[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/is_partitioned.cu

@@ -0,0 +1,96 @@
+#include <thrust/functional.h>
+#include <thrust/iterator/retag.h>
+#include <thrust/partition.h>
+
+#include <unittest/unittest.h>
+
+template <typename T>
+struct is_even
+{
+  _CCCL_HOST_DEVICE bool operator()(T x) const
+  {
+    return ((int) x % 2) == 0;
+  }
+};
+
+template <typename Vector>
+void TestIsPartitionedSimple()
+{
+  Vector v{1, 1, 1, 0};
+
+  // empty partition
+  ASSERT_EQUAL_QUIET(true, thrust::is_partitioned(v.begin(), v.begin(), ::cuda::std::identity{}));
+
+  // one element true partition
+  ASSERT_EQUAL_QUIET(true, thrust::is_partitioned(v.begin(), v.begin() + 1, ::cuda::std::identity{}));
+
+  // just true partition
+  ASSERT_EQUAL_QUIET(true, thrust::is_partitioned(v.begin(), v.begin() + 2, ::cuda::std::identity{}));
+
+  // both true & false partitions
+  ASSERT_EQUAL_QUIET(true, thrust::is_partitioned(v.begin(), v.end(), ::cuda::std::identity{}));
+
+  // one element false partition
+  ASSERT_EQUAL_QUIET(true, thrust::is_partitioned(v.begin() + 3, v.end(), ::cuda::std::identity{}));
+
+  v = {1, 0, 1, 1};
+
+  // not partitioned
+  ASSERT_EQUAL_QUIET(false, thrust::is_partitioned(v.begin(), v.end(), ::cuda::std::identity{}));
+}
+DECLARE_VECTOR_UNITTEST(TestIsPartitionedSimple);
+
+template <class Vector>
+void TestIsPartitioned()
+{
+  using T = typename Vector::value_type;
+
+  const size_t n = (1 << 16) + 13;
+
+  Vector v = unittest::random_integers<T>(n);
+
+  v[0] = 1;
+  v[1] = 0;
+
+  ASSERT_EQUAL(false, thrust::is_partitioned(v.begin(), v.end(), is_even<T>()));
+
+  thrust::partition(v.begin(), v.end(), is_even<T>());
+
+  ASSERT_EQUAL(true, thrust::is_partitioned(v.begin(), v.end(), is_even<T>()));
+}
+DECLARE_INTEGRAL_VECTOR_UNITTEST(TestIsPartitioned);
+
+template <typename InputIterator, typename Predicate>
+bool is_partitioned(my_system& system, InputIterator /*first*/, InputIterator, Predicate)
+{
+  system.validate_dispatch();
+  return false;
+}
+
+void TestIsPartitionedDispatchExplicit()
+{
+  thrust::device_vector<int> vec(1);
+
+  my_system sys(0);
+  thrust::is_partitioned(sys, vec.begin(), vec.end(), 0);
+
+  ASSERT_EQUAL(true, sys.is_valid());
+}
+DECLARE_UNITTEST(TestIsPartitionedDispatchExplicit);
+
+template <typename InputIterator, typename Predicate>
+bool is_partitioned(my_tag, InputIterator first, InputIterator, Predicate)
+{
+  *first = 13;
+  return false;
+}
+
+void TestIsPartitionedDispatchImplicit()
+{
+  thrust::device_vector<int> vec(1);
+
+  thrust::is_partitioned(thrust::retag<my_tag>(vec.begin()), thrust::retag<my_tag>(vec.end()), 0);
+
+  ASSERT_EQUAL(13, vec.front());
+}
+DECLARE_UNITTEST(TestIsPartitionedDispatchImplicit);