ps_gpu_wrapper.cc 43.5 KB
Newer Older
T
Thunderbrook 已提交
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
// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
//
// 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
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// 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.

/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.

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

  http://www.apache.org/licenses/LICENSE-2.0

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. */

T
Thunderbrook 已提交
29
#ifdef PADDLE_WITH_HETERPS
Y
yaoxuefeng 已提交
30

T
Thunderbrook 已提交
31
#include <algorithm>
Y
yaoxuefeng 已提交
32 33
#include <deque>

T
Thunderbrook 已提交
34 35 36 37 38 39
#include "paddle/fluid/framework/fleet/ps_gpu_wrapper.h"
#include "paddle/fluid/platform/timer.h"

namespace paddle {
namespace framework {

T
Thunderbrook 已提交
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
#ifdef PADDLE_WITH_PSLIB
void AfsWrapper::init(const std::string& fs_name, const std::string& fs_user,
                      const std::string& pass_wd, const std::string& conf) {
  int ret = afs_handler_.init(fs_name.c_str(), fs_user.c_str(), pass_wd.c_str(),
                              conf.c_str());
  if (ret != 0) {
    LOG(ERROR) << "AFS Init Error";
  }
}

int AfsWrapper::remove(const std::string& path) {
  return afs_handler_.remove(path);
}

int AfsWrapper::mkdir(const std::string& path) {
  return afs_handler_.mkdir(path);
}

std::vector<std::string> AfsWrapper::list(const std::string& path) {
  return afs_handler_.list(path);
}

int AfsWrapper::exist(const std::string& path) {
  return afs_handler_.exist(path);
}

int AfsWrapper::upload(const std::string& local_file,
                       const std::string& afs_file) {
  return afs_handler_.upload_file(local_file, afs_file);
}

int AfsWrapper::download(const std::string& local_file,
                         const std::string& afs_file) {
  return afs_handler_.download_file(local_file, afs_file);
}
75 76 77 78 79 80 81 82 83 84 85 86

int AfsWrapper::touchz(const std::string& path) {
  return afs_handler_.touchz(path);
}

std::string AfsWrapper::cat(const std::string& path) {
  return afs_handler_.cat(path);
}

int AfsWrapper::mv(const std::string& old_path, const std::string& dest_path) {
  return afs_handler_.mv(old_path, dest_path);
}
T
Thunderbrook 已提交
87 88
#endif

T
Thunderbrook 已提交
89 90
std::shared_ptr<PSGPUWrapper> PSGPUWrapper::s_instance_ = NULL;
bool PSGPUWrapper::is_initialized_ = false;
T
Thunderbrook 已提交
91 92 93 94 95 96 97 98
#ifdef PADDLE_WITH_PSLIB
void PSGPUWrapper::InitAfsApi(const std::string& fs_name,
                              const std::string& fs_user,
                              const std::string& pass_wd,
                              const std::string& conf) {
  int ret = afs_handler_.init(fs_name.c_str(), fs_user.c_str(), pass_wd.c_str(),
                              conf.c_str());
  if (ret != 0) {
99
    VLOG(0) << "AFS Init Error";
T
Thunderbrook 已提交
100 101 102 103
  }
  use_afs_api_ = 1;
}
#endif
104
void PSGPUWrapper::PreBuildTask(std::shared_ptr<HeterContext> gpu_task) {
Y
yaoxuefeng 已提交
105
  VLOG(3) << "PSGPUWrapper::BuildGPUPSTask begin";
T
Thunderbrook 已提交
106 107
  platform::Timer timeline;
  timeline.Start();
108
  int device_num = heter_devices_.size();
109 110 111 112 113
  if (!multi_mf_dim_) {
    gpu_task->init(thread_keys_shard_num_, device_num);
  } else {
    gpu_task->init(thread_keys_shard_num_, device_num, multi_mf_dim_);
  }
114

Y
yaoxuefeng 已提交
115
  std::vector<std::thread> threads;
116 117 118 119 120 121 122 123 124 125 126 127 128
  if (!multi_mf_dim_) {
    thread_keys_.resize(thread_keys_thread_num_);
    for (int i = 0; i < thread_keys_thread_num_; i++) {
      thread_keys_[i].resize(thread_keys_shard_num_);
    }
  } else {
    thread_dim_keys_.resize(thread_keys_thread_num_);
    for (int i = 0; i < thread_keys_thread_num_; i++) {
      thread_dim_keys_[i].resize(thread_keys_shard_num_);
      for (int j = 0; j < thread_keys_shard_num_; j++) {
        thread_dim_keys_[i][j].resize(multi_mf_dim_);
      }
    }
Y
yaoxuefeng 已提交
129
  }
Y
yaoxuefeng 已提交
130 131 132 133

  size_t total_len = 0;
  size_t len_per_thread = 0;
  int remain = 0;
Y
yaoxuefeng 已提交
134
  size_t begin = 0;
Y
yaoxuefeng 已提交
135 136 137 138 139 140

  std::string data_set_name = std::string(typeid(*dataset_).name());

  if (data_set_name.find("SlotRecordDataset") != std::string::npos) {
    SlotRecordDataset* dataset = dynamic_cast<SlotRecordDataset*>(dataset_);
    auto input_channel = dataset->GetInputChannel();
Y
yaoxuefeng 已提交
141
    VLOG(0) << "psgpu wrapperinputslotchannle size: " << input_channel->Size();
Y
yaoxuefeng 已提交
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
    const std::deque<SlotRecord>& vec_data = input_channel->GetData();
    total_len = vec_data.size();
    len_per_thread = total_len / thread_keys_thread_num_;
    remain = total_len % thread_keys_thread_num_;
    VLOG(0) << "total len: " << total_len;
    auto gen_func = [this](const std::deque<SlotRecord>& total_data,
                           int begin_index, int end_index, int i) {
      for (auto iter = total_data.begin() + begin_index;
           iter != total_data.begin() + end_index; iter++) {
        const auto& ins = *iter;
        const auto& feasign_v = ins->slot_uint64_feasigns_.slot_values;
        for (const auto feasign : feasign_v) {
          int shard_id = feasign % thread_keys_shard_num_;
          this->thread_keys_[i][shard_id].insert(feasign);
        }
Y
yaoxuefeng 已提交
157
      }
Y
yaoxuefeng 已提交
158
    };
159 160 161 162 163 164 165 166 167 168 169 170 171
    auto gen_dynamic_mf_func = [this](const std::deque<SlotRecord>& total_data,
                                      int begin_index, int end_index, int i) {
      for (auto iter = total_data.begin() + begin_index;
           iter != total_data.begin() + end_index; iter++) {
        const auto& ins = *iter;
        const auto& feasign_v = ins->slot_uint64_feasigns_.slot_values;
        const auto& slot_offset = ins->slot_uint64_feasigns_.slot_offsets;
        for (size_t slot_idx = 0; slot_idx < slot_offset_vector_.size();
             slot_idx++) {
          for (size_t j = slot_offset[slot_offset_vector_[slot_idx]];
               j < slot_offset[slot_offset_vector_[slot_idx] + 1]; j++) {
            int shard_id = feasign_v[j] % thread_keys_shard_num_;
            int dim_id = slot_index_vec_[slot_idx];
Y
yaoxuefeng 已提交
172 173 174
            if (feasign_v[j] != 0) {
              this->thread_dim_keys_[i][shard_id][dim_id].insert(feasign_v[j]);
            }
175 176 177 178
          }
        }
      }
    };
Y
yaoxuefeng 已提交
179
    for (int i = 0; i < thread_keys_thread_num_; i++) {
180 181 182 183 184 185 186 187 188 189 190
      if (!multi_mf_dim_) {
        VLOG(0) << "yxf::psgpu wrapper genfunc";
        threads.push_back(
            std::thread(gen_func, std::ref(vec_data), begin,
                        begin + len_per_thread + (i < remain ? 1 : 0), i));
      } else {
        VLOG(0) << "yxf::psgpu wrapper genfunc with dynamic mf";
        threads.push_back(
            std::thread(gen_dynamic_mf_func, std::ref(vec_data), begin,
                        begin + len_per_thread + (i < remain ? 1 : 0), i));
      }
Y
yaoxuefeng 已提交
191
      begin += len_per_thread + (i < remain ? 1 : 0);
Y
yaoxuefeng 已提交
192
    }
Y
yaoxuefeng 已提交
193 194 195 196
    for (std::thread& t : threads) {
      t.join();
    }
    timeline.Pause();
T
Thunderbrook 已提交
197
    VLOG(0) << "GpuPs build task cost " << timeline.ElapsedSec() << " seconds.";
Y
yaoxuefeng 已提交
198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
  } else {
    CHECK(data_set_name.find("MultiSlotDataset") != std::string::npos);
    VLOG(0) << "ps_gpu_wrapper use MultiSlotDataset";
    MultiSlotDataset* dataset = dynamic_cast<MultiSlotDataset*>(dataset_);
    auto input_channel = dataset->GetInputChannel();

