On Mon, 30 Jul 2012, Dave Abrahams wrote:
on Fri Jul 13 2012, Alex Hagen-Zanker <ahh34-AT-cam.ac.uk> wrote:
https://svn.boost.org/trac/boost/ticket/7130.
To see for yourself: http://people.pwf.cam.ac.uk/ahh34/dijkstra_test.zip
I am interested to hear your opinion.
It might make sense to consider whether Boost.Context could be used to do this job without major restructuring of the existing dijkstra code. http://lists.boost.org/boost-announce/2011/05/0310.php
It would work, but I think using a lower-level, more general library like that would have a higher performance impact than doing something algorithm-specific. I think doing something like a manual translation of N3328 resumable functions or similar to get generators (with event points as yield statements) would be faster, even though it would tangle the algorithm specification more. I have written macros that help to make that kind of code cleaner for previous projects, so a variant of them might help for this use case as well. -- Jeremiah Willcock