-
|
Hi all, What's the difference between e.g. FastMoore and CompactMoore? When should I prefer one over the other? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
The compact implementations typically store adjacency information in a single array whereas the fast implementations store adjacency information at each state individually. Therefore, compact implementations primarily scale with size, whereas fast implementations primarily scale with density. Unless you work with very large/sparse automata (e.g., one million states+), the compact implementations are a safe default. Ironically, the oftentimes lower memory consumption of the compact implementations can make them faster, too (at least in micro benchmarks). |
Beta Was this translation helpful? Give feedback.
The compact implementations typically store adjacency information in a single array whereas the fast implementations store adjacency information at each state individually. Therefore, compact implementations primarily scale with size, whereas fast implementations primarily scale with density.
Unless you work with very large/sparse automata (e.g., one million states+), the compact implementations are a safe default. Ironically, the oftentimes lower memory consumption of the compact implementations can make them faster, too (at least in micro benchmarks).