[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/cudax/test/stf/utility/source_location_map.cu

@@ -0,0 +1,51 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of CUDASTF in CUDA C++ Core Libraries,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2022-2025 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#include <cuda/std/source_location>
+
+#include <cuda/experimental/__stf/utility/source_location.cuh>
+#include <cuda/experimental/stf.cuh>
+
+using namespace cuda::experimental::stf;
+
+// Create a map indexed by source locations
+::std::unordered_map<::cuda::std::source_location, int, reserved::source_location_hash, reserved::source_location_equal>
+  stats_map;
+
+void update_counter(::cuda::std::source_location loc = ::cuda::std::source_location::current())
+{
+  stats_map[loc]++;
+}
+
+void funcA()
+{
+  update_counter();
+}
+
+void funcB()
+{
+  update_counter();
+}
+
+int main()
+{
+  for (size_t i = 0; i < 10; i++)
+  {
+    funcA();
+    funcB();
+    funcB();
+  }
+
+  for (auto& e : stats_map)
+  {
+    auto& loc = e.first;
+    fprintf(stderr, "loc.function_name() %s : count %d\n", loc.function_name(), e.second);
+  }
+}

cccl_upstream/cudax/test/stf/utility/timing_with_fences.cu

@@ -0,0 +1,101 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of CUDASTF in CUDA C++ Core Libraries,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+// SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#include <cuda/experimental/stf.cuh>
+
+using namespace cuda::experimental::stf;
+
+static __global__ void cuda_sleep_kernel(long long int clock_cnt)
+{
+  long long int start_clock  = clock64();
+  long long int clock_offset = 0;
+  while (clock_offset < clock_cnt)
+  {
+    clock_offset = clock64() - start_clock;
+  }
+}
+
+void cuda_sleep(double ms, cudaStream_t stream)
+{
+  int device;
+  cudaGetDevice(&device);
+
+  // cudaDevAttrClockRate: Peak clock frequency in kilohertz;
+  int clock_rate;
+  cudaDeviceGetAttribute(&clock_rate, cudaDevAttrClockRate, device);
+
+  long long int clock_cnt = (long long int) (ms * clock_rate);
+  cuda_sleep_kernel<<<1, 1, 0, stream>>>(clock_cnt);
+}
+
+template <typename Ctx_t>
+void run(int NTASKS, int ms)
+{
+  Ctx_t ctx;
+
+  int dummy[1];
+  auto handle = ctx.logical_data(dummy);
+
+  cudaEvent_t start, stop;
+  cuda_safe_call(cudaEventCreate(&start));
+  cuda_safe_call(cudaEventCreate(&stop));
+
+  // warm-up
+  ctx.task(handle.rw())->*[ms](cudaStream_t stream, auto) {
+    cuda_sleep(ms, stream);
+  };
+
+  cuda_safe_call(cudaEventRecord(start, ctx.fence()));
+
+  for (int iter = 0; iter < NTASKS; iter++)
+  {
+    ctx.task(handle.rw())->*[ms](cudaStream_t stream, auto) {
+      cuda_sleep(ms, stream);
+    };
+  }
+
+  cuda_safe_call(cudaEventRecord(stop, ctx.fence()));
+
+  ctx.finalize();
+
+  [[maybe_unused]] float elapsed;
+  cuda_safe_call(cudaEventElapsedTime(&elapsed, start, stop));
+
+  [[maybe_unused]] float expected = 1.0f * NTASKS * ms;
+
+  /* We cannot really expect this measurement to be accurate because the
+   * thread(s) executing the code might be preempted on a system with a high load
+   * (as during unit tests). So the best we can expect is that the elapsed time
+   * is larger than the sleep time, but event the timer on the GPU is not
+   * perfectly accurate so we do not make any strict assumptions about the
+   * test, and just keep this test to demonstrate how to use the mechanisms,
+   * and ensure they are functional . */
+  // EXPECT(elapsed >= expected);
+}
+
+int main(int argc, char** argv)
+{
+  int NTASKS = 25;
+  int ms     = 200;
+
+  if (argc > 1)
+  {
+    NTASKS = atoi(argv[1]);
+  }
+
+  if (argc > 2)
+  {
+    ms = atoi(argv[2]);
+  }
+
+  run<context>(NTASKS, ms);
+  run<stream_ctx>(NTASKS, ms);
+  run<graph_ctx>(NTASKS, ms);
+}