CMakeLists.txt 29.7 KB
Newer Older
H
heqiaozhi 已提交
1 2
#windows treat symbolic file as a real file, which is different with unix
#We create a hidden file and compile it instead of origin source file.
D
dzhwinter 已提交
3 4
function(windows_symbolic TARGET)
  set(oneValueArgs "")
P
peizhilin 已提交
5
  set(multiValueArgs SRCS PATH)
6 7
  cmake_parse_arguments(windows_symbolic "${options}" "${oneValueArgs}"
                        "${multiValueArgs}" ${ARGN})
P
peizhilin 已提交
8
  set(final_path ${CMAKE_CURRENT_SOURCE_DIR}/${windows_symbolic_PATH})
D
dzhwinter 已提交
9
  foreach(src ${windows_symbolic_SRCS})
P
peizhilin 已提交
10
    get_filename_component(src ${src} NAME_WE)
11 12 13 14 15
    if(NOT EXISTS ${final_path}/${src}.cc OR NOT EXISTS ${final_path}/${src}.cu)
      message(
        FATAL
        " ${src}.cc and ${src}.cu must exsits, and ${src}.cu must be symbolic file."
      )
P
peizhilin 已提交
16
    endif()
D
dzhwinter 已提交
17

18 19 20 21
    file(
      GENERATE
      OUTPUT ${final_path}/.${src}.cu
      INPUT ${final_path}/${src}.cc)
P
peizhilin 已提交
22

23 24 25 26 27
    add_custom_command(
      OUTPUT ${final_path}/.${src}.cu
      COMMAND ${CMAKE_COMMAND} -E copy_if_different "${final_path}/${src}.cc"
              "${final_path}/.${src}.cu"
      COMMENT "create hidden file of ${src}.cu")
28
    add_custom_target(${TARGET} ALL DEPENDS ${final_path}/.${src}.cu)
D
dzhwinter 已提交
29 30 31
  endforeach()
endfunction()

X
start  
Xin Pan 已提交
32
add_subdirectory(ir)
D
dzhwinter 已提交
33
add_subdirectory(details)
D
dongdaxiang 已提交
34
add_subdirectory(fleet)
35
add_subdirectory(io)
36
add_subdirectory(new_executor)
37
if(WITH_CINN)
38 39
  add_subdirectory(paddle2cinn)
endif()
H
heqiaozhi 已提交
40
#ddim lib
41
proto_library(framework_proto SRCS framework.proto)
42
proto_library(pass_desc_proto SRCS pass_desc.proto DEPS framework_proto)
43

44
proto_library(op_def_proto SRCS op_def.proto DEPS framework_proto)
45 46 47
cc_library(
  op_def_api
  SRCS op_def_api.cc
R
Ruibiao Chen 已提交
48
  DEPS op_def_proto)
49 50 51 52

file(GLOB OP_DEF_FILES
     ${PADDLE_SOURCE_DIR}/paddle/fluid/operators/compat/*.pbtxt)
file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/op_def.pbtxt
53 54 55
     "namespace { \n"
     "const std::unordered_map<std::string, std::string> op_def_map =  { \n")
foreach(OP_DEF_FILE ${OP_DEF_FILES})
56 57 58 59
  file(READ ${OP_DEF_FILE} OP_DEF_CONTENT)
  get_filename_component(OP_NAME ${OP_DEF_FILE} NAME_WE)
  file(APPEND ${CMAKE_CURRENT_BINARY_DIR}/op_def.pbtxt
       "{\"${OP_NAME}\",R\"(${OP_DEF_CONTENT})\"},\n")
60
endforeach()
61
file(APPEND ${CMAKE_CURRENT_BINARY_DIR}/op_def.pbtxt "{\"\",\"\"}};\n}")
62

T
Thunderbrook 已提交
63
proto_library(heter_service_proto SRCS heter_service.proto)
D
dongdaxiang 已提交
64
proto_library(data_feed_proto SRCS data_feed.proto)
H
hutuxian 已提交
65
proto_library(trainer_desc_proto SRCS trainer_desc.proto DEPS framework_proto
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
              data_feed_proto)

cc_library(
  string_array
  SRCS string_array.cc
  DEPS utf8proc)

