Closed juroland closed 7 years ago
This would be better as two PRs (can you please use kernel style commit messages - the bodies are good, but they should have a header line).
Do you have a provenance for the optimisation, or is it purely from your analysis - it partially invalidates the comment above, so it would be good to say where the optimisation is from in that comment as an addendum.
When these are merged, we need to have you added to A+C in gonum/license.
Yes, as you guessed, this optimization is only based on my analysis.
Ok for the two PRs with commit messages that use the kernel style. They will arrive shortly.
The first commit changes complexity based on the reference provided in comments (i.e. http://www.cs.utexas.edu/ftp/techreports/tr07-54.pdf). The second one is a first proposal to reduce unnecessary work due to the use of duplicated nodes in the priority queue. Better alternatives could be discussed. What do you think about this ?