Benchmarks on GCC 16.1 show C++26 reflection’s compile cost comes almost entirely from <meta> inclusion (~155ms/TU), not the reflection algorithm itself.
Key Takeaways
X-macro with const char* is fastest: ~27ms/TU for N=4, beating even the <meta> include-only cost of ~181ms.
Reflection algorithm scales at ~0.07ms/enumerator, nearly identical to a hand-rolled switch, once header cost is excluded.
PCH for <meta> cuts reflection compile time ~2.3x (187ms to 81ms for N=4); C++20 modules in GCC 16 do the opposite, adding ~2.2x slowdown.
enchantum scales with configured scan range, not actual enum size, making it expensive for small enums (~24ms overhead regardless of N).
At 500 TUs, reflection costs ~94 CPU-seconds vs ~13 seconds for the const char* X-macro approach, though parallelism reduces wall-clock impact.
Hacker News Comment Review
Commenters flagged that GCC 16 results may not generalize: Clang’s C++26 reflection is not yet available for comparison, and module behavior in particular is compiler-specific.
The C++26 reflection syntax drew sharp criticism for readability; several commenters noted it looks foreign even to experienced C++11/14 developers, undercutting the “no boilerplate” framing.
Compile-time debugging of consteval reflection code remains an open gap; static_assert is the main tool, with no step-through debugger support available.
Notable Comments
@dataflow: Questions whether enchantum can handle duplicate enum values, discontiguous ranges, or MSVC – limits not addressed in the benchmark.
@king_geedorah: “Another win for X macros and for C style in general” – frames results as a vindication of preprocessor approaches despite ergonomic costs.
