.travis.yml

language: c
sudo: true
dist: xenial

env:
  global:
    - PATH=/usr/local/cuda/bin:${PATH}
    - LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH}

addons:
 apt:
   packages:
     - openmpi-bin
     - openmpi-common
     - libopenmpi-dev
     - gcc
     - automake
     - autoconf
     - libtool
     - flex

before_install:

 # Install CUDA
 - wget https://developer.nvidia.com/compute/cuda/10.0/Prod/local_installers/cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64
 - sudo dpkg -i cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64
 - sudo apt-key add /var/cuda-repo-10-0-local-10.0.130-410.48/7fa2af80.pub
 - sudo apt-get -y update
 - rm -f cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64
 - sudo apt-get install -y cuda

install:
 # Install libfabric
 - git clone https://github.com/ofiwg/libfabric.git
 - pushd libfabric
 - ./autogen.sh
 - ./configure --enable-debug --disable-psm3; make -j8; sudo make install
 - popd

 # Install NCCL
 - git clone https://github.com/NVIDIA/nccl.git
 - pushd nccl
 - make src.build NVCC_GENCODE="-gencode=arch=compute_70,code=sm_70 -gencode=arch=compute_60,code=sm_60"
 - popd

before_script: sudo ldconfig

script:
 # Build aws-ofi-nccl plugin
 - ./autogen.sh
 - ./configure --with-nccl=${PWD}/nccl/build --with-cuda=/usr/local/cuda
 - make && sudo make install