perf(canvas): AA toggle, GPU sync for benchmarks, AA cost findings#633
Merged
softmarshmallow merged 5 commits intomainfrom Apr 6, 2026
Merged
perf(canvas): AA toggle, GPU sync for benchmarks, AA cost findings#633softmarshmallow merged 5 commits intomainfrom
softmarshmallow merged 5 commits intomainfrom
Conversation
- Renamed scenarios to remove "BUG" prefix and reflect their purpose as no-cache baselines. - Updated comments in the code to clarify the intent of passing `stable=true` for full draw measurements. - Adjusted the Markdown documentation to align with the new scenario naming conventions.
…dings - Add `force_no_aa` field to `RenderPolicy` with `anti_alias()` method, allowing A/B measurement of anti-aliasing cost at different zoom levels. Paint functions (`sk_solid_paint`, `sk_paint_stack`, `sk_paint_stack_without_images`) now accept an `aa: bool` parameter instead of hardcoding `true`. - Add `sync_gpu` config on `RuntimeRendererConfig`. When enabled, `gpu_flush()` calls `flush_submit_and_sync_cpu()` instead of the async `flush_and_submit()`, making per-stage timing in `FrameFlushStats` reflect actual GPU execution cost. Benchmarks (`bench` and `bench-report`) enable this by default. - Add `--no-aa` flag to the bench CLI for A/B AA cost measurement. - Add `skia_bench_subpixel` isolated GPU benchmark measuring AA cost at sub-pixel scale (0.02x zoom). Key finding: AA on sub-pixel geometry is 3.2x more expensive than AA off. - Add `docs/wg/feat-2d/aa-cost-findings.md` documenting the investigation results, including the GPU flush timing discovery. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
6 Skipped Deployments
|
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
WalkthroughAdds anti-aliasing control and GPU sync to the canvas stack, updates paint helper APIs and their call sites, introduces a GPU-timed Skia microbenchmark example, updates benchmark runner/CLI and scenario naming, and adds documentation of AA cost findings. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
rect rgba(200,200,255,0.5)
participant CLI
participant BenchRunner
participant Renderer
participant GPU
end
CLI->>BenchRunner: start benchmark (args, e.g. --no-aa)
BenchRunner->>Renderer: init(policy: force_no_aa = args.no_aa)
BenchRunner->>Renderer: set_sync_gpu(true)
loop per-scenario
BenchRunner->>Renderer: render_frame(pics, zoom, aa=policy.anti_alias())
Renderer->>GPU: submit draw commands
alt sync_gpu = true
Renderer->>GPU: flush_submit_and_sync_cpu() -- blocks until GPU complete
GPU-->>Renderer: completed
else
Renderer->>GPU: flush_and_submit() -- returns immediately
end
Renderer-->>BenchRunner: report elapsed (GPU-synced if enabled)
end
BenchRunner->>CLI: print results (AA on/off, baseline_nocache_zoom)
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested labels
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
# Conflicts: # .agents/skills/cg-perf/SKILL.md
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (6)
crates/grida-canvas/src/painter/paint.rs (2)
27-31: Document the newaaparameter in the function's doc comment.The
# Argumentssection listspaints,size, andimages, but the newly addedaaparameter is not documented. Consider adding it for API clarity.📝 Suggested documentation update
/// # Arguments /// * `paints` - Array of paints to blend together /// * `size` - Container size for paint calculations /// * `images` - Image repository for image paint resolution +/// * `aa` - Whether to enable anti-aliasing on the resulting paint /// /// # Returns🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/src/painter/paint.rs` around lines 27 - 31, The doc comment for sk_paint_stack is missing documentation for the new aa parameter; update the function's Rust doc comment (the /// block above sk_paint_stack) to add aa under the # Arguments section, describing that aa is a boolean that toggles anti-aliasing (e.g., true = enable anti-aliasing, false = disable) and any effects on rendering behavior so callers understand its purpose alongside paints, size, and images.
