extension_utils.py 26.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
# Copyright (c) 2021 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.

import os
import re
import six
import sys
import glob
20
import logging
21 22 23 24 25 26 27 28 29 30
import collections
import textwrap
import warnings
import subprocess

from contextlib import contextmanager
from setuptools.command import bdist_egg

from .. import load_op_library
from ...fluid import core
31
from ...fluid.framework import OpProtoHolder
32 33
from ...sysconfig import get_include, get_lib

34 35 36 37 38 39
logging.basicConfig(
    format='%(asctime)s - %(levelname)s - %(message)s', level=logging.INFO)
logger = logging.getLogger("utils.cpp_extension")

OS_NAME = sys.platform
IS_WINDOWS = OS_NAME.startswith('win')
40 41 42

MSVC_COMPILE_FLAGS = [
    '/MT', '/wd4819', '/wd4251', '/wd4244', '/wd4267', '/wd4275', '/wd4018',
43 44
    '/wd4190', '/EHsc', '/w', '/DGOOGLE_GLOG_DLL_DECL',
    '/DBOOST_HAS_STATIC_ASSERT', '/DNDEBUG', '/DPADDLE_USE_DSO'
45 46
]

47
MSVC_LINK_FLAGS = ['/MACHINE:X64', 'paddle_custom_op.lib']
48

49
COMMON_NVCC_FLAGS = ['-DPADDLE_WITH_CUDA', '-DEIGEN_USE_GPU', '-O3']
50

51 52 53 54
GCC_MINI_VERSION = (5, 4, 0)
# Give warning if using wrong compiler
WRONG_COMPILER_WARNING = '''
                        *************************************
55
                        *  Compiler Compatibility WARNING   *
56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77
                        *************************************

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Found that your compiler ({user_compiler}) is not compatible with the compiler 
built Paddle for this platform, which is {paddle_compiler} on {platform}. Please
use {paddle_compiler} to compile your custom op. Or you may compile Paddle from
source using {user_compiler}, and then also use it compile your custom op.

See https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/2.0/install/compile/linux-compile.html
for help with compiling Paddle from source.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
'''
# Give warning if used compiler version is incompatible
ABI_INCOMPATIBILITY_WARNING = '''
                            **********************************
                            *    ABI Compatibility WARNING   *
                            **********************************

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

78
Found that your compiler ({user_compiler} == {version}) may be ABI-incompatible with pre-installed Paddle!
79 80 81 82 83 84 85
Please use compiler that is ABI-compatible with GCC >= 5.4 (Recommended 8.2).

See https://gcc.gnu.org/onlinedocs/libstdc++/manual/abi.html for ABI Compatibility
information

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
'''
86 87 88
USING_NEW_CUSTOM_OP_LOAD_METHOD = True


89 90
# NOTE(chenweihang): In order to be compatible with
# the two custom op define method, after removing
91 92 93 94 95 96 97 98 99
# old method, we can remove them together
def use_new_custom_op_load_method(*args):
    global USING_NEW_CUSTOM_OP_LOAD_METHOD
    if len(args) == 0:
        return USING_NEW_CUSTOM_OP_LOAD_METHOD
    else:
        assert len(args) == 1 and isinstance(args[0], bool)
        USING_NEW_CUSTOM_OP_LOAD_METHOD = args[0]

100 101 102 103 104 105 106 107 108 109 110 111 112

@contextmanager
def bootstrap_context():
    """
    Context to manage how to write `__bootstrap__` code in .egg
    """
    origin_write_stub = bdist_egg.write_stub
    bdist_egg.write_stub = custom_write_stub
    yield

    bdist_egg.write_stub = origin_write_stub


113 114 115 116 117 118 119 120
def load_op_meta_info_and_register_op(lib_filename):
    if USING_NEW_CUSTOM_OP_LOAD_METHOD:
        core.load_op_meta_info_and_register_op(lib_filename)
    else:
        core.load_op_library(lib_filename)
    return OpProtoHolder.instance().update_op_proto()


