Failing to install on Google Colab

[wolfram77]

Hello, I am trying to install nvGraph on Google Colab. It was working fine yesterday, but Colab is now upgraded to CUDA 11.0 from 10.1 and the ./build.sh command is failing with error message from thrust. Similar errors are reported from a number of other source files.

In file included from /usr/local/cuda/targets/x86_64-linux/include/thrust/detail/config/config.h:27:0,
                 from /usr/local/cuda/targets/x86_64-linux/include/thrust/detail/config.h:23,
                 from /usr/local/cuda/targets/x86_64-linux/include/thrust/sequence.h:24,
                 from /root/nvgraph/cpp/src/triangles_counting.cpp:18:
/usr/local/cuda/targets/x86_64-linux/include/thrust/detail/config/cpp_dialect.h:104:13: error: Thrust requires C++14. Please pass -std=c++14 to your compiler. Define THRUST_IGNORE_DEPRECATED_CPP_DIALECT to suppress this message. [-Werror]
   THRUST_COMPILER_DEPRECATION(C++14, pass -std=c++14 to your compiler);
             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~                                                                            


In file included from /usr/local/cuda/targets/x86_64-linux/include/thrust/system/cuda/detail/execution_policy.h:33:0,
                 from /usr/local/cuda/targets/x86_64-linux/include/thrust/iterator/detail/device_system_tag.h:23,
                 from /usr/local/cuda/targets/x86_64-linux/include/thrust/iterator/iterator_traits.h:111,
                 from /usr/local/cuda/targets/x86_64-linux/include/thrust/detail/sequence.inl:24,
                 from /usr/local/cuda/targets/x86_64-linux/include/thrust/sequence.h:295,
                 from /root/nvgraph/cpp/src/triangles_counting.cpp:18:
/usr/local/cuda/targets/x86_64-linux/include/thrust/system/cuda/config.h:78:2: error: #error The version of CUB in your include path is not compatible with this release of Thrust. CUB is now included in the CUDA Toolkit, so you no longer need to use your own checkout of CUB. Define THRUST_IGNORE_CUB_VERSION_CHECK to ignore this.
 #error The version of CUB in your include path is not compatible with this release of Thrust. CUB is now included in the CUDA Toolkit, so you no longer need to use your own checkout of CUB. Define THRUST_IGNORE_CUB_VERSION_CHECK to ignore this.
  ^~~~~

That stops cusparse build and then the identifiers are listed as undefined:

...
/root/nvgraph/cpp/include/matrix.hxx(772): error: identifier "cusparseScsrsm_solve" is undefined
/root/nvgraph/cpp/include/matrix.hxx(775): error: identifier "cusparseSolveAnalysisInfo_t" is undefined
/root/nvgraph/cpp/include/matrix.hxx(776): error: identifier "cusparseDcsrsm_solve" is undefined

[dmikushin]

nvGraph was designed to work with CUDA 9. In recent versions of CUDA (e.g. CUDA 11) NVIDIA broke the compatibility of cusparse API. As I understood, they wanted some kind of unification with other math libraries. But the way they did it looks pretty much like a stupid management decision. They don't give a thing about breaking hundreds of very useful computing softwares, such as PetSC. Why not to make a compatibility layer instead, NVIDIA? Sure, we have a lot of time to catch up with your changes, because you make a generous free donation, to us the poor people.

[wolfram77]

🤣 I was able to install it from anaconda though. https://anaconda.org/nvidia/nvgraph

[dmikushin]

Yes, because there at Anaconda they still use old CUDA runtimes for nvgraph.