|
|||||||||
|
|||||||||
| |||||||||
|
|||||||||
|
|||||||||
| |||||||||
|
|
|
| |||||||||
 |
|
|
«
Previous Thread
|
Next Thread
»
|
Thread Tools | Search this Thread | Rate Thread | Display Modes |
Dev Shed Forums Sponsor:
|
|
Stop making mediocre tutorials.The best tutorials are video! Camtasia Studio makes it easy to create engaging, buzz-building screen videos at any size, in any popular format. Download the free trial!
|
|
#1
September 3rd, 2003, 08:23 PM
|
|||
|
|||
Unused memory unrecoverable?
So I'm writing a program in Python, and comparing the memory usage of arrays to hash tables... And when testing arrays, I write the following two pieces of code.
The first: Code:
from random import randint, seed, sample from bisect import insort_left from time import time array = [] seed(5) # record when operation begins t_beg = time() # for 10 iterations for i in range(10): # maintaining order, insert 2000 random [x, y] pairs into the array for j in xrange(2000): insort_left(array, [randint(0, 0xFFFFFFF), randint(0, 0xFFFFFFF)]) # now delete 500 randomly chosen pairs rem = sample(xrange(len(array)), 500) rem.sort() rem.reverse() for j in rem: del array[j] # record when operation ends t_end = time() # print statistics print "took " + str(t_end - t_beg) + " seconds" print "size is " + str(len(array)) The secod piece of code is just like the first, but I delete 12000 randomly chosen elements of the array at the end: Code:
from random import randint, seed, sample from bisect import insort_left from time import time array = [] seed(5) # record when operation begins t_beg = time() # for 10 iterations for i in range(10): # maintaining order, insert 2000 random [x, y] pairs into the array for j in xrange(2000): insort_left(array, [randint(0, 0xFFFFFFF), randint(0, 0xFFFFFFF)]) # now delete 500 randomly chosen pairs rem = sample(xrange(len(array)), 500) rem.sort() rem.reverse() for j in rem: del array[j] # record when operation ends t_end = time() # 10*(2000-500) = 15000 entries exist; delete 12000 (80%) at random rem = sample(xrange(len(array)), 12000) rem.sort() rem.reverse() for j in rem: del array[j] # print statistics print "took " + str(t_end - t_beg) + " seconds" print "size is " + str(len(array)) Now after deleting 12000 elements of the array, you'd think that memory would be recovered, right? Well through the pseudo-scientific method of bringing up task manager in Windows XP and watching the memory usage of python.exe in the Processes tab, it seems the memory isn't reclaimed -- both code segments occupy approximately the same amount of memory after execution (1.5 MB). For the second piece of code, I even tried importing the garbage collector module and running gc.collect(), but to no avail. What gives? Thanks, - theperfectsoup
|
|
#2
September 4th, 2003, 06:42 AM
|
||||
|
||||
|
Oh ok, i'll have to test this myself when I get home.. just a few questions - always seem to be asking Q's - what version of Python are you using? Does isenabled() return true?
 also, when the program has finished the memory is returned anyway right?Have fun, Mark.
|
|
#3
September 4th, 2003, 02:32 PM
|
|||
|
|||
|
I'm using Python version 2.3. The call gc.isenabled() returns 1 (True). And when I quit Python, the memory is all recovered.
After experimenting a little more, here are some other interesting facts I've found... 1. After you delete the 12,000 elements of the array, if you then allocate 12,000 elements (bringing the total back up to 15,000), then memory usage does NOT increase. So when you delete those 12,000 elements, the memory is put aside for future allocated elements to use, but it is not released by the Python interpreter. 2. If you use the Array module, which basically creates underlying arrays in C, allocate 15,000 elements, and then delete 12,000 you can clearly see the memory being freed via task manager. This is the behavior I want, but the problem is I need to store 80-bit numbers and Array only typically allows 32-bit (64-bit for floating point). I've thought about splitting up the 80-bit values across three Array variables -- one for the highest 32-bits, the next for the middle 32-bits, and the last for the lowest 16-bits, but because of the way I'm storing data (including multidimensional arrays, which Array does not allow), that falls through. Oh well. - theperfectsoup
|
|
#4
September 4th, 2003, 06:05 PM
|
||||
|
||||
|
I surpose in a way it makes a strange kinda sence for the interpriter to try hold onto the memory it's used untill it's done - not that thats what it does - that way it doesn't have to play tag with other programs, once it has it it doesn't have to let it go and get it back again
, Just one point of view  Strange problem your having there, what exactly are you trying to do that requires putting numbers that long in a list? sounds interesting though..by 32 bit you mean a number with 32 trailing 0's? Have fun, Mark.
|
|
| Viewing: Dev Shed Forums > Programming Languages > Python Programming > Unused memory unrecoverable? |
| Thread Tools | Search this Thread |
| Display Modes | Rate This Thread |
|
|
 |
|
|
Del.icio.us
Digg
Google
Spurl
Blink
Furl
Simpy
Y! MyWeb
Linear Mode
