Changeset 561 for trunk/src/gui/painting/qdrawhelper.cpp

Timestamp:

Feb 11, 2010, 11:19:06 PM (15 years ago)

Author:

Dmitry A. Kuminov

Message:

trunk: Merged in qt 4.6.1 sources.

Location:

trunk

Files:

• . (modified) (1 prop)
• src/gui/painting/qdrawhelper.cpp (modified) (188 diffs)

trunk/src/gui/painting/qdrawhelper.cpp

@@ -2 +2 @@
 **
 ** Copyright (C) 2009 Nokia Corporation and/or its subsidiary(-ies).
-** Contact: Qt Software Information (qt-info@nokia.com)
+** All rights reserved.
+** Contact: Nokia Corporation (qt-info@nokia.com)
 **
 ** This file is part of the QtGui module of the Qt Toolkit.
@@ -21 +22 @@
 ** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.
 **
-** In addition, as a special exception, Nokia gives you certain
-** additional rights. These rights are described in the Nokia Qt LGPL
-** Exception version 1.0, included in the file LGPL_EXCEPTION.txt in this
-** package.
+** In addition, as a special exception, Nokia gives you certain additional
+** rights.  These rights are described in the Nokia Qt LGPL Exception
+** version 1.1, included in the file LGPL_EXCEPTION.txt in this package.
 **
 ** GNU General Public License Usage
@@ -34 +34 @@
 ** met: http://www.gnu.org/copyleft/gpl.html.
 **
-** If you are unsure which license is appropriate for your use, please
-** contact the sales department at qt-sales@nokia.com.
+** If you have questions regarding the use of this file, please contact
+** Nokia at qt-info@nokia.com.
 ** $QT_END_LICENSE$
 **
 ****************************************************************************/
+
 #include <private/qdrawhelper_p.h>
 #include <private/qpaintengine_raster_p.h>
 #include <private/qpainter_p.h>
 #include <private/qdrawhelper_x86_p.h>
+#include <private/qdrawhelper_armv6_p.h>
+#include <private/qdrawhelper_neon_p.h>
 #include <private/qmath_p.h>
 #include <qmath.h>
 
 QT_BEGIN_NAMESPACE
+
 
 #define MASK(src, a) src = BYTE_MUL(src, a)
@@ -158 +162 @@
 }
 
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-template <typename EnumType, int value>
-class QEnumToType
-{
-public:
-    inline EnumType operator()() const
-    {
-        return EnumType(value);
-    }
-};
-template <QImage::Format format>
-class QImageFormatToType
-{
-public:
-    inline QImage::Format operator()() const
-    {
-        return format;
-    }
-};
-// Would have used QEnumToType instead of creating a specialized version for QImageFormatToType,
-// but that causes internal compiler error on VC6
-#define Q_TEMPLATE_IMAGEFORMAT_FIX(format) , const QImageFormatToType<format> &imageFormatType
-#define Q_TEMPLATE_IMAGEFORMAT_CALL(format) , QImageFormatToType<format>()
-#define Q_TEMPLATE_ENUM_FIX(Type, Value) , const QEnumToType<Type, Value> &enumTemplateType
-#define Q_TEMPLATE_ENUM_CALL(Type, Value) , QEnumToType<Type, Value>()
-#define Q_TEMPLATE_FIX(Type) , const QTypeInfo<Type> &templateType
-#define Q_TEMPLATE_CALL(Type) , QTypeInfo<Type>()
-#else
-#define Q_TEMPLATE_IMAGEFORMAT_FIX(format)
-#define Q_TEMPLATE_IMAGEFORMAT_CALL(format)
-#define Q_TEMPLATE_ENUM_FIX(Type, Value)
-#define Q_TEMPLATE_ENUM_CALL(Type, Value)
-#define Q_TEMPLATE_FIX(Type)
-#define Q_TEMPLATE_CALL(Type)
-#endif
-
 template <class DST>
 Q_STATIC_TEMPLATE_FUNCTION uint * QT_FASTCALL destFetch(uint *buffer, QRasterBuffer *rasterBuffer,
-                                    int x, int y, int length
-                                    Q_TEMPLATE_FIX(DST))
+                                    int x, int y, int length)
 {
     const DST *src = reinterpret_cast<DST*>(rasterBuffer->scanLine(y)) + x;
@@ -206 +173 @@
 }
 
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-#define DEST_FETCH_DECL(DST)                                            \
-    static uint * QT_FASTCALL destFetch_##DST(uint *buffer,             \
-                                              QRasterBuffer *rasterBuffer, \
-                                              int x, int y, int length) \
-    {                                                                   \
-        return destFetch<DST>(buffer, rasterBuffer, x, y, length Q_TEMPLATE_CALL(DST));      \
-    }
-
-DEST_FETCH_DECL(qargb8565)
-DEST_FETCH_DECL(qrgb666)
-DEST_FETCH_DECL(qargb6666)
-DEST_FETCH_DECL(qrgb555)
-DEST_FETCH_DECL(qrgb888)
-DEST_FETCH_DECL(qargb8555)
-DEST_FETCH_DECL(qrgb444)
-DEST_FETCH_DECL(qargb4444)
-#undef DEST_FETCH_DECL
-# define SPANFUNC_POINTER_DESTFETCH(Arg) destFetch_##Arg
-#else // !VC6 && !VC2002
 # define SPANFUNC_POINTER_DESTFETCH(Arg) destFetch<Arg>
-#endif
 
 static const DestFetchProc destFetchProc[QImage::NImageFormats] =
@@ -367 +313 @@
 Q_STATIC_TEMPLATE_FUNCTION void QT_FASTCALL destStore(QRasterBuffer *rasterBuffer,
                                   int x, int y,
-                                  const uint *buffer, int length
-                                  Q_TEMPLATE_FIX(DST))
+                                  const uint *buffer, int length)
 {
     DST *dest = reinterpret_cast<DST*>(rasterBuffer->scanLine(y)) + x;
-    const quint32 *src = reinterpret_cast<const quint32*>(buffer);
+    const quint32p *src = reinterpret_cast<const quint32p*>(buffer);
     while (length--)
         *dest++ = DST(*src++);
 }
 
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-# define DEST_STORE_DECL(DST)                                           \
-    static void QT_FASTCALL destStore_##DST(QRasterBuffer *rasterBuffer, \
-                                            int x, int y,               \
-                                            const uint *buffer, int length) \
-    {                                                                   \
-        destStore<DST>(rasterBuffer, x, y, buffer, length Q_TEMPLATE_CALL(DST));             \
-    }
-
-DEST_STORE_DECL(qargb8565)
-DEST_STORE_DECL(qrgb555)
-DEST_STORE_DECL(qrgb666)
-DEST_STORE_DECL(qargb6666)
-DEST_STORE_DECL(qargb8555)
-DEST_STORE_DECL(qrgb888)
-DEST_STORE_DECL(qrgb444)
-DEST_STORE_DECL(qargb4444)
-# undef DEST_FETCH_DECL
-# define SPANFUNC_POINTER_DESTSTORE(DEST) destStore_##DEST
-#else // !VC6 && !VC2002
 # define SPANFUNC_POINTER_DESTSTORE(DEST) destStore<DEST>
-#endif
 
 static const DestStoreProc destStoreProc[QImage::NImageFormats] =
@@ -426 +350 @@
   We need 5 fetch methods per surface type:
   untransformed
-  transformed
-  transformed tiled
-  transformed bilinear
-  transformed bilinear tiled
+  transformed (tiled and not tiled)
+  transformed bilinear (tiled and not tiled)
 
   We don't need bounds checks for untransformed, but we need them for the other ones.
@@ -437 +359 @@
 
 template <QImage::Format format>
-Q_STATIC_TEMPLATE_FUNCTION uint QT_FASTCALL qt_fetchPixel(const uchar *scanLine, int x, const QVector<QRgb> *rgb
-                                   Q_TEMPLATE_IMAGEFORMAT_FIX(format));
+Q_STATIC_TEMPLATE_FUNCTION uint QT_FASTCALL qt_fetchPixel(const uchar *scanLine, int x, const QVector<QRgb> *rgb);
 
