[INFRA] Import NVIDIA/CCCL upstream as optimization reference library

CCCL (CUDA C++ Core Libraries) provides:
- CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk)
- Thrust: high-level parallel algorithms (transform_reduce, sort, scan)
- libcudacxx: CUDA C++ standard library (atomics, barriers, memory)
- cudax: experimental features (memory resources, allocators)
- Tuning policies: per-SM hardware-specific algorithm parameters

Competition optimization vectors mapped to CCCL:
- Output TPS (83% weight): warp_reduce, block_reduce, device_topk
- Input TPS (14% weight): device_scan, block_load, prefetch
- Cache TPS (3% weight): prefix caching strategy patterns
- Memory (0.9 util): pooled/cached/buddy allocators

Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only)
License: Apache-2.0

cccl_upstream/cudax/test/group/mapping/composite_mapping.cu

@@ -0,0 +1,196 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of CUDA Experimental in CUDA C++ Core Libraries,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#include <cuda/devices>
+#include <cuda/hierarchy>
+#include <cuda/launch>
+#include <cuda/std/cstddef>
+#include <cuda/std/numeric>
+#include <cuda/std/tuple>
+#include <cuda/std/type_traits>
+#include <cuda/std/utility>
+#include <cuda/stream>
+#include <cuda/warp>
+
+#include <cuda/experimental/group.cuh>
+
+#include "group_testing.cuh"
+
+namespace
+{
+template <class Mapping1, class Mapping2, class Config>
+__device__ void test_composite_mapping(const Mapping1& mapping1, const Mapping2& mapping2, Config config)
+{
+  using Mapping = cudax::composite_mapping<Mapping1, Mapping2>;
+
+  // Test construction from 2 mappings.
+  {
+    cudax::composite_mapping mapping{mapping1, mapping2};
+    static_assert(cuda::std::is_same_v<decltype(mapping), Mapping>);
+    static_assert(cuda::std::is_nothrow_constructible_v<Mapping, Mapping1, Mapping2>
+                  == (cuda::std::is_nothrow_copy_constructible_v<Mapping1>
+                      && cuda::std::is_nothrow_copy_constructible_v<Mapping2>) );
+  }
+
+  // Test get().
+  {
+    const cudax::composite_mapping mapping{mapping1, mapping2};
+    static_assert(cuda::std::is_same_v<decltype(mapping.get()), const cuda::std::tuple<Mapping1, Mapping2>&>);
+    static_assert(noexcept(mapping.get()));
+
+    const auto& mapping1_ref = cuda::std::get<0>(mapping.get());
+    CHECK(mapping1_ref.unit_count() == 4);
+
+    const auto& mapping2_ref = cuda::std::get<1>(mapping.get());
+    CHECK(mapping2_ref.unit_count(0) == 1);
+    CHECK(mapping2_ref.unit_count(1) == 3);
+  }
+
+  // Test map(...).
+  {
+    const cudax::this_warp parent_group{config};
+    const ThreadsInWarpMappingResult prev_mapping_result;
+    const cudax::composite_mapping mapping{mapping1, mapping2};
+
+    static_assert(
+      cudax::__group_mapping_result<decltype(mapping.map(cuda::gpu_thread, parent_group, prev_mapping_result))>);
+
+    auto result  = mapping.map(cuda::gpu_thread, parent_group, prev_mapping_result);
+    using Result = decltype(result);
+
+    const auto rank_in_warp = cuda::gpu_thread.rank_as<unsigned>(parent_group);
+
+    if constexpr (Mapping1::static_unit_count() != cuda::std::dynamic_extent
+                  && Mapping2::static_group_count() != cuda::std::dynamic_extent)
+    {
+      static_assert(Result::static_group_count() == 16);
+    }
+    else
+    {
+      static_assert(Result::static_group_count() == cuda::std::dynamic_extent);
+    }
+    CHECK(result.group_count() == 16);
+    CHECK(result.group_rank() == (rank_in_warp / 4 * 2 + (rank_in_warp % 4 > 0)));
+
+    static_assert(Result::static_unit_count() == cuda::std::dynamic_extent);
+    CHECK(result.unit_count() == ((rank_in_warp % 4 > 0) ? 3 : 1));
+    CHECK(result.unit_rank() == ((rank_in_warp % 4 > 0) ? (rank_in_warp % 4 - 1) : 0));
