Merge remote-tracking branch 'upstream/branch-22.02' into non-owning-…

…cuda-async-mr

CHANGELOG.md

@@ -1,3 +1,7 @@
+# RMM 22.02.00 (Date TBD)
+
+Please see https://github.com/rapidsai/rmm/releases/tag/v22.02.00a for the latest changes to this development branch.
+
 # RMM 21.12.00 (Date TBD)
 
 Please see https://github.com/rapidsai/rmm/releases/tag/v21.12.00a for the latest changes to this development branch.

CMakeLists.txt

@@ -14,7 +14,7 @@
 
 cmake_minimum_required(VERSION 3.20.1 FATAL_ERROR)
 
-file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-21.12/RAPIDS.cmake
+file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-22.02/RAPIDS.cmake
      ${CMAKE_BINARY_DIR}/RAPIDS.cmake)
 include(${CMAKE_BINARY_DIR}/RAPIDS.cmake)
 
@@ -25,7 +25,7 @@ include(rapids-find)
 
 project(
   RMM
-  VERSION 21.12.00
+  VERSION 22.02.00
   LANGUAGES CXX)
 
 # Write the version header

benchmarks/CMakeLists.txt

@@ -41,6 +41,10 @@ function(ConfigureBench BENCH_NAME)
     target_compile_definitions(${BENCH_NAME} PUBLIC CUDA_API_PER_THREAD_DEFAULT_STREAM)
   endif()
 
+  if(DEFINED CUDA_MALLOC_ASYNC_SUPPORT AND NOT CUDA_MALLOC_ASYNC_SUPPORT)
+    target_compile_definitions(${TEST_NAME} PUBLIC "RMM_DISABLE_CUDA_MALLOC_ASYNC")
+  endif()
+
   target_compile_options(${BENCH_NAME} PUBLIC $<$<COMPILE_LANG_AND_ID:CXX,GNU,Clang>:-Wall -Werror
                                               -Wno-error=deprecated-declarations>)
   if(DISABLE_DEPRECATION_WARNING)

build.sh

@@ -18,8 +18,8 @@ ARGS=$*
 # script, and that this script resides in the repo dir!
 REPODIR=$(cd $(dirname $0); pwd)
 
-VALIDARGS="clean librmm rmm -v -g -n -s --ptds -h"
-HELP="$0 [clean] [librmm] [rmm] [-v] [-g] [-n] [-s] [--ptds] [-h]
+VALIDARGS="clean librmm rmm -v -g -n -s --ptds --no-cudamallocasync -h"
+HELP="$0 [clean] [librmm] [rmm] [-v] [-g] [-n] [-s] [--ptds] [--no-cudamallocasync] [--cmake-args=\"<args>\"] [-h]
    clean  - remove all existing build artifacts and configuration (start over)
    librmm - build and install the librmm C++ code
    rmm    - build and install the rmm Python package
@@ -28,6 +28,8 @@ HELP="$0 [clean] [librmm] [rmm] [-v] [-g] [-n] [-s] [--ptds] [-h]
    -n     - no install step
    -s     - statically link against cudart
    --ptds - enable per-thread default stream
+   --no-cudamallocasync  - disable CUDA malloc async support
+   --cmake-args=\\\"<args>\\\"   - pass arbitrary list of CMake configuration options (escape all quotes in argument)
    -h     - print this text
 
    default action (no args) is to build and install 'librmm' and 'rmm' targets
@@ -42,6 +44,7 @@ BUILD_TYPE=Release
 INSTALL_TARGET=install
 CUDA_STATIC_RUNTIME=OFF
 PER_THREAD_DEFAULT_STREAM=OFF
+CUDA_MALLOC_ASYNC_SUPPORT=ON
 RAN_CMAKE=0
 
 # Set defaults for vars that may not have been defined externally
@@ -54,16 +57,42 @@ function hasArg {
     (( NUMARGS != 0 )) && (echo " ${ARGS} " | grep -q " $1 ")
 }
 
