PTX: Add multimem instructions (#3603)

* Add multimem.ld_reduce * Add multimem.red * Add multimem.st Co-authored-by: Allard Hendriksen <[email protected]>
NVIDIA · Jan 30, 2025 · afa2ca2 · afa2ca2
1 parent b1f2e63
commit afa2ca2
Show file tree

Hide file tree

Showing 11 changed files with 210 additions and 0 deletions.
diff --git a/docs/libcudacxx/ptx/instructions.rst b/docs/libcudacxx/ptx/instructions.rst
@@ -22,6 +22,9 @@ PTX Instructions
    instructions/mbarrier_expect_tx
    instructions/mbarrier_test_wait
    instructions/mbarrier_try_wait
+   instructions/multimem_ld_reduce
+   instructions/multimem_red
+   instructions/multimem_st
    instructions/red_async
    instructions/st_async
    instructions/st_bulk

diff --git a/docs/libcudacxx/ptx/instructions/multimem_ld_reduce.rst b/docs/libcudacxx/ptx/instructions/multimem_ld_reduce.rst
@@ -0,0 +1,9 @@
+.. _libcudacxx-ptx-instructions-multimem-ld_reduce:
+
+multimem.ld_reduce
+==================
+
+-  PTX ISA:
+   `multimem.ld_reduce <https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#data-movement-and-conversion-instructions-multimem-ld-reduce-multimem-st-multimem-red>`__
+
+.. include:: generated/multimem_ld_reduce.rst
diff --git a/docs/libcudacxx/ptx/instructions/multimem_red.rst b/docs/libcudacxx/ptx/instructions/multimem_red.rst
@@ -0,0 +1,9 @@
+.. _libcudacxx-ptx-instructions-multimem-red:
+
+multimem.red
+============
+
+-  PTX ISA:
+   `multimem.red <https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#data-movement-and-conversion-instructions-multimem-ld-reduce-multimem-st-multimem-red>`__
+
+.. include:: generated/multimem_red.rst
diff --git a/docs/libcudacxx/ptx/instructions/multimem_st.rst b/docs/libcudacxx/ptx/instructions/multimem_st.rst
@@ -0,0 +1,9 @@
+.. _libcudacxx-ptx-instructions-multimem-st:
+
+multimem.st
+===========
+
+-  PTX ISA:
+   `multimem.st <https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#data-movement-and-conversion-instructions-multimem-ld-reduce-multimem-st-multimem-red>`__
+
+.. include:: generated/multimem_st.rst
diff --git a/libcudacxx/include/cuda/__ptx/instructions/multimem_ld_reduce.h b/libcudacxx/include/cuda/__ptx/instructions/multimem_ld_reduce.h
@@ -0,0 +1,37 @@
+// -*- 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_MULTIMEM_LD_REDUCE_H_
+#define _CUDA_PTX_MULTIMEM_LD_REDUCE_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/multimem_ld_reduce.h>
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_PTX
+
+#endif // _CUDA_PTX_MULTIMEM_LD_REDUCE_H_
diff --git a/libcudacxx/include/cuda/__ptx/instructions/multimem_red.h b/libcudacxx/include/cuda/__ptx/instructions/multimem_red.h
@@ -0,0 +1,37 @@
+// -*- 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_MULTIMEM_RED_H_
+#define _CUDA_PTX_MULTIMEM_RED_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/multimem_red.h>
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_PTX
+
+#endif // _CUDA_PTX_MULTIMEM_RED_H_
diff --git a/libcudacxx/include/cuda/__ptx/instructions/multimem_st.h b/libcudacxx/include/cuda/__ptx/instructions/multimem_st.h
@@ -0,0 +1,37 @@
+// -*- 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_MULTIMEM_ST_H_
+#define _CUDA_PTX_MULTIMEM_ST_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/multimem_st.h>
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_PTX
+
+#endif // _CUDA_PTX_MULTIMEM_ST_H_
diff --git a/libcudacxx/include/cuda/ptx b/libcudacxx/include/cuda/ptx
@@ -84,6 +84,9 @@
 #include <cuda/__ptx/instructions/mbarrier_expect_tx.h>
 #include <cuda/__ptx/instructions/mbarrier_init.h>
 #include <cuda/__ptx/instructions/mbarrier_wait.h>
+#include <cuda/__ptx/instructions/multimem_ld_reduce.h>
+#include <cuda/__ptx/instructions/multimem_red.h>
+#include <cuda/__ptx/instructions/multimem_st.h>
 #include <cuda/__ptx/instructions/red_async.h>
 #include <cuda/__ptx/instructions/st_async.h>
 #include <cuda/__ptx/instructions/st_bulk.h>

diff --git a/libcudacxx/test/libcudacxx/cuda/ptx/ptx.multimem.ld_reduce.compile.pass.cpp b/libcudacxx/test/libcudacxx/cuda/ptx/ptx.multimem.ld_reduce.compile.pass.cpp
@@ -0,0 +1,22 @@
+//===----------------------------------------------------------------------===//
+//
+// 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/multimem_ld_reduce.h"
+
+int main(int, char**)
+{
+  return 0;
+}
diff --git a/libcudacxx/test/libcudacxx/cuda/ptx/ptx.multimem.red.compile.pass.cpp b/libcudacxx/test/libcudacxx/cuda/ptx/ptx.multimem.red.compile.pass.cpp
@@ -0,0 +1,22 @@
+//===----------------------------------------------------------------------===//
+//
+// 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/multimem_red.h"
+
+int main(int, char**)
+{
+  return 0;
+}
diff --git a/libcudacxx/test/libcudacxx/cuda/ptx/ptx.multimem.st.compile.pass.cpp b/libcudacxx/test/libcudacxx/cuda/ptx/ptx.multimem.st.compile.pass.cpp
@@ -0,0 +1,22 @@
+//===----------------------------------------------------------------------===//
+//
+// 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/multimem_st.h"
+
+int main(int, char**)
+{
+  return 0;
+}