121 122 123 124 125 126 127 128
def custom_write_stub(resource, pyfile):
    """
    Customized write_stub function to allow us to inject generated python
    api codes into egg python file.
    """
    _stub_template = textwrap.dedent("""
        import os
        import sys
129
        import types
130 131
        import paddle
        
132
        def inject_ext_module(module_name, api_names):
133 134 135
            if module_name in sys.modules:
                return sys.modules[module_name]

136
            new_module = types.ModuleType(module_name)
137 138
            for api_name in api_names:
                setattr(new_module, api_name, eval(api_name))
139 140 141 142 143 144 145 146 147 148

            return new_module

        def __bootstrap__():
            cur_dir = os.path.dirname(os.path.abspath(__file__))
            so_path = os.path.join(cur_dir, "{resource}")

            assert os.path.exists(so_path)

            # load custom op shared library with abs path
149 150
            new_custom_ops = paddle.utils.cpp_extension.load_op_meta_info_and_register_op(so_path)
            m = inject_ext_module(__name__, new_custom_ops)
151 152 153 154 155 156 157 158
        
        __bootstrap__()

        {custom_api}
        """).lstrip()

    # Parse registerring op information
    _, op_info = CustomOpInfo.instance().last()
159
    so_path = op_info.so_path
160

161 162 163 164 165
    new_custom_ops = load_op_meta_info_and_register_op(so_path)
    assert len(
        new_custom_ops
    ) > 0, "Required at least one custom operators, but received len(custom_op) =  %d" % len(
        new_custom_ops)
166 167 168 169 170 171

    # NOTE: To avoid importing .so file instead of python file because they have same name,
    # we rename .so shared library to another name, see EasyInstallCommand.
    filename, ext = os.path.splitext(resource)
    resource = filename + "_pd_" + ext

172 173 174 175
    api_content = []
    for op_name in new_custom_ops:
        api_content.append(_custom_api_content(op_name))

176 177 178
    with open(pyfile, 'w') as f:
        f.write(
            _stub_template.format(
179
                resource=resource, custom_api='\n\n'.join(api_content)))
180 181


182
OpInfo = collections.namedtuple('OpInfo', ['so_name', 'so_path'])
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202


class CustomOpInfo:
    """
    A global Singleton map to record all compiled custom ops information.
    """

    @classmethod
    def instance(cls):
        if not hasattr(cls, '_instance'):
            cls._instance = cls()
        return cls._instance

    def __init__(self):
        assert not hasattr(
            self.__class__,
            '_instance'), 'Please use `instance()` to get CustomOpInfo object!'
        # NOTE(Aurelius84): Use OrderedDict to save more order information
        self.op_info_map = collections.OrderedDict()

203 204
    def add(self, op_name, so_name, so_path=None):
        self.op_info_map[op_name] = OpInfo(so_name, so_path)
205 206 207 208 209 210 211 212 213

    def last(self):
        """
        Return the lastest insert custom op info.
        """
        assert len(self.op_info_map) > 0
        return next(reversed(self.op_info_map.items()))


214
def prepare_unix_cudaflags(cflags):
215 216 217
    """
    Prepare all necessary compiled flags for nvcc compiling CUDA files.
    """
218 219 220 221 222 223 224 225
    cflags = COMMON_NVCC_FLAGS + [
        '-ccbin', 'cc', '-Xcompiler', '-fPIC', '-w', '--expt-relaxed-constexpr',
        '-DNVCC'
    ] + cflags + get_cuda_arch_flags(cflags)

    return cflags


226
def prepare_win_cudaflags(cflags):
227 228 229
    """
    Prepare all necessary compiled flags for nvcc compiling CUDA files.
    """
230
    cflags = COMMON_NVCC_FLAGS + ['-w'] + cflags + get_cuda_arch_flags(cflags)
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

    return cflags


def add_std_without_repeat(cflags, compiler_type, use_std14=False):
    """
    Append -std=c++11/14 in cflags if without specific it before.
    """
    cpp_flag_prefix = '/std:' if compiler_type == 'msvc' else '-std='
    if not any(cpp_flag_prefix in flag for flag in cflags):
        suffix = 'c++14' if use_std14 else 'c++11'
        cpp_flag = cpp_flag_prefix + suffix
        cflags.append(cpp_flag)


