gsan: > On Friday July 11 2008, Andre Nathan wrote: > > On Thu, 2008-07-10 at 16:52 -0700, Don Stewart wrote: > > > Well, they're radically different graph representations, and fgl > > > hasn't been designed for large graphs. > > > > Do you know if King and Launchbury's implementation (Data.Graph) scales > > better? > > Looks like it. > > I now did a rough benchmark on fully connected graphs with 25 and 50 nodes. > Data.Graph.Inductive used 28MB and 365MB respectively (x13 increase) compared > to Data.Graph's 486KB and 2MB (x4). And Data.Graph seems to be much faster > (basic operations up to 200 times), although it might be related to GC. >
Do you have the bencmark code? I'd like to try a couple of variants on the underlying structures. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