 template<>
 Q_STATIC_TEMPLATE_SPECIALIZATION
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_Mono>(const uchar *scanLine,
-                                                 int x, const QVector<QRgb> *rgb
-                                                 Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_Mono))
+                                                 int x, const QVector<QRgb> *rgb)
 {
     bool pixel = scanLine[x>>3] & (0x80 >> (x & 7));
@@ -454 +374 @@
 Q_STATIC_TEMPLATE_SPECIALIZATION
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_MonoLSB>(const uchar *scanLine,
-                                                    int x, const QVector<QRgb> *rgb
-                                                    Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_MonoLSB))
+                                                    int x, const QVector<QRgb> *rgb)
 {
     bool pixel = scanLine[x>>3] & (0x1 << (x & 7));
@@ -465 +384 @@
 Q_STATIC_TEMPLATE_SPECIALIZATION
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_Indexed8>(const uchar *scanLine,
-                                                     int x, const QVector<QRgb> *rgb
-                                                     Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_Indexed8))
+                                                     int x, const QVector<QRgb> *rgb)
 {
     return PREMUL(rgb->at(scanLine[x]));
@@ -474 +392 @@
 Q_STATIC_TEMPLATE_SPECIALIZATION
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_ARGB32>(const uchar *scanLine,
-                                                   int x, const QVector<QRgb> *
-                                                   Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB32))
+                                                   int x, const QVector<QRgb> *)
 {
     return PREMUL(((const uint *)scanLine)[x]);
@@ -483 +400 @@
 Q_STATIC_TEMPLATE_SPECIALIZATION
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_ARGB32_Premultiplied>(const uchar *scanLine,
-                                                                 int x, const QVector<QRgb> *
-                                                                 Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB32_Premultiplied))
+                                                                 int x, const QVector<QRgb> *)
 {
     return ((const uint *)scanLine)[x];
@@ -492 +408 @@
 Q_STATIC_TEMPLATE_SPECIALIZATION
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_RGB16>(const uchar *scanLine,
-                                                  int x, const QVector<QRgb> *
-                                                  Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_RGB16))
+                                                  int x, const QVector<QRgb> *)
 {
     return qConvertRgb16To32(((const ushort *)scanLine)[x]);
@@ -502 +417 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_ARGB8565_Premultiplied>(const uchar *scanLine,
                                                      int x,
-                                                     const QVector<QRgb> *
-                                                     Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB8565_Premultiplied))
+                                                     const QVector<QRgb> *)
 {
     const qargb8565 color = reinterpret_cast<const qargb8565*>(scanLine)[x];
@@ -513 +427 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_RGB666>(const uchar *scanLine,
                                                    int x,
-                                                   const QVector<QRgb> *
-                                                   Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_RGB666))
+                                                   const QVector<QRgb> *)
 {
     const qrgb666 color = reinterpret_cast<const qrgb666*>(scanLine)[x];
@@ -524 +437 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_ARGB6666_Premultiplied>(const uchar *scanLine,
                                                    int x,
-                                                   const QVector<QRgb> *
-                                                   Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB6666_Premultiplied))
+                                                   const QVector<QRgb> *)
 {
     const qargb6666 color = reinterpret_cast<const qargb6666*>(scanLine)[x];
@@ -535 +447 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_RGB555>(const uchar *scanLine,
                                                    int x,
-                                                   const QVector<QRgb> *
-                                                   Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_RGB555))
+                                                   const QVector<QRgb> *)
 {
     const qrgb555 color = reinterpret_cast<const qrgb555*>(scanLine)[x];
@@ -546 +457 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_ARGB8555_Premultiplied>(const uchar *scanLine,
                                                      int x,
-                                                     const QVector<QRgb> *
-                                                     Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB8555_Premultiplied))
+                                                     const QVector<QRgb> *)
 {
     const qargb8555 color = reinterpret_cast<const qargb8555*>(scanLine)[x];
@@ -557 +467 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_RGB888>(const uchar *scanLine,
                                                    int x,
-                                                   const QVector<QRgb> *
-                                                   Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_RGB888))
+                                                   const QVector<QRgb> *)
 {
     const qrgb888 color = reinterpret_cast<const qrgb888*>(scanLine)[x];
@@ -568 +477 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_RGB444>(const uchar *scanLine,
                                                    int x,
-                                                   const QVector<QRgb> *
-                                                   Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_RGB444))
+                                                   const QVector<QRgb> *)
 {
     const qrgb444 color = reinterpret_cast<const qrgb444*>(scanLine)[x];
@@ -579 +487 @@
 uint QT_FASTCALL qt_fetchPixel<QImage::Format_ARGB4444_Premultiplied>(const uchar *scanLine,
                                                      int x,
-                                                     const QVector<QRgb> *
-                                                     Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB4444_Premultiplied))
+                                                     const QVector<QRgb> *)
 {
     const qargb4444 color = reinterpret_cast<const qargb4444*>(scanLine)[x];
@@ -586 +493 @@
 }
 
+template<>
+Q_STATIC_TEMPLATE_SPECIALIZATION
+uint QT_FASTCALL qt_fetchPixel<QImage::Format_Invalid>(const uchar *,
+                                                     int ,
+                                                     const QVector<QRgb> *)
+{
+    return 0;
+}
+
 typedef uint (QT_FASTCALL *FetchPixelProc)(const uchar *scanLine, int x, const QVector<QRgb> *);
 
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-
-// explicit template instantiations needed to compile with VC6 and VC2002
-
-#define SPANFUNC_INSTANTIATION_FETCHPIXEL(Arg)  \
-        static inline uint fetchPixel_##Arg(const uchar * scanLine, int x, const QVector<QRgb> * rgb) \
-{ \
-        return qt_fetchPixel<QImage::Arg>(scanLine, x, rgb Q_TEMPLATE_IMAGEFORMAT_CALL(QImage::Arg)); \
-}
-
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_Mono);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_MonoLSB);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_Indexed8);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_ARGB32_Premultiplied);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_ARGB32);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_RGB16);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_ARGB8565_Premultiplied);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_RGB666);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_ARGB6666_Premultiplied);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_RGB555);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_ARGB8555_Premultiplied);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_RGB888);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_RGB444);
-SPANFUNC_INSTANTIATION_FETCHPIXEL(Format_ARGB4444_Premultiplied);
-
-#undef SPANFUNC_INSTANTIATION_FETCHPIXEL
-
-#define SPANFUNC_POINTER_FETCHPIXEL(Arg) fetchPixel_##Arg
-
-#else // !VC6 && !VC2002
-# define SPANFUNC_POINTER_FETCHPIXEL(Arg) qt_fetchPixel<QImage::Arg>
-#endif
+#define SPANFUNC_POINTER_FETCHPIXEL(Arg) qt_fetchPixel<QImage::Arg>
 
 
@@ -654 +539 @@
 template <QImage::Format format>
 Q_STATIC_TEMPLATE_FUNCTION const uint * QT_FASTCALL qt_fetchUntransformed(uint *buffer, const Operator *, const QSpanData *data,
-                                             int y, int x, int length Q_TEMPLATE_IMAGEFORMAT_FIX(format))
+                                             int y, int x, int length)
 {
     const uchar *scanLine = data->texture.scanLine(y);
     for (int i = 0; i < length; ++i)
-        buffer[i] = qt_fetchPixel<format>(scanLine, x + i, data->texture.colorTable Q_TEMPLATE_IMAGEFORMAT_CALL(format));
+        buffer[i] = qt_fetchPixel<format>(scanLine, x + i, data->texture.colorTable);
     return buffer;
 }
@@ -666 +551 @@
 qt_fetchUntransformed<QImage::Format_ARGB32_Premultiplied>(uint *, const Operator *,
                                                          const QSpanData *data,
-                                                         int y, int x, int Q_TEMPLATE_IMAGEFORMAT_FIX(QImage::Format_ARGB32_Premultiplied))
+                                                         int y, int x, int)
 {
     const uchar *scanLine = data->texture.scanLine(y);
@@ -672 +557 @@
 }
 
-static const uint * QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, const QSpanData *data,
+template<TextureBlendType blendType>  /* either BlendTransformed or BlendTransformedTiled */
+Q_STATIC_TEMPLATE_FUNCTION
+const uint * QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, const QSpanData *data,
                                                          int y, int x, int length)
 {
@@ -699 +586 @@
             int py = fy >> 16;
 
-            bool out = (px < 0) || (px >= image_width)
-                       || (py < 0) || (py >= image_height);
-
-            const uchar *scanLine = data->texture.scanLine(py);
-            *b = out ? uint(0) : fetch(scanLine, px, data->texture.colorTable);
+            if (blendType == BlendTransformedTiled) {
+                px %= image_width;
+                py %= image_height;
+                if (px < 0) px += image_width;
+                if (py < 0) py += image_height;
+
+                const uchar *scanLine = data->texture.scanLine(py);
+                *b = fetch(scanLine, px, data->texture.colorTable);
+            } else {
+                if ((px < 0) || (px >= image_width)
+                    || (py < 0) || (py >= image_height)) {
+                    *b = uint(0);
+                } else {
+                    const uchar *scanLine = data->texture.scanLine(py);
+                    *b = fetch(scanLine, px, data->texture.colorTable);
+                }
+            }
             fx += fdx;
             fy += fdy;
@@ -721 +620 @@
             const qreal tx = fx * iw;
             const qreal ty = fy * iw;
-            const int px = int(tx) - (tx < 0);
-            const int py = int(ty) - (ty < 0);
-
-            bool out = (px < 0) || (px >= image_width)
-                       || (py < 0) || (py >= image_height);
-
-            const uchar *scanLine = data->texture.scanLine(py);
-            *b = out ? uint(0) : fetch(scanLine, px, data->texture.colorTable);
+            int px = int(tx) - (tx < 0);
+            int py = int(ty) - (ty < 0);
+
+            if (blendType == BlendTransformedTiled) {
+                px %= image_width;
+                py %= image_height;
+                if (px < 0) px += image_width;
+                if (py < 0) py += image_height;
+
+                const uchar *scanLine = data->texture.scanLine(py);
+                *b = fetch(scanLine, px, data->texture.colorTable);
+            } else {
+                if ((px < 0) || (px >= image_width)
+                    || (py < 0) || (py >= image_height)) {
+                    *b = uint(0);
+                } else {
+                    const uchar *scanLine = data->texture.scanLine(py);
+                    *b = fetch(scanLine, px, data->texture.colorTable);
+                }
+            }
             fx += fdx;
             fy += fdy;
@@ -743 +654 @@
 }
 
-static const uint * QT_FASTCALL fetchTransformedTiled(uint *buffer, const Operator *, const QSpanData *data,
-                                                              int y, int x, int length)
-{
-    FetchPixelProc fetch = fetchPixelProc[data->texture.format];
+template<TextureBlendType blendType, QImage::Format format> /* blendType = BlendTransformedBilinear or BlendTransformedBilinearTiled */
+Q_STATIC_TEMPLATE_FUNCTION
+const uint * QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Operator *, const QSpanData *data,
+                                                  int y, int x, int length)
+{
+#ifdef Q_CC_RVCT // needed to avoid compiler crash in RVCT 2.2
+    FetchPixelProc fetch;
+    if (format != QImage::Format_Invalid)
+        fetch = qt_fetchPixel<format>;
+    else
+        fetch = fetchPixelProc[data->texture.format];
+#else
+    FetchPixelProc fetch = (format != QImage::Format_Invalid) ? FetchPixelProc(qt_fetchPixel<format>) : fetchPixelProc[data->texture.format];
+#endif
 
     int image_width = data->texture.width;
@@ -766 +687 @@
                       + data->m12 * cx + data->dy) * fixed_scale);
 