def get_cuda_arch_flags(cflags):
    """
    For an arch, say "6.1", the added compile flag will be
    ``-gencode=arch=compute_61,code=sm_61``.
    For an added "+PTX", an additional
    ``-gencode=arch=compute_xx,code=compute_xx`` is added.
    """
    # TODO(Aurelius84):
    return []


def normalize_extension_kwargs(kwargs, use_cuda=False):
258
    """
259 260 261 262 263 264
    Normalize include_dirs, library_dir and other attributes in kwargs.
    """
    assert isinstance(kwargs, dict)
    # append necessary include dir path of paddle
    include_dirs = kwargs.get('include_dirs', [])
    include_dirs.extend(find_paddle_includes(use_cuda))
265

266 267 268 269 270 271 272
    kwargs['include_dirs'] = include_dirs

    # append necessary lib path of paddle
    library_dirs = kwargs.get('library_dirs', [])
    library_dirs.extend(find_paddle_libraries(use_cuda))
    kwargs['library_dirs'] = library_dirs

273 274 275 276 277 278 279
    # append compile flags and check settings of compiler
    extra_compile_args = kwargs.get('extra_compile_args', [])
    if isinstance(extra_compile_args, dict):
        for compiler in ['cxx', 'nvcc']:
            if compiler not in extra_compile_args:
                extra_compile_args[compiler] = []

280 281 282 283 284 285
    if IS_WINDOWS:
        # TODO(zhouwei): may append compile flags in future
        pass
        # append link flags
        extra_link_args = kwargs.get('extra_link_args', [])
        extra_link_args.extend(MSVC_LINK_FLAGS)
286 287
        if use_cuda:
            extra_link_args.extend(['cudadevrt.lib', 'cudart_static.lib'])
288 289 290
        kwargs['extra_link_args'] = extra_link_args
    else:
        # append compile flags
291
        add_compile_flag(extra_compile_args, ['-g', '-w'])  # disable warnings
292

293 294
        # append link flags
        extra_link_args = kwargs.get('extra_link_args', [])
295 296 297 298
        if use_new_custom_op_load_method():
            extra_link_args.append('-lpaddle_custom_op')
        else:
            extra_link_args.append('-lpaddle_framework')
299 300
        if use_cuda:
            extra_link_args.append('-lcudart')
301

302
        kwargs['extra_link_args'] = extra_link_args
303

304 305 306 307
        # add runtime library dirs
        runtime_library_dirs = kwargs.get('runtime_library_dirs', [])
        runtime_library_dirs.extend(find_paddle_libraries(use_cuda))
        kwargs['runtime_library_dirs'] = runtime_library_dirs
308

309 310
    kwargs['extra_compile_args'] = extra_compile_args

311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331
    kwargs['language'] = 'c++'
    return kwargs


def find_cuda_home():
    """
    Use heuristic method to find cuda path
    """
    # step 1. find in $CUDA_HOME or $CUDA_PATH
    cuda_home = os.environ.get('CUDA_HOME') or os.environ.get('CUDA_PATH')

    # step 2.  find path by `which nvcc`
    if cuda_home is None:
        which_cmd = 'where' if IS_WINDOWS else 'which'
        try:
            with open(os.devnull, 'w') as devnull:
                nvcc_path = subprocess.check_output(
                    [which_cmd, 'nvcc'], stderr=devnull)
                if six.PY3:
                    nvcc_path = nvcc_path.decode()
                nvcc_path = nvcc_path.rstrip('\r\n')
332