    const std::deque<Record>& vec_data = input_channel->GetData();
    total_len = vec_data.size();
    len_per_thread = total_len / thread_keys_thread_num_;
    remain = total_len % thread_keys_thread_num_;
    auto gen_func = [this](const std::deque<Record>& total_data,
                           int begin_index, int end_index, int i) {
      for (auto iter = total_data.begin() + begin_index;
           iter != total_data.begin() + end_index; iter++) {
        const auto& ins = *iter;
        const auto& feasign_v = ins.uint64_feasigns_;
        for (const auto feasign : feasign_v) {
          uint64_t cur_key = feasign.sign().uint64_feasign_;
          int shard_id = cur_key % thread_keys_shard_num_;
          this->thread_keys_[i][shard_id].insert(cur_key);
        }
      }
    };
    for (int i = 0; i < thread_keys_thread_num_; i++) {
      threads.push_back(
          std::thread(gen_func, std::ref(vec_data), begin,
                      begin + len_per_thread + (i < remain ? 1 : 0), i));
      begin += len_per_thread + (i < remain ? 1 : 0);
    }
    for (std::thread& t : threads) {
      t.join();
    }
    timeline.Pause();
T
Thunderbrook 已提交
231
    VLOG(0) << "GpuPs build task cost " << timeline.ElapsedSec() << " seconds.";
Y
yaoxuefeng 已提交
232 233 234 235
  }

  timeline.Start();

236
  threads.clear();
Y
yaoxuefeng 已提交
237
  // merge thread_keys to shard_keys
238 239 240 241
  auto merge_ins_func = [this, gpu_task](int shard_num) {
    for (int i = 0; i < thread_keys_thread_num_; ++i) {
      gpu_task->batch_add_keys(shard_num, thread_keys_[i][shard_num]);
      thread_keys_[i][shard_num].clear();
Y
yaoxuefeng 已提交
242
    }
243
  };
244 245 246 247 248 249 250
  auto merge_ins_dynamic_mf_func = [this, gpu_task](int shard_num, int dim_id) {
    for (int i = 0; i < thread_keys_thread_num_; ++i) {
      gpu_task->batch_add_keys(shard_num, dim_id,
                               thread_dim_keys_[i][shard_num][dim_id]);
      thread_dim_keys_[i][shard_num][dim_id].clear();
    }
  };
251
  for (int i = 0; i < thread_keys_shard_num_; ++i) {
252 253 254 255 256 257 258
    if (!multi_mf_dim_) {
      threads.push_back(std::thread(merge_ins_func, i));
    } else {
      for (int j = 0; j < multi_mf_dim_; j++) {
        threads.push_back(std::thread(merge_ins_dynamic_mf_func, i, j));
      }
    }
259 260 261
  }
  for (auto& t : threads) {
    t.join();
Y
yaoxuefeng 已提交
262 263 264
  }
  timeline.Pause();

265
  VLOG(0) << "GpuPs task add keys cost " << timeline.ElapsedSec()
Y
yaoxuefeng 已提交
266 267 268 269 270
          << " seconds.";
  timeline.Start();
  gpu_task->UniqueKeys();
  timeline.Pause();

271
  VLOG(0) << "GpuPs task unique cost " << timeline.ElapsedSec() << " seconds.";
Y
yaoxuefeng 已提交
272 273 274 275
  for (int i = 0; i < thread_keys_shard_num_; i++) {
    for (int j = 0; j < multi_mf_dim_; j++) {
      if (i == 0 && j == multi_mf_dim_ - 1) {
        gpu_task->feature_dim_keys_[i][j].push_back(0);
276
      }
Y
yaoxuefeng 已提交
277 278 279 280
      VLOG(0) << "GpuPs shard: " << i << "mf dim: " << index_dim_vec_[j]
              << " key len: " << gpu_task->feature_dim_keys_[i][j].size();
      gpu_task->value_dim_ptr_[i][j].resize(
          gpu_task->feature_dim_keys_[i][j].size());
281
    }
Y
yaoxuefeng 已提交
282
  }
283 284 285 286
}

void PSGPUWrapper::BuildPull(std::shared_ptr<HeterContext> gpu_task) {
  platform::Timer timeline;
T
Thunderbrook 已提交
287
  std::vector<std::future<void>> task_futures;
288 289 290 291
  int device_num = heter_devices_.size();
  auto& local_keys = gpu_task->feature_keys_;
  auto& local_ptr = gpu_task->value_ptr_;

292 293 294
  auto& local_dim_keys = gpu_task->feature_dim_keys_;
  auto& local_dim_ptr = gpu_task->value_dim_ptr_;

295 296
  auto& device_keys = gpu_task->device_keys_;
  auto& device_vals = gpu_task->device_values_;
297 298 299 300 301 302 303 304 305
  auto& device_dim_keys = gpu_task->device_dim_keys_;
  auto& device_dim_ptr = gpu_task->device_dim_ptr_;
  auto& device_dim_mutex = gpu_task->dim_mutex_;
  if (multi_mf_dim_) {
    for (size_t dev = 0; dev < device_dim_keys.size(); dev++) {
      device_dim_keys[dev].resize(multi_mf_dim_);
      device_dim_ptr[dev].resize(multi_mf_dim_);
    }
  }
T
Thunderbrook 已提交
306
  // auto& device_mutex = gpu_task->mutex_;
307 308 309 310 311 312

  std::vector<std::thread> threads(thread_keys_shard_num_);
#ifdef PADDLE_WITH_PSLIB
  auto fleet_ptr = FleetWrapper::GetInstance();
#endif
#ifdef PADDLE_WITH_PSCORE
313
  auto fleet_ptr = paddle::distributed::FleetWrapper::GetInstance();
314
#endif
315

