SSE2 implementation of convert_ARGB_to_ARGB_PM_inplace for QImage

Merge-request: 725
Reviewed-by: Benjamin Poulain <benjamin.poulain@nokia.com>

4 files changed, 130 insertions, 2 deletions

diff --git a/src/gui/image/image.pri b/src/gui/image/image.pri
index 9f0c87e..3a02d56 100644
--- a/src/gui/image/image.pri
+++ b/src/gui/image/image.pri
@@ -92,3 +92,6 @@ contains(QT_CONFIG, jpeg):include($$PWD/qjpeghandler.pri)
 contains(QT_CONFIG, mng):include($$PWD/qmnghandler.pri)
 contains(QT_CONFIG, tiff):include($$PWD/qtiffhandler.pri)
 contains(QT_CONFIG, gif):include($$PWD/qgifhandler.pri)
+
+# SIMD
+SSE2_SOURCES += image/qimage_sse2.cpp
diff --git a/src/gui/image/qimage.cpp b/src/gui/image/qimage.cpp
index 79f266d..88a0366 100644
--- a/src/gui/image/qimage.cpp
+++ b/src/gui/image/qimage.cpp
@@ -58,6 +58,7 @@
 #include <private/qmemrotate_p.h>
 #include <private/qpixmapdata_p.h>
 #include <private/qimagescale_p.h>
+#include <private/qsimd_p.h>
 
 #include <qhash.h>
 
@@ -209,7 +210,7 @@ QImageData * QImageData::create(const QSize &size, QImage::Format format, int nu
         break;
     }
 
-    const int bytes_per_line = ((width * depth + 31) >> 5) << 2; // bytes per scanline (must be multiple of 8)
+    const int bytes_per_line = ((width * depth + 31) >> 5) << 2; // bytes per scanline (must be multiple of 4)
 
     // sanity check for potential overflows
     if (INT_MAX/depth < width
@@ -3630,7 +3631,7 @@ static const Image_Converter converter_map[QImage::NImageFormats][QImage::NImage
     } // Format_ARGB4444_Premultiplied
 };
 
-static const InPlace_Image_Converter inplace_converter_map[QImage::NImageFormats][QImage::NImageFormats] =
+static InPlace_Image_Converter inplace_converter_map[QImage::NImageFormats][QImage::NImageFormats] =
 {
     {
         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
@@ -3727,6 +3728,19 @@ static const InPlace_Image_Converter inplace_converter_map[QImage::NImageFormats
     } // Format_ARGB4444_Premultiplied
 };
 
+void qInitImageConversions()
+{
+    const uint features = qDetectCPUFeatures();
+    Q_UNUSED(features);
+
+#ifdef QT_HAVE_SSE2
+    if (features & SSE2) {
+        extern bool convert_ARGB_to_ARGB_PM_inplace_SSE2(QImageData *data, Qt::ImageConversionFlags);
+        inplace_converter_map[QImage::Format_ARGB32][QImage::Format_ARGB32_Premultiplied] = convert_ARGB_to_ARGB_PM_inplace_SSE2;
+    }
+#endif
+}
+
 /*!
     Returns a copy of the image in the given \a format.
 
diff --git a/src/gui/image/qimage_p.h b/src/gui/image/qimage_p.h
index f1a0c47..5272848 100644
--- a/src/gui/image/qimage_p.h
+++ b/src/gui/image/qimage_p.h
@@ -108,6 +108,8 @@ struct Q_GUI_EXPORT QImageData {        // internal image data
     QPaintEngine *paintEngine;
 };
 
+void qInitImageConversions();
+
 QT_END_NAMESPACE
 
 #endif
diff --git a/src/gui/image/qimage_sse2.cpp b/src/gui/image/qimage_sse2.cpp
new file mode 100644
index 0000000..e2b89b9
--- /dev/null
+++ b/src/gui/image/qimage_sse2.cpp
@@ -0,0 +1,109 @@
+/****************************************************************************
+**
+** 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.
+**
+**
+**
+**
+**
+**
+**
+**
+** $QT_END_LICENSE$
+**
+****************************************************************************/
+
+#include "qimage.h"
+#include <private/qimage_p.h>
+#include <private/qsimd_p.h>
+#include <private/qdrawhelper_p.h>
+#include <private/qdrawingprimitive_sse2_p.h>
+
+#ifdef QT_HAVE_SSE2
+
+QT_BEGIN_NAMESPACE
+
+bool convert_ARGB_to_ARGB_PM_inplace_SSE2(QImageData *data, Qt::ImageConversionFlags)
+{
+    Q_ASSERT(data->format == QImage::Format_ARGB32);
+
+    // extra pixels on each line
+    const int spare = data->width & 3;
+    // width in pixels of the pad at the end of each line
+    const int pad = (data->bytes_per_line >> 2) - data->width;
+    const int iter = data->width >> 2;
+    int height = data->height;
+
+    const __m128i alphaMask = _mm_set1_epi32(0xff000000);
+    const __m128i nullVector = _mm_setzero_si128();
+    const __m128i half = _mm_set1_epi16(0x80);
+    const __m128i colorMask = _mm_set1_epi32(0x00ff00ff);
+
+    __m128i *d = reinterpret_cast<__m128i*>(data->data);
+    while (height--) {
+        const __m128i *end = d + iter;
+
+        for (; d != end; ++d) {
+            const __m128i srcVector = _mm_loadu_si128(d);
+            const __m128i srcVectorAlpha = _mm_and_si128(srcVector, alphaMask);
+            if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, alphaMask)) == 0xffff) {
+                // opaque, data is unchanged
+            } else if (_mm_movemask_epi8(_mm_cmpeq_epi32(srcVectorAlpha, nullVector)) == 0xffff) {
+                // fully transparent
+                _mm_storeu_si128(d, nullVector);
+            } else {
+                __m128i alphaChannel = _mm_srli_epi32(srcVector, 24);
+                alphaChannel = _mm_or_si128(alphaChannel, _mm_slli_epi32(alphaChannel, 16));
+
+                __m128i result;
+                BYTE_MUL_SSE2(result, srcVector, alphaChannel, colorMask, half);
+                result = _mm_or_si128(_mm_andnot_si128(alphaMask, result), srcVectorAlpha);
+                _mm_storeu_si128(d, result);
+            }
+        }
+
+        QRgb *p = reinterpret_cast<QRgb*>(d);
+        QRgb *pe = p+spare;
+        for (; p != pe; ++p) {
+            if (*p < 0x00ffffff)
+                *p = 0;
+            else if (*p < 0xff000000)
+                *p = PREMUL(*p);
+        }
+
+        d = reinterpret_cast<__m128i*>(p+pad);
+    }
+
+    data->format = QImage::Format_ARGB32_Premultiplied;
+    return true;
+}
+
+QT_END_NAMESPACE
+
+#endif // QT_HAVE_SSE2