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

8 files changed, 141 insertions, 100 deletions

diff --git a/src/gui/painting/qdrawhelper.cpp b/src/gui/painting/qdrawhelper.cpp
index 03ed597..bd5b0bd 100644
--- a/src/gui/painting/qdrawhelper.cpp
+++ b/src/gui/painting/qdrawhelper.cpp
@@ -659,19 +659,59 @@ const uint * QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, const
   interpolate 4 argb pixels with the distx and disty factor.
   distx and disty bust be between 0 and 16
  */
-static inline uint interpolate_4_pixels_16(uint tl, uint tr, uint bl, uint br, int distx, int disty, int idistx, int idisty)
-{
-    uint tlrb = ((tl & 0x00ff00ff) * idistx * idisty);
-    uint tlag = (((tl & 0xff00ff00) >> 8) * idistx * idisty);
-    uint trrb = ((tr & 0x00ff00ff) * distx * idisty);
-    uint trag = (((tr & 0xff00ff00) >> 8) * distx * idisty);
-    uint blrb = ((bl & 0x00ff00ff) * idistx * disty);
-    uint blag = (((bl & 0xff00ff00) >> 8) * idistx * disty);
-    uint brrb = ((br & 0x00ff00ff) * distx * disty);
-    uint brag = (((br & 0xff00ff00) >> 8) * distx * disty);
+static inline uint interpolate_4_pixels_16(uint tl, uint tr, uint bl, uint br, int distx, int disty)
+{
+    uint distxy = distx * disty;
+    //idistx * disty = (16-distx) * disty = 16*disty - distxy
+    //idistx * idisty = (16-distx) * (16-disty) = 16*16 - 16*distx -16*dity + distxy
+    uint tlrb = (tl & 0x00ff00ff)         * (16*16 - 16*distx - 16*disty + distxy);
+    uint tlag = ((tl & 0xff00ff00) >> 8)  * (16*16 - 16*distx - 16*disty + distxy);
+    uint trrb = ((tr & 0x00ff00ff)        * (distx*16 - distxy));
+    uint trag = (((tr & 0xff00ff00) >> 8) * (distx*16 - distxy));
+    uint blrb = ((bl & 0x00ff00ff)        * (disty*16 - distxy));
+    uint blag = (((bl & 0xff00ff00) >> 8) * (disty*16 - distxy));
+    uint brrb = ((br & 0x00ff00ff)        * (distxy));
+    uint brag = (((br & 0xff00ff00) >> 8) * (distxy));
     return (((tlrb + trrb + blrb + brrb) >> 8) & 0x00ff00ff) | ((tlag + trag + blag + brag) & 0xff00ff00);
 }
 
+#if defined(QT_ALWAYS_HAVE_SSE2)
+#define interpolate_4_pixels_16_sse2(tl, tr, bl, br, distx, disty, colorMask, v_256, b)  \
+{ \
+    const __m128i dxdy = _mm_mullo_epi16 (distx, disty); \
+    const __m128i distx_ = _mm_slli_epi16(distx, 4); \
+    const __m128i disty_ = _mm_slli_epi16(disty, 4); \
+    const __m128i idxidy =  _mm_add_epi16(dxdy, _mm_sub_epi16(v_256, _mm_add_epi16(distx_, disty_))); \
+    const __m128i dxidy =  _mm_sub_epi16(distx_, dxdy); \
+    const __m128i idxdy =  _mm_sub_epi16(disty_, dxdy); \
+ \
+    __m128i tlAG = _mm_srli_epi16(tl, 8); \
+    __m128i tlRB = _mm_and_si128(tl, colorMask); \
+    __m128i trAG = _mm_srli_epi16(tr, 8); \
+    __m128i trRB = _mm_and_si128(tr, colorMask); \
+    __m128i blAG = _mm_srli_epi16(bl, 8); \
+    __m128i blRB = _mm_and_si128(bl, colorMask); \
+    __m128i brAG = _mm_srli_epi16(br, 8); \
+    __m128i brRB = _mm_and_si128(br, colorMask); \
+ \
+    tlAG = _mm_mullo_epi16(tlAG, idxidy); \
+    tlRB = _mm_mullo_epi16(tlRB, idxidy); \
+    trAG = _mm_mullo_epi16(trAG, dxidy); \
+    trRB = _mm_mullo_epi16(trRB, dxidy); \
+    blAG = _mm_mullo_epi16(blAG, idxdy); \
+    blRB = _mm_mullo_epi16(blRB, idxdy); \
+    brAG = _mm_mullo_epi16(brAG, dxdy); \
+    brRB = _mm_mullo_epi16(brRB, dxdy); \
+ \
+    /* Add the values, and shift to only keep 8 significant bits per colors */ \
+    __m128i rAG =_mm_add_epi16(_mm_add_epi16(tlAG, trAG), _mm_add_epi16(blAG, brAG)); \
+    __m128i rRB =_mm_add_epi16(_mm_add_epi16(tlRB, trRB), _mm_add_epi16(blRB, brRB)); \
+    rAG = _mm_andnot_si128(colorMask, rAG); \
+    rRB = _mm_srli_epi16(rRB, 8); \
+    _mm_storeu_si128((__m128i*)(b), _mm_or_si128(rAG, rRB)); \
+}
+#endif
+
 