316
#if (defined PADDLE_WITH_PSLIB) && (defined PADDLE_WITH_HETERPS)
317 318 319 320 321 322 323 324 325 326 327
  // get day_id: day nums from 1970
  struct std::tm b;
  b.tm_year = year_ - 1900;
  b.tm_mon = month_ - 1;
  b.tm_mday = day_;
  b.tm_min = b.tm_hour = b.tm_sec = 0;
  std::time_t seconds_from_1970 = std::mktime(&b);
  int day_id = seconds_from_1970 / 86400;
  fleet_ptr->pslib_ptr_->_worker_ptr->set_day_id(table_id_, day_id);
#endif

328
  timeline.Start();
329 330 331 332 333 334 335 336 337

  auto ptl_dynamic_mf_func = [this, &local_dim_keys, &local_dim_ptr,
                              &fleet_ptr](int i, int j) {
#ifdef PADDLE_WITH_PSLIB
    size_t key_size = local_dim_keys[i][j].size();
    int32_t status = -1;
    int32_t cnt = 0;
    while (true) {
      auto tt = fleet_ptr->pslib_ptr_->_worker_ptr->pull_sparse_ptr(
T
Thunderbrook 已提交
338 339
          i, reinterpret_cast<char**>(local_dim_ptr[i][j].data()),
          this->table_id_, local_dim_keys[i][j].data(), key_size);
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
      bool flag = true;

      tt.wait();

      try {
        status = tt.get();
      } catch (const std::future_error& e) {
        VLOG(0) << "Caught a future_error with code" << e.code()
                << ", Message:" << e.what();
      }
      if (status != 0) {
        VLOG(0) << "fleet pull sparse failed, status[" << status << "]";
        sleep(sleep_seconds_before_fail_exit_);
        flag = false;
        cnt++;
      }
      if (cnt > 3) {
        VLOG(0) << "fleet pull sparse failed, retry 3 times";
        exit(-1);
      }

      if (flag) {
        break;
      }
    }
    if (status != 0) {
      LOG(ERROR) << "fleet pull sparse failed, status[" << status << "]";
      sleep(300);
      exit(-1);
    } else {
      VLOG(0) << "FleetWrapper Pull sparse to local done with table size: "
              << local_dim_keys[i][j].size();
    }
#endif
  };
Y
yaoxuefeng 已提交
375 376 377 378 379 380

  threads.resize(thread_keys_shard_num_ * multi_mf_dim_);
  for (int i = 0; i < thread_keys_shard_num_; i++) {
    for (int j = 0; j < multi_mf_dim_; j++) {
      task_futures.emplace_back(
          pull_thread_pool_[i]->enqueue(ptl_dynamic_mf_func, i, j));
381
    }
382
  }
Y
yaoxuefeng 已提交
383 384
  for (auto& f : task_futures) {
    f.wait();
385
  }
Y
yaoxuefeng 已提交
386
  task_futures.clear();
387
  timeline.Pause();
T
Thunderbrook 已提交
388
  VLOG(0) << "pull sparse from CpuPS into GpuPS cost " << timeline.ElapsedSec()
389
          << " seconds.";
Y
yaoxuefeng 已提交
390 391 392 393 394 395 396 397
  if (multi_node_) {
    auto gloo_wrapper = paddle::framework::GlooWrapper::GetInstance();
    if (!gloo_wrapper->IsInitialized()) {
      VLOG(0) << "GLOO is not inited";
      gloo_wrapper->Init();
    }
    gloo_wrapper->Barrier();
  }
398 399

  timeline.Start();
Y
yaoxuefeng 已提交
400 401 402
  std::vector<std::vector<std::pair<uint64_t, char*>>> pass_values;

  bool record_status = false;
T
Thunderbrook 已提交
403 404
  auto& device_task_keys = gpu_task->device_task_keys_;
  auto& device_task_ptrs = gpu_task->device_task_ptr_;
Y
yaoxuefeng 已提交
405 406 407 408
  auto build_pull_dynamic_mf_func = [this, device_num, &local_dim_keys,
                                     &local_dim_ptr, &device_dim_keys,
                                     &device_dim_ptr,
                                     &device_dim_mutex](int i, int j) {
409 410 411 412 413 414 415 416 417 418
#ifdef PADDLE_WITH_PSLIB
    std::vector<std::vector<FeatureKey>> task_keys(device_num);
    std::vector<std::vector<paddle::ps::DownpourFixedFeatureValue*>> task_ptrs(
        device_num);
    for (size_t k = 0; k < local_dim_keys[i][j].size(); k++) {
      int shard = local_dim_keys[i][j][k] % device_num;
      task_keys[shard].push_back(local_dim_keys[i][j][k]);
      task_ptrs[shard].push_back(local_dim_ptr[i][j][k]);
    }
    for (int dev = 0; dev < device_num; dev++) {
Y
yaoxuefeng 已提交
419 420 421 422 423 424 425 426
      device_dim_mutex[dev][j]->lock();
      int len = task_keys[dev].size();
      int cur = device_dim_keys[dev][j].size();
      device_dim_keys[dev][j].resize(device_dim_keys[dev][j].size() + len);
      device_dim_ptr[dev][j].resize(device_dim_ptr[dev][j].size() + len);
      for (int k = 0; k < len; ++k) {
        device_dim_keys[dev][j][cur + k] = task_keys[dev][k];
        device_dim_ptr[dev][j][cur + k] = task_ptrs[dev][k];
427
      }
Y
yaoxuefeng 已提交
428
      device_dim_mutex[dev][j]->unlock();
429 430 431
    }
#endif
  };
Y
yaoxuefeng 已提交
432
  auto build_func = [device_num, record_status, &pass_values, &local_keys,
T
Thunderbrook 已提交
433 434
                     &local_ptr, &device_task_keys, &device_task_ptrs](int i) {
    auto& task_keys = device_task_keys[i];
T
Thunderbrook 已提交
435
#ifdef PADDLE_WITH_PSLIB
T
Thunderbrook 已提交
436
    auto& task_ptrs = device_task_ptrs[i];
T
Thunderbrook 已提交
437 438 439
#endif

#ifdef PADDLE_WITH_PSCORE
T
Thunderbrook 已提交
440
    auto& task_ptrs = device_task_ptrs[i];
T
Thunderbrook 已提交
441
#endif
442 443 444 445 446 447

