Merge branch '4.7' of scm.dev.nokia.troll.no:qt/oslo-staging-2 into 4.7-integration

* '4.7' of scm.dev.nokia.troll.no:qt/oslo-staging-2:
  Fix the blending of ARGB_PM image when using palignr to load the data
  Adjusted define styling
  Fixes QWS window decoration offset issues
  tst_QDataStream::stream_QPixmap: actualy test QPixmap.
  Keyboard navigation regression in QTreeView
  Implement the general blending of ARGB32_pm with SSSE3
  QtDeclarative debugging: Add an option not to stream the properties of an object.

6 files changed, 282 insertions, 6 deletions

diff --git a/src/gui/itemviews/qabstractitemview.cpp b/src/gui/itemviews/qabstractitemview.cpp
index 97499f3..4ffd284 100644
--- a/src/gui/itemviews/qabstractitemview.cpp
+++ b/src/gui/itemviews/qabstractitemview.cpp
@@ -2270,9 +2270,11 @@ void QAbstractItemView::keyPressEvent(QKeyEvent *event)
             } else {
                 d->selectionModel->setCurrentIndex(newCurrent, command);
                 d->pressedPosition = visualRect(newCurrent).center() + d->offset();
-                // We copy the same behaviour as for mousePressEvent().
-                QRect rect(d->pressedPosition - d->offset(), QSize(1, 1));
-                setSelection(rect, command);
+                if (newCurrent.isValid()) {
+                    // We copy the same behaviour as for mousePressEvent().
+                    QRect rect(d->pressedPosition - d->offset(), QSize(1, 1));
+                    setSelection(rect, command);
+                }
             }
             event->accept();
             return;
diff --git a/src/gui/painting/painting.pri b/src/gui/painting/painting.pri
index dfa4a48..793d380 100644
--- a/src/gui/painting/painting.pri
+++ b/src/gui/painting/painting.pri
@@ -212,6 +212,7 @@ if(mmx|3dnow|sse|sse2|iwmmxt) {
     SSE3DNOW_SOURCES += painting/qdrawhelper_sse3dnow.cpp
     SSE_SOURCES += painting/qdrawhelper_sse.cpp
     SSE2_SOURCES += painting/qdrawhelper_sse2.cpp
+    SSSE3_SOURCES += painting/qdrawhelper_ssse3.cpp
     IWMMXT_SOURCES += painting/qdrawhelper_iwmmxt.cpp
 }
 
diff --git a/src/gui/painting/qbackingstore.cpp b/src/gui/painting/qbackingstore.cpp
index 83751ed..ac9b994 100644
--- a/src/gui/painting/qbackingstore.cpp
+++ b/src/gui/painting/qbackingstore.cpp
@@ -352,7 +352,7 @@ void QWidgetBackingStore::beginPaint(QRegion &toClean, QWidget *widget, QWindowS
     // Always flush repainted areas.