cc_library(
  data_type
  SRCS data_type.cc
  DEPS framework_proto ddim device_context)
cc_test(
  data_type_test
  SRCS data_type_test.cc
  DEPS data_type place tensor)
81
if(WITH_GPU)
82
  if(WIN32)
D
dzhwinter 已提交
83
    windows_symbolic(tensor_util SRCS tensor_util.cu)
84 85 86 87
    nv_library(
      tensor
      SRCS tensor.cc .tensor_util.cu
      DEPS place memory data_type device_context dense_tensor)
D
dzhwinter 已提交
88 89
    add_dependencies(tensor tensor_util)
  else()
90 91 92 93
    nv_library(
      tensor
      SRCS tensor.cc tensor_util.cu
      DEPS place memory data_type device_context profiler dense_tensor)
94
  endif()
95
elseif(WITH_ROCM)
96 97 98 99
  hip_library(
    tensor
    SRCS tensor.cc tensor_util.cu
    DEPS place memory data_type device_context profiler dense_tensor)
Y
Yang Yu 已提交
100
else()
101 102 103 104
  cc_library(
    tensor
    SRCS tensor.cc tensor_util.cc
    DEPS place memory data_type device_context profiler dense_tensor)
105
endif()
D
dzhwinter 已提交
106

107 108 109 110
cc_test(
  tensor_test
  SRCS tensor_test.cc
  DEPS tensor)
111
if(WITH_GPU)
112 113 114 115
  nv_test(
    tensor_util_test
    SRCS tensor_util_test.cc tensor_util_test.cu
    DEPS tensor dlpack_tensor)
116
elseif(WITH_ROCM)
117 118 119 120
  hip_test(
    tensor_util_test
    SRCS tensor_util_test.cc tensor_util_test.cu
    DEPS tensor dlpack_tensor)
Y
Yang Yu 已提交
121
else()
122 123 124 125
  cc_test(
    tensor_util_test
    SRCS tensor_util_test.cc
    DEPS tensor dlpack_tensor)
Y
Yang Yu 已提交
126
endif()
D
dzhwinter 已提交
127

128 129 130 131
cc_test(
  copy_same_tensor_test
  SRCS copy_same_tensor_test.cc
  DEPS tensor)
132

133 134 135 136 137 138 139 140
cc_test(
  eigen_test
  SRCS eigen_test.cc
  DEPS tensor)
cc_library(
  mixed_vector
  SRCS mixed_vector.cc
  DEPS device_context place memory)
L
liaogang 已提交
141

142
if(WITH_GPU)
143 144 145 146
  nv_test(
    mixed_vector_test
    SRCS mixed_vector_test.cc mixed_vector_test.cu
    DEPS mixed_vector place memory device_context tensor)
147
elseif(WITH_ROCM)
148 149 150 151
  hip_test(
    mixed_vector_test
    SRCS mixed_vector_test.cc mixed_vector_test.cu
    DEPS mixed_vector place memory device_context tensor)
152
else()
153 154 155 156
  cc_test(
    mixed_vector_test
    SRCS mixed_vector_test.cc
    DEPS mixed_vector place memory device_context tensor)
157
endif()
158 159 160 161
cc_library(
  lod_tensor
  SRCS lod_tensor.cc
  DEPS ddim mixed_vector place tensor framework_proto version)
D
dzhwinter 已提交
162

163 164 165 166
cc_test(
  lod_tensor_test
  SRCS lod_tensor_test.cc
  DEPS lod_utils lod_tensor memory)
167 168

if(WITH_GPU)
169 170 171 172
  nv_test(
    lod_tensor_gpu_test
    SRCS lod_tensor_test.cu
    DEPS lod_tensor)
173
elseif(WITH_ROCM)
174 175 176 177
  hip_test(
    lod_tensor_gpu_test
    SRCS lod_tensor_test.cu
    DEPS lod_tensor)
178
endif()
179

180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
cc_library(
  garbage_collector
  SRCS garbage_collector.cc
  DEPS device_context memory gflags glog)

cc_library(
  reader
  SRCS reader.cc
  DEPS lod_tensor ddim)
cc_test(
  reader_test
  SRCS reader_test.cc
  DEPS reader)

cc_library(
  threadpool
  SRCS threadpool.cc
  DEPS enforce)
cc_test(
  threadpool_test
  SRCS threadpool_test.cc
  DEPS threadpool)

cc_library(
  var_type_traits
  SRCS var_type_traits.cc
  DEPS lod_tensor selected_rows_utils framework_proto scope)
if(WITH_GPU)
S
sneaxiy 已提交
208
  target_link_libraries(var_type_traits dynload_cuda)
S
sneaxiy 已提交
209
endif()
210 211 212 213
cc_test(
  var_type_traits_test
  SRCS var_type_traits_test.cc
  DEPS var_type_traits)
S
sneaxiy 已提交
214

T
Thunderbrook 已提交
215
set(BRPC_DEPS "")
216
if(WITH_PSCORE)
217
  set(BRPC_DEPS brpc ssl crypto)
218 219
endif()
if(WITH_PSLIB)
220 221 222 223 224 225 226 227
  if(WITH_PSLIB_BRPC)
    set(BRPC_DEPS pslib_brpc)
  elseif(NOT WITH_HETERPS)
    set(BRPC_DEPS brpc ssl crypto)
  endif()
  if(WITH_ARM_BRPC)
    set(BRPC_DEPS arm_brpc)
  endif()
T
Thunderbrook 已提交
228 229
endif()

