Merge pull request #14422 from bdice/branch-24.02-merge-23.12

Forward-merge branch-23.12 to branch-24.02
rapidsai · Nov 16, 2023 · 6b4deeb · 6b4deeb
2 parents deeabc9 + e4e6975
commit 6b4deeb
Show file tree

Hide file tree

Showing 84 changed files with 1,153 additions and 586 deletions.
diff --git a/build.sh b/build.sh
@@ -369,7 +369,7 @@ fi
 # build cudf_kafka Python package
 if hasArg cudf_kafka; then
     cd ${REPODIR}/python/cudf_kafka
-    SKBUILD_CONFIGURE_OPTIONS="-DCMAKE_LIBRARY_PATH=${LIBCUDF_BUILD_DIR}" \
+    SKBUILD_CONFIGURE_OPTIONS="-DCMAKE_PREFIX_PATH=${INSTALL_PREFIX} -DCMAKE_LIBRARY_PATH=${LIBCUDF_BUILD_DIR} ${EXTRA_CMAKE_ARGS}" \
         SKBUILD_BUILD_OPTIONS="-j${PARALLEL_LEVEL:-1}" \
         python -m pip install --no-build-isolation --no-deps .
 fi

diff --git a/ci/build_wheel.sh b/ci/build_wheel.sh
@@ -36,6 +36,8 @@ fi
 
 if [[ ${package_name} == "dask_cudf" ]]; then
     sed -r -i "s/cudf==(.*)\"/cudf${PACKAGE_CUDA_SUFFIX}==\1${alpha_spec}\"/g" ${pyproject_file}
+    sed -r -i "s/dask-cuda==(.*)\"/dask-cuda==\1${alpha_spec}\"/g" ${pyproject_file}
+    sed -r -i "s/rapids-dask-dependency==(.*)\"/rapids-dask-dependency==\1${alpha_spec}\"/g" ${pyproject_file}
 else
     sed -r -i "s/rmm(.*)\"/rmm${PACKAGE_CUDA_SUFFIX}\1${alpha_spec}\"/g" ${pyproject_file}
     # ptxcompiler and cubinlinker aren't version constrained

diff --git a/ci/release/update-version.sh b/ci/release/update-version.sh
@@ -43,6 +43,7 @@ sed_runner 's/'"VERSION ${CURRENT_SHORT_TAG}.*"'/'"VERSION ${NEXT_FULL_TAG}"'/g'
 
 # Python CMakeLists updates
 sed_runner 's/'"cudf_version .*)"'/'"cudf_version ${NEXT_FULL_TAG})"'/g' python/cudf/CMakeLists.txt
+sed_runner 's/'"cudf_kafka_version .*)"'/'"cudf_kafka_version ${NEXT_FULL_TAG})"'/g' python/cudf_kafka/CMakeLists.txt
 
 # cpp libcudf_kafka update
 sed_runner 's/'"VERSION ${CURRENT_SHORT_TAG}.*"'/'"VERSION ${NEXT_FULL_TAG}"'/g' cpp/libcudf_kafka/CMakeLists.txt
@@ -80,6 +81,7 @@ DEPENDENCIES=(
   kvikio
   libkvikio
   librmm
+  rapids-dask-dependency
   rmm
 )
 for DEP in "${DEPENDENCIES[@]}"; do

diff --git a/ci/test_wheel_dask_cudf.sh b/ci/test_wheel_dask_cudf.sh
@@ -23,9 +23,6 @@ manylinux="manylinux_${manylinux_version}"
 RAPIDS_PY_WHEEL_NAME="cudf_${manylinux}_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-cudf-dep
 python -m pip install --no-deps ./local-cudf-dep/cudf*.whl
 
-# Always install latest dask for testing
-python -m pip install git+https://github.com/dask/dask.git@main git+https://github.com/dask/distributed.git@main git+https://github.com/rapidsai/dask-cuda.git@branch-24.02
-
 # echo to expand wildcard before adding `[extra]` requires for pip
 python -m pip install $(echo ./dist/dask_cudf*.whl)[test]
 

diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -24,11 +24,8 @@ dependencies:
 - cudatoolkit
 - cupy>=12.0.0
 - cxx-compiler
-- cython>=3.0.0
-- dask-core>=2023.9.2
+- cython>=3.0.3
 - dask-cuda==24.2.*
-- dask>=2023.9.2
-- distributed>=2023.9.2
 - dlpack>=0.5,<0.6.0a0
 - doxygen=1.9.1
 - fastavro>=0.22.9
@@ -60,7 +57,7 @@ dependencies:
 - numpy>=1.21,<1.25
 - numpydoc
 - nvcc_linux-64=11.8
-- nvcomp==2.6.1
+- nvcomp==3.0.4
 - nvtx>=0.2.1
 - packaging
 - pandas>=1.3,<1.6.0dev0
@@ -80,6 +77,7 @@ dependencies:
 - python-snappy>=0.6.0
 - python>=3.9,<3.11
 - pytorch<1.12.0
+- rapids-dask-dependency==24.2.*
 - rich
 - rmm==24.2.*
 - s3fs>=2022.3.0

