diff options
Diffstat (limited to 'src/gui/painting')
41 files changed, 4845 insertions, 2663 deletions
diff --git a/src/gui/painting/WEBGRADIENTS_LICENSE.txt b/src/gui/painting/WEBGRADIENTS_LICENSE.txt new file mode 100644 index 0000000000..1a4d527a4d --- /dev/null +++ b/src/gui/painting/WEBGRADIENTS_LICENSE.txt @@ -0,0 +1,21 @@ +MIT License + +Copyright (c) 2017 itmeo + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. diff --git a/src/gui/painting/painting.pri b/src/gui/painting/painting.pri index 29bef14f0c..c3585a4647 100644 --- a/src/gui/painting/painting.pri +++ b/src/gui/painting/painting.pri @@ -99,8 +99,13 @@ SOURCES += \ painting/qplatformbackingstore.cpp \ painting/qpathsimplifier.cpp +webgradients.files = painting/webgradients.binaryjson +webgradients.prefix = qgradient +webgradients.base = painting + RESOURCES += \ - painting/qpdf.qrc + painting/qpdf.qrc \ + webgradients darwin { HEADERS += painting/qcoregraphics_p.h @@ -122,7 +127,7 @@ SSE2_SOURCES += painting/qdrawhelper_sse2.cpp SSSE3_SOURCES += painting/qdrawhelper_ssse3.cpp SSE4_1_SOURCES += painting/qdrawhelper_sse4.cpp \ painting/qimagescale_sse4.cpp -AVX2_SOURCES += painting/qdrawhelper_avx2.cpp +ARCH_HASWELL_SOURCES += painting/qdrawhelper_avx2.cpp NEON_SOURCES += painting/qdrawhelper_neon.cpp painting/qimagescale_neon.cpp NEON_HEADERS += painting/qdrawhelper_neon_p.h diff --git a/src/gui/painting/qblendfunctions.cpp b/src/gui/painting/qblendfunctions.cpp index a4a091a29f..2dd5144e40 100644 --- a/src/gui/painting/qblendfunctions.cpp +++ b/src/gui/painting/qblendfunctions.cpp @@ -430,28 +430,28 @@ struct Blend_RGB32_on_RGB32_ConstAlpha { }; struct Blend_ARGB32_on_ARGB32_SourceAlpha { - inline void write(quint32 *dst, quint32 src) { - *dst = src + BYTE_MUL(*dst, qAlpha(~src)); + inline void write(quint32 *dst, quint32 src) + { + blend_pixel(*dst, src); } inline void flush(void *) {} }; struct Blend_ARGB32_on_ARGB32_SourceAndConstAlpha { - inline Blend_ARGB32_on_ARGB32_SourceAndConstAlpha(quint32 alpha) { + inline Blend_ARGB32_on_ARGB32_SourceAndConstAlpha(quint32 alpha) + { m_alpha = (alpha * 255) >> 8; - m_ialpha = 255 - m_alpha; } - inline void write(quint32 *dst, quint32 src) { - src = BYTE_MUL(src, m_alpha); - *dst = src + BYTE_MUL(*dst, qAlpha(~src)); + inline void write(quint32 *dst, quint32 src) + { + blend_pixel(*dst, src, m_alpha); } inline void flush(void *) {} quint32 m_alpha; - quint32 m_ialpha; }; void qt_scale_image_rgb32_on_rgb32(uchar *destPixels, int dbpl, diff --git a/src/gui/painting/qbrush.cpp b/src/gui/painting/qbrush.cpp index 5c13308d94..860653cc4c 100644 --- a/src/gui/painting/qbrush.cpp +++ b/src/gui/painting/qbrush.cpp @@ -46,9 +46,13 @@ #include "qvariant.h" #include "qline.h" #include "qdebug.h" +#include <QtCore/qjsondocument.h> +#include <QtCore/qjsonarray.h> #include <QtCore/qcoreapplication.h> #include "private/qhexstring_p.h" #include <QtCore/qnumeric.h> +#include <QtCore/qfile.h> +#include <QtCore/qmutex.h> QT_BEGIN_NAMESPACE @@ -1073,7 +1077,10 @@ QDataStream &operator<<(QDataStream &s, const QBrush &b) s << type_as_int; if (s.version() >= QDataStream::Qt_4_3) { s << int(gradient->spread()); - s << int(gradient->coordinateMode()); + QGradient::CoordinateMode co_mode = gradient->coordinateMode(); + if (s.version() < QDataStream::Qt_5_12 && co_mode == QGradient::ObjectMode) + co_mode = QGradient::ObjectBoundingMode; + s << int(co_mode); } if (s.version() >= QDataStream::Qt_4_5) @@ -1330,6 +1337,72 @@ QGradient::QGradient() { } +/*! + \enum QGradient::Preset + \since 5.12 + + This enum specifies a set of predefined presets for QGradient, + based on the gradients from https://webgradients.com/. +*/ + +/*! + \fn QGradient::QGradient(QGradient::Preset preset) + \since 5.12 + + Constructs a gradient based on a predefined \a preset. + + The coordinate mode of the resulting gradient is + QGradient::ObjectMode, allowing the preset + to be applied to arbitrary object sizes. +*/ +QGradient::QGradient(Preset preset) + : QGradient() +{ + static QHash<int, QGradient> cachedPresets; + static QMutex cacheMutex; + QMutexLocker locker(&cacheMutex); + if (cachedPresets.contains(preset)) { + const QGradient &cachedPreset = cachedPresets.value(preset); + m_type = cachedPreset.m_type; + m_data = cachedPreset.m_data; + m_stops = cachedPreset.m_stops; + m_spread = cachedPreset.m_spread; + dummy = cachedPreset.dummy; + } else { + static QJsonDocument jsonPresets = []() { + QFile webGradients(QLatin1String(":/qgradient/webgradients.binaryjson")); + webGradients.open(QFile::ReadOnly); + return QJsonDocument::fromBinaryData(webGradients.readAll()); + }(); + + const QJsonValue presetData = jsonPresets[preset - 1]; + if (!presetData.isObject()) + return; + + m_type = LinearGradient; + setCoordinateMode(ObjectMode); + setSpread(PadSpread); + + const QJsonValue start = presetData[QLatin1Literal("start")]; + const QJsonValue end = presetData[QLatin1Literal("end")]; + m_data.linear.x1 = start[QLatin1Literal("x")].toDouble(); + m_data.linear.y1 = start[QLatin1Literal("y")].toDouble(); + m_data.linear.x2 = end[QLatin1Literal("x")].toDouble(); + m_data.linear.y2 = end[QLatin1Literal("y")].toDouble(); + + for (const QJsonValue &stop : presetData[QLatin1String("stops")].toArray()) { + setColorAt(stop[QLatin1Literal("position")].toDouble(), + QColor(QRgb(stop[QLatin1Literal("color")].toInt()))); + } + + cachedPresets.insert(preset, *this); + } +} + +QT_END_NAMESPACE +static void initGradientPresets() { Q_INIT_RESOURCE(qmake_webgradients); } +Q_CONSTRUCTOR_FUNCTION(initGradientPresets); +QT_BEGIN_NAMESPACE /*! \enum QGradient::Type @@ -1492,14 +1565,19 @@ QGradientStops QGradient::stops() const \value LogicalMode This is the default mode. The gradient coordinates are specified logical space just like the object coordinates. + \value ObjectMode In this mode the gradient coordinates are + relative to the bounding rectangle of the object being drawn, with + (0,0) in the top left corner, and (1,1) in the bottom right corner + of the object's bounding rectangle. This value was added in Qt + 5.12. \value StretchToDeviceMode In this mode the gradient coordinates are relative to the bounding rectangle of the paint device, with (0,0) in the top left corner, and (1,1) in the bottom right corner of the paint device. - \value ObjectBoundingMode In this mode the gradient coordinates are - relative to the bounding rectangle of the object being drawn, with - (0,0) in the top left corner, and (1,1) in the bottom right corner - of the object's bounding rectangle. + \value ObjectBoundingMode This mode is the same as ObjectMode, except that + the {QBrush::transform()} {brush transform}, if any, is applied relative to + the logical space instead of the object space. This enum value is + deprecated and should not be used in new code. */ /*! diff --git a/src/gui/painting/qbrush.h b/src/gui/painting/qbrush.h index e5cff9cc9b..aee51c526d 100644 --- a/src/gui/painting/qbrush.h +++ b/src/gui/painting/qbrush.h @@ -194,7 +194,8 @@ public: enum CoordinateMode { LogicalMode, StretchToDeviceMode, - ObjectBoundingMode + ObjectBoundingMode, + ObjectMode }; Q_ENUM(CoordinateMode) @@ -203,7 +204,180 @@ public: ComponentInterpolation }; + enum Preset { + WarmFlame = 1, + NightFade = 2, + SpringWarmth = 3, + JuicyPeach = 4, + YoungPassion = 5, + LadyLips = 6, + SunnyMorning = 