230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259
cc_library(
  scope
  SRCS scope.cc
  DEPS glog threadpool xxhash var_type_traits)
cc_library(
  device_worker
  SRCS device_worker.cc
  DEPS trainer_desc_proto lod_tensor scope ${BRPC_DEPS})
cc_test(
  device_worker_test
  SRCS device_worker_test.cc
  DEPS device_worker)

cc_library(
  scope_pool
  SRCS scope_pool.cc
  DEPS scope)
cc_test(
  scope_test
  SRCS scope_test.cc
  DEPS scope)
cc_test(
  variable_test
  SRCS variable_test.cc
  DEPS tensor var_type_traits)

cc_library(
  data_device_transform
  SRCS data_device_transform.cc
  DEPS tensor)
260
if(WITH_GPU)
261 262 263 264
  nv_test(
    data_device_transform_test
    SRCS data_device_transform_test.cu
    DEPS operator op_registry device_context math_function scope)
265
elseif(WITH_ROCM)
266 267 268 269
  hip_test(
    data_device_transform_test
    SRCS data_device_transform_test.cu
    DEPS operator op_registry device_context math_function scope)
270
endif()
271

272
if(WITH_GPU)
273 274 275 276 277 278 279 280 281
  if(WIN32)
    #windows treat symbolic file as a real file, which is different with unix
    #We create a hidden file and compile it instead of origin source file.
    windows_symbolic(hidden_file SRCS data_type_transform.cu)
    nv_library(
      data_type_transform
      SRCS .data_type_transform.cu
      DEPS tensor)
    add_dependencies(data_type_transform hidden_file)
D
dzhwinter 已提交
282
  else()
283 284 285 286
    nv_library(
      data_type_transform
      SRCS data_type_transform.cu
      DEPS tensor)
287
  endif()
288 289 290 291
  nv_test(
    data_type_transform_test
    SRCS data_type_transform_test.cc data_type_transform_test.cu
    DEPS data_type_transform)
292
elseif(WITH_ROCM)
293 294 295 296 297 298 299 300
  hip_library(
    data_type_transform
    SRCS data_type_transform.cu
    DEPS tensor)
  hip_test(
    data_type_transform_test
    SRCS data_type_transform_test.cc data_type_transform_test.cu
    DEPS data_type_transform)
301 302 303 304 305 306 307 308 309
elseif(WITH_XPU)
  cc_library(
    data_type_transform
    SRCS data_type_transform.cc
    DEPS tensor xpulib)
  cc_test(
    data_type_transform_test
    SRCS data_type_transform_test.cc
    DEPS data_type_transform)
310
else()
311 312 313 314 315 316 317 318
  cc_library(
    data_type_transform
    SRCS data_type_transform.cc
    DEPS tensor)
  cc_test(
    data_type_transform_test
    SRCS data_type_transform_test.cc
    DEPS data_type_transform)
319
endif()
320

321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343
cc_library(
  data_layout_transform
  SRCS data_layout_transform.cc
  DEPS tensor math_function)
cc_test(
  data_layout_transform_test
  SRCS data_layout_transform_test.cc
  DEPS data_layout_transform)

cc_library(
  data_transform
  SRCS data_transform.cc
  DEPS math_function
       tensor
       framework_proto
       selected_rows_utils
       data_device_transform
       data_type_transform
       data_layout_transform)

cc_library(
  attribute
  SRCS attribute.cc
R
Ruibiao Chen 已提交
344
  DEPS framework_proto enforce)
345 346 347 348 349 350 351 352 353 354 355 356
cc_test(
  attribute_test
  SRCS attribute_test.cc
  DEPS attribute framework_proto proto_desc)
cc_test(
  program_desc_test
  SRCS program_desc_test.cc
  DEPS proto_desc device_context)

cc_library(
  op_version_proto
  SRCS op_version_proto.cc
R
Ruibiao Chen 已提交
357
  DEPS framework_proto)
358 359 360 361

cc_library(
  op_version_registry
  SRCS op_version_registry.cc
R
Ruibiao Chen 已提交
362
  DEPS op_version_proto framework_proto)
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
cc_test(
  op_version_registry_test
  SRCS op_version_registry_test.cc
  DEPS op_version_registry)