+        fx -= half_point;
+        fy -= half_point;
         while (b < end) {
-            int px = fx >> 16;
-            int py = fy >> 16;
-
-            px %= image_width;
-            py %= image_height;
-            if (px < 0) px += image_width;
-            if (py < 0) py += image_height;
-
-            const uchar *scanLine = data->texture.scanLine(py);
-            *b = fetch(scanLine, px, data->texture.colorTable);
+            int x1 = (fx >> 16);
+            int x2 = x1 + 1;
+            int y1 = (fy >> 16);
+            int y2 = y1 + 1;
+
+            int distx = ((fx - (x1 << 16)) >> 8);
+            int disty = ((fy - (y1 << 16)) >> 8);
+            int idistx = 256 - distx;
+            int idisty = 256 - disty;
+
+            if (blendType == BlendTransformedBilinearTiled) {
+                x1 %= image_width;
+                x2 %= image_width;
+                y1 %= image_height;
+                y2 %= image_height;
+
+                if (x1 < 0) x1 += image_width;
+                if (x2 < 0) x2 += image_width;
+                if (y1 < 0) y1 += image_height;
+                if (y2 < 0) y2 += image_height;
+
+                Q_ASSERT(x1 >= 0 && x1 < image_width);
+                Q_ASSERT(x2 >= 0 && x2 < image_width);
+                Q_ASSERT(y1 >= 0 && y1 < image_height);
+                Q_ASSERT(y2 >= 0 && y2 < image_height);
+            } else {
+                x1 = qBound(0, x1, image_width - 1);
+                x2 = qBound(0, x2, image_width - 1);
+                y1 = qBound(0, y1, image_height - 1);
+                y2 = qBound(0, y2, image_height - 1);
+            }
+
+            const uchar *s1 = data->texture.scanLine(y1);
+            const uchar *s2 = data->texture.scanLine(y2);
+
+            uint tl = fetch(s1, x1, data->texture.colorTable);
+            uint tr = fetch(s1, x2, data->texture.colorTable);
+            uint bl = fetch(s2, x1, data->texture.colorTable);
+            uint br = fetch(s2, x2, data->texture.colorTable);
+
+            uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
+            uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
+            *b = INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
+
             fx += fdx;
             fy += fdy;
@@ -792 +749 @@
         while (b < end) {
             const qreal iw = fw == 0 ? 1 : 1 / fw;
-            const qreal tx = fx * iw;
-            const qreal ty = fy * iw;
-            int px = int(tx) - (tx < 0);
-            int py = int(ty) - (ty < 0);
-
-            px %= image_width;
-            py %= image_height;
-            if (px < 0) px += image_width;
-            if (py < 0) py += image_height;
-
-            const uchar *scanLine = data->texture.scanLine(py);
-            *b = fetch(scanLine, px, data->texture.colorTable);
+            const qreal px = fx * iw - 0.5;
+            const qreal py = fy * iw - 0.5;
+
+            int x1 = int(px) - (px < 0);
+            int x2 = x1 + 1;
+            int y1 = int(py) - (py < 0);
+            int y2 = y1 + 1;
+
+            int distx = int((px - x1) * 256);
+            int disty = int((py - y1) * 256);
+            int idistx = 256 - distx;
+            int idisty = 256 - disty;
+
+            if (blendType == BlendTransformedBilinearTiled) {
+                x1 %= image_width;
+                x2 %= image_width;
+                y1 %= image_height;
+                y2 %= image_height;
+
+                if (x1 < 0) x1 += image_width;
+                if (x2 < 0) x2 += image_width;
+                if (y1 < 0) y1 += image_height;
+                if (y2 < 0) y2 += image_height;
+
+                Q_ASSERT(x1 >= 0 && x1 < image_width);
+                Q_ASSERT(x2 >= 0 && x2 < image_width);
+                Q_ASSERT(y1 >= 0 && y1 < image_height);
+                Q_ASSERT(y2 >= 0 && y2 < image_height);
+            } else {
+                x1 = qBound(0, x1, image_width - 1);
+                x2 = qBound(0, x2, image_width - 1);
+                y1 = qBound(0, y1, image_height - 1);
+                y2 = qBound(0, y2, image_height - 1);
+            }
+
+            const uchar *s1 = data->texture.scanLine(y1);
+            const uchar *s2 = data->texture.scanLine(y2);
+
+            uint tl = fetch(s1, x1, data->texture.colorTable);
+            uint tr = fetch(s1, x2, data->texture.colorTable);
+            uint bl = fetch(s2, x1, data->texture.colorTable);
+            uint br = fetch(s2, x2, data->texture.colorTable);
+
+            uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
+            uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
+            *b = INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
+
             fx += fdx;
             fy += fdy;
@@ -818 +810 @@
 }
 
-static const uint * QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Operator *, const QSpanData *data,
-                                                                 int y, int x, int length)
-{
-    FetchPixelProc fetch = fetchPixelProc[data->texture.format];
-
-    int image_width = data->texture.width;
-    int image_height = data->texture.height;
-
-    const qreal cx = x + 0.5;
-    const qreal cy = y + 0.5;
-
-    const uint *end = buffer + length;
-    uint *b = buffer;
-    if (data->fast_matrix) {
-        // The increment pr x in the scanline
-        int fdx = (int)(data->m11 * fixed_scale);
-        int fdy = (int)(data->m12 * fixed_scale);
-
-        int fx = int((data->m21 * cy
-                      + data->m11 * cx + data->dx) * fixed_scale);
-        int fy = int((data->m22 * cy
-                      + data->m12 * cx + data->dy) * fixed_scale);
-
-        fx -= half_point;
-        fy -= half_point;
-        while (b < end) {
-            int x1 = (fx >> 16);
-            int x2 = x1 + 1;
-            int y1 = (fy >> 16);
-            int y2 = y1 + 1;
-
-            int distx = ((fx - (x1 << 16)) >> 8);
-            int disty = ((fy - (y1 << 16)) >> 8);
-            int idistx = 256 - distx;
-            int idisty = 256 - disty;
-
-            x1 = qBound(0, x1, image_width - 1);
-            x2 = qBound(0, x2, image_width - 1);
-            y1 = qBound(0, y1, image_height - 1);
-            y2 = qBound(0, y2, image_height - 1);
-
-            const uchar *s1 = data->texture.scanLine(y1);
-            const uchar *s2 = data->texture.scanLine(y2);
-
-            uint tl = fetch(s1, x1, data->texture.colorTable);
-            uint tr = fetch(s1, x2, data->texture.colorTable);
-            uint bl = fetch(s2, x1, data->texture.colorTable);
-            uint br = fetch(s2, x2, data->texture.colorTable);
-
-            uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
-            uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
-            *b = INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
-
-            fx += fdx;
-            fy += fdy;
-            ++b;
-        }
-    } else {
-        const qreal fdx = data->m11;
-        const qreal fdy = data->m12;
-        const qreal fdw = data->m13;
-
-        qreal fx = data->m21 * cy + data->m11 * cx + data->dx;
-        qreal fy = data->m22 * cy + data->m12 * cx + data->dy;
-        qreal fw = data->m23 * cy + data->m13 * cx + data->m33;
-
-        while (b < end) {
-            const qreal iw = fw == 0 ? 1 : 1 / fw;
-            const qreal px = fx * iw - 0.5;
-            const qreal py = fy * iw - 0.5;
-
-            int x1 = int(px) - (px < 0);
-            int x2 = x1 + 1;
-            int y1 = int(py) - (py < 0);
-            int y2 = y1 + 1;
-
-            int distx = int((px - x1) * 256);
-            int disty = int((py - y1) * 256);
-            int idistx = 256 - distx;
-            int idisty = 256 - disty;
-
-            x1 = qBound(0, x1, image_width - 1);
-            x2 = qBound(0, x2, image_width - 1);
-            y1 = qBound(0, y1, image_height - 1);
-            y2 = qBound(0, y2, image_height - 1);
-
-            const uchar *s1 = data->texture.scanLine(y1);
-            const uchar *s2 = data->texture.scanLine(y2);
-
-            uint tl = fetch(s1, x1, data->texture.colorTable);
-            uint tr = fetch(s1, x2, data->texture.colorTable);
-            uint bl = fetch(s2, x1, data->texture.colorTable);
-            uint br = fetch(s2, x2, data->texture.colorTable);
-
-            uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
-            uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
-            *b = INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
-
-            fx += fdx;
-            fy += fdy;
-            fw += fdw;
-            //force increment to avoid /0
-            if (!fw) {
-                fw += fdw;
-            }
-            ++b;
-        }
-    }
-
-    return buffer;
-}
-
-static const uint * QT_FASTCALL fetchTransformedBilinearTiled(uint *buffer, const Operator *, const QSpanData *data,
-                                                                     int y, int x, int length)
-{
-    FetchPixelProc fetch = fetchPixelProc[data->texture.format];
-
-    int image_width = data->texture.width;
-    int image_height = data->texture.height;
-
-    const qreal cx = x + 0.5;
-    const qreal cy = y + 0.5;
-
-    const uint *end = buffer + length;
-    uint *b = buffer;
-    if (data->fast_matrix) {
-        // The increment pr x in the scanline
-        int fdx = (int)(data->m11 * fixed_scale);
-        int fdy = (int)(data->m12 * fixed_scale);
-
-        int fx = int((data->m21 * cy + data->m11 * cx + data->dx) * fixed_scale);
-        int fy = int((data->m22 * cy + data->m12 * cx + data->dy) * fixed_scale);
-
-        fx -= half_point;
-        fy -= half_point;
-        while (b < end) {
-            int x1 = (fx >> 16);
-            int x2 = x1 + 1;
-            int y1 = (fy >> 16);
-            int y2 = y1 + 1;
-
-            int distx = ((fx - (x1 << 16)) >> 8);
-            int disty = ((fy - (y1 << 16)) >> 8);
-            int idistx = 256 - distx;
-            int idisty = 256 - disty;
-
-            x1 %= image_width;
-            x2 %= image_width;
-            y1 %= image_height;
-            y2 %= image_height;
-
-            if (x1 < 0) x1 += image_width;
-            if (x2 < 0) x2 += image_width;
-            if (y1 < 0) y1 += image_height;
-            if (y2 < 0) y2 += image_height;
-
-            Q_ASSERT(x1 >= 0 && x1 < image_width);
-            Q_ASSERT(x2 >= 0 && x2 < image_width);
-            Q_ASSERT(y1 >= 0 && y1 < image_height);
-            Q_ASSERT(y2 >= 0 && y2 < image_height);
-
-            const uchar *s1 = data->texture.scanLine(y1);
-            const uchar *s2 = data->texture.scanLine(y2);
-
-            uint tl = fetch(s1, x1, data->texture.colorTable);
-            uint tr = fetch(s1, x2, data->texture.colorTable);
-            uint bl = fetch(s2, x1, data->texture.colorTable);
-            uint br = fetch(s2, x2, data->texture.colorTable);
-
-            uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
-            uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
-            *b = INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
-
-            fx += fdx;
-            fy += fdy;
-            ++b;
-        }
-    } else {
-        const qreal fdx = data->m11;
-        const qreal fdy = data->m12;
-        const qreal fdw = data->m13;
-
-        qreal fx = data->m21 * cy + data->m11 * cx + data->dx;
-        qreal fy = data->m22 * cy + data->m12 * cx + data->dy;
-        qreal fw = data->m23 * cy + data->m13 * cx + data->m33;
-
-        while (b < end) {
-            const qreal iw = fw == 0 ? 1 : 1 / fw;
-            const qreal px = fx * iw - 0.5;
-            const qreal py = fy * iw - 0.5;
-
-            int x1 = int(px) - (px < 0);
-            int x2 = x1 + 1;
-            int y1 = int(py) - (py < 0);
-            int y2 = y1 + 1;
-
-            int distx = int((px - x1) * 256);
-            int disty = int((py - y1) * 256);
-            int idistx = 256 - distx;
-            int idisty = 256 - disty;
-
-            x1 %= image_width;
-            x2 %= image_width;
-            y1 %= image_height;
-            y2 %= image_height;
-
-            if (x1 < 0) x1 += image_width;
-            if (x2 < 0) x2 += image_width;
-            if (y1 < 0) y1 += image_height;
-            if (y2 < 0) y2 += image_height;
-
-            Q_ASSERT(x1 >= 0 && x1 < image_width);
-            Q_ASSERT(x2 >= 0 && x2 < image_width);
-            Q_ASSERT(y1 >= 0 && y1 < image_height);
-            Q_ASSERT(y2 >= 0 && y2 < image_height);
-
-            const uchar *s1 = data->texture.scanLine(y1);
-            const uchar *s2 = data->texture.scanLine(y2);
-
-            uint tl = fetch(s1, x1, data->texture.colorTable);
-            uint tr = fetch(s1, x2, data->texture.colorTable);
-            uint bl = fetch(s2, x1, data->texture.colorTable);
-            uint br = fetch(s2, x2, data->texture.colorTable);
-
-            uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
-            uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
-            *b = INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
-
-            fx += fdx;
-            fy += fdy;
-            fw += fdw;
-            //force increment to avoid /0
-            if (!fw) {
-                fw += fdw;
-            }
-            ++b;
-        }
-    }
-
-    return buffer;
-}
-
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-
-// explicit template instantiations needed to compile with VC6 and VC2002
-
-#define SPANFUNC_POINTER_FETCHUNTRANSFORMED(Arg)  \
-        const uint *qt_fetchUntransformed_##Arg(uint *buffer, const Operator *op, const QSpanData *data, \
-                                             int y, int x, int length) \
-{ \
-        return qt_fetchUntransformed<QImage::Arg>(buffer, op, data, y, x, length Q_TEMPLATE_IMAGEFORMAT_CALL(QImage::Arg)); \
-}
-
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_Mono);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_MonoLSB);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_Indexed8);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_ARGB32_Premultiplied);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_ARGB32);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_RGB16);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_ARGB8565_Premultiplied);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_RGB666);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_ARGB6666_Premultiplied);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_RGB555);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_ARGB8555_Premultiplied);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_RGB888);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_RGB444);
-SPANFUNC_POINTER_FETCHUNTRANSFORMED(Format_ARGB4444_Premultiplied);
-
-#undef SPANFUNC_POINTER_FETCHUNTRANSFORMED
-
-#define SPANFUNC_POINTER_FETCHHUNTRANSFORMED(Arg) qt_fetchUntransformed_##Arg
-
-#else // !VC6 && !VC2002
-# define SPANFUNC_POINTER_FETCHHUNTRANSFORMED(Arg) qt_fetchUntransformed<QImage::Arg>
-#endif
+#define SPANFUNC_POINTER_FETCHHUNTRANSFORMED(Arg) qt_fetchUntransformed<QImage::Arg>
 
