init.cc 10.1 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
D
dzhwinter 已提交
2

L
Luo Tao 已提交
3 4 5
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
D
dzhwinter 已提交
6

L
Luo Tao 已提交
7
    http://www.apache.org/licenses/LICENSE-2.0
D
dzhwinter 已提交
8

L
Luo Tao 已提交
9 10 11 12 13
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
Y
Yang Yu 已提交
14
#include <string.h>  // for strdup
L
Leo Chen 已提交
15

D
dzhwinter 已提交
16
#include <algorithm>
C
chengduo 已提交
17 18
#include <fstream>
#include <iostream>
Q
qingqing01 已提交
19 20
#include <memory>
#include <set>
D
dzhwinter 已提交
21
#include <stdexcept>
D
dzhwinter 已提交
22 23
#include <string>

Y
Yi Wang 已提交
24
#include "paddle/fluid/framework/operator.h"
T
tensor-tang 已提交
25
#include "paddle/fluid/platform/cpu_helper.h"
T
tensor-tang 已提交
26
#include "paddle/fluid/platform/cpu_info.h"
27
#include "paddle/fluid/string/split.h"
S
sneaxiy 已提交
28
#ifdef PADDLE_WITH_CUDA
Y
Yu Yang 已提交
29
#include "paddle/fluid/platform/cuda_device_guard.h"
30
#include "paddle/fluid/platform/dynload/cupti.h"
S
sneaxiy 已提交
31
#endif
Y
Yi Wang 已提交
32
#include "paddle/fluid/platform/device_context.h"
33
#include "paddle/fluid/platform/init.h"
Y
Yi Wang 已提交
34
#include "paddle/fluid/platform/place.h"
35
#include "paddle/fluid/string/piece.h"
36

37 38 39 40 41
#ifdef PADDLE_WITH_XPU
#include "paddle/fluid/platform/xpu_header.h"
#include "paddle/fluid/platform/xpu_info.h"
#endif

42
DECLARE_int32(paddle_num_threads);
43 44 45
DEFINE_int32(multiple_of_cupti_buffer_size, 1,
             "Multiple of the CUPTI device buffer size. If the timestamps have "
             "been dropped when you are profiling, try increasing this value.");
T
tensor-tang 已提交
46

47 48 49 50 51 52 53 54 55 56
namespace paddle {
namespace platform {

void ParseCommandLineFlags(int argc, char **argv, bool remove) {
  google::ParseCommandLineFlags(&argc, &argv, remove);
}

}  // namespace platform
}  // namespace paddle

D
dzhwinter 已提交
57 58 59
namespace paddle {
namespace framework {

60 61 62 63
#ifdef _WIN32
#define strdup _strdup
#endif

D
dzhwinter 已提交
64
std::once_flag gflags_init_flag;
65
std::once_flag glog_init_flag;
D
dzhwinter 已提交
66

67
bool InitGflags(std::vector<std::string> args) {
68
  bool successed = false;
D
dzhwinter 已提交
69
  std::call_once(gflags_init_flag, [&]() {
C
chengduo 已提交
70
    FLAGS_logtostderr = true;
L
Leo Chen 已提交
71 72 73 74 75
    // NOTE(zhiqiu): dummy is needed, since the function
    // ParseNewCommandLineFlags in gflags.cc starts processing
    // commandline strings from idx 1.
    // The reason is, it assumes that the first one (idx 0) is
    // the filename of executable file.
76 77
    args.insert(args.begin(), "dummy");
    std::vector<char *> argv;
D
dzhwinter 已提交
78
    std::string line;
79 80 81 82
    int argc = args.size();
    for (auto &arg : args) {
      argv.push_back(const_cast<char *>(arg.data()));
      line += arg;
D
dzhwinter 已提交
83 84
      line += ' ';
    }
L
Leo Chen 已提交
85 86
    VLOG(1) << "Before Parse: argc is " << argc
            << ", Init commandline: " << line;
87 88

    char **arr = argv.data();
D
dzhwinter 已提交
89
    google::ParseCommandLineFlags(&argc, &arr, true);
90
    successed = true;
91 92

    VLOG(1) << "After Parse: argc is " << argc;
D
dzhwinter 已提交
93
  });
94
  return successed;
D
dzhwinter 已提交
95 96
}

