Many-to-all shortest path
Hi, I went over the documentation and some foruns but I was not able to find the answer for a question that is crucial for me. Can I compute many-to-all Shortest paths (about 5,000 origins in a 100,000 nodes network) taking advantage of all computer processors? Looking at the code, it looks like Dijkstra is computed with the GIL locked, but I was not sure how it goes for the other algorithms. Thanks, Pedro
On 19.01.2017 05:09, Pedro Camargo wrote:
Can I compute many-to-all Shortest paths (about 5,000 origins in a 100,000 nodes network) taking advantage of all computer processors?
This is not implemented, but can be achieved easily with threads or multiprocessing.
Looking at the code, it looks like Dijkstra is computed with the GIL locked, but I was not sure how it goes for the other algorithms.
I don't think this is true, since the library dispatches to C++. -- Tiago de Paula Peixoto <tiago@skewed.de>
Thanks, Thiago. In that case, I guess I need ond copy of the graph per thread. Correct? And since we are at it, does anybody have experience in using graph-tool wrapped inside Cython code? Cheers, Pedro On 19 Jan 2017 7:54 PM, "Tiago de Paula Peixoto" <tiago@skewed.de> wrote:
On 19.01.2017 05:09, Pedro Camargo wrote:
Can I compute many-to-all Shortest paths (about 5,000 origins in a 100,000 nodes network) taking advantage of all computer processors?
This is not implemented, but can be achieved easily with threads or multiprocessing.
Looking at the code, it looks like Dijkstra is computed with the GIL locked, but I was not sure how it goes for the other algorithms.
I don't think this is true, since the library dispatches to C++.
-- Tiago de Paula Peixoto <tiago@skewed.de> _______________________________________________ graph-tool mailing list graph-tool@skewed.de https://lists.skewed.de/mailman/listinfo/graph-tool
participants (2)
-
Pedro Camargo -
Tiago de Paula Peixoto