333 334 335 336 337 338
                # for example: /usr/local/cuda/bin/nvcc
                cuda_home = os.path.dirname(os.path.dirname(nvcc_path))
        except:
            if IS_WINDOWS:
                # search from default NVIDIA GPU path
                candidate_paths = glob.glob(
339 340
                    'C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v*.*'
                )
341 342 343 344 345
                if len(candidate_paths) > 0:
                    cuda_home = candidate_paths[0]
            else:
                cuda_home = "/usr/local/cuda"
    # step 3. check whether path is valid
346 347
    if cuda_home and not os.path.exists(
            cuda_home) and core.is_compiled_with_cuda():
348 349 350 351 352 353 354 355
        cuda_home = None
        warnings.warn(
            "Not found CUDA runtime, please use `export CUDA_HOME= XXX` to specific it."
        )

    return cuda_home


356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
def find_cuda_includes():
    """
    Use heuristic method to find cuda include path
    """
    cuda_home = find_cuda_home()
    if cuda_home is None:
        raise ValueError(
            "Not found CUDA runtime, please use `export CUDA_HOME=XXX` to specific it."
        )

    return [os.path.join(cuda_home, 'include')]


def find_paddle_includes(use_cuda=False):
    """
    Return Paddle necessary include dir path.
    """
    # pythonXX/site-packages/paddle/include
    paddle_include_dir = get_include()
    third_party_dir = os.path.join(paddle_include_dir, 'third_party')
    include_dirs = [paddle_include_dir, third_party_dir]

378 379 380
    if use_cuda:
        cuda_include_dir = find_cuda_includes()
        include_dirs.extend(cuda_include_dir)
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401

    return include_dirs


def find_cuda_libraries():
    """
    Use heuristic method to find cuda static lib path
    """
    cuda_home = find_cuda_home()
    if cuda_home is None:
        raise ValueError(
            "Not found CUDA runtime, please use `export CUDA_HOME=XXX` to specific it."
        )
    if IS_WINDOWS:
        cuda_lib_dir = [os.path.join(cuda_home, 'lib', 'x64')]
    else:
        cuda_lib_dir = [os.path.join(cuda_home, 'lib64')]

    return cuda_lib_dir


402 403 404 405 406 407
def find_paddle_libraries(use_cuda=False):
    """
    Return Paddle necessary library dir path.
    """
    # pythonXX/site-packages/paddle/libs
    paddle_lib_dirs = [get_lib()]
408

409 410 411
    if use_cuda:
        cuda_lib_dir = find_cuda_libraries()
        paddle_lib_dirs.extend(cuda_lib_dir)
412

413 414 415
    return paddle_lib_dirs


416 417
def add_compile_flag(extra_compile_args, flags):
    assert isinstance(flags, list)
418 419
    if isinstance(extra_compile_args, dict):
        for args in extra_compile_args.values():
420
            args.extend(flags)
421
    else:
422
        extra_compile_args.extend(flags)
423 424 425 426 427 428 429 430 431 432


def is_cuda_file(path):

    cuda_suffix = set(['.cu'])
    items = os.path.splitext(path)
    assert len(items) > 1
    return items[-1] in cuda_suffix


433
def get_build_directory(verbose=False):
434 435 436 437 438 439
    """
    Return paddle extension root directory, default specific by `PADDLE_EXTENSION_DIR`
    """
    root_extensions_directory = os.environ.get('PADDLE_EXTENSION_DIR')
    if root_extensions_directory is None:
        dir_name = "paddle_extensions"
440 441 442 443 444 445 446 447
        root_extensions_directory = os.path.join(
            os.path.expanduser('~/.cache'), dir_name)
        if IS_WINDOWS:
            root_extensions_directory = os.path.normpath(
                root_extensions_directory)
        elif OS_NAME.startswith('darwin'):
            # TODO(Aurelius84): consider macOs
            raise NotImplementedError("Not support Mac now.")
448

449 450
        log_v("$PADDLE_EXTENSION_DIR is not set, using path: {} by default.".
              format(root_extensions_directory), verbose)
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465

    if not os.path.exists(root_extensions_directory):
        os.makedirs(root_extensions_directory)

    return root_extensions_directory


def parse_op_info(op_name):
    """
    Parse input names and outpus detail information from registered custom op
    from OpInfoMap.
    """
    from paddle.fluid.framework import OpProtoHolder
    if op_name not in OpProtoHolder.instance().op_proto_map:
        raise ValueError(
466
            "Please load {} shared library file firstly by `paddle.utils.cpp_extension.load_op_meta_info_and_register_op(...)`".
467 468 469 470
            format(op_name))
    op_proto = OpProtoHolder.instance().get_op_proto(op_name)

    in_names = [x.name for x in op_proto.inputs]
471
    out_names = [x.name for x in op_proto.outputs]
472

