Your IP : 216.73.216.61
,|�W�7 � @ s� d Z d d l Z d d l Z d d l Z d d l Z d d l Z d d l m Z m Z d d l m
Z
d d l m Z m
Z
d d l m Z d d l m Z e d k r� d d l m Z d d
l m Z Gd d � d e e � � Z Gd
d � d e � Z Gd d � d e � Z Gd d � d e � Z Gd d � d e e � Z Gd d � d e e e � Z Gd d � d e � Z Gd d � d e � Z Gd d � d e � Z d S) z,
Backends for embarrassingly parallel code.
� N)�ABCMeta�abstractmethod� )�
format_exc)�WorkerInterrupt�TransportableException)�mp)�with_metaclass)�
MemmapingPool)�
ThreadPoolc @ s� e Z d Z d Z e d d � � Z e d d d � � Z d d d d � Z d
d � Z d d
� Z d d � Z
d d � Z d d d � Z d S)�ParallelBackendBasezEHelper abc which defines all methods a ParallelBackend must implementc C s d S)a� Determine the number of jobs that can actually run in parallel
n_jobs is the number of workers requested by the callers. Passing
n_jobs=-1 means requesting all available workers for instance matching
the number of CPU cores on the worker host(s).
This method should return a guesstimate of the number of workers that
can actually perform work concurrently. The primary use case is to make
it possible for the caller to know in how many chunks to slice the
work.
In general working on larger data chunks is more efficient (less
scheduling overhead and better use of CPU cache prefetching heuristics)
as long as all the workers have enough work to do.
N� )�self�n_jobsr
r
�J/opt/alt/python35/lib/python3.5/site-packages/joblib/_parallel_backends.py�effective_n_jobs s z$ParallelBackendBase.effective_n_jobsNc C s d S)zSchedule a func to be runNr
)r �func�callbackr
r
r � apply_async* s z ParallelBackendBase.apply_asyncr c K s | | _ | j | � S)z�Reconfigure the backend and return the number of workers.
This makes it possible to reuse an existing backend instance for
successive independent calls to Parallel with different parameters.
)�parallelr )r r r � backend_argsr
r
r � configure. s z ParallelBackendBase.configurec C s d S)z#Shutdown the process or thread poolNr
)r r
r
r � terminate7 s z ParallelBackendBase.terminatec C s d S)z Determine the optimal batch sizer r
)r r
r
r �compute_batch_size: s z&ParallelBackendBase.compute_batch_sizec C s d S)z1Callback indicate how long it took to run a batchNr
)r �
batch_size�durationr
r
r �batch_completed> s z#ParallelBackendBase.batch_completedc C s g S)z'List of exception types to be captured.r
)r r
r
r �get_exceptionsA s z"ParallelBackendBase.get_exceptionsTc C s d S)a} Abort any running tasks
This is called when an exception has been raised when executing a tasks
and all the remaining tasks will be ignored and can therefore be
aborted to spare computation resources.
If ensure_ready is True, the backend should be left in an operating
state as future tasks might be re-submitted via that same backend
instance.
If ensure_ready is False, the implementer of this method can decide
to leave the backend in a closed / terminated state as no new task
are expected to be submitted to this backend.
Setting ensure_ready to False is an optimization that can be leveraged
when aborting tasks via killing processes from a local process pool
managed by the backend it-self: if we expect no new tasks, there is no
point in re-creating a new working pool.
Nr
)r � ensure_readyr
r
r �abort_everythingE s z$ParallelBackendBase.abort_everything)
�__name__�
__module__� __qualname__�__doc__r r r r r r r r r r
r
r
r r s r c @ s1 e Z d Z d Z d d � Z d d d � Z d S)�SequentialBackendz�A ParallelBackend which will execute all batches sequentially.
Does not use/create any threading objects, and hence has minimal
overhead. Used when n_jobs == 1.
c C s | d k r t d � � d S)z?Determine the number of jobs which are going to run in parallelr z&n_jobs == 0 in Parallel has no meaningr )�
ValueError)r r r
r
r r e s z"SequentialBackend.effective_n_jobsNc C s t | � } | r | | � | S)zSchedule a func to be run)�ImmediateResult)r r r �resultr
r
r r k s
z SequentialBackend.apply_async)r r! r" r# r r r
r
r
r r$ ^ s r$ c @ sL e Z d Z d Z d d � Z d d � Z d d d � Z d d
d � Z d S) �PoolManagerMixinz,A helper class for managing pool of workers.c C sd | d k r t d � � nE t d k s3 | d k r7 d S| d k r` t t j � d | d � } | S)z?Determine the number of jobs which are going to run in parallelr z&n_jobs == 0 in Parallel has no meaningNr )r% r �max� cpu_count)r r r
r
r r v s z!PoolManagerMixin.effective_n_jobsc C s6 | j d k r2 | j j � | j j � d | _ d S)z#Shutdown the process or thread poolN)�_pool�closer )r r
r
r r � s
zPoolManagerMixin.terminateNc C s | j j t | � d | �S)zSchedule a func to be runr )r+ r � SafeFunction)r r r r
r
r r � s z PoolManagerMixin.apply_asyncTc C s< | j � | r8 | j d | j j d | j | j j � d S)z@Shutdown the pool and restart a new one with the same parametersr r N)r r r r �
_backend_args)r r r
r
r r � s
z!PoolManagerMixin.abort_everything)r r! r" r# r r r r r
r
r
r r( s s
r( c @ sF e Z d Z d Z d Z d Z d Z d Z d d � Z d d � Z d
S) �AutoBatchingMixinz/A helper class for automagically batching jobs.g�������?� r g c C s | j } | j } | d k r� | | j k r� t | | j | � } t d | d � } | | _ | j j d k r� | j j d | | f � nb | | j k r� | d k r� | d } | | _ | j j d k r� | j j d | | f � n | } | | k rd | _ | S)z Determine the optimal batch sizer r0 r �
z:Batch computation too fast (%.4fs.) Setting batch_size=%d.z:Batch computation too slow (%.4fs.) Setting batch_size=%d.) �_effective_batch_size�_smoothed_batch_duration�MIN_IDEAL_BATCH_DURATION�intr) r �verbose�_print�MAX_IDEAL_BATCH_DURATION)r �old_batch_size�batch_duration�ideal_batch_sizer r
r
r r � s2
z$AutoBatchingMixin.compute_batch_sizec C sL | | j k rH | j } | d k r- | } n d | d | } | | _ d S)z1Callback indicate how long it took to run a batchr g�������?g�������?N)r2 r3 )r r r � old_duration� new_durationr
r
r r � s z!AutoBatchingMixin.batch_completedN)
r r! r" r# r4 r8 r2 r3 r r r
r
r
r r/ � s .r/ c @ s( e Z d Z d Z d d d d � Z d S)�ThreadingBackenda� A ParallelBackend which will use a thread pool to execute batches in.