87-91: Document the newaaparameter for consistency.Similar to
sk_paint_stack, the# Argumentssection should include theaaparameter.📝 Suggested documentation update
/// # Arguments /// * `paints` - Array of paints to blend together (image paints will be ignored) /// * `size` - Container size for paint calculations +/// * `aa` - Whether to enable anti-aliasing on the resulting paint /// /// # Returns🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/src/painter/paint.rs` around lines 87 - 91, The public function sk_paint_stack_without_images has a new boolean parameter aa that isn't documented; update its doc comment (the /// doc block above sk_paint_stack_without_images) to include an entry in the Arguments/# Arguments section describing aa (e.g., "aa: enable/disable anti-aliasing for the returned skia_safe::Paint") matching the style and wording used for sk_paint_stack so docs stay consistent.crates/grida-canvas/src/text/attributed_paragraph.rs (1)
38-48: Text rendering always uses AA regardless of render policy.This helper always passes
truefor anti-aliasing. Unlike shape rendering inpainter.rswhich usesself.policy.anti_alias(), text will always render with AA enabled—even whenforce_no_aais set for benchmarking.If this is intentional (text quality often requires AA), consider adding a comment explaining the decision. If text should also respect the policy, this function would need to accept an
aaparameter.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/src/text/attributed_paragraph.rs` around lines 38 - 48, The resolve_fill_paint helper always forces anti-aliasing by passing `true` to paint_util::sk_paint_stack and paint_util::sk_paint_stack_without_images; update resolve_fill_paint to accept an `aa: bool` parameter (or a descriptive name like `anti_alias`) and forward that boolean to those two calls so text respects rendering policy, then update all callers (text layout/render paths) to pass the renderer's policy value (e.g., self.policy.anti_alias() or force_no_aa inversion) instead of hardcoding true; alternatively, if AA-for-text is intentional, add a clear comment inside resolve_fill_paint explaining why AA is always enabled.docs/wg/feat-2d/aa-cost-findings.md (2)
112-116: Consider using a relative link for internal docs reference.The reference to
docs/wg/feat-2d/optimization.mdcould be a relative link for better navigation within the documentation site:📝 Proposed change
## Related - `crates/grida-canvas/examples/skia_bench/skia_bench_subpixel.rs` — isolated benchmark -- `docs/wg/feat-2d/optimization.md` — master optimization catalog +- [optimization.md](./optimization.md) — master optimization catalog - Chromium pinch-zoom: reduced rasterization quality during interaction🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/wg/feat-2d/aa-cost-findings.md` around lines 112 - 116, Update the internal docs reference to use a relative link instead of an absolute-looking path: replace the `docs/wg/feat-2d/optimization.md` entry in the "Related" list with a relative link such as `./optimization.md` or `../feat-2d/optimization.md` (choose the correct relative path from the current file's location) so the link resolves correctly within the docs site; ensure the link text remains `optimization.md` (or keep the existing descriptive text) and verify it navigates properly after the change.
1-5: Addformat: mdfrontmatter for MDX compatibility.Per coding guidelines, documentation files in
docs/{wg,reference}/**that don't use JSX/MDX features should includeformat: mdfrontmatter to opt out of MDX parsing. This prevents potential build issues with angle brackets or other special characters.📝 Proposed fix
+--- +format: md +--- + # Anti-Aliasing Cost at Sub-Pixel Scale **Date:** 2026-04-07As per coding guidelines: "For MDX compatibility, add
format: mdfrontmatter to documentation files that don't use JSX/MDX features to opt out of MDX parsing and prevent build issues with angle brackets."🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/wg/feat-2d/aa-cost-findings.md` around lines 1 - 5, This markdown document "Anti-Aliasing Cost at Sub-Pixel Scale" is missing MDX opt-out frontmatter; add a top-matter block containing format: md (i.e., a frontmatter section with format: md) to the very top of the file so the MDX parser treats it as plain Markdown and avoids parsing angle-bracket content, ensuring the new frontmatter appears before the existing title/header.crates/grida-canvas/examples/skia_bench/skia_bench_subpixel.rs (1)
135-144: Nitpick: Float equality comparison.Direct float comparison
zoom != 1.0works here because the values are hardcoded literals (1.0 and 0.02), but it's a minor code smell. A threshold comparison would be more defensive:💡 Alternative approach
- if zoom != 1.0 { + if (zoom - 1.0).abs() > f32::EPSILON { canvas.save(); canvas.scale((zoom, zoom)); } for pic in pics { canvas.draw_picture(pic, None, None); } - if zoom != 1.0 { + if (zoom - 1.0).abs() > f32::EPSILON { canvas.restore(); }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/examples/skia_bench/skia_bench_subpixel.rs` around lines 135 - 144, The code uses direct float equality checks `zoom != 1.0` around canvas.save()/scale() and canvas.restore(); change these to an epsilon-based comparison (e.g., `(zoom - 1.0).abs() > 1e-6` or use `f32::EPSILON`/`f64::EPSILON` depending on zoom's type) to defensively detect when zoom meaningfully differs from 1.0 before calling canvas.scale((zoom, zoom)), ensuring both the pre-scale conditional and the matching restore check use the same epsilon comparison.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@crates/grida-canvas/src/painter/painter.rs`:
- Line 350: Helpers are still hardcoding anti-aliasing; update
shadow::draw_drop_shadow, shadow::draw_inner_shadow,
text_stroke::draw_text_stroke, text_stroke::draw_text_stroke_outside_fast_pre to
take an anti_alias: bool (or read PainterPolicy) and call
set_anti_alias(anti_alias) instead of set_anti_alias(true), then propagate that
flag from the caller sites in painter.rs and from VNPainter (add an anti_alias
field or read self.policy.anti_alias()) so all render paths (stroked text,
shadows, vector-network) honor self.policy.anti_alias(); ensure public
signatures and call sites are updated consistently.
