On Apr 24, 2006, at 10:59 AM, xiao-hui.wu@exxonmobil.com wrote:
Has anyone compared the performance of Boost Graph Library with LEDA?
The following is a link I received that contains some comparisons. I would like to know your experience since I am trying to decide which one is more appropriate for my problems.
That comparison is based on the max-flow algorithm, only, and the BGL implementation of that particular algorithm has not been optimized. We have done some tests in the past that show the BGL to provide better performance than LEDA in other cases (e.g., I'm looking at some charts that show the BGL running 6x faster than an old version of LEDA on connected components). Nobody can draw any defensible conclusions without a real study involving many different algorithms. We can't do that study without a LEDA license; they chose not to expand their max-flow study to other algorithms.
I like the generic interface of BGL, although LEDA has a C# counterpart called GOAL, which uses C# interfaces. I am not sure if the performance of GOAL matches LEDA, but it seems to provide some customization capability.
There is a C# counterpart to the BGL called QuickGraph, but I don't know how far along it is. There are also Python bindings for the BGL, which wrap the C++ BGL. They provide "most" of the performance of the BGL and "most" of its customizability. Doug