[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/libcudacxx/include/cuda/std/__barrier/barrier.h

@@ -0,0 +1,227 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// 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) 2024 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CUDA_STD___BARRIER_BARRIER_H
+#define __CUDA_STD___BARRIER_BARRIER_H
+
+#include <cuda/std/detail/__config>
+
+#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
+#  pragma GCC system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
+#  pragma clang system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
+#  pragma system_header
+#endif // no system header
+
+#include <cuda/std/__barrier/empty_completion.h>
+#include <cuda/std/__barrier/poll_tester.h>
+#include <cuda/std/__chrono/duration.h>
+#include <cuda/std/__new_>
+#include <cuda/std/atomic>
+#include <cuda/std/cstddef>
+
+#include <cuda/std/__cccl/prologue.h>
+
+_CCCL_BEGIN_NAMESPACE_CUDA_STD
+
+_CCCL_DIAG_PUSH
+_CCCL_DIAG_SUPPRESS_MSVC(4324) // structure was padded due to alignment specifier
+
+template <class _CompletionF, thread_scope _Sco = thread_scope_system>
+class __barrier_base
+{
+  __atomic_impl<ptrdiff_t, _Sco> __expected, __arrived;
+  _CompletionF __completion;
+  __atomic_impl<bool, _Sco> __phase;
+
+public:
+  using arrival_token = bool;
+
+private:
+  template <typename _Barrier>
+  friend class __barrier_poll_tester_phase;
+  template <typename _Barrier>
+  friend class __barrier_poll_tester_parity;
+  template <typename _Barrier>
+  _CCCL_HOST_DEVICE_API friend bool __call_try_wait(const _Barrier& __b, typename _Barrier::arrival_token&& __phase);
+  template <typename _Barrier>
+  _CCCL_HOST_DEVICE_API friend bool __call_try_wait_parity(const _Barrier& __b, bool __parity);
+
+  [[nodiscard]] _CCCL_HOST_DEVICE_API bool __try_wait(arrival_token __old) const
+  {
+    return __phase.load(memory_order_acquire) != __old;
+  }
+  [[nodiscard]] _CCCL_HOST_DEVICE_API bool __try_wait_parity(bool __parity) const
+  {
+    return __try_wait(__parity);
+  }
+
+public:
+  _CCCL_HIDE_FROM_ABI __barrier_base() = default;
+
+  _CCCL_HOST_DEVICE_API __barrier_base(ptrdiff_t __expected, _CompletionF __completion = _CompletionF())
+      : __expected(__expected)
+      , __arrived(__expected)
+      , __completion(__completion)
+      , __phase(false)
+  {}
+
+  _CCCL_HIDE_FROM_ABI ~__barrier_base() = default;
+
+  __barrier_base(__barrier_base const&)            = delete;
+  __barrier_base& operator=(__barrier_base const&) = delete;
+
+  /*discard*/ _CCCL_HOST_DEVICE_API arrival_token arrive(ptrdiff_t __update = 1)
+  {
+    auto const __old_phase    = __phase.load(memory_order_relaxed);
+    auto const __result       = __arrived.fetch_sub(__update, memory_order_acq_rel) - __update;
+    auto const __new_expected = __expected.load(memory_order_relaxed);
+
+    _CCCL_ASSERT(__result >= 0, "");
+
+    if (0 == __result)
+    {
+      __completion();
+      __arrived.store(__new_expected, memory_order_relaxed);
+      __phase.store(!__old_phase, memory_order_release);
+      __atomic_notify_all(&__phase.__a, __scope_to_tag<_Sco>{});
+    }
+    return __old_phase;
+  }
+  _CCCL_HOST_DEVICE_API void wait(arrival_token&& __old_phase) const
+  {
+    __phase.wait(__old_phase, memory_order_acquire);
+  }
+  _CCCL_HOST_DEVICE_API void arrive_and_wait()
+  {
+    wait(arrive());
+  }
+  _CCCL_HOST_DEVICE_API void arrive_and_drop()
+  {
+    __expected.fetch_sub(1, memory_order_relaxed);
+    arrive();
+  }
+
+  [[nodiscard]] _CCCL_HOST_DEVICE_API static constexpr ptrdiff_t max() noexcept
+  {
+    return numeric_limits<ptrdiff_t>::max();
+  }
+};
+
+_CCCL_DIAG_POP
+
+template <thread_scope _Sco>
+class __barrier_base<__empty_completion, _Sco>
