std::array. This builds directly on the fix I added in Fix: Add static_assert to prevent segfault in execute() with array arguments (#1690) #1691 to permanently stop the runtime crashes.
2. Dynamic Containers: Extend the code generator to support std::vector and std::span. This implies synthesizing loops in the derivative code rather than unrolling them, which is critical for variable-length inputs.
3. Full Math Support: Implement the missing 18 special functions (Bessel, Legendre, Zeta, etc.) to hit 100% coverage of C++17 <cmath>.

I’ve drafted the full technical approach and timeline here:
[Link]

Let me know if this scope feels right to you.

Best,
Ayush