Merge branch 'main' into test-mr-pr/01-foundation

Author: bdice

conda/environments/all_cuda-129_arch-aarch64.yaml

@@ -17,12 +17,12 @@ dependencies:
 - cmake>=4.0
 - cramjam
 - croaring==4.4.2
+- cuda-bindings>=12.9.2,<13.0
 - cuda-cudart-dev
 - cuda-cupti-dev
 - cuda-nvcc
 - cuda-nvrtc-dev
 - cuda-nvtx-dev
-- cuda-python>=12.9.2,<13.0
 - cuda-sanitizer-api
 - cuda-version=12.9
 - cupy>=14.0.1,!=14.1.0

conda/environments/all_cuda-129_arch-x86_64.yaml

@@ -17,12 +17,12 @@ dependencies:
 - cmake>=4.0
 - cramjam
 - croaring==4.4.2
+- cuda-bindings>=12.9.2,<13.0
 - cuda-cudart-dev
 - cuda-cupti-dev
 - cuda-nvcc
 - cuda-nvrtc-dev
 - cuda-nvtx-dev
-- cuda-python>=12.9.2,<13.0
 - cuda-sanitizer-api
 - cuda-version=12.9
 - cupy>=14.0.1,!=14.1.0

conda/environments/all_cuda-133_arch-aarch64.yaml

@@ -17,12 +17,12 @@ dependencies:
 - cmake>=4.0
 - cramjam
 - croaring==4.4.2
+- cuda-bindings>=13.0.1,<14.0
 - cuda-cudart-dev
 - cuda-cupti-dev
 - cuda-nvcc
 - cuda-nvrtc-dev
 - cuda-nvtx-dev
-- cuda-python>=13.0.1,<14.0
 - cuda-sanitizer-api
 - cuda-version=13.3
 - cupy>=14.0.1,!=14.1.0

conda/environments/all_cuda-133_arch-x86_64.yaml

@@ -17,12 +17,12 @@ dependencies:
 - cmake>=4.0
 - cramjam
 - croaring==4.4.2
+- cuda-bindings>=13.0.1,<14.0
 - cuda-cudart-dev
 - cuda-cupti-dev
 - cuda-nvcc
 - cuda-nvrtc-dev
 - cuda-nvtx-dev
-- cuda-python>=13.0.1,<14.0
 - cuda-sanitizer-api
 - cuda-version=13.3
 - cupy>=14.0.1,!=14.1.0

conda/recipes/cudf-polars/recipe.yaml

@@ -42,8 +42,8 @@ requirements:
     - packaging
     - ${{ pin_compatible("cuda-version", upper_bound="x", lower_bound="x") }}
     - if: cuda_major == "12"
-      then: cuda-python >=12.9.2,<13.0
-      else: cuda-python >=13.0.1,<14.0
+      then: cuda-bindings >=12.9.2,<13.0
+      else: cuda-bindings >=13.0.1,<14.0
   run_constraints:
     - ray-default >=2.55.1
   ignore_run_exports:

conda/recipes/cudf/recipe.yaml

@@ -102,8 +102,8 @@ requirements:
     - ${{ pin_compatible("rmm", upper_bound="x.x") }}
     - fsspec >=0.6.0
     - if: cuda_major == "12"
-      then: cuda-python >=12.9.2,<13.0
-      else: cuda-python >=13.0.1,<14.0
+      then: cuda-bindings >=12.9.2,<13.0
+      else: cuda-bindings >=13.0.1,<14.0
     - if: linux and x86_64
       then:
         - libcufile

conda/recipes/pylibcudf/recipe.yaml

@@ -90,8 +90,8 @@ requirements:
     - libcudf =${{ version }}
     - ${{ pin_compatible("rmm", upper_bound="x.x") }}
     - if: cuda_major == "12"
-      then: cuda-python >=12.9.2,<13.0
-      else: cuda-python >=13.0.1,<14.0
+      then: cuda-bindings >=12.9.2,<13.0
+      else: cuda-bindings >=13.0.1,<14.0
     - nvtx >=0.2.1
   run_constraints:
     - numpy >=2.0,<3.0

cpp/benchmarks/reshape/interleave.cpp

@@ -1,13 +1,12 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
 #include <benchmarks/common/generate_input.hpp>
 #include <benchmarks/common/memory_stats.hpp>
 
