pipeline_process.py revision 5fe6dc81886f49eb2089c883c6862f2e896155cf 
1"""Pipeline process that encapsulates the actual content.
2
3The actual stages include the builder and the executor.
4"""
5
6__author__ = 'yuhenglong@google.com (Yuheng Long)'
7
8import multiprocessing
9
10# Pick an integer at random.
11POISONPILL = 975
12
13
14class PipelineProcess(multiprocessing.Process):
15  """A process that encapsulates the actual content pipeline stage.
16
17  The actual pipeline stage can be the builder or the tester.  This process
18  continuously pull tasks from the queue until a poison pill is received.
19  Once a job is received, it will hand it to the actual stage for processing.
20
21  Each pipeline stage contains three modules.
22  The first module continuously pulls task from the input queue. It searches the
23  cache to check whether the task has encountered before. If so, duplicate
24  computation can be avoided.
25  The second module consists of a pool of workers that do the actual work, e.g.,
26  the worker will compile the source code and get the image in the builder
27  pipeline stage.
28  The third module is a helper that put the result cost to the cost field of the
29  duplicate tasks. For example, if two tasks are equivalent, only one task, say
30  t1 will be executed and the other task, say t2 will not be executed. The third
31  mode gets the result from t1, when it is available and set the cost of t2 to
32  be the same as that of t1.
33  """
34
35  def __init__(self, num_processes, name, cache, stage, task_queue, helper,
36               worker, result_queue):
37    """Set up input/output queue and the actual method to be called.
38
39    Args:
40      num_processes: Number of helpers subprocessors this stage has.
41      name: The name of this stage.
42      cache: The computed tasks encountered before.
43      stage: An int value that specifies the stage for this pipeline stage, for
44        example, build stage or test stage. This value will be used to retrieve
45        the keys in different stage. I.e., the flags set is the key in build
46        stage and the checksum is the key in the test stage. The key is used to
47        detect duplicates.
48      task_queue: The input task queue for this pipeline stage.
49      helper: The method hosted by the helper module to fill up the cost of the
50        duplicate tasks.
51      worker: The method hosted by the worker pools to do the actual work, e.g.,
52        compile the image.
53      result_queue: The output task queue for this pipeline stage.
54    """
55
56    multiprocessing.Process.__init__(self)
57
58    self._name = name
59    self._task_queue = task_queue
60    self._result_queue = result_queue
61
62    self._helper = helper
63    self._worker = worker
64
65    self._cache = cache
66    self._stage = stage
67    self._num_processes = num_processes
68
69    # the queues used by the modules for communication
70    manager = multiprocessing.Manager()
71    self._helper_queue = manager.Queue()
72    self._work_queue = manager.Queue()
73
74  def run(self):
75    """Busy pulling the next task from the queue for execution.
76
77    Once a job is pulled, this stage invokes the actual stage method and submits
78    the result to the next pipeline stage.
79
80    The process will terminate on receiving the poison pill from previous stage.
81    """
82
83    # the worker pool
84    work_pool = multiprocessing.Pool(self._num_processes)
85
86    # the helper process
87    helper_process = multiprocessing.Process(target=self._helper,
88                                             args=(self._cache,
89                                                   self._helper_queue,
90                                                   self._work_queue,
91                                                   self._result_queue))
92    helper_process.start()
93    mycache = self._cache.keys()
94
95    while True:
96      task = self._task_queue.get()
97      if task == POISONPILL:
98        # Poison pill means shutdown
99        self._result_queue.put(POISONPILL)
100        break
101
102      task_key = task.get_key(self._stage)
103      if task_key in mycache:
104        # The task has been encountered before. It will be sent to the helper
105        # module for further processing.
106        self._helper_queue.put(task)
107      else:
108        # Let the workers do the actual work.
109        work_pool.apply_async(self._worker, args=(task, self._work_queue,
110                                                   self._result_queue))
111        mycache.append(task_key)
112
113    # Shutdown the workers pool and the helper process.
114    work_pool.close()
115    work_pool.join()
116
117    self._helper_queue.put(POISONPILL)
118    helper_process.join()
119