[flang-commits] [flang] [flang][cuda] Allocate local descriptor in managed memory (PR #102060)

[Valentin Clement バレンタイン クレメン via flang-commits]

================
@@ -0,0 +1,30 @@
+//===-- include/flang/Runtime/CUDA/descriptor.h -----------------*- C++ -*-===//
+//
+// Part of the LLVM Project, 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef FORTRAN_RUNTIME_CUDA_DESCRIPTOR_H_
+#define FORTRAN_RUNTIME_CUDA_DESCRIPTOR_H_
+
+#include "flang/Runtime/descriptor.h"
+#include "flang/Runtime/entry-names.h"
+#include <cstddef>
+
+namespace Fortran::runtime::cuf {
+
+extern "C" {
+
+// Allocate a descriptor in managed or unified memory.
----------------
clementval wrote:

My understanding is that the same cuda call is issued for both. It depends on what the underlying system support to choose one or the other.

https://github.com/llvm/llvm-project/pull/102060