PTX: Add cp.async.mbarrier.arrive{.noinc}

docs/libcudacxx/ptx/instructions.rst

@@ -11,6 +11,7 @@ PTX Instructions
    instructions/cp_async_bulk_commit_group
    instructions/cp_async_bulk_wait_group
    instructions/cp_async_bulk_tensor
+   instructions/cp_async_mbarrier_arrive
    instructions/cp_reduce_async_bulk
    instructions/cp_reduce_async_bulk_tensor
    instructions/fence

docs/libcudacxx/ptx/instructions/cp_async_mbarrier_arrive.rst

@@ -0,0 +1,10 @@
+.. _libcudacxx-ptx-instructions-cp-async-mbarrier-arrive:
+
+cp.async.mbarrier.arrive
+========================
+
+-  PTX ISA:
+   `cp.async.mbarrier.arrive <https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#parallel-synchronization-and-communication-instructions-cp-async-mbarrier-arrive>`__
+
+.. include:: generated/cp_async_mbarrier_arrive.rst
+.. include:: generated/cp_async_mbarrier_arrive_noinc.rst

libcudacxx/include/cuda/__ptx/instructions/cp_async_mbarrier_arrive.h

@@ -0,0 +1,38 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// 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_PTX_CP_ASYNC_MBARRIER_ARRIVE_H_
+#define _CUDA_PTX_CP_ASYNC_MBARRIER_ARRIVE_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/__ptx/ptx_dot_variants.h>
+#include <cuda/__ptx/ptx_helper_functions.h>
+#include <cuda/std/cstdint>
+
+#include <nv/target> // __CUDA_MINIMUM_ARCH__ and friends
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA_PTX
+
+#include <cuda/__ptx/instructions/generated/cp_async_mbarrier_arrive.h>
+#include <cuda/__ptx/instructions/generated/cp_async_mbarrier_arrive_noinc.h>
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_PTX
+
+#endif // _CUDA_PTX_CP_ASYNC_MBARRIER_ARRIVE_H_

libcudacxx/include/cuda/ptx

@@ -74,6 +74,7 @@
 #include <cuda/__ptx/instructions/cp_async_bulk_commit_group.h>
 #include <cuda/__ptx/instructions/cp_async_bulk_tensor.h>
 #include <cuda/__ptx/instructions/cp_async_bulk_wait_group.h>
+#include <cuda/__ptx/instructions/cp_async_mbarrier_arrive.h>
 #include <cuda/__ptx/instructions/cp_reduce_async_bulk.h>
 #include <cuda/__ptx/instructions/cp_reduce_async_bulk_tensor.h>
 #include <cuda/__ptx/instructions/fence.h>

libcudacxx/test/libcudacxx/cuda/ptx/ptx.cp.async.mbarrier.arrive.compile.pass.cpp

@@ -0,0 +1,23 @@
+//===----------------------------------------------------------------------===//
+//
+// 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.
+//
+//===----------------------------------------------------------------------===//
+// UNSUPPORTED: libcpp-has-no-threads
+
+// <cuda/ptx>
+
+#include <cuda/ptx>
+#include <cuda/std/utility>
+
+#include "generated/cp_async_mbarrier_arrive.h"
+#include "generated/cp_async_mbarrier_arrive_noinc.h"
+
+int main(int, char**)
+{
+  return 0;
+}