cc_library(
  op_proto_maker
  SRCS op_proto_maker.cc
  DEPS framework_proto attribute glog)
cc_test(
  op_proto_maker_test
  SRCS op_proto_maker_test.cc
  DEPS op_proto_maker)
cc_library(
  no_need_buffer_vars_inference
  SRCS no_need_buffer_vars_inference.cc
  DEPS attribute device_context)
cc_library(
  op_info
  SRCS op_info.cc
  DEPS attribute framework_proto no_need_buffer_vars_inference)
cc_library(
  shape_inference
  SRCS shape_inference.cc
  DEPS ddim attribute device_context)

cc_test(
  no_need_buffer_vars_inference_test
  SRCS no_need_buffer_vars_inference_test.cc
  DEPS no_need_buffer_vars_inference layer)

cc_library(
  transfer_scope_cache
  SRCS transfer_scope_cache.cc
  DEPS scope framework_proto device_context)

cc_library(
  unused_var_check
  SRCS unused_var_check.cc
  DEPS glog no_need_buffer_vars_inference)

cc_library(
  op_kernel_type
  SRCS op_kernel_type.cc
  DEPS device_context place)

if(WITH_XPU)
  cc_library(
    phi_utils
    SRCS phi_utils.cc
    DEPS lod_tensor
         selected_rows_utils
         place
         phi
         var_type_traits
         phi_api_utils
         op_info
         xpu_op_list)
else()
  cc_library(
    phi_utils
    SRCS phi_utils.cc
    DEPS lod_tensor
         selected_rows_utils
         place
         phi
         var_type_traits
         phi_api_utils
         op_info)
endif()
433

434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
if(WITH_XPU)
  cc_library(
    operator
    SRCS operator.cc
    DEPS xpu_op_list
         op_info
         device_context
         tensor
         scope
         glog
         trainer_desc_proto
         data_feed_proto
         shape_inference
         data_transform
         lod_tensor
         profiler
         transfer_scope_cache
         op_kernel_type
         op_call_stack
         unused_var_check
         nan_inf_utils
         phi_utils
         kernel_factory
         infershape_utils
         op_utils)
else()
  cc_library(
    operator
    SRCS operator.cc
    DEPS op_info
         device_context
         tensor
         scope
         glog
         trainer_desc_proto
         data_feed_proto
         shape_inference
         data_transform
         lod_tensor
         profiler
         transfer_scope_cache
         op_kernel_type
         op_call_stack
         unused_var_check
         nan_inf_utils
         phi_utils
         kernel_factory
         infershape_utils
         op_utils)
endif()
D
dzhwinter 已提交
484

485 486 487 488 489 490 491 492
cc_test(
  operator_test
  SRCS operator_test.cc
  DEPS operator op_registry device_context)
cc_test(
  operator_exception_test
  SRCS operator_exception_test.cc
  DEPS operator op_registry device_context)
X
version  
Xin Pan 已提交
493 494

cc_library(version SRCS version.cc)
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
cc_test(
  version_test
  SRCS version_test.cc
  DEPS version)

cc_library(
  proto_desc
  SRCS var_desc.cc op_desc.cc block_desc.cc program_desc.cc process_mesh_desc.cc
  DEPS attribute shape_inference op_info operator glog version)

cc_library(
  op_registry
  SRCS op_registry.cc
  DEPS op_proto_maker op_info operator glog proto_desc)

cc_library(
  op_call_stack
  SRCS op_call_stack.cc
  DEPS op_proto_maker enforce)
cc_test(
  op_call_stack_test
  SRCS op_call_stack_test.cc
  DEPS op_call_stack)

cc_library(
520 521
  program_utils
  SRCS program_utils.cc
R
Ruibiao Chen 已提交
522
  DEPS proto_desc)
523
cc_test(
524 525 526
  program_utils_test
  SRCS program_utils_test.cc
  DEPS proto_desc program_utils)
527

528
if(WITH_GPU)
529 530 531 532
  nv_test(
    op_registry_test
    SRCS op_registry_test.cc
    DEPS op_registry)
533
elseif(WITH_ROCM)
534 535 536 537
  hip_test(
    op_registry_test
    SRCS op_registry_test.cc
    DEPS op_registry)
538
endif()
539

540 541 542
if(WITH_PYTHON)
  py_proto_compile(framework_py_proto SRCS framework.proto data_feed.proto)
  py_proto_compile(trainer_py_proto SRCS trainer_desc.proto data_feed.proto)
543 544
  py_proto_compile(distributed_strategy_py_proto SRCS
                   distributed_strategy.proto)
545
  py_proto_compile(pass_desc_py_proto SRCS pass_desc.proto)
546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571
  #Generate an empty \
  #__init__.py to make framework_py_proto as a valid python module.
  add_custom_target(
    fleet_proto_init ALL
    COMMAND ${CMAKE_COMMAND} -E make_directory
            ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
    COMMAND
      ${CMAKE_COMMAND} -E touch
      ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto/__init__.py)
  add_custom_target(framework_py_proto_init ALL COMMAND ${CMAKE_COMMAND} -E
                                                        touch __init__.py)
  add_dependencies(
    framework_py_proto
    framework_py_proto_init
    trainer_py_proto
    distributed_strategy_py_proto
    fleet_proto_init
    pass_desc_py_proto
    ps_py_proto
    ps_py_proto_init)
  if(NOT WIN32)
    add_custom_command(
      TARGET framework_py_proto
      POST_BUILD
      COMMAND ${CMAKE_COMMAND} -E make_directory
              ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto
P
peizhilin 已提交
572
      COMMAND cp *.py ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/
573 574
      COMMAND cp distributed_strategy_*.py
              ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
P
peizhilin 已提交
575 576
      COMMENT "Copy generated python proto into directory paddle/fluid/proto."
      WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
577 578 579 580 581 582 583 584 585 586
    add_custom_target(
      fleet_executor_proto_init ALL
      DEPENDS fleet_proto_init fleet_executor_desc_py_proto
      COMMAND
        cp
        ${PADDLE_BINARY_DIR}/paddle/fluid/distributed/fleet_executor/fleet_executor_*.py
        ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
      COMMENT
        "Copy generated python proto into directory paddle/distributed/fleet/proto."
    )
587
  else()
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603
    string(REPLACE "/" "\\" proto_dstpath
                   "${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/")
    string(
      REPLACE "/" "\\" fleet_proto_dstpath
              "${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto/")
    add_custom_command(
      TARGET framework_py_proto
      POST_BUILD
      COMMAND ${CMAKE_COMMAND} -E make_directory
              ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto
      COMMAND copy /Y *.py ${proto_dstpath}
      COMMAND copy /Y distributed_strategy_*.py ${fleet_proto_dstpath}
      COMMENT "Copy generated python proto into directory paddle/fluid/proto."
      COMMENT
        "Copy generated python proto into directory paddle/distributed/fleet/proto."
      WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
604
  endif()
605
endif()
606

607 608 609 610 611 612 613 614 615 616 617
if(WITH_PSCORE)
  add_custom_target(
    index_dataset_proto_init ALL
    DEPENDS fleet_proto_init index_dataset_py_proto
    COMMAND
      cp
      ${PADDLE_BINARY_DIR}/paddle/fluid/distributed/index_dataset/index_dataset_*.py
      ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
    COMMENT
      "Copy generated python proto into directory paddle/distributed/fleet/proto."
  )
618
endif()
1
123malin 已提交
619

620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
cc_library(
  lod_rank_table
  SRCS lod_rank_table.cc
  DEPS lod_tensor)

cc_library(
  feed_fetch_method
  SRCS feed_fetch_method.cc
  DEPS lod_tensor scope glog)
cc_library(
  variable_helper
  SRCS variable_helper.cc
  DEPS lod_tensor)

if(TENSORRT_FOUND)
  cc_library(
    naive_executor
    SRCS naive_executor.cc
    DEPS op_registry
         denormal
         device_context
         scope
         framework_proto
         glog
         lod_rank_table
         feed_fetch_method
         graph_to_program_pass
         variable_helper
         tensorrt_engine_op)
W
wenbin 已提交
649
else()
650 651 652 653 654 655 656 657 658 659 660 661 662
  cc_library(
    naive_executor
    SRCS naive_executor.cc
    DEPS op_registry
         denormal
         device_context
         scope
         framework_proto
         glog
         lod_rank_table
         feed_fetch_method
         graph_to_program_pass
         variable_helper)
663
endif()
664

