Refactor calculations in SkEdge to be more understandable
I've parsed some of the mysteries in the approximation code
but otherwise given up for now, leaving some notes about the
best I can figure this out.
I (re)learned while writing this that shifting negative numbers
left by 2 behaves different than dividing by 4 (shift rounds
towards negative infinity than towards 0). Dividing by 4
involves a few more assembly instructions *and* causes differences
in the diffs (because we might choose to approximate some quads with
4 instead of 2 segments. [1]
[1] https://godbolt.org/z/qx7oEYPcY
Change-Id: I62832abf9c5a2ca1339cc045b32182f82d8ae28f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/976996
Reviewed-by: Daniel Dilan <danieldilan@google.com>
1 file changed
