Sajina PK
15c82d6da8
## Motivation Enable UCX communication tracing and communication metadata ## Technical Details Implement UCX API wrappers to trace transport-layer communication. This adds communication data tracking and exposes “UCX Comm Send/Recv” timelines, enabling detailed analysis of MPI, OpenSHMEM, and other UCX-based runtime communication patterns. - Implements function interception for UCX functions across multiple categories using gotcha component. - Extended comm_data component to track UCX send/recv operations - Added ucx_send and ucx_recv labels for Perfetto counter tracks. Integrated UCX data tracking with existing MPI/RCCL tracking infrastructure. - Added ROCPROFSYS_USE_UCX configuration option (enabled by default). - Created FindUCX.cmake module for UCX header detection. Falls back to internal UCX headers if system headers not found. - Updated all Dockerfiles to include UCX dependencies.
62 wiersze
2.5 KiB
Docker
62 wiersze
2.5 KiB
Docker
ARG DISTRO=rockylinux/rockylinux
|
|
ARG VERSION=8
|
|
FROM ${DISTRO}:${VERSION}
|
|
|
|
ENV HOME /root
|
|
ENV SHELL /bin/bash
|
|
ENV BASH_ENV /etc/bash.bashrc
|
|
ENV DEBIAN_FRONTEND noninteractive
|
|
|
|
WORKDIR /tmp
|
|
SHELL [ "/bin/bash", "-c" ]
|
|
|
|
ENV PATH /usr/lib64/openmpi/bin:/usr/local/bin:${PATH}
|
|
ENV LIBRARY_PATH ${LIBRARY_PATH}:/opt/amdgpu/lib64
|
|
|
|
RUN yum groupinstall -y "Development Tools" && \
|
|
yum install -y epel-release && crb enable && \
|
|
yum install -y --allowerasing chrpath cmake curl dpkg-devel gmock-devel gtest-devel \
|
|
iproute json-devel libdrm-devel ninja-build numactl-devel openmpi-devel \
|
|
papi-devel python3-pip sqlite-devel texinfo wget which zlib-devel ucx-devel && \
|
|
yum clean all && \
|
|
python3 -m pip install 'cmake==3.21' && \
|
|
python3 -m pip install 'perfetto'
|
|
|
|
ARG ROCM_VERSION=0.0
|
|
ARG ROCM_MAJOR=0
|
|
ARG ROCM_MINOR=0
|
|
ARG ROCM_PATCH=0
|
|
ARG ROCM_VERSION_URL=0.0
|
|
ARG ROCM_VERSN=0
|
|
|
|
RUN if [ "${ROCM_MAJOR}" != "0" ] || [ "${ROCM_MINOR}" != "0" ]; then \
|
|
OS_VERSION=$(grep '^VERSION_ID=' /etc/os-release | cut -d'=' -f2 | tr -d '"') && \
|
|
OS_VERSION_MAJOR=$(echo "$OS_VERSION" | cut -d'.' -f1) && \
|
|
RPM_TAG=".el${OS_VERSION_MAJOR}" && \
|
|
if [ "${OS_VERSION_MAJOR}" -eq 8 ]; then PERL_REPO=powertools; else PERL_REPO=crb; fi && \
|
|
dnf -y --enablerepo=${PERL_REPO} install perl-File-BaseDir && \
|
|
yum install -y https://repo.radeon.com/amdgpu-install/${ROCM_VERSION_URL}/rhel/${OS_VERSION}/amdgpu-install-${ROCM_MAJOR}.${ROCM_MINOR}.${ROCM_VERSN}-1${RPM_TAG}.noarch.rpm && \
|
|
yum install -y rocm-dev && \
|
|
yum clean all; \
|
|
fi
|
|
|
|
ARG PYTHON_VERSIONS="6 7 8 9 10 11 12 13"
|
|
|
|
RUN wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh -O miniforge.sh && \
|
|
bash miniforge.sh -b -p /opt/conda && \
|
|
export PATH="/opt/conda/bin:${PATH}" && \
|
|
conda config --set always_yes yes --set changeps1 no --set solver classic && \
|
|
conda update -c conda-forge -n base conda && \
|
|
for i in ${PYTHON_VERSIONS}; do conda create -n py3.${i} -c conda-forge python=3.${i} pip; done && \
|
|
for i in ${PYTHON_VERSIONS}; do /opt/conda/envs/py3.${i}/bin/python -m pip install numpy perfetto dataclasses; done && \
|
|
conda clean -a -y && \
|
|
conda init
|
|
|
|
RUN if [ "${ROCM_VERSION}" != "0.0" ]; then ln -sf /opt/rocm-${ROCM_VERSION}* /opt/rocm; fi
|
|
|
|
WORKDIR /home
|
|
ENV LC_ALL C.UTF-8
|
|
SHELL [ "/bin/bash", "--login", "-c" ]
|
|
COPY ./entrypoint-rhel.sh /docker-entrypoint.sh
|
|
ENTRYPOINT [ "/docker-entrypoint.sh" ]
|