Suggestion: Three.js example on on homepage: https://elalish.github.io/manifold/ about manifold

[elalish]

Again, I think you miss the point of what this library does; it's not intended to be a real-time solution. It's intended to make very complicated designs possible without fear of crashing the CAD program or taking forever to generate.

[bhouston]

What is the performance of this library? Is there numbers you can share?

[FishOrBear]

I think we are all evaluating whether this library can be compiled to wasm and then used on a web 3d editor.

[elalish]

That would be great. Do you have any experience with building WASM?

[fire]

It should be technically possible to compile the library to wasm and then execute from three js rendering mesh buffers.

@elalish Did you remove the CUDA dependency?

[elalish]

CUDA is not exactly a dependency; this library uses Nvidia's Thrust which can target either CUDA, OMP, or standard C++ backends. For instance, the tests in Github actions all use the C++ backend as they don't give us machines with GPUs. A CMake flag chooses the backend you wish to build.

It may be possible to build the C++ version as WASM, though it would be single-threaded (still pretty fast though). Long term, I think the ideal thing would be to build WASM using WebGPU compute kernels. I would imagine someone is working on making C++17 parallel algorithms (which is based on and very similar to Thrust) compile to WASM with WebGPU, but it hasn't happened yet as far as I know. I would rather wait for a compiler stage like that to exist than try to jump right onto the bleeding edge myself.