    for (size_t j = 0; j < local_keys[i].size(); j++) {
      int shard = local_keys[i][j] % device_num;
      task_keys[shard].push_back(local_keys[i][j]);
      task_ptrs[shard].push_back(local_ptr[i][j]);
    }
448
#ifdef PADDLE_WITH_PSLIB
Y
yaoxuefeng 已提交
449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
    if (record_status) {
      size_t local_keys_size = local_keys.size();
      size_t pass_values_size = pass_values.size();
      for (size_t j = 0; j < pass_values_size; j += local_keys_size) {
        auto& shard_values = pass_values[j];
        for (size_t pair_idx = 0; pair_idx < pass_values[j].size();
             pair_idx++) {
          auto& cur_pair = shard_values[pair_idx];
          int shard = cur_pair.first % device_num;
          task_keys[shard].push_back(cur_pair.first);
          task_ptrs[shard].push_back(
              (paddle::ps::DownpourFixedFeatureValue*)cur_pair.second);
        }
      }
    }
464
#endif
T
Thunderbrook 已提交
465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
  };
  if (!multi_mf_dim_) {
    for (int i = 0; i < thread_keys_shard_num_; i++) {
      task_futures.emplace_back(hbm_thread_pool_[i]->enqueue(build_func, i));
    }
    for (auto& f : task_futures) {
      f.wait();
    }
    task_futures.clear();
    VLOG(0) << "GpuPs build hbmps done";
  }
  std::vector<std::vector<int>> prefix_sum;
  prefix_sum.resize(device_num);
  for (int i = 0; i < device_num; i++) {
    prefix_sum[i].resize(thread_keys_shard_num_ + 1);
    prefix_sum[i][0] = 0;
  }
  auto calc_prefix_func = [this, &prefix_sum, &device_keys, &device_vals,
                           &device_task_keys](int device_num) {
    for (int j = 0; j < thread_keys_shard_num_; j++) {
      prefix_sum[device_num][j + 1] =
          prefix_sum[device_num][j] + device_task_keys[j][device_num].size();
    }
    device_keys[device_num].resize(
        prefix_sum[device_num][thread_keys_shard_num_]);
    device_vals[device_num].resize(
        prefix_sum[device_num][thread_keys_shard_num_]);
  };
  if (!multi_mf_dim_) {
    for (int i = 0; i < device_num; i++) {
      task_futures.emplace_back(
          hbm_thread_pool_[i]->enqueue(calc_prefix_func, i));
    }
    for (auto& f : task_futures) {
      f.wait();
    }
    task_futures.clear();
  }
  VLOG(0) << "prefix done";
  auto prepare_dev_value_func = [device_num, &prefix_sum, &device_keys,
                                 &device_vals, &device_task_keys,
                                 &device_task_ptrs](int dev, int shard_id) {
    auto& task_keys = device_task_keys[shard_id];
#ifdef PADDLE_WITH_PSLIB
    auto& task_ptrs = device_task_ptrs[shard_id];
#endif

#ifdef PADDLE_WITH_PSCORE
513
    auto& task_ptrs = device_task_ptrs[shard_id];
T
Thunderbrook 已提交
514
#endif
515

T
Thunderbrook 已提交
516 517
    int len = prefix_sum[dev][shard_id + 1] - prefix_sum[dev][shard_id];
    int cur = prefix_sum[dev][shard_id];
T
Thunderbrook 已提交
518
#ifdef PADDLE_WITH_PSLIB
T
Thunderbrook 已提交
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
    for (int j = 0; j < len; ++j) {
      device_keys[dev][cur + j] = task_keys[dev][j];
      float* ptr_val = task_ptrs[dev][j]->data();
      FeatureValue& val = device_vals[dev][cur + j];
      size_t dim = task_ptrs[dev][j]->size();

      val.delta_score = ptr_val[1];
      val.show = ptr_val[2];
      val.clk = ptr_val[3];
      val.slot = ptr_val[6];
      val.lr = ptr_val[4];
      val.lr_g2sum = ptr_val[5];
      val.cpu_ptr = (uint64_t)(task_ptrs[dev][j]);

      if (dim > 7) {
        val.mf_size = MF_DIM + 1;
        for (int x = 0; x < val.mf_size; x++) {
          val.mf[x] = ptr_val[x + 7];
        }
      } else {
        val.mf_size = 0;
        for (int x = 0; x < MF_DIM + 1; x++) {
          val.mf[x] = 0;
Y
yaoxuefeng 已提交
542 543
        }
      }
T
Thunderbrook 已提交
544
    }
T
Thunderbrook 已提交
545 546
#endif
#ifdef PADDLE_WITH_PSCORE
T
Thunderbrook 已提交
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568
    for (int j = 0; j < len; ++j) {
      device_keys[dev][cur + j] = task_keys[dev][j];
      float* ptr_val = task_ptrs[dev][j]->data();
      FeatureValue& val = device_vals[dev][cur + j];
      size_t dim = task_ptrs[dev][j]->size();
      val.delta_score = ptr_val[2];
      val.show = ptr_val[3];
      val.clk = ptr_val[4];
      val.slot = ptr_val[0];
      val.lr = ptr_val[5];
      val.lr_g2sum = ptr_val[6];
      val.cpu_ptr = (uint64_t)(task_ptrs[dev][j]);

      if (dim > 7) {
        val.mf_size = MF_DIM + 1;
        for (int x = 0; x < val.mf_size; x++) {
          val.mf[x] = ptr_val[x + 7];
        }
      } else {
        val.mf_size = 0;
        for (int x = 0; x < MF_DIM + 1; x++) {
          val.mf[x] = 0;
T
Thunderbrook 已提交
569 570
        }
      }
T
Thunderbrook 已提交
571
    }
T
Thunderbrook 已提交
572
#endif
T
Thunderbrook 已提交
573
    VLOG(3) << "GpuPs build hbmps done";
Y
yaoxuefeng 已提交
574
  };
575

T
Thunderbrook 已提交
576
  if (multi_mf_dim_) {
577 578 579
    for (int i = 0; i < thread_keys_shard_num_; i++) {
      for (int j = 0; j < multi_mf_dim_; j++) {
        threads[i * multi_mf_dim_ + j] =
Y
yaoxuefeng 已提交
580
            std::thread(build_pull_dynamic_mf_func, i, j);
581 582
      }
    }
T
Thunderbrook 已提交
583 584 585 586 587 588 589 590 591 592 593 594 595 596
    for (std::thread& t : threads) {
      t.join();
    }
  } else {
    for (int i = 0; i < thread_keys_shard_num_; i++) {
      for (int j = 0; j < device_num; j++) {
        task_futures.emplace_back(
            hbm_thread_pool_[i]->enqueue(prepare_dev_value_func, j, i));
      }
    }
    for (auto& f : task_futures) {
      f.wait();
    }
    task_futures.clear();
Y
yaoxuefeng 已提交
597 598
  }
  timeline.Pause();
T
Thunderbrook 已提交
599
  VLOG(0) << "GpuPs prepare for build hbm cost " << timeline.ElapsedSec()
600
          << " seconds.";
Y
yaoxuefeng 已提交
601 602
}

603
void PSGPUWrapper::BuildGPUTask(std::shared_ptr<HeterContext> gpu_task) {
604
  int device_num = heter_devices_.size();
Y
yaoxuefeng 已提交
605 606
  platform::Timer timeline;
  timeline.Start();
T
Thunderbrook 已提交
607

