Merge branch 'branch-25.02' into add-aarch64

wphicks · web-flow · commit c16906ca38ce · 2025-01-21T09:59:51.000-05:00
diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -75,7 +75,7 @@ dependencies:
 - sphinx-copybutton
 - sphinx-markdown-tables
 - statsmodels
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 - treelite==4.3.0
 - umap-learn==0.5.6
 - xgboost>=2.1.0
diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml
@@ -25,7 +25,7 @@ dependencies:
 - dask-ml
 - doxygen=1.9.1
 - fmt>=11.0.2,<12
-- gcc_linux-64=11.*
+- gcc_linux-64=13.*
 - graphviz
 - hdbscan>=0.8.39,<0.8.40
 - hypothesis>=6.0,<7
@@ -71,7 +71,7 @@ dependencies:
 - sphinx-copybutton
 - sphinx-markdown-tables
 - statsmodels
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 - treelite==4.3.0
 - umap-learn==0.5.6
 - xgboost>=2.1.0
diff --git a/conda/environments/clang_tidy_cuda-118_arch-x86_64.yaml b/conda/environments/clang_tidy_cuda-118_arch-x86_64.yaml
@@ -33,6 +33,6 @@ dependencies:
 - ninja
 - nvcc_linux-64=11.8
 - spdlog>=1.14.1,<1.15
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 - tomli
 name: clang_tidy_cuda-118_arch-x86_64
diff --git a/conda/environments/cpp_all_cuda-118_arch-x86_64.yaml b/conda/environments/cpp_all_cuda-118_arch-x86_64.yaml
@@ -31,5 +31,5 @@ dependencies:
 - ninja
 - nvcc_linux-64=11.8
 - spdlog>=1.14.1,<1.15
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 name: cpp_all_cuda-118_arch-x86_64
diff --git a/conda/environments/cpp_all_cuda-125_arch-x86_64.yaml b/conda/environments/cpp_all_cuda-125_arch-x86_64.yaml
@@ -15,7 +15,7 @@ dependencies:
 - cuda-version=12.5
 - cxx-compiler
 - fmt>=11.0.2,<12
-- gcc_linux-64=11.*
+- gcc_linux-64=13.*
 - libcublas-dev
 - libcufft-dev
 - libcumlprims==25.2.*,>=0.0.0a0
@@ -27,5 +27,5 @@ dependencies:
 - librmm==25.2.*,>=0.0.0a0
 - ninja
 - spdlog>=1.14.1,<1.15
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 name: cpp_all_cuda-125_arch-x86_64
diff --git a/conda/recipes/cuml-cpu/conda_build_config.yaml b/conda/recipes/cuml-cpu/conda_build_config.yaml
@@ -1,8 +1,10 @@
 c_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cxx_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cmake_version:
   - ">=3.26.4,!=3.30.0"
@@ -11,4 +13,4 @@ c_stdlib:
   - sysroot
 
 c_stdlib_version:
-  - "=2.17"
+  - "=2.28"
diff --git a/conda/recipes/cuml/conda_build_config.yaml b/conda/recipes/cuml/conda_build_config.yaml
@@ -1,14 +1,14 @@
 c_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cxx_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cuda_compiler:
-  - cuda-nvcc
-
-cuda11_compiler:
-  - nvcc
+  - cuda-nvcc  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - nvcc  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cmake_version:
   - ">=3.26.4,!=3.30.0"
@@ -17,7 +17,7 @@ c_stdlib:
   - sysroot
 
 c_stdlib_version:
-  - "=2.17"
+  - "=2.28"
 
 treelite_version:
   - "=4.3.0"
diff --git a/conda/recipes/cuml/meta.yaml b/conda/recipes/cuml/meta.yaml
@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -33,10 +33,8 @@ build:
     - SCCACHE_S3_KEY_PREFIX=cuml-linux64 # [linux64]
     - SCCACHE_S3_USE_SSL
   ignore_run_exports_from:
-    {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }}
-    {% else %}
     - {{ compiler('cuda') }}
+    {% if cuda_major != "11" %}
     - cuda-cudart-dev
     {% endif %}
     - cuda-python
@@ -46,7 +44,7 @@ requirements:
     - {{ compiler('c') }}
     - {{ compiler('cxx') }}
     {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }} ={{ cuda_version }}
+    - {{ compiler('cuda') }} ={{ cuda_version }}
     {% else %}
     - {{ compiler('cuda') }}
     {% endif %}
diff --git a/conda/recipes/libcuml/conda_build_config.yaml b/conda/recipes/libcuml/conda_build_config.yaml
@@ -1,20 +1,20 @@
 c_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cxx_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cuda_compiler:
-  - cuda-nvcc
-
-cuda11_compiler:
-  - nvcc
+  - cuda-nvcc  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - nvcc  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 c_stdlib:
   - sysroot
 
