[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Discuss-gnuradio] Virtual memory problem assumably caused by thread
Re: [Discuss-gnuradio] Virtual memory problem assumably caused by threaded process
Mon, 13 Aug 2007 00:13:54 -0700
On Fri, Aug 10, 2007 at 07:41:04PM +0200, Christian Sokolowski wrote:
> Hi all,
> the receiver of a point-to-point communication system is causing some memory
> rx_graph() architecture:
> usrp.board - some signal processing blocks - packet sink
> Arriving packets are fed to a message_queue. A thread watches this
> queue and arriving packets are proceeded to main().
> Pseudo-code of the receiver:
> class rx_graph (gr.flow_graph):
> def __init__(self, callback):
> self.dst_queue = gr.msg_queue(1)
> self._watcher = _queue_watcher_thread (self.dst_queue, callback)
> def configure(self, _k, _RXfreq):
> 'all signal processing blocks are connect here...'
> class _queue_watcher_thread(_threading.Thread):
> def __init__(self, rcvd_pktq, callback):
> self.rcvd_pktq = rcvd_pktq
> self.callback = callback
> self.keep_running = True
> def run(self):
> while self.keep_running:
> msg = self.rcvd_pktq.delete_head()
> payload = msg.to_string()
> if self.callback:
> self.callback(payload, bits_correct, bits_wrong)
> def callback(payload, bits_correct, bits_wrong):
> "payload is processed here"
> while True:
> # a new object of rx_graph is created after every 100 transmitted
> # packets in order to perform rate adaptation
> rx = rx_graph(callback) # creates an object of rx_graph
> # receiving modulation and carrier frequency via TCP socket"
> # modulation order and carrier frequency have been separately
> # transmitted with TCP before starting to receive payload packets
> rx.configure( modulation_order, RXfreq )
> ...now receiving packets...
> rx = 0 # <------------ works somehow, but this does not free the
> # memory and stop the thread!?
> The application works as I wish, but some memory is not freed properly.
> For some unknown reason virtual memory usage is continously increasing and
> the python application crashes after exceeding 1.6GB with the following
> error message:
> File "./GA_receiver_ver7.py", line 237, in <module>
> File "./GA_receiver_ver7.py", line 192, in main
> File "/usr/local/lib/python2.5/site-packages/gnuradio/gr/flow_graph.py",
> line 93, in start
> self.scheduler.start ()
> File "/usr/local/lib/python2.5/site-packages/gnuradio/gr/scheduler.py",
> line 57, in start
> File "/usr/lib/python2.5/threading.py", line 434, in start
> _start_new_thread(self.__bootstrap, ())
> thread.error: can't start new thread
> Exception exceptions.AssertionError: AssertionError('cannot join thread
> before it is started',) in <bound method rx_graph.__del__ of
> <__main__.rx_graph object at 0x82903ec>> ignored
> I wonder if the _queue_watcher_thread is closed each time at the end of the
> while loop in main properly when I set the object rx=0. Do I need to
> consider some memory deallocation issues?
> I greatly appreciate any kind of help.
> The full receiver script can be seen here:
> Thanks, Chris
I think that you are right that something is leaking memory, and that
it could be the _queue_watcher_thread. You should be able to confirm
the failure to delete the thread by watching with ps. The threads
associated with python will continue to grow.
To get info about threads on GNU/Linux:
Other things to check:
Is anybody is setting the _queue_watcher_thread self.keep_running to False?
The loop could be hung in the rcvd_pktq.delete_head(). Somebody may
need to to insert_tail a special message that says "exit loop".
msg = self.rcvd_pktq.delete_head()
ok, payload = packet_utils.unmake_packet(msg.to_string(),
Thanks for looking at this,