Whitepapers and Technical Write-ups
The benchmark is the proof. These papers document the methods and measurements behind the libraries published here — what was tried, what worked, what didn't, and what the numbers actually say.
Each paper is written as an engineering write-up rather than an academic submission: problem, approach, measurements with caveats, and an honest read of when the technique helps versus when it doesn't.
consteval — field values appear as immediates in the output. Includes the one subtlety that makes or breaks the guarantee: why the uniqueness concept must constrain the whole argument pack, not each argument individually.
consteval · type routing · API design
-O1 and above. Includes compile-time scaling measurements across GCC, Clang, and MSVC, and an honest account of where the pattern degrades and when a plain std::tuple is the better tool.
IDIV from indexing-bound kernels by hoisting divisors into the type system. Measured 12–30× operation-level speedup on RTX 5070 Ti across 16 divisors at 32-bit and 64-bit widths. Includes the honest finding that all three magic-number delivery mechanisms are statistically tied — the win is Granlund–Montgomery itself, not the compile-time encoding.
The papers here are not peer-reviewed in the academic sense. They are engineering documents: the techniques they describe are in production use in the libraries listed on the projects page, the benchmarks are reproducible from the harnesses in the organization's repositories, and the numbers are taken on hardware that anyone can buy.
Where a technique has limits — where it doesn't apply, where the compiler already does the equivalent, or where the win is smaller than it looks in isolation — those limits are stated in the paper rather than left for the reader to discover. A speedup number without its caveats is not a result; it's a marketing claim.
If you spot something that looks wrong, the channels to flag it are the GitHub issues on the relevant repository and the Discord.