Merge pull request #786 from rapidsai/branch-0.22

[RELEASE] ucx-py v0.22

ci/gpu/build.sh

@@ -26,7 +26,7 @@ export HOME=$WORKSPACE
 cd $WORKSPACE
 export GIT_DESCRIBE_TAG=`git describe --tags`
 export MINOR_VERSION=`echo $GIT_DESCRIBE_TAG | grep -o -E '([0-9]+\.[0-9]+)'`
-export RAPIDS_VERSION="21.08"
+export RAPIDS_VERSION="21.10"
 export UCX_PATH=$CONDA_PREFIX
 
 ################################################################################

docker/Dockerfile

@@ -0,0 +1,24 @@
+FROM python:3
+
+RUN apt-get update && \
+    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends tzdata && \
+    apt-get install -y \
+        automake \
+        dh-make \
+        g++ \
+        git \
+        libcap2 \
+        libhwloc-dev \
+        libnuma-dev \
+        libtool \
+        make \
+        udev \
+        wget \
+    && apt-get remove -y openjdk-11-* || apt-get autoremove -y \
+    && apt-get clean && rm -rf /var/lib/apt/lists/*
+
+COPY run.sh /root
+
+WORKDIR /root
+
+CMD [ "/root/run.sh" ]

docker/README.md

@@ -0,0 +1,24 @@
+# Docker container
+
+## Summary
+
+Contains reference dockerfile and build script to run UCX-Py tests and benchmarks. This is a minimal setup, without support for CUDA, MOFED or rdma-core.
+
+## Building Docker image
+
+To begin, it's necessary to build the image, this is done as follows:
+
+```bash
+cd docker
+docker build -t ucx-py -f Dockerfile .
+```
+
+## Running
+
+Once building the Docker image is complete, the container can be started with the following command:
+
+```bash
+docker run ucx-py
+```
+
+The container above will run UCX-Py tests and benchmarks.

docker/run.sh

@@ -0,0 +1,70 @@
+#!/bin/bash
+# Copyright (c) 2021, NVIDIA CORPORATION.
+set -e
+
+function logger {
+    echo -e "\n$@\n"
+}
+
+PYTHON_PREFIX=$(python -c "import distutils.sysconfig; print(distutils.sysconfig.PREFIX)")
+
+################################################################################
+# SETUP - Install python packages and check environment
+################################################################################
+
+pip install \
+    "pytest" "pytest-asyncio" \
+    "dask" "distributed" \
+	"cython"
+
+logger "Check versions"
+python --version
+pip list
+
+################################################################################
+# BUILD - Build UCX master, UCX-Py and run tests
+################################################################################
+logger "Build UCX master"
+cd $HOME
+git clone https://github.com/openucx/ucx
+cd ucx
+./autogen.sh
+./contrib/configure-devel \
+    --prefix=$PYTHON_PREFIX \
+    --enable-gtest=no \
+    --with-valgrind=no
+make -j install
+
+echo $PYTHON_PREFIX >> /etc/ld.so.conf.d/python.conf
+ldconfig
+
+logger "UCX Version and Build Information"
+ucx_info -v
+
+
+################################################################################
+# TEST - Run pytests for ucx-py
+################################################################################
+logger "Clone and Build UCX-Py"
+cd $HOME
+git clone https://github.com/rapidsai/ucx-py
+cd ucx-py
+python setup.py build_ext --inplace
+python -m pip install -e .
+
+for tls in "tcp" "all"; do
+    export UCX_TLS=$tls
+
+    logger "Python pytest for ucx-py"
+
+    # Test with TCP/Sockets
+    logger "Tests (UCX_TLS=$UCX_TLS)"
+    pytest --cache-clear -vs ucp/_libs/tests
+    pytest --cache-clear -vs tests/
+
+    logger "Benchmarks (UCX_TLS=$UCX_TLS)"
+    python benchmarks/send-recv.py -o numpy \
+        --server-dev 0 --client-dev 0 --reuse-alloc
+    python benchmarks/send-recv-core.py -o numpy \
+        --server-dev 0 --client-dev 0 --reuse-alloc
+done

docs/source/install.rst

@@ -55,10 +55,12 @@ Build Dependencies
 
     conda create -n ucx -c conda-forge \
         automake make libtool pkg-config \
-        libhwloc psutil \
+        psutil \
         "python=3.7" setuptools "cython>=0.29.14,<3.0.0a0"
 