 static const SourceFetchProc sourceFetch[NBlendTypes][QImage::NImageFormats] = {
@@ -1136 +854 @@
     {
         0, // Invalid
-        fetchTransformed,   // Mono
-        fetchTransformed,   // MonoLsb
-        fetchTransformed,   // Indexed8
-        fetchTransformed,   // RGB32
-        fetchTransformed,   // ARGB32
-        fetchTransformed,   // ARGB32_Premultiplied
-        fetchTransformed,   // RGB16
-        fetchTransformed,   // ARGB8565_Premultiplied
-        fetchTransformed,   // RGB666
-        fetchTransformed,   // ARGB6666_Premultiplied
-        fetchTransformed,   // RGB555
-        fetchTransformed,   // ARGB8555_Premultiplied
-        fetchTransformed,   // RGB888
-        fetchTransformed,   // RGB444
-        fetchTransformed,   // ARGB4444_Premultiplied
+        fetchTransformed<BlendTransformed>,   // Mono
+        fetchTransformed<BlendTransformed>,   // MonoLsb
+        fetchTransformed<BlendTransformed>,   // Indexed8
+        fetchTransformed<BlendTransformed>,   // RGB32
+        fetchTransformed<BlendTransformed>,   // ARGB32
+        fetchTransformed<BlendTransformed>,   // ARGB32_Premultiplied
+        fetchTransformed<BlendTransformed>,   // RGB16
+        fetchTransformed<BlendTransformed>,   // ARGB8565_Premultiplied
+        fetchTransformed<BlendTransformed>,   // RGB666
+        fetchTransformed<BlendTransformed>,   // ARGB6666_Premultiplied
+        fetchTransformed<BlendTransformed>,   // RGB555
+        fetchTransformed<BlendTransformed>,   // ARGB8555_Premultiplied
+        fetchTransformed<BlendTransformed>,   // RGB888
+        fetchTransformed<BlendTransformed>,   // RGB444
+        fetchTransformed<BlendTransformed>,   // ARGB4444_Premultiplied
     },
     {
         0, // TransformedTiled
-        fetchTransformedTiled,   // Mono
-        fetchTransformedTiled,   // MonoLsb
-        fetchTransformedTiled,   // Indexed8
-        fetchTransformedTiled,   // RGB32
-        fetchTransformedTiled,   // ARGB32
-        fetchTransformedTiled,   // ARGB32_Premultiplied
-        fetchTransformedTiled,   // RGB16
-        fetchTransformedTiled,   // ARGB8565_Premultiplied
-        fetchTransformedTiled,   // RGB666
-        fetchTransformedTiled,   // ARGB6666_Premultiplied
-        fetchTransformedTiled,   // RGB555
-        fetchTransformedTiled,   // ARGB8555_Premultiplied
-        fetchTransformedTiled,   // RGB888
-        fetchTransformedTiled,   // RGB444
-        fetchTransformedTiled,   // ARGB4444_Premultiplied
+        fetchTransformed<BlendTransformedTiled>,   // Mono
+        fetchTransformed<BlendTransformedTiled>,   // MonoLsb
+        fetchTransformed<BlendTransformedTiled>,   // Indexed8
+        fetchTransformed<BlendTransformedTiled>,   // RGB32
+        fetchTransformed<BlendTransformedTiled>,   // ARGB32
+        fetchTransformed<BlendTransformedTiled>,   // ARGB32_Premultiplied
+        fetchTransformed<BlendTransformedTiled>,   // RGB16
+        fetchTransformed<BlendTransformedTiled>,   // ARGB8565_Premultiplied
+        fetchTransformed<BlendTransformedTiled>,   // RGB666
+        fetchTransformed<BlendTransformedTiled>,   // ARGB6666_Premultiplied
+        fetchTransformed<BlendTransformedTiled>,   // RGB555
+        fetchTransformed<BlendTransformedTiled>,   // ARGB8555_Premultiplied
+        fetchTransformed<BlendTransformedTiled>,   // RGB888
+        fetchTransformed<BlendTransformedTiled>,   // RGB444
+        fetchTransformed<BlendTransformedTiled>,   // ARGB4444_Premultiplied
     },
     {
         0, // Bilinear
-        fetchTransformedBilinear,   // Mono
-        fetchTransformedBilinear,   // MonoLsb
-        fetchTransformedBilinear,   // Indexed8
-        fetchTransformedBilinear,   // RGB32
-        fetchTransformedBilinear,   // ARGB32
-        fetchTransformedBilinear,   // ARGB32_Premultiplied
-        fetchTransformedBilinear,   // RGB16
-        fetchTransformedBilinear,   // ARGB8565_Premultiplied
-        fetchTransformedBilinear,   // RGB666
-        fetchTransformedBilinear,   // ARGB6666_Premultiplied
-        fetchTransformedBilinear,   // RGB555
-        fetchTransformedBilinear,   // ARGB8555_Premultiplied
-        fetchTransformedBilinear,   // RGB888
-        fetchTransformedBilinear,   // RGB444
-        fetchTransformedBilinear    // ARGB4444_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // Mono
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // MonoLsb
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // Indexed8
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_ARGB32_Premultiplied>,   // RGB32
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_ARGB32>,   // ARGB32
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_ARGB32_Premultiplied>,   // ARGB32_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // RGB16
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // ARGB8565_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // RGB666
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // ARGB6666_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // RGB555
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // ARGB8555_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // RGB888
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>,   // RGB444
+        fetchTransformedBilinear<BlendTransformedBilinear, QImage::Format_Invalid>    // ARGB4444_Premultiplied
     },
     {
         0, // BilinearTiled
-        fetchTransformedBilinearTiled,   // Mono
-        fetchTransformedBilinearTiled,   // MonoLsb
-        fetchTransformedBilinearTiled,   // Indexed8
-        fetchTransformedBilinearTiled,   // RGB32
-        fetchTransformedBilinearTiled,   // ARGB32
-        fetchTransformedBilinearTiled,   // ARGB32_Premultiplied
-        fetchTransformedBilinearTiled,   // RGB16
-        fetchTransformedBilinearTiled,   // ARGB8565_Premultiplied
-        fetchTransformedBilinearTiled,   // RGB666
-        fetchTransformedBilinearTiled,   // ARGB6666_Premultiplied
-        fetchTransformedBilinearTiled,   // RGB555
-        fetchTransformedBilinearTiled,   // ARGB8555_Premultiplied
-        fetchTransformedBilinearTiled,   // RGB888
-        fetchTransformedBilinearTiled,   // RGB444
-        fetchTransformedBilinearTiled    // ARGB4444_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // Mono
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // MonoLsb
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // Indexed8
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_ARGB32_Premultiplied>,   // RGB32
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_ARGB32>,   // ARGB32
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_ARGB32_Premultiplied>,   // ARGB32_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // RGB16
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // ARGB8565_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // RGB666
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // ARGB6666_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // RGB555
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // ARGB8555_Premultiplied
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // RGB888
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>,   // RGB444
+        fetchTransformedBilinear<BlendTransformedBilinearTiled, QImage::Format_Invalid>    // ARGB4444_Premultiplied
     },
 };
@@ -1466 +1184 @@
         ry -= data->gradient.conical.center.y;
         while (buffer < end) {
-            qreal angle = atan2(ry, rx) + data->gradient.conical.angle;
+            qreal angle = qAtan2(ry, rx) + data->gradient.conical.angle;
 
