Re: [igraph] shortest paths and ram

[Tamas Nepusz]

Hi Davide,

> another (stupid? ;-) ) question: i'm trying to find the average
> distance for a weighted strongly connected graph. I think that i
> should use Graph.shortest_paths() looping over each node, but step
> after step the ram consuption increases.
Python's garbage collection mechanisms might cause that, but I'm not
100% sure. Currently I'm running your code in Valgrind to check for
possible memory leaks in the igraph core. In the meanwhile, you might
try calling Python's garbage collection mechanism explicitly using
gc.collect(). What I did is: I called g.shortest_paths_dijkstra() 10
times on your graph with a single vertex, noted that the memory usage
went up (cat /proc/<pid>/status | grep VmSize), then called
gc.collect(), called g.shortest_paths_dijkstra() 10 times again and
noted that the memory usage stayed the same.

Note that gc.collect() will never release memory back to the OS, this is
because Python's memory manager uses an internal pool and it never gives
back anything to the OS in the hope that it might be useful later.
That's why the memory usage didn't decrease for me after calling
gc.collect()

-- 
Tamas

pgpbNCw2nQj_P.pgp
Description: PGP signature