     dirtyOnScreen += toClean;
 
-#ifdef Q_WS_QWS
+#if defined(Q_WS_QWS) && !defined(Q_BACKINGSTORE_SUBSURFACES)
     toClean.translate(tlwOffset);
 #endif
 
@@ -1294,7 +1294,12 @@ void QWidgetBackingStore::sync()
 #ifdef Q_BACKINGSTORE_SUBSURFACES
         QWindowSurface *subSurface = w->windowSurface();
         BeginPaintInfo beginPaintInfo;
-        beginPaint(toBePainted, w, subSurface, &beginPaintInfo, false);
+
+        QPoint off = w->mapTo(tlw, QPoint());
+        toBePainted.translate(off);
+        beginPaint(toBePainted, w, subSurface, &beginPaintInfo, true);
+        toBePainted.translate(-off);
+
         if (beginPaintInfo.nothingToPaint)
             continue;
 
diff --git a/src/gui/painting/qdrawhelper.cpp b/src/gui/painting/qdrawhelper.cpp
index 1ff3d7b..276da93 100644
--- a/src/gui/painting/qdrawhelper.cpp
+++ b/src/gui/painting/qdrawhelper.cpp
@@ -7840,6 +7840,17 @@ void qInitDrawhelperAsm()
             qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_RGB32] = qt_blend_rgb32_on_rgb32_sse2;
             qBlendFunctions[QImage::Format_RGB32][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_sse2;
             qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_sse2;
+
+#if defined(QT_HAVE_SSSE3)
+            if (features & SSSE3) {
+                extern void qt_blend_argb32_on_argb32_ssse3(uchar *destPixels, int dbpl,
+                                                            const uchar *srcPixels, int sbpl,
+                                                            int w, int h,
+                                                            int const_alpha);
+                qBlendFunctions[QImage::Format_RGB32][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_ssse3;
+                qBlendFunctions[QImage::Format_ARGB32_Premultiplied][QImage::Format_ARGB32_Premultiplied] = qt_blend_argb32_on_argb32_ssse3;
+            }
+#endif // QT_HAVE_SSSE3
         } else
 #endif
         {
diff --git a/src/gui/painting/qdrawhelper_ssse3.cpp b/src/gui/painting/qdrawhelper_ssse3.cpp
new file mode 100644
index 0000000..9c02009
--- /dev/null
+++ b/src/gui/painting/qdrawhelper_ssse3.cpp
@@ -0,0 +1,253 @@
+/****************************************************************************
+**
+** Copyright (C) 2010 Nokia Corporation and/or its subsidiary(-ies).
+** All rights reserved.
+** Contact: Nokia Corporation (qt-info@nokia.com)
+**
+** This file is part of the QtGui module of the Qt Toolkit.
+**
+** $QT_BEGIN_LICENSE:LGPL$
+** No Commercial Usage
+** This file contains pre-release code and may not be distributed.
+** You may use this file in accordance with the terms and conditions
+** contained in the Technology Preview License Agreement accompanying
+** this package.
+**
+** GNU Lesser General Public License Usage
+** Alternatively, this file may be used under the terms of the GNU Lesser
+** General Public License version 2.1 as published by the Free Software
+** Foundation and appearing in the file LICENSE.LGPL included in the
+** packaging of this file.  Please review the following information to
+** ensure the GNU Lesser General Public License version 2.1 requirements
+** 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 1.1, included in the file LGPL_EXCEPTION.txt in this package.
+**
+** If you have questions regarding the use of this file, please contact
+** Nokia at qt-info@nokia.com.
+**
+**
+**
+**
+**
+**
+**
+**
+** $QT_END_LICENSE$
+**
+****************************************************************************/
+
+
+#ifdef QT_HAVE_SSSE3
+
+#include <private/qsimd_p.h>
+#include <private/qdrawhelper_x86_p.h>
+#include <private/qdrawingprimitive_sse2_p.h>
+
+QT_BEGIN_NAMESPACE
+
+inline static void blend_pixel(quint32 &dst, const quint32 src)
+{
+    if (src >= 0xff000000)
+        dst = src;
+    else if (src != 0)
+        dst = src + BYTE_MUL(dst, qAlpha(~src));
+}
+
+
+/* The instruction palignr uses direct arguments, so we have to generate the code fo the different
+   shift (4, 8, 12). Checking the alignment inside the loop is unfortunatelly way too slow.
