[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/inner_product.cu

@@ -0,0 +1,155 @@
+#include <thrust/device_free.h>
+#include <thrust/device_malloc.h>
+#include <thrust/device_vector.h>
+#include <thrust/functional.h>
+#include <thrust/inner_product.h>
+#include <thrust/iterator/retag.h>
+
+#include <unittest/unittest.h>
+
+template <class Vector>
+void TestInnerProductSimple()
+{
+  using T = typename Vector::value_type;
+
+  Vector v1{1, -2, 3};
+  Vector v2{-4, 5, 6};
+
+  T init   = 3;
+  T result = thrust::inner_product(v1.begin(), v1.end(), v2.begin(), init);
+  ASSERT_EQUAL(result, 7);
+}
+DECLARE_VECTOR_UNITTEST(TestInnerProductSimple);
+
+template <typename InputIterator1, typename InputIterator2, typename OutputType>
+int inner_product(my_system& system, InputIterator1, InputIterator1, InputIterator2, OutputType)
+{
+  system.validate_dispatch();
+  return 13;
+}
+
+void TestInnerProductDispatchExplicit()
+{
+  thrust::device_vector<int> vec;
+
+  my_system sys(0);
+  thrust::inner_product(sys, vec.begin(), vec.end(), vec.begin(), 0);
+
+  ASSERT_EQUAL(true, sys.is_valid());
+}
+DECLARE_UNITTEST(TestInnerProductDispatchExplicit);
+
+template <typename InputIterator1, typename InputIterator2, typename OutputType>
+int inner_product(my_tag, InputIterator1, InputIterator1, InputIterator2, OutputType)
+{
+  return 13;
+}
+
+void TestInnerProductDispatchImplicit()
+{
+  thrust::device_vector<int> vec;
+
+  int result = thrust::inner_product(
+    thrust::retag<my_tag>(vec.begin()), thrust::retag<my_tag>(vec.end()), thrust::retag<my_tag>(vec.begin()), 0);
+
+  ASSERT_EQUAL(13, result);
+}
+DECLARE_UNITTEST(TestInnerProductDispatchImplicit);
+
+template <class Vector>
+void TestInnerProductWithOperator()
+{
+  using T = typename Vector::value_type;
+
+  Vector v1{1, -2, 3};
+  Vector v2{-1, 3, 6};
+
+  // compute (v1 - v2) and perform a multiplies reduction
+  T init   = 3;
+  T result = thrust::inner_product(
+    v1.begin(), v1.end(), v2.begin(), init, ::cuda::std::multiplies<T>(), ::cuda::std::minus<T>());
+  ASSERT_EQUAL(result, 90);
+}
+DECLARE_VECTOR_UNITTEST(TestInnerProductWithOperator);
+
+template <typename T>
+struct TestInnerProduct
+{
+  void operator()(const size_t n)
+  {
+    thrust::host_vector<T> h_v1 = unittest::random_integers<T>(n);
+    thrust::host_vector<T> h_v2 = unittest::random_integers<T>(n);
+
+    thrust::device_vector<T> d_v1 = h_v1;
+    thrust::device_vector<T> d_v2 = h_v2;
+
+    T init = 13;
+
+    T expected = thrust::inner_product(h_v1.begin(), h_v1.end(), h_v2.begin(), init);
+    T result   = thrust::inner_product(d_v1.begin(), d_v1.end(), d_v2.begin(), init);
+
+    ASSERT_EQUAL(expected, result);
+  }
+};
+VariableUnitTest<TestInnerProduct, IntegralTypes> TestInnerProductInstance;
+
+struct only_set_when_both_expected
+{
+  long long expected;
+  bool* flag;
+
+  _CCCL_DEVICE long long operator()(long long x, long long y)
+  {
+    if (x == expected && y == expected)
+    {
+      *flag = true;
+    }
+
+    return x == y;
+  }
+};
+
+void TestInnerProductWithBigIndexesHelper(int magnitude)
+{
+  thrust::counting_iterator<long long> begin(1);
+  thrust::counting_iterator<long long> end = begin + (1ll << magnitude);
+  ASSERT_EQUAL(::cuda::std::distance(begin, end), 1ll << magnitude);
+
+  thrust::device_ptr<bool> has_executed = thrust::device_malloc<bool>(1);
+  *has_executed                         = false;
+
+  only_set_when_both_expected fn = {(1ll << magnitude) - 1, thrust::raw_pointer_cast(has_executed)};
+
+  ASSERT_EQUAL(thrust::inner_product(thrust::device, begin, end, begin, 0ll, ::cuda::std::plus<long long>(), fn),
+               (1ll << magnitude));
+
+  bool has_executed_h = *has_executed;
+  thrust::device_free(has_executed);
+
+  ASSERT_EQUAL(has_executed_h, true);
+}
+
+void TestInnerProductWithBigIndexes()
+{
+  TestInnerProductWithBigIndexesHelper(30);
+#ifndef THRUST_FORCE_32_BIT_OFFSET_TYPE
+  TestInnerProductWithBigIndexesHelper(31);
+  TestInnerProductWithBigIndexesHelper(32);
+  TestInnerProductWithBigIndexesHelper(33);
+#endif
+}
+DECLARE_UNITTEST(TestInnerProductWithBigIndexes);
+
+void TestInnerProductPlaceholders()
+{ // Regression test for NVIDIA/thrust#1178
+  using namespace thrust::placeholders;
+
+  thrust::device_vector<float> v1(100, 1.f);
+  thrust::device_vector<float> v2(100, 1.f);
+
+  auto result =
+    thrust::inner_product(v1.begin(), v1.end(), v2.begin(), 0.0f, ::cuda::std::plus<float>{}, _1 * _2 + 1.0f);
+
+  ASSERT_ALMOST_EQUAL(result, 200.f);
+}
+DECLARE_UNITTEST(TestInnerProductPlaceholders);