When testing Compose 1.7.0-alpha02 (Kotlin 2.0.20-...

# compose-web

[Robert Jaros]

When testing Compose 1.7.0-alpha02 (Kotlin 2.0.20-RC2) with web targets (both js and wasm) I see noticeable performance degradation compared to Compose 1.6.11 (Kotlin 2.0.0) (~15-20%) when running benchmarks with lot of components. Is this to be expected?

[Michael Paus]

Why did you compare this old version and not the current 1.7.0-alpha03 one?

[Robert Jaros]

03 was released just now

[Robert Jaros]

I'll try with it in a moment

[Robert Jaros]

There is no difference between alpha02 and alpha03

[Robert Jaros]

I've re-run my benchmark with compose 1.6.11 and 1.7.0-alpha01 to 03, all using Kotlin 2.0.20-RC2. The results are (lower is better):
1.6.11 - 5.19
1.7.0-alpha01 - 5.42
1.7.0-alpha02 - 6.36
1.7.0-alpha03 - 6.42
There is apparently some major regression between alpha 01 and 02. That's a bit disappointing and I'm really interested what is wrong.

[Robert Jaros]

In the above test I've used Compose HTML targeting JS. But I have similar problem with Kilua targeting WasmJS.

[Oleksandr Karpovich [JB]]

Could you please share your benchmark?

[Robert Jaros]

I'm using https://github.com/krausest/js-framework-benchmark

[Robert Jaros]

I'm building a simple compose html application and adding the resulting code to the frameworks dir.

[Robert Jaros]

Are you interested in the sources of my app or distribution js to run the benchmark?

[Oleksandr Karpovich [JB]]

Sources of your app might be helpful

[Robert Jaros]

The code is not really mine. It was posted some time ago in this slack by someone. I've just optimized it a bit.
https://github.com/rjaros/compose-html-js-framework-benchmark/tree/main