Skip to main content

dggs

The dggs shader module provides WGSL helpers for compact 64-bit DGGS cell IDs stored as two u32 words. It is intended for WebGPU storage-buffer paths that need to compare, decode, or expand geohash, quadkey, S2, A5, or H3 cell keys without first materializing CPU polygons.

Usage​

import {dggs} from '@luma.gl/shadertools';

const modules = [dggs];

Arrow BigUint64Array storage is little-endian on supported browser and Node WebGPU runtimes. Convert storage reads before using the canonical DGGS helpers:

let cellKey = dggs_u64_from_little_endian_words(cellKeys[cellIndex]);

The module exposes Uint64 word helpers plus encoding-specific boundary and coordinate helpers used by @luma.gl/arrow DGGS preparation. Application code usually reaches it through prepareDggsCellKeyGPUVector() and prepareDggsCellPathGPUVector() rather than calling every WGSL helper directly.

Boundary helpers also expose *_fp64_split wrappers that return vec4f(longitudeHigh, latitudeHigh, longitudeLow, latitudeLow). These wrappers are layout compatibility helpers only: the DGGS decode math currently still computes Float32 longitude/latitude values, so the low components are zero until true higher-precision DGGS decode math is added.

Remarks​

  • dggs is WGSL-only.
  • Canonical DGGS helper word order is vec2<u32>(high, low).
  • Arrow Uint64 storage reads should pass through dggs_u64_from_little_endian_words() before DGGS comparisons or decoders.
  • The Global Grids example shows UTF-8 DGGS IDs parsed into Uint64 GPU keys and expanded into Float32 cell boundaries.