473
    return in_names, out_names
474 475


476
def _import_module_from_library(module_name, build_directory, verbose=False):
477
    """
478
    Load shared library and import it as callable python module.
479
    """
480 481 482 483 484
    if IS_WINDOWS:
        dynamic_suffix = '.pyd'
    else:
        dynamic_suffix = '.so'
    ext_path = os.path.join(build_directory, module_name + dynamic_suffix)
485 486 487 488 489
    if not os.path.exists(ext_path):
        raise FileNotFoundError("Extension path: {} does not exist.".format(
            ext_path))

    # load custom op_info and kernels from .so shared library
490
    log_v('loading shared library from: {}'.format(ext_path), verbose)
491
    op_names = load_op_meta_info_and_register_op(ext_path)
492 493

    # generate Python api in ext_path
494 495
    return _generate_python_module(module_name, op_names, build_directory,
                                   verbose)
496 497


498 499 500 501
def _generate_python_module(module_name,
                            op_names,
                            build_directory,
                            verbose=False):
502 503 504
    """
    Automatically generate python file to allow import or load into as module
    """
505
    api_file = os.path.join(build_directory, module_name + '.py')
506
    log_v("generate api file: {}".format(api_file), verbose)
507 508

    # write into .py file
509
    api_content = [_custom_api_content(op_name) for op_name in op_names]
510
    with open(api_file, 'w') as f:
511
        f.write('\n\n'.join(api_content))
512 513

    # load module
514 515
    custom_module = _load_module_from_file(api_file, verbose)
    return custom_module
516 517 518


def _custom_api_content(op_name):
519
    params_str, ins_str, outs_str = _get_api_inputs_str(op_name)
520 521 522 523 524 525 526

    API_TEMPLATE = textwrap.dedent("""
        from paddle.fluid.layer_helper import LayerHelper

        def {op_name}({inputs}):
            helper = LayerHelper("{op_name}", **locals())

527
            # prepare inputs and outputs
528 529
            ins = {ins}
            outs = {{}}
530 531 532 533 534
            out_names = {out_names}
            for out_name in out_names:
                # Set 'float32' temporarily, and the actual dtype of output variable will be inferred
                # in runtime.
                outs[out_name] = helper.create_variable(dtype='float32')
535 536 537
            
            helper.append_op(type="{op_name}", inputs=ins, outputs=outs)

538 539 540
            res = [outs[out_name] for out_name in out_names]

            return res[0] if len(res)==1 else res
541 542 543 544
            """).lstrip()

    # generate python api file
    api_content = API_TEMPLATE.format(
545
        op_name=op_name, inputs=params_str, ins=ins_str, out_names=outs_str)
546 547 548 549

    return api_content


550
def _load_module_from_file(api_file_path, verbose=False):
551 552 553 554 555 556 557 558
    """
    Load module from python file.
    """
    if not os.path.exists(api_file_path):
        raise FileNotFoundError("File : {} does not exist.".format(
            api_file_path))

    # Unique readable module name to place custom api.
559
    log_v('import module from file: {}'.format(api_file_path), verbose)
560 561 562 563 564 565 566 567 568
    ext_name = "_paddle_cpp_extension_"
    if six.PY2:
        import imp
        module = imp.load_source(ext_name, api_file_path)
    else:
        from importlib import machinery
        loader = machinery.SourceFileLoader(ext_name, api_file_path)
        module = loader.load_module()

569
    return module
570 571 572 573 574 575


def _get_api_inputs_str(op_name):
    """
    Returns string of api parameters and inputs dict.
    """
576
    in_names, out_names = parse_op_info(op_name)
577 578 579 580 581
    # e.g: x, y, z
    params_str = ','.join([p.lower() for p in in_names])
    # e.g: {'X': x, 'Y': y, 'Z': z}
    ins_str = "{%s}" % ','.join(
        ["'{}' : {}".format(in_name, in_name.lower()) for in_name in in_names])
582 583 584
    # e.g: ['Out', 'Index']
    outs_str = "[%s]" % ','.join(["'{}'".format(name) for name in out_names])
    return params_str, ins_str, outs_str
585 586


