From b183774063fcebf1438ffff56a8c3d33a8c9e3bb Mon Sep 17 00:00:00 2001
From: Dominik Honnef <dominik@honnef.co>
Date: Thu, 15 Jun 2023 01:31:40 +0200
Subject: [PATCH] internal/stroke, gpu: reuse slice when splitting cubics

When building GPU vertices from paths, we call stroke.SplitCubic once
per OpCubic. Before this change, each call to stroke.SplitCubic would
allocate a slice, which we would only use to iterate over.

This allocation can be easily avoided by reusing the slice. We can
conveniently store it in gpu.quadSplitter.

In a real application that renders hundreds of paths with tens of
rounded rectangles per path, this saved roughly 4500 allocations (or 1
MB worth) per frame.

Signed-off-by: Dominik Honnef <dominik@honnef.co>
---
 gpu/clip.go               | 3 +++
 gpu/gpu.go                | 4 +++-
 internal/stroke/stroke.go | 8 +++++---
 3 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/gpu/clip.go b/gpu/clip.go
index 292a6c04..3ba61d00 100644
--- a/gpu/clip.go
+++ b/gpu/clip.go
@@ -12,6 +12,9 @@ type quadSplitter struct {
 	bounds  f32.Rectangle
 	contour uint32
 	d       *drawOps
+
+	// scratch space used by calls to stroke.SplitCubic
+	scratch []stroke.QuadSegment
 }
 
 func encodeQuadTo(data []byte, meta uint32, from, ctrl, to f32.Point) {
diff --git a/gpu/gpu.go b/gpu/gpu.go
index 1b92ccee..ab850fa4 100644
--- a/gpu/gpu.go
+++ b/gpu/gpu.go
@@ -1327,7 +1327,9 @@ func decodeToOutlineQuads(qs *quadSplitter, tr f32.Affine2D, pathData []byte) {
 			q = q.Transform(tr)
 			qs.splitAndEncode(q)
 		case scene.OpCubic:
-			for _, q := range stroke.SplitCubic(scene.DecodeCubic(cmd)) {
+			from, ctrl0, ctrl1, to := scene.DecodeCubic(cmd)
+			qs.scratch = stroke.SplitCubic(from, ctrl0, ctrl1, to, qs.scratch[:0])
+			for _, q := range qs.scratch {
 				q = q.Transform(tr)
 				qs.splitAndEncode(q)
 			}
diff --git a/internal/stroke/stroke.go b/internal/stroke/stroke.go
index cc18be73..daf472d2 100644
--- a/internal/stroke/stroke.go
+++ b/internal/stroke/stroke.go
@@ -621,6 +621,7 @@ func StrokePathCommands(style StrokeStyle, scene []byte) StrokeQuads {
 // decodeToStrokeQuads decodes scene commands to quads ready to stroke.
 func decodeToStrokeQuads(pathData []byte) StrokeQuads {
 	quads := make(StrokeQuads, 0, 2*len(pathData)/(scene.CommandSize+4))
+	scratch := make([]QuadSegment, 0, 10)
 	for len(pathData) >= scene.CommandSize+4 {
 		contour := binary.LittleEndian.Uint32(pathData)
 		cmd := ops.DecodeCommand(pathData[4:])
@@ -645,7 +646,9 @@ func decodeToStrokeQuads(pathData []byte) StrokeQuads {
 			}
 			quads = append(quads, quad)
 		case scene.OpCubic:
-			for _, q := range SplitCubic(scene.DecodeCubic(cmd)) {
+			from, ctrl0, ctrl1, to := scene.DecodeCubic(cmd)
+			scratch = SplitCubic(from, ctrl0, ctrl1, to, scratch[:0])
+			for _, q := range scratch {
 				quad := StrokeQuad{
 					Contour: contour,
 					Quad:    q,
@@ -660,8 +663,7 @@ func decodeToStrokeQuads(pathData []byte) StrokeQuads {
 	return quads
 }
 
-func SplitCubic(from, ctrl0, ctrl1, to f32.Point) []QuadSegment {
-	quads := make([]QuadSegment, 0, 10)
+func SplitCubic(from, ctrl0, ctrl1, to f32.Point, quads []QuadSegment) []QuadSegment {
 	// Set the maximum distance proportionally to the longest side
 	// of the bounding rectangle.
 	hull := f32.Rectangle{