             *buffer = qt_gradient_pixel(&data->gradient, 1 - angle / (2*Q_PI));
@@ -1480 +1198 @@
             rw = 1;
         while (buffer < end) {
-            qreal angle = atan2(ry/rw - data->gradient.conical.center.x,
+            qreal angle = qAtan2(ry/rw - data->gradient.conical.center.x,
                                 rx/rw - data->gradient.conical.center.y)
                           + data->gradient.conical.angle;
@@ -1498 +1216 @@
 }
 
-
+#if defined(Q_CC_RVCT)
+// Force ARM code generation for comp_func_* -methods
+#  pragma push
+#  pragma arm
+#  if defined(QT_HAVE_ARMV6)
+static __forceinline void preload(const uint *start)
+{
+    asm( "pld [start]" );
+}
+static const uint L2CacheLineLength = 32;
+static const uint L2CacheLineLengthInInts = L2CacheLineLength/sizeof(uint);
+#    define PRELOAD_INIT(x) preload(x);
+#    define PRELOAD_INIT2(x,y) PRELOAD_INIT(x) PRELOAD_INIT(y)
+#    define PRELOAD_COND(x) if (((uint)&x[i])%L2CacheLineLength == 0) preload(&x[i] + L2CacheLineLengthInInts);
+// Two consecutive preloads stall, so space them out a bit by using different modulus.
+#    define PRELOAD_COND2(x,y) if (((uint)&x[i])%L2CacheLineLength == 0) preload(&x[i] + L2CacheLineLengthInInts); \
+         if (((uint)&y[i])%L2CacheLineLength == 16) preload(&y[i] + L2CacheLineLengthInInts);
+#  endif // QT_HAVE_ARMV6
+#endif // Q_CC_RVCT
+
+#if !defined(Q_CC_RVCT) || !defined(QT_HAVE_ARMV6)
+#    define PRELOAD_INIT(x)
+#    define PRELOAD_INIT2(x,y)
+#    define PRELOAD_COND(x)
+#    define PRELOAD_COND2(x,y)
+#endif
 
 /* The constant alpha factor describes an alpha factor that gets applied
@@ -1531 +1274 @@
     } else {
         int ialpha = 255 - const_alpha;
-        for (int i = 0; i < length; ++i)
+        PRELOAD_INIT(dest)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             dest[i] = BYTE_MUL(dest[i], ialpha);
+        }
     }
 }
@@ -1542 +1288 @@
     } else {
         int ialpha = 255 - const_alpha;
-        for (int i = 0; i < length; ++i)
+        PRELOAD_INIT(dest)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             dest[i] = BYTE_MUL(dest[i], ialpha);
+        }
     }
 }
@@ -1558 +1307 @@
         int ialpha = 255 - const_alpha;
         color = BYTE_MUL(color, const_alpha);
-        for (int i = 0; i < length; ++i)
+        PRELOAD_INIT(dest)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             dest[i] = color + BYTE_MUL(dest[i], ialpha);
+        }
     }
 }
@@ -1569 +1321 @@
     } else {
         int ialpha = 255 - const_alpha;
-        for (int i = 0; i < length; ++i)
+        PRELOAD_INIT2(dest, src)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             dest[i] = INTERPOLATE_PIXEL_255(src[i], const_alpha, dest[i], ialpha);
+        }
     }
 }
@@ -1595 +1350 @@
         if (const_alpha != 255)
             color = BYTE_MUL(color, const_alpha);
-        for (int i = 0; i < length; ++i)
+        PRELOAD_INIT(dest)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             dest[i] = color + BYTE_MUL(dest[i], qAlpha(~color));
+        }
     }
 }
@@ -1602 +1360 @@
 static void QT_FASTCALL comp_func_SourceOver(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = src[i];
             dest[i] = s + BYTE_MUL(dest[i], qAlpha(~s));
@@ -1609 +1369 @@
     } else {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = BYTE_MUL(src[i], const_alpha);
             dest[i] = s + BYTE_MUL(dest[i], qAlpha(~s));
@@ -1624 +1385 @@
     if (const_alpha != 255)
         color = BYTE_MUL(color, const_alpha);
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         dest[i] = d + BYTE_MUL(color, qAlpha(~d));
@@ -1632 +1395 @@
 static void QT_FASTCALL comp_func_DestinationOver(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint d = dest[i];
             dest[i] = d + BYTE_MUL(src[i], qAlpha(~d));
@@ -1639 +1404 @@
     } else {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint d = dest[i];
             uint s = BYTE_MUL(src[i], const_alpha);
@@ -1652 +1418 @@
 static void QT_FASTCALL comp_func_solid_SourceIn(uint *dest, int length, uint color, uint const_alpha)
 {
+    PRELOAD_INIT(dest)
     if (const_alpha == 255) {
-        for (int i = 0; i < length; ++i)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             dest[i] = BYTE_MUL(color, qAlpha(dest[i]));
+        }
     } else {
         color = BYTE_MUL(color, const_alpha);
         uint cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             uint d = dest[i];
             dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(d), d, cia);
@@ -1667 +1437 @@
 static void QT_FASTCALL comp_func_SourceIn(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
-        for (int i = 0; i < length; ++i)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             dest[i] = BYTE_MUL(src[i], qAlpha(dest[i]));
+        }
     } else {
         uint cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint d = dest[i];
             uint s = BYTE_MUL(src[i], const_alpha);
@@ -1691 +1465 @@
         a = BYTE_MUL(a, const_alpha) + 255 - const_alpha;
     }
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         dest[i] = BYTE_MUL(dest[i], a);
     }
@@ -1698 +1474 @@
 static void QT_FASTCALL comp_func_DestinationIn(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
-        for (int i = 0; i < length; ++i)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             dest[i] = BYTE_MUL(dest[i], qAlpha(src[i]));
+        }
     } else {
         int cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint a = BYTE_MUL(qAlpha(src[i]), const_alpha) + cia;
             dest[i] = BYTE_MUL(dest[i], a);
@@ -1717 +1497 @@
 static void QT_FASTCALL comp_func_solid_SourceOut(uint *dest, int length, uint color, uint const_alpha)
 {
+    PRELOAD_INIT(dest)
     if (const_alpha == 255) {
-        for (int i = 0; i < length; ++i)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             dest[i] = BYTE_MUL(color, qAlpha(~dest[i]));
+        }
     } else {
         color = BYTE_MUL(color, const_alpha);
         int cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND(dest)
             uint d = dest[i];
             dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(~d), d, cia);
@@ -1732 +1516 @@
 static void QT_FASTCALL comp_func_SourceOut(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
-        for (int i = 0; i < length; ++i)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             dest[i] = BYTE_MUL(src[i], qAlpha(~dest[i]));
+        }
     } else {
         int cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = BYTE_MUL(src[i], const_alpha);
             uint d = dest[i];
@@ -1755 +1543 @@
     if (const_alpha != 255)
         a = BYTE_MUL(a, const_alpha) + 255 - const_alpha;
-    for (int i = 0; i < length; ++i)
+    PRELOAD_INIT(dest)
+    for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         dest[i] = BYTE_MUL(dest[i], a);
+    }
 }
 
 static void QT_FASTCALL comp_func_DestinationOut(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
-        for (int i = 0; i < length; ++i)
+        for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             dest[i] = BYTE_MUL(dest[i], qAlpha(~src[i]));
+        }
     } else {
         int cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint sia = BYTE_MUL(qAlpha(~src[i]), const_alpha) + cia;
             dest[i] = BYTE_MUL(dest[i], sia);
@@ -1785 +1580 @@
     }
     uint sia = qAlpha(~color);
-    for (int i = 0; i < length; ++i)
+    PRELOAD_INIT(dest)
+    for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(dest[i]), dest[i], sia);
+    }
 }
 
 static void QT_FASTCALL comp_func_SourceAtop(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = src[i];
             uint d = dest[i];
@@ -1799 +1599 @@
     } else {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = BYTE_MUL(src[i], const_alpha);
             uint d = dest[i];
@@ -1818 +1619 @@
         a = qAlpha(color) + 255 - const_alpha;
     }
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         dest[i] = INTERPOLATE_PIXEL_255(d, a, color, qAlpha(~d));
@@ -1826 +1629 @@
 static void QT_FASTCALL comp_func_DestinationAtop(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = src[i];
             uint d = dest[i];
@@ -1835 +1640 @@
         int cia = 255 - const_alpha;
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint s = BYTE_MUL(src[i], const_alpha);
             uint d = dest[i];
@@ -1855 +1661 @@
     uint sia = qAlpha(~color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(~d), d, sia);
@@ -1863 +1671 @@
 static void QT_FASTCALL comp_func_XOR(uint *dest, const uint *src, int length, uint const_alpha)
 {
+    PRELOAD_INIT2(dest, src)
     if (const_alpha == 255) {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint d = dest[i];
             uint s = src[i];
@@ -1871 +1681 @@
     } else {
         for (int i = 0; i < length; ++i) {
+            PRELOAD_COND2(dest, src)
             uint d = dest[i];
             uint s = BYTE_MUL(src[i], const_alpha);
@@ -1921 +1732 @@
     uint s = color;
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
 #define MIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
@@ -1941 +1754 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Plus_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -1977 +1792 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2003 +1820 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Multiply_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2041 +1860 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2067 +1888 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Screen_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2116 +1939 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2142 +1967 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Overlay_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2185 +2012 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2211 +2040 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Darken_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2254 +2085 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2280 +2113 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Lighten_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2333 +2168 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2359 +2196 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_ColorDodge_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2401 +2240 @@
     if (src == 0 || src_da + dst_sa <= sa_da)
         return qt_div_255(temp);
-    else
-        return qt_div_255(sa * (src_da + dst_sa - sa_da) / src + temp);
+    return qt_div_255(sa * (src_da + dst_sa - sa_da) / src + temp);
 }
 
