[PiterLewis]

Description

Fixes a compilation error with a modern c++20 toolchains caused by the lerp functions defined in 'cxx/herlper_math.h'. The global 'leap' overloads conflict with the std library 'std::lerp' declaration when compiling CUDA extension.

Changes

Renamed the four internal helper overloads. From lerp to -> mc_lerp.

Validation

Tested in Kaggle environment with:

• Python 3.12
• PyTorch 2.10.0+cu128
• CUDA available
• NVIDIA Tesla T4

Before the change, torchmcubes failed during compilation due to the
lerp/std::lerp conflict.

After the change:

• torchmcubes builds successfully
• import torchmcubes succeeds
• torchmcubes.marching_cubes() executes successfully on CUDA
• Valid vertices and triangles are produced

[tatsy]

Thank you for your PR.
Finally, I found that "helper_math.h" is no longer necessary, as "cuda_runtime.h" provides vector arithmetics.
I have thoroughly removed the dependency on "helper_math.h" and think now it works on your environment.

[PiterLewis]

Going to check it! Happy coding!