NVlabs / nvdiffrast

Nvdiffrast - Modular Primitives for High-Performance Differentiable Rendering

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Native CPP support

tperl-cg opened this issue · comments

Hi,
Can I build nvdiffrast to work with cpp and ceres optimizer?

No, the code currently supports PyTorch/TensorFlow only. However, if you're willing to port nvdiffrast ops to support a different framework, all you need is a new set of C++ stubs that act as an interface to the Cuda kernels, similar to how the C++ code in nvdiffrast/torch interfaces between PyTorch and Cuda.

Can It work with iOS metal? or on CPU? how can I support this with the tool chain? can you give an example?

No, it won't work without a Cuda-compatible GPU. All the important bits are implemented using Cuda.

I am confused, it is written the it supports open-gl