[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/thrust/testing/cpp/CMakeLists.txt

@@ -0,0 +1,20 @@
+file(
+  GLOB test_srcs
+  RELATIVE "${CMAKE_CURRENT_LIST_DIR}"
+  CONFIGURE_DEPENDS
+  *.cu
+  *.cpp
+)
+
+foreach (thrust_target IN LISTS THRUST_TARGETS)
+  thrust_get_target_property(config_device ${thrust_target} DEVICE)
+  if (NOT config_device STREQUAL "CPP")
+    continue()
+  endif()
+
+  foreach (test_src IN LISTS test_srcs)
+    get_filename_component(test_name "${test_src}" NAME_WLE)
+    string(PREPEND test_name "cpp.")
+    thrust_add_test(test_target ${test_name} "${test_src}" ${thrust_target})
+  endforeach()
+endforeach()

cccl_upstream/thrust/testing/cpp/adjacent_difference.cu

@@ -0,0 +1,72 @@
+#include <thrust/adjacent_difference.h>
+#include <thrust/device_free.h>
+#include <thrust/device_malloc.h>
+#include <thrust/execution_policy.h>
+
+#include <unittest/unittest.h>
+
+struct detect_wrong_difference
+{
+  using difference_type   = void;
+  using value_type        = long long;
+  using pointer           = void;
+  using reference         = detect_wrong_difference;
+  using iterator_category = ::cuda::std::output_iterator_tag;
+
+  bool* flag;
+
+  _CCCL_HOST_DEVICE detect_wrong_difference operator++() const
+  {
+    return *this;
+  }
+  _CCCL_HOST_DEVICE detect_wrong_difference operator*() const
+  {
+    return *this;
+  }
+  template <typename Difference>
+  _CCCL_HOST_DEVICE detect_wrong_difference operator+(Difference) const
+  {
+    return *this;
+  }
+  template <typename Index>
+  _CCCL_HOST_DEVICE detect_wrong_difference operator[](Index) const
+  {
+    return *this;
+  }
+
+  _CCCL_DEVICE void operator=(long long difference) const
+  {
+    if (difference != 1)
+    {
+      *flag = false;
+    }
+  }
+};
+
+void TestAdjacentDifferenceWithBigIndexesHelper(int magnitude)
+{
+  thrust::counting_iterator<long long> begin(1);
+  thrust::counting_iterator<long long> end = begin + (1ll << magnitude);
+  ASSERT_EQUAL(::cuda::std::distance(begin, end), 1ll << magnitude);
+
+  thrust::device_ptr<bool> all_differences_correct = thrust::device_malloc<bool>(1);
+  *all_differences_correct                         = true;
+
+  detect_wrong_difference out = {thrust::raw_pointer_cast(all_differences_correct)};
+
+  thrust::adjacent_difference(thrust::device, begin, end, out);
+
+  bool all_differences_correct_h = *all_differences_correct;
+  thrust::device_free(all_differences_correct);
+
+  ASSERT_EQUAL(all_differences_correct_h, true);
+}
+
+void TestAdjacentDifferenceWithBigIndexes()
+{
+  TestAdjacentDifferenceWithBigIndexesHelper(30);
+  TestAdjacentDifferenceWithBigIndexesHelper(31);
+  TestAdjacentDifferenceWithBigIndexesHelper(32);
+  TestAdjacentDifferenceWithBigIndexesHelper(33);
+}
+DECLARE_UNITTEST(TestAdjacentDifferenceWithBigIndexes);