@@ -2413 +2251 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2439 +2279 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_ColorBurn_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2489 +2331 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2515 +2359 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_HardLight_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2542 +2388 @@
 
 /*
-   if 2.Sca < Sa
-       Dca' = Dca.(Sa - (1 - Dca/Da).(2.Sca - Sa)) + Sca.(1 - Da) + Dca.(1 - Sa)
-   otherwise if 8.Dca <= Da
-       Dca' = Dca.(Sa - (1 - Dca/Da).(2.Sca - Sa).(3 - 8.Dca/Da)) + Sca.(1 - Da) + Dca.(1 - Sa)
-   otherwise
-       Dca' = (Dca.Sa + ((Dca/Da)^(0.5).Da - Dca).(2.Sca - Sa)) + Sca.(1 - Da) + Dca.(1 - Sa)
+    if 2.Sca <= Sa
+        Dca' = Dca.(Sa + (2.Sca - Sa).(1 - Dca/Da)) + Sca.(1 - Da) + Dca.(1 - Sa)
+    otherwise if 2.Sca > Sa and 4.Dca <= Da
+        Dca' = Dca.Sa + Da.(2.Sca - Sa).(4.Dca/Da.(4.Dca/Da + 1).(Dca/Da - 1) + 7.Dca/Da) + Sca.(1 - Da) + Dca.(1 - Sa)
+    otherwise if 2.Sca > Sa and 4.Dca > Da
+        Dca' = Dca.Sa + Da.(2.Sca - Sa).((Dca/Da)^0.5 - Dca/Da) + Sca.(1 - Da) + Dca.(1 - Sa)
 */
 static inline int soft_light_op(int dst, int src, int da, int sa)
@@ -2556 +2402 @@
 
     if (src2 < sa)
-        return (dst * ((sa * 255) - (255 - dst_np) * (src2 - sa)) + temp) / 65025;
-    else if (8 * dst <= da)
-        return (dst * ((sa * 255) - ((255 - dst_np) * (src2 - sa) * ((3 * 255) - 8 * dst_np)) / 255) + temp) / 65025;
+        return (dst * (sa * 255 + (src2 - sa) * (255 - dst_np)) + temp) / 65025;
+    else if (4 * dst <= da)
+        return (dst * sa * 255 + da * (src2 - sa) * ((((16 * dst_np - 12 * 255) * dst_np + 3 * 65025) * dst_np) / 65025) + temp) / 65025;
     else {
-        // sqrt is too expensive to do three times per pixel, so skipping it for now
-        // a future possibility is to use a LUT
-        return ((dst * sa * 255) + (int(dst_np) * da - (dst * 255)) * (src2 - sa) + temp) / 65025;
+        return (dst * sa * 255 + da * (src2 - sa) * (int(sqrt(qreal(dst_np * 255))) - dst_np) + temp) / 65025;
     }
 }
@@ -2574 +2418 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2600 +2446 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_SoftLight_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2643 +2491 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2669 +2519 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Difference_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2706 +2558 @@
     int sb = qBlue(color);
 
+    PRELOAD_INIT(dest)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND(dest)
         uint d = dest[i];
         int da = qAlpha(d);
@@ -2732 +2586 @@
 Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Exclusion_impl(uint *dest, const uint *src, int length, const T &coverage)
 {
+    PRELOAD_INIT2(dest, src)
     for (int i = 0; i < length; ++i) {
+        PRELOAD_COND2(dest, src)
         uint d = dest[i];
         uint s = src[i];
@@ -2757 +2613 @@
         comp_func_Exclusion_impl(dest, src, length, QPartialCoverage(const_alpha));
 }
+
+#if defined(Q_CC_RVCT)
+// Restore pragma state from previous #pragma arm
+#  pragma pop
+#endif
 
 static void QT_FASTCALL rasterop_solid_SourceOrDestination(uint *dest,
@@ -3221 +3082 @@
 
 template <class T>
-Q_STATIC_TEMPLATE_FUNCTION void blendColor(int count, const QSpan *spans, void *userData
-                       Q_TEMPLATE_FIX(T))
+Q_STATIC_TEMPLATE_FUNCTION void blendColor(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
@@ -3268 +3128 @@
 }
 
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-#define BLEND_COLOR_DECL(DST)                                           \
-    static void blendColor_##DST(int count,                             \
-                                 const QSpan *spans,                    \
-                                 void *userData)                        \
-    {                                                                   \
-        blendColor<DST>(count, spans, userData Q_TEMPLATE_CALL(DST));   \
-    }
-
-BLEND_COLOR_DECL(qargb8565)
-BLEND_COLOR_DECL(qrgb666)
-BLEND_COLOR_DECL(qargb6666)
-BLEND_COLOR_DECL(qrgb555)
-BLEND_COLOR_DECL(qargb8555)
-BLEND_COLOR_DECL(qrgb888)
-BLEND_COLOR_DECL(qrgb444)
-BLEND_COLOR_DECL(qargb4444)
-#undef DEST_FETCH_DECL
-#define SPANFUNC_POINTER_BLENDCOLOR(DST) blendColor_##DST
-#else // !VC6 && !VC2002
-# define SPANFUNC_POINTER_BLENDCOLOR(DST) blendColor<DST>
-#endif
+#define SPANFUNC_POINTER_BLENDCOLOR(DST) blendColor<DST>
 
 static void blend_color_rgb16(int count, const QSpan *spans, void *userData)
@@ -3367 +3206 @@
 }
 
-template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_src_generic(int count, const QSpan *spans, void *userData
-                              Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
-{
-    QSpanData *data = reinterpret_cast<QSpanData *>(userData);
-
-    uint buffer[buffer_size];
-    uint src_buffer[buffer_size];
-    Operator op = getOperator(data, spans, count);
-
+template <typename T>
+void handleSpans(int count, const QSpan *spans, const QSpanData *data, T &handler)
+{
     uint const_alpha = 256;
     if (data->type == QSpanData::Texture)
         const_alpha = data->texture.const_alpha;
 
-    while (count--) {
+    int coverage = 0;
+    while (count) {
         int x = spans->x;
-        int length = spans->len;
-        const int coverage = (spans->coverage * const_alpha) >> 8;
+        const int y = spans->y;
+        int right = x + spans->len;
+
+        // compute length of adjacent spans
+        for (int i = 1; i < count && spans[i].y == y && spans[i].x == right; ++i)
+            right += spans[i].len;
+        int length = right - x;
+
         while (length) {
             int l = qMin(buffer_size, length);
-            const uint *src = op.src_fetch(src_buffer, &op, data, spans->y, x, l);
-            if (spanMethod == RegularSpans) {
-                uint *dest = op.dest_fetch ? op.dest_fetch(buffer, data->rasterBuffer, x, spans->y, l) : buffer;
-                op.func(dest, src, l, coverage);
-                if (op.dest_store)
-                    op.dest_store(data->rasterBuffer, x, spans->y, dest, l);
-            } else {
-                drawBufferSpan(data, src, l, x, spans->y, l, coverage);
+            length -= l;
+
+            int process_length = l;
+            int process_x = x;
+
+            const uint *src = handler.fetch(process_x, y, process_length);
+            int offset = 0;
+            while (l > 0) {
+                if (x == spans->x) // new span?
+                    coverage = (spans->coverage * const_alpha) >> 8;
+
+                int right = spans->x + spans->len;
+                int len = qMin(l, right - x);
+
+                handler.process(x, y, len, coverage, src, offset);
+
+                l -= len;
+                x += len;
+                offset += len;
+
+                if (x == right) { // done with current span?
+                    ++spans;
+                    --count;
+                }
             }
-            x += l;
-            length -= l;
-        }
-        ++spans;
-    }
-}
+            handler.store(process_x, y, process_length);
+        }
+    }
+}
+
+struct QBlendBase
+{
+    QBlendBase(QSpanData *d, Operator o)
+        : data(d)
+        , op(o)
+        , dest(0)
+    {
+    }
+
+    QSpanData *data;
+    Operator op;
+
+    uint *dest;
+
+    uint buffer[buffer_size];
+    uint src_buffer[buffer_size];
+};
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_untransformed_generic(int count, const QSpan *spans, void *userData
-                                        Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+class BlendSrcGeneric : public QBlendBase
+{
+public:
+    BlendSrcGeneric(QSpanData *d, Operator o)
+        : QBlendBase(d, o)
+    {
+    }
+
+    const uint *fetch(int x, int y, int len)
+    {
+        if (spanMethod == RegularSpans)
+            dest = op.dest_fetch ? op.dest_fetch(buffer, data->rasterBuffer, x, y, len) : buffer;
+
+        return op.src_fetch(src_buffer, &op, data, y, x, len);
+    }
+
+    void process(int x, int y, int len, int coverage, const uint *src, int offset)
+    {
+        if (spanMethod == RegularSpans)
+            op.func(dest + offset, src + offset, len, coverage);
+        else
+            drawBufferSpan(data, src + offset, len, x, y, len, coverage);
+    }
+
+    void store(int x, int y, int len)
+    {
+        if (spanMethod == RegularSpans && op.dest_store) {
+            op.dest_store(data->rasterBuffer, x, y, dest, len);
+        }
+    }
+};
+
+template <SpanMethod spanMethod>
+Q_STATIC_TEMPLATE_FUNCTION void blend_src_generic(int count, const QSpan *spans, void *userData)
+{
+    QSpanData *data = reinterpret_cast<QSpanData *>(userData);
+    BlendSrcGeneric<spanMethod> blend(data, getOperator(data, spans, count));
+    handleSpans(count, spans, data, blend);
+}
+
+template <SpanMethod spanMethod>
+Q_STATIC_TEMPLATE_FUNCTION void blend_untransformed_generic(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
@@ -3456 +3367 @@
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_untransformed_argb(int count, const QSpan *spans, void *userData
-                                     Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_untransformed_argb(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
     if (data->texture.format != QImage::Format_ARGB32_Premultiplied
         && data->texture.format != QImage::Format_RGB32) {
-        blend_untransformed_generic<spanMethod>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, spanMethod));
+        blend_untransformed_generic<spanMethod>(count, spans, userData);
         return;
     }
@@ -4708 +4618 @@
         mode != QPainter::CompositionMode_Source)
     {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
         return;
     }
@@ -4764 +4673 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_24)
+#if defined(QT_QWS_DEPTH_24)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4771 +4680 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4778 +4686 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4787 +4695 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4794 +4701 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4803 +4710 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4810 +4716 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4819 +4725 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4826 +4731 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4835 +4740 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4842 +4746 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4851 +4755 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4858 +4761 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4867 +4770 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4874 +4776 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4883 +4785 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -4890 +4791 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -4899 +4800 @@
     else
 #endif
-        blend_untransformed_generic<RegularSpans>(count, spans, userData
-                                                  Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_untransformed_generic<RegularSpans>(count, spans, userData);
 }
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_tiled_generic(int count, const QSpan *spans, void *userData
-                                Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_tiled_generic(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
@@ -4959 +4858 @@
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_tiled_argb(int count, const QSpan *spans, void *userData
-                             Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_tiled_argb(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
     if (data->texture.format != QImage::Format_ARGB32_Premultiplied
         && data->texture.format != QImage::Format_RGB32) {
-        blend_tiled_generic<spanMethod>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, spanMethod));
+        blend_tiled_generic<spanMethod>(count, spans, userData);
         return;
     }
@@ -5021 +4919 @@
         mode != QPainter::CompositionMode_Source)
     {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
         return;
     }
@@ -5054 +4951 @@
             sy += image_height;
 
