[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/cub/test/catch2_test_util_arch.cu

@@ -0,0 +1,117 @@
+// SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <cub/util_arch.cuh>
+
+#include <c2h/catch2_test_helper.h>
+
+template <auto V>
+struct show;
+
+template <int Nominal4ByteThreadsPerBlock,
+          int Nominal4ByteItemsPerThread,
+          typename ComputeT,
+          int ExpectedThreadsPerBlock,
+          int ExpectedItemsPerThread>
+void check_mem_bound_scaling()
+{
+  using mbs = cub::detail::MemBoundScaling<Nominal4ByteThreadsPerBlock, Nominal4ByteItemsPerThread, ComputeT>;
+
+  if constexpr (mbs::ITEMS_PER_THREAD != ExpectedItemsPerThread)
+  {
+    show<mbs::ITEMS_PER_THREAD>::asdf();
+  }
+  STATIC_REQUIRE(mbs::ITEMS_PER_THREAD == ExpectedItemsPerThread);
+
+  if constexpr (mbs::BLOCK_THREADS != ExpectedThreadsPerBlock)
+  {
+    show<mbs::BLOCK_THREADS>::asdf();
+  }
+  STATIC_REQUIRE(mbs::BLOCK_THREADS == ExpectedThreadsPerBlock);
+}
+
+C2H_TEST("MemBoundScaling", "[util][arch]")
+{
+  check_mem_bound_scaling<256, 1, char, 256, 2>();
+  check_mem_bound_scaling<256, 16, char, 256, 32>();
+  check_mem_bound_scaling<256, 20, char, 256, 40>();
+  check_mem_bound_scaling<256, 100, char, 256, 200>();
+  check_mem_bound_scaling<256, 500, char, 64, 1000>();
+  check_mem_bound_scaling<256, 10000, char, 32, 20000>();
+
+  check_mem_bound_scaling<256, 1, int, 256, 1>();
+  check_mem_bound_scaling<256, 16, int, 256, 16>();
+  check_mem_bound_scaling<256, 20, int, 256, 20>();
+  check_mem_bound_scaling<256, 100, int, 128, 100>();
+  check_mem_bound_scaling<256, 500, int, 32, 500>();
+  check_mem_bound_scaling<256, 10000, int, 32, 10000>();
+
+  check_mem_bound_scaling<256, 1, int4, 256, 1>();
+  check_mem_bound_scaling<256, 16, int4, 256, 4>();
+  check_mem_bound_scaling<256, 20, int4, 256, 5>();
+  check_mem_bound_scaling<256, 100, int4, 128, 25>();
+  check_mem_bound_scaling<256, 500, int4, 32, 125>();
+  check_mem_bound_scaling<256, 10000, int4, 32, 2500>();
+
+  using large_t = char[1024];
+  check_mem_bound_scaling<256, 1, large_t, 64, 1>();
+  check_mem_bound_scaling<256, 16, large_t, 64, 1>();
+  check_mem_bound_scaling<256, 20, large_t, 64, 1>();
+  check_mem_bound_scaling<256, 100, large_t, 64, 1>();
+  check_mem_bound_scaling<256, 500, large_t, 64, 1>();
+  check_mem_bound_scaling<256, 10000, large_t, 32, 39>();
+}
+
+template <int Nominal4ByteThreadsPerBlock,
+          int Nominal4ByteItemsPerThread,
+          typename ComputeT,
+          int ExpectedThreadsPerBlock,
+          int ExpectedItemsPerThread>
+void check_reg_bound_scaling()
+{
+  using mbs = cub::detail::RegBoundScaling<Nominal4ByteThreadsPerBlock, Nominal4ByteItemsPerThread, ComputeT>;
+
+  if constexpr (mbs::ITEMS_PER_THREAD != ExpectedItemsPerThread)
+  {
+    show<mbs::ITEMS_PER_THREAD>::asdf();
+  }
+  STATIC_REQUIRE(mbs::ITEMS_PER_THREAD == ExpectedItemsPerThread);
+
+  if constexpr (mbs::BLOCK_THREADS != ExpectedThreadsPerBlock)
+  {
+    show<mbs::BLOCK_THREADS>::asdf();
+  }
+  STATIC_REQUIRE(mbs::BLOCK_THREADS == ExpectedThreadsPerBlock);
+}
+
+C2H_TEST("RegBoundScaling", "[util][arch]")
+{
+  check_reg_bound_scaling<256, 1, char, 256, 1>();
+  check_reg_bound_scaling<256, 16, char, 256, 16>();
+  check_reg_bound_scaling<256, 20, char, 256, 20>();
+  check_reg_bound_scaling<256, 100, char, 256, 100>();
+  check_reg_bound_scaling<256, 500, char, 128, 500>();
+  check_reg_bound_scaling<256, 10000, char, 32, 10000>();
+
+  check_reg_bound_scaling<256, 1, int, 256, 1>();
+  check_reg_bound_scaling<256, 16, int, 256, 16>();
+  check_reg_bound_scaling<256, 20, int, 256, 20>();
+  check_reg_bound_scaling<256, 100, int, 128, 100>();
+  check_reg_bound_scaling<256, 500, int, 32, 500>();
+  check_reg_bound_scaling<256, 10000, int, 32, 10000>();
+
+  check_reg_bound_scaling<256, 1, int4, 256, 1>();
+  check_reg_bound_scaling<256, 16, int4, 256, 4>();
+  check_reg_bound_scaling<256, 20, int4, 256, 5>();
+  check_reg_bound_scaling<256, 100, int4, 128, 25>();
+  check_reg_bound_scaling<256, 500, int4, 32, 125>();
+  check_reg_bound_scaling<256, 10000, int4, 32, 2500>();
+
+  using large_t = char[1024];
+  check_reg_bound_scaling<256, 1, large_t, 64, 1>();
+  check_reg_bound_scaling<256, 16, large_t, 64, 1>();
+  check_reg_bound_scaling<256, 20, large_t, 64, 1>();
+  check_reg_bound_scaling<256, 100, large_t, 64, 1>();
+  check_reg_bound_scaling<256, 500, large_t, 64, 1>();
+  check_reg_bound_scaling<256, 10000, large_t, 32, 39>();
+}