587 588 589
def _write_setup_file(name,
                      sources,
                      file_path,
590
                      build_dir,
591
                      include_dirs,
592 593
                      extra_cxx_cflags,
                      extra_cuda_cflags,
594 595
                      link_args,
                      verbose=False):
596 597 598 599 600 601 602
    """
    Automatically generate setup.py and write it into build directory.
    """
    template = textwrap.dedent("""
    import os
    from paddle.utils.cpp_extension import CppExtension, CUDAExtension, BuildExtension, setup
    from paddle.utils.cpp_extension import get_build_directory
603 604 605 606
    from paddle.utils.cpp_extension.extension_utils import use_new_custom_op_load_method

    use_new_custom_op_load_method({use_new_method})

607 608 609 610 611 612
    setup(
        name='{name}',
        ext_modules=[
            {prefix}Extension(
                sources={sources},
                include_dirs={include_dirs},
613
                extra_compile_args={{'cxx':{extra_cxx_cflags}, 'nvcc':{extra_cuda_cflags}}},
614 615
                extra_link_args={extra_link_args})],
        cmdclass={{"build_ext" : BuildExtension.with_options(
616 617
            output_dir=r'{build_dir}',
            no_python_abi_suffix=True)
618 619 620 621 622
        }})""").lstrip()

    with_cuda = False
    if any([is_cuda_file(source) for source in sources]):
        with_cuda = True
623
    log_v("with_cuda: {}".format(with_cuda), verbose)
624 625 626 627 628 629

    content = template.format(
        name=name,
        prefix='CUDA' if with_cuda else 'Cpp',
        sources=list2str(sources),
        include_dirs=list2str(include_dirs),
630 631
        extra_cxx_cflags=list2str(extra_cxx_cflags),
        extra_cuda_cflags=list2str(extra_cuda_cflags),
632
        extra_link_args=list2str(link_args),
633
        build_dir=build_dir,
634
        use_new_method=use_new_custom_op_load_method())
635 636

    log_v('write setup.py into {}'.format(file_path), verbose)
637 638 639 640 641 642
    with open(file_path, 'w') as f:
        f.write(content)


def list2str(args):
    """
643
    Convert list[str] into string. For example: ['x', 'y'] -> "['x', 'y']"
644 645 646
    """
    if args is None: return '[]'
    assert isinstance(args, (list, tuple))
647 648
    args = ["{}".format(arg) for arg in args]
    return repr(args)
649 650


651
def _jit_compile(file_path, interpreter=None, verbose=False):
652 653 654 655 656
    """
    Build shared library in subprocess
    """
    ext_dir = os.path.dirname(file_path)
    setup_file = os.path.basename(file_path)
657 658 659 660

    if interpreter is None:
        interpreter = 'python'
    try:
661 662
        which = 'where' if IS_WINDOWS else 'which'
        py_path = subprocess.check_output([which, interpreter])
663 664 665 666 667 668 669 670 671 672 673 674
        py_version = subprocess.check_output([interpreter, '-V'])
        if six.PY3:
            py_path = py_path.decode()
            py_version = py_version.decode()
        log_v("Using Python interpreter: {}, version: {}".format(
            py_path.strip(), py_version.strip()), verbose)
    except Exception:
        _, error, _ = sys.exc_info()
        raise RuntimeError(
            'Failed to check Python interpreter with `{}`, errors: {}'.format(
                interpreter, error))

675 676 677 678 679 680 681
    if IS_WINDOWS:
        compile_cmd = 'cd /d {} && {} {} build'.format(ext_dir, interpreter,
                                                       setup_file)
    else:
        compile_cmd = 'cd {} && {} {} build'.format(ext_dir, interpreter,
                                                    setup_file)

