File: allocator.cpp

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.8-1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,696 kB
  • sloc: cpp: 7,438,781; ansic: 1,393,871; asm: 1,012,926; python: 241,771; f90: 86,635; objc: 75,411; lisp: 42,144; pascal: 17,286; sh: 8,596; ml: 5,082; perl: 4,730; makefile: 3,591; awk: 3,523; javascript: 2,251; xml: 892; fortran: 672
file content (71 lines) | stat: -rw-r--r-- 2,114 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
//===-- runtime/CUDA/allocator.cpp ----------------------------------------===//
//
// 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
//
//===----------------------------------------------------------------------===//

#include "flang/Runtime/CUDA/allocator.h"
#include "../derived.h"
#include "../stat.h"
#include "../terminator.h"
#include "../type-info.h"
#include "flang/Common/Fortran.h"
#include "flang/ISO_Fortran_binding_wrapper.h"
#include "flang/Runtime/CUDA/common.h"
#include "flang/Runtime/allocator-registry.h"

#include "cuda_runtime.h"

namespace Fortran::runtime::cuda {
extern "C" {

void RTDEF(CUFRegisterAllocator)() {
  allocatorRegistry.Register(
      kPinnedAllocatorPos, {&CUFAllocPinned, CUFFreePinned});
  allocatorRegistry.Register(
      kDeviceAllocatorPos, {&CUFAllocDevice, CUFFreeDevice});
  allocatorRegistry.Register(
      kManagedAllocatorPos, {&CUFAllocManaged, CUFFreeManaged});
  allocatorRegistry.Register(
      kUnifiedAllocatorPos, {&CUFAllocUnified, CUFFreeUnified});
}
}

void *CUFAllocPinned(std::size_t sizeInBytes) {
  void *p;
  CUDA_REPORT_IF_ERROR(cudaMallocHost((void **)&p, sizeInBytes));
  return p;
}

void CUFFreePinned(void *p) { CUDA_REPORT_IF_ERROR(cudaFreeHost(p)); }

void *CUFAllocDevice(std::size_t sizeInBytes) {
  void *p;
  CUDA_REPORT_IF_ERROR(cudaMalloc(&p, sizeInBytes));
  return p;
}

void CUFFreeDevice(void *p) { CUDA_REPORT_IF_ERROR(cudaFree(p)); }

void *CUFAllocManaged(std::size_t sizeInBytes) {
  void *p;
  CUDA_REPORT_IF_ERROR(
      cudaMallocManaged((void **)&p, sizeInBytes, cudaMemAttachGlobal));
  return reinterpret_cast<void *>(p);
}

void CUFFreeManaged(void *p) { CUDA_REPORT_IF_ERROR(cudaFree(p)); }

void *CUFAllocUnified(std::size_t sizeInBytes) {
  // Call alloc managed for the time being.
  return CUFAllocManaged(sizeInBytes);
}

void CUFFreeUnified(void *p) {
  // Call free managed for the time being.
  CUFFreeManaged(p);
}

} // namespace Fortran::runtime::cuda