-        while (length) {
-            int l = qMin(image_width - sx, length);
-            if (buffer_size < l)
-                l = buffer_size;
-
-            DST *dest = ((DST*)data->rasterBuffer->scanLine(spans->y)) + x;
-            const SRC *src = (SRC*)data->texture.scanLine(sy) + sx;
-            if (modeSource && coverage == 255) {
+        if (modeSource && coverage == 255) {
+            // Copy the first texture block
+            length = qMin(image_width,length);
+            int tx = x;
+            while (length) {
+                int l = qMin(image_width - sx, length);
+                if (buffer_size < l)
+                    l = buffer_size;
+                DST *dest = ((DST*)data->rasterBuffer->scanLine(spans->y)) + tx;
+                const SRC *src = (SRC*)data->texture.scanLine(sy) + sx;
+
                 qt_memconvert<DST, SRC>(dest, src, l);
-            } else if (sizeof(DST) == 3 && sizeof(SRC) == 3 && l >= 4 &&
-                       (quintptr(dest) & 3) == (quintptr(src) & 3))
-            {
-                blendUntransformed_dest24(dest, src, coverage, l);
-            } else if (sizeof(DST) == 2 && sizeof(SRC) == 2 && l >= 2 &&
-                       (quintptr(dest) & 3) == (quintptr(src) & 3))
-            {
-                blendUntransformed_dest16(dest, src, coverage, l);
-            } else {
-                blendUntransformed_unaligned(dest, src, coverage, l);
+                length -= l;
+                tx += l;
+                sx = 0;
             }
 
-            x += l;
-            length -= l;
-            sx = 0;
+            // Now use the rasterBuffer as the source of the texture,
+            // We can now progressively copy larger blocks
+            // - Less cpu time in code figuring out what to copy
+            // We are dealing with one block of data
+            // - More likely to fit in the cache
+            // - can use memcpy
+            int copy_image_width = qMin(image_width, int(spans->len));
+            length = spans->len - copy_image_width;
+            DST *src = ((DST*)data->rasterBuffer->scanLine(spans->y)) + x;
+            DST *dest = src + copy_image_width;
+            while (copy_image_width < length) {
+                qt_memconvert(dest, src, copy_image_width);
+                dest += copy_image_width;
+                length -= copy_image_width;
+                copy_image_width *= 2;
+            }
+            qt_memconvert(dest, src, length);
+        } else {
+            while (length) {
+                int l = qMin(image_width - sx, length);
+                if (buffer_size < l)
+                    l = buffer_size;
+                DST *dest = ((DST*)data->rasterBuffer->scanLine(spans->y)) + x;
+                const SRC *src = (SRC*)data->texture.scanLine(sy) + sx;
+                if (sizeof(DST) == 3 && sizeof(SRC) == 3 && l >= 4 &&
+                           (quintptr(dest) & 3) == (quintptr(src) & 3))
+                {
+                    blendUntransformed_dest24(dest, src, coverage, l);
+                } else if (sizeof(DST) == 2 && sizeof(SRC) == 2 && l >= 2 &&
+                           (quintptr(dest) & 3) == (quintptr(src) & 3))
+                {
+                    blendUntransformed_dest16(dest, src, coverage, l);
+                } else {
+                    blendUntransformed_unaligned(dest, src, coverage, l);
+                }
+
+                x += l;
+                length -= l;
+                sx = 0;
+            }
         }
         ++spans;
@@ -5085 +5015 @@
 static void blend_tiled_rgb888(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_24)
+#if defined(QT_QWS_DEPTH_24)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5092 +5022 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_argb6666(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5107 +5036 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_rgb666(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5122 +5050 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_argb8565(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5137 +5064 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_rgb565(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5152 +5078 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_argb8555(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5167 +5092 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_rgb555(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5182 +5106 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_argb4444(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5197 +5120 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_tiled_rgb444(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5212 +5134 @@
     else
 #endif
-        blend_tiled_generic<RegularSpans>(count, spans, userData
-                                          Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const QSpan *spans, void *userData
-                                            Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
     if (data->texture.format != QImage::Format_ARGB32_Premultiplied
         && data->texture.format != QImage::Format_RGB32) {
-        blend_src_generic<spanMethod>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, spanMethod));
+        blend_src_generic<spanMethod>(count, spans, userData);
         return;
     }
@@ -5403 +5323 @@
 
     if (mode != QPainter::CompositionMode_SourceOver) {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
         return;
     }
@@ -5600 +5519 @@
 static void blend_transformed_bilinear_rgb888(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_24)
+#if defined(QT_QWS_DEPTH_24)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5607 +5526 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_argb6666(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5622 +5540 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_rgb666(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5637 +5554 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_argb8565(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5652 +5568 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -5659 +5574 @@
                                               void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5668 +5583 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_argb8555(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5683 +5597 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_rgb555(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5698 +5611 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_argb4444(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5713 +5625 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 static void blend_transformed_bilinear_rgb444(int count, const QSpan *spans, void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -5728 +5639 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_tiled_argb(int count, const QSpan *spans, void *userData
-                                                  Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_tiled_argb(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
     if (data->texture.format != QImage::Format_ARGB32_Premultiplied
         && data->texture.format != QImage::Format_RGB32) {
-        blend_src_generic<spanMethod>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, spanMethod));
+        blend_src_generic<spanMethod>(count, spans, userData);
         return;
     }
@@ -5925 +5834 @@
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_argb(int count, const QSpan *spans, void *userData
-                                   Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_argb(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
     if (data->texture.format != QImage::Format_ARGB32_Premultiplied
         && data->texture.format != QImage::Format_RGB32) {
-        blend_src_generic<spanMethod>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, spanMethod));
+        blend_src_generic<spanMethod>(count, spans, userData);
         return;
     }
@@ -6053 +5961 @@
 
     if (mode != QPainter::CompositionMode_SourceOver) {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
         return;
     }
@@ -6196 +6103 @@
                                      void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_24)
+#if defined(QT_QWS_DEPTH_24)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6203 +6110 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6210 +6116 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6219 +6125 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6226 +6131 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6235 +6140 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6242 +6146 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6251 +6155 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6258 +6161 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6267 +6170 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6274 +6176 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6283 +6185 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6290 +6191 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6299 +6200 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6306 +6206 @@
                                          void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6315 +6215 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6322 +6221 @@
                                        void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6331 +6230 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
 template <SpanMethod spanMethod>
-Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_tiled_argb(int count, const QSpan *spans, void *userData
-                                         Q_TEMPLATE_ENUM_FIX(SpanMethod, spanMethod))
+Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_tiled_argb(int count, const QSpan *spans, void *userData)
 {
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
     if (data->texture.format != QImage::Format_ARGB32_Premultiplied
         && data->texture.format != QImage::Format_RGB32) {
-        blend_src_generic<spanMethod>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, spanMethod));
+        blend_src_generic<spanMethod>(count, spans, userData);
         return;
     }
@@ -6476 +6373 @@
 
     if (mode != QPainter::CompositionMode_SourceOver) {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
         return;
     }
@@ -6620 +6516 @@
                                            void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_24)
+#if defined(QT_QWS_DEPTH_24)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6627 +6523 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6634 +6529 @@
                                              void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6643 +6538 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6650 +6544 @@
                                            void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_18)
+#if defined(QT_QWS_DEPTH_18)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6659 +6553 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6666 +6559 @@
                                              void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6675 +6568 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6682 +6574 @@
                                            void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_16)
+#if defined(QT_QWS_DEPTH_16)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6691 +6583 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6698 +6589 @@
                                              void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6707 +6598 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6714 +6604 @@
                                            void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_15)
+#if defined(QT_QWS_DEPTH_15)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6723 +6613 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6730 +6619 @@
                                              void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6739 +6628 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
 }
 
@@ -6746 +6634 @@
                                            void *userData)
 {
-#if !defined(Q_WS_QWS) || defined(QT_QWS_DEPTH_12)
+#if defined(QT_QWS_DEPTH_12)
     QSpanData *data = reinterpret_cast<QSpanData *>(userData);
 
@@ -6755 +6643 @@
     else
 #endif
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
-}
-
-#if defined(Q_CC_MSVC) && _MSC_VER <= 1300 && !defined(Q_CC_INTEL)
-
-// explicit template instantiations needed to compile with VC6 and VC2002
-
-#define SPANFUNC_INSTANTIATION(Name, Arg)  \
-static inline void Name##_##Arg(int count, const QSpan *spans, void *userData) \
-{ \
-    Name<Arg>(count, spans, userData Q_TEMPLATE_ENUM_CALL(SpanMethod, Arg)); \
-}
-
-SPANFUNC_INSTANTIATION(blend_untransformed_generic, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_untransformed_argb, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_tiled_generic, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_tiled_argb, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_src_generic, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_transformed_argb, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_transformed_tiled_argb, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_transformed_bilinear_argb, RegularSpans);
-SPANFUNC_INSTANTIATION(blend_transformed_bilinear_tiled_argb, RegularSpans);
-#undef SPANFUNC_INSTANTIATION
-
-#define SPANFUNC_POINTER(Name, Arg) Name##_##Arg
-
-#else // !VC6 && !VC2002
+        blend_src_generic<RegularSpans>(count, spans, userData);
+}
+
 # define SPANFUNC_POINTER(Name, Arg) Name<Arg>
-#endif
 