608
  std::vector<size_t> feature_keys_count(device_num);
T
Thunderbrook 已提交
609
  size_t size_max = 0;
Y
yaoxuefeng 已提交
610 611 612 613 614 615 616

  for (int i = 0; i < device_num; i++) {
    for (int j = 0; j < multi_mf_dim_; j++) {
      feature_keys_count[i] += gpu_task->device_dim_ptr_[i][j].size();
      VLOG(1) << i << " card with dynamic mf dim: " << index_dim_vec_[j]
              << " dim index: " << j << " contains feasign nums: "
              << gpu_task->device_dim_ptr_[i][j].size();
617
    }
Y
yaoxuefeng 已提交
618 619 620
    VLOG(1) << i << " card with dynamic mf contains feasign nums total: "
            << feature_keys_count[i];
    size_max = std::max(size_max, feature_keys_count[i]);
T
Thunderbrook 已提交
621 622
  }
  if (HeterPs_) {
623 624
    delete HeterPs_;
    HeterPs_ = nullptr;
T
Thunderbrook 已提交
625
  }
626
  if (size_max <= 0) {
627
    VLOG(0) << "Skip build gpu ps cause feasign nums = " << size_max;
628 629
    return;
  }
630
  std::vector<std::thread> threads(device_num);
T
Thunderbrook 已提交
631
  HeterPs_ = HeterPsBase::get_instance(size_max, resource_);
F
Fan Zhang 已提交
632
#ifdef PADDLE_WITH_CUDA
633
  HeterPs_->set_nccl_comm_and_size(inner_comms_, inter_comms_, node_size_);
F
Fan Zhang 已提交
634
#endif
Y
yaoxuefeng 已提交
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695
  auto build_dynamic_mf_func = [this, &gpu_task](int i, int j) {
    this->HeterPs_->set_multi_mf_dim(multi_mf_dim_, max_mf_dim_);
    int mf_dim = this->index_dim_vec_[j];
    VLOG(0) << "building table: " << i << "with mf dim: " << mf_dim;
    size_t feature_value_size =
        TYPEALIGN(8, sizeof(FeatureValue) + ((mf_dim + 1) * sizeof(float)));
    auto& device_dim_keys = gpu_task->device_dim_keys_[i][j];
    auto& device_dim_ptrs = gpu_task->device_dim_ptr_[i][j];
    size_t len = device_dim_keys.size();
    CHECK(len == device_dim_ptrs.size());
    this->mem_pools_[i * this->multi_mf_dim_ + j] =
        new MemoryPool(len, feature_value_size);
    auto& mem_pool = this->mem_pools_[i * this->multi_mf_dim_ + j];
    for (size_t k = 0; k < len; k++) {
      FeatureValue* val = (FeatureValue*)(mem_pool->mem_address(k));
      float* ptr_val = device_dim_ptrs[k]->data();
      size_t dim = device_dim_ptrs[k]->size();
#ifdef PADDLE_WITH_PSLIB
      val->delta_score =
          ptr_val[paddle::ps::DownpourCtrDymfAccessor::
                      DownpourCtrDymfFeatureValue::delta_score_index()];
      val->show = ptr_val[paddle::ps::DownpourCtrDymfAccessor::
                              DownpourCtrDymfFeatureValue::show_index()];
      val->clk = ptr_val[paddle::ps::DownpourCtrDymfAccessor::
                             DownpourCtrDymfFeatureValue::click_index()];
      val->slot = int(ptr_val[paddle::ps::DownpourCtrDymfAccessor::
                                  DownpourCtrDymfFeatureValue::slot_index()]);
      val->lr = ptr_val[paddle::ps::DownpourCtrDymfAccessor::
                            DownpourCtrDymfFeatureValue::embed_w_index()];
      val->lr_g2sum =
          ptr_val[paddle::ps::DownpourCtrDymfAccessor::
                      DownpourCtrDymfFeatureValue::embed_g2sum_index()];
      val->cpu_ptr = (uint64_t)(device_dim_ptrs[k]);
      ptr_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  mf_dim_index()] = float(mf_dim);
      val->mf_dim = mf_dim;
#endif
      if (dim > 8) {  // CpuPS alreay expand as mf_dim
        val->mf_size = mf_dim + 1;
        for (int x = 0; x < val->mf_dim + 1; x++) {
          val->mf[x] = ptr_val[x + 8];
        }
      } else {
        val->mf_size = 0;
        for (int x = 0; x < val->mf_dim + 1; x++) {
          val->mf[x] = 0;
        }
      }
    }
    platform::CUDADeviceGuard guard(resource_->dev_id(i));
    this->hbm_pools_[i * this->multi_mf_dim_ + j] = new HBMMemoryPool(mem_pool);
    auto& cur_pool = this->hbm_pools_[i * this->multi_mf_dim_ + j];
    this->HeterPs_->build_ps(i, device_dim_keys.data(), cur_pool->mem(), len,
                             feature_value_size, 500000, 2);
    if (device_dim_keys.size() > 0) {
      VLOG(0) << "show ptr table: " << i
              << " table kv size: " << device_dim_keys.size()
              << "dim: " << mf_dim << " len: " << len;
      this->HeterPs_->show_one_table(i);
    }
    delete mem_pool;
Y
yaoxuefeng 已提交
696
  };
Y
yaoxuefeng 已提交
697 698 699 700 701
  threads.resize(device_num * multi_mf_dim_);
  for (int i = 0; i < device_num; i++) {
    for (int j = 0; j < multi_mf_dim_; j++) {
      threads[i + j * device_num] = std::thread(build_dynamic_mf_func, i, j);
    }
Y
yaoxuefeng 已提交
702 703 704
  }
  for (std::thread& t : threads) {
    t.join();
T
Thunderbrook 已提交
705 706
  }
  timeline.Pause();
707
  VLOG(0) << "GpuPs build table total costs: " << timeline.ElapsedSec()
T
Thunderbrook 已提交
708
          << " s.";
709 710 711 712 713 714 715 716 717 718 719 720 721 722
}

void PSGPUWrapper::LoadIntoMemory(bool is_shuffle) {
  platform::Timer timer;
  VLOG(3) << "Begin LoadIntoMemory(), dataset[" << dataset_ << "]";
  timer.Start();
  dataset_->LoadIntoMemory();
  timer.Pause();
  VLOG(0) << "LoadIntoMemory cost: " << timer.ElapsedSec() << "s";

  // local shuffle
  if (is_shuffle) {
    dataset_->LocalShuffle();
  }
Y
yaoxuefeng 已提交
723
  InitSlotInfo();
724 725 726 727 728 729 730 731
  std::shared_ptr<HeterContext> gpu_task = gpu_task_pool_.Get();
  gpu_task->Reset();
  data_ready_channel_->Put(gpu_task);
  VLOG(3) << "End LoadIntoMemory(), dataset[" << dataset_ << "]";
}

void PSGPUWrapper::start_build_thread() {
  running_ = true;
732
  VLOG(3) << "start build CPU ps thread.";
733
  pre_build_threads_ = std::thread([this] { pre_build_thread(); });
734 735
}

736 737
void PSGPUWrapper::pre_build_thread() {
  // prebuild: process load_data
738 739 740 741 742
  while (running_) {
    std::shared_ptr<HeterContext> gpu_task = nullptr;
    if (!data_ready_channel_->Get(gpu_task)) {
      continue;
    }
743
    VLOG(3) << "thread PreBuildTask start.";
744 745 746
    platform::Timer timer;
    timer.Start();
    // build cpu ps data process
747
    PreBuildTask(gpu_task);
748
    timer.Pause();
749
    VLOG(0) << "thread PreBuildTask end, cost time: " << timer.ElapsedSec()
T
Thunderbrook 已提交
750
            << " s";
751 752 753 754 755
    buildcpu_ready_channel_->Put(gpu_task);
  }
  VLOG(3) << "build cpu thread end";
}

756 757 758 759 760 761 762 763 764 765
void PSGPUWrapper::build_task() {
  // build_task: build_pull + build_gputask
  std::shared_ptr<HeterContext> gpu_task = nullptr;
  // train end, gpu free
  if (!gpu_free_channel_->Get(gpu_task)) {
    return;
  }
  // ins and pre_build end
  if (!buildcpu_ready_channel_->Get(gpu_task)) {
    return;
766
  }
767