+ */
+#define BLENDING_LOOP(palignrOffset, length)\
+    for (; x < length-3; x += 4) { \
+        const __m128i srcVectorLastLoaded = _mm_load_si128((__m128i *)&src[x - minusOffsetToAlignSrcOn16Bytes + 4]);\
+        const __m128i srcVector = _mm_alignr_epi8(srcVectorLastLoaded, srcVectorPrevLoaded, palignrOffset); \
+        const __m128i srcVectorAlpha = _mm_and_si128(srcVector, alphaMask); \
+        if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, alphaMask)) == 0xffff) { \
+            _mm_store_si128((__m128i *)&dst[x], srcVector); \
+        } else if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, nullVector)) != 0xffff) { \
+            __m128i alphaChannel = _mm_shuffle_epi8(srcVector, alphaShuffleMask); \
+            alphaChannel = _mm_sub_epi16(one, alphaChannel); \
+            const __m128i dstVector = _mm_load_si128((__m128i *)&dst[x]); \
+            __m128i destMultipliedByOneMinusAlpha; \
+            BYTE_MUL_SSE2(destMultipliedByOneMinusAlpha, dstVector, alphaChannel, colorMask, half); \
+            const __m128i result = _mm_add_epi8(srcVector, destMultipliedByOneMinusAlpha); \
+            _mm_store_si128((__m128i *)&dst[x], result); \
+        } \
+        srcVectorPrevLoaded = srcVectorLastLoaded;\
+    }
+
+
+#define BLEND_SOURCE_OVER_ARGB32_FIRST_ROW_SSSE3(dst, src, length, nullVector, half, one, colorMask, alphaMask) { \
+    int x = 0; \
+\
+    /* First, get dst aligned. */ \
+    const int offsetToAlignOn16Bytes = (4 - ((reinterpret_cast<quintptr>(dst) >> 2) & 0x3)) & 0x3;\
+    const int prologLength = qMin(length, offsetToAlignOn16Bytes);\
+\
+    for (; x < prologLength; ++x) {\
+        blend_pixel(dst[x], src[x]); \
+    } \
+\
+    const int minusOffsetToAlignSrcOn16Bytes = (reinterpret_cast<quintptr>(&(src[x])) >> 2) & 0x3;\
+\
+    if (!minusOffsetToAlignSrcOn16Bytes) {\
+        /* src is aligned, usual algorithm but with aligned operations.\
+           See the SSE2 version for more documentation on the algorithm itself. */\
+        const __m128i alphaShuffleMask = _mm_set_epi8(0xff,15,0xff,15,0xff,11,0xff,11,0xff,7,0xff,7,0xff,3,0xff,3);\
+        for (; x < length-3; x += 4) { \
+            const __m128i srcVector = _mm_load_si128((__m128i *)&src[x]); \
+            const __m128i srcVectorAlpha = _mm_and_si128(srcVector, alphaMask); \
+            if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, alphaMask)) == 0xffff) { \
+                _mm_store_si128((__m128i *)&dst[x], srcVector); \
+            } else if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, nullVector)) != 0xffff) { \
+                __m128i alphaChannel = _mm_shuffle_epi8(srcVector, alphaShuffleMask); \
+                alphaChannel = _mm_sub_epi16(one, alphaChannel); \
+                const __m128i dstVector = _mm_load_si128((__m128i *)&dst[x]); \
+                __m128i destMultipliedByOneMinusAlpha; \
+                BYTE_MUL_SSE2(destMultipliedByOneMinusAlpha, dstVector, alphaChannel, colorMask, half); \
+                const __m128i result = _mm_add_epi8(srcVector, destMultipliedByOneMinusAlpha); \
+                _mm_store_si128((__m128i *)&dst[x], result); \
+            } \
+        } /* end for() */\
+    } else if ((length - x) >= 8) {\
+        /* We are at the first line, so "x - minusOffsetToAlignSrcOn16Bytes" could go before src, and\
+           generate an invalid access. */\
+\
+        /* We use two vectors to extract the src: prevLoaded for the first pixels, lastLoaded for the current pixels. */\
+        __m128i srcVectorPrevLoaded;\
+        if (minusOffsetToAlignSrcOn16Bytes > prologLength) {\
+            /* We go forward 4 pixels to avoid reading before src. */\
+            for (; x < prologLength + 4; ++x)\
+                blend_pixel(dst[x], src[x]); \
+        }\
+        srcVectorPrevLoaded = _mm_load_si128((__m128i *)&src[x - minusOffsetToAlignSrcOn16Bytes]);\
+        const int palignrOffset = minusOffsetToAlignSrcOn16Bytes << 2;\
+\
+        const __m128i alphaShuffleMask = _mm_set_epi8(0xff,15,0xff,15,0xff,11,0xff,11,0xff,7,0xff,7,0xff,3,0xff,3);\
+        switch (palignrOffset) {\
+        case 4:\
+            BLENDING_LOOP(4, length)\
+            break;\
+        case 8:\
+            BLENDING_LOOP(8, length)\
+            break;\
+        case 12:\
+            BLENDING_LOOP(12, length)\
+            break;\
+        }\
+    }\
+    for (; x < length; ++x) \
+        blend_pixel(dst[x], src[x]); \
+}
+
+// Basically blend src over dst with the const alpha defined as constAlphaVector.
