[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/mismatch.cu

@@ -0,0 +1,148 @@
+#include <thrust/iterator/retag.h>
+#include <thrust/mismatch.h>
+
+#include <unittest/unittest.h>
+template <class Vector>
+void TestMismatchSimple()
+{
+  Vector a{1, 2, 3, 4};
+  Vector b{1, 2, 4, 3};
+
+  ASSERT_EQUAL(thrust::mismatch(a.begin(), a.end(), b.begin()).first - a.begin(), 2);
+  ASSERT_EQUAL(thrust::mismatch(a.begin(), a.end(), b.begin()).second - b.begin(), 2);
+
+  b[2] = 3;
+
+  ASSERT_EQUAL(thrust::mismatch(a.begin(), a.end(), b.begin()).first - a.begin(), 3);
+  ASSERT_EQUAL(thrust::mismatch(a.begin(), a.end(), b.begin()).second - b.begin(), 3);
+
+  b[3] = 4;
+
+  ASSERT_EQUAL(thrust::mismatch(a.begin(), a.end(), b.begin()).first - a.begin(), 4);
+  ASSERT_EQUAL(thrust::mismatch(a.begin(), a.end(), b.begin()).second - b.begin(), 4);
+}
+DECLARE_VECTOR_UNITTEST(TestMismatchSimple);
+
+template <class Vector>
+void TestMismatchBoundedSimple()
+{
+  using T = typename Vector::value_type;
+
+  // Equal-length ranges with mismatch
+  {
+    Vector a{1, 2, 3, 4};
+    Vector b{1, 2, 4, 3};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end());
+    ASSERT_EQUAL(result.first - a.begin(), 2);
+    ASSERT_EQUAL(result.second - b.begin(), 2);
+  }
+
+  // Equal-length ranges, no mismatch
+  {
+    Vector a{1, 2, 3};
+    Vector b{1, 2, 3};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end());
+    ASSERT_EQUAL(result.first - a.begin(), 3);
+    ASSERT_EQUAL(result.second - b.begin(), 3);
+  }
+
+  // Range1 shorter: stops when range1 exhausted
+  {
+    Vector a{1, 2};
+    Vector b{1, 2, 99};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end());
+    ASSERT_EQUAL(result.first - a.begin(), 2); // exhausted range1
+    ASSERT_EQUAL(result.second - b.begin(), 2);
+  }
+
+  // Range2 shorter: stops when range2 exhausted
+  {
+    Vector a{1, 2, 99};
+    Vector b{1, 2};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end());
+    ASSERT_EQUAL(result.first - a.begin(), 2);
+    ASSERT_EQUAL(result.second - b.begin(), 2); // exhausted range2
+  }
+
+  // Mismatch before either range ends (range2 shorter)
+  {
+    Vector a{1, 9, 3};
+    Vector b{1, 2};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end());
+    ASSERT_EQUAL(result.first - a.begin(), 1);
+    ASSERT_EQUAL(result.second - b.begin(), 1);
+  }
+
+  // With binary predicate
+  {
+    Vector a{1, 2, 3};
+    Vector b{1, 2};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end(), ::cuda::std::equal_to<T>());
+    ASSERT_EQUAL(result.first - a.begin(), 2);
+    ASSERT_EQUAL(result.second - b.begin(), 2);
+  }
+
+  // Empty ranges
+  {
+    Vector a;
+    Vector b{1, 2};
+    auto result = thrust::mismatch(a.begin(), a.end(), b.begin(), b.end());
+    ASSERT_EQUAL(result.first - a.begin(), 0);
+    ASSERT_EQUAL(result.second - b.begin(), 0);
+  }
+}
+DECLARE_VECTOR_UNITTEST(TestMismatchBoundedSimple);
+
+void TestMismatchBoundedWithExec()
+{
+  thrust::device_vector<int> a{1, 2, 3, 4};
+  thrust::device_vector<int> b{1, 2, 99};
+
+  // range1 longer, range2 shorter — stops at range2 end
+  auto result = thrust::mismatch(thrust::device, a.begin(), a.end(), b.begin(), b.end());
+  ASSERT_EQUAL(result.first - a.begin(), 2);
+  ASSERT_EQUAL(result.second - b.begin(), 2);
+
+  // With predicate
+  result = thrust::mismatch(thrust::device, a.begin(), a.end(), b.begin(), b.end(), ::cuda::std::equal_to<int>());
+  ASSERT_EQUAL(result.first - a.begin(), 2);
+  ASSERT_EQUAL(result.second - b.begin(), 2);
+}
+DECLARE_UNITTEST(TestMismatchBoundedWithExec);
+
+template <typename InputIterator1, typename InputIterator2>
+cuda::std::pair<InputIterator1, InputIterator2>
+mismatch(my_system& system, InputIterator1 first, InputIterator1, InputIterator2)
+{
+  system.validate_dispatch();
+  return cuda::std::make_pair(first, first);
+}
+
+void TestMismatchDispatchExplicit()
+{
+  thrust::device_vector<int> vec(1);
+
+  my_system sys(0);
+  thrust::mismatch(sys, vec.begin(), vec.begin(), vec.begin());
+
+  ASSERT_EQUAL(true, sys.is_valid());
+}
+DECLARE_UNITTEST(TestMismatchDispatchExplicit);
+
+template <typename InputIterator1, typename InputIterator2>
+cuda::std::pair<InputIterator1, InputIterator2> mismatch(my_tag, InputIterator1 first, InputIterator1, InputIterator2)
+{
+  *first = 13;
+  return cuda::std::make_pair(first, first);
+}
+
+void TestMismatchDispatchImplicit()
+{
+  thrust::device_vector<int> vec(1);
+
+  thrust::mismatch(
+    thrust::retag<my_tag>(vec.begin()), thrust::retag<my_tag>(vec.begin()), thrust::retag<my_tag>(vec.begin()));
+
+  ASSERT_EQUAL(13, vec.front());
+}
+DECLARE_UNITTEST(TestMismatchDispatchImplicit);