+
+    const auto lane_mask_ref = ((rank_in_warp % 4 > 0) ? 0b1110u : 0b0001u) << ((rank_in_warp / 4) * 4);
+    CHECK(result.lane_mask() == cuda::device::lane_mask{lane_mask_ref});
+
+    CHECK(result.is_valid());
+    static_assert(Result::is_always_exhaustive());
+    static_assert(Result::is_always_contiguous());
+  }
+
+  // Test operator|.
+  {
+    auto mapping = mapping1 | mapping2;
+
+    static_assert(cuda::std::is_same_v<Mapping, decltype(mapping)>);
+    static_assert(noexcept(mapping1 | mapping2));
+
+    const auto& mapping1_ref = cuda::std::get<0>(mapping.get());
+    CHECK(mapping1_ref.unit_count() == 4);
+
+    const auto& mapping2_ref = cuda::std::get<1>(mapping.get());
+    CHECK(mapping2_ref.unit_count(0) == 1);
+    CHECK(mapping2_ref.unit_count(1) == 3);
+  }
+  {
+    auto mapping = cudax::composite_mapping{mapping1} | mapping2;
+
+    static_assert(cuda::std::is_same_v<Mapping, decltype(mapping)>);
+    static_assert(noexcept(cudax::composite_mapping{mapping1} | mapping2));
+
+    const auto& mapping1_ref = cuda::std::get<0>(mapping.get());
+    CHECK(mapping1_ref.unit_count() == 4);
+
+    const auto& mapping2_ref = cuda::std::get<1>(mapping.get());
+    CHECK(mapping2_ref.unit_count(0) == 1);
+    CHECK(mapping2_ref.unit_count(1) == 3);
+  }
+  {
+    auto mapping = mapping1 | cudax::composite_mapping{mapping2};
+
+    static_assert(cuda::std::is_same_v<Mapping, decltype(mapping)>);
+    static_assert(noexcept(mapping1 | cudax::composite_mapping{mapping2}));
+
+    const auto& mapping1_ref = cuda::std::get<0>(mapping.get());
+    CHECK(mapping1_ref.unit_count() == 4);
+
+    const auto& mapping2_ref = cuda::std::get<1>(mapping.get());
+    CHECK(mapping2_ref.unit_count(0) == 1);
+    CHECK(mapping2_ref.unit_count(1) == 3);
+  }
+  {
+    auto mapping = cudax::composite_mapping{mapping1} | cudax::composite_mapping{mapping2};
+
+    static_assert(cuda::std::is_same_v<Mapping, decltype(mapping)>);
+    static_assert(noexcept(cudax::composite_mapping{mapping1} | cudax::composite_mapping{mapping2}));
+
+    const auto& mapping1_ref = cuda::std::get<0>(mapping.get());
+    CHECK(mapping1_ref.unit_count() == 4);
+
+    const auto& mapping2_ref = cuda::std::get<1>(mapping.get());
+    CHECK(mapping2_ref.unit_count(0) == 1);
+    CHECK(mapping2_ref.unit_count(1) == 3);
+  }
+}
+
+struct TestKernel
+{
+  template <class Config>
+  __device__ void operator()(const Config& config)
+  {
+    {
+      const cudax::group_by<4> mapping1{};
+      const cudax::group_as mapping2{cuda::std::integer_sequence<cuda::std::size_t, 1, 3>{}};
+      test_composite_mapping(mapping1, mapping2, config);
+    }
+    {
+      const cudax::group_by mapping1{4};
+      const cudax::group_as mapping2{cuda::std::integer_sequence<cuda::std::size_t, 1, 3>{}};
+      test_composite_mapping(mapping1, mapping2, config);
+    }
+    {
+      const cudax::group_by<4> mapping1{};
+      constexpr unsigned counts2[]{1, 3};
+      const cudax::group_as mapping2{counts2};
+      test_composite_mapping(mapping1, mapping2, config);
+    }
+    {
+      const cudax::group_by mapping1{4};
+      constexpr unsigned counts2[]{1, 3};
+      const cudax::group_as mapping2{counts2};
+      test_composite_mapping(mapping1, mapping2, config);
+    }
+  }
+};
+} // namespace
+
+C2H_TEST("Composite mapping", "[group]")
+{
+  const auto device = cuda::devices[0];
+
+  const cuda::stream stream{device};
+
+  {
+    const auto config = cuda::make_config(cuda::grid_dims<1>(), cuda::block_dims<8, 4>());
+    cuda::launch(stream, config, TestKernel{});
+  }
+  {
+    const auto config = cuda::make_config(cuda::grid_dims<1>(), cuda::block_dims(dim3{8, 4}));
+    cuda::launch(stream, config, TestKernel{});
+  }
+
+  stream.sync();
+}