boost graph removing edges vs filtered_graph performance

Question

A single threaded boost graph that repeatedly removes and then adds same edges later seems to be faster than a multi-threaded filtered_graph that uses edge predicates to concurrently do the same on multiple filtered graphs of the same underlying graph.

Possibly, filtered_graph traverses each and every edge and applies the predicate.

Is there a way to directly generate a filtered_graph without predicates i.e. use functions like remove_edge but actually create a filtered_graph?

score 2 · Accepted Answer · answered Feb 23 '16 at 06:50

2

Just use copy_graph on the filtered graph:

boost graph copy and removing vertex

answered Feb 23 '16 at 06:50

sehe

374,641
47
450
633

This indeed seems to be faster than using the filtered graph itself. Is it really the case that the filter predicates are re-evaluated each time an edge is accessed? Or what is the reason for this speed difference? – lucidbrot Jan 25 '21 at 08:28
2

@lucidbrot I'm not sure whether optimizations are in place, but I imagine it would be very hard to get that correctly generically (graph concepts are _very_ generic) so, yeah, it makes a bit of sense. – sehe Jan 25 '21 at 13:02

boost graph removing edges vs filtered_graph performance

1 Answers1