[NumPy-Tickets] [NumPy] #1668: Masked array calculations memory leak

NumPy Trac numpy-tickets@scipy....
Tue Nov 9 03:38:29 CST 2010


#1668: Masked array calculations memory leak
-------------------------+--------------------------------------------------
 Reporter:  storpipfugl  |       Owner:  somebody
     Type:  defect       |      Status:  new     
 Priority:  normal       |   Milestone:  2.0.0   
Component:  Other        |     Version:  1.5.0   
 Keywords:               |  
-------------------------+--------------------------------------------------
 Test code: [[BR]]
 --- [[BR]]
 mem_leak_ma.py[[BR]]
 ---

 Description:[[BR]]
 Use mem_leak_ma.py. At the raw_input('check_mem') line the array
 data_array and masked array cal_data_array are the only arrays in scope.
 Base memory consumption before calculation is 95 MB. Expected memory
 consumption at end of test program: 3712**2*(8*2 + 1)/10**6 + 95 = 329 MB.
 Readout from /proc/<pid>/status at line raw_input('check_mem'):
 {{{
 VmPeak:   848092 kB
 VmSize:   484768 kB
 }}}
 This leaves 156 MB unaccounted for (VmSize).

 Expected behavior:
 The memory usage at the end of the test program is expected to be
 comparable to the corresponding example with a regular ndarray plus memory
 usage for the mask.

 Working code:[[BR]]
 ---[[BR]]
 mem_leak.py[[BR]]
 ---

 Use mem_leak.py. Readout from /proc/<pid>/status at line
 raw_input('check_mem'):
 {{{
 VmPeak:   417416 kB
 VmSize:   309768 kB
 }}}
 as expected (VmSize).

 System:[[BR]]
 Ubuntu linux 10.04 64 bit[[BR]]
 Python 2.6.5[[BR]]
 Reproduced with numpy 1.3.0 and 1.5.0  [[BR]]

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1668>
NumPy <http://projects.scipy.org/numpy>
My example project


More information about the NumPy-Tickets mailing list