7, + RainyAshville = 8, + FrozenDreams = 9, + WinterNeva = 10, + DustyGrass = 11, + TemptingAzure = 12, + HeavyRain = 13, + AmyCrisp = 14, + MeanFruit = 15, + DeepBlue = 16, + RipeMalinka = 17, + CloudyKnoxville = 18, + MalibuBeach = 19, + NewLife = 20, + TrueSunset = 21, + MorpheusDen = 22, + RareWind = 23, + NearMoon = 24, + WildApple = 25, + SaintPetersburg = 26, + PlumPlate = 28, + EverlastingSky = 29, + HappyFisher = 30, + Blessing = 31, + SharpeyeEagle = 32, + LadogaBottom = 33, + LemonGate = 34, + ItmeoBranding = 35, + ZeusMiracle = 36, + OldHat = 37, + StarWine = 38, + HappyAcid = 41, + AwesomePine = 42, + NewYork = 43, + ShyRainbow = 44, + MixedHopes = 46, + FlyHigh = 47, + StrongBliss = 48, + FreshMilk = 49, + SnowAgain = 50, + FebruaryInk = 51, + KindSteel = 52, + SoftGrass = 53, + GrownEarly = 54, + SharpBlues = 55, + ShadyWater = 56, + DirtyBeauty = 57, + GreatWhale = 58, + TeenNotebook = 59, + PoliteRumors = 60, + SweetPeriod = 61, + WideMatrix = 62, + SoftCherish = 63, + RedSalvation = 64, + BurningSpring = 65, + NightParty = 66, + SkyGlider = 67, + HeavenPeach = 68, + PurpleDivision = 69, + AquaSplash = 70, + SpikyNaga = 72, + LoveKiss = 73, + CleanMirror = 75, + PremiumDark = 76, + ColdEvening = 77, + CochitiLake = 78, + SummerGames = 79, + PassionateBed = 80, + MountainRock = 81, + DesertHump = 82, + JungleDay = 83, + PhoenixStart = 84, + OctoberSilence = 85, + FarawayRiver = 86, + AlchemistLab = 87, + OverSun = 88, + PremiumWhite = 89, + MarsParty = 90, + EternalConstance = 91, + JapanBlush = 92, + SmilingRain = 93, + CloudyApple = 94, + BigMango = 95, + HealthyWater = 96, + AmourAmour = 97, + RiskyConcrete = 98, + StrongStick = 99, + ViciousStance = 100, + PaloAlto = 101, + HappyMemories = 102, + MidnightBloom = 103, + Crystalline = 104, + PartyBliss = 106, + ConfidentCloud = 107, + LeCocktail = 108, + RiverCity = 109, + FrozenBerry = 110, + ChildCare = 112, + FlyingLemon = 113, + NewRetrowave = 114, + HiddenJaguar = 115, + AboveTheSky = 116, + Nega = 117, + DenseWater = 118, + Seashore = 120, + MarbleWall = 121, + CheerfulCaramel = 122, + NightSky = 123, + MagicLake = 124, + YoungGrass = 125, + ColorfulPeach = 126, + GentleCare = 127, + PlumBath = 128, + HappyUnicorn = 129, + AfricanField = 131, + SolidStone = 132, + OrangeJuice = 133, + GlassWater = 134, + NorthMiracle = 136, + FruitBlend = 137, + MillenniumPine = 138, + HighFlight = 139, + MoleHall = 140, + SpaceShift = 142, + ForestInei = 143, + RoyalGarden = 144, + RichMetal = 145, + JuicyCake = 146, + SmartIndigo = 147, + SandStrike = 148, + NorseBeauty = 149, + AquaGuidance = 150, + SunVeggie = 151, + SeaLord = 152, + BlackSea = 153, + GrassShampoo = 154, + LandingAircraft = 155, + WitchDance = 156, + SleeplessNight = 157, + AngelCare = 158, + CrystalRiver = 159, + SoftLipstick = 160, + SaltMountain = 161, + PerfectWhite = 162, + FreshOasis = 163, + StrictNovember = 164, + MorningSalad = 165, + DeepRelief = 166, + SeaStrike = 167, + NightCall = 168, + SupremeSky = 169, + LightBlue = 170, + MindCrawl = 171, + LilyMeadow = 172, + SugarLollipop = 173, + SweetDessert = 174, + MagicRay = 175, + TeenParty = 176, + FrozenHeat = 177, + GagarinView = 178, + FabledSunset = 179, + PerfectBlue = 180, + }; + Q_ENUM(Preset) + QGradient(); + QGradient(Preset); Type type() const { return m_type; } diff --git a/src/gui/painting/qcolor.cpp b/src/gui/painting/qcolor.cpp index c55bcb12c9..1d7375d1df 100644 --- a/src/gui/painting/qcolor.cpp +++ b/src/gui/painting/qcolor.cpp @@ -304,11 +304,6 @@ static const int rgbTblSize = sizeof(rgbTbl) / sizeof(RGBData); #undef rgb -#if defined(Q_CC_MSVC) && _MSC_VER < 1600 -inline bool operator<(const RGBData &data1, const RGBData &data2) -{ return qstrcmp(data1.name, data2.name) < 0; } -#endif - inline bool operator<(const char *name, const RGBData &data) { return qstrcmp(name, data.name) < 0; } inline bool operator<(const RGBData &data, const char *name) diff --git a/src/gui/painting/qcolorprofile_p.h b/src/gui/painting/qcolorprofile_p.h index ca1786ee6d..425e9abace 100644 --- a/src/gui/painting/qcolorprofile_p.h +++ b/src/gui/painting/qcolorprofile_p.h @@ -55,6 +55,11 @@ #include <QtGui/qrgb.h> #include <QtGui/qrgba64.h> +#if defined(__SSE2__) +#include <emmintrin.h> +#elif defined(__ARM_NEON__) || defined(__ARM_NEON) +#include <arm_neon.h> +#endif QT_BEGIN_NAMESPACE class Q_GUI_EXPORT QColorProfile @@ -67,82 +72,165 @@ public: QRgba64 toLinear64(QRgb rgb32) const { - ushort r = m_toLinear[qRed(rgb32) << 4]; - ushort g = m_toLinear[qGreen(rgb32) << 4]; - ushort b = m_toLinear[qBlue(rgb32) << 4]; +#if defined(__SSE2__) + __m128i v = _mm_cvtsi32_si128(rgb32); + v = _mm_unpacklo_epi8(v, _mm_setzero_si128()); + const __m128i vidx = _mm_slli_epi16(v, 4); + const int ridx = _mm_extract_epi16(vidx, 2); + const int gidx = _mm_extract_epi16(vidx, 1); + const int bidx = _mm_extract_epi16(vidx, 0); + v = _mm_slli_epi16(v, 8); // a * 256 + v = _mm_insert_epi16(v, m_toLinear[ridx], 0); + v = _mm_insert_epi16(v, m_toLinear[gidx], 1); + v = _mm_insert_epi16(v, m_toLinear[bidx], 2); + v = _mm_add_epi16(v, _mm_srli_epi16(v, 8)); + QRgba64 rgba64; + _mm_storel_epi64(reinterpret_cast<__m128i *>(&rgba64), v); + return rgba64; +#elif (defined(__ARM_NEON__) || defined(__ARM_NEON)) && Q_BYTE_ORDER == Q_LITTLE_ENDIAN + uint8x8_t v8 = vreinterpret_u8_u32(vmov_n_u32(rgb32)); + uint16x4_t v16 = vget_low_u16(vmovl_u8(v8)); + const uint16x4_t vidx = vshl_n_u16(v16, 4); + const int ridx = vget_lane_u16(vidx, 2); + const int gidx = vget_lane_u16(vidx, 1); + const int bidx = vget_lane_u16(vidx, 0); + v16 = vshl_n_u16(v16, 8); // a * 256 + v16 = vset_lane_u16(m_toLinear[ridx], v16, 0); + v16 = vset_lane_u16(m_toLinear[gidx], v16, 1); + v16 = vset_lane_u16(m_toLinear[bidx], v16, 2); + v16 = vadd_u16(v16, vshr_n_u16(v16, 8)); + return QRgba64::fromRgba64(vget_lane_u64(vreinterpret_u64_u16(v16), 0)); +#else + uint r = m_toLinear[qRed(rgb32) << 4]; + uint g = m_toLinear[qGreen(rgb32) << 4]; + uint b = m_toLinear[qBlue(rgb32) << 4]; r = r + (r >> 8); g = g + (g >> 8); b = b + (b >> 8); return QRgba64::fromRgba64(r, g, b, qAlpha(rgb32) * 257); +#endif } QRgb toLinear(QRgb rgb32) const { - uchar r = (m_toLinear[qRed(rgb32) << 4] + 0x80) >> 8; - uchar g = (m_toLinear[qGreen(rgb32) << 4] + 0x80) >> 8; - uchar b = (m_toLinear[qBlue(rgb32) << 4] + 0x80) >> 8; - return qRgba(r, g, b, qAlpha(rgb32)); + return convertWithTable(rgb32, m_toLinear); } QRgba64 toLinear(QRgba64 rgb64) const { - ushort r = rgb64.red(); - ushort g = rgb64.green(); - ushort b = rgb64.blue(); - r = r - (r >> 8); - g = g - (g >> 8); - b = b - (b >> 8); - r = m_toLinear[r >> 4]; - g = m_toLinear[g >> 4]; - b = m_toLinear[b >> 4]; - r = r + (r >> 8); - g = g + (g >> 8); - b = b + (b >> 8); - return QRgba64::fromRgba64(r, g, b, rgb64.alpha()); + return convertWithTable(rgb64, m_toLinear); } QRgb fromLinear64(QRgba64 rgb64) const { - ushort r = rgb64.red(); - ushort g = rgb64.green(); - ushort b = rgb64.blue(); +#if defined(__SSE2__) + __m128i v = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(&rgb64)); + v = _mm_sub_epi16(v, _mm_srli_epi16(v, 8)); + const __m128i vidx = _mm_srli_epi16(v, 4); + const int ridx = _mm_extract_epi16(vidx, 0); + const int gidx = _mm_extract_epi16(vidx, 1); + const int bidx = _mm_extract_epi16(vidx, 2); + v = _mm_insert_epi16(v, m_fromLinear[ridx], 2); + v = _mm_insert_epi16(v, m_fromLinear[gidx], 1); + v = _mm_insert_epi16(v, m_fromLinear[bidx], 0); + v = _mm_add_epi16(v, _mm_set1_epi16(0x80)); + v = _mm_srli_epi16(v, 8); + v = _mm_packus_epi16(v, v); + return _mm_cvtsi128_si32(v); +#elif (defined(__ARM_NEON__) || defined(__ARM_NEON)) && Q_BYTE_ORDER == Q_LITTLE_ENDIAN + uint16x4_t v = vreinterpret_u16_u64(vmov_n_u64(rgb64)); + v = vsub_u16(v, vshr_n_u16(v, 8)); + const uint16x4_t vidx = vshr_n_u16(v, 4); + const int ridx = vget_lane_u16(vidx, 0); + const int gidx = vget_lane_u16(vidx, 1); + const int bidx = vget_lane_u16(vidx, 2); + v = vset_lane_u16(m_fromLinear[ridx], v, 2); + v = vset_lane_u16(m_fromLinear[gidx], v, 1); + v = vset_lane_u16(m_fromLinear[bidx], v, 0); + uint8x8_t v8 = vrshrn_n_u16(vcombine_u16(v, v), 8); + return vget_lane_u32(vreinterpret_u32_u8(v8), 0); +#else + uint a = rgb64.alpha(); + uint r = rgb64.red(); + uint g = rgb64.green(); + uint b = rgb64.blue(); + a = a - (a >> 8); r = r - (r >> 8); g = g - (g >> 8); b = b - (b >> 8); + a = (a + 0x80) >> 8; r = (m_fromLinear[r >> 4] + 0x80) >> 8; g = (m_fromLinear[g >> 4] + 0x80) >> 8; b = (m_fromLinear[b >> 4] + 0x80) >> 8; - return qRgba(r, g, b, rgb64.alpha8()); + return (a << 24) | (r << 16) | (g << 8) | b; +#endif } QRgb fromLinear(QRgb rgb32) const { - uchar r = (m_fromLinear[qRed(rgb32) << 4] + 0x80) >> 8; - uchar g = (m_fromLinear[qGreen(rgb32) << 4] + 0x80) >> 8; - uchar b = (m_fromLinear[qBlue(rgb32) << 4] + 0x80) >> 8; - return qRgba(r, g, b, qAlpha(rgb32)); + return convertWithTable(rgb32, m_fromLinear); } QRgba64 fromLinear(QRgba64 rgb64) const { + return convertWithTable(rgb64, m_fromLinear); + } + +private: + QColorProfile() { } + + Q_ALWAYS_INLINE static QRgb convertWithTable(QRgb rgb32, const ushort *table) + { + const int r = (table[qRed(rgb32) << 4] + 0x80) >> 8; + const int g = (table[qGreen(rgb32) << 4] + 0x80) >> 8; + const int b = (table[qBlue(rgb32) << 4] + 0x80) >> 8; + return (rgb32 & 0xff000000) | (r << 16) | (g << 8) | b; + } + Q_ALWAYS_INLINE static QRgba64 convertWithTable(QRgba64 rgb64, const ushort *table) + { +#if defined(__SSE2__) + __m128i v = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(&rgb64)); + v = _mm_sub_epi16(v, _mm_srli_epi16(v, 8)); + const __m128i vidx = _mm_srli_epi16(v, 4); + const int ridx = _mm_extract_epi16(vidx, 2); + const int gidx = _mm_extract_epi16(vidx, 1); + const int bidx = _mm_extract_epi16(vidx, 0); + v = _mm_insert_epi16(v, table[ridx], 2); + v = _mm_insert_epi16(v, table[gidx], 1); + v = _mm_insert_epi16(v, table[bidx], 0); + v = _mm_add_epi16(v, _mm_srli_epi16(v, 8)); + QRgba64 rgba64; + _mm_storel_epi64(reinterpret_cast<__m128i *>(&rgba64), v); + return rgba64; +#elif (defined(__ARM_NEON__) || defined(__ARM_NEON)) && Q_BYTE_ORDER == Q_LITTLE_ENDIAN + uint16x4_t v = vreinterpret_u16_u64(vmov_n_u64(rgb64)); + v = vsub_u16(v, vshr_n_u16(v, 8)); + const uint16x4_t vidx = vshr_n_u16(v, 4); + const int ridx = vget_lane_u16(vidx, 2); + const int gidx = vget_lane_u16(vidx, 1); + const int bidx = vget_lane_u16(vidx, 0); + v = vset_lane_u16(table[ridx], v, 2); + v = vset_lane_u16(table[gidx], v, 1); + v = vset_lane_u16(table[bidx], v, 0); + v = vadd_u16(v, vshr_n_u16(v, 8)); + return QRgba64::fromRgba64(vget_lane_u64(vreinterpret_u64_u16(v), 0)); +#else ushort r = rgb64.red(); ushort g = rgb64.green(); ushort b = rgb64.blue(); r = r - (r >> 8); g = g - (g >> 8); b = b - (b >> 8); - r = m_fromLinear[r >> 4]; - g = m_fromLinear[g >> 4]; - b = m_fromLinear[b >> 4]; + r = table[r >> 4]; + g = table[g >> 4]; + b = table[b >> 4]; r = r + (r >> 8); g = g + (g >> 8); b = b + (b >> 8); return QRgba64::fromRgba64(r, g, b, rgb64.alpha()); +#endif } -private: - QColorProfile() { } - // We translate to 0-65280 (255*256) instead to 0-65535 to make simple // shifting an accurate conversion. // We translate from 0-4080 (255*16) for the same speed up, and to keep diff --git a/src/gui/painting/qcompositionfunctions.cpp b/src/gui/painting/qcompositionfunctions.cpp index 339a9749b8..027bf23115 100644 --- a/src/gui/painting/qcompositionfunctions.cpp +++ b/src/gui/painting/qcompositionfunctions.cpp @@ -64,125 +64,354 @@ QT_BEGIN_NAMESPACE where the source is an array of pixels. */ -#if defined __SSE2__ -# define LOAD(ptr) _mm_loadl_epi64(reinterpret_cast<const __m128i *>(ptr)) +struct Argb32OperationsC +{ + typedef QRgb Type; + typedef quint8 Scalar; + typedef QRgb OptimalType; + typedef quint8 OptimalScalar; + + static const Type clear; + static bool isOpaque(Type val) + { return qAlpha(val) == 255; } + static bool isTransparent(Type val) + { return qAlpha(val) == 0; } + static Scalar scalarFrom8bit(uint8_t a) + { return a; } + static void memfill(Type *ptr, Type value, qsizetype len) + { qt_memfill32(ptr, value, len); } + static void memcpy(Type *Q_DECL_RESTRICT dest, const Type *Q_DECL_RESTRICT src, qsizetype len) + { ::memcpy(dest, src, len * sizeof(Type)); } + + static OptimalType load(const Type *ptr) + { return *ptr; } + static OptimalType convert(const Type &val) + { return val; } + static void store(Type *ptr, OptimalType value) + { *ptr = value; } + static OptimalType add(OptimalType a, OptimalType b) + { return a + b; } + static OptimalScalar add(OptimalScalar a, OptimalScalar b) + { return a + b; } + static OptimalType plus(OptimalType a, OptimalType b) + { return comp_func_Plus_one_pixel(a, b); } + static OptimalScalar alpha(OptimalType val) + { return qAlpha(val); } + static OptimalScalar invAlpha(OptimalScalar c) + { return 255 - c; } + static OptimalScalar invAlpha(OptimalType val) + { return alpha(~val); } + static OptimalScalar scalar(Scalar v) + { return v; } + static OptimalType multiplyAlpha8bit(OptimalType val, uint8_t a) + { return BYTE_MUL(val, a); } + static OptimalType interpolate8bit(OptimalType x, uint8_t a1, OptimalType y, uint8_t a2) + { return INTERPOLATE_PIXEL_255(x, a1, y, a2); } + static OptimalType multiplyAlpha(OptimalType val, OptimalScalar a) + { return BYTE_MUL(val, a); } + static OptimalScalar multiplyAlpha8bit(OptimalScalar val, uint8_t a) + { return qt_div_255(val * a); } + static OptimalType interpolate(OptimalType x, OptimalScalar a1, OptimalType y, OptimalScalar a2) + { return INTERPOLATE_PIXEL_255(x, a1, y, a2); } +}; + +const Argb32OperationsC::Type Argb32OperationsC::clear = 0; + +struct Rgba64OperationsBase +{ + typedef QRgba64 Type; + typedef quint16 Scalar; + + static const Type clear; + + static bool isOpaque(Type val) + { return val.isOpaque(); } + static bool isTransparent(Type val) + { return val.isTransparent(); } + static Scalar scalarFrom8bit(uint8_t a) + { return a * 257; } + + static void memfill(Type *ptr, Type value, qsizetype len) + { qt_memfill64((quint64*)ptr, value, len); } + static void memcpy(Type *Q_DECL_RESTRICT dest, const Type *Q_DECL_RESTRICT src, qsizetype len) + { ::memcpy(dest, src, len * sizeof(Type)); } +}; + +const Rgba64OperationsBase::Type Rgba64OperationsBase::clear = QRgba64::fromRgba64(0); + +struct Rgba64OperationsC : public Rgba64OperationsBase +{ + typedef QRgba64 OptimalType; + typedef quint16 OptimalScalar; + + static OptimalType load(const Type *ptr) + { return *ptr; } + static OptimalType convert(const Type &val) + { return val; } + static void store(Type *ptr, OptimalType value) + { *ptr = value; } + static OptimalType add(OptimalType a, OptimalType b) + { return QRgba64::fromRgba64((quint64)a + (quint64)b); } + static OptimalScalar add(OptimalScalar a, OptimalScalar b) + { return a + b; } + static OptimalType plus(OptimalType a, OptimalType b) + { return addWithSaturation(a, b); } + static OptimalScalar alpha(OptimalType val) + { return val.alpha(); } + static OptimalScalar invAlpha(Scalar c) + { return 65535 - c; } + static OptimalScalar invAlpha(OptimalType val) + { return 65535 - alpha(val); } + static OptimalScalar scalar(Scalar v) + { return v; } + static OptimalType multiplyAlpha8bit(OptimalType val, uint8_t a) + { return multiplyAlpha255(val, a); } + static OptimalScalar multiplyAlpha8bit(OptimalScalar val, uint8_t a) + { return qt_div_255(val * a); } + static OptimalType interpolate8bit(OptimalType x, uint8_t a1, OptimalType y, uint8_t a2) + { return interpolate255(x, a1, y, a2); } + static OptimalType multiplyAlpha(OptimalType val, OptimalScalar a) + { return multiplyAlpha65535(val, a); } + static OptimalType interpolate(OptimalType x, OptimalScalar a1, OptimalType y, OptimalScalar a2) + { return interpolate65535(x, a1, y, a2); } +}; + +#if defined(__SSE2__) +struct Rgba64OperationsSSE2 : public Rgba64OperationsBase +{ + typedef __m128i OptimalType; + typedef __m128i OptimalScalar; + + static OptimalType load(const Type *ptr) + { + return _mm_loadl_epi64(reinterpret_cast<const __m128i *>(ptr)); + } + static OptimalType convert(const Type &value) + { #ifdef Q_PROCESSOR_X86_64 -# define CONVERT(value) _mm_cvtsi64_si128(value) + return _mm_cvtsi64_si128(value); #else -# define CONVERT(value) LOAD(&value) + return load(&value); #endif -# define STORE(ptr, value) _mm_storel_epi64(reinterpret_cast<__m128i *>(ptr), value) -# define ADD(p, q) _mm_add_epi32(p, q) -# define ALPHA(c) _mm_shufflelo_epi16(c, _MM_SHUFFLE(3, 3, 3, 3)) -# define CONST(n) _mm_shufflelo_epi16(_mm_cvtsi32_si128(n), _MM_SHUFFLE(0, 0, 0, 0)) -# define INVALPHA(c) _mm_sub_epi32(CONST(65535), ALPHA(c)) -#elif defined __ARM_NEON__ -# define LOAD(ptr) vreinterpret_u16_u64(vld1_u64(reinterpret_cast<const uint64_t *>(ptr))) -# define CONVERT(value) vreinterpret_u16_u64(vmov_n_u64(value)) -# define STORE(ptr, value) vst1_u64(reinterpret_cast<uint64_t *>(ptr), vreinterpret_u64_u16(value)) -# define ADD(p, q) vadd_u16(p, q) -# define ALPHA(c) vdup_lane_u16(c, 3) -# define CONST(n) vdup_n_u16(n) -# define INVALPHA(c) vmvn_u16(ALPHA(c)) -#else -# define LOAD(ptr) *ptr -# define CONVERT(value) value -# define STORE(ptr, value) *ptr = value -# define ADD(p, q) (p + q) -# define ALPHA(c) (c).alpha() -# define CONST(n) n -# define INVALPHA(c) (65535 - ALPHA(c)) + } + static void store(Type *ptr, OptimalType value) + { + _mm_storel_epi64(reinterpret_cast<__m128i *>(ptr), value); + } + static OptimalType add(OptimalType a, OptimalType b) + { + return _mm_add_epi16(a, b); + } +// same as above: +// static OptimalScalar add(OptimalScalar a, OptimalScalar b) + static OptimalType plus(OptimalType a, OptimalType b) + { + return _mm_adds_epu16(a, b); + } + static OptimalScalar alpha(OptimalType c) + { + return _mm_shufflelo_epi16(c, _MM_SHUFFLE(3, 3, 3, 3)); + } + static OptimalScalar invAlpha(Scalar c) + { + return scalar(65535 - c); + } + static OptimalScalar invAlpha(OptimalType c) + { + return _mm_xor_si128(_mm_set1_epi16(-1), alpha(c)); + } + static OptimalScalar scalar(Scalar n) + { + return _mm_shufflelo_epi16(_mm_cvtsi32_si128(n), _MM_SHUFFLE(0, 0, 0, 0)); + } + static OptimalType multiplyAlpha8bit(OptimalType val, uint8_t a) + { + return multiplyAlpha255(val, a); + } +// same as above: +// static OptimalScalar multiplyAlpha8bit(OptimalScalar a, uint8_t a) + static OptimalType interpolate8bit(OptimalType x, uint8_t a1, OptimalType y, uint8_t a2) + { + return interpolate255(x, a1, y, a2); + } + static OptimalType multiplyAlpha(OptimalType val, OptimalScalar a) + { + return multiplyAlpha65535(val, a); + } + // a2 is const-ref because otherwise MSVC2015@x86 complains that it can't 16-byte align the argument. + static OptimalType interpolate(OptimalType x, OptimalScalar a1, OptimalType y, const OptimalScalar &a2) + { + return interpolate65535(x, a1, y, a2); + } +}; #endif +#if defined(__ARM_NEON__) +struct Rgba64OperationsNEON : public Rgba64OperationsBase +{ + typedef uint16x4_t OptimalType; + typedef uint16x4_t OptimalScalar; + + static OptimalType load(const Type *ptr) + { + return vreinterpret_u16_u64(vld1_u64(reinterpret_cast<const uint64_t *>(ptr))); + } + static OptimalType convert(const Type &val) + { + return vreinterpret_u16_u64(vmov_n_u64(val)); + } + static void store(Type *ptr, OptimalType value) + { + vst1_u64(reinterpret_cast<uint64_t *>(ptr), vreinterpret_u64_u16(value)); + } + static OptimalType add(OptimalType a, OptimalType b) + { + return vadd_u16(a, b); + } +// same as above: +// static OptimalScalar add(OptimalScalar a, OptimalScalar b) + static OptimalType plus(OptimalType a, OptimalType b) + { + return vqadd_u16(a, b); + } + static OptimalScalar alpha(OptimalType c) + { + return vdup_lane_u16(c, 3); + } + static OptimalScalar invAlpha(Scalar c) + { + return scalar(65535 - c); + } + static OptimalScalar invAlpha(OptimalType c) + { + return vmvn_u16(alpha(c)); + } + static OptimalScalar scalar(Scalar n) + { + return vdup_n_u16(n); + } + static OptimalType multiplyAlpha8bit(OptimalType val, uint8_t a) + { + return multiplyAlpha255(val, a); + } +// same as above: +// static OptimalScalar multiplyAlpha8bit(OptimalScalar a, uint8_t a) + static OptimalType interpolate8bit(OptimalType x, uint8_t a1, OptimalType y, uint8_t a2) + { + return interpolate255(x, a1, y, a2); + } + static OptimalType multiplyAlpha(OptimalType val, OptimalScalar a) + { + return multiplyAlpha65535(val, a); + } + static OptimalType interpolate(OptimalType x, OptimalScalar a1, OptimalType y, OptimalScalar a2) + { + return interpolate65535(x, a1, y, a2); + } +}; + +#endif + +typedef Argb32OperationsC Argb32Operations; +#if defined(__SSE2__) +typedef Rgba64OperationsSSE2 Rgba64Operations; +#elif defined(__ARM_NEON__) +typedef Rgba64OperationsNEON Rgba64Operations; +#else +typedef Rgba64OperationsC Rgba64Operations; +#endif /* result = 0 d = d * cia */ -void QT_FASTCALL comp_func_solid_Clear(uint *dest, int length, uint, uint const_alpha) +template<class Ops> +inline static void comp_func_Clear_template(typename Ops::Type *dest, int length, uint const_alpha) { - if (const_alpha == 255) { - qt_memfill32(dest, 0, length); - } else { + if (const_alpha == 255) + Ops::memfill(dest, Ops::clear, length); + else { uint ialpha = 255 - const_alpha; - for (int i = 0; i < length; ++i) - dest[i] = BYTE_MUL(dest[i], ialpha); + for (int i = 0; i < length; ++i) { + Ops::store(&dest[i], Ops::multiplyAlpha8bit(Ops::load(&dest[i]), ialpha)); + } } } +void QT_FASTCALL comp_func_solid_Clear(uint *dest, int length, uint, uint const_alpha) +{ + comp_func_Clear_template<Argb32Operations>(dest, length, const_alpha); +} + void QT_FASTCALL comp_func_solid_Clear_rgb64(QRgba64 *dest, int length, QRgba64, uint const_alpha) { - if (const_alpha == 255) { - qt_memfill64((quint64*)dest, 0, length); - } else { - uint ialpha = 255 - const_alpha; - for (int i = 0; i < length; ++i) - STORE(&dest[i], multiplyAlpha255(LOAD(&dest[i]), ialpha)); - } + comp_func_Clear_template<Rgba64Operations>(dest, length, const_alpha); } void QT_FASTCALL comp_func_Clear(uint *dest, const uint *, int length, uint