+// nullVector, half, one, colorMask are constant accross the whole image/texture, and should be defined as:
+//const __m128i nullVector = _mm_set1_epi32(0);
+//const __m128i half = _mm_set1_epi16(0x80);
+//const __m128i one = _mm_set1_epi16(0xff);
+//const __m128i colorMask = _mm_set1_epi32(0x00ff00ff);
+//const __m128i alphaMask = _mm_set1_epi32(0xff000000);
+//
+// The computation being done is:
+// result = s + d * (1-alpha)
+// with shortcuts if fully opaque or fully transparent.
+#define BLEND_SOURCE_OVER_ARGB32_MAIN_SSSE3(dst, src, length, nullVector, half, one, colorMask, alphaMask) { \
+    int x = 0; \
+\
+    /* First, get dst aligned. */ \
+    ALIGNMENT_PROLOGUE_16BYTES(dst, x, length) { \
+        blend_pixel(dst[x], src[x]); \
+    } \
+\
+    const int minusOffsetToAlignSrcOn16Bytes = (reinterpret_cast<quintptr>(&(src[x])) >> 2) & 0x3;\
+\
+    if (!minusOffsetToAlignSrcOn16Bytes) {\
+        /* src is aligned, usual algorithm but with aligned operations.\
+           See the SSE2 version for more documentation on the algorithm itself. */\
+        const __m128i alphaShuffleMask = _mm_set_epi8(0xff,15,0xff,15,0xff,11,0xff,11,0xff,7,0xff,7,0xff,3,0xff,3);\
+        for (; x < length-3; x += 4) { \
+            const __m128i srcVector = _mm_load_si128((__m128i *)&src[x]); \
+            const __m128i srcVectorAlpha = _mm_and_si128(srcVector, alphaMask); \
+            if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, alphaMask)) == 0xffff) { \
+                _mm_store_si128((__m128i *)&dst[x], srcVector); \
+            } else if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, nullVector)) != 0xffff) { \
+                __m128i alphaChannel = _mm_shuffle_epi8(srcVector, alphaShuffleMask); \
+                alphaChannel = _mm_sub_epi16(one, alphaChannel); \
+                const __m128i dstVector = _mm_load_si128((__m128i *)&dst[x]); \
+                __m128i destMultipliedByOneMinusAlpha; \
+                BYTE_MUL_SSE2(destMultipliedByOneMinusAlpha, dstVector, alphaChannel, colorMask, half); \
+                const __m128i result = _mm_add_epi8(srcVector, destMultipliedByOneMinusAlpha); \
+                _mm_store_si128((__m128i *)&dst[x], result); \
+            } \
+        } /* end for() */\
+    } else if ((length - x) >= 8) {\
+        /* We use two vectors to extract the src: prevLoaded for the first pixels, lastLoaded for the current pixels. */\
+        __m128i srcVectorPrevLoaded = _mm_load_si128((__m128i *)&src[x - minusOffsetToAlignSrcOn16Bytes]);\
+        const int palignrOffset = minusOffsetToAlignSrcOn16Bytes << 2;\
+\
+        const __m128i alphaShuffleMask = _mm_set_epi8(0xff,15,0xff,15,0xff,11,0xff,11,0xff,7,0xff,7,0xff,3,0xff,3);\
+        switch (palignrOffset) {\
+        case 4:\
+            BLENDING_LOOP(4, length)\
+            break;\
+        case 8:\
+            BLENDING_LOOP(8, length)\
+            break;\
+        case 12:\
+            BLENDING_LOOP(12, length)\
+            break;\
