Skip to content

docs: document SVG rasterisation cost as a Known gap (confirmed on two pages) - #272

Merged
tannevaled merged 1 commit into
mainfrom
docs-svg-raster-cost-known-gap
Oct 11, 2026
Merged

tannevaled merged 1 commit into
mainfrom
docs-svg-raster-cost-known-gap

Conversation

@tannevaled

Copy link
Copy Markdown
Contributor

Summary

  • Round 169's own follow-up CPU profiling of foundation.zurb.com (one of the slow pages from the 500-page bench run, confirmed round 168's font-face fix didn't apply here) pinned its ~30s render time to SVG rasterisation: 84% of CPU samples inside rasterx's Draw path, mostly per-pixel gradient-color sampling. DisableJS=true drops the same page to ~6.2s not because JS execution itself is CPU-slow, but because the no-JS render has less SVG content to rasterise than the JS-settled page actually ends up with.
  • This independently confirms an earlier, never-written-down finding from investigating docs.docker.com's own slow render (a decorative dot-pattern.svg background, cost scaling with output area) — the same architectural question, now with two real-world triggers instead of one.
  • Not fixed: both live options (cap output resolution, budget rasterisation time) change rendering output, a decision this session has twice deferred rather than pick unilaterally.
  • Docs-only, no code change.

Test plan

  • N/A (documentation only)

🤖 Generated with Claude Code

…o pages)

Round 169's own follow-up CPU profiling of foundation.zurb.com (one
of the slow pages from the 500-page bench run, confirmed round 168's
font-face fix didn't apply here) pinned its ~30s render time to SVG
rasterisation: 84% of CPU samples inside rasterx's Draw path, mostly
per-pixel gradient-color sampling. DisableJS=true drops the SAME page
to ~6.2s not because JS execution itself is CPU-slow, but because the
no-JS render has less SVG content to rasterise than the JS-settled
page actually ends up with.

This independently confirms an earlier, never-written-down finding
from investigating docs.docker.com's own slow render (a decorative
dot-pattern.svg background, cost scaling with output area) -- the
same architectural question, now with two real-world triggers
instead of one. Not fixed: both live options (cap output resolution,
budget rasterisation time) change rendering output, a decision this
session has twice deferred rather than pick unilaterally.

Also tightens round 168's own "JS-settle-related" phrasing for the
same page now that the actual cause is pinned down precisely.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@tannevaled
tannevaled merged commit 67425a9 into main Oct 11, 2026
7 checks passed
@tannevaled
tannevaled deleted the docs-svg-raster-cost-known-gap branch October 11, 2026 16:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant