Merge branch 'branch-24.02' into fea/add_ci_check_for_external_kernels

ci/build_docs.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 set -euo pipefail
 
@@ -40,8 +40,8 @@ popd
 
 rapids-logger "Build Python docs"
 pushd docs/cudf
-make dirhtml
-make text
+make dirhtml O="-j 4"
+make text O="-j 4"
 mkdir -p "${RAPIDS_DOCS_DIR}/cudf/"{html,txt}
 mv build/dirhtml/* "${RAPIDS_DOCS_DIR}/cudf/html"
 mv build/text/* "${RAPIDS_DOCS_DIR}/cudf/txt"

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -12,6 +12,7 @@ dependencies:
 - benchmark==1.8.0
 - boto3>=1.21.21
 - botocore>=1.24.21
+- breathe>=4.35.0
 - c-compiler
 - cachetools
 - clang-tools=16.0.6
@@ -74,7 +75,7 @@ dependencies:
 - pydata-sphinx-theme!=0.14.2
 - pytest
 - pytest-benchmark
-- pytest-cases<3.8.2
+- pytest-cases>=3.8.2
 - pytest-cov
 - pytest-xdist
 - python-confluent-kafka>=1.9.0,<1.10.0a0

conda/environments/all_cuda-120_arch-x86_64.yaml

@@ -12,6 +12,7 @@ dependencies:
 - benchmark==1.8.0
 - boto3>=1.21.21
 - botocore>=1.24.21
+- breathe>=4.35.0
 - c-compiler
 - cachetools
 - clang-tools=16.0.6
@@ -71,7 +72,7 @@ dependencies:
 - pydata-sphinx-theme!=0.14.2
 - pytest
 - pytest-benchmark
-- pytest-cases<3.8.2
+- pytest-cases>=3.8.2
 - pytest-cov
 - pytest-xdist
 - python-confluent-kafka>=1.9.0,<1.10.0a0

cpp/CMakeLists.txt

@@ -380,6 +380,8 @@ add_library(
   src/io/orc/dict_enc.cu
   src/io/orc/orc.cpp
   src/io/orc/reader_impl.cu
+  src/io/orc/reader_impl_helpers.cpp
+  src/io/orc/reader_impl_preprocess.cu
   src/io/orc/stats_enc.cu
   src/io/orc/stripe_data.cu
   src/io/orc/stripe_enc.cu

cpp/benchmarks/CMakeLists.txt

@@ -1,5 +1,5 @@
 # =============================================================================
-# Copyright (c) 2018-2023, NVIDIA CORPORATION.
+# Copyright (c) 2018-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except
 # in compliance with the License. You may obtain a copy of the License at
@@ -183,6 +183,11 @@ ConfigureNVBench(
   sort/sort_lists.cpp sort/sort_structs.cpp
 )
 
+# ##################################################################################################
+# * structs benchmark
+# --------------------------------------------------------------------------------
+ConfigureNVBench(STRUCT_CREATION_NVBENCH structs/create_structs.cpp)
+
 # ##################################################################################################
 # * quantiles benchmark
 # --------------------------------------------------------------------------------

cpp/benchmarks/structs/create_structs.cpp

@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2024, NVIDIA CORPORATION.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <benchmarks/common/generate_nested_types.hpp>
+
+#include <nvbench/nvbench.cuh>
+
+void nvbench_create_structs(nvbench::state& state)
+{
+  state.exec(nvbench::exec_tag::sync,
+             [&](nvbench::launch& launch) { auto const table_ptr = create_structs_data(state); });
+}
+
+NVBENCH_BENCH(nvbench_create_structs)
+  .set_name("create_structs")
+  .add_int64_power_of_two_axis("NumRows", {10, 18, 26})
+  .add_int64_axis("Depth", {1, 8, 16})
+  .add_int64_axis("Nulls", {0, 1});

cpp/doxygen/developer_guide/TESTING.md

@@ -464,9 +464,9 @@ the host (`to_host`).
 
