[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/examples/remove_points2d.cu

@@ -0,0 +1,78 @@
+#include <thrust/host_vector.h>
+#include <thrust/random.h>
+#include <thrust/remove.h>
+
+#include <iostream>
+
+// This example generates random points in the
+// unit square [0,1)x[0,1) and then removes all
+// points where x^2 + y^2 > 1
+//
+// The x and y coordinates are stored in separate arrays
+// and a zip_iterator is used to combine them together
+
+template <typename T>
+struct is_outside_circle
+{
+  template <typename Tuple>
+  inline __host__ __device__ bool operator()(const Tuple& tuple) const
+  {
+    // unpack the tuple into x and y coordinates
+    const T x = cuda::std::get<0>(tuple);
+    const T y = cuda::std::get<1>(tuple);
+
+    if (x * x + y * y > 1)
+    {
+      return true;
+    }
+    else
+    {
+      return false;
+    }
+  }
+};
+
+int main()
+{
+  const size_t N = 20;
+
+  // generate random points in the unit square on the host
+  thrust::default_random_engine rng;
+  thrust::uniform_real_distribution<float> u01(0.0f, 1.0f);
+  thrust::host_vector<float> x(N);
+  thrust::host_vector<float> y(N);
+  for (size_t i = 0; i < N; i++)
+  {
+    x[i] = u01(rng);
+    y[i] = u01(rng);
+  }
+
+  // print the initial points
+  std::cout << std::fixed;
+  std::cout << "Generated " << N << " points" << '\n';
+  for (size_t i = 0; i < x.size(); i++)
+  {
+    std::cout << "(" << x[i] << "," << y[i] << ")" << '\n';
+  }
+  std::cout << '\n';
+
+  // remove points where x^2 + y^2 > 1 and determine new array sizes
+  size_t new_size =
+    thrust::remove_if(thrust::make_zip_iterator(x.begin(), y.begin()),
+                      thrust::make_zip_iterator(x.end(), y.end()),
+                      is_outside_circle<float>())
+    - thrust::make_zip_iterator(x.begin(), y.begin());
+
+  // resize the vectors (note: this does not free any memory)
+  x.resize(new_size);
+  y.resize(new_size);
+
+  // print the filtered points
+  std::cout << "After stream compaction, " << new_size << " points remain" << '\n';
+  for (size_t i = 0; i < x.size(); i++)
+  {
+    std::cout << "(" << x[i] << "," << y[i] << ")" << '\n';
+  }
+
+  return 0;
+}