diff --git a/conda/environments/all_cuda-120_arch-x86_64.yaml b/conda/environments/all_cuda-120_arch-x86_64.yaml
@@ -17,7 +17,6 @@ dependencies:
 - cmake>=3.26.4
 - cramjam
 - cuda-cudart-dev
-- cuda-gdb
 - cuda-nvcc
 - cuda-nvrtc-dev
 - cuda-nvtx-dev
@@ -26,11 +25,8 @@ dependencies:
 - cuda-version=12.0
 - cupy>=12.0.0
 - cxx-compiler
-- cython>=3.0.0
-- dask-core>=2023.9.2
+- cython>=3.0.3
 - dask-cuda==24.2.*
-- dask>=2023.9.2
-- distributed>=2023.9.2
 - dlpack>=0.5,<0.6.0a0
 - doxygen=1.9.1
 - fastavro>=0.22.9
@@ -59,7 +55,7 @@ dependencies:
 - numba>=0.57,<0.58
 - numpy>=1.21,<1.25
 - numpydoc
-- nvcomp==2.6.1
+- nvcomp==3.0.4
 - nvtx>=0.2.1
 - packaging
 - pandas>=1.3,<1.6.0dev0
@@ -78,6 +74,7 @@ dependencies:
 - python-snappy>=0.6.0
 - python>=3.9,<3.11
 - pytorch<1.12.0
+- rapids-dask-dependency==24.2.*
 - rich
 - rmm==24.2.*
 - s3fs>=2022.3.0

diff --git a/conda/recipes/cudf/meta.yaml b/conda/recipes/cudf/meta.yaml
@@ -57,7 +57,7 @@ requirements:
   host:
     - protobuf ==4.24.*
     - python
-    - cython >=3.0.0
+    - cython >=3.0.3
     - scikit-build >=0.13.1
     - setuptools
     - dlpack >=0.5,<0.6.0a0

diff --git a/conda/recipes/cudf_kafka/build.sh b/conda/recipes/cudf_kafka/build.sh
@@ -1,16 +1,3 @@
 # Copyright (c) 2020-2023, NVIDIA CORPORATION.
 
-# This assumes the script is executed from the root of the repo directory
-# Need to set CUDA_HOME inside conda environments because the hacked together
-# setup.py for cudf-kafka searches that way.
-# TODO: Remove after https://github.com/rapidsai/cudf/pull/14292 updates
-# cudf_kafka to use scikit-build
-CUDA_MAJOR=${RAPIDS_CUDA_VERSION%%.*}
-if [[ ${CUDA_MAJOR} == "12" ]]; then
-    target_name="x86_64-linux"
-    if [[ ! $(arch) == "x86_64" ]]; then
-        target_name="sbsa-linux"
-    fi
-    export CUDA_HOME="${PREFIX}/targets/${target_name}/"
-fi
 ./build.sh -v cudf_kafka
diff --git a/conda/recipes/cudf_kafka/conda_build_config.yaml b/conda/recipes/cudf_kafka/conda_build_config.yaml
@@ -9,3 +9,9 @@ sysroot_version:
 
 cmake_version:
   - ">=3.26.4"
+
+cuda_compiler:
+  - cuda-nvcc
+
+cuda11_compiler:
+  - nvcc
diff --git a/conda/recipes/cudf_kafka/meta.yaml b/conda/recipes/cudf_kafka/meta.yaml
@@ -33,28 +33,31 @@ build:
     - SCCACHE_S3_KEY_PREFIX=cudf-kafka-linux64 # [linux64]
     - SCCACHE_S3_USE_SSL
     - SCCACHE_S3_NO_CREDENTIALS
-    # TODO: Remove after https://github.com/rapidsai/cudf/pull/14292 updates
-    # cudf_kafka to use scikit-build
-    - RAPIDS_CUDA_VERSION
+  ignore_run_exports_from:
+    {% if cuda_major == "11" %}
+    - {{ compiler('cuda11') }}
+    {% endif %}
 
 requirements:
   build:
     - cmake {{ cmake_version }}
+    - ninja
     - {{ compiler('c') }}
     - {{ compiler('cxx') }}
-    - ninja
-    - sysroot_{{ target_platform }} {{ sysroot_version }}
-    # TODO: Remove after https://github.com/rapidsai/cudf/pull/14292 updates
-    # cudf_kafka to use scikit-build
-    {% if cuda_major == "12" %}
-    - cuda-gdb
+    {% if cuda_major == "11" %}
+    - {{ compiler('cuda11') }} ={{ cuda_version }}
+    {% else %}
+    - {{ compiler('cuda') }}
     {% endif %}
+    - cuda-version ={{ cuda_version }}
+    - sysroot_{{ target_platform }} {{ sysroot_version }}
   host:
     - python
-    - cython >=3.0.0
+    - cython >=3.0.3
     - cuda-version ={{ cuda_version }}
     - cudf ={{ version }}
     - libcudf_kafka ={{ version }}
+    - scikit-build >=0.13.1
     - setuptools
     {% if cuda_major == "12" %}
     - cuda-cudart-dev

