Re: [Boost-users] newbie BGL problem: core dump due to graph size

From: Jeremy Siek (jsiek_at_[hidden])
Date: 2003-12-12 09:24:13

Hi Duccio,

The only intrinsic size limitations, say for adjacency_list, that I am
aware of are in the range of numbers representable by an "int". As for the
algorithms, I am not aware of any size limitations. Of course, there's
always the possibility of a bug in there somewhere...

Cheers,
Jeremy

On Thu, 11 Dec 2003 duccio.medini_at_[hidden] wrote:
duccio> Hi all.
duccio>
duccio> I am calculating the connected components of af graph with a c++ code
duccio> written with BGL.
duccio> My graph is huge (4E+05 vertexes, 10^6 ~ 10^8 edges).
duccio> The code uses the connected_components algorithm (or, in a different
duccio> version, directly the depth_first_visit one) and works well up to a
duccio> "reasonable" number of edges in the graph.
duccio> Over this threshold it core dumps.
duccio> I have the same problem both on SPARC Solaris 9 g++3.2.2 and on Intel Linux
duccio> 2.4 g++3.1 .
duccio> I know that it should not be a problem of the RAM memory available on the
duccio> system (or addressable form a single istance). I am able to use up to 6-7
duccio> GBs of RAM, and it tipically core dumps at 2.5-5 GBs.
duccio>
duccio> I suspect that I'm violating some intrinsic size limit in the BGL
duccio> adjacency_list  class or similar.
duccio> Are there such limits, and can they be extended?
duccio> Any clue would be really appreciated.
duccio>
duccio> Thanks in advance!
duccio>
duccio> Duccio Medini

----------------------------------------------------------------------
 Jeremy Siek                          http://php.indiana.edu/~jsiek/
 Ph.D. Student, Indiana Univ. B'ton   email: jsiek_at_[hidden]
 C++ Booster (http://www.boost.org)   office phone: [phone]
----------------------------------------------------------------------