Build issue with Cuda 10.2

[aricursion]

Hi! I'm getting a build issue on a system with cuda-10.2 and multiple GPUS (however only one is visible) The visible GPU is a Titan X Pascal so I set TCNN_CUDA_ARCHITECTURES=61. I ran the following to build: cmake . -B build -DCMAKE_CUDA_COMPILER="/usr/local/cuda-10.2/bin/nvcc", but was met with

-- The C compiler identification is GNU 9.4.0
-- The CXX compiler identification is GNU 9.4.0
-- The CUDA compiler identification is unknown
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detecting CUDA compiler ABI info
CMake Error in CMakeLists.txt:
  CUDA_ARCHITECTURES is empty for target "cmTC_a5ce1".


CMake Error in CMakeLists.txt:
  CUDA_ARCHITECTURES is empty for target "cmTC_a5ce1".


CMake Error at /usr/local/share/cmake-3.19/Modules/CMakeDetermineCompilerABI.cmake:48 (try_compile):
  Failed to generate test project build system.
Call Stack (most recent call first):
  /usr/local/share/cmake-3.19/Modules/CMakeTestCUDACompiler.cmake:19 (CMAKE_DETERMINE_COMPILER_ABI)
  CMakeLists.txt:11 (project)

Please let me know if its neccisary for me to post the logs. Thanks!

[psakamoori]

Below setting - fixed the above error for me...

export CUDACXX="/usr/local/cuda-11.0/bin/nvcc"

Looks its the same your trying with cmake args...But, just give it a try...

[reynoldscem]

@psakamoori the same fix worked for me, export CUDACXX=/usr/local/cuda-11.7/bin/nvcc, in my case.