[flohofwoe]

WebGPU currently doesn't support the "bindless" resource access model (see: https://github.com/gpuweb/gpuweb/issues/380).

The "max number of sampled texture per shader stage" is a runtime device limit, and the minimal value for that seems to be 16. So texture atlasses are still a thing in WebGPU.

WebGPU has render bundles, which allow to pre-record command sequences, but even with that you don't want to change resource bindings thousands of times per frame (or even hundreds of times).

It might make sense though to build texture atlases dynamically (basically use one very big texture as "tile cache") and update that via writeTexture() calls (just don't rebuild the entire atlas each frame).