CUDA_ARCHITECTURES is empty for target (Fedora 35/36)

[alkavan]

I am trying to build this on Fedora 36, I have the latest available cuda from the official NVIDIA repository. Had to use the fedora35 repository, because fedora36 does not contain the cuda package! I assume this is not related to the problem at hand (although it might be).

$ cmake . -B build
-- The CUDA compiler identification is unknown
-- Check for working CUDA compiler: /usr/local/cuda-11.7/bin/nvcc
CMake Error in /home/foo/Develop/instant-ngp/build/CMakeFiles/CMakeTmp/CMakeLists.txt:
  CUDA_ARCHITECTURES is empty for target "cmTC_fcd43".


CMake Error in /home/foo/Develop/instant-ngp/build/CMakeFiles/CMakeTmp/CMakeLists.txt:
  CUDA_ARCHITECTURES is empty for target "cmTC_fcd43".


CMake Error at /usr/share/cmake/Modules/CMakeTestCUDACompiler.cmake:43 (try_compile):
  Failed to generate test project build system.
Call Stack (most recent call first):
  CMakeLists.txt:11 (project)

This is the nvcc output of --version:

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Jun__8_16:49:14_PDT_2022
Cuda compilation tools, release 11.7, V11.7.99
Build cuda_11.7.r11.7/compiler.31442593_0

It's clear what the problem is, the compiler ABI is not identified. But I don't understand why that should be the case?

I saw two issue on this subject, the fixes suggested there do not work for me? I mean adding -GNinja to CMake command. Any known solution to overcome this? I any case, this does seemed like a bug in instant-ngp ? The extra development packages I had to install for Fedora 36 are:

sudo dnf install openexr-devel liblxi-devel glfw-devel glew-devel libomp-devel libXinerama-devel libXcursor-devel

I would like to add then to README once I can confirm this works.

Also setting the TCNN_CUDA_ARCHITECTURES environment variable, like suggested does not work.