fix: perf degradation in devtool plugin

[SyMind]

Summary

Fix perf degradation in #8136

Checklist

• Tests updated (or not required).
• Documentation updated (or not required).

[SyMind]

!bench

[netlify]

✅ Deploy Preview for rspack canceled.

Name	Link
🔨 Latest commit	5bc6002
🔍 Latest deploy log	https://app.netlify.com/sites/rspack/deploys/670fbf12f372500008f03fbf

[rspack-bot]

📝 Benchmark detail: Open

Name	Base (2024-10-16 4ea491e)	Current	Change
10000_development-mode + exec	2.11 s ± 23 ms	2.11 s ± 26 ms	+0.16 %
10000_development-mode_hmr + exec	662 ms ± 17 ms	670 ms ± 14 ms	+1.14 %
10000_production-mode + exec	2.64 s ± 19 ms	2.68 s ± 24 ms	+1.84 %
arco-pro_development-mode + exec	1.79 s ± 88 ms	1.82 s ± 44 ms	+1.71 %
arco-pro_development-mode_hmr + exec	426 ms ± 1.5 ms	427 ms ± 0.83 ms	+0.30 %
arco-pro_production-mode + exec	3.07 s ± 44 ms	3.19 s ± 62 ms	+3.85 %
arco-pro_production-mode_generate-package-json-webpack-plugin + exec	3.13 s ± 84 ms	3.24 s ± 43 ms	+3.41 %
threejs_development-mode_10x + exec	1.61 s ± 22 ms	1.61 s ± 17 ms	-0.11 %
threejs_development-mode_10x_hmr + exec	759 ms ± 9.2 ms	760 ms ± 9.9 ms	+0.17 %
threejs_production-mode_10x + exec	4.9 s ± 23 ms	4.96 s ± 21 ms	+1.34 %

[chenjiahan]

How was this perf degradation discovered?

[SyMind]

@chenjiahan I updated rspack-sources yesterday, and I expected it to cause some performance degradation. However, today I noticed that the extent of the performance degradation was not as expected, so I checked yesterday's PR.

[chenjiahan]

Get 👍

[Narretz]

Can someone explain how the benchmark works? Because I see "current" here, and it has +% almost everywhere. Doesn't it mean the change is slower than before?

[SyMind]

@Narretz The benchmark differs from 4ea491e.

Within this range, there is a specific performance degradation caused by rolling back a particular algorithm in rspack-sources. However, after rolling back the code, I found that the degradation was greater than expected, with a maximum degradation of 6.83%.

I reviewed other commits and discovered another unexpected degradation. This commit is intended to fix that, and then the degradation of this benchmark is as expected.