Updated Containerfile for OSU+OpenMPI with explicit git refs

eth-cscs · Feb 14, 2025 · e4d0b51 · e4d0b51
1 parent 58346bf
commit e4d0b51
Show file tree

Hide file tree

Showing 2 changed files with 23 additions and 17 deletions.
diff --git a/examples/container/osu-ompi/Containerfile b/examples/container/osu-ompi/Containerfile
@@ -1,7 +1,4 @@
-FROM docker.io/nvidia/cuda:12.6.3-devel-ubuntu24.04
-
-ARG OMPI_VER=5.0.6
-ARG GDRCOPY_VER=2.3
+FROM docker.io/nvidia/cuda:12.8.0-devel-ubuntu24.04
 
 RUN apt-get update \
     && DEBIAN_FRONTEND=noninteractive \
@@ -33,29 +30,35 @@ RUN apt-get update \
         --no-install-recommends \
     && rm -rf /var/lib/apt/lists/*
 
-RUN git clone --depth 1 -b v${GDRCOPY_VER} https://github.com/NVIDIA/gdrcopy.git && \
-    cd gdrcopy && \
-    export CUDA_PATH=${CUDA_HOME:-$(echo $(which nvcc) | grep -o '.*cuda')} && \
-    make CC=gcc CUDA=$CUDA_PATH lib && \
-    make lib_install && \
-    cd ../ && rm -rf gdrcopy
+ARG GDRCOPY_VER=2.4.4
+RUN git clone --depth 1 --branch v${GDRCOPY_VER} https://github.com/NVIDIA/gdrcopy.git \
+    && cd gdrcopy \
+    && export CUDA_PATH=${CUDA_HOME:-$(echo $(which nvcc) | grep -o '.*cuda')} \
+    && make CC=gcc CUDA=$CUDA_PATH lib \
+    && make lib_install \
+    && cd ../ && rm -rf gdrcopy
 
-RUN git clone --depth 1 https://github.com/HewlettPackard/shs-cassini-headers.git \
+ARG cassini_headers_ref=9a8a738a879f007849fbc69be8e3487a4abf0952
+RUN git clone https://github.com/HewlettPackard/shs-cassini-headers.git \
     && cd shs-cassini-headers \
+    && git checkout ${cassini_headers_ref} \
     && cp -r include/* /usr/include/ \
     && cp -r share/* /usr/share/ \
     && cd .. \
     && rm -r shs-cassini-headers
 
-RUN git clone --depth 1 https://github.com/HewlettPackard/shs-cxi-driver.git \
+ARG cxi_driver_ref=caa8bf41a25817111f137bb7e8be1e45c4e6758f
+RUN git clone https://github.com/HewlettPackard/shs-cxi-driver.git \
     && cd shs-cxi-driver \
+    && git checkout ${cxi_ddriver_ref} \
     && cp -r include/* /usr/include/ \
     && cd .. \
     && rm -r shs-cxi-driver
 
-RUN git clone --depth 1 https://github.com/HewlettPackard/shs-libcxi.git \
+ARG libcxi_ref=31a183b521d1da670574e2a1bf59a91cb579b105
+RUN git clone https://github.com/HewlettPackard/shs-libcxi.git \
     && cd shs-libcxi \
-    && git checkout release/shs-12.0 \
+    && git checkout ${libcxi_ref} \
     && ./autogen.sh \
     && ./configure --prefix=/usr --with-cuda=/usr/local/cuda \
     && make -j$(nproc) \
@@ -65,10 +68,12 @@ RUN git clone --depth 1 https://github.com/HewlettPackard/shs-libcxi.git \
     && rm -r shs-libcxi
 
 # Install libfabric
+ARG libfabric_version=2.0.0
+ARG libfabric_ref=36b974dc27c67fd0268e0ab575fd01592f756992
 # Libfabric 2.0.0 seems to have a bug preventing compilation of the CXI provider, using main branch for now
-ARG libfabric_version=main
-RUN git clone --depth 1 --branch ${libfabric_version} https://github.com/ofiwg/libfabric.git \
+RUN git clone --branch main --single-branch https://github.com/ofiwg/libfabric.git \
     && cd libfabric \
+    && git checkout ${libfabric_ref} \
     && ./autogen.sh \
     && ./configure --prefix=/usr --with-cuda=/usr/local/cuda --enable-cuda-dlopen --enable-gdrcopy-dlopen --enable-cxi --enable-lnx --enable-efa \
     && make -j$(nproc) \
@@ -77,6 +82,7 @@ RUN git clone --depth 1 --branch ${libfabric_version} https://github.com/ofiwg/l
     && cd .. \
     && rm -rf libfabric
 
+ARG OMPI_VER=5.0.6
 RUN wget -q  https://download.open-mpi.org/release/open-mpi/v5.0/openmpi-${OMPI_VER}.tar.gz \
     && tar xf openmpi-${OMPI_VER}.tar.gz \
     && cd openmpi-${OMPI_VER} \

diff --git a/examples/container/osu-ompi/omb-cxi-ompi.toml b/examples/container/osu-ompi/omb-cxi-ompi.toml
@@ -1,4 +1,4 @@
-image="/capstor/scratch/cscs/amadonna/localhost+omb-cxi.sqsh"
+image="/capstor/scratch/cscs/amadonna/localhost+omb-cxi-ompi.sqsh"
 writable=true
 entrypoint=false