Smarter opaque/transparent splits and transformed AA

[kvark]

Related to #1384, #1797, #1297
Also TODO: Gecko try push https://treeherder.mozilla.org/#/jobs?repo=try&revision=e4e8ff793bcf938c2cb9ca414368f4bc808a66ad

Currently blocked on the rotated-image.yaml breakage, investigation is ongoing - solved!

---

This change is 

[glennw]

Reviewed 21 of 21 files at r1.
Review status: all files reviewed at latest revision, 5 unresolved discussions.

---

webrender/res/prim_shared.glsl, line 772 at r1 (raw file):

Rectangle fetch_rectangle(int address) {
    vec4 data[2] = fetch_from_resource_cache_2(address);
    vec4 mask = vec4((int(data[1].x) & ivec4(1,2,4,8)) != ivec4(0));

Neat!

---

webrender/src/frame.rs, line 46 at r1 (raw file):

    pipeline_epochs: Vec<(PipelineId, Epoch)>,
    replacements: Vec<(ClipId, ClipId)>,
    opaque_parts: Vec<LayoutRect>,

Perhaps add a comment here saying that these are part of the context to avoid reallocations?

---

webrender/src/frame.rs, line 638 at r1 (raw file):

    /// the rounded parts of the clip. This is pretty simple now, so has a lot of
    /// potential for further optimizations.
    fn try_to_add_rectangle_splitting_on_clip(

It might be worth only calling this function if the size of the rectangle is over some arbitrary threshold (maybe 256?).

---

webrender/src/util.rs, line 305 at r1 (raw file):

    }

    fn split_rectangles(

This seems like it would still result in t-junctions if the radii are not the same at each corner? I might just be not reading this carefully enough...

---

webrender_api/src/display_item.rs, line 42 at r1 (raw file):

}

bitflags! {

NOTE: Once we make the primitive / clip segments work internally, this shouldn't need to be in the public API, and we can move it from the api crate to the main crate. But we need it here for now since we're actually creating new primitives when splitting via the public API.

---

Comments from Reviewable

[kvark]

Review points are addressed now, and the try push looks green.

---

Review status: 20 of 21 files reviewed at latest revision, 5 unresolved discussions.

---

webrender/src/frame.rs, line 46 at r1 (raw file):

Previously, glennw (Glenn Watson) wrote…

Perhaps add a comment here saying that these are part of the context to avoid reallocations?

fixed

---

webrender/src/frame.rs, line 638 at r1 (raw file):

Previously, glennw (Glenn Watson) wrote…

It might be worth only calling this function if the size of the rectangle is over some arbitrary threshold (maybe 256?).

good idea, added now

---

webrender/src/util.rs, line 305 at r1 (raw file):

Previously, glennw (Glenn Watson) wrote…

This seems like it would still result in t-junctions if the radii are not the same at each corner? I might just be not reading this carefully enough...

this implementation is sub-optimal (still, even if much better than the old one) for the fact it works off the inner rectangle that is conservative: it takes the biggest of the radii affecting each side of the produced rectangle, so we just waste some space if one of the corners have lesser radii than the others.
My plan was to leave this for further optimization, given that the framework is set (framework of producing the list of opaque/transparent parts).

---

webrender_api/src/display_item.rs, line 42 at r1 (raw file):

Previously, glennw (Glenn Watson) wrote…

NOTE: Once we make the primitive / clip segments work internally, this shouldn't need to be in the public API, and we can move it from the api crate to the main crate. But we need it here for now since we're actually creating new primitives when splitting via the public API.

good point!

---

Comments from Reviewable

[glennw]

Yep, it's definitely an improvement. We can look into the t-junction with non uniform radii issue as a follow up. r=me once we have a green gecko try run.

[kvark]

Thanks, it is all green.
@bors-servo r+

[bors-servo]

📌 Commit 98b0991 has been approved by kvark

[bors-servo]

⌛ Testing commit 98b0991 with merge 6de1b7c...

[kvark]

@bors-servo r-
@bors-servo r=glennw

[bors-servo]

💡 This pull request was already approved, no need to approve it again.

• This pull request is currently being tested. If there's no response from the continuous integration service, you may use retry to trigger a build again.

[bors-servo]

📌 Commit 98b0991 has been approved by glennw

[bors-servo]

⌛ Testing commit 98b0991 with merge 8e4f5c1...

[bors-servo]

☀️ Test successful - status-appveyor, status-travis
Approved by: glennw
Pushing 8e4f5c1 to master...