Merge branch 'branch-25.02' into cln/column_empty/masked

.github/workflows/pr.yaml

@@ -52,6 +52,7 @@ jobs:
       OTEL_SERVICE_NAME: 'pr-cudf'
     steps:
       - name: Telemetry setup
+        if: ${{ vars.TELEMETRY_ENABLED == 'true' }}
         uses: rapidsai/shared-actions/telemetry-dispatch-stash-base-env-vars@main
   changed-files:
     secrets: inherit
@@ -329,7 +330,7 @@ jobs:
   telemetry-summarize:
     runs-on: ubuntu-latest
     needs: pr-builder
-    if: always()
+    if: ${{ vars.TELEMETRY_ENABLED == 'true' && !cancelled() }}
     continue-on-error: true
     steps:
       - name: Load stashed telemetry env vars

.github/workflows/trigger-breaking-change-alert.yaml

@@ -12,7 +12,7 @@ jobs:
   trigger-notifier:
     if: contains(github.event.pull_request.labels.*.name, 'breaking')
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/breaking-change-alert.yaml@branch-24.12
+    uses: rapidsai/shared-workflows/.github/workflows/breaking-change-alert.yaml@branch-25.02
     with:
       sender_login: ${{ github.event.sender.login }}
       sender_avatar: ${{ github.event.sender.avatar_url }}

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

@@ -19,7 +19,7 @@ dependencies:
 - cramjam
 - cubinlinker
 - cuda-nvtx=11.8
-- cuda-python>=11.7.1,<12.0a0
+- cuda-python>=11.8.5,<12.0a0
 - cuda-sanitizer-api=11.8.86
 - cuda-version=11.8
 - cudatoolkit
@@ -87,7 +87,6 @@ dependencies:
 - s3fs>=2022.3.0
 - scikit-build-core>=0.10.0
 - scipy
-- spdlog>=1.14.1,<1.15
 - sphinx
 - sphinx-autobuild
 - sphinx-copybutton

conda/environments/all_cuda-125_arch-x86_64.yaml

@@ -21,7 +21,7 @@ dependencies:
 - cuda-nvcc
 - cuda-nvrtc-dev
 - cuda-nvtx-dev
-- cuda-python>=12.0,<13.0a0
+- cuda-python>=12.6.2,<13.0a0
 - cuda-sanitizer-api
 - cuda-version=12.5
 - cupy>=12.0.0
@@ -86,7 +86,6 @@ dependencies:
 - s3fs>=2022.3.0
 - scikit-build-core>=0.10.0
 - scipy
-- spdlog>=1.14.1,<1.15
 - sphinx
 - sphinx-autobuild
 - sphinx-copybutton

conda/recipes/cudf/meta.yaml

@@ -91,7 +91,7 @@ requirements:
     - cudatoolkit
     - ptxcompiler >=0.7.0
     - cubinlinker  # CUDA enhanced compatibility.
-    - cuda-python >=11.7.1,<12.0a0
+    - cuda-python >=11.8.5,<12.0a0
     {% else %}
     - cuda-cudart
     - libcufile  # [linux64]
@@ -100,7 +100,7 @@ requirements:
     # TODO: Add nvjitlink here
     # xref: https://github.com/rapidsai/cudf/issues/12822
     - cuda-nvrtc
-    - cuda-python >=12.0,<13.0a0
+    - cuda-python >=12.6.2,<13.0a0
     - pynvjitlink
     {% endif %}
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}

conda/recipes/libcudf/conda_build_config.yaml

@@ -31,9 +31,6 @@ fmt_version:
 flatbuffers_version:
   - "=24.3.25"
 
-spdlog_version:
-  - ">=1.14.1,<1.15"
-
 nvcomp_version:
   - "=4.1.0.6"
 

conda/recipes/libcudf/meta.yaml

@@ -68,7 +68,6 @@ requirements:
     - librdkafka {{ librdkafka_version }}
     - fmt {{ fmt_version }}
     - flatbuffers {{ flatbuffers_version }}
-    - spdlog {{ spdlog_version }}
     - zlib {{ zlib_version }}
 
 outputs:

conda/recipes/pylibcudf/meta.yaml

@@ -83,9 +83,9 @@ requirements:
     - {{ pin_compatible('rmm', max_pin='x.x') }}
     - fsspec >=0.6.0
     {% if cuda_major == "11" %}
-    - cuda-python >=11.7.1,<12.0a0
+    - cuda-python >=11.8.5,<12.0a0
     {% else %}
-    - cuda-python >=12.0,<13.0a0
+    - cuda-python >=12.6.2,<13.0a0
     {% endif %}
     - nvtx >=0.2.1
     - packaging

cpp/CMakeLists.txt

@@ -273,6 +273,14 @@ endif()
 
