TheRock/rocm-systems/projects/rocshmem/docker/Dockerfile.ubuntu

FROM rocm/dev-ubuntu-24.04:latest
# You can also pick a specific rocm version, Ubuntu version
#FROM rocm/dev-ubuntu-24.04:7.2.1
#FROM rocm/dev-ubuntu-22.04:7.2.1
# If you plan on testing VLLM/DeepEP this is a good base
#FROM rocm/pytorch:rocm7.2_ubuntu24.04_py3.12_pytorch_release_2.9.1

LABEL org.opencontainers.image.title="ROCm rocSHMEM Container"
LABEL org.opencontainers.image.description="Container for rocSHMEM development with MPI support"
LABEL org.opencontainers.image.authors="rocshmem-maintainers@amd.com"
LABEL org.opencontainers.image.documentation="https://rocm.docs.amd.com/projects/rocSHMEM/en/latest/docker"
LABEL org.opencontainers.image.source="https://github.com/ROCm/rocm-systems/projects/rocshmem"

WORKDIR /app
COPY README.md help perf-compare /app/
ENV PATH="${PATH}:/app/"

# Enable CRB and install system dependencies
RUN apt-get update \
 && apt-get install -y software-properties-common \
 && apt-get update \
 && apt-get install -y \
      build-essential autoconf libtool flex bison pkg-config cmake git wget curl ca-certificates \
      libibverbs-dev ibverbs-utils libibumad-dev libpci-dev librdmacm-dev rdma-core infiniband-diags \
      python3-venv \
# OPTIONAL: required to use mpirun launcher
      openssh-server \
# OPTIONAL: required to build VLLM/DeepEP
#      libsqlite3-dev libfmt-dev libmsgpack-dev libsuitesparse-dev \
#      libboost-all-dev \
#      libgrpc++-dev protobuf-compiler-grpc protobuf-compiler libprotobuf-dev \
#      libaio-dev liburing-dev pybind11-dev ninja-build libgflags-dev \
# OPTIONAL: devel tooling to modify code and tweak inside the image
#      vim cmake-curses-gui \
# Cleanup
 && apt-get autoremove -y \
 && apt-get clean && rm -rf /var/lib/apt/lists/* \
# Optional: remove sccache if present
 && (rm -f "$(which sccache)" 2>/dev/null || true) \
 && (python3 -m pip uninstall -y sccache || true)
# Install Python dependencies for VLLM/DeepEP
#RUN pip3 install --upgrade pip && \
#    pip3 install "pybind11[global]" meson==0.64.0 && \
#    pip install setuptools==75.1.0 &&

# OPTIONAL: Setup ssh (only needed to use mpirun launch method, e.g., rocshmem functional tests)
# The following block is OPTIONAL and can be commented out if you only want to run 1-node testing
# Note: all deployed containers must share the SAME KEY!
#   Same docker image build must be deployed on all nodes (docker save, docker load)
ARG SSH_PORT=2222
RUN ssh-keygen -t ed25519 -f ~/.ssh/id_ed25519
# Alternative: comment line above, uncomment lines below:
#   import an invariant ssh-key from outside the docker (so you can build the docker on each nodes)
#COPY id_ed25519 /root/.ssh/id_ed25519
#COPY id_ed25519.pub /root/.ssh/id_ed25519.pub
RUN cat ~/.ssh/id_ed25519.pub >~/.ssh/authorized_keys && \
    sed -i "s/#Port 22/Port $SSH_PORT/" /etc/ssh/sshd_config && \
    mkdir -p /run/sshd && ssh-keygen -A && \
    echo " StrictHostKeyChecking=accept-new" >>/root/.ssh/config
CMD ["sh", "-c", "\
        set -e; \
        src/tools/rocshmem_info \
     && echo \"Starting SSH server for mpiexec with OMPI_MCA_plm_rsh_arg=$OMPI_MCA_plm_rsh_args\" \
     && exec /usr/sbin/sshd -D"]
# End of ssh setup

# OPTIONAL: Install rocm-enabled IB perftest (used to validate if network config is sane)
# The following block is OPTIONAL and can be commented out if you only want to run rocSHMEM
WORKDIR /app
RUN git clone https://github.com/ROCm/rdma-perftest
WORKDIR /app/rdma-perftest/
RUN ./autogen.sh \
 && ./configure --enable-rocm --with-rocm=/opt/rocm \
 && make -j 8 #install

# Clone rocSHMEM
WORKDIR /app
RUN git clone --no-checkout --filter=blob:none https://github.com/ROCm/rocm-systems.git && cd rocm-systems \
 && git sparse-checkout set --cone projects/rocshmem \
 && git checkout develop

# Install dependencies
# OPTIONAL: install UCX/MPI to test RO (MPI with host-proxy) conduit
RUN INSTALL_DIR=/app/ompi bash /app/rocm-systems/projects/rocshmem/scripts/install_dependencies.sh
ENV PATH="/app/ompi/bin:${PATH}"

# Build rocSHMEM develop
WORKDIR /app/rocm-systems/projects/rocshmem
RUN mkdir -p /app/build
WORKDIR /app/build
RUN /app/rocm-systems/projects/rocshmem/scripts/build_configs/all_backends \
      -DUSE_EXTERNAL_MPI=OFF \
      -DMPI_ROOT=/app/ompi

# OPTIONAL: Build a rocSHMEM PR for easy comparison with develop
ARG PR_NUM
WORKDIR /app/rocm-systems/projects/rocshmem
RUN test -n "$PR_NUM" \
 && git fetch origin pull/$PR_NUM/head \
 && git checkout FETCH_HEAD \
#&& git rebase develop \
 || test -z "$PR_NUM"
RUN test -n "$PR_NUM" \
 && mkdir -p /app/build$PR_NUM && cd /app/build$PR_NUM \
 && /app/rocm-systems/projects/rocshmem/scripts/build_configs/all_backends \
      -DUSE_EXTERNAL_MPI=OFF \
      -DMPI_ROOT=/app/ompi \
 || test -z "$PR_NUM"
# WORKDIR defaults to the PR build, or develop build if PR not built
WORKDIR /app/build${PR_NUM}

# Add environment for running, tune as needed
ENV ROCSHMEM_TEST_UUID=1 ROCSHMEM_HEAP_SIZE=6442450944
# FOR One node GDA testing
#ENV ROCSHMEM_DISABLE_MIXED_IPC=1
ENV OMPI_MCA_plm_rsh_args=-p$SSH_PORT \
    OMPI_ALLOW_RUN_AS_ROOT=1 OMPI_ALLOW_RUN_AS_ROOT_CONFIRM=1
# System dependent: you may need to select a different PML on some network types, note RO should not be used in this configuration
#ENV OMPI_MCA_pml=ob1 OMPI_MCA_osc=rdma
# System dependent: you may need to force the bootstrap TCP interface if autodetection picks a non-routable iface
#ENV OMPI_MCA_oob_tcp_if_include=
#ENV ROCSHMEM_BOOTSTRAP_SOCKET_IFNAME=
# System dependent: Ignore service/backend RoCE ifaces
#ENV ROCSHMEM_HCA_LIST=^mlx5_6