Due to floating point inaccuracy, when intersecting edges, the
intersection point may fall above one of the edges' top vertices
or below one of the bottom vertices. In these cases, we were simply
splitting one edge on the relevant endpoint of the other edge. This
is incorrect if the intersection is far from the endpoint (e.g.,
the test case in the linked bug, where one of the intersected edges
is near-horizontal but the intersection falls below both of its
endpoints, in the middle of the edge.)
The correct solution is to split both edges as normal, and take care
to produce edges with the correct ordering where the intersection is
above or below an edge. However, since the new vertex may be above
the current vertex, simply restarting intersection checks at the
current vertex won't work. We need to process the intersection
vertex before the current one.
This introduces another problem: unlike all other splitting modes
(which always shorten edges), splitting an edge above the top or
below the bottom can lengthen it, causing it to violate the AEL
with an adjacent edge which then shortens it back to the original
point (in cleanup_active_edges()). Since the splitting and merging
code can't agree, we loop forever.
Instead of simply fusing neighboring edges in cleanup_active_edges(),
the proper fix to this problem is to detect the AEL violation and
rewind all processing to the vertex above it. For performance, we
only rewind when we detect that a split edge is no longer ordered
within the mesh (merge_enclosing_edges()) or within the the AEL
(rewind_if_necessary()). We also store the enclosing edges of each
vertex, which allows us to rewind quickly, since we know exactly which
edges need to be added/removed from the AEL.
cleanup_active_edges(), fix_active_state() and Vertex::fProcessed have
been removed. In their place are rewind_active_edges() and
rewind_if_necessary(), which uses the same logic as
cleanup_active_edges() but uses it to know when to rewind.
Bug: skia:5026
Change-Id: I3638a429f5428498d6df6bb7b98c67374dc291aa
Reviewed-on: https://skia-review.googlesource.com/18900
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
When sanitizing contours, if the first and last vertices coincide,
continue with the previous vertex, not the next vertex, since we
may otherwise exit prematurely. Also, round the last vertex before
entering the loop, just in case it coincides with the first.
Add a test case to exercise the above, and another one which exercises
the intruding-vertex workaround.
BUG=691593
Change-Id: Ic28a9308a21164d185edef0ee6fbc29b40742149
Reviewed-on: https://skia-review.googlesource.com/8364
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Should be no user- or test-visible changes.
BUG=skia:
Change-Id: I6499dc978a41fee344b847c118f84227271561c5
Reviewed-on: https://skia-review.googlesource.com/6906
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Stephan White <senorblanco@chromium.org>
Use quads rather than triangles for the edge geometry. This allows
us to perform a simpler edge categorization (see below). It also
improves performance by reducing the number of edges processed during
the simplify and tessellate steps.
Label AA edges as three types: inner, outer, and connector. This
results in correct alpha values for intersected edges, even when
the top or bottom vertex has been merged with a vertex on edges
of different types.
Changed the "collinear edges" sample from the concavepaths GM for a
"fast-foward" shape, which more clearly shows the problem being fixed
here. (The collinearity from the "collinear edges" was actually being
removed earlier up the stack, causing the path to become convex and
not exercise the concave path renderers anyway.)
NOTE: this will cause changes in the "concavepaths" GM results, and
minor pixel diffs in a number of other tests.
Change-Id: I6c2b0cdb35cda42b01cf1100621271fef5be35b0
Reviewed-on: https://skia-review.googlesource.com/6430
Reviewed-by: Stephan White <senorblanco@chromium.org>
Commit-Queue: Stephan White <senorblanco@chromium.org>
This reverts commit d4b2155248.
Reason for revert: accidentally added some unwanted changes
Original change's description:
> Quality and performance fixes for AA tessellating path renderer.
>
> Use quads rather than triangles for the edge geometry. This allows
> us to perform a simpler edge categorization (see below). It also
> improves performance by reducing the number of edges processed during
> the simplify and tessellate steps.
>
> Label AA edges as three types: inner, outer, and connector. This
> results in correct alpha values for intersected edges, even when
> the top or bottom vertex has been merged with a vertex on edges
> of different types.