682 683
    print("Compiling user custom op, it will cost a few seconds.....")
    run_cmd(compile_cmd, verbose)
684 685 686 687 688 689 690 691


def parse_op_name_from(sources):
    """
    Parse registerring custom op name from sources.
    """

    def regex(content):
692
        if USING_NEW_CUSTOM_OP_LOAD_METHOD:
693
            pattern = re.compile(r'PD_BUILD_OP\(([^,\)]+)\)')
694 695
        else:
            pattern = re.compile(r'REGISTER_OPERATOR\(([^,]+),')
696 697 698 699 700 701 702 703 704 705 706 707 708

        content = re.sub(r'\s|\t|\n', '', content)
        op_name = pattern.findall(content)
        op_name = set([re.sub('_grad', '', name) for name in op_name])

        return op_name

    op_names = set()
    for source in sources:
        with open(source, 'r') as f:
            content = f.read()
            op_names |= regex(content)

709
    return list(op_names)
710 711


712
def run_cmd(command, verbose=False):
713 714 715
    """
    Execute command with subprocess.
    """
716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765
    # logging
    log_v("execute command: {}".format(command), verbose)
    try:
        from subprocess import DEVNULL  # py3
    except ImportError:
        DEVNULL = open(os.devnull, 'wb')

    # execute command
    try:
        if verbose:
            return subprocess.check_call(
                command, shell=True, stderr=subprocess.STDOUT)
        else:
            return subprocess.check_call(command, shell=True, stdout=DEVNULL)
    except Exception:
        _, error, _ = sys.exc_info()
        raise RuntimeError("Failed to run command: {}, errors: {}".format(
            compile, error))


def check_abi_compatibility(compiler, verbose=False):
    """
    Check whether GCC version on user local machine is compatible with Paddle in
    site-packages.
    """
    # TODO(Aurelius84): After we support windows, remove IS_WINDOWS in following code.
    if os.environ.get('PADDLE_SKIP_CHECK_ABI') in ['True', 'true', '1'
                                                   ] or IS_WINDOWS:
        return True

    cmd_out = subprocess.check_output(
        ['which', compiler], stderr=subprocess.STDOUT)
    compiler_path = os.path.realpath(cmd_out.decode()
                                     if six.PY3 else cmd_out).strip()
    # step 1. if not found any suitable compiler, raise error
    if not any(name in compiler_path
               for name in _expected_compiler_current_platform()):
        warnings.warn(
            WRONG_COMPILER_WARNING.format(
                user_compiler=compiler,
                paddle_compiler=_expected_compiler_current_platform()[0],
                platform=OS_NAME))
        return False

    # clang++ have no ABI compatibility problem
    if OS_NAME.startswith('darwin'):
        return True
    try:
        if OS_NAME.startswith('linux'):
            version_info = subprocess.check_output(
766
                [compiler, '-dumpfullversion', '-dumpversion'])
767 768 769 770 771 772 773 774 775 776 777 778
            if six.PY3:
                version_info = version_info.decode()
            version = version_info.strip().split('.')
            assert len(version) == 3
            # check version compatibility
            if tuple(map(int, version)) >= GCC_MINI_VERSION:
                return True
            else:
                warnings.warn(
                    ABI_INCOMPATIBILITY_WARNING.format(
                        user_compiler=compiler, version=version_info.strip()))
        elif IS_WINDOWS:
779
            # TODO(zhouwei): support check abi compatibility on windows
780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797
            warnings.warn("We don't support Windows now.")
    except Exception:
        _, error, _ = sys.exc_info()
        warnings.warn('Failed to check compiler version for {}: {}'.format(
            compiler, error))

    return False


def _expected_compiler_current_platform():
    """
    Returns supported compiler string on current platform
    """
    expect_compilers = ['clang', 'clang++'] if OS_NAME.startswith(
        'darwin') else ['gcc', 'g++', 'gnu-c++', 'gnu-cc']
    return expect_compilers


798
def log_v(info, verbose=True):
799 800 801 802 803
    """
    Print log information on stdout.
    """
    if verbose:
        logging.info(info)