pycuda.driver.CompileError about deepdrr HOT 1 CLOSED

Hi,

It spent me few days to build the CUDA environment, and now I have successfully built required CUDA environment with PyCUDA on both Windows and Linux Ubuntu 16.04. Here I would like to put some recaps, to help anyone if needed.

1. Windows:

If you are new to Windows like me, and don't know how to add C compiler to path, this link below might be very helpful:
https://code.visualstudio.com/docs/cpp/config-mingw

2. Ubuntu 16.04

I strongly recommend to build DeepDRR required CUDA environment using NVIDIA's official installation guide:
https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html

It's suggested to follow the steps strictly, and not jumping between steps to avoid dependency errors. It's important to check that local driver(can be checked by nvidia-smi), and CUDA complier driver(can be checked by nvcc --version) are matched, and don't install additional driver if the machine already has one. Also, disable secure boot when doing the installation. Here is an useful link to solve the local driver issue:
https://askubuntu.com/questions/902636/nvidia-smi-command-not-found-ubuntu-16-04

Sometimes, unfortunately, it might need to remove/reinstall CUDA toolkit to necessarily solve issues, it depends on how you installed CUDA toolkit on your machine, ways to remove CUDA toolkit on Ubuntu are different from using runfile or deb for installation.

It's also very useful to read official installation guide section 7 to verify installation. Try to compile example provided by CUDA and running the binaries, this is often overlooked by the posts I went through online.
https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions

Here are my specs to setup CUDA environment for DeepDRR:

Python = 3.6.10
PyCuda = 2019.1.2
NVIDIA PyCuda = 2019.1.2
NVIDIAGPU = GTX1660
CUDA Toolkit = 10.1
Driver version = 430.64

As suggested in project README, DeepDRR works well with CUDA 8.0. And @gaocong13 suggested that it also worked with CUDA 9.0. So this time I have confirmed that CUDA 10.1 works.

I think we can close this issue for now. @mathiasunberath please also feel free to share this to other users for reference.

Cheers,

from deepdrr.