[llvm-branch-commits] [flang] [flang][cuda] Add CUF allocator (PR #101216)

[Slava Zakharin via llvm-branch-commits]

================
@@ -0,0 +1,43 @@
+//===-- include/flang/Runtime/CUDA/allocator.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_ALLOCATOR_H_
+#define FORTRAN_RUNTIME_CUDA_ALLOCATOR_H_
+
+#include "flang/Runtime/descriptor.h"
+
+static constexpr unsigned kPinnedAllocatorPos = 1;
+static constexpr unsigned kDeviceAllocatorPos = 2;
+static constexpr unsigned kManagedAllocatorPos = 3;
+
+#define CUDA_REPORT_IF_ERROR(expr) \
+  [](CUresult result) { \
+    if (!result) \
+      return; \
+    const char *name = nullptr; \
+    cuGetErrorName(result, &name); \
+    if (!name) \
+      name = "<unknown>"; \
+    fprintf(stderr, "'%s' failed with '%s'\n", #expr, name); \
----------------
vzakhari wrote:

It might be best to use `FortranRuntime` `terminator` class to report and terminate in case of an error. Ideally, we would also provide the source code location from the compiler to the runtime.

https://github.com/llvm/llvm-project/pull/101216