Test | Setup | Benchmark |
---|---|---|
Get | - Create an empty arena. - Insert 10,000 entities. |
Get 10,000 entities randomly. |
Insert | - Create an empty arena. | Insert 10,000 entities. |
InsertUsed | - Create an empty arena. - Insert 10,000 entities. - Remove all entities sequentially. |
Insert 10,000 entities. |
Iter | - Create an empty arena. - Insert 10,000 entities. |
Iterate over the arena sequentially. |
IterHalf | - Create an empty arena. - Insert 10,000 entities. - Remove 5,000 entities randomly. |
Iterate over the arena sequentially. |
Remove | - Create an empty arena. - Insert 10,000 entities. |
Remove all entities randomly. |
The lower bound of the 95% confidence interval is used to reduce run to run noise. This likely under-penalizes indirection overhead, so take measurements with a grain of salt.
This includes benchmarks for crates typically considered to be generational arenas. idvec
is
omitted from these charts because some of its operations take a long time, blowing up the scale.
This was run on an AMD 3950x.
This includes benchmarks for things that aren't considered generational arenas. A sperate chart is
available with idvec
removed.