const_alpha) { - comp_func_solid_Clear(dest, length, 0, const_alpha); + comp_func_Clear_template<Argb32Operations>(dest, length, const_alpha); } void QT_FASTCALL comp_func_Clear_rgb64(QRgba64 *dest, const QRgba64 *, int length, uint const_alpha) { - comp_func_solid_Clear_rgb64(dest, length, QRgba64(), const_alpha); + comp_func_Clear_template<Rgba64Operations>(dest, length, const_alpha); } /* result = s dest = s * ca + d * cia */ -void QT_FASTCALL comp_func_solid_Source(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_Source_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - if (const_alpha == 255) { - qt_memfill32(dest, color, length); - } else { - uint ialpha = 255 - const_alpha; - color = BYTE_MUL(color, const_alpha); + if (const_alpha == 255) + Ops::memfill(dest, color, length); + else { + const uint ialpha = 255 - const_alpha; + auto s = Ops::multiplyAlpha8bit(Ops::convert(color), const_alpha); for (int i = 0; i < length; ++i) { - dest[i] = color + BYTE_MUL(dest[i], ialpha); + auto d = Ops::multiplyAlpha8bit(Ops::load(&dest[i]), ialpha); + Ops::store(&dest[i], Ops::add(s, d)); } } } -void QT_FASTCALL comp_func_solid_Source_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +template<class Ops> +inline static void comp_func_Source_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) - qt_memfill64((quint64*)dest, color, length); + Ops::memcpy(dest, src, length); else { - uint ialpha = 255 - const_alpha; - auto c = multiplyAlpha255(CONVERT(color), const_alpha); + const uint ialpha = 255 - const_alpha; for (int i = 0; i < length; ++i) { - STORE(&dest[i], ADD(c, multiplyAlpha255(LOAD(&dest[i]), ialpha))); + auto s = Ops::load(src + i); + auto d = Ops::load(dest + i); + Ops::store(&dest[i], Ops::interpolate8bit(s, const_alpha, d, ialpha)); } } } +void QT_FASTCALL comp_func_solid_Source(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_Source_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_Source_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_Source_template<Rgba64Operations>(dest, length, color, const_alpha); +} + void QT_FASTCALL comp_func_Source(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - ::memcpy(dest, src, size_t(length) * sizeof(uint)); - } else { - uint ialpha = 255 - const_alpha; - for (int i = 0; i < length; ++i) { - dest[i] = INTERPOLATE_PIXEL_255(src[i], const_alpha, dest[i], ialpha); - } - } + comp_func_Source_template<Argb32Operations>(dest, src, length, const_alpha); } void QT_FASTCALL comp_func_Source_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) - ::memcpy(dest, src, size_t(length) * sizeof(quint64)); - else { - uint ialpha = 255 - const_alpha; - for (int i = 0; i < length; ++i) { - STORE(&dest[i], interpolate255(LOAD(&src[i]), const_alpha, LOAD(&dest[i]), ialpha)); - } - } + comp_func_Source_template<Rgba64Operations>(dest, src, length, const_alpha); } void QT_FASTCALL comp_func_solid_Destination(uint *, int, uint, uint) @@ -207,68 +436,66 @@ void QT_FASTCALL comp_func_Destination_rgb64(QRgba64 *, const QRgba64 *, int, ui = s * ca + d * (sia * ca + cia) = s * ca + d * (1 - sa*ca) */ -void QT_FASTCALL comp_func_solid_SourceOver(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_SourceOver_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - if ((const_alpha & qAlpha(color)) == 255) { - qt_memfill32(dest, color, length); - } else { + if (const_alpha == 255 && Ops::isOpaque(color)) + Ops::memfill(dest, color, length); + else { + auto c = Ops::convert(color); if (const_alpha != 255) - color = BYTE_MUL(color, const_alpha); + c = Ops::multiplyAlpha8bit(c, const_alpha); + auto cAlpha = Ops::invAlpha(c); for (int i = 0; i < length; ++i) { - dest[i] = color + BYTE_MUL(dest[i], qAlpha(~color)); + auto d = Ops::multiplyAlpha(Ops::load(&dest[i]), cAlpha); + Ops::store(&dest[i], Ops::add(c, d)); } } } -void QT_FASTCALL comp_func_solid_SourceOver_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +template<class Ops> +inline static void comp_func_SourceOver_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { - if (const_alpha == 255 && color.isOpaque()) { - qt_memfill64((quint64*)dest, color, length); + if (const_alpha == 255) { + for (int i = 0; i < length; ++i) { + auto c = src[i]; + if (Ops::isOpaque(c)) + Ops::store(&dest[i], Ops::convert(c)); + else if (!Ops::isTransparent(c)) { + auto s = Ops::convert(c); + auto d = Ops::multiplyAlpha(Ops::load(&dest[i]), Ops::invAlpha(s)); + Ops::store(&dest[i], Ops::add(s, d)); + } + } } else { - auto c = CONVERT(color); - if (const_alpha != 255) - c = multiplyAlpha255(c, const_alpha); - auto cAlpha = INVALPHA(c); for (int i = 0; i < length; ++i) { - STORE(&dest[i], ADD(c, multiplyAlpha65535(LOAD(&dest[i]), cAlpha))); + auto s = Ops::multiplyAlpha8bit(Ops::load(&src[i]), const_alpha); + auto d = Ops::multiplyAlpha(Ops::load(&dest[i]), Ops::invAlpha(s)); + Ops::store(&dest[i], Ops::add(s, d)); } } } +void QT_FASTCALL comp_func_solid_SourceOver(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_SourceOver_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_SourceOver_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_SourceOver_template<Rgba64Operations>(dest, length, color, const_alpha); +} + void QT_FASTCALL comp_func_SourceOver(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - uint s = src[i]; - if (s >= 0xff000000) - dest[i] = s; - else if (s != 0) - dest[i] = s + BYTE_MUL(dest[i], qAlpha(~s)); - } - } else { - for (int i = 0; i < length; ++i) { - uint s = BYTE_MUL(src[i], const_alpha); - dest[i] = s + BYTE_MUL(dest[i], qAlpha(~s)); - } - } + comp_func_SourceOver_template<Argb32Operations>(dest, src, length, const_alpha); } void QT_FASTCALL comp_func_SourceOver_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - QRgba64 s = src[i]; - if (s.isOpaque()) - dest[i] = s; - else if (!s.isTransparent()) - STORE(&dest[i], ADD(CONVERT(s), multiplyAlpha65535(LOAD(&dest[i]), 65535 - s.alpha()))); - } - } else { - for (int i = 0; i < length; ++i) { - auto s = multiplyAlpha255(LOAD(&src[i]), const_alpha); - STORE(&dest[i], ADD(s, multiplyAlpha65535(LOAD(&dest[i]), INVALPHA(s)))); - } - } + comp_func_SourceOver_template<Rgba64Operations>(dest, src, length, const_alpha); } /* @@ -276,128 +503,122 @@ void QT_FASTCALL comp_func_SourceOver_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const dest = (d + s * dia) * ca + d * cia = d + s * dia * ca */ -void QT_FASTCALL comp_func_solid_DestinationOver(uint *dest, int length, uint color, uint const_alpha) -{ - if (const_alpha != 255) - color = BYTE_MUL(color, const_alpha); - for (int i = 0; i < length; ++i) { - uint d = dest[i]; - dest[i] = d + BYTE_MUL(color, qAlpha(~d)); - } -} - -void QT_FASTCALL comp_func_solid_DestinationOver_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_DestinationOver_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - auto c = CONVERT(color); + auto c = Ops::convert(color); if (const_alpha != 255) - c = multiplyAlpha255(c, const_alpha); + c = Ops::multiplyAlpha8bit(c, const_alpha); for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - STORE(&dest[i], ADD(d, multiplyAlpha65535(c, INVALPHA(d)))); + auto d = Ops::load(&dest[i]); + auto s = Ops::multiplyAlpha(c, Ops::invAlpha(d)); + Ops::store(&dest[i], Ops::add(s, d)); } } -void QT_FASTCALL comp_func_DestinationOver(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +template<class