Python cuda methods not working despite BUILD_VGICP_CUDA=ON

[AndreejS]

Hey,
I have build everything as mentioned in the Non-ROS section of the README with -DBUILD_VGICP_CUDA=ON and applied what was mentioned in #66 to create the python bindings. GICP and VGICP are working without any issues, however for the CUDA methods (VGICP_CUDA and NDT_CUDA) i get "error: you need to build fast_gicp with BUILD_VGICP_CUDA=ON" and receive the identity as a result.

I verified that the build was successful with CUDA:

As well as verifying that all cuda based folders within gicp are created afterwards.

Additionally i checked if there is any issue with CUDA:

Everything was done on Ubuntu 18.04 with Python 3.9.
I would be grateful for any hints to solve this.

[AndreejS]

For those who experience the same issue, it can be fixed by enabling "BUILD_VGICP_CUDA" in the Python bindings, here.

Since the CUDA approaches are disabled by default in python and "cmake ... -DCMAKE_BUILD_TYPE=Release -DBUILD_VGICP_CUDA=ON" will obviously not fix this.

[Zhao-hangtian]

how to config fastgcpi to use CUDA in python? please