768
  VLOG(0) << "BuildPull start.";
769 770 771 772 773
  platform::Timer timer;
  timer.Start();
  BuildPull(gpu_task);
  BuildGPUTask(gpu_task);
  timer.Pause();
774
  VLOG(0) << "BuildPull + BuildGPUTask end, cost time: " << timer.ElapsedSec()
775 776 777
          << "s";

  current_task_ = gpu_task;
778 779 780 781 782 783 784 785 786
}

void PSGPUWrapper::BeginPass() {
  platform::Timer timer;
  timer.Start();
  if (current_task_) {
    PADDLE_THROW(
        platform::errors::Fatal("[BeginPass] current task is not ended."));
  }
787 788

  build_task();
789
  timer.Pause();
790 791 792 793 794 795

  if (current_task_ == nullptr) {
    PADDLE_THROW(platform::errors::Fatal(
        "[BeginPass] after build_task, current task is not null."));
  }

T
Thunderbrook 已提交
796
  VLOG(0) << "BeginPass end, cost time: " << timer.ElapsedSec() << "s";
797 798 799 800 801 802 803 804 805 806 807 808
}

void PSGPUWrapper::EndPass() {
  if (!current_task_) {
    PADDLE_THROW(
        platform::errors::Fatal("[EndPass] current task has been ended."));
  }
  platform::Timer timer;
  timer.Start();
  size_t keysize_max = 0;
  // in case of feasign_num = 0, skip dump_to_cpu
  for (size_t i = 0; i < heter_devices_.size(); i++) {
Y
yaoxuefeng 已提交
809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
    for (int j = 0; j < multi_mf_dim_; j++) {
      keysize_max =
          std::max(keysize_max, current_task_->device_dim_keys_[i][j].size());
    }
  }

  auto dump_pool_to_cpu_func = [this](int i, int j) {
    PADDLE_ENFORCE_GPU_SUCCESS(cudaSetDevice(this->resource_->dev_id(i)));
    auto& hbm_pool = this->hbm_pools_[i * this->multi_mf_dim_ + j];
    auto& device_keys = this->current_task_->device_dim_keys_[i][j];
    size_t len = device_keys.size();
    int mf_dim = this->index_dim_vec_[j];
    VLOG(0) << "dump pool to cpu table: " << i << "with mf dim: " << mf_dim;
    size_t feature_value_size =
        TYPEALIGN(8, sizeof(FeatureValue) + ((mf_dim + 1) * sizeof(float)));
    char* test_build_values = (char*)malloc(feature_value_size * len);
    cudaMemcpy(test_build_values, hbm_pool->mem(), feature_value_size * len,
               cudaMemcpyDeviceToHost);
    CHECK(len == hbm_pool->capacity());
#ifdef PADDLE_WITH_PSLIB
    uint64_t unuse_key = std::numeric_limits<uint64_t>::max();
    for (size_t i = 0; i < len; ++i) {
      if (device_keys[i] == unuse_key) {
        continue;
      }
      size_t offset = i * feature_value_size;
      FeatureValue* gpu_val = (FeatureValue*)(test_build_values + offset);
      auto* downpour_value =
          (paddle::ps::DownpourFixedFeatureValue*)(gpu_val->cpu_ptr);
      int downpour_value_size = downpour_value->size();
      if (gpu_val->mf_size > 0 && downpour_value_size == 8) {
        downpour_value->resize(gpu_val->mf_dim + 1 + downpour_value_size);
      }
      float* cpu_val = downpour_value->data();
      cpu_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  delta_score_index()] = gpu_val->delta_score;
      cpu_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  show_index()] = gpu_val->show;
      cpu_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  click_index()] = gpu_val->clk;
      cpu_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  embed_w_index()] = gpu_val->lr;
      cpu_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  embed_g2sum_index()] = gpu_val->lr_g2sum;
      cpu_val[paddle::ps::DownpourCtrDymfAccessor::DownpourCtrDymfFeatureValue::
                  slot_index()] = gpu_val->slot;
      if (gpu_val->mf_size > 0) {
        for (int x = 0; x < gpu_val->mf_dim + 1; x++) {
          cpu_val[x + 8] = gpu_val->mf[x];
        }
      }
    }
#endif
    free(test_build_values);
  };
  if (multi_mf_dim_) {
    VLOG(0) << "psgpu wrapper dump pool: multi_mf_dim_: " << multi_mf_dim_;
    size_t device_num = heter_devices_.size();
    std::vector<std::thread> threads(device_num * multi_mf_dim_);
    for (size_t i = 0; i < device_num; i++) {
      for (int j = 0; j < multi_mf_dim_; j++) {
        threads[i + j * device_num] = std::thread(dump_pool_to_cpu_func, i, j);
      }
    }
    for (std::thread& t : threads) {
      t.join();
    }
876 877 878 879
  }
  if (keysize_max != 0) {
    HeterPs_->end_pass();
  }
880

Y
yaoxuefeng 已提交
881 882 883
  for (size_t i = 0; i < hbm_pools_.size(); i++) {
    delete hbm_pools_[i];
  }
884
  gpu_task_pool_.Push(current_task_);
885 886 887
  current_task_ = nullptr;
  gpu_free_channel_->Put(current_task_);
  timer.Pause();
Y
yaoxuefeng 已提交
888
  VLOG(1) << "EndPass end, cost time: " << timer.ElapsedSec() << "s";
T
Thunderbrook 已提交
889 890 891 892 893 894 895 896 897 898 899 900 901
}

void PSGPUWrapper::PullSparse(const paddle::platform::Place& place,
                              const int table_id,
                              const std::vector<const uint64_t*>& keys,
                              const std::vector<float*>& values,
                              const std::vector<int64_t>& slot_lengths,
                              const int hidden_size) {
  platform::Timer all_timer;
  platform::Timer pull_gpups_timer;
  all_timer.Start();
  int64_t total_length =
      std::accumulate(slot_lengths.begin(), slot_lengths.end(), 0UL);
F
Fan Zhang 已提交
902
  VLOG(3) << "Begine Gpu/Xpu Ps PullSparse";
903
  auto buf = memory::Alloc(place, total_length * sizeof(FeatureValue));
T
Thunderbrook 已提交
904 905 906 907 908
  FeatureValue* total_values_gpu = reinterpret_cast<FeatureValue*>(buf->ptr());
  if (platform::is_cpu_place(place)) {
    PADDLE_THROW(platform::errors::Unimplemented(
        "Warning:: CPUPlace is not supported in GpuPs now."));
  } else if (platform::is_gpu_place(place)) {
F
Fan Zhang 已提交
909
#ifdef PADDLE_WITH_CUDA
T
Thunderbrook 已提交
910
    VLOG(3) << "Begin copy keys, key_num[" << total_length << "]";
911
    int device_id = place.GetDeviceId();
T
Thunderbrook 已提交
912 913 914 915 916 917 918 919 920 921
    int devid_2_index = HeterPs_->get_index_by_devid(device_id);
    LoDTensor& total_keys_tensor = keys_tensor[devid_2_index];
    uint64_t* total_keys = reinterpret_cast<uint64_t*>(
        total_keys_tensor.mutable_data<int64_t>({total_length, 1}, place));