+        }\
+    }\
+    for (; x < length; ++x) \
+        blend_pixel(dst[x], src[x]); \
+}
+
+void qt_blend_argb32_on_argb32_ssse3(uchar *destPixels, int dbpl,
+                                     const uchar *srcPixels, int sbpl,
+                                     int w, int h,
+                                     int const_alpha)
+{
+    const quint32 *src = (const quint32 *) srcPixels;
+    quint32 *dst = (quint32 *) destPixels;
+    if (const_alpha == 256) {
+        const __m128i alphaMask = _mm_set1_epi32(0xff000000);
+        const __m128i nullVector = _mm_setzero_si128();
+        const __m128i half = _mm_set1_epi16(0x80);
+        const __m128i one = _mm_set1_epi16(0xff);
+        const __m128i colorMask = _mm_set1_epi32(0x00ff00ff);
+
+        // We have to unrol the first row in order to deal with the load on unaligned data
+        // prior to the src pointer.
+        BLEND_SOURCE_OVER_ARGB32_FIRST_ROW_SSSE3(dst, src, w, nullVector, half, one, colorMask, alphaMask);
+        dst = (quint32 *)(((uchar *) dst) + dbpl);
+        src = (const quint32 *)(((const uchar *) src) + sbpl);
+
+        for (int y = 1; y < h; ++y) {
+            BLEND_SOURCE_OVER_ARGB32_MAIN_SSSE3(dst, src, w, nullVector, half, one, colorMask, alphaMask);
+            dst = (quint32 *)(((uchar *) dst) + dbpl);
+            src = (const quint32 *)(((const uchar *) src) + sbpl);
+        }
+    } else if (const_alpha != 0) {
+        // dest = (s + d * sia) * ca + d * cia
+        //      = s * ca + d * (sia * ca + cia)
+        //      = s * ca + d * (1 - sa*ca)
+        const_alpha = (const_alpha * 255) >> 8;
+        const __m128i nullVector = _mm_setzero_si128();
+        const __m128i half = _mm_set1_epi16(0x80);
+        const __m128i one = _mm_set1_epi16(0xff);
+        const __m128i colorMask = _mm_set1_epi32(0x00ff00ff);
+        const __m128i constAlphaVector = _mm_set1_epi16(const_alpha);
+        for (int y = 0; y < h; ++y) {
+            BLEND_SOURCE_OVER_ARGB32_WITH_CONST_ALPHA_SSE2(dst, src, w, nullVector, half, one, colorMask, constAlphaVector)
+            dst = (quint32 *)(((uchar *) dst) + dbpl);
+            src = (const quint32 *)(((const uchar *) src) + sbpl);
+        }
+    }
+}
+
+QT_END_NAMESPACE
+
+#endif // QT_HAVE_SSSE3
diff --git a/src/gui/painting/qwindowsurface_qws.cpp b/src/gui/painting/qwindowsurface_qws.cpp
index a816ed2..170668b 100644
--- a/src/gui/painting/qwindowsurface_qws.cpp
+++ b/src/gui/painting/qwindowsurface_qws.cpp
@@ -882,7 +882,11 @@ void QWSMemorySurface::beginPaint(const QRegion &rgn)
         const QVector<QRect> rects = rgn.rects();
         const QColor blank = Qt::transparent;
         for (QVector<QRect>::const_iterator it = rects.begin(); it != rects.end(); ++it) {
-            p.fillRect(*it, blank);
+            QRect r = *it;
+#ifdef Q_BACKINGSTORE_SUBSURFACES
+            r.translate(painterOffset());
+#endif
+            p.fillRect(r, blank);
         }
     }
     QWSWindowSurface::beginPaint(rgn);