2052391 - Align tile-invalidation quantize with the rasterizer's pixel rounding

Status: RESOLVED FIXED

(Core :: Graphics: WebRender, task)

Comment 1

[Glenn Watson [:gw]]

Attachment: Bug 2052391 - Align tile-invalidation quantize with the rasterizer's pixel rounding

The picture-cache invalidation stores prim corners quantized to 1/4px via
round(v * 4). Those buckets are centered on the .5 boundary, which is exactly
where the rasterizer's whole-pixel round() flips device pixels. So two corners
that render to different device pixels (e.g. 20.4 -> 20 and 20.5 -> 21) can share
one invalidation key, and a tile is not re-rasterized when it should be, leaving
stale content. This was masked while prims were device-snapped (integer corners
quantize unambiguously); it surfaces once prims carry sub-pixel positions.

Truncate toward zero instead. This places the 1/4px bucket boundaries on every
half-integer with the same sign-aware tie-ownership as round() (half away from
zero), so a quant key uniquely determines the rendered device pixel for both
positive and negative coordinates, while keeping the 1/4px tolerance for
anti-aliased content. Verified: a full sweep over [-300, 300] including exact
half-integers has zero keys mapping to two device pixels (600 with the old
round).

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/2d5d6f15f148
https://hg.mozilla.org/integration/autoland/rev/20aa7c3784ad
Align tile-invalidation quantize with the rasterizer's pixel rounding r=gfx-reviewers,lsalzman

Comment 3

[amarc]

https://hg.mozilla.org/mozilla-central/rev/20aa7c3784ad

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/c26d2b38a389
https://hg.mozilla.org/integration/autoland/rev/c8d58387f362
Revert "Bug 2052391 - Align tile-invalidation quantize with the rasterizer's pixel rounding r=gfx-reviewers,lsalzman" for causing wr failures @ offset-rotate-005.html

Comment 5

[amarc]

Backed out for causing wr failures

• Backout link
• Push with failures
• Failure Log

Comment 6

[rperta]

(In reply to amarc from comment #5)

Backed out for causing wr failures

• Backout link
• Push with failures
• Failure Log

Hi, @Glenn, it also caused a permanent failure on Beta simulations

• Failure line: TEST-UNEXPECTED-FAIL | /_mozilla/mathml/menclose/menclose-2-roundedbox.html | Testing http://web-platform.test:8000/_mozilla/mathml/menclose/menclose-2-roundedbox.html == http://web-platform.test:8000/_mozilla/mathml/menclose/menclose-2-roundedbox-ref.html
• Failure log
• Beta sim push with failures
• How to run these simulations

Comment 7

[Iulian Moraru]

Backout merged to central: https://hg-edge.mozilla.org/mozilla-central/rev/c8d58387f362

Comment 8

[Glenn Watson [:gw]]

Confirmed that the first failure is an extra 1 pixel fuzziness, so updated test annotation for that. The beta failure was not previously fuzzy, but I was able to reproduce the exact same failure on my local machine on main/origin and then confirm it wasn't affected by this patch - so I've added fuzziness annotation for this.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/0e084350ec08
https://hg.mozilla.org/integration/autoland/rev/2795d316fad1
Align tile-invalidation quantize with the rasterizer's pixel rounding r=gfx-reviewers,lsalzman

Comment 10

[Iulian Moraru]

https://hg.mozilla.org/mozilla-central/rev/2795d316fad1