+function cmakeArgs {
+    # Check for multiple cmake args options
+    if [[ $(echo $ARGS | { grep -Eo "\-\-cmake\-args" || true; } | wc -l ) -gt 1 ]]; then
+        echo "Multiple --cmake-args options were provided, please provide only one: ${ARGS}"
+        exit 1
+    fi
+
+    # Check for cmake args option
+    if [[ -n $(echo $ARGS | { grep -E "\-\-cmake\-args" || true; } ) ]]; then
+        # There are possible weird edge cases that may cause this regex filter to output nothing and fail silently
+        # the true pipe will catch any weird edge cases that may happen and will cause the program to fall back
+        # on the invalid option error
+        CMAKE_ARGS=$(echo $ARGS | { grep -Eo "\-\-cmake\-args=\".+\"" || true; })
+        if [[ -n ${CMAKE_ARGS} ]]; then
+            # Remove the full  CMAKE_ARGS argument from list of args so that it passes validArgs function
+            ARGS=${ARGS//$CMAKE_ARGS/}
+            # Filter the full argument down to just the extra string that will be added to cmake call
+            CMAKE_ARGS=$(echo $CMAKE_ARGS | grep -Eo "\".+\"" | sed -e 's/^"//' -e 's/"$//')
+        fi
+    fi
+}
+
+
 # Runs cmake if it has not been run already for build directory
 # LIBRMM_BUILD_DIR
 function ensureCMakeRan {
     mkdir -p "${LIBRMM_BUILD_DIR}"
     if (( RAN_CMAKE == 0 )); then
         echo "Executing cmake for librmm..."
-        cmake -DCMAKE_INSTALL_PREFIX="${INSTALL_PREFIX}" \
+        cmake -B "${LIBRMM_BUILD_DIR}" -S . \
+              -DCMAKE_INSTALL_PREFIX="${INSTALL_PREFIX}" \
               -DCUDA_STATIC_RUNTIME="${CUDA_STATIC_RUNTIME}" \
               -DPER_THREAD_DEFAULT_STREAM="${PER_THREAD_DEFAULT_STREAM}" \
-              -DCMAKE_BUILD_TYPE=${BUILD_TYPE} -B "${LIBRMM_BUILD_DIR}" -S .
+              -DCUDA_MALLOC_ASYNC_SUPPORT="${CUDA_MALLOC_ASYNC_SUPPORT}" \
+              -DCMAKE_BUILD_TYPE=${BUILD_TYPE} \
+              ${CMAKE_ARGS}
         RAN_CMAKE=1
     fi
 }
@@ -74,12 +103,14 @@ if hasArg -h; then
 fi
 
 # Check for valid usage
-if (( NUMARGS != 0 )); then
+if (( ${NUMARGS} != 0 )); then
+    # Check for cmake args
+    cmakeArgs
     for a in ${ARGS}; do
-	if ! (echo " ${VALIDARGS} " | grep -q " ${a} "); then
-	    echo "Invalid option: ${a}"
-	    exit 1
-	fi
+    if ! (echo " ${VALIDARGS} " | grep -q " ${a} "); then
+        echo "Invalid option or formatting, check --help: ${a}"
+        exit 1
+    fi
     done
 fi
 
@@ -100,6 +131,9 @@ fi
 if hasArg --ptds; then
     PER_THREAD_DEFAULT_STREAM=ON
 fi
+if hasArg --no-cudamallocasync; then
+    CUDA_MALLOC_ASYNC_SUPPORT=OFF
+fi
 
 # If clean given, run it prior to any other steps
 if hasArg clean; then
@@ -131,14 +165,16 @@ fi
 if (( NUMARGS == 0 )) || hasArg rmm; then
     cd "${REPODIR}/python"
     export INSTALL_PREFIX
-    if [[ ${INSTALL_TARGET} != "" ]]; then
-        echo "building rmm..."
+    echo "building rmm..."
+    if [[ ${CUDA_MALLOC_ASYNC_SUPPORT} == OFF ]]; then
+        python setup.py build_ext_no_async --inplace
+    else
         python setup.py build_ext --inplace
+    fi
+
+    if [[ ${INSTALL_TARGET} != "" ]]; then
         echo "installing rmm..."
         python setup.py install --single-version-externally-managed --record=record.txt
-    else
-        echo "building rmm..."
-        python setup.py build_ext --inplace
     fi
 
 fi

ci/cpu/upload.sh

@@ -29,8 +29,8 @@ fi
 
 gpuci_logger "Get conda file output locations"
 
-export LIBRMM_FILE=`conda build conda/recipes/librmm --output`
-export RMM_FILE=`conda build conda/recipes/rmm --python=$PYTHON --output`
+export LIBRMM_FILE=$(conda build conda/recipes/librmm --output)
+export RMM_FILES=$(conda build conda/recipes/rmm --python=$PYTHON --output)
 
 ################################################################################
 # UPLOAD - Conda packages
@@ -46,9 +46,11 @@ if [[ "$BUILD_LIBRMM" == "1" && "$UPLOAD_LIBRMM" == "1" ]]; then
 fi
 
 if [[ "$BUILD_RMM" == "1" && "$UPLOAD_RMM" == "1" ]]; then
-  test -e ${RMM_FILE}
-  echo "Upload rmm"
-  echo ${RMM_FILE}
-  gpuci_retry anaconda -t ${MY_UPLOAD_KEY} upload -u ${CONDA_USERNAME:-rapidsai} ${LABEL_OPTION} --skip-existing ${RMM_FILE} --no-progress
+  while read -r RMM_FILE; do
+    test -e ${RMM_FILE}
+    echo "Upload rmm"
+    echo ${RMM_FILE}
+    gpuci_retry anaconda -t ${MY_UPLOAD_KEY} upload -u ${CONDA_USERNAME:-rapidsai} ${LABEL_OPTION} --skip-existing ${RMM_FILE} --no-progress
+  done <<< "${RMM_FILES}"
 fi
 

ci/gpu/build.sh

@@ -23,7 +23,9 @@ export HOME=$WORKSPACE
 cd $WORKSPACE
 
 # Determine CUDA release version
-export CUDA_REL=${CUDA_VERSION%.*}
+export CUDA_MAJOR_VER=$(echo "${CUDA_VERSION}" | cut -f 1 -d.)
+export CUDA_MINOR_VER=$(echo "${CUDA_VERSION}" | cut -f 2 -d.)
+export CUDA_REL="${CUDA_MAJOR_VER}.${CUDA_MINOR_VER}"
 
 # Get latest tag and number of commits since tag
 export GIT_DESCRIBE_TAG=`git describe --abbrev=0 --tags`
@@ -46,7 +48,7 @@ gpuci_mamba_retry install -y \
                   "cudatoolkit=$CUDA_REL" \
                   "rapids-build-env=${MINOR_VERSION}.*"
 
-# https://docs.rapids.ai/maintainers/depmgmt/ 
+# https://docs.rapids.ai/maintainers/depmgmt/
 # conda remove --force rapids-build-env
 # gpuci_mamba_retry install "your-pkg=1.0.0"
 
@@ -110,18 +112,18 @@ else
     done
 
     cd $WORKSPACE/python
-    
+
     CONDA_FILE=`find $WORKSPACE/ci/artifacts/rmm/cpu/conda-bld/ -name "librmm*.tar.bz2"`
     CONDA_FILE=`basename "$CONDA_FILE" .tar.bz2` #get filename without extension
     CONDA_FILE=${CONDA_FILE//-/=} #convert to conda install
     gpuci_logger "Installing $CONDA_FILE"
     gpuci_mamba_retry install -c $WORKSPACE/ci/artifacts/rmm/cpu/conda-bld/ "$CONDA_FILE"
 
     export LIBRMM_BUILD_DIR="$WORKSPACE/ci/artifacts/rmm/cpu/conda_work/build"
-    
+
     gpuci_logger "Building rmm"
     "$WORKSPACE/build.sh" -v rmm
-    
+
     gpuci_logger "pytest rmm"
     py.test --cache-clear --junitxml=${WORKSPACE}/test-results/junit-rmm.xml -v --cov-config=.coveragerc --cov=rmm --cov-report=xml:${WORKSPACE}/python/rmm-coverage.xml --cov-report term
     exitcode=$?

cmake/thirdparty/get_spdlog.cmake

@@ -16,13 +16,12 @@
 function(find_and_configure_spdlog)
 
   include(${rapids-cmake-dir}/cpm/spdlog.cmake)
-  rapids_cpm_spdlog()
+  rapids_cpm_spdlog(INSTALL_EXPORT_SET rmm-exports)
 
   if(spdlog_ADDED)
     install(TARGETS spdlog_header_only EXPORT rmm-exports)
   else()
     rapids_export_package(BUILD spdlog rmm-exports)
-    rapids_export_package(INSTALL spdlog rmm-exports)
   endif()
 endfunction()
 

conda/environments/rmm_dev_cuda10.1.yml

@@ -18,3 +18,4 @@ dependencies:
 - cudatoolkit=10.1
 - spdlog>=1.8.5,<1.9
 - cython>=0.29,<0.30
+- gcovr>=5.0

conda/environments/rmm_dev_cuda10.2.yml

@@ -18,3 +18,4 @@ dependencies:
 - cudatoolkit=10.2
 - spdlog>=1.8.5,<1.9
 - cython>=0.29,<0.30
+- gcovr>=5.0

conda/environments/rmm_dev_cuda11.0.yml

@@ -18,3 +18,4 @@ dependencies:
 - cudatoolkit=11.0
 - spdlog>=1.8.5,<1.9
 - cython>=0.29,<0.30
+- gcovr>=5.0

conda/recipes/librmm/build.sh

@@ -1,4 +1,4 @@
 # Copyright (c) 2018-2019, NVIDIA CORPORATION.
 
