question about the algorithm

[xlqqqqq]

Why does it show that Dijkstra is faster than Barrier after I run the demo.py file? I also come in to test like the Quick Start section, randomly generated a map with 10,000 nodes and 40,000 edges, and the test results showed that Dijkstra is still faster. The test.py file contains my testing process, where did I go wrong?

test.py

[arman0z]

This is actually the expected behavior.

The algorithm is designed for very large graphs (100k+ nodes). On your 10,000-node test, Dijkstra is faster because it has lower overhead.

I just pushed fixes that disables the advanced features on smaller graphs and matches Dijkstra's performance. The theoretical O(m log^(2/3) n) advantage only shows up on massive graphs where the asymptotic improvement outweighs the implementation overhead.