 ### Background
 
-libcudf employs a custom-built [preload library
-docs](https://man7.org/linux/man-pages/man8/ld.so.8.html) to validate its internal stream usage (the
-code may be found
+libcudf employs a custom-built [preload
+library](https://man7.org/linux/man-pages/man8/ld.so.8.html) to validate its internal stream usage
+(the code may be found
 [`here`](https://github.com/rapidsai/cudf/blob/main/cpp/tests/utilities/identify_stream_usage.cpp)).
 This library wraps every asynchronous CUDA runtime API call that accepts a stream with a check to
 ensure that the passed CUDA stream is a valid one, immediately throwing an exception if an invalid

cpp/include/cudf/io/orc.hpp

@@ -393,13 +393,15 @@ class orc_reader_options_builder {
  * @endcode
  *
  * @param options Settings for controlling reading behavior
+ * @param stream CUDA stream used for device memory operations and kernel launches
  * @param mr Device memory resource used to allocate device memory of the table in the returned
  * table_with_metadata.
  *
  * @return The set of columns
  */
 table_with_metadata read_orc(
   orc_reader_options const& options,
+  rmm::cuda_stream_view stream        = cudf::get_default_stream(),
   rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
 
 /** @} */  // end of group
@@ -864,8 +866,10 @@ class orc_writer_options_builder {
  * @endcode
  *
  * @param options Settings for controlling reading behavior
+ * @param stream CUDA stream used for device memory operations and kernel launches
  */
-void write_orc(orc_writer_options const& options);
+void write_orc(orc_writer_options const& options,
+               rmm::cuda_stream_view stream = cudf::get_default_stream());
 
 /**
  * @brief Builds settings to use for `write_orc_chunked()`.
@@ -1287,8 +1291,10 @@ class orc_chunked_writer {
    * @brief Constructor with chunked writer options
    *
    * @param[in] options options used to write table
+   * @param[in] stream CUDA stream used for device memory operations and kernel launches
    */
-  orc_chunked_writer(chunked_orc_writer_options const& options);
+  orc_chunked_writer(chunked_orc_writer_options const& options,
+                     rmm::cuda_stream_view stream = cudf::get_default_stream());
 
   /**
    * @brief Writes table to output.

cpp/include/cudf/io/orc_metadata.hpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
+ * Copyright (c) 2019-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -63,10 +63,12 @@ struct raw_orc_statistics {
  * @endcode
  *
  * @param src_info Dataset source
+ * @param stream CUDA stream used for device memory operations and kernel launches
  *
  * @return Column names and encoded ORC statistics
  */
-raw_orc_statistics read_raw_orc_statistics(source_info const& src_info);
+raw_orc_statistics read_raw_orc_statistics(
+  source_info const& src_info, rmm::cuda_stream_view stream = cudf::get_default_stream());
 
 /**
  * @brief Monostate type alias for the statistics variant.
@@ -207,10 +209,12 @@ struct parsed_orc_statistics {
  * @ingroup io_readers
  *
  * @param src_info Dataset source
+ * @param stream CUDA stream used for device memory operations and kernel launches
  *
  * @return Column names and decoded ORC statistics
  */
-parsed_orc_statistics read_parsed_orc_statistics(source_info const& src_info);
+parsed_orc_statistics read_parsed_orc_statistics(
+  source_info const& src_info, rmm::cuda_stream_view stream = cudf::get_default_stream());
 
 /**
  * @brief Schema of an ORC column, including the nested columns.
@@ -368,10 +372,12 @@ class orc_metadata {
  * @ingroup io_readers
  *
  * @param src_info Dataset source
+ * @param stream CUDA stream used for device memory operations and kernel launches
  *
  * @return orc_metadata with ORC schema, number of rows and number of stripes.
  */
-orc_metadata read_orc_metadata(source_info const& src_info);
+orc_metadata read_orc_metadata(source_info const& src_info,
+                               rmm::cuda_stream_view stream = cudf::get_default_stream());
 