665 666 667 668 669 670 671 672 673 674 675
cc_library(
  executor_gc_helper
  SRCS executor_gc_helper.cc
  DEPS scope
       proto_desc
       operator
       garbage_collector
       op_registry
       while_op_helper
       recurrent_op_helper
       conditional_block_op_helper)
676
if(WITH_DISTRIBUTE)
T
Thunderbrook 已提交
677
  if(WITH_PSLIB)
678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732
    cc_library(
      executor
      SRCS executor.cc
           multi_trainer.cc
           pipeline_trainer.cc
           dataset_factory.cc
           dist_multi_trainer.cc
           trainer_factory.cc
           trainer.cc
           data_feed_factory.cc
           heterxpu_trainer.cc
           data_feed.cc
           device_worker.cc
           hogwild_worker.cc
           hetercpu_worker.cc
           ps_gpu_worker.cc
           ps_gpu_trainer.cc
           downpour_worker.cc
           downpour_worker_opt.cc
           data_feed.cu
           pull_dense_worker.cc
           section_worker.cc
           device_worker_factory.cc
           data_set.cc
      DEPS op_registry
           device_context
           scope
           framework_proto
           trainer_desc_proto
           glog
           fs
           shell
           fleet_wrapper
           heter_wrapper
           ps_gpu_wrapper
           box_wrapper
           metrics
           lodtensor_printer
           lod_rank_table
           feed_fetch_method
           collective_helper
           ${GLOB_DISTRIBUTE_DEPS}
           graph_to_program_pass
           variable_helper
           data_feed_proto
           timer
           monitor
           heter_service_proto
           fleet_executor
           ${BRPC_DEP})
    set(DISTRIBUTE_COMPILE_FLAGS
        "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor -Wno-error=parentheses"
    )
    if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)
      set(DISTRIBUTE_COMPILE_FLAGS "${DISTRIBUTE_COMPILE_FLAGS} -faligned-new")
T
Thunderbrook 已提交
733
    endif()
734 735 736 737 738 739 740 741
    set_source_files_properties(
      executor.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      device_worker.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      hetercpu_worker.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      heterxpu_trainer.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
T
tangwei12 已提交
742
  elseif(WITH_PSCORE)
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802
    cc_library(
      executor
      SRCS executor.cc
           multi_trainer.cc
           pipeline_trainer.cc
           dataset_factory.cc
           dist_multi_trainer.cc
           trainer_factory.cc
           trainer.cc
           data_feed_factory.cc
           heterxpu_trainer.cc
           heter_pipeline_trainer.cc
           data_feed.cc
           device_worker.cc
           hogwild_worker.cc
           hetercpu_worker.cc
           downpour_worker.cc
           downpour_lite_worker.cc
           downpour_worker_opt.cc
           data_feed.cu
           pull_dense_worker.cc
           section_worker.cc
           heter_section_worker.cc
           device_worker_factory.cc
           data_set.cc
      DEPS op_registry
           device_context
           scope
           framework_proto
           data_feed_proto
           heter_service_proto
           trainer_desc_proto
           glog
           index_sampler
           index_wrapper
           sampler
           index_dataset_proto
           lod_rank_table
           fs
           shell
           fleet_wrapper
           heter_wrapper
           box_wrapper
           metrics
           lodtensor_printer
           feed_fetch_method
           graph_to_program_pass
           variable_helper
           timer
           monitor
           heter_service_proto
           fleet
           heter_server
           brpc
           fleet_executor)
    set(DISTRIBUTE_COMPILE_FLAGS
        "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor -Wno-error=parentheses"
    )
    if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)
      set(DISTRIBUTE_COMPILE_FLAGS "${DISTRIBUTE_COMPILE_FLAGS} -faligned-new")
803
    endif()
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
    set_source_files_properties(
      executor.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      device_worker.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      multi_trainer.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      hogwild_worker.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      downpour_lite_worker.cc PROPERTIES COMPILE_FLAGS
                                         ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      heter_section_worker.cc PROPERTIES COMPILE_FLAGS
                                         ${DISTRIBUTE_COMPILE_FLAGS})
    set_source_files_properties(
      heter_pipeline_trainer.cc PROPERTIES COMPILE_FLAGS
                                           ${DISTRIBUTE_COMPILE_FLAGS})
T
tangwei12 已提交
821
  else()
822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868
    cc_library(
      executor
      SRCS executor.cc
           multi_trainer.cc
           pipeline_trainer.cc
           dataset_factory.cc
           dist_multi_trainer.cc
           trainer_factory.cc
           trainer.cc
           data_feed_factory.cc
           heterxpu_trainer.cc
           data_feed.cc
           device_worker.cc
           hogwild_worker.cc
           hetercpu_worker.cc
           ps_gpu_worker.cc
           ps_gpu_trainer.cc
           downpour_worker.cc
           downpour_worker_opt.cc
           data_feed.cu
           pull_dense_worker.cc
           section_worker.cc
           device_worker_factory.cc
           data_set.cc
      DEPS op_registry
           device_context
           scope
           framework_proto
           data_feed_proto
           heter_service_proto
           trainer_desc_proto
           glog
           lod_rank_table
           fs
           shell
           fleet_wrapper
           heter_wrapper
           ps_gpu_wrapper
           box_wrapper
           metrics
           lodtensor_printer
           feed_fetch_method
           graph_to_program_pass
           variable_helper
           timer
           monitor
           fleet_executor)
T
Thunderbrook 已提交
869
  endif()
T
Thunderbrook 已提交
870
elseif(WITH_PSLIB)
871 872 873 874 875
  set(DISTRIBUTE_COMPILE_FLAGS
      "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor"
  )
  if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)
    set(DISTRIBUTE_COMPILE_FLAGS "${DISTRIBUTE_COMPILE_FLAGS} -faligned-new")
T
Thunderbrook 已提交
876
  endif()
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
  set_source_files_properties(
    executor.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
  set_source_files_properties(
    device_worker.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
  set_source_files_properties(
    hetercpu_worker.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
  set_source_files_properties(
    heterxpu_trainer.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
  cc_library(
    executor
    SRCS executor.cc
         multi_trainer.cc
         pipeline_trainer.cc
         dataset_factory.cc
         dist_multi_trainer.cc
         trainer_factory.cc
         trainer.cc
         data_feed_factory.cc
         heterxpu_trainer.cc
         data_feed.cc
         device_worker.cc
         hogwild_worker.cc
         hetercpu_worker.cc
         ps_gpu_worker.cc
         ps_gpu_trainer.cc
         downpour_worker.cc
         downpour_worker_opt.cc
         data_feed.cu
         pull_dense_worker.cc
         section_worker.cc
         device_worker_factory.cc
         data_set.cc
    DEPS op_registry
         device_context
         scope
         framework_proto
         data_feed_proto
         heter_service_proto
         trainer_desc_proto
         glog
         lod_rank_table
         fs
         shell
         fleet_wrapper
         heter_wrapper
         ps_gpu_wrapper
         box_wrapper
         lodtensor_printer
         feed_fetch_method
         graph_to_program_pass
         variable_helper
         timer
         monitor
         fleet_executor
         ${BRPC_DEP})
W
Wu Yi 已提交
932
else()
933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978
  cc_library(
    executor
    SRCS executor.cc
         multi_trainer.cc
         pipeline_trainer.cc
         dataset_factory.cc
         dist_multi_trainer.cc
         trainer_factory.cc
         trainer.cc
         data_feed_factory.cc
         heterxpu_trainer.cc
         data_feed.cc
         device_worker.cc
         hogwild_worker.cc
         hetercpu_worker.cc
         ps_gpu_worker.cc
         ps_gpu_trainer.cc
         downpour_worker.cc
         downpour_worker_opt.cc
         data_feed.cu
         pull_dense_worker.cc
         section_worker.cc
         device_worker_factory.cc
         data_set.cc
    DEPS op_registry
         device_context
         scope
         framework_proto
         data_feed_proto
         heter_service_proto
         trainer_desc_proto
         glog
         lod_rank_table
         fs
         shell
         fleet_wrapper
         heter_wrapper
         ps_gpu_wrapper
         box_wrapper
         lodtensor_printer
         feed_fetch_method
         graph_to_program_pass
         variable_helper
         timer
         monitor
         fleet_executor)
W
Wu Yi 已提交
979
endif()
B
baojun-nervana 已提交
980