 template<TextureBlendType blendType>
 Q_STATIC_TEMPLATE_FUNCTION inline void fetchTransformedBilinear_pixelBounds(int max, int l1, int l2, int &v1, int &v2)
@@ -721,7 +761,7 @@ const uint * QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Operator *
     const qreal cx = x + 0.5;
     const qreal cy = y + 0.5;
 
-    const uint *end = buffer + length;
+    uint *end = buffer + length;
     uint *b = buffer;
     if (data->fast_matrix) {
         // The increment pr x in the scanline
@@ -879,7 +919,75 @@ const uint * QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Operator *
                 const uchar *s1 = data->texture.scanLine(y1);
                 const uchar *s2 = data->texture.scanLine(y2);
                 int disty = (fy & 0x0000ffff) >> 12;
-                int idisty = 16 - disty;
+
+#if defined(QT_ALWAYS_HAVE_SSE2)
+                if (blendType != BlendTransformedBilinearTiled &&
+                    (format == QImage::Format_ARGB32_Premultiplied || format == QImage::Format_RGB32)) {
+
+                    //prolog to get into the bounds
+                    while (b < end) {
+                        int x1 = (fx >> 16);
+                        int x2;
+                        fetchTransformedBilinear_pixelBounds<blendType>(image_width, image_x1, image_x2, x1, x2);
+                        if (x1 != x2) //break if we are insided the bounds.
+                            break;
+                        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);
+                        int distx = (fx & 0x0000ffff) >> 12;
+                        *b = interpolate_4_pixels_16(tl, tr, bl, br, distx, disty);
+                        fx += fdx;
+                        ++b;
+                    }
+                    uint *boundedEnd;
+                    if (fdx > 0)
+                        boundedEnd = qMin(end, buffer + uint((image_x2 - (fx >> 16)) / data->m11));
+                    else
+                        boundedEnd = qMin(end, buffer + uint((image_x1 - (fx >> 16)) / data->m11));
+                    boundedEnd -= 3;
+
+                    const __m128i colorMask = _mm_set1_epi32(0x00ff00ff);
+                    //const __m128i distShuffleMask = _mm_set_epi8(13, 12, 13, 12, 9, 8, 9, 8, 5, 4, 5, 4, 1, 0, 1, 0);
+                    const __m128i v_256 = _mm_set1_epi16(256);
+                    const __m128i v_disty = _mm_set1_epi16(disty);
+                    __m128i v_fdx = _mm_set1_epi32(fdx*4);
+
+                    ptrdiff_t secondLine = reinterpret_cast<const uint *>(s2) - reinterpret_cast<const uint *>(s1);
+
+                    union Vect_buffer { __m128i vect; quint32 i[4]; };
+                    Vect_buffer v_fx;
+
+                    for (int i = 0; i < 4; i++) {
+                        v_fx.i[i] = fx;
+                        fx += fdx;
+                    }
+
+                    while (b < boundedEnd) {
+
+                        Vect_buffer tl, tr, bl, br;
+
+                        for (int i = 0; i < 4; i++) {
+                            int x1 = v_fx.i[i] >> 16;
+                            const uint *addr_tl = reinterpret_cast<const uint *>(s1) + x1;
+                            const uint *addr_tr = addr_tl + 1;
+                            tl.i[i] = *addr_tl;
+                            tr.i[i] = *addr_tr;
+                            bl.i[i] = *(addr_tl+secondLine);
+                            br.i[i] = *(addr_tr+secondLine);
+                        }
+                        __m128i v_distx = _mm_srli_epi16(v_fx.vect, 12);      //distx = (fx & 0x0000ffff) >> 12;
+                        //v_distx = _mm_shuffle_epi8(v_disty, distShuffleMask); //distx |= distx << 16;
+                        v_distx = _mm_shufflehi_epi16(v_distx, _MM_SHUFFLE(2,2,0,0));
+                        v_distx = _mm_shufflelo_epi16(v_distx, _MM_SHUFFLE(2,2,0,0));
+
+                        interpolate_4_pixels_16_sse2(tl.vect, tr.vect, bl.vect, br.vect, v_distx, v_disty, colorMask, v_256, b);
+                        b+=4;
+                        v_fx.vect = _mm_add_epi32(v_fx.vect, v_fdx);
+                    }
+                    fx = v_fx.i[0];
+                }
+#endif
                 while (b < end) {
                     int x1 = (fx >> 16);
                     int x2;
@@ -889,8 +997,7 @@ const uint * QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Operator *
                     uint bl = fetch(s2, x1, data->texture.colorTable);
                     uint br = fetch(s2, x2, data->texture.colorTable);
                     int distx = (fx & 0x0000ffff) >> 12;
-                    int idistx = 16 - distx;
-                    *b = interpolate_4_pixels_16(tl, tr, bl, br, distx, disty, idistx, idisty);
+                    *b = interpolate_4_pixels_16(tl, tr, bl, br, distx, disty);
                     fx += fdx;
                     ++b;
                 }
@@ -949,10 +1056,8 @@ const uint * QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Operator *
 