 #include <cudf/reshape.hpp>
-#include <cudf/strings/strings_column_view.hpp>
 #include <cudf/utilities/default_stream.hpp>
 
 #include <nvbench/nvbench.cuh>
@@ -33,10 +32,9 @@ static void bench_interleave(nvbench::state& state)
   auto const stream      = cudf::get_default_stream();
 
   state.set_cuda_stream(nvbench::make_cuda_stream_view(stream.value()));
-  auto chars_size = cudf::strings_column_view(source_view.column(0)).chars_size(stream) +
-                    cudf::strings_column_view(source_view.column(1)).chars_size(stream);
-  state.add_global_memory_reads<nvbench::int8_t>(chars_size);   // all bytes are read
-  state.add_global_memory_writes<nvbench::int8_t>(chars_size);  // all bytes are written
+  auto input_bytes = source_table->alloc_size();
+  state.add_global_memory_reads<nvbench::int8_t>(input_bytes);   // all bytes are read
+  state.add_global_memory_writes<nvbench::int8_t>(input_bytes);  // all bytes are written
 
   auto const mem_stats_logger = cudf::memory_stats_logger();
   state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) {

cpp/libcudf_streaming/benchmarks/bench_shuffle.cpp

@@ -8,6 +8,7 @@
 #include <rapidsmpf/bootstrap/bootstrap.hpp>
 #include <rapidsmpf/bootstrap/utils.hpp>
 #include <rapidsmpf/communicator/communicator.hpp>
+#include <rapidsmpf/communicator/logger.hpp>
 #include <rapidsmpf/error.hpp>
 #include <rapidsmpf/memory/spill.hpp>
 #include <rapidsmpf/nvtx.hpp>
@@ -516,6 +517,8 @@ int main(int argc, char** argv)
   auto& stat_enabled_mr = br->device_mr_adaptor();
   rmm::mr::set_current_device_resource(stat_enabled_mr);
 
+  auto log = rapidsmpf::Logger::from_options(options);
+
   std::shared_ptr<rapidsmpf::Communicator> comm;
   auto progress_thread = std::make_shared<rapidsmpf::ProgressThread>(stats);
   if (args.comm_type == "mpi") {
@@ -527,7 +530,7 @@ int main(int argc, char** argv)
       return 1;
     }
     rapidsmpf::mpi::init(&argc, &argv);
-    comm = std::make_shared<rapidsmpf::MPI>(MPI_COMM_WORLD, options, progress_thread);
+    comm = std::make_shared<rapidsmpf::MPI>(MPI_COMM_WORLD, progress_thread, log);
 #else
     std::cerr << "Error: MPI communicator is not available in this build." << std::endl;
     return 1;
@@ -537,11 +540,11 @@ int main(int argc, char** argv)
     if (use_bootstrap) {
       // Launched with rrun - use bootstrap backend
       comm = rapidsmpf::bootstrap::create_ucxx_comm(
-        progress_thread, rapidsmpf::bootstrap::BackendType::AUTO, options);
+        progress_thread, rapidsmpf::bootstrap::BackendType::AUTO, options, log);
     } else {
 #ifdef CUDF_STREAMING_HAVE_MPI
       // Launched with mpirun - use MPI bootstrap
-      comm = rapidsmpf::ucxx::init_using_mpi(MPI_COMM_WORLD, options, progress_thread);
+      comm = rapidsmpf::ucxx::init_using_mpi(MPI_COMM_WORLD, options, progress_thread, log);
 #else
       std::cerr << "Error: UCXX without MPI support requires bootstrap mode." << std::endl;
       return 1;
@@ -558,7 +561,6 @@ int main(int argc, char** argv)
 
   args.pprint(*comm);
 
-  auto& log                    = comm->logger();
   rmm::cuda_stream_view stream = cudf::get_default_stream();
 
   // Print benchmark/hardware info.

