Benchmarks for Various Data Structures

Results

Table of results (with no compiler optimizations: -O0). All times are in seconds.

Data structure insert
(random) delete
(random) insert
(sorted) delete
(sorted) insert
(sorted rev.) delete
(sorted rev.) find
(random)

AVL tree 1.420 2.015 0.917 1.002 0.902 0.994 0.712

Red-black tree 0.750 0.832 0.354 0.207 0.347 0.197 0.717

Skiplist 1.104 1.177 0.319 0.183 0.245 0.282 0.135

BList (2048) 4.335 3.299 9.377 (no opt)
3.398 (w/opt) 1.172 2.914 3.490 2.452

Hash table (OA) 0.269 2.914 (PACK)
1.014 (MARK) 0.251

Hash table (chaining) 0.391 0.392 0.367

Data structure	insert (random)	delete (random)	insert (sorted)	delete (sorted)	insert (sorted rev.)	delete (sorted rev.)	find (random)
AVL tree	1.420	2.015	0.917	1.002	0.902	0.994	0.712
Red-black tree	0.750	0.832	0.354	0.207	0.347	0.197	0.717
Skiplist	1.104	1.177	0.319	0.183	0.245	0.282	0.135
BList (2048)	4.335	3.299	9.377 (no opt) 3.398 (w/opt)	1.172	2.914	3.490	2.452
Hash table (OA)	0.269	2.914 (PACK) 1.014 (MARK)					0.251
Hash table (chaining)	0.391	0.392					0.367

Table of results (with compiler optimization level -O2). All times are in seconds.

Data structure insert^*
(random) delete^*
(random) insert
(sorted) delete
(sorted) insert
(sorted rev.) delete
(sorted rev.) find^*
(random)

AVL tree 0.689 0.842 0.186 0.197 0.185 0.196 0.345

Red-black tree 0.653 0.601 0.223 0.110 0.228 0.109 0.328

Skiplist 0.790 0.768 0.138 0.066 0.111 0.111 0.054

BList¹ 1.837 2.461 6.297 (no opt)²
0.691 (w/opt)³ 0.175 0.475 3.371 2.402

Hash table (OA)⁴ 0.213 0.655 (PACK)⁵
0.187 (MARK)⁶ 0.222

Hash table (chaining)⁷ 0.360 0.345 0.346

Data structure	insert^* (random)	delete^* (random)	insert (sorted)	delete (sorted)	insert (sorted rev.)	delete (sorted rev.)	find^* (random)
AVL tree	0.689	0.842	0.186	0.197	0.185	0.196	0.345
Red-black tree	0.653	0.601	0.223	0.110	0.228	0.109	0.328
Skiplist	0.790	0.768	0.138	0.066	0.111	0.111	0.054
BList¹	1.837	2.461	6.297 (no opt)² 0.691 (w/opt)³	0.175	0.475	3.371	2.402
Hash table (OA)⁴	0.213	0.655 (PACK)⁵ 0.187 (MARK)⁶					0.222
Hash table (chaining)⁷	0.360	0.345					0.346

^* These columns are the most relevant because they represent the average/typical case.

¹The BList was tested with 2,048 items per node.
²No special optimizations were made and the normal search code was used to find the position.
³Before walking the list looking for the insertion point, a check was made to see if the value being inserted was larger than the largest element in the list. This meant no searching was required and it could simply be inserted at the end.
⁴The confguration: load factor is 0.667, hash function is Super Fast, and the collision resolution policy is linear probing.
⁵When an item is deleted, the remaining items in the cluster are reinserted into the table.
⁶When an item is deleted, the slot is simply marked as DELETED.
⁷The confguration: load factor is 5, hash function is Super Fast.

Here are some blist benchmarks that show the differences in performance.