Description

Replaces uses of .forEach() with regular for loops. This has significant impact on performance in hot paths such as Component.updateTree() and Component.renderTree().

In the graph above, you see 50 runs of benchmark/update_components_benchmark.dart. The forEach results are blue, the for-loop results are green. I could see this effect after just replacing the forEach calls in component.dart. Data here.

Aside for posterity: for i in {1..50}; do flutter test benchmark/main.dart --no-pub -r silent 2>/dev/null >> benchmarks_for_loop.txt; done, then get the data from the text file.

I went ahead and replaced additional forEach calls elsewhere in the engine codebase, but there was no additional effect on the benchmark. Still, I kept those changes in. I only replaced forEach in places that seemed relatively hot (e.g. ComponentSet.reorder()). There are more forEach calls in the codebase but those seem fine to me as they aren't likely to be called too often.

It should be noted that I needed to update the benchmark to add children to the components. Every _BenchmarkComponent now has 10 children. This feels a bit more realistic use of the framework than having a flat array of components with no children. By changing the benchmark code in this way, I made it a bit slower, so I'm not sure if the effect will be seen in the CI/CD.

I also tried whether the change will have effect on my game's benchmark (which is a lot more involved and uses flutter driver to test the whole game in AOT mode). For the game, the effect is negligible but that was kind of expected since my game spends a significant amount of its CPU time on AI, raycasting, smoke simulation and drawVertices, none of which really depend on the speed of the engine update() mechanism.

Checklist

• I have followed the Contributor Guide when preparing my PR.
• I have updated/added tests for ALL new/updated/fixed functionality.
• I have updated/added relevant documentation in docs and added dartdoc comments with ///.
• I have updated/added relevant examples in examples or docs.

Breaking Change?

• Yes, this PR is a breaking change.
• No, this PR is not a breaking change.

Related Issues