[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/cuda/offset_iterator.cu

@@ -0,0 +1,125 @@
+#include <thrust/distance.h>
+#include <thrust/functional.h>
+#include <thrust/iterator/offset_iterator.h>
+#include <thrust/iterator/transform_iterator.h>
+
+#include <cuda/std/iterator>
+
+#include <unittest/unittest.h>
+
+struct device_only_iterator
+{
+  using iterator_category = cuda::std::random_access_iterator_tag;
+  using difference_type   = cuda::std::ptrdiff_t;
+  using value_type        = int;
+  using pointer           = int*;
+  using reference         = int&;
+
+  _CCCL_HOST_DEVICE device_only_iterator(pointer ptr)
+      : m_ptr(ptr)
+  {}
+
+  _CCCL_DEVICE reference operator*() const
+  {
+    return *m_ptr;
+  }
+
+  _CCCL_DEVICE device_only_iterator& operator++()
+  {
+    ++m_ptr;
+    return *this;
+  }
+
+  _CCCL_DEVICE device_only_iterator operator++(int)
+  {
+    device_only_iterator tmp = *this;
+    ++*this;
+    return tmp;
+  }
+
+  _CCCL_DEVICE device_only_iterator& operator--()
+  {
+    --m_ptr;
+    return *this;
+  }
+
+  _CCCL_DEVICE device_only_iterator operator--(int)
+  {
+    device_only_iterator tmp = *this;
+    --*this;
+    return tmp;
+  }
+
+  _CCCL_DEVICE device_only_iterator& operator+=(difference_type n)
+  {
+    m_ptr += n;
+    return *this;
+  }
+
+  _CCCL_DEVICE friend device_only_iterator operator+(device_only_iterator it, difference_type d)
+  {
+    return it += d;
+  }
+
+  _CCCL_DEVICE friend difference_type operator-(const device_only_iterator& a, const device_only_iterator& b)
+  {
+    return a.m_ptr - b.m_ptr;
+  }
+
+  _CCCL_DEVICE friend bool operator==(const device_only_iterator& a, const device_only_iterator& b)
+  {
+    return a.m_ptr == b.m_ptr;
+  }
+
+  _CCCL_DEVICE friend bool operator!=(const device_only_iterator& a, const device_only_iterator& b)
+  {
+    return a.m_ptr != b.m_ptr;
+  }
+
+private:
+  pointer m_ptr;
+};
+
+_CCCL_HOST_DEVICE void TestOffsetIteratorBoth(thrust::offset_iterator<device_only_iterator> iter)
+{
+  assert(iter.offset() == 0);
+  ++iter;
+  assert(iter.offset() == 1);
+  iter++;
+  assert(iter.offset() == 2);
+  --iter;
+  assert(iter.offset() == 1);
+  iter--;
+  assert(iter.offset() == 0);
+  iter += 100;
+  assert(iter.offset() == 100);
+}
+
+__global__ void TestOffsetIteratorDevice(thrust::offset_iterator<device_only_iterator> iter)
+{
+  TestOffsetIteratorBoth(iter);
+
+  // access
+  assert(*iter == 1);
+
+  auto iter2 = iter;
+  iter2 += 3;
+  assert(*iter2 == 1);
+
+  // difference
+  assert(iter2 - iter == 3);
+
+  // comparison
+  assert(!(iter2 == iter));
+  assert(iter2 != iter);
+}
+
+void TestOffsetIteratorWithDeviceOnlyIterator()
+{
+  thrust::device_vector<int> v{1, 2, 3, 4, 5};
+  device_only_iterator base(thrust::raw_pointer_cast(v.data()));
+  thrust::offset_iterator iter(base);
+  TestOffsetIteratorBoth(iter);
+  TestOffsetIteratorDevice<<<1, 1>>>(iter);
+}
+DECLARE_UNITTEST(TestOffsetIteratorWithDeviceOnlyIterator);