nccl-tests: init at 2.13.6

{ lib

, stdenv

, fetchFromGitHub

, which

, cudaPackages

, mpiSupport ? false

, mpi

}:

assert mpiSupport -> mpi != null;

with cudaPackages;

cudaPackages.backendStdenv.mkDerivation rec {

  pname = "nccl-tests";

  version = "2.13.6";

  src = fetchFromGitHub {

    owner = "NVIDIA";

    repo = pname;

    rev = "v${version}";

    hash = "sha256-3gSBQ0g6mnQ/MFXGflE+BqqrIUoiBgp8+fWRQOvLVkw=";

  };

  nativeBuildInputs = [ which cuda_nvcc ];

  buildInputs = [

    cuda_cudart

    nccl

  ] ++ lib.optional mpiSupport mpi;

  makeFlags = [

    "CUDA_HOME=${cudatoolkit}"

    "NCCL_HOME=${nccl}"

  ] ++ lib.optionals mpiSupport [

    "MPI=1"

  ];

  enableParallelBuilding = true;

  installPhase = ''

    mkdir -p $out/bin

    cp -r build/* $out/bin/

  '';

  meta = with lib; {

    description = "Tests to check both the performance and the correctness of NVIDIA NCCL operations.";

    homepage = "https://github.com/NVIDIA/nccl-tests";

    platforms = [ "x86_64-linux" ];

    license = licenses.bsd3;

    maintainers = with maintainers; [ jmillerpdt ];

  };

}

    nccl = final.callPackage ../development/libraries/science/math/nccl { };

    nccl-tests = final.callPackage ../development/libraries/science/math/nccl/tests.nix { };

    autoAddOpenGLRunpathHook = final.callPackage ( { makeSetupHook, addOpenGLRunpath }:

      makeSetupHook {

        name = "auto-add-opengl-runpath-hook";