From 5673fc8b9c0785537d64b1356e47a4574b6b8852 Mon Sep 17 00:00:00 2001
From: Gunnar Sletta <gunnar.sletta@digia.com>
Date: Fri, 26 Apr 2013 14:54:04 +0200
Subject: [PATCH] Optimize particle mask extruder
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Change-Id: Ideed246f49301f826935d3fe4a25a3b593ee6b07
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
---
src/particles/qquickmaskextruder.cpp | 35 +++++++++++++++++++---------
1 file changed, 24 insertions(+), 11 deletions(-)
diff --git a/src/particles/qquickmaskextruder.cpp b/src/particles/qquickmaskextruder.cpp
index 927cf97006..f8ed042a9b 100644
--- a/src/particles/qquickmaskextruder.cpp
+++ b/src/particles/qquickmaskextruder.cpp
@@ -114,12 +114,18 @@ bool QQuickMaskExtruder::contains(const QRectF &bounds, const QPointF &point)
ensureInitialized(bounds);//###Current usage patterns WILL lead to different bounds/r calls. Separate list?
if (m_img.isNull())
return false;
- QPoint p = point.toPoint() - bounds.topLeft().toPoint();
+
+ QPointF pt = point - bounds.topLeft();
+ QPoint p(pt.x() * m_img.width() / bounds.width(),
+ pt.y() * m_img.height() / bounds.height());
return m_img.rect().contains(p) && (bool)m_img.pixelIndex(p);
}
-void QQuickMaskExtruder::ensureInitialized(const QRectF &r)
+void QQuickMaskExtruder::ensureInitialized(const QRectF &rf)
{
+ // Convert to integer coords to avoid comparing floats and ints which would
+ // often result in rounding errors.
+ QRect r = rf.toRect();
if (m_lastWidth == r.width() && m_lastHeight == r.height())
return;//Same as before
if (!m_pix.isReady())
@@ -129,15 +135,22 @@ void QQuickMaskExtruder::ensureInitialized(const QRectF &r)
m_mask.clear();
- m_img = m_pix.image().createAlphaMask();
- m_img = m_img.convertToFormat(QImage::Format_Mono);//Else LSB, but I think that's easier
- m_img = m_img.scaled(r.size().toSize());//TODO: Do they need aspect ratio stuff? Or tiling?
- int w = m_img.width();
- int h = m_img.height();
- for (int i=0; i<w; i++){
- for (int j=0; j<h; j++){
- if (m_img.pixelIndex(i,j))//Direct bit manipulation is presumably more efficient
- m_mask << QPointF(i,j);
+ m_img = m_pix.image();
+ // Image will in all likelyhood be in this format already, so
+ // no extra memory or conversion takes place
+ if (m_img.format() != QImage::Format_ARGB32 && m_img.format() != QImage::Format_ARGB32_Premultiplied)
+ m_img = m_img.convertToFormat(QImage::Format_ARGB32_Premultiplied);
+
+ // resample on the fly using 16-bit
+ int sx = (m_img.width() << 16) / r.width();
+ int sy = (m_img.height() << 16) / r.height();
+ int w = r.width();
+ int h = r.height();
+ for (int y=0; y<h; ++y) {
+ const uint *sl = (const uint *) m_img.constScanLine((y * sy) >> 16);
+ for (int x=0; x<w; ++x) {
+ if (sl[(x * sx) >> 16] & 0xff000000)
+ m_mask << QPointF(x, y);
}
}
}
--
GitLab