 # add third party dependencies using CPM
 rapids_cpm_init()
+
+# Not using rapids-cmake since we never want to find, always download.
+CPMAddPackage(
+  NAME rapids_logger GITHUB_REPOSITORY rapidsai/rapids-logger GIT_SHALLOW TRUE GIT_TAG
+  c510947ae9d3a67530cfe3e5eaccb5a3b8ea0e55 VERSION c510947ae9d3a67530cfe3e5eaccb5a3b8ea0e55
+)
+rapids_make_logger(cudf EXPORT_SET cudf-exports)
+
 # find jitify
 include(cmake/thirdparty/get_jitify.cmake)
 # find NVTX
@@ -299,8 +307,6 @@ include(cmake/Modules/JitifyPreprocessKernels.cmake)
 include(cmake/thirdparty/get_kvikio.cmake)
 # find fmt
 include(cmake/thirdparty/get_fmt.cmake)
-# find spdlog
-include(cmake/thirdparty/get_spdlog.cmake)
 # find nanoarrow
 include(cmake/thirdparty/get_nanoarrow.cmake)
 # find thread_pool
@@ -772,7 +778,6 @@ add_library(
   src/utilities/default_stream.cpp
   src/utilities/host_memory.cpp
   src/utilities/linked_column.cpp
-  src/utilities/logger.cpp
   src/utilities/prefetch.cpp
   src/utilities/stacktrace.cpp
   src/utilities/stream_pool.cpp
@@ -910,11 +915,8 @@ if(CUDF_LARGE_STRINGS_DISABLED)
   target_compile_definitions(cudf PRIVATE CUDF_LARGE_STRINGS_DISABLED)
 endif()
 
-# Define RMM logging level
-target_compile_definitions(cudf PRIVATE "RMM_LOGGING_LEVEL=LIBCUDF_LOGGING_LEVEL")
-
-# Define spdlog level
-target_compile_definitions(cudf PUBLIC "SPDLOG_ACTIVE_LEVEL=SPDLOG_LEVEL_${LIBCUDF_LOGGING_LEVEL}")
+# Define logging level
+target_compile_definitions(cudf PRIVATE "CUDF_LOG_ACTIVE_LEVEL=${LIBCUDF_LOGGING_LEVEL}")
 
 # Enable remote IO through KvikIO
 target_compile_definitions(cudf PRIVATE $<$<BOOL:${CUDF_KVIKIO_REMOTE_IO}>:CUDF_KVIKIO_REMOTE_IO>)
@@ -928,14 +930,17 @@ if(TARGET CUDA::cuFile${_cufile_suffix})
   target_compile_definitions(cudf PRIVATE CUDF_CUFILE_FOUND)
 endif()
 
+# Remove this after upgrading to a CCCL that has a proper CMake option. See
+# https://github.com/NVIDIA/cccl/pull/2844
+target_compile_definitions(cudf PRIVATE THRUST_FORCE_32_BIT_OFFSET_TYPE=1)
+
 # Compile stringified JIT sources first
 add_dependencies(cudf jitify_preprocess_run)
 
 # Specify the target module library dependencies
 target_link_libraries(
   cudf
-  PUBLIC CCCL::CCCL rmm::rmm rmm::rmm_logger $<BUILD_LOCAL_INTERFACE:BS::thread_pool>
-         spdlog::spdlog_header_only
+  PUBLIC CCCL::CCCL rmm::rmm rmm::rmm_logger $<BUILD_LOCAL_INTERFACE:BS::thread_pool> cudf_logger
   PRIVATE $<BUILD_LOCAL_INTERFACE:nvtx3::nvtx3-cpp>
           cuco::cuco
           ZLIB::ZLIB
@@ -944,6 +949,7 @@ target_link_libraries(
           $<TARGET_NAME_IF_EXISTS:CUDA::cuFile${_cufile_suffix}>
           nanoarrow
           rmm::rmm_logger_impl
+          cudf_logger_impl
 )
 
 # Add Conda library, and include paths if specified

cpp/benchmarks/io/cuio_common.cpp

@@ -17,7 +17,7 @@
 #include <benchmarks/io/cuio_common.hpp>
 
 #include <cudf/detail/utilities/integer_utils.hpp>
-#include <cudf/detail/utilities/logger.hpp>
+#include <cudf/logger.hpp>
 #include <cudf/utilities/memory_resource.hpp>
 
 #include <rmm/mr/pinned_host_memory_resource.hpp>

cpp/benchmarks/stream_compaction/apply_boolean_mask.cpp

@@ -63,8 +63,8 @@ void apply_boolean_mask_benchmark(nvbench::state& state, nvbench::type_list<Data
   data_profile profile  = data_profile_builder().cardinality(0).no_validity().distribution(
     input_type, distribution_id::UNIFORM, 0, 20);
 
