Skip to content

Always propagate completion errors to NCCL #4

Always propagate completion errors to NCCL

Always propagate completion errors to NCCL #4

Workflow file for this run

name: PR CI
on: [push, pull_request]
env:
APT_PACKAGES: >-
build-essential
clang
gcc
git
libhwloc-dev
make
jobs:
distcheck:
runs-on: ubuntu-22.04
strategy:
matrix:
cc:
- gcc
- clang
fail-fast: false
steps:
- name: Install Dependencies
run: |
sudo apt-get install -y ${{ env.APT_PACKAGES }}
- name: Install CUDA
run: |
sudo apt-get install -y nvidia-cuda-toolkit
- uses: actions/checkout@v4
- name: Build and Distribution Checks
run: |
set -x
# We're just doing distchecks, so it is fine if we
# just grab the latest master.
git clone --depth 1 https://github.com/ofiwg/libfabric.git
pushd libfabric
./autogen.sh
./configure --prefix=$PWD/install CC=${{ matrix.cc }}
make -j $(nproc)
make install
popd
# actions/checkout@v4 would drop the plugin source in $PWD,
# so go ahead and build it
./autogen.sh
./configure --with-libfabric=$PWD/libfabric/install --with-cuda=/usr/local/cuda/ CC=${{ matrix.cc }}
make -j $(nproc)
# Run Unit tests
make check
# Run dist tarball checks
make distcheck
- name: Upload build logs
if: failure()
uses: actions/upload-artifact@v4
with:
name: ${{ matrix.cc }}-config.log
path: config.log