Ops> +inline static void comp_func_DestinationOver_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - uint d = dest[i]; - dest[i] = d + BYTE_MUL(src[i], qAlpha(~d)); + auto d = Ops::load(&dest[i]); + auto s = Ops::multiplyAlpha(Ops::load(&src[i]), Ops::invAlpha(d)); + Ops::store(&dest[i], Ops::add(s, d)); } } else { for (int i = 0; i < length; ++i) { - uint d = dest[i]; - uint s = BYTE_MUL(src[i], const_alpha); - dest[i] = d + BYTE_MUL(s, qAlpha(~d)); + auto d = Ops::load(&dest[i]); + auto s = Ops::multiplyAlpha8bit(Ops::load(&src[i]), const_alpha); + s = Ops::multiplyAlpha(s, Ops::invAlpha(d)); + Ops::store(&dest[i], Ops::add(s, d)); } } } +void QT_FASTCALL comp_func_solid_DestinationOver(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_DestinationOver_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_DestinationOver_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_DestinationOver_template<Rgba64Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_DestinationOver(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +{ + comp_func_DestinationOver_template<Argb32Operations>(dest, src, length, const_alpha); +} + void QT_FASTCALL comp_func_DestinationOver_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - STORE(&dest[i], ADD(d, multiplyAlpha65535(LOAD(&src[i]), INVALPHA(d)))); - } - } else { - for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - auto s = multiplyAlpha255(LOAD(&src[i]), const_alpha); - STORE(&dest[i], ADD(d, multiplyAlpha65535(s, INVALPHA(d)))); - } - } + comp_func_DestinationOver_template<Rgba64Operations>(dest, src, length, const_alpha); } /* result = s * da dest = s * da * ca + d * cia */ -void QT_FASTCALL comp_func_solid_SourceIn(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_SourceIn_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { if (const_alpha == 255) { + auto c = Ops::convert(color); for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(color, qAlpha(dest[i])); + Ops::store(&dest[i], Ops::multiplyAlpha(c, Ops::alpha(Ops::load(&dest[i])))); } } else { - color = BYTE_MUL(color, const_alpha); - uint cia = 255 - const_alpha; + auto c = Ops::multiplyAlpha8bit(Ops::convert(color), const_alpha); + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); for (int i = 0; i < length; ++i) { - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(d), d, cia); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(c, Ops::alpha(d), d, cia)); } } } -void QT_FASTCALL comp_func_solid_SourceIn_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +template<class Ops> +inline static void comp_func_SourceIn_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { - auto c = CONVERT(color); for (int i = 0; i < length; ++i) { - STORE(&dest[i], multiplyAlpha65535(c, dest[i].alpha())); + auto s = Ops::load(&src[i]); + Ops::store(&dest[i], Ops::multiplyAlpha(s, Ops::alpha(Ops::load(&dest[i])))); } } else { - uint ca = const_alpha * 257; - auto cia = CONST(65535 - ca); - auto c = multiplyAlpha65535(CONVERT(color), ca); + auto ca = Ops::scalarFrom8bit(const_alpha); + auto cia = Ops::invAlpha(ca); + auto cav = Ops::scalar(ca); for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - STORE(&dest[i], interpolate65535(c, ALPHA(d), d, cia)); + auto d = Ops::load(&dest[i]); + auto s = Ops::multiplyAlpha(Ops::load(&src[i]), cav); + Ops::store(&dest[i], Ops::interpolate(s, Ops::alpha(d), d, cia)); } } } +void QT_FASTCALL comp_func_solid_SourceIn(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_SourceIn_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_SourceIn_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_SourceIn_template<Rgba64Operations>(dest, length, color, const_alpha); +} + void QT_FASTCALL comp_func_SourceIn(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(src[i], qAlpha(dest[i])); - } - } else { - uint cia = 255 - const_alpha; - for (int i = 0; i < length; ++i) { - uint d = dest[i]; - uint s = BYTE_MUL(src[i], const_alpha); - dest[i] = INTERPOLATE_PIXEL_255(s, qAlpha(d), d, cia); - } - } + comp_func_SourceIn_template<Argb32Operations>(dest, src, length, const_alpha); } void QT_FASTCALL comp_func_SourceIn_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - STORE(&dest[i], multiplyAlpha65535(LOAD(&src[i]), dest[i].alpha())); - } - } else { - uint ca = const_alpha * 257; - auto cia = CONST(65535 - ca); - for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - auto s = multiplyAlpha65535(LOAD(&src[i]), ca); - STORE(&dest[i], interpolate65535(s, ALPHA(d), d, cia)); - } - } + comp_func_SourceIn_template<Rgba64Operations>(dest, src, length, const_alpha); } /* @@ -405,128 +626,120 @@ void QT_FASTCALL comp_func_SourceIn_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const Q dest = d * sa * ca + d * cia = d * (sa * ca + cia) */ -void QT_FASTCALL comp_func_solid_DestinationIn(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_DestinationIn_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - uint a = qAlpha(color); + auto sa = Ops::alpha(Ops::convert(color)); if (const_alpha != 255) { - a = BYTE_MUL(a, const_alpha) + 255 - const_alpha; - } - for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(dest[i], a); + sa = Ops::multiplyAlpha8bit(sa, const_alpha); + sa = Ops::add(sa, Ops::invAlpha(Ops::scalarFrom8bit(const_alpha))); } -} -void QT_FASTCALL comp_func_solid_DestinationIn_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) -{ - uint a = color.alpha(); - uint ca64k = const_alpha * 257; - if (const_alpha != 255) - a = qt_div_65535(a * ca64k) + 65535 - ca64k; for (int i = 0; i < length; ++i) { - STORE(&dest[i], multiplyAlpha65535(LOAD(&dest[i]), a)); + Ops::store(&dest[i], Ops::multiplyAlpha(Ops::load(&dest[i]), sa)); } } -void QT_FASTCALL comp_func_DestinationIn(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +template<class Ops> +inline static void comp_func_DestinationIn_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(dest[i], qAlpha(src[i])); + auto a = Ops::alpha(Ops::load(&src[i])); + Ops::store(&dest[i], Ops::multiplyAlpha(Ops::load(&dest[i]), a)); } } else { - uint cia = 255 - const_alpha; + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); for (int i = 0; i < length; ++i) { - uint a = BYTE_MUL(qAlpha(src[i]), const_alpha) + cia; - dest[i] = BYTE_MUL(dest[i], a); + auto sa = Ops::multiplyAlpha8bit(Ops::alpha(Ops::load(&src[i])), const_alpha); + sa = Ops::add(sa, cia); + Ops::store(&dest[i], Ops::multiplyAlpha(Ops::load(&dest[i]), sa)); } } } +void QT_FASTCALL comp_func_solid_DestinationIn(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_DestinationIn_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_DestinationIn_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_DestinationIn_template<Rgba64Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_DestinationIn(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +{ + comp_func_DestinationIn_template<Argb32Operations>(dest, src, length, const_alpha); +} + void QT_FASTCALL comp_func_DestinationIn_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - STORE(&dest[i], multiplyAlpha65535(LOAD(&dest[i]), src[i].alpha())); - } - } else { - uint ca = const_alpha * 257; - uint cia = 65535 - ca; - for (int i = 0; i < length; ++i) { - uint a = qt_div_65535(src[i].alpha() * ca) + cia; - STORE(&dest[i], multiplyAlpha65535(LOAD(&dest[i]), a)); - } - } + comp_func_DestinationIn_template<Rgba64Operations>(dest, src, length, const_alpha); } /* result = s * dia dest = s * dia * ca + d * cia */ - -void QT_FASTCALL comp_func_solid_SourceOut(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_SourceOut_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { + auto c = Ops::convert(color); if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(color, qAlpha(~dest[i])); - } + for (int i = 0; i < length; ++i) + Ops::store(&dest[i], Ops::multiplyAlpha(c, Ops::invAlpha(Ops::load(&dest[i])))); } else { - color = BYTE_MUL(color, const_alpha); - uint cia = 255 - const_alpha; + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); + c = Ops::multiplyAlpha8bit(c, const_alpha); for (int i = 0; i < length; ++i) { - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(~d), d, cia); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(c, Ops::invAlpha(d), d, cia)); } } } -void QT_FASTCALL comp_func_solid_SourceOut_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +template<class Ops> +inline static void comp_func_SourceOut_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - dest[i] = multiplyAlpha65535(color, 65535 - dest[i].alpha()); + auto s = Ops::load(&src[i]); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::multiplyAlpha(s, Ops::invAlpha(d))); } } else { - uint ca = const_alpha * 257; - uint cia = 65535 - ca; - color = multiplyAlpha65535(color, ca); + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); for (int i = 0; i < length; ++i) { - QRgba64 d = dest[i]; - dest[i] = interpolate65535(color, 65535 - d.alpha(), d, cia); + auto s = Ops::multiplyAlpha8bit(Ops::load(&src[i]), const_alpha); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(s, Ops::invAlpha(d), d, cia)); } } } +void QT_FASTCALL comp_func_solid_SourceOut(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_SourceOut_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_SourceOut_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_SourceOut_template<Rgba64Operations>(dest, length, color, const_alpha); +} + void QT_FASTCALL comp_func_SourceOut(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(src[i], qAlpha(~dest[i])); - } - } else { - uint cia = 255 - const_alpha; - for (int i = 0; i < length; ++i) { - uint s = BYTE_MUL(src[i], const_alpha); - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(s, qAlpha(~d), d, cia); - } - } + comp_func_SourceOut_template<Argb32Operations>(dest, src, length, const_alpha); } void QT_FASTCALL comp_func_SourceOut_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - dest[i] = multiplyAlpha65535(src[i], 65535 - dest[i].alpha()); - } - } else { - uint ca = const_alpha * 257; - uint cia = 65535 - ca; - for (int i = 0; i < length; ++i) { - QRgba64 d = dest[i]; - QRgba64 s = multiplyAlpha65535(src[i], ca); - dest[i] = interpolate65535(s, 65535 - d.alpha(), d, cia); - } - } + comp_func_SourceOut_template<Rgba64Operations>(dest, src, length, const_alpha); } /* @@ -534,56 +747,58 @@ void QT_FASTCALL comp_func_SourceOut_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const dest = d * sia * ca + d * cia = d * (sia * ca + cia) */ -void QT_FASTCALL comp_func_solid_DestinationOut(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_DestinationOut_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - uint a = qAlpha(~color); - if (const_alpha != 255) - a = BYTE_MUL(a, const_alpha) + 255 - const_alpha; - for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(dest[i], a); + auto sai = Ops::invAlpha(Ops::convert(color)); + if (const_alpha != 255) { + sai = Ops::multiplyAlpha8bit(sai, const_alpha); + sai = Ops::add(sai, Ops::invAlpha(Ops::scalarFrom8bit(const_alpha))); } -} -void QT_FASTCALL comp_func_solid_DestinationOut_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) -{ - uint a = 65535 - color.alpha(); - uint ca64k = const_alpha * 257; - if (const_alpha != 255) - a = qt_div_65535(a * ca64k) + 65535 - ca64k; for (int i = 0; i < length; ++i) { - dest[i] = multiplyAlpha65535(dest[i], a); + Ops::store(&dest[i], Ops::multiplyAlpha(Ops::load(&dest[i]), sai)); } } -void QT_FASTCALL comp_func_DestinationOut(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +template<class Ops> +inline static void comp_func_DestinationOut_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - dest[i] = BYTE_MUL(dest[i], qAlpha(~src[i])); + auto sia = Ops::invAlpha(Ops::load(&src[i])); + Ops::store(&dest[i], Ops::multiplyAlpha(Ops::load(&dest[i]), sia)); } } else { - uint cia = 255 - const_alpha; + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); for (int i = 0; i < length; ++i) { - uint sia = BYTE_MUL(qAlpha(~src[i]), const_alpha) + cia; - dest[i] = BYTE_MUL(dest[i], sia); + auto sia = Ops::multiplyAlpha8bit(Ops::invAlpha(Ops::load(&src[i])), const_alpha); + sia = Ops::add(sia, cia); + Ops::store(&dest[i], Ops::multiplyAlpha(Ops::load(&dest[i]), sia)); } } } +void QT_FASTCALL comp_func_solid_DestinationOut(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_DestinationOut_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_DestinationOut_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_DestinationOut_template<Rgba64Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_DestinationOut(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +{ + comp_func_DestinationOut_template<Argb32Operations>(dest, src, length, const_alpha); +} + void QT_FASTCALL comp_func_DestinationOut_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - dest[i] = multiplyAlpha65535(dest[i], 65535 - src[i].alpha()); - } - } else { - uint ca = const_alpha * 257; - uint cia = 65535 - ca; - for (int i = 0; i < length; ++i) { - uint a = qt_div_65535((65535 - src[i].alpha()) * ca) + cia; - dest[i] = multiplyAlpha65535(dest[i], a); - } - } + comp_func_DestinationOut_template<Rgba64Operations>(dest, src, length, const_alpha); } /* @@ -592,59 +807,57 @@ void QT_FASTCALL comp_func_DestinationOut_rgb64(QRgba64 *Q_DECL_RESTRICT dest, c = s*ca * da + d * (sia*ca + cia) = s*ca * da + d * (1 - sa*ca) */ -void QT_FASTCALL comp_func_solid_SourceAtop(uint *dest, int length, uint color, uint const_alpha) -{ - if (const_alpha != 255) { - color = BYTE_MUL(color, const_alpha); - } - uint sia = qAlpha(~color); - for (int i = 0; i < length; ++i) { - dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(dest[i]), dest[i], sia); - } -} - -void QT_FASTCALL comp_func_solid_SourceAtop_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_SourceAtop_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { + auto c = Ops::convert(color); if (const_alpha != 255) - color = multiplyAlpha255(color, const_alpha); - uint sia = 65535 - color.alpha(); + c = Ops::multiplyAlpha8bit(c, const_alpha); + auto sia = Ops::invAlpha(c); for (int i = 0; i < length; ++i) { - dest[i] = interpolate65535(color, dest[i].alpha(), dest[i], sia); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(c, Ops::alpha(d), d, sia)); } } -void QT_FASTCALL comp_func_SourceAtop(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +template<class Ops> +inline static void comp_func_SourceAtop_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - uint s = src[i]; - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(s, qAlpha(d), d, qAlpha(~s)); + auto s = Ops::load(&src[i]); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(s, Ops::alpha(d), d, Ops::invAlpha(s))); } } else { for (int