-  auto source_table =
-    create_random_table(cycle_dtypes({input_type}, n_cols), row_count{n_rows}, profile);
+  auto source_table = create_random_table(
+    cycle_dtypes({input_type, cudf::type_id::STRING}, n_cols), row_count{n_rows}, profile);
 
   profile.set_bool_probability_true(percent_true / 100.0);
   profile.set_null_probability(std::nullopt);  // no null mask
@@ -85,6 +85,6 @@ using data_type = nvbench::type_list<int32_t, int64_t, double, cudf::string_view
 NVBENCH_BENCH_TYPES(apply_boolean_mask_benchmark, NVBENCH_TYPE_AXES(data_type))
   .set_name("apply_boolean_mask")
   .set_type_axes_names({"type"})
-  .add_int64_axis("columns", {1, 4})
+  .add_int64_axis("columns", {1, 4, 9})
   .add_int64_axis("rows", {100'000, 1'000'000, 10'000'000})
-  .add_int64_axis("hits_%", {10, 50, 100});
+  .add_int64_axis("hits_%", {10, 20, 50, 80, 90, 100});

cpp/benchmarks/stream_compaction/distinct.cpp

@@ -34,6 +34,7 @@ void nvbench_distinct(nvbench::state& state, nvbench::type_list<Type>)
   cudf::size_type const num_rows    = state.get_int64("NumRows");
   auto const keep                   = get_keep(state.get_string("keep"));
   cudf::size_type const cardinality = state.get_int64("cardinality");
+  auto const null_probability       = state.get_float64("null_probability");
 
   if (cardinality > num_rows) {
     state.skip("cardinality > num_rows");
@@ -42,7 +43,7 @@ void nvbench_distinct(nvbench::state& state, nvbench::type_list<Type>)
 
   data_profile profile = data_profile_builder()
                            .cardinality(cardinality)
-                           .null_probability(0.01)
+                           .null_probability(null_probability)
                            .distribution(cudf::type_to_id<Type>(),
                                          distribution_id::UNIFORM,
                                          static_cast<Type>(0),
@@ -65,6 +66,7 @@ using data_type = nvbench::type_list<int32_t, int64_t>;
 NVBENCH_BENCH_TYPES(nvbench_distinct, NVBENCH_TYPE_AXES(data_type))
   .set_name("distinct")
   .set_type_axes_names({"Type"})
+  .add_float64_axis("null_probability", {0.01})
   .add_string_axis("keep", {"any", "first", "last", "none"})
   .add_int64_axis("cardinality", {100, 100'000, 10'000'000, 1'000'000'000})
   .add_int64_axis("NumRows", {100, 100'000, 10'000'000, 1'000'000'000});

cpp/benchmarks/text/minhash.cpp

@@ -54,9 +54,8 @@ static void bench_minhash(nvbench::state& state)
   state.add_global_memory_writes<nvbench::int32_t>(num_rows);  // output are hashes
 
   state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) {
-    auto result = base64
-                    ? nvtext::minhash64_permuted(input, 0, parameters_a, parameters_b, hash_width)
-                    : nvtext::minhash_permuted(input, 0, parameters_a, parameters_b, hash_width);
+    auto result = base64 ? nvtext::minhash64(input, 0, parameters_a, parameters_b, hash_width)
+                         : nvtext::minhash(input, 0, parameters_a, parameters_b, hash_width);
   });
 }
 

cpp/cmake/thirdparty/patches/cccl_override.json

@@ -3,11 +3,6 @@
   "packages" : {
     "CCCL" : {
       "patches" : [
-        {
-          "file" : "${current_json_dir}/thrust_disable_64bit_dispatching.diff",
-          "issue" : "Remove 64bit dispatching as not needed by libcudf and results in compiling twice as many kernels [https://github.com/rapidsai/cudf/pull/11437]",
-          "fixed_in" : ""
-        },
         {
           "file" : "${current_json_dir}/thrust_faster_sort_compile_times.diff",
           "issue" : "Improve Thrust sort compile times by not unrolling loops for inlined comparators [https://github.com/rapidsai/cudf/pull/10577]",

cpp/doxygen/developer_guide/DEVELOPER_GUIDE.md

@@ -1082,15 +1082,15 @@ initialization. If this setting is higher than the compile-time CMake variable,
 in between the two settings will be excluded from the written log. The available levels are the same
 as for the CMake variable.
 * Global logger object exposed via `cudf::logger()` - sets the minimum logging level at runtime.
-For example, calling `cudf::logger().set_level(spdlog::level::err)`, will exclude any messages that
+For example, calling `cudf::default_logger().set_level(level_enum::err)`, will exclude any messages that
 are not errors or critical errors. This API should not be used within libcudf to manipulate logging,
 its purpose is to allow upstream users to configure libcudf logging to fit their application.
 
 By default, logging messages are output to stderr.
 Setting the environment variable `LIBCUDF_DEBUG_LOG_FILE` redirects the log to a file with the
 specified path (can be relative to the current directory).
-Upstream users can also manipulate `cudf::logger().sinks()` to add sinks or divert the log to
-standard output or even a custom spdlog sink.
+Upstream users can also manipulate `cudf::default_logger().sinks()` to add sinks or divert the log to
+standard output.
 
 # Data Types