In `@crates/grida-canvas/src/runtime/render_policy.rs`:
- Around line 105-107: RenderPolicy's force_no_aa field is not preserved across
the flags bridge; add a bit for force_no_aa to the RenderPolicyFlags
enum/bitflags and update the conversion logic in RenderPolicy::to_flags (or
policy.to_flags()) and RenderPolicy::from_flags so the bit is set when
force_no_aa is true and read back when reconstructing the policy, removing the
hardcoded false that currently appears in the from_flags path.
In `@crates/grida-dev/src/bench/runner.rs`:
- Line 1794: Summary: The serialized "kind" value for ScenarioResult was renamed
from "zoom_forced_stable" to "baseline_nocache_zoom", which is a breaking change
for downstream consumers. Fix: restore backward compatibility by either (a)
reverting the literal back to "zoom_forced_stable" in the assignment that sets
kind, or (b) update ScenarioResult's serde serialization to accept/emit both
names (use serde rename/alias or a custom Serialize/Deserialize impl) so
existing reports still parse; reference the affected symbol "kind" on
ScenarioResult and the serialization logic in report.rs when making the change,
and if the rename is intentional ensure you also bump the package/report format
version and update any CI comparison scripts/tests that depend on the old value.
---
Nitpick comments:
In `@crates/grida-canvas/examples/skia_bench/skia_bench_subpixel.rs`:
- Around line 135-144: The code uses direct float equality checks `zoom != 1.0`
around canvas.save()/scale() and canvas.restore(); change these to an
epsilon-based comparison (e.g., `(zoom - 1.0).abs() > 1e-6` or use
`f32::EPSILON`/`f64::EPSILON` depending on zoom's type) to defensively detect
when zoom meaningfully differs from 1.0 before calling canvas.scale((zoom,
zoom)), ensuring both the pre-scale conditional and the matching restore check
use the same epsilon comparison.
In `@crates/grida-canvas/src/painter/paint.rs`:
- Around line 27-31: The doc comment for sk_paint_stack is missing documentation
for the new aa parameter; update the function's Rust doc comment (the /// block
above sk_paint_stack) to add aa under the # Arguments section, describing that
aa is a boolean that toggles anti-aliasing (e.g., true = enable anti-aliasing,
false = disable) and any effects on rendering behavior so callers understand its
purpose alongside paints, size, and images.
- Around line 87-91: The public function sk_paint_stack_without_images has a new
boolean parameter aa that isn't documented; update its doc comment (the /// doc
block above sk_paint_stack_without_images) to include an entry in the
Arguments/# Arguments section describing aa (e.g., "aa: enable/disable
anti-aliasing for the returned skia_safe::Paint") matching the style and wording
used for sk_paint_stack so docs stay consistent.
In `@crates/grida-canvas/src/text/attributed_paragraph.rs`:
- Around line 38-48: The resolve_fill_paint helper always forces anti-aliasing
by passing `true` to paint_util::sk_paint_stack and
paint_util::sk_paint_stack_without_images; update resolve_fill_paint to accept
an `aa: bool` parameter (or a descriptive name like `anti_alias`) and forward
that boolean to those two calls so text respects rendering policy, then update
all callers (text layout/render paths) to pass the renderer's policy value
(e.g., self.policy.anti_alias() or force_no_aa inversion) instead of hardcoding
true; alternatively, if AA-for-text is intentional, add a clear comment inside
resolve_fill_paint explaining why AA is always enabled.