                     int distx = (fx & 0x0000ffff) >> 12;
                     int disty = (fy & 0x0000ffff) >> 12;
-                    int idistx = 16 - distx;
-                    int idisty = 16 - disty;
 
-                    *b = interpolate_4_pixels_16(tl, tr, bl, br, distx, disty, idistx, idisty);
+                    *b = interpolate_4_pixels_16(tl, tr, bl, br, distx, disty);
 
                     fx += fdx;
                     fy += fdy;
diff --git a/src/gui/painting/qdrawhelper_sse2.cpp b/src/gui/painting/qdrawhelper_sse2.cpp
index ba2ba32..f97d865 100644
--- a/src/gui/painting/qdrawhelper_sse2.cpp
+++ b/src/gui/painting/qdrawhelper_sse2.cpp
@@ -266,10 +266,10 @@ void qt_memfill32_sse2(quint32 *dest, quint32 value, int count)
 
     int n = (count128 + 3) / 4;
     switch (count128 & 0x3) {
-    case 0: do { _mm_store_si128(dst128++, value128);
-    case 3:      _mm_store_si128(dst128++, value128);
-    case 2:      _mm_store_si128(dst128++, value128);
-    case 1:      _mm_store_si128(dst128++, value128);
+    case 0: do { _mm_stream_si128(dst128++, value128);
+    case 3:      _mm_stream_si128(dst128++, value128);
+    case 2:      _mm_stream_si128(dst128++, value128);
+    case 1:      _mm_stream_si128(dst128++, value128);
     } while (--n > 0);
     }
 
@@ -300,11 +300,14 @@ void QT_FASTCALL comp_func_solid_SourceOver_sse2(uint *destPixels, int length, u
         const __m128i half = _mm_set1_epi16(0x80);
         const __m128i minusAlphaOfColorVector = _mm_set1_epi16(minusAlphaOfColor);
 
+        ALIGNMENT_PROLOGUE_16BYTES(dst, x, length)
+            destPixels[x] = color + BYTE_MUL(destPixels[x], minusAlphaOfColor);
+
         for (; x < length-3; x += 4) {
-            __m128i dstVector = _mm_loadu_si128((__m128i *)&dst[x]);
+            __m128i dstVector = _mm_load_si128((__m128i *)&dst[x]);
             BYTE_MUL_SSE2(dstVector, dstVector, minusAlphaOfColorVector, colorMask, half);
             dstVector = _mm_add_epi8(colorVector, dstVector);
-            _mm_storeu_si128((__m128i *)&dst[x], dstVector);
+            _mm_store_si128((__m128i *)&dst[x], dstVector);
         }
         for (;x < length; ++x)
             destPixels[x] = color + BYTE_MUL(destPixels[x], minusAlphaOfColor);
diff --git a/src/gui/painting/qdrawhelper_ssse3.cpp b/src/gui/painting/qdrawhelper_ssse3.cpp
index 4cb4089..fb5602e 100644
--- a/src/gui/painting/qdrawhelper_ssse3.cpp
+++ b/src/gui/painting/qdrawhelper_ssse3.cpp
@@ -79,69 +79,6 @@ inline static void blend_pixel(quint32 &dst, const quint32 src)
     }
 
 
-#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);
@@ -153,7 +90,7 @@ inline static void blend_pixel(quint32 &dst, const quint32 src)
 // 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) { \
+#define BLEND_SOURCE_OVER_ARGB32_SSSE3(dst, src, length, nullVector, half, one, colorMask, alphaMask) { \
     int x = 0; \
 \
     /* First, get dst aligned. */ \
@@ -218,14 +155,8 @@ void qt_blend_argb32_on_argb32_ssse3(uchar *destPixels, int dbpl,
         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);
+        for (int y = 0; y < h; ++y) {
+            BLEND_SOURCE_OVER_ARGB32_SSSE3(dst, src, w, nullVector, half, one, colorMask, alphaMask);
             dst = (quint32 *)(((uchar *) dst) + dbpl);
             src = (const quint32 *)(((const uchar *) src) + sbpl);
         }
diff --git a/src/gui/painting/qgraphicssystem_runtime.cpp b/src/gui/painting/qgraphicssystem_runtime.cpp
index 2828e9d..a9fbbee 100644
--- a/src/gui/painting/qgraphicssystem_runtime.cpp
+++ b/src/gui/painting/qgraphicssystem_runtime.cpp
@@ -94,7 +94,8 @@ QRuntimePixmapData::QRuntimePixmapData(const QRuntimeGraphicsSystem *gs, PixelTy
 
 QRuntimePixmapData::~QRuntimePixmapData()
 {
-    m_graphicsSystem->removePixmapData(this);
+    if (QApplicationPrivate::graphics_system)
+        m_graphicsSystem->removePixmapData(this);
     delete m_data;
 }
 
@@ -258,7 +259,8 @@ QRuntimeWindowSurface::QRuntimeWindowSurface(const QRuntimeGraphicsSystem *gs, Q
 
 QRuntimeWindowSurface::~QRuntimeWindowSurface()
 {
-    m_graphicsSystem->removeWindowSurface(this);
+    if (QApplicationPrivate::graphics_system)
+        m_graphicsSystem->removeWindowSurface(this);
 }
 
 QPaintDevice *QRuntimeWindowSurface::paintDevice()
diff --git a/src/gui/painting/qgraphicssystem_runtime_p.h b/src/gui/painting/qgraphicssystem_runtime_p.h
index 0232241..421fbeb 100644
--- a/src/gui/painting/qgraphicssystem_runtime_p.h
+++ b/src/gui/painting/qgraphicssystem_runtime_p.h
@@ -177,6 +177,7 @@ private:
 
     friend class QRuntimePixmapData;
     friend class QRuntimeWindowSurface;
+    friend class QMeeGoGraphicsSystem;
 };
 
 QT_END_NAMESPACE
diff --git a/src/gui/painting/qpaintengineex.cpp b/src/gui/painting/qpaintengineex.cpp
index 881bd6e..1e857e4 100644
--- a/src/gui/painting/qpaintengineex.cpp
+++ b/src/gui/painting/qpaintengineex.cpp
@@ -768,7 +768,7 @@ void QPaintEngineEx::drawRoundedRect(const QRectF &rect, qreal xRadius, qreal yR
         x1, y2 - (1 - KAPPA) * yRadius,
         x1, y2 - yRadius,
         x1, y1 + yRadius,                   // LineTo
-        x1, y1 + KAPPA * yRadius,           // CurveTo
+        x1, y1 + (1 - KAPPA) * yRadius,           // CurveTo
         x1 + (1 - KAPPA) * xRadius, y1,
         x1 + xRadius, y1
     };
diff --git a/src/gui/painting/qstroker.cpp b/src/gui/painting/qstroker.cpp
index eabbd8a..9cff339 100644
--- a/src/gui/painting/qstroker.cpp
+++ b/src/gui/painting/qstroker.cpp
@@ -609,7 +609,7 @@ void QStroker::joinPoints(qfixed focal_x, qfixed focal_y, const QLineF &nextLine
             }
             QLineF miterLine(QPointF(qt_fixed_to_real(focal_x),
                                      qt_fixed_to_real(focal_y)), isect);
-            if (miterLine.length() > qt_fixed_to_real(m_strokeWidth * m_miterLimit) / 2) {
+            if (type == QLineF::NoIntersection || miterLine.length() > qt_fixed_to_real(m_strokeWidth * m_miterLimit) / 2) {
                 emitLineTo(qt_real_to_fixed(nextLine.x1()),
                            qt_real_to_fixed(nextLine.y1()));
             } else {
diff --git a/src/gui/painting/qtransform.h b/src/gui/painting/qtransform.h
index 212a582..11f2673 100644
--- a/src/gui/painting/qtransform.h
+++ b/src/gui/painting/qtransform.h
@@ -64,7 +64,6 @@ class QVariant;
 
 class Q_GUI_EXPORT QTransform
 {
-    Q_ENUMS(TransformationType)
 public:
     enum TransformationType {
         TxNone      = 0x00,