981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001
target_link_libraries(executor while_op_helper executor_gc_helper
                      recurrent_op_helper conditional_block_op_helper)

cc_library(
  parallel_executor
  SRCS parallel_executor.cc
  DEPS threaded_ssa_graph_executor
       scope_buffered_ssa_graph_executor
       parallel_ssa_graph_executor
       async_ssa_graph_executor
       graph
       build_strategy
       bind_threaded_ssa_graph_executor
       collective_helper
       fast_threaded_ssa_graph_executor
       variable_helper)

cc_library(
  executor_cache
  SRCS executor_cache.cc
  DEPS parallel_executor)
T
Thunderbrook 已提交
1002
if(WITH_PSCORE)
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
  get_property(RPC_DEPS GLOBAL PROPERTY RPC_DEPS)
  cc_test(
    dist_multi_trainer_test
    SRCS dist_multi_trainer_test.cc
    DEPS conditional_block_op executor gloo_wrapper ${RPC_DEPS})
  cc_test(
    heter_pipeline_trainer_test
    SRCS heter_pipeline_trainer_test.cc
    DEPS conditional_block_op
         scale_op
         heter_listen_and_serv_op
         executor
         heter_server
         gloo_wrapper
         eigen_function
         ${RPC_DEPS})
T
Thunderbrook 已提交
1019
else()
1020 1021 1022 1023
  cc_test(
    dist_multi_trainer_test
    SRCS dist_multi_trainer_test.cc
    DEPS conditional_block_op executor gloo_wrapper)
T
Thunderbrook 已提交
1024
endif()
1025 1026 1027
cc_library(
  prune
  SRCS prune.cc
R
Ruibiao Chen 已提交
1028
  DEPS framework_proto)
1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
cc_test(
  prune_test
  SRCS prune_test.cc
  DEPS op_info prune recurrent_op device_context)
cc_test(
  var_type_inference_test
  SRCS var_type_inference_test.cc
  DEPS op_registry proto_desc)
cc_library(
  selected_rows_utils
  SRCS selected_rows_utils.cc
  DEPS selected_rows)
cc_test(
  selected_rows_utils_test
  SRCS selected_rows_utils_test.cc
  DEPS selected_rows_utils)

cc_test(
  op_kernel_type_test
  SRCS op_kernel_type_test.cc
  DEPS place device_context framework_proto op_kernel_type)
Y
Yang Yu 已提交
1050
cc_test(cow_ptr_tests SRCS details/cow_ptr_test.cc)
M
minqiyang 已提交
1051

