Next set of ClipScrollTree optimizations

[mrobinson]

This seems to cut the prominence of ClipScrollTree updates in my profile a great deal.

---

This change is 

[mrobinson]

Here is the Gecko try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=51369c294686d137fc80ace6ef8032bd5208ad50&group_state=expanded

97% done and no failures.

[glennw]

I'd like to get #2146 landed before we land this, if possible (doing a gecko try now, hopefully it will be ready after that), since they will conflict a bit.

This looks awesome from a quick scan though, it should help with CPU performance a lot!

[glennw]

Reviewed 6 of 6 files at r1.
Review status: all files reviewed at latest revision, 1 unresolved discussion.

---

webrender/src/prim_store.rs, line 1661 at r1 (raw file):

}

fn get_local_bounding_rect(

This uses transform_point3d while the version it's replaced with in util.rs uses transform_point2d. Perhaps that's not actually relevant here?

---

Comments from Reviewable

[glennw]

Just one question, otherwise this looks good to me (it will need a rebase after #2146 lands too).

[bors-servo]

☔ The latest upstream changes (presumably #2146) made this pull request unmergeable. Please resolve the merge conflicts.

[mrobinson]

Review status: all files reviewed at latest revision, 1 unresolved discussion, some commit checks failed.

---

webrender/src/prim_store.rs, line 1661 at r1 (raw file):

Previously, glennw (Glenn Watson) wrote…

This uses transform_point3d while the version it's replaced with in util.rs uses transform_point2d. Perhaps that's not actually relevant here?

I think this should be okay. Both methods are using the same code to calculate the x and y coordinates of the final point.

    #[inline]
    pub fn transform_point3d(&self, p: &TypedPoint3D<T, Src>) -> TypedPoint3D<T, Dst> {
        let x = p.x * self.m11 + p.y * self.m21 + p.z * self.m31 + self.m41;
        let y = p.x * self.m12 + p.y * self.m22 + p.z * self.m32 + self.m42;
        let z = p.x * self.m13 + p.y * self.m23 + p.z * self.m33 + self.m43;
        let w = p.x * self.m14 + p.y * self.m24 + p.z * self.m34 + self.m44;

        point3(x/w, y/w, z/w)
    }

    #[inline]
    pub fn transform_point2d(&self, p: &TypedPoint2D<T, Src>) -> TypedPoint2D<T, Dst> {
        let x = p.x * self.m11 + p.y * self.m21 + self.m41;
        let y = p.x * self.m12 + p.y * self.m22 + self.m42;

        let w = p.x * self.m14 + p.y * self.m24 + self.m44;

        point2(x/w, y/w)
    }

transform_point2d should be slightly more efficient.

---

Comments from Reviewable

[mrobinson]

Okay. I've rebased this branch as well as adding another, very simple, cleanup commit for util.rs.

[glennw]

Nice!

[glennw]

@bors-servo r+

[bors-servo]

📌 Commit 2901a69 has been approved by glennw

[bors-servo]

⌛ Testing commit 2901a69 with merge c76f272...

[bors-servo]

☀️ Test successful - status-appveyor, status-travis
Approved by: glennw
Pushing c76f272 to master...