Issue with Build Server

[AhmedaliElgabry]

i am runing the server in AWS EC2 instance with g4dn.xlarge GPU instance ,

when i am bulding the application using
==> make -j$(nproc)

I get an error related to CUDA_ARCHITECTURES :

-- CMAKE_CUDA_ARCHITECTURES:
-- The CUDA compiler identification is unknown
-- Detecting CUDA compiler ABI info
CMake Error in /home/ubuntu/multiscan/build/external/open3d_build/CMakeFiles/CMakeTmp/CMakeLists.txt:
CUDA_ARCHITECTURES is empty for target "cmTC_ec848".

CMake Error in /home/ubuntu/multiscan/build/external/open3d_build/CMakeFiles/CMakeTmp/CMakeLists.txt:
CUDA_ARCHITECTURES is empty for target "cmTC_ec848".

CMake Error at /home/ubuntu/anaconda3/envs/multiscan/share/cmake-3.22/Modules/CMakeDetermineCompilerABI.cmake:49 (try_compile):
Failed to generate test project build system.
Call Stack (most recent call first):
/home/ubuntu/anaconda3/envs/multiscan/share/cmake-3.22/Modules/CMakeTestCUDACompiler.cmake:19 (CMAKE_DETERMINE_COMPILER_ABI)
CMakeLists.txt:441 (enable_language)

can you help ,please ? Regards

[SamMaoYS]

Hi @AhmedaliElgabry,

This error seems to be related to your CUDA compiler. Here are some things that I think you can have a check:

1. Is the CUDA compiler installed on the machine you are using? With command nvcc --version, if it shows your cuda compiler driver is installed, than you can check if adding -DCMAKE_CUDA_COMPILER=/path/to/your cuda version/bin/nvcc in the cmake build command works for you.
2. If above doesn't work, I also found this recently, without build from source, the open3d cuda accerlated reconstruction pipeline works with the default pip wheel. So you can bypass the cmake build from source for open3d with cmake option -DMTS_BUILD_OPEN3D=OFF, and directly do pip install open3d==0.16.0. I tested the reconstruction pipeline with version 0.16.0. I hope this also work for other and newer versions, but I haven't tested it myself.

All the best,
Yongsen Mao