파일
Sajina PK 15c82d6da8 [rocprofiler-system]: Enable UCX Communication API tracing (#2306)
## 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.
2026-01-20 13:16:43 -05:00

68 라인
2.6 KiB
Docker

ARG DISTRO
ARG VERSION
FROM ${DISTRO}:${VERSION}
ENV HOME=/root
ENV LANG=en_US.UTF-8
ENV LANGUAGE=en_US
ENV LC_ALL=C
ENV SHELL=/bin/bash
ENV BASH_ENV=/etc/bash.bashrc
ENV DEBIAN_FRONTEND=noninteractive
WORKDIR /tmp
SHELL [ "/bin/bash", "-c" ]
ARG EXTRA_PACKAGES=""
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
ENV PATH=${HOME}/.local/bin:${PATH}
RUN apt-get update && \
apt-get dist-upgrade -y && \
apt-get install -y apt-utils autoconf autotools-dev bash-completion bison \
build-essential chrpath cmake curl flex gettext git-core gnupg2 iproute2 \
libgmock-dev libgtest-dev libnuma1 libopenmpi-dev libpapi-dev libpfm4-dev \
librpm-dev libsqlite3-dev libtool libudev1 lsb-release m4 ninja-build \
nlohmann-json3-dev python3-pip rpm texinfo wget \
libucx-dev ucx-utils && \
OS_VERSION=$(grep '^VERSION_ID=' /etc/os-release | cut -d'=' -f2 | tr -d '"') && \
OS_ID=$(grep '^ID=' /etc/os-release | cut -d'=' -f2 | tr -d '"') && \
if [ "${OS_ID}" == "ubuntu" ] && [ "${OS_VERSION}" == "22.04" ]; then \
python3 -m pip install 'cmake==3.21'; \
else \
python3 -m pip install --break-system-packages 'cmake==3.21' perfetto; \
fi
RUN if [ "${ROCM_MAJOR}" != "0" ] || [ "${ROCM_MINOR}" != "0" ]; then \
OS_VERSION=$(grep '^VERSION_ID=' /etc/os-release | cut -d'=' -f2 | tr -d '"') && \
OS_CODENAME=$(grep '^VERSION_CODENAME=' /etc/os-release | cut -d'=' -f2) && \
AMDGPU_DEB="amdgpu-install_${ROCM_MAJOR}.${ROCM_MINOR}.${ROCM_VERSN}-1_all.deb" && \
wget https://repo.radeon.com/amdgpu-install/${ROCM_VERSION_URL}/ubuntu/${OS_CODENAME}/${AMDGPU_DEB} && \
apt-get install -y ./${AMDGPU_DEB} && \
apt-get update && \
apt-get install -y rocm-dev rccl-dev libpciaccess0 ${EXTRA_PACKAGES} && \
apt-get autoclean; \
fi
ARG PYTHON_VERSIONS="6 7 8 9 10 11 12 13"
RUN wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh -O miniforge.sh && \
bash miniforge.sh -b -p /opt/conda && \
export PATH="/opt/conda/bin:${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
ENV LC_ALL=C.UTF-8
WORKDIR /home
SHELL [ "/bin/bash", "--login", "-c" ]