This PR will add a few small alterations to the new constileation provider, and more importantly adds some new benchmarks using BenchmarkDotNet to show some stats around each of the providers available. These can be used in the documentation, instead of needing to include in the changelog.
Below are some of the measurements ran on my own PC, of which are only some basic benchmarks thus far. I'm sure people are heavily opinionated, and I invite them to also contribute ;)
Update: if there is a certain bench we need, please do suggest it. i've refrained from adding too many more at this point until we know of any specific requirements/requests.
Bench: how fast each provider can call .active, for every tile in a small world
Method
Mean
Error
StdDev
Ratio
RatioSD
Active_Stock
31.78 ms
0.456 ms
0.426 ms
1.00
0.00
Active_Heap
122.07 ms
2.413 ms
2.370 ms
3.84
0.09
Active_Constileation
40.41 ms
0.575 ms
0.510 ms
1.27
0.02
Active_1d
52.94 ms
0.731 ms
0.648 ms
1.66
0.03
Active_2d
54.96 ms
1.083 ms
1.064 ms
1.73
0.05
Active_Struct
61.59 ms
0.651 ms
0.544 ms
1.93
0.03
Bench: how fast the provider can issue and store its own data, for every tile in a small world
Method
Mean
Error
StdDev
Ratio
RatioSD
AssignFromSelf_Stock
34.30 ms
0.486 ms
0.454 ms
1.00
0.00
AssignFromSelf_Heap
267.03 ms
3.455 ms
3.062 ms
7.79
0.15
AssignFromSelf_Constileation
42.30 ms
0.664 ms
0.621 ms
1.23
0.03
AssignFromSelf_1d
208.46 ms
2.823 ms
2.640 ms
6.08
0.10
AssignFromSelf_2d
242.42 ms
4.669 ms
6.234 ms
7.11
0.24
AssignFromSelf_Struct
278.45 ms
4.580 ms
4.284 ms
8.12
0.13
Bench: how fast the provider can translate other tile data, such as Terraria.Tile, for every tile in a small world
Method
Mean
Error
StdDev
Ratio
AssignFromTile_Stock
433.3 ms
7.22 ms
6.40 ms
1.00
AssignFromTile_Heap
210.4 ms
3.33 ms
3.11 ms
0.49
AssignFromTile_Constileation
195.7 ms
2.45 ms
2.29 ms
0.45
AssignFromTile_1d
239.0 ms
2.07 ms
1.83 ms
0.55
AssignFromTile_2d
240.5 ms
3.03 ms
2.69 ms
0.56
AssignFromTile_Struct
277.6 ms
3.74 ms
3.49 ms
0.64
Bench: how fast a provider can clear tile data (using methods), for every tile in a small world
Method
Mean
Error
StdDev
Ratio
RatioSD
Clear_Stock
308.8 ms
3.90 ms
3.65 ms
1.00
0.00
Clear_Heap
479.1 ms
9.20 ms
9.84 ms
1.55
0.03
Clear_Constileation
331.2 ms
2.16 ms
1.92 ms
1.07
0.01
Clear_1d
557.8 ms
3.24 ms
3.03 ms
1.81
0.03
Clear_2d
692.9 ms
20.61 ms
59.78 ms
2.14
0.16
Clear_Struct
789.2 ms
13.95 ms
22.93 ms
2.60
0.11
Bench: how fast a provider can run similar logic to the clear world function found in vanilla, for every tile in a small world
Method
Mean
Error
StdDev
Ratio
RatioSD
ClearWorld_Stock
79.27 ms
0.669 ms
0.593 ms
1.00
0.00
ClearWorld_Heap
239.87 ms
2.324 ms
2.060 ms
3.03
0.03
ClearWorld_Constileation
62.48 ms
0.948 ms
0.887 ms
0.79
0.01
ClearWorld_1d
139.73 ms
1.621 ms
1.516 ms
1.76
0.02
ClearWorld_2d
151.27 ms
2.341 ms
2.190 ms
1.91
0.03
ClearWorld_Struct
177.16 ms
3.190 ms
2.828 ms
2.23
0.05
Bench: how fast a provider can issue tile data, for every tile in a small world
Method
Mean
Error
StdDev
Ratio
RatioSD
Gets_Stock
17.49 ms
0.224 ms
0.210 ms
1.00
0.00
Gets_Heap
100.88 ms
2.007 ms
2.465 ms
5.76
0.17
Gets_Constileation
27.15 ms
0.513 ms
0.455 ms
1.55
0.03
Gets_1d
29.46 ms
0.431 ms
0.403 ms
1.68
0.03
Gets_2d
27.93 ms
0.258 ms
0.242 ms
1.60
0.02
Gets_Struct
33.77 ms
0.671 ms
1.719 ms
1.85
0.08
Bench: how fast a provider can change the type of a tile, for every tile in a small world
Method
Mean
Error
StdDev
Ratio
RatioSD
Type_Stock
28.41 ms
0.420 ms
0.393 ms
1.00
0.00
Type_Heap
120.99 ms
2.317 ms
2.480 ms
4.25
0.10
Type_Constileation
36.95 ms
0.732 ms
0.977 ms
1.30
0.04
Type_1d
38.65 ms
0.686 ms
1.027 ms
1.38
0.04
Type_2d
39.64 ms
0.405 ms
0.359 ms
1.40
0.02
Type_Struct
43.09 ms
0.749 ms
0.664 ms
1.52
0.03
Bench: test how fast a provider can call basic actions, like clearing, type change and .active, for every tile in a small world
hakusaro
commented
1 year ago
This PR will add a few small alterations to the new constileation provider, and more importantly adds some new benchmarks using BenchmarkDotNet to show some stats around each of the providers available. These can be used in the documentation, instead of needing to include in the changelog.
Below are some of the measurements ran on my own PC, of which are only some basic benchmarks thus far. I'm sure people are heavily opinionated, and I invite them to also contribute ;)
Update: if there is a certain bench we need, please do suggest it. i've refrained from adding too many more at this point until we know of any specific requirements/requests.
Great job, you read this far 🥇