init.cc 10.2 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
D
dzhwinter 已提交
15
#include <algorithm>
C
chengduo 已提交
16 17
#include <fstream>
#include <iostream>
Q
qingqing01 已提交
18 19
#include <memory>
#include <set>
D
dzhwinter 已提交
20
#include <stdexcept>
D
dzhwinter 已提交
21 22
#include <string>

Y
Yi Wang 已提交
23
#include "paddle/fluid/framework/operator.h"
T
tensor-tang 已提交
24
#include "paddle/fluid/platform/cpu_helper.h"
T
tensor-tang 已提交
25
#include "paddle/fluid/platform/cpu_info.h"
26
#include "paddle/fluid/string/split.h"
S
sneaxiy 已提交
27
#ifdef PADDLE_WITH_CUDA
Y
Yu Yang 已提交
28
#include "paddle/fluid/platform/cuda_device_guard.h"
29
#include "paddle/fluid/platform/dynload/cupti.h"
S
sneaxiy 已提交
30
#endif
Y
Yi Wang 已提交
31
#include "paddle/fluid/platform/device_context.h"
32
#include "paddle/fluid/platform/init.h"
Y
Yi Wang 已提交
33
#include "paddle/fluid/platform/place.h"
34
#include "paddle/fluid/string/piece.h"
35

36
DECLARE_int32(paddle_num_threads);
37 38 39
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
40

41 42 43 44 45 46 47 48 49 50
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
51 52 53
namespace paddle {
namespace framework {

54 55 56 57
#ifdef _WIN32
#define strdup _strdup
#endif

D
dzhwinter 已提交
58
std::once_flag gflags_init_flag;
59
std::once_flag glog_init_flag;
X
Xin Pan 已提交
60
std::once_flag p2p_init_flag;
61
std::once_flag glog_warning_once_flag;
D
dzhwinter 已提交
62

63
bool InitGflags(std::vector<std::string> args) {
64
  bool successed = false;
D
dzhwinter 已提交
65
  std::call_once(gflags_init_flag, [&]() {
C
chengduo 已提交
66
    FLAGS_logtostderr = true;
L
Leo Chen 已提交
67 68 69 70 71
    // 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.
72 73
    args.insert(args.begin(), "dummy");
    std::vector<char *> argv;
D
dzhwinter 已提交
74
    std::string line;
75 76 77 78
    int argc = args.size();
    for (auto &arg : args) {
      argv.push_back(const_cast<char *>(arg.data()));
      line += arg;
D
dzhwinter 已提交
79 80
      line += ' ';
    }
L
Leo Chen 已提交
81 82
    VLOG(1) << "Before Parse: argc is " << argc
            << ", Init commandline: " << line;
83 84

    char **arr = argv.data();
D
dzhwinter 已提交
85
    google::ParseCommandLineFlags(&argc, &arr, true);
86
    successed = true;
87 88

    VLOG(1) << "After Parse: argc is " << argc;
D
dzhwinter 已提交
89
  });
90
  return successed;
D
dzhwinter 已提交
91 92
}

93 94 95 96 97 98 99 100
void InitP2P(std::vector<int> devices) {
#ifdef PADDLE_WITH_CUDA
  std::call_once(p2p_init_flag, [&]() {
    int count = devices.size();
    for (int i = 0; i < count; ++i) {
      for (int j = 0; j < count; ++j) {
        if (devices[i] == devices[j]) continue;
        int can_acess = -1;
G
GaoWei8 已提交
101 102
        PADDLE_ENFORCE_CUDA_SUCCESS(
            cudaDeviceCanAccessPeer(&can_acess, devices[i], devices[j]));
103 104 105 106
        if (can_acess != 1) {
          LOG(WARNING) << "Cannot enable P2P access from " << devices[i]
                       << " to " << devices[j];
        } else {
Y
Yu Yang 已提交
107
          platform::CUDADeviceGuard guard(devices[i]);
108 109 110 111 112 113 114 115
          cudaDeviceEnablePeerAccess(devices[j], 0);
        }
      }
    }
  });
#endif
}

116 117 118 119
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
120 121 122 123 124 125 126 127 128 129 130 131
#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.")); \
132 133 134 135 136 137 138 139 140 141
  }
  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
}

X
Xin Pan 已提交
142
void InitDevices(bool init_p2p) {
143 144 145
  // CUPTI attribute should be set before any CUDA context is created (see CUPTI
  // documentation about CUpti_ActivityAttribute).
  InitCupti();
146
  /*Init all available devices by default */
147
  std::vector<int> devices;
148
#ifdef PADDLE_WITH_CUDA
D
dzhwinter 已提交
149
  try {
150 151
    // use user specified GPUs in single-node multi-process mode.
    devices = platform::GetSelectedDevices();
D
dzhwinter 已提交
152 153
  } catch (const std::exception &exp) {
    LOG(WARNING) << "Compiled with WITH_GPU, but no GPU found in runtime.";
154
  }
D
dzhwinter 已提交
155
#endif
156
  InitDevices(init_p2p, devices);
D
dzhwinter 已提交
157 158
}

