[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/zip_iterator_scan.cu

@@ -0,0 +1,102 @@
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/scan.h>
+
+#include <unittest/unittest.h>
+
+#if THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
+#  include <unittest/cuda/testframework.h>
+#endif
+
+using namespace unittest;
+
+template <typename Tuple>
+struct TuplePlus
+{
+  _CCCL_HOST_DEVICE Tuple operator()(Tuple x, Tuple y) const
+  {
+    return cuda::std::make_tuple(
+      cuda::std::get<0>(x) + cuda::std::get<0>(y), cuda::std::get<1>(x) + cuda::std::get<1>(y));
+  }
+}; // end SumTuple
+
+template <typename T>
+struct TestZipIteratorScan
+{
+  void operator()(const size_t n)
+  {
+    thrust::host_vector<T> h_data0 = unittest::random_samples<T>(n);
+    thrust::host_vector<T> h_data1 = unittest::random_samples<T>(n);
+
+    thrust::device_vector<T> d_data0 = h_data0;
+    thrust::device_vector<T> d_data1 = h_data1;
+
+    using Tuple = cuda::std::tuple<T, T>;
+
+    thrust::host_vector<Tuple> h_result(n);
+    thrust::device_vector<Tuple> d_result(n);
+
+    // inclusive_scan (tuple output)
+    thrust::inclusive_scan(
+      thrust::make_zip_iterator(h_data0.begin(), h_data1.begin()),
+      thrust::make_zip_iterator(h_data0.end(), h_data1.end()),
+      h_result.begin(),
+      TuplePlus<Tuple>());
+    thrust::inclusive_scan(
+      thrust::make_zip_iterator(d_data0.begin(), d_data1.begin()),
+      thrust::make_zip_iterator(d_data0.end(), d_data1.end()),
+      d_result.begin(),
+      TuplePlus<Tuple>());
+    ASSERT_EQUAL_QUIET(h_result, d_result);
+
+    // exclusive_scan (tuple output)
+    thrust::exclusive_scan(
+      thrust::make_zip_iterator(h_data0.begin(), h_data1.begin()),
+      thrust::make_zip_iterator(h_data0.end(), h_data1.end()),
+      h_result.begin(),
+      cuda::std::make_tuple<T, T>(0, 0),
+      TuplePlus<Tuple>());
+    thrust::exclusive_scan(
+      thrust::make_zip_iterator(d_data0.begin(), d_data1.begin()),
+      thrust::make_zip_iterator(d_data0.end(), d_data1.end()),
+      d_result.begin(),
+      cuda::std::make_tuple<T, T>(0, 0),
+      TuplePlus<Tuple>());
+    ASSERT_EQUAL_QUIET(h_result, d_result);
+
+    thrust::host_vector<T> h_result0(n);
+    thrust::host_vector<T> h_result1(n);
+    thrust::device_vector<T> d_result0(n);
+    thrust::device_vector<T> d_result1(n);
+
+    // inclusive_scan (zip_iterator output)
+    thrust::inclusive_scan(
+      thrust::make_zip_iterator(h_data0.begin(), h_data1.begin()),
+      thrust::make_zip_iterator(h_data0.end(), h_data1.end()),
+      thrust::make_zip_iterator(h_result0.begin(), h_result1.begin()),
+      TuplePlus<Tuple>());
+    thrust::inclusive_scan(
+      thrust::make_zip_iterator(d_data0.begin(), d_data1.begin()),
+      thrust::make_zip_iterator(d_data0.end(), d_data1.end()),
+      thrust::make_zip_iterator(d_result0.begin(), d_result1.begin()),
+      TuplePlus<Tuple>());
+    ASSERT_EQUAL_QUIET(h_result0, d_result0);
+    ASSERT_EQUAL_QUIET(h_result1, d_result1);
+
+    // exclusive_scan (zip_iterator output)
+    thrust::exclusive_scan(
+      thrust::make_zip_iterator(h_data0.begin(), h_data1.begin()),
+      thrust::make_zip_iterator(h_data0.end(), h_data1.end()),
+      thrust::make_zip_iterator(h_result0.begin(), h_result1.begin()),
+      cuda::std::make_tuple<T, T>(0, 0),
+      TuplePlus<Tuple>());
+    thrust::exclusive_scan(
+      thrust::make_zip_iterator(d_data0.begin(), d_data1.begin()),
+      thrust::make_zip_iterator(d_data0.end(), d_data1.end()),
+      thrust::make_zip_iterator(d_result0.begin(), d_result1.begin()),
+      cuda::std::make_tuple<T, T>(0, 0),
+      TuplePlus<Tuple>());
+    ASSERT_EQUAL_QUIET(h_result0, d_result0);
+    ASSERT_EQUAL_QUIET(h_result1, d_result1);
+  }
+};
+VariableUnitTest<TestZipIteratorScan, SignedIntegralTypes> TestZipIteratorScanInstance;