[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/vector_insert.cu

@@ -0,0 +1,251 @@
+#include <thrust/device_malloc_allocator.h>
+#include <thrust/sequence.h>
+
+#include <unittest/unittest.h>
+
+template <class Vector>
+struct TestVectorRangeInsertSimple
+{
+  void operator()(size_t)
+  {
+    Vector v1(5);
+    thrust::sequence(v1.begin(), v1.end());
+
+    // test when insertion range fits inside capacity
+    // and the size of the insertion is greater than the number
+    // of displaced elements
+    Vector v2(3);
+    v2.reserve(10);
+    thrust::sequence(v2.begin(), v2.end());
+
+    size_t new_size       = v2.size() + v1.size();
+    size_t insertion_size = v1.end() - v1.begin();
+    size_t num_displaced  = v2.end() - (v2.begin() + 1);
+
+    ASSERT_EQUAL(true, v2.capacity() >= new_size);
+    ASSERT_EQUAL(true, insertion_size > num_displaced);
+
+    v2.insert(v2.begin() + 1, v1.begin(), v1.end());
+
+    Vector ref{0, 0, 1, 2, 3, 4, 1, 2};
+    ASSERT_EQUAL(ref, v2);
+
+    ASSERT_EQUAL(8lu, v2.size());
+    ASSERT_EQUAL(10lu, v2.capacity());
+
+    // test when insertion range fits inside capacity
+    // and the size of the insertion is equal to the number
+    // of displaced elements
+    Vector v3(5);
+    v3.reserve(10);
+    thrust::sequence(v3.begin(), v3.end());
+
+    new_size       = v3.size() + v1.size();
+    insertion_size = v1.end() - v1.begin();
+    num_displaced  = v3.end() - v3.begin();
+
+    ASSERT_EQUAL(true, v3.capacity() >= new_size);
+    ASSERT_EQUAL(true, insertion_size == num_displaced);
+
+    v3.insert(v3.begin(), v1.begin(), v1.end());
+    ref = {0, 1, 2, 3, 4, 0, 1, 2, 3, 4};
+    ASSERT_EQUAL(ref, v3);
+
+    ASSERT_EQUAL(10lu, v3.size());
+    ASSERT_EQUAL(10lu, v3.capacity());
+
+    // test when insertion range fits inside capacity
+    // and the size of the insertion is less than the
+    // number of displaced elements
+    Vector v4(5);
+    v4.reserve(10);
+    thrust::sequence(v4.begin(), v4.end());
+
+    new_size       = v4.size() + v1.size();
+    insertion_size = (v1.begin() + 3) - v1.begin();
+    num_displaced  = v4.end() - (v4.begin() + 1);
+
+    ASSERT_EQUAL(true, v4.capacity() >= new_size);
+    ASSERT_EQUAL(true, insertion_size < num_displaced);
+
+    v4.insert(v4.begin() + 1, v1.begin(), v1.begin() + 3);
+
+    ref = {0, 0, 1, 2, 1, 2, 3, 4};
+    ASSERT_EQUAL(ref, v4);
+
+    ASSERT_EQUAL(8lu, v4.size());
+    ASSERT_EQUAL(10lu, v4.capacity());
+
+    // test when insertion range does not fit inside capacity
+    Vector v5(5);
+    thrust::sequence(v5.begin(), v5.end());
+
+    new_size = v5.size() + v1.size();
+
+    ASSERT_EQUAL(true, v5.capacity() < new_size);
+
+    v5.insert(v5.begin() + 1, v1.begin(), v1.end());
+
+    ref = {0, 0, 1, 2, 3, 4, 1, 2, 3, 4};
+    ASSERT_EQUAL(ref, v5);
+
+    ASSERT_EQUAL(10lu, v5.size());
+  }
+}; // end TestVectorRangeInsertSimple
+VectorUnitTest<TestVectorRangeInsertSimple, NumericTypes, thrust::device_vector, thrust::device_malloc_allocator>
+  TestVectorRangeInsertSimpleDeviceInstance;
+VectorUnitTest<TestVectorRangeInsertSimple, NumericTypes, thrust::host_vector, std::allocator>
+  TestVectorRangeInsertSimpleHostInstance;
+
+template <class T>
+struct TestVectorRangeInsert
+{
+  void operator()(size_t n)
+  {
+    thrust::host_vector<T> h_src = unittest::random_samples<T>(n + 3);
+    thrust::host_vector<T> h_dst = unittest::random_samples<T>(n);
+
+    thrust::device_vector<T> d_src = h_src;
+    thrust::device_vector<T> d_dst = h_dst;
+
+    // choose insertion range at random
+    size_t begin = n > 0 ? (size_t) h_src[n] % n : 0;
+    size_t end   = n > 0 ? (size_t) h_src[n + 1] % n : 0;
+    if (end < begin)
+    {
+      using ::cuda::std::swap;
+      swap(begin, end);
+    }
+
+    // choose insertion position at random
+    size_t position = n > 0 ? (size_t) h_src[n + 2] % n : 0;
+
+    // insert on host
+    h_dst.insert(h_dst.begin() + position, h_src.begin() + begin, h_src.begin() + end);
