[notkainoa]

Text overlays are 6.7–8.2x cheaper per frame. The text layer reshaped every overlay from scratch on every frame, even when the text hadn't changed. Now it reuses last frame's layout.

Case	Before (p50)	After (p50)	Speedup
1 title with shadow	34.0 µs	5.1 µs	6.7x
Title + paragraph with background + outlined label	197.0 µs	24.1 µs	8.2x

Runs on every frame of GPUI editor playback, scrubbing and export.

What changed

• A layout is reused when the content, font, size, letter spacing, alignment and wrap width all match the previous frame.
• Opacity, position and animations still update every frame. They don't need a reshape.
• Gradient and staggered reveal passes still reshape, as before.

Output is unchanged

• 29,615 rendered frames hash identical to main with perf(rendering): reuse shaped text overlay layouts between frames (6.7-8.2x faster) #2524, perf(rendering): reshape the keyboard overlay only when its text changes (2.8x faster) #2525 and perf(rendering): reshape captions only when their text or style changes (up to 7.6x faster) #2526 applied together. Covers all 14 text animations, text styles, mid-playback edits, keyboard and caption styles, playback, backward scrubbing, random seeks, pauses and resizes. A planted cache-key bug makes thousands of frames differ, so the check catches mistakes.
• New GPU test: a reused layout renders byte-identical pixels to a fresh one (shadow, outline and gradient included), and editing the text drops the old layout.
• A second GPU test changes each shaping input on its own: font size, family, weight, italic, letter spacing, alignment, line height and wrap width. Removing any one of them from the cache key makes it fail.
• cargo test -p cap-rendering --release --lib: 305 passed on current main. The GPU tests pass with --include-ignored.
• cargo clippy -p cap-rendering --all-targets -- -D warnings and cargo fmt are clean.

Benchmark setup

TextLayer::prepare at 1920×1080. p50 over 300 frames after 30 warmup frames. Each number is the median of 3 runs. Apple M1, release build, opt-level 3. "Before" is the same benchmark on main (3d19c5c).

CARGO_PROFILE_RELEASE_OPT_LEVEL=3 cargo test -p cap-rendering --release --lib benchmark_text_layer_prepare -- --ignored --nocapture

Confidence Score: 5/5

The PR appears safe to merge; no blocking issue was found.

Summary

TextLayer now reuses the previous frame’s plain text layouts when all shaping inputs match.

• Text overlays reuse the previous frame’s layout when its shaping inputs match.

No new actionable issues found. GPU tests were inspected but not run.

_{Reviews (2) · Last reviewed commit: "test(rendering): check the text layout c..." · Reviewed by Greptile}

[vercel]

Someone is attempting to deploy a commit to the Cap Software Inc Team on Vercel.

A member of the Team first needs to authorize it.

[richiemcilroy]

@greptileai review

[richiemcilroy]

Thanks for this! The same per-frame reuse landed on main in 164bfef (crates/rendering/src/layers/text.rs, keyed on the full shaping attributes and colours, so the gradient and stagger passes hit the cache too), which is why this now conflicts. Your reused_layouts_follow_each_shaping_input test is more thorough than main's, since it varies every shaping input in turn. Would you be up for a small PR that ports just that test onto main's cache?

[notkainoa]

Opened #2739 with just that test on main's cache @richiemcilroy

[richiemcilroy]

Thanks so much for this contribution, @notkainoa! The same per-frame shaped-text reuse landed on main in 164bfef, so this one is covered. Closing this one, but really appreciate the time you put into it.