 /** @} */  // end of group
 }  // namespace io

cpp/include/cudf/strings/strings_column_view.hpp

@@ -106,7 +106,7 @@ class strings_column_view : private column_view {
   /**
    * @brief Returns the internal column of chars
    *
-   * @throw cudf::logic error if this is an empty column
+   * @throw cudf::logic_error if this is an empty column
    * @param stream CUDA stream used for device memory operations and kernel launches
    * @return The chars column
    */

cpp/src/io/functions.cpp

@@ -269,9 +269,9 @@ void write_csv(csv_writer_options const& options,
     mr);
 }
 
-raw_orc_statistics read_raw_orc_statistics(source_info const& src_info)
+raw_orc_statistics read_raw_orc_statistics(source_info const& src_info,
+                                           rmm::cuda_stream_view stream)
 {
-  auto stream = cudf::get_default_stream();
   // Get source to read statistics from
   std::unique_ptr<datasource> source;
   if (src_info.type() == io_type::FILEPATH) {
@@ -342,9 +342,10 @@ column_statistics::column_statistics(orc::column_statistics&& cs)
   }
 }
 
-parsed_orc_statistics read_parsed_orc_statistics(source_info const& src_info)
+parsed_orc_statistics read_parsed_orc_statistics(source_info const& src_info,
+                                                 rmm::cuda_stream_view stream)
 {
-  auto const raw_stats = read_raw_orc_statistics(src_info);
+  auto const raw_stats = read_raw_orc_statistics(src_info, stream);
 
   parsed_orc_statistics result;
   result.column_names = raw_stats.column_names;
@@ -395,12 +396,12 @@ orc_column_schema make_orc_column_schema(host_span<orc::SchemaType const> orc_sc
 }
 };  // namespace
 
-orc_metadata read_orc_metadata(source_info const& src_info)
+orc_metadata read_orc_metadata(source_info const& src_info, rmm::cuda_stream_view stream)
 {
   auto sources = make_datasources(src_info);
 
   CUDF_EXPECTS(sources.size() == 1, "Only a single source is currently supported.");
-  auto const footer = orc::metadata(sources.front().get(), cudf::detail::default_stream_value).ff;
+  auto const footer = orc::metadata(sources.front().get(), stream).ff;
 
   return {{make_orc_column_schema(footer.types, 0, "")},
           static_cast<size_type>(footer.numberOfRows),
@@ -410,21 +411,21 @@ orc_metadata read_orc_metadata(source_info const& src_info)
 /**
  * @copydoc cudf::io::read_orc
  */
-table_with_metadata read_orc(orc_reader_options const& options, rmm::mr::device_memory_resource* mr)
+table_with_metadata read_orc(orc_reader_options const& options,
+                             rmm::cuda_stream_view stream,
+                             rmm::mr::device_memory_resource* mr)
 {
   CUDF_FUNC_RANGE();
 
   auto datasources = make_datasources(options.get_source());
-  auto reader      = std::make_unique<orc::detail::reader>(
-    std::move(datasources), options, cudf::get_default_stream(), mr);
-
+  auto reader = std::make_unique<orc::detail::reader>(std::move(datasources), options, stream, mr);
   return reader->read(options);
 }
 
 /**
  * @copydoc cudf::io::write_orc
  */
-void write_orc(orc_writer_options const& options)
+void write_orc(orc_writer_options const& options, rmm::cuda_stream_view stream)
 {
   namespace io_detail = cudf::io::detail;
 
@@ -434,8 +435,7 @@ void write_orc(orc_writer_options const& options)
   CUDF_EXPECTS(sinks.size() == 1, "Multiple sinks not supported for ORC writing");
 
   auto writer = std::make_unique<orc::detail::writer>(
-    std::move(sinks[0]), options, io_detail::single_write_mode::YES, cudf::get_default_stream());
-
+    std::move(sinks[0]), options, io_detail::single_write_mode::YES, stream);
   try {
     writer->write(options.get_table());
   } catch (...) {
@@ -451,15 +451,16 @@ void write_orc(orc_writer_options const& options)
 /**
  * @copydoc cudf::io::orc_chunked_writer::orc_chunked_writer
  */
-orc_chunked_writer::orc_chunked_writer(chunked_orc_writer_options const& options)
+orc_chunked_writer::orc_chunked_writer(chunked_orc_writer_options const& options,
+                                       rmm::cuda_stream_view stream)
 {
   namespace io_detail = cudf::io::detail;
 