    // construct slot_level lod info
    auto slot_lengths_lod = slot_lengths;
    for (size_t i = 1; i < slot_lengths_lod.size(); i++) {
      slot_lengths_lod[i] += slot_lengths_lod[i - 1];
    }
922
    auto buf_key = memory::Alloc(place, keys.size() * sizeof(uint64_t*));
T
Thunderbrook 已提交
923
    auto buf_length =
924
        memory::Alloc(place, slot_lengths.size() * sizeof(int64_t));
T
Thunderbrook 已提交
925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946
    uint64_t** gpu_keys = reinterpret_cast<uint64_t**>(buf_key->ptr());
    int64_t* gpu_len = reinterpret_cast<int64_t*>(buf_length->ptr());
    cudaMemcpy(gpu_keys, keys.data(), keys.size() * sizeof(uint64_t*),
               cudaMemcpyHostToDevice);
    cudaMemcpy(gpu_len, slot_lengths_lod.data(),
               slot_lengths.size() * sizeof(int64_t), cudaMemcpyHostToDevice);

    this->CopyKeys(place, gpu_keys, total_keys, gpu_len,
                   static_cast<int>(slot_lengths.size()),
                   static_cast<int>(total_length));
    VLOG(3) << "Begin call PullSparseGPU in GPUPS, dev: " << devid_2_index
            << " len: " << total_length;
    pull_gpups_timer.Start();
    HeterPs_->pull_sparse(devid_2_index, total_keys, total_values_gpu,
                          static_cast<int>(total_length));
    pull_gpups_timer.Pause();

    VLOG(3) << "Begin Copy result to tensor, total_length[" << total_length
            << "]";
    this->CopyForPull(place, gpu_keys, values, total_values_gpu, gpu_len,
                      static_cast<int>(slot_lengths.size()), hidden_size,
                      total_length);
Y
yaoxuefeng 已提交
947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
  } else {
    PADDLE_THROW(platform::errors::PreconditionNotMet(
        "GpuPs: PullSparse Only Support CUDAPlace Now."));
  }
  all_timer.Pause();
  VLOG(3) << "GpuPs PullSparse total costs: " << all_timer.ElapsedSec()
          << " s, of which GPUPS costs: " << pull_gpups_timer.ElapsedSec()
          << " s";
  VLOG(3) << "End PullSparse";
}

void PSGPUWrapper::PullSparse(const paddle::platform::Place& place,
                              const int table_id,
                              const std::vector<const uint64_t*>& keys,
                              const std::vector<float*>& values,
                              const std::vector<int64_t>& slot_lengths,
                              const std::vector<int>& slot_dim,
                              const int hidden_size) {
  VLOG(3) << "Begine Gpu Ps PullSparse";
  platform::Timer all_timer;
  platform::Timer pull_gpups_timer;
  all_timer.Start();
  size_t total_length =
      std::accumulate(slot_lengths.begin(), slot_lengths.end(), 0UL);
  size_t feature_value_size = 0;

  feature_value_size = TYPEALIGN(
      8, sizeof(FeatureValue) + sizeof(float) * (index_dim_vec_.back() + 1));
  VLOG(0) << "yxf pull sparse feature_value_size: " << feature_value_size;

#ifdef PADDLE_WITH_CUDA
  VLOG(3) << "Begine Gpu Ps PullSparse";
  auto buf = memory::Alloc(place, total_length * feature_value_size);
  FeatureValue* total_values_gpu = reinterpret_cast<FeatureValue*>(buf->ptr());
#endif
#ifdef PADDLE_WITH_XPU_KP
  VLOG(3) << "Begine Xpu Ps PullSparse";
  FeatureValue* total_values_gpu = nullptr;
  xpu_malloc(reinterpret_cast<void**>(&total_values_gpu),
             total_length * feature_value_size);
#endif
  if (platform::is_cpu_place(place)) {
    PADDLE_THROW(platform::errors::Unimplemented(
        "Warning:: CPUPlace is not supported in GpuPs now."));
  } else if (platform::is_gpu_place(place)) {
    VLOG(3) << "Begin copy keys, key_num[" << total_length << "]";
    int device_id = place.GetDeviceId();
    int devid_2_index = HeterPs_->get_index_by_devid(device_id);
    LoDTensor& total_keys_tensor = keys_tensor[devid_2_index];
    uint64_t* total_keys =
        reinterpret_cast<uint64_t*>(total_keys_tensor.mutable_data<int64_t>(
            {int64_t(total_length), 1}, place));

    // construct slot_level lod info
    auto slot_lengths_lod = slot_lengths;
    for (size_t i = 1; i < slot_lengths_lod.size(); i++) {
      slot_lengths_lod[i] += slot_lengths_lod[i - 1];
    }
    auto buf_key = memory::Alloc(place, keys.size() * sizeof(uint64_t*));
    auto buf_length =
        memory::Alloc(place, slot_lengths.size() * sizeof(int64_t));
    uint64_t** gpu_keys = reinterpret_cast<uint64_t**>(buf_key->ptr());
    int64_t* gpu_len = reinterpret_cast<int64_t*>(buf_length->ptr());
    cudaMemcpy(gpu_keys, keys.data(), keys.size() * sizeof(uint64_t*),
               cudaMemcpyHostToDevice);
    cudaMemcpy(gpu_len, slot_lengths_lod.data(),
               slot_lengths.size() * sizeof(int64_t), cudaMemcpyHostToDevice);

    auto buf_dim = memory::Alloc(place, slot_dim.size() * sizeof(int));
    int* gpu_dim = reinterpret_cast<int*>(buf_dim->ptr());
    cudaMemcpy(gpu_dim, slot_dim.data(), slot_dim.size() * sizeof(int),
               cudaMemcpyHostToDevice);

    this->CopyKeys(place, gpu_keys, total_keys, gpu_len,
                   static_cast<int>(slot_lengths.size()),
                   static_cast<int>(total_length));
    VLOG(3) << "Begin call PullSparseGPU in GPUPS, dev: " << devid_2_index
            << " len: " << total_length;

    pull_gpups_timer.Start();
    HeterPs_->pull_sparse(devid_2_index, total_keys, total_values_gpu,
                          total_length);

    VLOG(3) << "Begin Copy result to tensor, total_length[" << total_length
            << "]";

    this->CopyForPull(place, gpu_keys, values, total_values_gpu, gpu_len,
                      static_cast<int>(slot_lengths.size()), hidden_size,
                      total_length, gpu_dim);

    pull_gpups_timer.Pause();

F
Fan Zhang 已提交
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
#endif
  } else if (platform::is_xpu_place(place)) {
#ifdef PADDLE_WITH_XPU_KP
    VLOG(3) << "Begin copy keys, key_num[" << total_length << "]";
    int device_id = place.GetDeviceId();
    int devid_2_index = HeterPs_->get_index_by_devid(device_id);
    LoDTensor& total_keys_tensor = keys_tensor[devid_2_index];
    uint64_t* total_keys = reinterpret_cast<uint64_t*>(
        total_keys_tensor.mutable_data<int64_t>({total_length, 1}, place));

    // construct slot_level lod info
    auto slot_lengths_lod = slot_lengths;
    for (size_t i = 1; i < slot_lengths_lod.size(); i++) {
      slot_lengths_lod[i] += slot_lengths_lod[i - 1];
    }

