doc

fzyzcjy · Oct 16, 2022 · ee39efe · ee39efe
1 parent d9a8335
commit ee39efe
Show file tree

Hide file tree

Showing 3 changed files with 4 additions and 10 deletions.
diff --git a/website/docs/benchmark/analyze/overhead/end-to-end.md b/website/docs/benchmark/analyze/overhead/end-to-end.md
diff --git a/...s/benchmark/analyze/overhead/component.md → ...docs/benchmark/analyze/overhead/result.md b/...s/benchmark/analyze/overhead/component.md → ...docs/benchmark/analyze/overhead/result.md
@@ -1,4 +1,4 @@
-# Result from component
+# Result
 
 :::caution
 
@@ -8,11 +8,9 @@ The `MaybePreemptRender` is slow (needs syscall) now, causing larger overhead. I
 
 Let us discuss the possible overhead components:
 
-1. `MaybePreemptRender` checks - overhead. In each preempt point, we have to check whether it is time to trigger a preempt render. Currently it is slow as it needs one syscall, but it can be (and should be) optimized in https://github.com/fzyzcjy/flutter_smooth/issues/110.
+1. `MaybePreemptRender` checks - overhead. In each preempt point, we have to check whether it is time to trigger a preempt render.
 2. Preempt render - not overhead. To my best knowledge, any solution that wants to achieve 60FPS has to run the build/layout/paint/etc pipeline for the part that wants to be changing. Therefore, it is inevitable to have this component. (Feel free to correct me if you come up with a faster solution without that.) By the way, this part strongly depends on your use case. If you have simple animations this will be fast, while for extremely fancy animations it can be slow.
 
-Thus, the question remains to be, how slow is `MaybePreemptRender` checks?
-
-[TODO]
+Thus, the question remains to be, how slow is `MaybePreemptRender` checks? By using the `overhead.py`, we know the average of one such check is 22.57 microseconds on my machine. In other words, since (by looking at tracing data) we call it roughly once per millisecond, it is **2.3%** overhead (again, the underlying implementation is super slow and will be optimized).
 
 <small>(To reproduce this, create your own tracing data using latest code, instead of using the sample tracing json in this repo - I recorded that prior to writing this section and that data does not include MaybePreemptRender timeline events.)</small>
diff --git a/website/sidebars.js b/website/sidebars.js
@@ -104,8 +104,7 @@ const sidebars = {
                             },
                             items: [
                                 'benchmark/analyze/overhead/definition',
-                                'benchmark/analyze/overhead/component',
-                                'benchmark/analyze/overhead/end-to-end',
+                                'benchmark/analyze/overhead/result',
                             ],
                         },
                         'benchmark/analyze/waste',