Three backends, one API
Switch between SVG, Canvas2D, and WebGL2 without touching your drawing code.
d3gl
GPU-accelerated rendering for d3 — SVG, Canvas2D, and WebGL2 from one API.
Familiar d3 contexts
Context-driven d3 generators (geoPath, d3-shape, d3-hierarchy links) render unchanged.
Project & tessellate once
Geometry is uploaded once; pan/zoom is a single transform-matrix uniform.
GPU recolor & picking
Recolor or show/hide is one texture write; hit-testing is GPU-accelerated.
GeoJSON & maps
Project any GeoJSON with any d3 projection and render the result on the GPU.
Large-scale networks
The network() engine draws millions of nodes and links — in-library force and nested module layouts (worker/GPU) that switch into each other live, streamed from where the nodes are, with smooth re-layout transitions that keep the camera framed, level-of-detail maps of networks that aggregate only where nodes would overlap and whose expanded modules keep their boundaries — in the module’s own colour, as wide as its flow — and a spatial level of detail that keeps a spread-out web graph of hundreds of thousands of nodes to a screenful of glyphs while it lays out.
GPU force layout
layout({ backend: "gpu" }) runs the force layout in WebGL2 on the renderer’s own device and streams it on screen within a per-frame GPU budget, without blocking the main thread; "auto" takes the GPU where it works and the worker elsewhere.
Publication export
Export crisp PNG and SVG straight from the same scene.