97

98

99 100 101 102
void InitCupti() {
#ifdef PADDLE_WITH_CUPTI
  if (FLAGS_multiple_of_cupti_buffer_size == 1) return;
  size_t attrValue = 0, attrValueSize = sizeof(size_t);
G
GaoWei8 已提交
103 104 105 106 107 108 109 110 111 112 113 114
#define MULTIPLY_ATTR_VALUE(attr)                                            \
  {                                                                          \
    PADDLE_ENFORCE_EQ(                                                       \
        !platform::dynload::cuptiActivityGetAttribute(attr, &attrValueSize,  \
                                                      &attrValue),           \
        true, platform::errors::Unavailable("Get cupti attribute failed.")); \
    attrValue *= FLAGS_multiple_of_cupti_buffer_size;                        \
    LOG(WARNING) << "Set " #attr " " << attrValue << " byte";                \
    PADDLE_ENFORCE_EQ(                                                       \
        !platform::dynload::cuptiActivitySetAttribute(attr, &attrValueSize,  \
                                                      &attrValue),           \
        true, platform::errors::Unavailable("Set cupti attribute failed.")); \
115 116 117 118 119 120 121 122 123 124
  }
  MULTIPLY_ATTR_VALUE(CUPTI_ACTIVITY_ATTR_DEVICE_BUFFER_SIZE);
  MULTIPLY_ATTR_VALUE(CUPTI_ACTIVITY_ATTR_DEVICE_BUFFER_SIZE_CDP);
#if CUDA_VERSION >= 9000
  MULTIPLY_ATTR_VALUE(CUPTI_ACTIVITY_ATTR_PROFILING_SEMAPHORE_POOL_SIZE);
#endif
#undef MULTIPLY_ATTR_VALUE
#endif
}

125
void InitDevices() {
126 127 128
  // CUPTI attribute should be set before any CUDA context is created (see CUPTI
  // documentation about CUpti_ActivityAttribute).
  InitCupti();
129
  /*Init all available devices by default */
130
  std::vector<int> devices;
131
#ifdef PADDLE_WITH_CUDA
D
dzhwinter 已提交
132
  try {
133 134
    // use user specified GPUs in single-node multi-process mode.
    devices = platform::GetSelectedDevices();
D
dzhwinter 已提交
135 136
  } catch (const std::exception &exp) {
    LOG(WARNING) << "Compiled with WITH_GPU, but no GPU found in runtime.";
137
  }
138 139 140 141 142 143 144 145
#endif
#ifdef PADDLE_WITH_XPU
  try {
    // use user specified XPUs in single-node multi-process mode.
    devices = platform::GetXPUSelectedDevices();
  } catch (const std::exception &exp) {
    LOG(WARNING) << "Compiled with WITH_XPU, but no XPU found in runtime.";
  }
D
dzhwinter 已提交
146
#endif
147
  InitDevices(devices);
D
dzhwinter 已提交
148 149
}

150
void InitDevices(const std::vector<int> devices) {
151 152 153
  std::vector<platform::Place> places;

  for (size_t i = 0; i < devices.size(); ++i) {
154 155 156
    // In multi process multi gpu mode, we may have gpuid = 7
    // but count = 1.
    if (devices[i] < 0) {
157 158 159
      LOG(WARNING) << "Invalid devices id.";
      continue;
    }
160 161

#ifdef PADDLE_WITH_CUDA
162
    places.emplace_back(platform::CUDAPlace(devices[i]));
163 164 165 166
#endif
#ifdef PADDLE_WITH_XPU
    places.emplace_back(platform::XPUPlace(devices[i]));
#endif
167 168
  }
  places.emplace_back(platform::CPUPlace());
169 170 171
#ifdef PADDLE_WITH_CUDA
  places.emplace_back(platform::CUDAPinnedPlace());
#endif
172
  platform::DeviceContextPool::Init(places);
Q
qingqing01 已提交
173

174
#ifndef PADDLE_WITH_MKLDNN
T
tensor-tang 已提交
175
  platform::SetNumThreads(FLAGS_paddle_num_threads);
176
#endif
T
tensor-tang 已提交
177

T
tensor-tang 已提交
178
#if !defined(_WIN32) && !defined(__APPLE__) && !defined(__OSX__)
T
tensor-tang 已提交
179
  if (platform::MayIUse(platform::avx)) {
T
tensor-tang 已提交
180 181 182 183
#ifndef __AVX__
    LOG(WARNING) << "AVX is available, Please re-compile on local machine";
#endif
  }
184 185

// Throw some informations when CPU instructions mismatch.
186 187 188 189 190 191
#define AVX_GUIDE(compiletime, runtime)                                  \
  PADDLE_THROW(platform::errors::Unavailable(                            \
      "This version is compiled on higher instruction(" #compiletime     \
      ") system, you may encounter illegal instruction error running on" \
      " your local CPU machine. Please reinstall the " #runtime          \
      " version or compile from source code."))
192 193

#ifdef __AVX512F__
T
tensor-tang 已提交
194 195
  if (!platform::MayIUse(platform::avx512f)) {
    if (platform::MayIUse(platform::avx2)) {
196
      AVX_GUIDE(AVX512, AVX2);
T
tensor-tang 已提交
197
    } else if (platform::MayIUse(platform::avx)) {
198 199 200 201 202
      AVX_GUIDE(AVX512, AVX);
    } else {
      AVX_GUIDE(AVX512, NonAVX);
    }
  }
T
tensor-tang 已提交
203
#endif
204 205