   auto sinks = make_datasinks(options.get_sink());
   CUDF_EXPECTS(sinks.size() == 1, "Multiple sinks not supported for ORC writing");
 
   writer = std::make_unique<orc::detail::writer>(
-    std::move(sinks[0]), options, io_detail::single_write_mode::NO, cudf::get_default_stream());
+    std::move(sinks[0]), options, io_detail::single_write_mode::NO, stream);
 }
 
 /**

cpp/src/io/orc/aggregate_orc_metadata.cpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -20,7 +20,6 @@
 
 #include <algorithm>
 #include <numeric>
-#include <optional>
 
 namespace cudf::io::orc::detail {
 
@@ -220,27 +219,25 @@ aggregate_orc_metadata::select_stripes(
   }
 
   // Read each stripe's stripefooter metadata
-  if (not selected_stripes_mapping.empty()) {
-    for (auto& mapping : selected_stripes_mapping) {
-      // Resize to all stripe_info for the source level
-      per_file_metadata[mapping.source_idx].stripefooters.resize(mapping.stripe_info.size());
-
-      for (size_t i = 0; i < mapping.stripe_info.size(); i++) {
-        auto const stripe         = mapping.stripe_info[i].first;
-        auto const sf_comp_offset = stripe->offset + stripe->indexLength + stripe->dataLength;
-        auto const sf_comp_length = stripe->footerLength;
-        CUDF_EXPECTS(
-          sf_comp_offset + sf_comp_length < per_file_metadata[mapping.source_idx].source->size(),
-          "Invalid stripe information");
-        auto const buffer =
-          per_file_metadata[mapping.source_idx].source->host_read(sf_comp_offset, sf_comp_length);
-        auto sf_data = per_file_metadata[mapping.source_idx].decompressor->decompress_blocks(
-          {buffer->data(), buffer->size()}, stream);
-        ProtobufReader(sf_data.data(), sf_data.size())
-          .read(per_file_metadata[mapping.source_idx].stripefooters[i]);
-        mapping.stripe_info[i].second = &per_file_metadata[mapping.source_idx].stripefooters[i];
-        if (stripe->indexLength == 0) { row_grp_idx_present = false; }
-      }
+  for (auto& mapping : selected_stripes_mapping) {
+    // Resize to all stripe_info for the source level
+    per_file_metadata[mapping.source_idx].stripefooters.resize(mapping.stripe_info.size());
+
+    for (size_t i = 0; i < mapping.stripe_info.size(); i++) {
+      auto const stripe         = mapping.stripe_info[i].first;
+      auto const sf_comp_offset = stripe->offset + stripe->indexLength + stripe->dataLength;
+      auto const sf_comp_length = stripe->footerLength;
+      CUDF_EXPECTS(
+        sf_comp_offset + sf_comp_length < per_file_metadata[mapping.source_idx].source->size(),
+        "Invalid stripe information");
+      auto const buffer =
+        per_file_metadata[mapping.source_idx].source->host_read(sf_comp_offset, sf_comp_length);
+      auto sf_data = per_file_metadata[mapping.source_idx].decompressor->decompress_blocks(
+        {buffer->data(), buffer->size()}, stream);
+      ProtobufReader(sf_data.data(), sf_data.size())
+        .read(per_file_metadata[mapping.source_idx].stripefooters[i]);
+      mapping.stripe_info[i].second = &per_file_metadata[mapping.source_idx].stripefooters[i];
+      if (stripe->indexLength == 0) { row_grp_idx_present = false; }
     }
   }
 
@@ -270,7 +267,7 @@ column_hierarchy aggregate_orc_metadata::select_columns(
       CUDF_EXPECTS(name_found, "Unknown column name: " + std::string(path));
     }
   }
-  return {std::move(selected_columns)};
+  return column_hierarchy{std::move(selected_columns)};
 }
 
 }  // namespace cudf::io::orc::detail