In `@docs/wg/feat-2d/aa-cost-findings.md`:
- Around line 112-116: Update the internal docs reference to use a relative link
instead of an absolute-looking path: replace the
`docs/wg/feat-2d/optimization.md` entry in the "Related" list with a relative
link such as `./optimization.md` or `../feat-2d/optimization.md` (choose the
correct relative path from the current file's location) so the link resolves
correctly within the docs site; ensure the link text remains `optimization.md`
(or keep the existing descriptive text) and verify it navigates properly after
the change.
- Around line 1-5: This markdown document "Anti-Aliasing Cost at Sub-Pixel
Scale" is missing MDX opt-out frontmatter; add a top-matter block containing
format: md (i.e., a frontmatter section with format: md) to the very top of the
file so the MDX parser treats it as plain Markdown and avoids parsing
angle-bracket content, ensuring the new frontmatter appears before the existing
title/header.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: a91e8ea8-f6e2-4753-b222-59fa3bbe9738
📒 Files selected for processing (16)
.agents/skills/cg-perf/SKILL.mdcrates/grida-canvas/Cargo.tomlcrates/grida-canvas/examples/golden_sk_paints.rscrates/grida-canvas/examples/skia_bench/skia_bench_subpixel.rscrates/grida-canvas/src/cache/paragraph.rscrates/grida-canvas/src/painter/paint.rscrates/grida-canvas/src/painter/painter.rscrates/grida-canvas/src/painter/text_stroke.rscrates/grida-canvas/src/runtime/config.rscrates/grida-canvas/src/runtime/render_policy.rscrates/grida-canvas/src/runtime/scene.rscrates/grida-canvas/src/text/attributed_paragraph.rscrates/grida-canvas/src/vectornetwork/vn_painter.rscrates/grida-dev/src/bench/args.rscrates/grida-dev/src/bench/runner.rsdocs/wg/feat-2d/aa-cost-findings.md
|
|
||
| let mut p = SkPaint::default(); | ||
| p.set_anti_alias(true); | ||
| p.set_anti_alias(self.policy.anti_alias()); |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
fd -i -0 'text_stroke\.rs|shadow\.rs|vn_painter\.rs' | while IFS= read -r -d '' f; do
echo "== $f =="
rg -n -C2 'fn draw_text_stroke|fn draw_text_stroke_outside_fast_pre|fn draw_drop_shadow|fn draw_inner_shadow|struct VNPainter|impl VNPainter|set_anti_alias' "$f"
done
echo "== call sites from painter =="
rg -n -C2 'draw_text_stroke(_outside_fast_pre)?\(|draw_drop_shadow\(|draw_inner_shadow\(|VNPainter::new_with_images|draw_vector_strokes\(' --type=rustRepository: gridaco/grida
Length of output: 10121
🏁 Script executed:
# Get full signatures and first 20 lines of each helper function
echo "=== text_stroke.rs functions ==="
sed -n '12,35p' crates/grida-canvas/src/painter/text_stroke.rs
echo ""
sed -n '96,120p' crates/grida-canvas/src/painter/text_stroke.rs
echo ""
echo "=== shadow.rs draw_drop_shadow ==="
sed -n '40,50p' crates/grida-canvas/src/painter/shadow.rs
echo ""
echo "=== shadow.rs draw_inner_shadow ==="
sed -n '90,100p' crates/grida-canvas/src/painter/shadow.rs
echo ""
echo "=== vn_painter.rs struct and impl ==="
sed -n '28,80p' crates/grida-canvas/src/vectornetwork/vn_painter.rsRepository: gridaco/grida
Length of output: 4478
--no-aa flag is bypassed by helper-driven render paths.
The inline paints in painter.rs now respect self.policy.anti_alias(), but the following helpers still ignore the policy:
shadow::draw_drop_shadow()andshadow::draw_inner_shadow()hardcodeset_anti_alias(true)and accept no AA parametertext_stroke::draw_text_stroke()andtext_stroke::draw_text_stroke_outside_fast_pre()accept no AA parameterVNPainterhas no AA mechanism
This means stroked text, shape shadows, and vector-network rendering will continue using anti-aliasing even when --no-aa is enabled. Update these helpers to accept or read an anti-alias flag from the rendering policy to make the benchmark toggle effective across all render paths.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@crates/grida-canvas/src/painter/painter.rs` at line 350, Helpers are still
hardcoding anti-aliasing; update shadow::draw_drop_shadow,
shadow::draw_inner_shadow, text_stroke::draw_text_stroke,
text_stroke::draw_text_stroke_outside_fast_pre to take an anti_alias: bool (or
read PainterPolicy) and call set_anti_alias(anti_alias) instead of
set_anti_alias(true), then propagate that flag from the caller sites in
painter.rs and from VNPainter (add an anti_alias field or read
self.policy.anti_alias()) so all render paths (stroked text, shadows,
vector-network) honor self.policy.anti_alias(); ensure public signatures and
call sites are updated consistently.
| results.push(ScenarioResult { | ||
| name: fss.name.to_string(), | ||
| kind: "zoom_forced_stable".to_string(), | ||
| kind: "baseline_nocache_zoom".to_string(), |
There was a problem hiding this comment.