 
+If you are using UCX 1.9 and older and using both CUDA and InfiniBand support, ensure ``libhwloc`` is also on the list above.
+
 Test Dependencies
 ~~~~~~~~~~~~~~~~~
 
@@ -70,17 +72,17 @@ Test Dependencies
         distributed
 
 
-UCX-1.11 (Development)
-~~~~~~~~~~~~~~~~~~~~~~
+UCX-1.11.1
+~~~~~~~~~~
 
-Instructions for building UCX 1.11 (current development version):
+Instructions for building UCX 1.11.1:
 
 ::
 
     conda activate ucx
     git clone https://github.com/openucx/ucx
     cd ucx
-    git checkout v1.11.x
+    git checkout v1.11.1
     ./autogen.sh
     mkdir build
     cd build
@@ -153,7 +155,12 @@ UCX-Py
     conda activate ucx
     git clone https://github.com/rapidsai/ucx-py.git
     cd ucx-py
-    python setup.py build_ext --inplace
-    pip install .
+    pip install -v .
     # or for develop build
     pip install -v -e .
+
+In UCX 1.10 and above, or for builds that don't need CUDA and InfiniBand support, users can disable building with hwloc support:
+
+::
+
+    UCXPY_DISABLE_HWLOC=1 pip install -v .

setup.py

@@ -21,10 +21,33 @@
 
 include_dirs = [os.path.dirname(get_python_inc())]
 library_dirs = [get_config_var("LIBDIR")]
-libraries = ["ucp", "uct", "ucm", "ucs", "hwloc"]
+libraries = ["ucp", "uct", "ucm", "ucs"]
 extra_compile_args = ["-std=c99", "-Werror"]
 
 
+DISABLE_HWLOC = int(os.environ.get("UCXPY_DISABLE_HWLOC", "0"))
+topological_distance_ext = []
+if DISABLE_HWLOC == 0:
+    libraries.append("hwloc")
+    topological_distance_ext = [
+        Extension(
+            "ucp._libs.topological_distance",
+            sources=[
+                "ucp/_libs/topological_distance.pyx",
+                "ucp/_libs/src/topological_distance.c",
+            ],
+            depends=[
+                "ucp/_libs/src/topological_distance.h",
+                "ucp/_libs/topological_distance_dep.pxd",
+            ],
+            include_dirs=include_dirs,
+            library_dirs=library_dirs,
+            libraries=libraries,
+            extra_compile_args=extra_compile_args,
+        ),
+    ]
+
+
 def get_ucp_version():
     with open(include_dirs[0] + "/ucp/api/ucp_version.h") as f:
         ftext = f.read()
@@ -59,22 +82,8 @@ def get_ucp_version():
             libraries=libraries,
             extra_compile_args=extra_compile_args,
         ),
-        Extension(
-            "ucp._libs.topological_distance",
-            sources=[
-                "ucp/_libs/topological_distance.pyx",
-                "ucp/_libs/src/topological_distance.c",
-            ],
-            depends=[
-                "ucp/_libs/src/topological_distance.h",
-                "ucp/_libs/topological_distance_dep.pxd",
-            ],
-            include_dirs=include_dirs,
-            library_dirs=library_dirs,
-            libraries=libraries,
-            extra_compile_args=extra_compile_args,
-        ),
-    ],
+    ]
+    + topological_distance_ext,
     compile_time_env={"CY_UCP_AM_SUPPORTED": _am_supported},
 )