Are there any known ways to confuse the garbage collector in skill? I have a moderately large simulation that uses Ocean to step through a number of conditions. For each condition, there is a transient analysis followed by some post-processing code which produces a waveform family (a waveform family for each transient sim). I don't think anything in my code should be expanding the memory used each time through the loop because in simplified form it looks like:
foreach(sim_setup all_sim_setups
;; close all viva windows (in case that was a memory killer)
foreach(win awvGetWindowList() awvCloseWindow(win))
;; reset the remaining window
awvResetWindow(awvGetCurrentWindow() ?force t)
;; configure the sim with desVar() calls, may be update some of the simulation analysis settings
printf("memoryAllocated() returned %g\n", memoryAllocated()) ;; check memory
gc() ;; force garbage collection
run()
;; access some results, do math (including a fairly large number of calls to dft()), plot various results,
;; write out some png files with saveGraphImage(), dump some text files with
;; fp=outfile("somefile.txt") fprintf(fp, "stuff") close(fp)
)
What I see is before the first run, the memory allocated is low and then it jumps a bunch after the first sim. But after that I'm growing by about 300Mb each time through:
memoryAllocated() returned 372.344
Done - free list cells = 2724041, bytes allocated for all types = 182042907
memoryAllocated() returned 5310.04
Done - free list cells = 2804791, bytes allocated for all types = 200569114
memoryAllocated() returned 5606.61
Done - free list cells = 2813000, bytes allocated for all types = 200666771
memoryAllocated() returned 5908.91
Done - free list cells = 2804808, bytes allocated for all types = 200776481
memoryAllocated() returned 6181.61
Done - free list cells = 2802109, bytes allocated for all types = 200945203
memoryAllocated() returned 6477.51
Done - free list cells = 2799522, bytes allocated for all types = 201107058
memoryAllocated() returned 6779.2
Done - free list cells = 2790640, bytes allocated for all types = 201284560
memoryAllocated() returned 7049.48
Done - free list cells = 2809618, bytes allocated for all types = 201976762
memoryAllocated() returned 7345.52
Done - free list cells = 2806987, bytes allocated for all types = 202142740
memoryAllocated() returned 7647.66
Done - free list cells = 2796874, bytes allocated for all types = 202318483
memoryAllocated() returned 7917.4
Done - free list cells = 2815944, bytes allocated for all types = 203002061
memoryAllocated() returned 8214.05
Done - free list cells = 2813363, bytes allocated for all types = 203163430
memoryAllocated() returned 8515.86
Done - free list cells = 2802934, bytes allocated for all types = 203276529
memoryAllocated() returned 8784.89
Done - free list cells = 2800332, bytes allocated for all types = 203505040
and then, *boom*, ocean fails with:
simulate...
*Error* ipcBeginProcess: Unable to fork() child process due to system error
(Consider exiting any unnecessary applications) - "Cannot allocate memory"
*
I've had the sense that maybe the simulation post processing code is where the leak is because I have a waveform calculator version and have noticed running in ADE-L that repeated sims seem to cause the memory use to keep growing.
So... that gets back to the original question. What should I be looking for that causes memory leaks? I would have thought it would be things like opening more and more windows without closing or creating new variables with new names each time through a loop or adding to a list or table or waveform family each time. But as near as I can tell I really shouldn't be needing any additional storage after the first time through.
Thanks
-Dan
![]()
