3D + 2D: Testing out my cross-platform WASM graphics engine

Hacker News Discussion ↗

Overview of the Project

  • A cross‑platform graphics engine where “renderlets” are WebAssembly modules that generate graphics data; “wander” is the C++ host/runtime that embeds and runs these modules.
  • Goal: high‑level, data‑driven graphics description compiled to portable Wasm, then mapped to platform GPU APIs. Positioned more as “graphics middleware / plugin system” than full game engine.

Relation to Existing Tech

  • Compared to sokol: sokol is seen as a minimal “STB-like” app/graphics layer; renderlet aims at higher‑level descriptions that generate runtime code.
  • Plans to build on WebGPU via wgpu and wasi-webgpu/wasi-gfx rather than re‑implementing low‑level backends.
  • Some see parallels to Flash/SWF, Rive, Godot, Flutter, Kha, and .NET/COM/CLS‑style intermediate layers.

Shaders, Languages, and Runtime

  • Current version requires host‑provided shaders; long‑term plan is automatic shader handling via higher‑level shader APIs (e.g., WGSL, possible Slang).
  • Today it’s a C++ library; intent is to ship a C ABI as .so/.dll so bindings for Go, Python, etc. can be generated easily.
  • Host application owns the event loop and window; renderlets are embedded modules with defined inputs/outputs.

Use Cases vs Game Engines

  • Supporters argue we need a lightweight, SDL‑like cross‑platform rendering/multimedia layer, not another Unity/Unreal.
  • Skeptics note Unity/Unreal already provide cross‑platform, full‑stack solutions and question whether a small team can match their practical value.
  • Some suggest collaborating with or building atop Godot or other existing engines.

Web, WASM, and Performance

  • Threading via SharedArrayBuffer is reported to work well now in modern browsers, though Safari has limitations.
  • Wasm graphics code currently runs at roughly ~80% of native in informal tests, considered acceptable for most non‑AAA 3D workloads.
  • Several comments highlight WebGPU’s limitations for AAA games (feature gaps, memory issues, overhead), but see strong potential for visualization and rich web apps.

Plugin/Renderlet Model & Skepticism

  • Proponents frame renderlets as sandboxed, reusable graphics plugins (e.g., TAA module) that can be dropped into pipelines across platforms.
  • Critics question why introduce an extra Wasm layer when C++ and WebGPU already exist, and see the main value only if there is a real ecosystem/marketplace of third‑party renderlets.

Future Features & Wishes

  • Strong interest in robust text and font support (likely via HarfBuzz), low‑level audio APIs, and broad input handling.
  • Some see this tech as a possible foundation for cross‑platform GUI frameworks or even Electron‑style app stacks, though that’s not its current focus.