+{
+  static constexpr uint64_t __expected_unit = 1ull;
+  static constexpr uint64_t __arrived_unit  = 1ull << 32;
+  static constexpr uint64_t __expected_mask = __arrived_unit - 1;
+  static constexpr uint64_t __phase_bit     = 1ull << 63;
+  static constexpr uint64_t __arrived_mask  = (__phase_bit - 1) & ~__expected_mask;
+
+  __atomic_impl<uint64_t, _Sco> __phase_arrived_expected;
+
+public:
+  using arrival_token = uint64_t;
+
+private:
+  template <typename _Barrier>
+  friend class __barrier_poll_tester_phase;
+  template <typename _Barrier>
+  friend class __barrier_poll_tester_parity;
+  template <typename _Barrier>
+  _CCCL_HOST_DEVICE_API friend bool __call_try_wait(const _Barrier& __b, typename _Barrier::arrival_token&& __phase);
+  template <typename _Barrier>
+  _CCCL_HOST_DEVICE_API friend bool __call_try_wait_parity(const _Barrier& __b, bool __parity);
+
+  static _CCCL_HOST_DEVICE_API constexpr uint64_t __init(ptrdiff_t __count) noexcept
+  {
+    _CCCL_ASSERT(__count >= 0, "Count must be non-negative.");
+    return (((1u << 31) - __count) << 32) | ((1u << 31) - __count);
+  }
+  [[nodiscard]] _CCCL_HOST_DEVICE_API bool __try_wait_phase(uint64_t __phase) const
+  {
+    uint64_t const __current = __phase_arrived_expected.load(memory_order_acquire);
+    return ((__current & __phase_bit) != __phase);
+  }
+  [[nodiscard]] _CCCL_HOST_DEVICE_API bool __try_wait(arrival_token __old) const
+  {
+    return __try_wait_phase(__old & __phase_bit);
+  }
+  [[nodiscard]] _CCCL_HOST_DEVICE_API bool __try_wait_parity(bool __parity) const
+  {
+    return __try_wait_phase(__parity ? __phase_bit : 0);
+  }
+
+public:
+  _CCCL_HIDE_FROM_ABI __barrier_base() = default;
+
+  _CCCL_HOST_DEVICE_API constexpr __barrier_base(ptrdiff_t __count, __empty_completion = __empty_completion())
+      : __phase_arrived_expected(__init(__count))
+  {
+    _CCCL_ASSERT(__count >= 0, "");
+  }
+
+  _CCCL_HIDE_FROM_ABI ~__barrier_base() = default;
+
+  __barrier_base(__barrier_base const&)            = delete;
+  __barrier_base& operator=(__barrier_base const&) = delete;
+
+  /*discard*/ _CCCL_HOST_DEVICE_API arrival_token arrive(ptrdiff_t __update = 1)
+  {
+    auto const __inc = __arrived_unit * __update;
+    auto const __old = __phase_arrived_expected.fetch_add(__inc, memory_order_acq_rel);
+    if ((__old ^ (__old + __inc)) & __phase_bit)
+    {
+      __phase_arrived_expected.fetch_add((__old & __expected_mask) << 32, memory_order_relaxed);
+      __phase_arrived_expected.notify_all();
+    }
+    return __old & __phase_bit;
+  }
+  _CCCL_HOST_DEVICE_API void wait(arrival_token&& __phase) const
+  {
+    ::cuda::std::__cccl_thread_poll_with_backoff(
+      __barrier_poll_tester_phase<__barrier_base>(this, ::cuda::std::move(__phase)));
+  }
+  _CCCL_HOST_DEVICE_API void wait_parity(bool __parity) const
+  {
+    ::cuda::std::__cccl_thread_poll_with_backoff(__barrier_poll_tester_parity<__barrier_base>(this, __parity));
+  }
+  _CCCL_HOST_DEVICE_API void arrive_and_wait()
+  {
+    wait(arrive());
+  }
+  _CCCL_HOST_DEVICE_API void arrive_and_drop()
+  {
+    __phase_arrived_expected.fetch_add(__expected_unit, memory_order_relaxed);
+    arrive();
+  }
+
+  [[nodiscard]] _CCCL_HOST_DEVICE_API static constexpr ptrdiff_t max() noexcept
+  {
+    return numeric_limits<int32_t>::max();
+  }
+};
+
+template <class _CompletionF = __empty_completion>
+class barrier : public __barrier_base<_CompletionF>
+{
+public:
+  _CCCL_HOST_DEVICE_API constexpr barrier(ptrdiff_t __count, _CompletionF __completion = _CompletionF())
+      : __barrier_base<_CompletionF>(__count, __completion)
+  {}
+};
+
+_CCCL_END_NAMESPACE_CUDA_STD
+
+#include <cuda/std/__cccl/epilogue.h>
+
+#endif // __CUDA_STD___BARRIER_BARRIER_H