cpp/libcudf_streaming/benchmarks/streaming/bench_streaming_shuffle.cpp

@@ -15,6 +15,7 @@
 #include <rapidsmpf/bootstrap/bootstrap.hpp>
 #include <rapidsmpf/bootstrap/utils.hpp>
 #include <rapidsmpf/communicator/communicator.hpp>
+#include <rapidsmpf/communicator/logger.hpp>
 #include <rapidsmpf/error.hpp>
 #include <rapidsmpf/nvtx.hpp>
 #include <rapidsmpf/shuffler/shuffler.hpp>
@@ -276,6 +277,7 @@ int main(int argc, char** argv)
 
   // Initialize configuration options from environment variables.
   rapidsmpf::config::Options options{rapidsmpf::config::get_environment_variables()};
+  auto log             = rapidsmpf::Logger::from_options(options);
   auto progress_thread = std::make_shared<rapidsmpf::ProgressThread>();
 
   std::shared_ptr<rapidsmpf::Communicator> comm;
@@ -288,7 +290,7 @@ int main(int argc, char** argv)
       return 1;
     }
     rapidsmpf::mpi::init(&argc, &argv);
-    comm = std::make_shared<rapidsmpf::MPI>(MPI_COMM_WORLD, options, progress_thread);
+    comm = std::make_shared<rapidsmpf::MPI>(MPI_COMM_WORLD, progress_thread, log);
 #else
     std::cerr << "Error: MPI communicator is not available in this build." << std::endl;
     return 1;
@@ -298,11 +300,11 @@ int main(int argc, char** argv)
     if (use_bootstrap) {
       // Launched with rrun - use bootstrap backend
       comm = rapidsmpf::bootstrap::create_ucxx_comm(
-        progress_thread, rapidsmpf::bootstrap::BackendType::AUTO, options);
+        progress_thread, rapidsmpf::bootstrap::BackendType::AUTO, options, log);
     } else {
 #ifdef CUDF_STREAMING_HAVE_MPI
       // Launched with mpirun - use MPI bootstrap
-      comm = rapidsmpf::ucxx::init_using_mpi(MPI_COMM_WORLD, options, progress_thread);
+      comm = rapidsmpf::ucxx::init_using_mpi(MPI_COMM_WORLD, options, progress_thread, log);
 #else
       std::cerr << "Error: UCXX without MPI support requires bootstrap mode." << std::endl;
       return 1;
@@ -344,7 +346,6 @@ int main(int argc, char** argv)
   auto& stat_enabled_mr = br->device_mr_adaptor();
   rmm::mr::set_current_device_resource(stat_enabled_mr);
 
-  auto& log                    = *comm->logger();
   rmm::cuda_stream_view stream = cudf::get_default_stream();
 
   // Print benchmark/hardware info.
@@ -361,7 +362,7 @@ int main(int argc, char** argv)
     ss << "    PCI Bus ID: " << pci_bus_id.substr(0, pci_bus_id.find('\0')) << "\n";
     ss << "    Total Memory: " << rapidsmpf::format_nbytes(properties.totalGlobalMem, 0) << "\n";
     ss << "  Comm: " << *comm << "\n";
-    log.print(ss.str());
+    log->print(ss.str());
   }
 
   auto ctx = std::make_shared<rapidsmpf::streaming::Context>(options, comm->logger(), br);
@@ -378,7 +379,7 @@ int main(int argc, char** argv)
        << "/s | global throughput: " << rapidsmpf::format_nbytes(args.total_nbytes / elapsed)
        << "/s";
     if (i < args.num_warmups) { ss << " (warmup run)"; }
-    log.print(ss.str());
+    log->print(ss.str());
     if (i >= args.num_warmups) { elapsed_vec.push_back(elapsed); }
   }
 
@@ -417,18 +418,18 @@ int main(int argc, char** argv)
          << rapidsmpf::format_nbytes(record.total() / static_cast<std::int64_t>(total_num_runs))
          << " (avg)";
     }
-    log.print(ss.str());
+    log->print(ss.str());
   }
 
   auto statistics = ctx->statistics();
   if (args.enable_memory_profiler) {
-    log.print(statistics->report({
+    log->print(statistics->report({
       .mr        = stat_enabled_mr,
       .pinned_mr = pinned_mr,
       .header    = "Statistics (of the last run):",
     }));
   } else {
-    log.print(statistics->report({.header = "Statistics (of the last run):"}));
+    log->print(statistics->report({.header = "Statistics (of the last run):"}));
   }
 
 #ifdef CUDF_STREAMING_HAVE_MPI