I noticed that insertion time in judyL is significantly larger compared to lookup. is it a correct observation? Is there anything that can be done to improve it?
My old design: 2 levels of AVL trees and 3-rd level is a linked list.
My new design: all 3 levels of JudyL.
The lookup with JudyL is definitely faster but while building my data structure (all insertions) and later doing insertion/deletion is slower than my old design.
1st level has max of 500 entries. each new entry unique and cannot be overlapping with existing entries.
2nd level has max of 1000 entries. each new entry unique and cannot be overlapping with existing entries.
3rd level has max of 2000 entries. new entry may overlap multiple entries. so entries corresponding to the holes are inserted.
In my initial tests where I'm seeing performance difference:
1st level has < 50 entries.
2nd level has < 100 entries
3rd level has < 128 entries.
Is this low performance expected from JudyL for insertions? Any suggestions. Thanks.