#ifdef __AVX2__
T
tensor-tang 已提交
206 207
  if (!platform::MayIUse(platform::avx2)) {
    if (platform::MayIUse(platform::avx)) {
208 209 210 211
      AVX_GUIDE(AVX2, AVX);
    } else {
      AVX_GUIDE(AVX2, NonAVX);
    }
T
tensor-tang 已提交
212 213
  }
#endif
214 215

#ifdef __AVX__
T
tensor-tang 已提交
216
  if (!platform::MayIUse(platform::avx)) {
217
    AVX_GUIDE(AVX, NonAVX);
T
tensor-tang 已提交
218
  }
219 220
#endif
#undef AVX_GUIDE
T
tensor-tang 已提交
221 222

#endif
223 224
}

C
chengduo 已提交
225
#ifndef _WIN32
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
// Description Quoted from
// https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/signal.h.html
const struct {
  const char *name;
  const char *error_string;
} SignalErrorStrings[] = {
    {"SIGSEGV", "Segmentation fault"},
    {"SIGILL", "Illegal instruction"},
    {"SIGFPE", "Erroneous arithmetic operation"},
    {"SIGABRT", "Process abort signal"},
    {"SIGBUS", "Access to an undefined portion of a memory object"},
    {"SIGTERM", "Termination signal"},
};

bool StartsWith(const char *str, const char *prefix) {
  size_t len_prefix = strlen(prefix);
  size_t len_str = strlen(str);
  return len_str < len_prefix ? false : memcmp(prefix, str, len_prefix) == 0;
}

const char *ParseSignalErrorString(const std::string &str) {
  for (size_t i = 0;
       i < (sizeof(SignalErrorStrings) / sizeof(*(SignalErrorStrings))); ++i) {
    if (std::string::npos != str.find(SignalErrorStrings[i].name)) {
      return SignalErrorStrings[i].error_string;
    }
  }
  return "Unknown signal";
}

// Handle SIGSEGV, SIGILL, SIGFPE, SIGABRT, SIGBUS, and SIGTERM.
257
void SignalHandle(const char *data, int size) {
C
chengduo 已提交
258
  try {
259 260
    // NOTE1: The glog FailureSignalHandler dumped messages
    //   are deal with line by line
261
    auto signal_msg_dunmer_ptr = SignalMessageDumper::Instance().Get();
262 263 264
    // NOTE2: we only deal with the time info ane signal info,
    //   the stack trace will generated by paddle self
    if (StartsWith(data, "*** Aborted at")) {
265
      *signal_msg_dunmer_ptr << "\n  [TimeInfo: " << std::string(data, size - 1)
266
                             << "]\n";
267 268 269 270 271
    } else if (StartsWith(data, "***")) {
      std::string signal_info(data, size - 1);
      std::string useless_substr("; stack trace:");
      size_t start_pos = signal_info.rfind(useless_substr);
      signal_info.replace(start_pos, useless_substr.length(), "");
272
      *signal_msg_dunmer_ptr << "  [SignalInfo: " << signal_info << "]\n";
273 274 275

      // NOTE3: Final singal error message print.
      // Here does not throw an exception,
276
      // otherwise it will casue "terminate called recursively"
277 278 279 280 281 282 283 284 285
      std::ostringstream sout;
      sout << platform::GetCurrentTraceBackString();
      sout << "\n----------------------\nError Message "
              "Summary:\n----------------------\n";
      sout << platform::errors::Fatal(
                  "`%s` is detected by the operating system.",
                  ParseSignalErrorString(signal_info))
                  .to_string();
      std::cout << sout.str() << (*signal_msg_dunmer_ptr).str() << std::endl;
286
    }
C
chengduo 已提交
287
  } catch (...) {
288 289 290
    // Since the program has already triggered a system error,
    // no further processing is required here, glog FailureSignalHandler
    // will Kill program by the default signal handler
C
chengduo 已提交
291 292 293 294
  }
}
#endif

Y
Yang Yu 已提交
295
void InitGLOG(const std::string &prog_name) {
296 297 298 299
  std::call_once(glog_init_flag, [&]() {
    // glog will not hold the ARGV[0] inside.
    // Use strdup to alloc a new string.
    google::InitGoogleLogging(strdup(prog_name.c_str()));
C
chengduo 已提交
300
#ifndef _WIN32
301 302
    google::InstallFailureSignalHandler();
    google::InstallFailureWriter(&SignalHandle);
C
chengduo 已提交
303
#endif
304
  });
Y
Yang Yu 已提交
305 306
}

D
dzhwinter 已提交
307 308
}  // namespace framework
}  // namespace paddle