 c_stdlib_version:
-  - "=2.17"
+  - "=2.28"
 
 cmake_version:
   - ">=3.26.4,!=3.30.0"
diff --git a/conda/recipes/libcuml/meta.yaml b/conda/recipes/libcuml/meta.yaml
@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -14,9 +14,7 @@ source:
 
 build:
   ignore_run_exports_from:
-    {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }}
-    {% endif %}
+    - {{ compiler('cuda') }}
   script_env:
     - AWS_ACCESS_KEY_ID
     - AWS_SECRET_ACCESS_KEY
@@ -38,7 +36,7 @@ requirements:
     - {{ compiler('c') }}
     - {{ compiler('cxx') }}
     {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }} ={{ cuda_version }}
+    - {{ compiler('cuda') }} ={{ cuda_version }}
     {% else %}
     - {{ compiler('cuda') }}
     {% endif %}
@@ -84,10 +82,8 @@ outputs:
       number: {{ GIT_DESCRIBE_NUMBER }}
       string: cuda{{ cuda_major }}_{{ date_string }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
       ignore_run_exports_from:
-        {% if cuda_major == "11" %}
-        - {{ compiler('cuda11') }}
-        {% else %}
         - {{ compiler('cuda') }}
+        {% if cuda_major != "11" %}
         - cuda-cudart-dev
         {% endif %}
     requirements:
@@ -131,10 +127,8 @@ outputs:
       number: {{ GIT_DESCRIBE_NUMBER }}
       string: cuda{{ cuda_major }}_{{ date_string }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
       ignore_run_exports_from:
-        {% if cuda_major == "11" %}
-        - {{ compiler('cuda11') }}
-        {% else %}
         - {{ compiler('cuda') }}
+        {% if cuda_major != "11" %}
         - cuda-cudart-dev
         {% endif %}
     requirements:
diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
@@ -422,6 +422,27 @@ if(BUILD_CUML_CPP_LIBRARY)
         src/hdbscan/hdbscan.cu
         src/hdbscan/condensed_hierarchy.cu
         src/hdbscan/prediction_data.cu)
+
+    # When using GCC 13, some maybe-uninitialized warnings appear from CCCL and are treated as errors.
+    # See this issue: https://github.com/rapidsai/cuml/issues/6225
+    set_property(
+      SOURCE src/hdbscan/condensed_hierarchy.cu
+      APPEND_STRING
+      PROPERTY COMPILE_FLAGS
+      " -Xcompiler=-Wno-maybe-uninitialized"
+    )
+    set_property(
+      SOURCE src/hdbscan/hdbscan.cu
+      APPEND_STRING
+      PROPERTY COMPILE_FLAGS
+      " -Xcompiler=-Wno-maybe-uninitialized"
+    )
+    set_property(
+      SOURCE src/hdbscan/prediction_data.cu
+      APPEND_STRING
+      PROPERTY COMPILE_FLAGS
+      " -Xcompiler=-Wno-maybe-uninitialized"
+    )
   endif()
 
   if(all_algo OR holtwinters_algo)
diff --git a/cpp/src/hdbscan/condensed_hierarchy.cu b/cpp/src/hdbscan/condensed_hierarchy.cu
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2024, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2025, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -27,6 +27,7 @@
 
 #include <cub/cub.cuh>
 #include <cuda/functional>
+#include <cuda/std/functional>
 #include <thrust/copy.h>
 #include <thrust/device_ptr.h>
 #include <thrust/execution_policy.h>
@@ -157,8 +158,9 @@ void CondensedHierarchy<value_idx, value_t>::condense(value_idx* full_parents,
     thrust::cuda::par.on(stream),
     full_sizes,
     full_sizes + size,
-    cuda::proclaim_return_type<bool>([=] __device__(value_idx a) -> bool { return a != -1; }),
-    0,
+    cuda::proclaim_return_type<value_idx>(
+      [=] __device__(value_idx a) -> value_idx { return static_cast<value_idx>(a != -1); }),
+    static_cast<value_idx>(0),
     thrust::plus<value_idx>());
 
   parents.resize(n_edges, stream);
diff --git a/cpp/src/hdbscan/detail/utils.h b/cpp/src/hdbscan/detail/utils.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021-2024, NVIDIA CORPORATION.
+ * Copyright (c) 2021-2025, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -34,6 +34,7 @@
 #include <rmm/exec_policy.hpp>
 