>
> Changed the "collinear edges" sample from the concavepaths GM for a
> "fast-foward" shape, which more clearly shows the problem being fixed
> here. (The collinearity from the "collinear edges" was actually being
> removed earlier up the stack, causing the path to become convex and
> not exercise the concave path renderers anyway.)
>
> NOTE: this will cause changes in the "concavepaths" GM results, and
> minor pixel diffs in a number of other tests.
>
> BUG=660893
>
> Change-Id: Ide49374d6d173404c7223f7316dd439df1435787
> Reviewed-on: https://skia-review.googlesource.com/6427
> Commit-Queue: Stephan White <senorblanco@chromium.org>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
>
TBR=bsalomon@google.com,senorblanco@chromium.org,reviews@skia.org
BUG=660893
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I06a36e397645bfc42442a5a9e7c27328f6048ab9
Reviewed-on: https://skia-review.googlesource.com/6428
Commit-Queue: Stephan White <senorblanco@chromium.org>
Reviewed-by: Stephan White <senorblanco@chromium.org>
Use quads rather than triangles for the edge geometry. This allows
us to perform a simpler edge categorization (see below). It also
improves performance by reducing the number of edges processed during
the simplify and tessellate steps.
Label AA edges as three types: inner, outer, and connector. This
results in correct alpha values for intersected edges, even when
the top or bottom vertex has been merged with a vertex on edges
of different types.
Changed the "collinear edges" sample from the concavepaths GM for a
"fast-foward" shape, which more clearly shows the problem being fixed
here. (The collinearity from the "collinear edges" was actually being
removed earlier up the stack, causing the path to become convex and
not exercise the concave path renderers anyway.)
NOTE: this will cause changes in the "concavepaths" GM results, and
minor pixel diffs in a number of other tests.
BUG=660893
Change-Id: Ide49374d6d173404c7223f7316dd439df1435787
Reviewed-on: https://skia-review.googlesource.com/6427
Commit-Queue: Stephan White <senorblanco@chromium.org>
Reviewed-by: Brian Salomon <bsalomon@google.com>
The vertices which are produced by stage 5 of the
tesselator are copied into the Polys and MonotonePolys it
produces. This is necessary because each vertex may have an
arbitrary valence, since it may participate in an arbitrary
number of Polys, so we can't use the vertex's prev/next
pointers to represent all the Monotones of which this
vertex may be a member.
However, each Edge can only be a member of two Polys (one
on each side of the edge). So by adding two prev/next
pointer pairs to each Edge, we can represent each Monotone
as a list of edges instead. Then we no longer need to copy
the vertices.
One wrinkle is that the ear-clipping stage (6) of the
tessellator does require prev/next pointers, in order to
remove vertices as their ears are clipped. So we convert
the edge list into a vertex list during Monotone::emit(),
using the prev/next pointers temporarily for that monotone.
This change improves performance by 7-20% on a non-caching
version of the tessellator, and reduces memory use.
Other notes:
1) Polys are initially constructed empty (no edges), but
with the top vertex, which is needed for splitting
Polys. Edges are added to Polys only after their bottom
vertex is seen.
2) MonotonePolys are always constructed with one edge, so
we always know their handedness (left/right).
MonotonePoly::addEdge() no longer detects when a monotone
is "done" (edge of opposite handedness); this is handled
by Poly::addEdge(), so MonotonePoly::addEdge() has no
return value.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2029243002
Review-Url: https://codereview.chromium.org/2029243002
This path renderer converts paths to linear contours, resolves intersections via Bentley-Ottman, implements a trapezoidal decomposition a la Fournier and Montuno to produce triangles, and renders those with a single draw call. It does not currently do antialiasing, so it must be used in conjunction with multisampling.
A fair amount of the code is to handle floating point edge cases in intersections. Rather than perform exact computations (which would require arbitrary precision arithmetic), we reconnect the mesh to reflect the intersection points. For example, intersections can occur above the current vertex, and force edges to be merged into the current vertex, requiring a restart of the intersections. Splitting edges for intersections can also force them to merge with formerly-distinct edges in the same polygon, or to violate the ordering of the active edge list, or the active edge state of split edges.
BUG=skia:
Review URL: https://codereview.chromium.org/855513004