+
+    // insert on device
+    d_dst.insert(d_dst.begin() + position, d_src.begin() + begin, d_src.begin() + end);
+
+    ASSERT_EQUAL(h_dst, d_dst);
+  }
+}; // end TestVectorRangeInsert
+VariableUnitTest<TestVectorRangeInsert, IntegralTypes> TestVectorRangeInsertInstance;
+
+template <class Vector>
+struct TestVectorFillInsertSimple
+{
+  void operator()(size_t)
+  {
+    // test when insertion range fits inside capacity
+    // and the size of the insertion is greater than the number
+    // of displaced elements
+    Vector v1(3);
+    v1.reserve(10);
+    thrust::sequence(v1.begin(), v1.end());
+
+    size_t insertion_size = 5;
+    size_t new_size       = v1.size() + insertion_size;
+    size_t num_displaced  = v1.end() - (v1.begin() + 1);
+
+    ASSERT_EQUAL(true, v1.capacity() >= new_size);
+    ASSERT_EQUAL(true, insertion_size > num_displaced);
+
+    v1.insert(v1.begin() + 1, insertion_size, 13);
+
+    Vector ref{0, 13, 13, 13, 13, 13, 1, 2};
+    ASSERT_EQUAL(ref, v1);
+
+    ASSERT_EQUAL(8lu, v1.size());
+    ASSERT_EQUAL(10lu, v1.capacity());
+
+    // test when insertion range fits inside capacity
+    // and the size of the insertion is equal to the number
+    // of displaced elements
+    Vector v2(5);
+    v2.reserve(10);
+    thrust::sequence(v2.begin(), v2.end());
+
+    insertion_size = 5;
+    new_size       = v2.size() + insertion_size;
+    num_displaced  = v2.end() - v2.begin();
+
+    ASSERT_EQUAL(true, v2.capacity() >= new_size);
+    ASSERT_EQUAL(true, insertion_size == num_displaced);
+
+    v2.insert(v2.begin(), insertion_size, 13);
+
+    ref = {13, 13, 13, 13, 13, 0, 1, 2, 3, 4};
+    ASSERT_EQUAL(ref, v2);
+
+    ASSERT_EQUAL(10lu, v2.size());
+    ASSERT_EQUAL(10lu, v2.capacity());
+
+    // test when insertion range fits inside capacity
+    // and the size of the insertion is less than the
+    // number of displaced elements
+    Vector v3(5);
+    v3.reserve(10);
+    thrust::sequence(v3.begin(), v3.end());
+
+    insertion_size = 3;
+    new_size       = v3.size() + insertion_size;
+    num_displaced  = v3.end() - (v3.begin() + 1);
+
+    ASSERT_EQUAL(true, v3.capacity() >= new_size);
+    ASSERT_EQUAL(true, insertion_size < num_displaced);
+
+    v3.insert(v3.begin() + 1, insertion_size, 13);
+
+    ref = {0, 13, 13, 13, 1, 2, 3, 4};
+    ASSERT_EQUAL(ref, v3);
+
+    ASSERT_EQUAL(8lu, v3.size());
+    ASSERT_EQUAL(10lu, v3.capacity());
+
+    // test when insertion range does not fit inside capacity
+    Vector v4(5);
+    thrust::sequence(v4.begin(), v4.end());
+
+    insertion_size = 5;
+    new_size       = v4.size() + insertion_size;
+
+    ASSERT_EQUAL(true, v4.capacity() < new_size);
+
+    v4.insert(v4.begin() + 1, insertion_size, 13);
+
+    ref = {0, 13, 13, 13, 13, 13, 1, 2, 3, 4};
+    ASSERT_EQUAL(ref, v4);
+
+    ASSERT_EQUAL(10lu, v4.size());
+  }
+}; // end TestVectorFillInsertSimple
+VectorUnitTest<TestVectorFillInsertSimple, NumericTypes, thrust::device_vector, thrust::device_malloc_allocator>
+  TestVectorFillInsertSimpleDeviceInstance;
+VectorUnitTest<TestVectorFillInsertSimple, NumericTypes, thrust::host_vector, std::allocator>
+  TestVectorFillInsertSimpleHostInstance;
+
+template <class T>
+struct TestVectorFillInsert
+{
+  void operator()(size_t n)
+  {
+    thrust::host_vector<T> h_dst = unittest::random_samples<T>(n + 2);
+
+    thrust::device_vector<T> d_dst = h_dst;
+
+    // choose insertion position at random
+    size_t position = n > 0 ? (size_t) h_dst[n] % n : 0;
+
+    // choose insertion size at random
+    size_t insertion_size = n > 0 ? (size_t) h_dst[n] % n : 13;
+
+    // insert on host
+    h_dst.insert(h_dst.begin() + position, insertion_size, 13);
+
+    // insert on device
+    d_dst.insert(d_dst.begin() + position, insertion_size, 13);
+
+    ASSERT_EQUAL(h_dst, d_dst);
+  }
+}; // end TestVectorFillInsert
+VariableUnitTest<TestVectorFillInsert, IntegralTypes> TestVectorFillInsertInstance;