File: pointer.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 (87 lines) | stat: -rw-r--r-- 2,981 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
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
//===-- runtime/CUDA/pointer.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/pointer.h"
#include "../assign-impl.h"
#include "../stat.h"
#include "../terminator.h"
#include "flang/Runtime/CUDA/descriptor.h"
#include "flang/Runtime/CUDA/memmove-function.h"
#include "flang/Runtime/pointer.h"

#include "cuda_runtime.h"

namespace Fortran::runtime::cuda {

extern "C" {
RT_EXT_API_GROUP_BEGIN

int RTDEF(CUFPointerAllocate)(Descriptor &desc, int64_t stream, bool hasStat,
    const Descriptor *errMsg, const char *sourceFile, int sourceLine) {
  if (desc.HasAddendum()) {
    Terminator terminator{sourceFile, sourceLine};
    // TODO: This require a bit more work to set the correct type descriptor
    // address
    terminator.Crash(
        "not yet implemented: CUDA descriptor allocation with addendum");
  }
  // Perform the standard allocation.
  int stat{
      RTNAME(PointerAllocate)(desc, hasStat, errMsg, sourceFile, sourceLine)};
  return stat;
}

int RTDEF(CUFPointerAllocateSync)(Descriptor &desc, int64_t stream,
    bool hasStat, const Descriptor *errMsg, const char *sourceFile,
    int sourceLine) {
  int stat{RTNAME(CUFPointerAllocate)(
      desc, stream, hasStat, errMsg, sourceFile, sourceLine)};
#ifndef RT_DEVICE_COMPILATION
  // Descriptor synchronization is only done when the allocation is done
  // from the host.
  if (stat == StatOk) {
    void *deviceAddr{
        RTNAME(CUFGetDeviceAddress)((void *)&desc, sourceFile, sourceLine)};
    RTNAME(CUFDescriptorSync)
    ((Descriptor *)deviceAddr, &desc, sourceFile, sourceLine);
  }
#endif
  return stat;
}

int RTDEF(CUFPointerAllocateSource)(Descriptor &pointer,
    const Descriptor &source, int64_t stream, bool hasStat,
    const Descriptor *errMsg, const char *sourceFile, int sourceLine) {
  int stat{RTNAME(CUFPointerAllocate)(
      pointer, stream, hasStat, errMsg, sourceFile, sourceLine)};
  if (stat == StatOk) {
    Terminator terminator{sourceFile, sourceLine};
    Fortran::runtime::DoFromSourceAssign(
        pointer, source, terminator, &MemmoveHostToDevice);
  }
  return stat;
}

int RTDEF(CUFPointerAllocateSourceSync)(Descriptor &pointer,
    const Descriptor &source, int64_t stream, bool hasStat,
    const Descriptor *errMsg, const char *sourceFile, int sourceLine) {
  int stat{RTNAME(CUFPointerAllocateSync)(
      pointer, stream, hasStat, errMsg, sourceFile, sourceLine)};
  if (stat == StatOk) {
    Terminator terminator{sourceFile, sourceLine};
    Fortran::runtime::DoFromSourceAssign(
        pointer, source, terminator, &MemmoveHostToDevice);
  }
  return stat;
}

RT_EXT_API_GROUP_END

} // extern "C"

} // namespace Fortran::runtime::cuda