 
@@ -7151 +7013 @@
 
     } else {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
     }
 }
@@ -7200 +7061 @@
 
     } else {
-        blend_src_generic<RegularSpans>(count, spans, userData
-                                        Q_TEMPLATE_ENUM_CALL(SpanMethod, RegularSpans));
+        blend_src_generic<RegularSpans>(count, spans, userData);
     }
 }
@@ -7260 +7120 @@
 
 void qt_build_pow_tables() {
-    qreal smoothing = 1.7;
+    qreal smoothing = qreal(1.7);
+
+#ifdef Q_WS_MAC
+    // decided by testing a few things on an iMac, should probably get this from the
+    // system...
+    smoothing = 2.0;
+#endif
 
 #ifdef Q_WS_WIN
@@ -7276 +7142 @@
 #else
     for (int i=0; i<256; ++i) {
-        qt_pow_rgb_gamma[i] = uchar(qRound(pow(i / 255.0, smoothing) * 255));
-        qt_pow_rgb_invgamma[i] = uchar(qRound(pow(i / 255.0, 1 / smoothing) * 255));
+        qt_pow_rgb_gamma[i] = uchar(qRound(qPow(i / qreal(255.0), smoothing) * 255));
+        qt_pow_rgb_invgamma[i] = uchar(qRound(qPow(i / qreal(255.), 1 / smoothing) * 255));
     }
 #endif
@@ -7284 +7150 @@
     const qreal gray_gamma = 2.31;
     for (int i=0; i<256; ++i)
-        qt_pow_gamma[i] = uint(qRound(pow(i / 255.0, gray_gamma) * 2047));
+        qt_pow_gamma[i] = uint(qRound(qPow(i / qreal(255.), gray_gamma) * 2047));
     for (int i=0; i<2048; ++i)
-        qt_pow_invgamma[i] = uchar(qRound(pow(i / 2047.0, 1 / gray_gamma) * 255));
+        qt_pow_invgamma[i] = uchar(qRound(qPow(i / 2047.0, 1 / gray_gamma) * 255));
 #endif
 }
@@ -7406 +7272 @@
                     {
                         int ialpha = 255 - coverage;
-                        dest[i] = BYTE_MUL(c, uint(coverage)) + BYTE_MUL(dest[i], ialpha);
+                        dest[i] = INTERPOLATE_PIXEL_255(c, coverage, dest[i], ialpha);
                     }
                 }
@@ -7447 +7313 @@
                         {
                             int ialpha = 255 - coverage;
-                            dest[xp] = BYTE_MUL(c, uint(coverage)) + BYTE_MUL(dest[xp], ialpha);
+                            dest[xp] = INTERPOLATE_PIXEL_255(c, coverage, dest[xp], ialpha);
                         }
                     }
@@ -7549 +7415 @@
 QT_RECTFILL(quint16)
 QT_RECTFILL(qargb8565)
-QT_RECTFILL(qrgb666);
-QT_RECTFILL(qargb6666);
+QT_RECTFILL(qrgb666)
+QT_RECTFILL(qargb6666)
 QT_RECTFILL(qrgb555)
 QT_RECTFILL(qargb8555)
 QT_RECTFILL(qrgb888)
-QT_RECTFILL(qrgb444);
-QT_RECTFILL(qargb4444);
+QT_RECTFILL(qrgb444)
+QT_RECTFILL(qargb4444)
 #undef QT_RECTFILL
+
+inline static void qt_rectfill_nonpremul_quint32(QRasterBuffer *rasterBuffer,
+                                                 int x, int y, int width, int height,
+                                                 quint32 color)
+{
+    qt_rectfill<quint32>(reinterpret_cast<quint32 *>(rasterBuffer->buffer()),
+                         INV_PREMUL(color), x, y, width, height, rasterBuffer->bytesPerLine());
+}
 
 
@@ -7600 +7474 @@
         qt_alphamapblit_quint32,
         qt_alphargbblit_quint32,
-        qt_rectfill_quint32
+        qt_rectfill_nonpremul_quint32
     },
     // Format_ARGB32_Premultiplied
@@ -7853 +7727 @@
     SSE2        = 0x20,
     CMOV        = 0x40,
-    IWMMXT      = 0x80
+    IWMMXT      = 0x80,
+    NEON        = 0x100
 };
 
@@ -7879 +7754 @@
     static const bool doIWMMXT = !qgetenv("QT_NO_IWMMXT").toInt();
     return doIWMMXT ? IWMMXT : 0;
+#elif defined(QT_HAVE_NEON)
+    static const bool doNEON = !qgetenv("QT_NO_NEON").toInt();
+    return doNEON ? NEON : 0;
 #else
     uint features = 0;
@@ -8033 +7911 @@
 }
 
+#if defined(Q_CC_RVCT) && defined(QT_HAVE_ARMV6)
+// Move these to qdrawhelper_arm.c when all
+// functions are implemented using arm assembly.
+static CompositionFunctionSolid qt_functionForModeSolid_ARMv6[numCompositionFunctions] = {
+        comp_func_solid_SourceOver,
+        comp_func_solid_DestinationOver,
+        comp_func_solid_Clear,
+        comp_func_solid_Source,
+        comp_func_solid_Destination,
+        comp_func_solid_SourceIn,
+        comp_func_solid_DestinationIn,
+        comp_func_solid_SourceOut,
+        comp_func_solid_DestinationOut,
+        comp_func_solid_SourceAtop,
+        comp_func_solid_DestinationAtop,
+        comp_func_solid_XOR,
+        comp_func_solid_Plus,
+        comp_func_solid_Multiply,
+        comp_func_solid_Screen,
+        comp_func_solid_Overlay,
+        comp_func_solid_Darken,
+        comp_func_solid_Lighten,
+        comp_func_solid_ColorDodge,
+        comp_func_solid_ColorBurn,
+        comp_func_solid_HardLight,
+        comp_func_solid_SoftLight,
+        comp_func_solid_Difference,
+        comp_func_solid_Exclusion,
+        rasterop_solid_SourceOrDestination,
+        rasterop_solid_SourceAndDestination,
+        rasterop_solid_SourceXorDestination,
+        rasterop_solid_NotSourceAndNotDestination,
+        rasterop_solid_NotSourceOrNotDestination,
+        rasterop_solid_NotSourceXorDestination,
+        rasterop_solid_NotSource,
+        rasterop_solid_NotSourceAndDestination,
+        rasterop_solid_SourceAndNotDestination
+};
+
+static CompositionFunction qt_functionForMode_ARMv6[numCompositionFunctions] = {
+        comp_func_SourceOver_armv6,
+        comp_func_DestinationOver,
+        comp_func_Clear,
+        comp_func_Source_armv6,
+        comp_func_Destination,
+        comp_func_SourceIn,
+        comp_func_DestinationIn,
+        comp_func_SourceOut,
+        comp_func_DestinationOut,
+        comp_func_SourceAtop,
+        comp_func_DestinationAtop,
+        comp_func_XOR,
+        comp_func_Plus,
+        comp_func_Multiply,
+        comp_func_Screen,
+        comp_func_Overlay,
+        comp_func_Darken,
+        comp_func_Lighten,
+        comp_func_ColorDodge,
+        comp_func_ColorBurn,
+        comp_func_HardLight,
+        comp_func_SoftLight,
+        comp_func_Difference,
+        comp_func_Exclusion,
+        rasterop_SourceOrDestination,
+        rasterop_SourceAndDestination,
+        rasterop_SourceXorDestination,
+        rasterop_NotSourceAndNotDestination,
+        rasterop_NotSourceOrNotDestination,
+        rasterop_NotSourceXorDestination,
+        rasterop_NotSource,
+        rasterop_NotSourceAndDestination,
+        rasterop_SourceAndNotDestination
+};
+
+static void qt_blend_color_argb_armv6(int count, const QSpan *spans, void *userData)
+{
+    QSpanData *data = reinterpret_cast<QSpanData *>(userData);
+
+    CompositionFunctionSolid func = qt_functionForModeSolid_ARMv6[data->rasterBuffer->compositionMode];
+    while (count--) {
+        uint *target = ((uint *)data->rasterBuffer->scanLine(spans->y)) + spans->x;
+        func(target, spans->len, data->solid.color, spans->coverage);
+        ++spans;
+    }
+}
+
+#endif // Q_CC_RVCT && QT_HAVE_ARMV6
+
+
 void qInitDrawhelperAsm()
 {
@@ -8148 +8116 @@
 #endif // QT_NO_DEBUG
 
+#if defined(Q_CC_RVCT) && defined(QT_HAVE_ARMV6)
+        functionForModeAsm = qt_functionForMode_ARMv6;
+        functionForModeSolidAsm = qt_functionForModeSolid_ARMv6;
+
+        qt_memfill32 = qt_memfill32_armv6;
+
+        qDrawHelper[QImage::Format_ARGB32_Premultiplied].blendColor = qt_blend_color_argb_armv6;
+
+        qBlendFunctions[QImage::Format_RGB32][QImage::Format_RGB32] = qt_blend_rgb32_on_rgb32_armv6;
+        qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_RGB32] = qt_blend_rgb32_on_rgb32_armv6;
+        qBlendFunctions[QImage::Format_RGB32][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_armv6;
+        qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_armv6;
+#elif defined(QT_HAVE_NEON)
+        if (features & NEON) {
+            qBlendFunctions[QImage::Format_RGB32][QImage::Format_RGB32] = qt_blend_rgb32_on_rgb32_neon;
+            qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_RGB32] = qt_blend_rgb32_on_rgb32_neon;
+            qBlendFunctions[QImage::Format_RGB32][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_neon;
+            qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_neon;
+        }
+#endif
+
     if (functionForModeSolidAsm) {
         const int destinationMode = QPainter::CompositionMode_Destination;