improve caching of worker pool for postcss

[sokra]

No description provided.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Updated
turbo-site	✅ Ready (Inspect)	Visit Preview	Jan 11, 2023 at 1:41PM (UTC)

9 Ignored Deployments

Name	Status	Preview	Updated
examples-basic-web	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-cra-web	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-designsystem-docs	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-kitchensink-blog	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-native-web	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-nonmonorepo	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-svelte-web	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
examples-tailwind-web	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)
turbo-vite-web	⬜️ Ignored (Inspect)		Jan 11, 2023 at 1:41PM (UTC)

[github-actions]

Benchmark for d988bd5

Test	Base	PR	%	Significant %
bench_hydration/Turbopack RSC/1000 modules	2845.43ms ± 3.56ms	2870.70ms ± 7.06ms	+0.89%	+0.14%

Click to view full benchmark

Test	Base	PR	%	Significant %
bench_hmr_to_commit/Turbopack CSR/1000 modules	8604.68µs ± 85.07µs	8568.98µs ± 72.49µs	-0.41%
bench_hmr_to_commit/Turbopack RCC/1000 modules	8944.82µs ± 47.08µs	8870.89µs ± 75.32µs	-0.83%
bench_hmr_to_commit/Turbopack RSC/1000 modules	465.13ms ± 2.61ms	467.97ms ± 1.25ms	+0.61%
bench_hmr_to_commit/Turbopack SSR/1000 modules	8744.28µs ± 94.13µs	8704.96µs ± 62.89µs	-0.45%
bench_hmr_to_eval/Turbopack CSR/1000 modules	7705.61µs ± 62.32µs	7686.00µs ± 55.06µs	-0.25%
bench_hmr_to_eval/Turbopack RCC/1000 modules	7751.06µs ± 75.60µs	7679.81µs ± 77.96µs	-0.92%
bench_hmr_to_eval/Turbopack SSR/1000 modules	7710.11µs ± 74.26µs	7720.03µs ± 63.29µs	+0.13%
bench_hydration/Turbopack RCC/1000 modules	3387.30ms ± 12.46ms	3394.85ms ± 9.93ms	+0.22%
bench_hydration/Turbopack RSC/1000 modules	2845.43ms ± 3.56ms	2870.70ms ± 7.06ms	+0.89%	+0.14%
bench_hydration/Turbopack SSR/1000 modules	2642.86ms ± 11.60ms	2663.15ms ± 12.74ms	+0.77%
bench_startup/Turbopack CSR/1000 modules	1651.34ms ± 5.78ms	1671.11ms ± 7.04ms	+1.20%
bench_startup/Turbopack RCC/1000 modules	2495.57ms ± 6.09ms	2507.99ms ± 9.16ms	+0.50%
bench_startup/Turbopack RSC/1000 modules	2410.78ms ± 9.35ms	2401.30ms ± 8.54ms	-0.39%
bench_startup/Turbopack SSR/1000 modules	2057.16ms ± 2.72ms	2058.22ms ± 8.24ms	+0.05%

[github-actions]

🟢 CI successful 🟢

Thanks

[jridgewell]

Can you explain how this solves PostCSS pool not being reused? These changes look equivalent to the old code.

[sokra]

Can you explain how this solves PostCSS pool not being reused? These changes look equivalent to the old code.

PostCssTransformedAssetVc::process is called for every transformed file. Putting something in a cell will lead to different Vcs (even if the content is the same, turbo engine only cares cell identity). This can be avoided by extracting the logic into a different turbo tasks function which is called with the same arguments (postcss_executor is called with context and postcss config which is often the same across every transformed file). This change leads to postcss_executor being the same Vc for every processed file (if it's affected by the same postcss.config.js). Calling evaluate with the same executor (and other args except for args) will lead to get_evaluate_pool being called with exactly the same arguments, which will lead to pool reuse.

The same problem exists in ModuleOptionsVc::new which is called for every directory. It also called import_map.cell() which leads to different context passed to PostCssTransformVc::new.

[jridgewell]

The same problem exists in ModuleOptionsVc::new which is called for every directory. It also called import_map.cell() which leads to different context passed to PostCssTransformVc::new.

The actionable feedback I'm getting from this is that we should extract code which prepares a cell into a helper function if the current function is too large?