diff --git a/conda/recipes/custreamz/meta.yaml b/conda/recipes/custreamz/meta.yaml
@@ -45,9 +45,7 @@ requirements:
     - streamz
     - cudf ={{ version }}
     - cudf_kafka ={{ version }}
-    - dask >=2023.9.2
-    - dask-core >=2023.9.2
-    - distributed >=2023.9.2
+    - rapids-dask-dependency ={{ minor_version }}
     - python-confluent-kafka >=1.9.0,<1.10.0a0
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
 

diff --git a/conda/recipes/dask-cudf/meta.yaml b/conda/recipes/dask-cudf/meta.yaml
@@ -37,17 +37,11 @@ build:
 requirements:
   host:
     - python
-    - cudf ={{ version }}
-    - dask >=2023.9.2
-    - dask-core >=2023.9.2
-    - distributed >=2023.9.2
     - cuda-version ={{ cuda_version }}
   run:
     - python
     - cudf ={{ version }}
-    - dask >=2023.9.2
-    - dask-core >=2023.9.2
-    - distributed >=2023.9.2
+    - rapids-dask-dependency ={{ minor_version }}
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
 
 test:

diff --git a/conda/recipes/dask-cudf/run_test.sh b/conda/recipes/dask-cudf/run_test.sh
diff --git a/conda/recipes/libcudf/conda_build_config.yaml b/conda/recipes/libcudf/conda_build_config.yaml
@@ -38,7 +38,7 @@ spdlog_version:
   - ">=1.11.0,<1.12"
 
 nvcomp_version:
-  - "=2.6.1"
+  - "=3.0.4"
 
 zlib_version:
   - ">=1.2.13"

diff --git a/cpp/benchmarks/text/vocab.cpp b/cpp/benchmarks/text/vocab.cpp
@@ -53,7 +53,7 @@ static void bench_vocab_tokenize(nvbench::state& state)
 
   auto const vocab_col = [] {
     data_profile const profile = data_profile_builder().no_validity().distribution(
-      cudf::type_id::STRING, distribution_id::NORMAL, 0, 5);
+      cudf::type_id::STRING, distribution_id::NORMAL, 0, 15);
     auto const col = create_random_column(cudf::type_id::STRING, row_count{100}, profile);
     return cudf::strings::filter_characters_of_type(
       cudf::strings_column_view(col->view()),

diff --git a/cpp/cmake/thirdparty/get_nvbench.cmake b/cpp/cmake/thirdparty/get_nvbench.cmake
@@ -21,7 +21,7 @@ function(find_and_configure_nvbench)
   set(cudf_patch_dir "${CMAKE_CURRENT_FUNCTION_LIST_DIR}/patches")
   rapids_cpm_package_override("${cudf_patch_dir}/nvbench_override.json")
 
-  rapids_cpm_nvbench()
+  rapids_cpm_nvbench(BUILD_STATIC)
 
 endfunction()
 

diff --git a/cpp/cmake/thirdparty/patches/nvbench_override.json b/cpp/cmake/thirdparty/patches/nvbench_override.json
@@ -9,8 +9,8 @@
           "fixed_in" : ""
         },
         {
-          "file" : "nvbench/use_existing_fmt.diff",
-          "issue" : "Fix add support for using an existing fmt [https://github.com/NVIDIA/nvbench/pull/125]",
+          "file" : "nvbench/nvml_with_static_builds.diff",
+          "issue" : "Add support for nvml with static nvbench [https://github.com/NVIDIA/nvbench/pull/148]",
           "fixed_in" : ""
         }
       ]

diff --git a/cpp/include/cudf/column/column_device_view.cuh b/cpp/include/cudf/column/column_device_view.cuh
@@ -16,6 +16,7 @@
 #pragma once
 
 #include <cudf/column/column_view.hpp>
+#include <cudf/detail/offsets_iterator.cuh>
 #include <cudf/detail/utilities/alignment.hpp>
 #include <cudf/fixed_point/fixed_point.hpp>
 #include <cudf/lists/list_view.hpp>
@@ -442,10 +443,11 @@ class alignas(16) column_device_view : public detail::column_device_view_base {
   __device__ T element(size_type element_index) const noexcept
   {
     size_type index       = element_index + offset();  // account for this view's _offset
-    auto const* d_offsets = d_children[strings_column_view::offsets_column_index].data<int32_t>();
     char const* d_strings = d_children[strings_column_view::chars_column_index].data<char>();
-    size_type offset      = d_offsets[index];
-    return string_view{d_strings + offset, d_offsets[index + 1] - offset};
+    auto const offsets    = d_children[strings_column_view::offsets_column_index];
+    auto const itr        = cudf::detail::input_offsetalator(offsets.head(), offsets.type());
+    auto const offset     = itr[index];
+    return string_view{d_strings + offset, static_cast<cudf::size_type>(itr[index + 1] - offset)};
   }
 
  private: