Add NVIDIA sm70 (V100) environment and CI workflow

.github/workflows/build-nvidia.yml

@@ -0,0 +1,96 @@
+name: Build NVIDIA Docker images
+
+on:
+  push:
+    branches:
+      - main
+    paths:
+      - envs/x86/sm70/**
+      - envs/x86/sm100/**
+      - .github/workflows/build-nvidia.yml
+  pull_request:
+    paths:
+      - envs/x86/sm70/**
+      - envs/x86/sm100/**
+      - .github/workflows/build-nvidia.yml
+  workflow_dispatch:
+
+env:
+  REGISTRY: docker.io
+  IMAGE_NAME: higherordermethods/selfish
+
+jobs:
+  build:
+    name: Build ${{ matrix.gpu_arch }} image
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - gpu_arch: sm70
+            cuda_version: "12.4"
+          - gpu_arch: sm100
+            cuda_version: "13.0"
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Log in to Docker Hub
+        if: github.event_name != 'pull_request'
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Set up Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Generate image metadata
+        id: meta
+        run: |
+          # Convert CUDA version 12.4 -> cuda124
+          VERSION_NO_DOTS=$(echo "${{ matrix.cuda_version }}" | tr -d '.')
+          GPU_BACKEND="cuda${VERSION_NO_DOTS}"
+
+          CPU_PLATFORM="x86"
+          GPU_ARCH="${{ matrix.gpu_arch }}"
+
+          # Tags: <version>-<cpu_platform>-<gpu_backend>-<gpu_arch>
+          echo "tags<<EOF" >> $GITHUB_OUTPUT
+          echo "${{ env.IMAGE_NAME }}:latest-${CPU_PLATFORM}-${GPU_BACKEND}-${GPU_ARCH}" >> $GITHUB_OUTPUT
+          echo "${{ env.IMAGE_NAME }}:${{ github.sha }}-${CPU_PLATFORM}-${GPU_BACKEND}-${GPU_ARCH}" >> $GITHUB_OUTPUT
+          echo "EOF" >> $GITHUB_OUTPUT
+
+          echo "gpu_backend=${GPU_BACKEND}" >> $GITHUB_OUTPUT
+
+      - name: Cache Docker layers
+        uses: actions/cache@v4
+        with:
+          path: /tmp/.buildx-cache
+          key: ${{ runner.os }}-buildx-${{ matrix.gpu_arch }}-${{ github.sha }}
+          restore-keys: |
+            ${{ runner.os }}-buildx-${{ matrix.gpu_arch }}-
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: envs/x86/${{ matrix.gpu_arch }}/Dockerfile
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          build-args: |
+            CUDA_VERSION=${{ matrix.cuda_version }}
+          cache-from: type=local,src=/tmp/.buildx-cache
+          cache-to: type=local,dest=/tmp/.buildx-cache-new,mode=max
+          labels: |
+            com.fluidnumerics.cuda.target=${{ matrix.gpu_arch }}
+            com.fluidnumerics.cuda.version=${{ matrix.cuda_version }}
+            org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}
+            org.opencontainers.image.revision=${{ github.sha }}
+
+      - name: Move cache
+        run: |
+          rm -rf /tmp/.buildx-cache
+          mv /tmp/.buildx-cache-new /tmp/.buildx-cache

envs/x86/sm100/Dockerfile

@@ -0,0 +1,182 @@
+FROM docker.io/rockylinux:9 AS bootstrap
+
+ARG CUDA_VERSION=13.0
+
+ENV SPACK_ROOT=/opt/spack \
+    CURRENTLY_BUILDING_DOCKER_IMAGE=1 \
+    container=docker
+
+RUN dnf update -y \
+ && dnf install -y epel-release \
+ && dnf update -y \
+ && dnf --enablerepo epel install -y \
+        bzip2 \
+        cmake \
+        curl-minimal \
+        file \
+        findutils \
+        gcc-c++ \
+        gcc \
+        gcc-gfortran \
+        git \
+        gnupg2 \
+        hg \
+        hostname \
+        iproute \
+        make \
+        patch \
+        python3 \
+        python3-pip \
+        python3-setuptools \
+        svn \
+        unzip \
+        xz \
+        zstd \
+ && pip3 install boto3 \
+ && rm -rf /var/cache/dnf \
+ && dnf clean all
+
+# Install CUDA toolkit from NVIDIA repo
+RUN dnf config-manager \
+        --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo \
+ && dnf clean all \
+ && dnf update -y \
+ && dnf install -y cuda-toolkit-$(echo ${CUDA_VERSION} | tr '.' '-')
+
+RUN ls -l /usr/local/cuda-${CUDA_VERSION}/include/cuda.h
+
+RUN mkdir $SPACK_ROOT && cd $SPACK_ROOT && \
+    git init --quiet && git remote add origin https://github.com/spack/spack.git && git fetch --depth=1 origin develop && git checkout --detach FETCH_HEAD && \
+    mkdir -p $SPACK_ROOT/opt/spack
+
+RUN ln -s $SPACK_ROOT/share/spack/docker/entrypoint.bash \
+          /usr/local/bin/docker-shell \
+ && ln -s $SPACK_ROOT/share/spack/docker/entrypoint.bash \
+          /usr/local/bin/interactive-shell \
+ && ln -s $SPACK_ROOT/share/spack/docker/entrypoint.bash \
+          /usr/local/bin/spack-env
+
+RUN mkdir -p /root/.spack \
+ && cp $SPACK_ROOT/share/spack/docker/modules.yaml \
+        /root/.spack/modules.yaml \
+ && rm -rf /root/*.* /run/nologin
+
+# [WORKAROUND]
+# https://superuser.com/questions/1241548/
+#     xubuntu-16-04-ttyname-failed-inappropriate-ioctl-for-device#1253889
+RUN [ -f ~/.profile ]                                               \
+ && sed -i 's/mesg n/( tty -s \\&\\& mesg n || true )/g' ~/.profile \
+ || true
+
+
+WORKDIR /root
+SHELL ["docker-shell"]
+
+# Creates the package cache
+RUN spack bootstrap now \
+    && spack bootstrap status --optional \
+    && spack spec hdf5+mpi
+
+ENTRYPOINT ["/bin/bash", "/opt/spack/share/spack/docker/entrypoint.bash"]
+CMD ["interactive-shell"]
+
+# Build stage with Spack pre-installed and ready to be used
+FROM bootstrap AS builder
+
+
+# What we want to install and how we want to install it
+# is specified in a manifest file (spack.yaml)
+RUN mkdir -p /opt/spack-environment && \
+set -o noclobber \
+&&  (echo spack: \
+&&   echo '  specs:' \
+&&   echo '  - feq-parse@2.2.2' \
+&&   echo '  - openmpi@5.0.8 +cuda cuda_arch=100' \
+&&   echo '  - hdf5@1.14.5 +fortran +mpi' \
+&&   echo '  - cmake@3.31.11'\
+&&   echo '  packages:' \
+&&   echo '    all:' \
+&&   echo '      require:' \
+&&   echo '      - target=x86_64_v3' \
+&&   echo '      prefer:' \
+&&   echo '      - cuda_arch=100' \
+&&   echo '    cuda:' \
+&&   echo '      buildable: false' \
+&&   echo '      externals:' \
+&&   echo "        - spec: \"cuda@${CUDA_VERSION}\"" \
+&&   echo "          prefix: \"/usr/local/cuda-${CUDA_VERSION}\"" \
+&&   echo '' \
+&&   echo '  concretizer:' \
+&&   echo '    unify: true' \
+&&   echo '  config:' \
+&&   echo '    install_tree:' \
+&&   echo '      root: /opt/software' \
+&&   echo '  view: /opt/views/view') > /opt/spack-environment/spack.yaml
+
+# Apply feq-parse patch to add "c" build dependency
+COPY ./envs/x86/sm100/feq-parse.patch /tmp/feq-parse.patch
+#
+RUN SPACK_PKGS_ROOT=$(spack repo list | awk '{print $NF}') &&\
+    SPACK_BUILTIN_PKGS_ROOT=${SPACK_PKGS_ROOT/repos\/spack_repo\/builtin} &&\
+    patch -p1 -d $SPACK_BUILTIN_PKGS_ROOT < /tmp/feq-parse.patch
+
+# Install the software, remove unnecessary deps
+RUN cd /opt/spack-environment && spack env activate . && spack repo list && spack install --fail-fast && spack gc -y
+
+# Strip all the binaries
+RUN find -L /opt/views/view/* -type f -exec readlink -f '{}' \; | \
+    xargs file -i | \
+    grep 'charset=binary' | \
+    grep 'x-executable\|x-archive\|x-sharedlib' | \
+    awk -F: '{print $1}' | xargs strip
+
+# Modifications to the environment that are necessary to run
+RUN cd /opt/spack-environment && \
+    spack env activate --sh -d . > activate.sh
+
+
+# Bare OS image to run the installed executables
+FROM docker.io/rockylinux:9
+
+COPY --from=builder /opt/spack-environment /opt/spack-environment
+COPY --from=builder /opt/software /opt/software
+
+RUN dnf update -y \
+ && dnf install -y epel-release \
+ && dnf update -y \
+ && dnf --enablerepo epel install -y \
+        bzip2 \
+        cmake \
+        curl-minimal \
+        file \
+        findutils \
+        gcc-c++ \
+        gcc \
+        gcc-gfortran \
+        lcov
+
+# Install CUDA runtime libraries
+ARG CUDA_VERSION=13.0
+RUN dnf config-manager \
+        --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo \
+ && dnf clean all \
+ && dnf update -y \
+ && dnf install -y \
+        cuda-libraries-$(echo ${CUDA_VERSION} | tr '.' '-') \
+        cuda-nvtx-$(echo ${CUDA_VERSION} | tr '.' '-')
+
+# paths.view is a symlink, so copy the parent to avoid dereferencing and duplicating it
+COPY --from=builder /opt/views /opt/views
+
+RUN { \
+      echo '#!/bin/sh' \
+      && echo '.' /opt/spack-environment/activate.sh \
+      && echo 'exec "$@"'; \
+    } > /entrypoint.sh \
+&& chmod a+x /entrypoint.sh \
+&& ln -s /opt/views/view /opt/view
+
+
+LABEL "mpi"="openmpi"
+ENTRYPOINT [ "/entrypoint.sh" ]
+CMD [ "/bin/bash" ]

envs/x86/sm100/feq-parse.patch

@@ -0,0 +1,12 @@
+diff --git a/repos/spack_repo/builtin/packages/feq_parse/package.py b/repos/spack_repo/builtin/packages/feq_parse/package.py
+index e4b960b7..bc0916b9 100644
+--- a/repos/spack_repo/builtin/packages/feq_parse/package.py
++++ b/repos/spack_repo/builtin/packages/feq_parse/package.py
+@@ -29,6 +29,7 @@ class FeqParse(CMakePackage):
+     version("1.0.2", sha256="1cd1db7562908ea16fc65dc5268b654405d0b3d9dcfe11f409949c431b48a3e8")
+
+     depends_on("fortran", type="build")  # generated
++    depends_on("c", type="build")  # generated
+
+     depends_on("cmake@3.0.2:", type="build")
+

envs/x86/sm100/spack.yaml

@@ -0,0 +1,24 @@
+spack:
+  specs:
+  - feq-parse@2.2.2
+  - openmpi@5.0.8 +cuda cuda_arch=100
+  - hdf5@1.14.5 +fortran +mpi
+
+  packages:
+    all:
+      require:
+        - "target=x86_64_v3"
+      prefer:
+        - "cuda_arch=100"
+
+  container:
+    format: docker
+    images:
+      os: rockylinux:9
+      spack:
+        ref: v1.0.2
+
+    strip: true
+
+    labels:
+      mpi: openmpi