August 5, 2026
Triangulation nation starts nitpicking
Exact, parallel 2D Delaunay triangulation for int32 coordinates
This new point-mapper is crazy fast, and the comments instantly wanted more
TLDR: Delaunay32 says it can process huge sets of grid-based 2D points much faster than popular alternatives, making it a big deal for maps, images, and other pixel-like data. The first community reaction was impressed but instantly pushy, asking why it’s limited to 32-bit numbers and whether it can handle live edits too.
A new C++ library called Delaunay32 just strutted into the room claiming it can turn giant clouds of grid-based 2D points into triangles way faster than some well-known rivals—and yes, the benchmark flex is the whole vibe. On big test sets, the project says it’s more than 10 times faster than one popular open-source option and around 3 times faster than another commercial favorite. Translation for normal humans: if you’ve got a mountain of pixel-like map points, this thing says it can sort them into neat little triangles at serious speed, without the random weirdness that can happen in trickier geometry tools.
But the real juice is in the early community reaction, where the applause came with instant interrogation. One commenter opened with the polite version of “cool story, but what about updates?” asking whether point insertion and deletion are supported too. That’s classic developer energy: someone drops a shiny speed monster, and the crowd immediately asks why it doesn’t also make coffee. The same commenter also poked at the project’s biggest limitation—why stop at 32-bit numbers? In plain English, why lock it to a smaller number size if bigger numbers could reduce rounding mistakes? The implied hot take: is this a brilliant focus choice, or a tradeoff that some users will side-eye?
So the mood is mostly impressed, but with that delicious undertone of nerdy suspicion. People love the speed, love the determinism, and love that it’s built for huge, messy datasets. But the comments are already doing what comments do best: turning a victory lap into a cross-examination.
Key Points
- •Delaunay32 is a C++17 library for exact, deterministic, parallel 2D Delaunay triangulation on signed 32-bit integer coordinates.
- •The library targets discrete or quantized spatial data such as pixels, raster samples, voxel projections, fixed-point maps, and similar integer-grid datasets.
- •Its implementation combines exact integer predicates, a Morton-ordered divide-and-conquer algorithm, compact two-dart topology, and optional multithreading.
- •For one million points, the article reports Delaunay32 automatic-threaded performance of about 10.7× to 11.3× faster than delaunator-cpp across uniform, clustered, and diagonal distributions.
- •The full benchmark was run on an Apple M1 using Apple Clang 21, and the article cautions that results are machine- and workload-specific.