search

jinnaiyuu / Hash-Distributed-Astar

Hash Distributed A*
https://sites.google.com/site/yuujinnaishomepage/home/parallel-search
MIT License
3 stars 1 forks source link

TODO: 2015/07/13 #103

Closed jinnaiyuu closed 9 years ago

jinnaiyuu commented 9 years ago

1 . analyze balance: exclude static predicates 2 . evaluate all hash methods 3 . build environment of writing paper and list the headings 4 . write script for funlucy

jinnaiyuu commented 9 years ago

3 . writing paper first thing I have to do is to revisit the results I have now.

24 puzzle: check pdb grid: revisit speedup as it is not that impressive MSA: revisit TSP: revisit

jinnaiyuu commented 9 years ago

5 . compare node per second vs fast-downward+pdb. mine is approximately 100,000/s for blocks.

jinnaiyuu commented 9 years ago

6 . Speedup Planner

Strips::apply_action 53 Strips::heuristic 22 Strips::ops 37

plans: 1 . exclude static predicates from trie 2 . delete static predicates from init state (after trie is built) 3 . delete static predicates from actions 4 . apply action efficiently (add&del effects)

jinnaiyuu commented 9 years ago

6 . Speedup Planner

apply_action: speedup add & del to vector

uniquelyMergeSortedVector: anyway to speedup this? called tons of times.

Trie: need to revisit this structure. what is the best implementation of the action trie? Trie::searchNodes Trie::findMatchingChildren

jinnaiyuu commented 9 years ago

apply_action: is it possible to speedup more than here, with this modeling?

Trie: root is the most expensive computation.