This is a low-overhead backend but it suffers from the Python Global
Interpreter Lock if the called function relies a lot on Python objects.
Mostly useful when the execution bottleneck is a compiled extension that
explicitly releases the GIL (for instance a Cython loop wrapped in a
"with nogil" block or an expensive call to a library such as NumPy).
r Nc K sF | j | � } | d k r* t t � � � | | _ t | � | _ | S)z?Build a process or thread pool and return the number of workersr )r �FallbackToBackendr$ r r r+ )r r r r r
r
r r � s zThreadingBackend.configure)r r! r" r# r r
r
r
r r> � s r> c sR e Z d Z d Z d Z � f d d � Z d d d d � Z � f d d
� Z � S) �MultiprocessingBackenda A ParallelBackend which will use a multiprocessing.Pool.
Will introduce some communication and memory overhead when exchanging
input and output data with the with the worker Python processes.
However, does not suffer from the Python Global Interpreter Lock.
�__JOBLIB_SPAWNED_PARALLEL__c s� t d k r d St j � j rB | d k r> t j d d d �d St j � j d k rn t j d d d �d St t | � j
| � S)z�Determine the number of jobs which are going to run in parallel.
This also checks if we are attempting to create a nested parallel
loop.
Nr zHMultiprocessing-backed parallel loops cannot be nested, setting n_jobs=1�
stacklevel� �
MainThreadzVMultiprocessing backed parallel loops cannot be nested below threads, setting n_jobs=1) r �current_process�daemon�warnings�warn� threading�current_thread�name�superr@ r )r r )� __class__r
r r s z'MultiprocessingBackend.effective_n_jobsr Nc K s� | j | � } | d k r* t t � � � t t j j | j d � � } | rZ t d � � d t j | j <t j
� t | | � | _ | | _
| S)z?Build a process or thread pool and return the number of workersr r a% [joblib] Attempting to do parallel computing without protecting your import on a system that does not support forking. To use parallel-computing in a script, you must protect your main loop using "if __name__ == '__main__'". Please see the joblib documentation on Parallel for more information�1)r r? r$ r5 �os�environ�get�JOBLIB_SPAWNED_PROCESS� ImportError�gc�collectr
r+ r )r r r r �already_forkedr
r
r r " s
z MultiprocessingBackend.configurec s6 t t | � j � | j t j k r2 t j | j =d S)z#Shutdown the process or thread poolN)rL r@ r rR rO rP )r )rM r
r r ; s z MultiprocessingBackend.terminate)r r! r" r# rR r r r r
r
)rM r r@ � s
r@ c @ s( e Z d Z d d � Z d d � Z d S)r& c C s | � | _ d S)N)�results)r �batchr
r
r �__init__C s zImmediateResult.__init__c C s | j S)N)rW )r r
r
r rQ H s zImmediateResult.getN)r r! r" rY rQ r
r
r
r r& B s r& c @ s. e Z d Z d Z d d � Z d d � Z d S)r- a Wrapper that handles the serialization of exception tracebacks.
If an exception is triggered when calling the inner function, a copy of
the full traceback is captured to make it possible to serialize
it so that it can be rendered in a different Python process.
c C s
| | _ d S)N)r )r r r
r
r rY S s zSafeFunction.__init__c O s y | j | | � SWnd t k
r1 t � � YnJ t j � \ } } } t | | | d d d d �} t | | � � Yn Xd S)N�contextr1 � tb_offsetr )r �KeyboardInterruptr �sys�exc_infor r )r �args�kwargs�e_type�e_value�e_tb�textr
r
r �__call__V s
zSafeFunction.__call__N)r r! r" r# rY re r
r
r
r r- L s r- c @ s" e Z d Z d Z d d � Z d S)r? z<Raised when configuration should fallback to another backendc C s
| | _ d S)N)�backend)r rf r
r
r rY g s zFallbackToBackend.__init__N)r r! r" r# rY r
r
r
r r? d s r? )!r# rT rO r] rG rI �abcr r � format_stackr �
my_exceptionsr r �_multiprocessing_helpersr �_compatr �poolr
�multiprocessing.poolr r r$ �objectr( r/ r> r@ r& r- � Exceptionr? r
r
r
r �<module> s. I"P G