 #include <cub/cub.cuh>
+#include <cuda/functional>
 #include <thrust/copy.h>
 #include <thrust/execution_policy.h>
 #include <thrust/for_each.h>
@@ -114,8 +115,9 @@ Common::CondensedHierarchy<value_idx, value_t> make_cluster_tree(
     thrust_policy,
     sizes,
     sizes + condensed_tree.get_n_edges(),
-    cuda::proclaim_return_type<bool>([=] __device__(value_idx a) -> bool { return a > 1; }),
-    0,
+    cuda::proclaim_return_type<value_idx>(
+      [=] __device__(value_idx a) -> value_idx { return static_cast<value_idx>(a > 1); }),
+    static_cast<value_idx>(0),
     thrust::plus<value_idx>());
 
   // remove leaves from condensed tree
diff --git a/cpp/test/CMakeLists.txt b/cpp/test/CMakeLists.txt
@@ -1,5 +1,5 @@
 #=============================================================================
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -149,6 +149,14 @@ if("${CMAKE_CUDA_COMPILER_VERSION}" VERSION_GREATER_EQUAL "11.2")
     # An HDBSCAN gtest is failing w/ CUDA 11.2 for some reason.
     if(all_algo OR hdbscan_algo)
       ConfigureTest(PREFIX SG NAME HDBSCAN_TEST  sg/hdbscan_test.cu ML_INCLUDE)
+      # When using GCC 13, some maybe-uninitialized warnings appear from CCCL and are treated as errors.
+      # See this issue: https://github.com/rapidsai/cuml/issues/6225
+      set_property(
+        SOURCE sg/hdbscan_test.cu
+        APPEND_STRING
+        PROPERTY COMPILE_FLAGS
+        " -Xcompiler=-Wno-maybe-uninitialized"
+      )
     endif()
 endif()
 
diff --git a/dependencies.yaml b/dependencies.yaml
@@ -190,29 +190,37 @@ dependencies:
           - matrix:
               arch: x86_64
             packages:
-              - gcc_linux-64=11.*
-              - sysroot_linux-64==2.17
+              - sysroot_linux-64==2.28
           - matrix:
               arch: aarch64
             packages:
-              - gcc_linux-aarch64=11.*
-              - sysroot_linux-aarch64==2.17
+              - sysroot_linux-aarch64==2.28
       - output_types: conda
         matrices:
           - matrix:
               arch: x86_64
               cuda: "11.8"
             packages:
+              - gcc_linux-64=11.*
               - nvcc_linux-64=11.8
           - matrix:
               arch: aarch64
               cuda: "11.8"
             packages:
+              - gcc_linux-64=11.*
               - nvcc_linux-aarch64=11.8
           - matrix:
+              arch: x86_64
+              cuda: "12.*"
+            packages:
+              - cuda-nvcc
+              - gcc_linux-64=13.*
+          - matrix:
+              arch: aarch64
               cuda: "12.*"
             packages:
               - cuda-nvcc
+              - gcc_linux-aarch64=13.*
   py_build_cuml:
     common:
       - output_types: [conda, requirements, pyproject]
diff --git a/python/cuml/cuml/internals/logger.pxd b/python/cuml/cuml/internals/logger.pxd
@@ -47,6 +47,7 @@ IF GPUBUILD == 1:
         cdef cppclass sink_vector:
             void push_back(const sink_ptr& sink) except +
             void pop_back() except +
+            void clear()
 
     cdef extern from "cuml/common/logger.hpp" namespace "ML" nogil:
         cdef cppclass logger:
diff --git a/python/cuml/cuml/internals/logger.pyx b/python/cuml/cuml/internals/logger.pyx
@@ -319,5 +319,6 @@ def flush():
 
 
 IF GPUBUILD == 1:
-    # Set callback functions to handle redirected sys.stdout in Python
+    # Clear existing sinks and add a callback sink to redirect to sys.stdout
+    default_logger().sinks().clear()
     default_logger().sinks().push_back(<sink_ptr> make_shared[callback_sink_mt](_log_callback, _log_flush))