cccl_upstream/libcudacxx/include/cuda/std/__barrier/empty_completion.h

@@ -0,0 +1,37 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// 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) 2024 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CUDA_STD___BARRIER_EMPTY_COMPLETION_H
+#define __CUDA_STD___BARRIER_EMPTY_COMPLETION_H
+
+#include <cuda/std/detail/__config>
+
+#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
+#  pragma GCC system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
+#  pragma clang system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
+#  pragma system_header
+#endif // no system header
+
+#include <cuda/std/__cccl/prologue.h>
+
+_CCCL_BEGIN_NAMESPACE_CUDA_STD
+
+struct __empty_completion
+{
+  _CCCL_HOST_DEVICE_API constexpr void operator()() noexcept {}
+};
+
+_CCCL_END_NAMESPACE_CUDA_STD
+
+#include <cuda/std/__cccl/epilogue.h>
+
+#endif // __CUDA_STD___BARRIER_EMPTY_COMPLETION_H

cccl_upstream/libcudacxx/include/cuda/std/__barrier/poll_tester.h

@@ -0,0 +1,84 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// 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) 2024 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CUDA_STD___BARRIER_POLL_TESTER_H
+#define __CUDA_STD___BARRIER_POLL_TESTER_H
+
+#include <cuda/std/detail/__config>
+
+#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
+#  pragma GCC system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
+#  pragma clang system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
+#  pragma system_header
+#endif // no system header
+
+#include <cuda/std/__utility/move.h>
+
+#include <cuda/std/__cccl/prologue.h>
+
+_CCCL_BEGIN_NAMESPACE_CUDA_STD
+
+template <class _Barrier>
+class __barrier_poll_tester_phase
+{
+  _Barrier const* __this;
+  typename _Barrier::arrival_token __phase;
+
+public:
+  _CCCL_HOST_DEVICE_API inline __barrier_poll_tester_phase(
+    _Barrier const* __this_, typename _Barrier::arrival_token&& __phase_)
+      : __this(__this_)
+      , __phase(::cuda::std::move(__phase_))
+  {}
+
+  [[nodiscard]] _CCCL_HOST_DEVICE_API inline bool operator()() const
+  {
+    return __this->__try_wait(__phase);
+  }
+};
+
+template <class _Barrier>
+class __barrier_poll_tester_parity
+{
+  _Barrier const* __this;
+  bool __parity;
+
+public:
+  _CCCL_HOST_DEVICE_API inline __barrier_poll_tester_parity(_Barrier const* __this_, bool __parity_)
+      : __this(__this_)
+      , __parity(__parity_)
+  {}
+
+  [[nodiscard]] _CCCL_HOST_DEVICE_API inline bool operator()() const
+  {
+    return __this->__try_wait_parity(__parity);
+  }
+};
+
+template <class _Barrier>
+[[nodiscard]] _CCCL_HOST_DEVICE_API inline bool
+__call_try_wait(const _Barrier& __b, typename _Barrier::arrival_token&& __phase)
+{
+  return __b.__try_wait(::cuda::std::move(__phase));
+}
+
+template <class _Barrier>
+[[nodiscard]] _CCCL_HOST_DEVICE_API inline bool __call_try_wait_parity(const _Barrier& __b, bool __parity)
+{
+  return __b.__try_wait_parity(__parity);
+}
+
+_CCCL_END_NAMESPACE_CUDA_STD
+
+#include <cuda/std/__cccl/epilogue.h>
+
+#endif // __CUDA_STD___BARRIER_POLL_TESTER_H