F
Fan Zhang 已提交
1055 1056 1057 1058 1059
    auto buf_key = memory::Alloc(place, keys.size() * sizeof(uint64_t*));
    auto buf_length =
        memory::Alloc(place, slot_lengths.size() * sizeof(int64_t));
    uint64_t** xpu_keys = reinterpret_cast<uint64_t**>(buf_key->ptr());
    int64_t* xpu_len = reinterpret_cast<int64_t*>(buf_length->ptr());
F
Fan Zhang 已提交
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
    PADDLE_ENFORCE_XPU_SUCCESS(xpu_memcpy(xpu_keys, keys.data(),
                                          keys.size() * sizeof(uint64_t*),
                                          XPU_HOST_TO_DEVICE));
    PADDLE_ENFORCE_XPU_SUCCESS(xpu_memcpy(xpu_len, slot_lengths_lod.data(),
                                          slot_lengths.size() * sizeof(int64_t),
                                          XPU_HOST_TO_DEVICE));

    this->CopyKeys(place, xpu_keys, total_keys, xpu_len,
                   static_cast<int>(slot_lengths.size()),
                   static_cast<int>(total_length));
    VLOG(3) << "Begin call PullSparseGPU in GPUPS, dev: " << devid_2_index
            << " len: " << total_length;
    pull_gpups_timer.Start();
    HeterPs_->pull_sparse(devid_2_index, total_keys, total_values_gpu,
                          static_cast<int>(total_length));
    pull_gpups_timer.Pause();

    VLOG(3) << "Begin Copy result to tensor, total_length[" << total_length
            << "]";
    this->CopyForPull(place, xpu_keys, values, total_values_gpu, xpu_len,
                      static_cast<int>(slot_lengths.size()), hidden_size,
                      total_length);
#endif
T
Thunderbrook 已提交
1083 1084
  } else {
    PADDLE_THROW(platform::errors::PreconditionNotMet(
F
Fan Zhang 已提交
1085
        "GpuPs/XpuPs: PullSparse Only Support CUDAPlace or XPUPlace Now."));
T
Thunderbrook 已提交
1086 1087
  }
  all_timer.Pause();
1088
  VLOG(3) << "GpuPs PullSparse total costs: " << all_timer.ElapsedSec()
T
Thunderbrook 已提交
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
          << " s, of which GPUPS costs: " << pull_gpups_timer.ElapsedSec()
          << " s";
  VLOG(3) << "End PullSparse";
}

void PSGPUWrapper::PushSparseGrad(const paddle::platform::Place& place,
                                  const int table_id,
                                  const std::vector<const uint64_t*>& keys,
                                  const std::vector<const float*>& grad_values,
                                  const std::vector<int64_t>& slot_lengths,
                                  const int hidden_size, const int batch_size) {
  platform::Timer all_timer;
  platform::Timer push_gpups_timer;
  all_timer.Start();
  int64_t total_length =
      std::accumulate(slot_lengths.begin(), slot_lengths.end(), 0UL);
F
Fan Zhang 已提交
1105
  // #ifdef PADDLE_WITH_CUDA
F
Fan Zhang 已提交
1106
  VLOG(3) << "Begin GPUPS PushSparseGrad";
Y
yaoxuefeng 已提交
1107 1108 1109 1110
  size_t grad_value_size =
      TYPEALIGN(8, sizeof(FeaturePushValue) + (max_mf_dim_ * sizeof(float)));
  auto buf = memory::Alloc(place, total_length * grad_value_size);
  VLOG(3) << "Push Sparse Max mf dimention: " << max_mf_dim_;
T
Thunderbrook 已提交
1111 1112 1113 1114 1115 1116
  FeaturePushValue* total_grad_values_gpu =
      reinterpret_cast<FeaturePushValue*>(buf->ptr());
  if (platform::is_cpu_place(place)) {
    PADDLE_THROW(platform::errors::Unimplemented(
        "Warning:: CPUPlace is not supported in GPUPS now."));
  } else if (platform::is_gpu_place(place)) {
F
Fan Zhang 已提交
1117
#ifdef PADDLE_WITH_CUDA
1118
    int device_id = place.GetDeviceId();
T
Thunderbrook 已提交
1119 1120 1121 1122 1123
    int devid_2_index = HeterPs_->get_index_by_devid(device_id);
    LoDTensor& cached_total_keys_tensor = keys_tensor[devid_2_index];
    uint64_t* total_keys =
        reinterpret_cast<uint64_t*>(cached_total_keys_tensor.data<int64_t>());
    VLOG(3) << "Begin copy grad tensor to gpups struct";
Y
yaoxuefeng 已提交
1124 1125 1126 1127 1128 1129 1130
    if (!multi_mf_dim_) {
      this->CopyForPush(place, grad_values, total_grad_values_gpu, slot_lengths,
                        hidden_size, total_length, batch_size);
    } else {
      this->CopyForPush(place, grad_values, total_grad_values_gpu, slot_lengths,
                        total_length, batch_size, grad_value_size);
    }
T
Thunderbrook 已提交
1131 1132 1133 1134 1135 1136 1137

    VLOG(3) << "Begin call PushSparseGPU in GPUPS, dev: " << devid_2_index
            << " len: " << total_length;
    push_gpups_timer.Start();
    HeterPs_->push_sparse(devid_2_index, total_keys, total_grad_values_gpu,
                          static_cast<int>(total_length));
    push_gpups_timer.Pause();
F
Fan Zhang 已提交
1138
#endif
F
Fan Zhang 已提交
1139
  } else if (platform::is_xpu_place(place)) {
F
Fan Zhang 已提交
1140
#ifdef PADDLE_WITH_XPU_KP
F
Fan Zhang 已提交
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
    int device_id = place.GetDeviceId();
    int devid_2_index = HeterPs_->get_index_by_devid(device_id);
    LoDTensor& cached_total_keys_tensor = keys_tensor[devid_2_index];
    uint64_t* total_keys =
        reinterpret_cast<uint64_t*>(cached_total_keys_tensor.data<int64_t>());
    VLOG(3) << "Begin copy grad tensor to xpups struct";
    this->CopyForPush(place, grad_values, total_grad_values_gpu, slot_lengths,
                      hidden_size, total_length, batch_size);

    VLOG(3) << "Begin call PushSparseXPU in XPUPS, dev: " << devid_2_index
            << " len: " << total_length;
    push_gpups_timer.Start();
    HeterPs_->push_sparse(devid_2_index, total_keys, total_grad_values_gpu,
                          static_cast<int>(total_length));
    push_gpups_timer.Pause();
F
Fan Zhang 已提交
1156
#endif
T
Thunderbrook 已提交
1157 1158 1159 1160 1161
  } else {
    PADDLE_THROW(platform::errors::PreconditionNotMet(
        "GPUPS: PushSparseGrad Only Support CUDAPlace Now."));
  }
  all_timer.Pause();
1162
  VLOG(3) << "PushSparseGrad total cost: " << all_timer.ElapsedSec()
T
Thunderbrook 已提交
1163 1164 1165 1166 1167 1168 1169 1170
          << " s, of which GPUPS cost: " << push_gpups_timer.ElapsedSec()
          << " s";
  VLOG(3) << "End PushSparseGrad";
}

}  // end namespace framework
}  // end namespace paddle
#endif