1052
cc_test(tuple_test SRCS tuple_test.cc)
1053

Z
Zeng Jinle 已提交
1054 1055
cc_test(inlined_vector_test SRCS inlined_vector_test.cc)

1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102
cc_library(
  dlpack_tensor
  SRCS dlpack_tensor.cc
  DEPS tensor dlpack)
cc_test(
  dlpack_tensor_test
  SRCS dlpack_tensor_test.cc
  DEPS dlpack_tensor glog)

cc_library(
  op_compatible_info
  SRCS op_compatible_info.cc
  DEPS string_helper proto_desc)
cc_test(
  op_compatible_info_test
  SRCS op_compatible_info_test.cc
  DEPS op_compatible_info proto_desc string_helper glog)

cc_library(
  save_load_util
  SRCS save_load_util.cc
  DEPS tensor scope layer)
cc_test(
  save_load_util_test
  SRCS save_load_util_test.cc
  DEPS save_load_util tensor scope layer)
cc_library(
  generator
  SRCS generator.cc
  DEPS enforce place)

cc_library(
  infershape_utils
  SRCS infershape_utils.cc
  DEPS lod_tensor
       selected_rows_utils
       attribute
       place
       var_type_traits
       phi
       phi_api_utils
       op_info
       shape_inference)
cc_test(
  infershape_utils_test
  SRCS infershape_utils_test.cc
  DEPS infershape_utils infermeta_utils meta_tensor)
1103

Y
Yan Chunwei 已提交
1104 1105 1106
# Get the current working branch
execute_process(
  COMMAND git rev-parse --abbrev-ref HEAD
1107 1108 1109
  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
  OUTPUT_VARIABLE PADDLE_BRANCH
  OUTPUT_STRIP_TRAILING_WHITESPACE)
Y
Yan Chunwei 已提交
1110 1111 1112 1113

# Get the latest abbreviated commit hash of the working branch
execute_process(
  COMMAND git log -1 --format=%h
1114 1115 1116
  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
  OUTPUT_VARIABLE PADDLE_COMMIT
  OUTPUT_STRIP_TRAILING_WHITESPACE)
Y
Yan Chunwei 已提交
1117 1118 1119 1120 1121

message(STATUS "commit: ${PADDLE_COMMIT}")
message(STATUS "branch: ${PADDLE_BRANCH}")

configure_file(commit.h.in commit.h)
1122

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
cc_library(
  custom_operator
  SRCS custom_operator.cc
  DEPS tensor
       attribute
       framework_proto
       op_registry
       operator
       dynamic_loader
       string_helper
       phi_tensor
       op_meta_info
       phi_api)
1136

H
hong 已提交
1137 1138 1139
#cc_binary(test_executor SRCS test_executor.cc DEPS executor op_registry ${GLOB_OP_LIB} ${GLOB_OPERATOR_DEPS} )
#cc_binary(new_executor SRCS new_exec_test.cc DEPS operator op_registry executor ${GLOB_OP_LIB} ${GLOB_OPERATOR_DEPS} profiler)

1140 1141 1142 1143 1144 1145 1146 1147 1148
set(FLUID_FRAMEWORK_MODULES
    proto_desc
    memory
    lod_tensor
    executor
    data_feed_proto
    layer
    dynamic_loader
    custom_operator)
1149

1150
cc_library(paddle_framework DEPS ${FLUID_FRAMEWORK_MODULES})
1151

1152 1153
if(WITH_TESTING AND TEST selected_rows_utils_test)
  set_tests_properties(selected_rows_utils_test PROPERTIES TIMEOUT 120)
Y
YUNSHEN XIE 已提交
1154
endif()
Z
Zeng Jinle 已提交
1155 1156

cc_test(scope_guard_test SRCS scope_guard_test.cc)
1157 1158 1159 1160
cc_test(
  phi_utils_test
  SRCS phi_utils_test.cc
  DEPS phi_utils)
1161 1162

if(WITH_GPU OR WITH_ROCM)
1163 1164 1165 1166
  cc_library(
    fluid_convert_utils
    SRCS convert_utils.cc
    DEPS data_type place gpu_info)
1167
else()
1168 1169 1170 1171
  cc_library(
    fluid_convert_utils
    SRCS convert_utils.cc
    DEPS data_type place)
1172
endif()
1173 1174 1175 1176
cc_test(
  convert_utils_test
  SRCS convert_utils_test.cc
  DEPS fluid_convert_utils)