Breaking change: kind field value renamed.
The kind field value changed from "zoom_forced_stable" to "baseline_nocache_zoom". Since ScenarioResult derives serde::Serialize (per report.rs:32-43), this is a breaking change for any downstream consumers—CI comparison scripts, report formatters, or dashboards—that match on this string value.
Consider whether this warrants a version bump or migration notice if existing benchmark reports are being compared across commits.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@crates/grida-dev/src/bench/runner.rs` at line 1794, Summary: The serialized
"kind" value for ScenarioResult was renamed from "zoom_forced_stable" to
"baseline_nocache_zoom", which is a breaking change for downstream consumers.
Fix: restore backward compatibility by either (a) reverting the literal back to
"zoom_forced_stable" in the assignment that sets kind, or (b) update
ScenarioResult's serde serialization to accept/emit both names (use serde
rename/alias or a custom Serialize/Deserialize impl) so existing reports still
parse; reference the affected symbol "kind" on ScenarioResult and the
serialization logic in report.rs when making the change, and if the rename is
intentional ensure you also bump the package/report format version and update
any CI comparison scripts/tests that depend on the old value.
There was a problem hiding this comment.
🧹 Nitpick comments (1)
crates/grida-canvas/src/runtime/scene.rs (1)
1540-1542: Consider documentingset_sync_gpuas benchmark-oriented API.A short doc comment here would help prevent accidental enabling in interactive rendering paths.
Suggested doc tweak
pub fn set_sync_gpu(&mut self, sync: bool) { + // Benchmark/timing mode: when true, flushes block until GPU completion. + // Keep disabled for normal interactive rendering. self.config.sync_gpu = sync; }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/grida-canvas/src/runtime/scene.rs` around lines 1540 - 1542, Add a short doc comment above the set_sync_gpu method explaining this API is intended for benchmarking only and should not be enabled in interactive rendering paths; mention that it toggles self.config.sync_gpu and can degrade real-time performance or block GPU work, so callers (tests/benchmarks) should use it rather than production UI code.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@crates/grida-canvas/src/runtime/scene.rs`:
- Around line 1540-1542: Add a short doc comment above the set_sync_gpu method
explaining this API is intended for benchmarking only and should not be enabled
in interactive rendering paths; mention that it toggles self.config.sync_gpu and
can degrade real-time performance or block GPU work, so callers
(tests/benchmarks) should use it rather than production UI code.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c413b52c-c7f5-43c0-a8ba-cc072f0d76f2
📒 Files selected for processing (3)
.agents/skills/cg-perf/SKILL.mdcrates/grida-canvas/Cargo.tomlcrates/grida-canvas/src/runtime/scene.rs
✅ Files skipped from review due to trivial changes (1)
- crates/grida-canvas/Cargo.toml
🚧 Files skipped from review as they are similar to previous changes (1)
- .agents/skills/cg-perf/SKILL.md
Add FLAG_FORCE_NO_AA to the RenderPolicyFlags bitmap so that from_flags/to_flags round-trips preserve the AA override across the WASM/host boundary. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
force_no_aafield +anti_alias()method. Paint functions acceptaa: boolinstead of hardcodingtrue. Bench CLI exposes--no-aafor A/B measurement.sync_gpuconfig makesgpu_flush()block until GPU completes, fixing per-stage timing accuracy (mid_flush_uswas measuring command submission, not GPU execution).skia_bench_subpixelmeasures AA cost at sub-pixel scale. Finding: AA on sub-pixel geometry is 3.2x more expensive than AA off.docs/wg/feat-2d/aa-cost-findings.mdwith benchmark data and optimization implications.Test plan
cargo test -p cg— 33 passedcargo check -p cg -p grida-canvas-wasm -p grida-dev— all compilecargo clippy -p cg --no-deps— no new warnings--no-aaproduces measurable difference on isolated bench (3.2x at 40K nodes, 0.02x zoom)🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Documentation