159 160 161 162
void InitDevices(bool init_p2p, const std::vector<int> devices) {
  std::vector<platform::Place> places;

  for (size_t i = 0; i < devices.size(); ++i) {
163 164 165
    // In multi process multi gpu mode, we may have gpuid = 7
    // but count = 1.
    if (devices[i] < 0) {
166 167 168 169 170 171 172 173 174
      LOG(WARNING) << "Invalid devices id.";
      continue;
    }
    places.emplace_back(platform::CUDAPlace(devices[i]));
  }
  if (init_p2p) {
    InitP2P(devices);
  }
  places.emplace_back(platform::CPUPlace());
175 176 177
#ifdef PADDLE_WITH_CUDA
  places.emplace_back(platform::CUDAPinnedPlace());
#endif
178
  platform::DeviceContextPool::Init(places);
Q
qingqing01 已提交
179

180
#ifndef PADDLE_WITH_MKLDNN
T
tensor-tang 已提交
181
  platform::SetNumThreads(FLAGS_paddle_num_threads);
182
#endif
T
tensor-tang 已提交
183

T
tensor-tang 已提交
184
#if !defined(_WIN32) && !defined(__APPLE__) && !defined(__OSX__)
T
tensor-tang 已提交
185
  if (platform::MayIUse(platform::avx)) {
T
tensor-tang 已提交
186 187 188 189
#ifndef __AVX__
    LOG(WARNING) << "AVX is available, Please re-compile on local machine";
#endif
  }
190 191

// Throw some informations when CPU instructions mismatch.
192 193 194 195 196 197
#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."))
198 199

#ifdef __AVX512F__
T
tensor-tang 已提交
200 201
  if (!platform::MayIUse(platform::avx512f)) {
    if (platform::MayIUse(platform::avx2)) {
202
      AVX_GUIDE(AVX512, AVX2);
T
tensor-tang 已提交
203
    } else if (platform::MayIUse(platform::avx)) {
204 205 206 207 208
      AVX_GUIDE(AVX512, AVX);
    } else {
      AVX_GUIDE(AVX512, NonAVX);
    }
  }
T
tensor-tang 已提交
209
#endif
210 211

#ifdef __AVX2__
T
tensor-tang 已提交
212 213
  if (!platform::MayIUse(platform::avx2)) {
    if (platform::MayIUse(platform::avx)) {
214 215 216 217
      AVX_GUIDE(AVX2, AVX);
    } else {
      AVX_GUIDE(AVX2, NonAVX);
    }
T
tensor-tang 已提交
218 219
  }
#endif
220 221

#ifdef __AVX__
T
tensor-tang 已提交
222
  if (!platform::MayIUse(platform::avx)) {
223
    AVX_GUIDE(AVX, NonAVX);
T
tensor-tang 已提交
224
  }
225 226
#endif
#undef AVX_GUIDE
T
tensor-tang 已提交
227 228

#endif
229 230
}

C
chengduo 已提交
231
#ifndef _WIN32
232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263
// 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.
std::ostringstream signal_msg_dumper;
264
void SignalHandle(const char *data, int size) {
C
chengduo 已提交
265
  try {
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
    // NOTE1: The glog FailureSignalHandler dumped messages
    //   are deal with line by line
    // NOTE2: we only deal with the time info ane signal info,
    //   the stack trace will generated by paddle self
    if (StartsWith(data, "*** Aborted at")) {
      signal_msg_dumper << "  [TimeInfo: " << std::string(data, size - 1)
                        << "]\n";
    } 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(), "");
      signal_msg_dumper << "  [SignalInfo: " << signal_info << "]\n";
      // NOTE3: Here does not throw an exception,
      // otherwise it will casue "terminate called recursively"
      auto exp = platform::EnforceNotMet(
          platform::errors::Fatal(
              "A serious error (%s) is detected by the operating system.",
              ParseSignalErrorString(signal_info)),
          __FILE__, __LINE__);
      std::cout << exp.what() << signal_msg_dumper.str() << std::endl;
    }
C
chengduo 已提交
288
  } catch (...) {
289 290 291
    // 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
292 293 294 295
  }
}
#endif

Y
Yang Yu 已提交
296
void InitGLOG(const std::string &prog_name) {
297 298 299 300
  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
301
#ifndef _WIN32
302 303
    google::InstallFailureSignalHandler();
    google::InstallFailureWriter(&SignalHandle);
C
chengduo 已提交
304
#endif
305
  });
Y
Yang Yu 已提交
306 307
}

D
dzhwinter 已提交
308 309
}  // namespace framework
}  // namespace paddle