 # This assumes the script is executed from the root of the repo directory
-./build.sh -v clean librmm
+./build.sh -v clean librmm --cmake-args=\"-DCMAKE_INSTALL_LIBDIR=lib\"

conda/recipes/librmm/meta.yaml

@@ -2,6 +2,8 @@
 
 {% set version = environ.get('GIT_DESCRIBE_TAG', '0.0.0.dev').lstrip('v') + environ.get('VERSION_SUFFIX', '') %}
 {% set cuda_version='.'.join(environ.get('CUDA', '9.2').split('.')[:2]) %}
+{% set cuda_major=cuda_version.split('.')[0] %}
+
 package:
   name: librmm
   version: {{ version }}
@@ -11,7 +13,7 @@ source:
 
 build:
   number: {{ GIT_DESCRIBE_NUMBER }}
-  string: cuda{{ cuda_version }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
+  string: cuda{{ cuda_major }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
   script_env:
     - CC
     - CXX
@@ -34,8 +36,8 @@ requirements:
   host:
     - cudatoolkit {{ cuda_version }}.*
   run:
-    - {{ pin_compatible('cudatoolkit', max_pin='x.x') }}
     - spdlog>=1.8.5,<1.9
+    - {{ pin_compatible('cudatoolkit', max_pin='x', min_pin='x') }}
 
 test:
   commands:

conda/recipes/rmm/build.sh

@@ -1,4 +1,9 @@
 # Copyright (c) 2018-2019, NVIDIA CORPORATION.
 
 # Script assumes the script is executed from the root of the repo directory
-./build.sh -v clean rmm
+BUILD_FLAGS=""
+if [ "${cudaMallocAsync}" == "no_cma" ]; then
+    BUILD_FLAGS="${BUILD_FLAGS} --no-cudamallocasync"
+fi
+
+./build.sh -v clean rmm ${BUILD_FLAGS}

conda/recipes/rmm/conda_build_config.yaml

@@ -0,0 +1,3 @@
+cudaMallocAsync:
+  - has_cma
+  - no_cma

conda/recipes/rmm/meta.yaml

@@ -3,6 +3,7 @@
 {% set version = environ.get('GIT_DESCRIBE_TAG', '0.0.0.dev').lstrip('v') + environ.get('VERSION_SUFFIX', '') %}
 {% set py_version=environ.get('CONDA_PY', 35) %}
 {% set cuda_version='.'.join(environ.get('CUDA', '10.1').split('.')[:2]) %}
+{% set cuda_major=cuda_version.split('.')[0] %}
 
 package:
   name: rmm
@@ -13,7 +14,7 @@ source:
 
 build:
   number: {{ GIT_DESCRIBE_NUMBER }}
-  string: cuda_{{ cuda_version }}_py{{ py_version }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
+  string: cuda{{ cuda_major }}_py{{ py_version }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}_{{ cudaMallocAsync }}
   script_env:
     - RMM_BUILD_NO_GPU_TEST
     - VERSION_SUFFIX
@@ -33,7 +34,11 @@ requirements:
     - python
     - numba >=0.49
     - numpy
-    - {{ pin_compatible('cudatoolkit', max_pin='x.x') }}
+{% if cudaMallocAsync == "has_cma" %}
+    - {{ pin_compatible('cudatoolkit', max_pin='x', lower_bound='11.2') }} # cudatoolkit >=11.2,<12.0.0
+{% else %}
+    - {{ pin_compatible('cudatoolkit', upper_bound='11.2', lower_bound='11.0') }} # cudatoolkit >=11.0,<11.2
+{% endif %}
 
 test:
   commands:

doxygen/Doxyfile

@@ -38,7 +38,7 @@ PROJECT_NAME           = "RMM"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = 21.12
+PROJECT_NUMBER         = 22.02
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

gcovr.cfg

@@ -0,0 +1,4 @@
+exclude=build/.*
+exclude=tests/.*
+html=yes
+html-details=yes