[1.49.0] New libraries in 1.49.0?
What are the new libraries in 1.49.0?
boost.heap [1] cheers, tim [1] http://boost-sandbox.sourceforge.net/doc/html/heap.html
On Sat, Jan 28, 2012 at 8:54 AM, Tim Blechmann <tim@klingt.org> wrote:
What are the new libraries in 1.49.0?
boost.heap [1]
cheers, tim
Thanks for the info! Boost.Heap is one I've been looking forward to! --Beman
on Sat Jan 28 2012, Tim Blechmann <tim-AT-klingt.org> wrote:
What are the new libraries in 1.49.0?
boost.heap [1]
cheers, tim
Nice; will Boost.Graph be using these its algorithms (e.g. Dijkstra)? -- Dave Abrahams BoostPro Computing http://www.boostpro.com
What are the new libraries in 1.49.0?
boost.heap [1]
cheers, tim
Nice; will Boost.Graph be using these its algorithms (e.g. Dijkstra)?
bgl has its own heap implementations, which are really optimized for its use case as they assume that priority keys are integer values. boost.heap is much more generic, so it might not make sense to change bgl ... cheers, tim
On Tue, 31 Jan 2012, Tim Blechmann wrote:
What are the new libraries in 1.49.0?
boost.heap [1]
cheers, tim
Nice; will Boost.Graph be using these its algorithms (e.g. Dijkstra)?
bgl has its own heap implementations, which are really optimized for its use case as they assume that priority keys are integer values. boost.heap is much more generic, so it might not make sense to change bgl ...
That's something someone should write a benchmark for -- the assumption that keys are integers allows for arrays as storage in some places, but other data structures can be used as well. -- Jeremiah Willcock
Jeremiah Willcock wrote:
On Tue, 31 Jan 2012, Tim Blechmann wrote:
bgl has its own heap implementations, which are really optimized for its use case as they assume that priority keys are integer values. boost.heap is much more generic, so it might not make sense to change bgl ...
That's something someone should write a benchmark for -- the assumption that keys are integers allows for arrays as storage in some places, but other data structures can be used as well.
I sometimes manipulate huge data (e.g. volume images), and space-efficiency is a very important factor there. So I would love Boost.Heap to have mutable heap with array-like storage. Currently I had to use my own heap class (or boost/pending things). Regards, Michel
On Thu, 2 Feb 2012, Michel Morin wrote:
Jeremiah Willcock wrote:
On Tue, 31 Jan 2012, Tim Blechmann wrote:
bgl has its own heap implementations, which are really optimized for its use case as they assume that priority keys are integer values. boost.heap is much more generic, so it might not make sense to change bgl ...
That's something someone should write a benchmark for -- the assumption that keys are integers allows for arrays as storage in some places, but other data structures can be used as well.
I sometimes manipulate huge data (e.g. volume images), and space-efficiency is a very important factor there. So I would love Boost.Heap to have mutable heap with array-like storage. Currently I had to use my own heap class (or boost/pending things).
Look at <boost/graph/detail/d_ary_heap.hpp> as well and see if that helps you; you'll need to look at <boost/graph/dijkstra_shortest_paths.hpp> for examples of how to use it. -- Jeremiah Willcock
Jeremiah Willcock wrote:
Look at <boost/graph/detail/d_ary_heap.hpp> as well and see if that helps you; you'll need to look at <boost/graph/dijkstra_shortest_paths.hpp> for examples of how to use it.
Great! I hope this will become public interface. One concern is that the implementation is specialized for monotonically decreasing costs (distances in shortest-path problems). In my applications, the cost can be increased. Regards, Michel
participants (5)
-
Beman Dawes -
Dave Abrahams -
Jeremiah Willcock -
Michel Morin -
Tim Blechmann