i = 0; i < length; ++i) { - uint s = BYTE_MUL(src[i], const_alpha); - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(s, qAlpha(d), d, qAlpha(~s)); + auto s = Ops::multiplyAlpha8bit(Ops::load(&src[i]), const_alpha); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(s, Ops::alpha(d), d, Ops::invAlpha(s))); } } } +void QT_FASTCALL comp_func_solid_SourceAtop(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_SourceAtop_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_SourceAtop_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_SourceAtop_template<Rgba64Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_SourceAtop(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +{ + comp_func_SourceAtop_template<Argb32Operations>(dest, src, length, const_alpha); +} + void QT_FASTCALL comp_func_SourceAtop_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - QRgba64 s = src[i]; - QRgba64 d = dest[i]; - dest[i] = interpolate65535(s, d.alpha(), d, 65535 - s.alpha()); - } - } else { - for (int i = 0; i < length; ++i) { - QRgba64 s = multiplyAlpha255(src[i], const_alpha); - QRgba64 d = dest[i]; - dest[i] = interpolate65535(s, d.alpha(), d, 65535 - s.alpha()); - } - } + comp_func_SourceAtop_template<Rgba64Operations>(dest, src, length, const_alpha); } /* @@ -652,69 +865,63 @@ void QT_FASTCALL comp_func_SourceAtop_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const dest = d*sa*ca + s*dia*ca + d *cia = s*ca * dia + d * (sa*ca + cia) */ -void QT_FASTCALL comp_func_solid_DestinationAtop(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_DestinationAtop_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - uint a = qAlpha(color); + auto c = Ops::convert(color); + auto sa = Ops::alpha(c); if (const_alpha != 255) { - color = BYTE_MUL(color, const_alpha); - a = qAlpha(color) + 255 - const_alpha; - } - for (int i = 0; i < length; ++i) { - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(d, a, color, qAlpha(~d)); + c = Ops::multiplyAlpha8bit(c, const_alpha); + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); + sa = Ops::add(Ops::alpha(c), cia); } -} -void QT_FASTCALL comp_func_solid_DestinationAtop_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) -{ - uint a = color.alpha(); - if (const_alpha != 255) { - color = multiplyAlpha255(color, const_alpha); - a = color.alpha() + 65535 - (const_alpha * 257); - } for (int i = 0; i < length; ++i) { - QRgba64 d = dest[i]; - dest[i] = interpolate65535(d, a, color, 65535 - d.alpha()); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(c, Ops::invAlpha(d), d, sa)); } } -void QT_FASTCALL comp_func_DestinationAtop(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +template<class Ops> +inline static void comp_func_DestinationAtop_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - uint s = src[i]; - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(d, qAlpha(s), s, qAlpha(~d)); + auto s = Ops::load(&src[i]); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(s, Ops::invAlpha(d), d, Ops::alpha(s))); } } else { - uint cia = 255 - const_alpha; + auto cia = Ops::invAlpha(Ops::scalarFrom8bit(const_alpha)); for (int i = 0; i < length; ++i) { - uint s = BYTE_MUL(src[i], const_alpha); - uint d = dest[i]; - uint a = qAlpha(s) + cia; - dest[i] = INTERPOLATE_PIXEL_255(d, a, s, qAlpha(~d)); + auto s = Ops::multiplyAlpha8bit(Ops::load(&src[i]), const_alpha); + auto d = Ops::load(&dest[i]); + auto sa = Ops::add(Ops::alpha(s), cia); + Ops::store(&dest[i], Ops::interpolate(s, Ops::invAlpha(d), d, sa)); } } } +void QT_FASTCALL comp_func_solid_DestinationAtop(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_DestinationAtop_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_DestinationAtop_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_DestinationAtop_template<Rgba64Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_DestinationAtop(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +{ + comp_func_DestinationAtop_template<Argb32Operations>(dest, src, length, const_alpha); +} + void QT_FASTCALL comp_func_DestinationAtop_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - QRgba64 s = src[i]; - QRgba64 d = dest[i]; - dest[i] = interpolate65535(d, s.alpha(), s, 65535 - d.alpha()); - } - } else { - uint ca = const_alpha * 257; - uint cia = 65535 - ca; - for (int i = 0; i < length; ++i) { - QRgba64 s = multiplyAlpha65535(src[i], ca); - QRgba64 d = dest[i]; - uint a = s.alpha() + cia; - dest[i] = interpolate65535(d, a, s, 65535 - d.alpha()); - } - } + comp_func_DestinationAtop_template<Rgba64Operations>(dest, src, length, const_alpha); } /* @@ -723,62 +930,58 @@ void QT_FASTCALL comp_func_DestinationAtop_rgb64(QRgba64 *Q_DECL_RESTRICT dest, = s*ca * dia + d * (sia*ca + cia) = s*ca * dia + d * (1 - sa*ca) */ -void QT_FASTCALL comp_func_solid_XOR(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_solid_XOR_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { + auto c = Ops::convert(color); if (const_alpha != 255) - color = BYTE_MUL(color, const_alpha); - uint sia = qAlpha(~color); + c = Ops::multiplyAlpha8bit(c, const_alpha); + auto sia = Ops::invAlpha(c); for (int i = 0; i < length; ++i) { - uint d = dest[i]; - dest[i] = INTERPOLATE_PIXEL_255(color, qAlpha(~d), d, sia); + auto d = Ops::load(&dest[i]); + Ops::store(&dest[i], Ops::interpolate(c, Ops::invAlpha(d), d, sia)); } } -void QT_FASTCALL comp_func_solid_XOR_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) -{ - if (const_alpha != 255) - color = multiplyAlpha255(color, const_alpha); - auto s = CONVERT(color); - auto sia = CONST(65535 - color.alpha()); - for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - STORE(&dest[i], interpolate65535(s, INVALPHA(d), d, sia)); - } -} - -void QT_FASTCALL comp_func_XOR(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +template<class Ops> +inline static void comp_func_XOR_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - uint d = dest[i]; - uint s = src[i]; - dest[i] = INTERPOLATE_PIXEL_255(s, qAlpha(~d), d, qAlpha(~s)); + auto d = Ops::load(&dest[i]); + auto s = Ops::load(&src[i]); + Ops::store(&dest[i], Ops::interpolate(s, Ops::invAlpha(d), d, Ops::invAlpha(s))); } } else { for (int i = 0; i < length; ++i) { - uint d = dest[i]; - uint s = BYTE_MUL(src[i], const_alpha); - dest[i] = INTERPOLATE_PIXEL_255(s, qAlpha(~d), d, qAlpha(~s)); + auto d = Ops::load(&dest[i]); + auto s = Ops::multiplyAlpha8bit(Ops::load(&src[i]), const_alpha); + Ops::store(&dest[i], Ops::interpolate(s, Ops::invAlpha(d), d, Ops::invAlpha(s))); } } } +void QT_FASTCALL comp_func_solid_XOR(uint *dest, int length, uint color, uint const_alpha) +{ + comp_func_solid_XOR_template<Argb32Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_solid_XOR_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_XOR_template<Rgba64Operations>(dest, length, color, const_alpha); +} + +void QT_FASTCALL comp_func_XOR(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) +{ + comp_func_XOR_template<Argb32Operations>(dest, src, length, const_alpha); +} + void QT_FASTCALL comp_func_XOR_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - auto s = LOAD(&src[i]); - STORE(&dest[i], interpolate65535(s, INVALPHA(d), d, INVALPHA(s))); - } - } else { - for (int i = 0; i < length; ++i) { - auto d = LOAD(&dest[i]); - auto s = multiplyAlpha255(LOAD(&src[i]), const_alpha); - STORE(&dest[i], interpolate65535(s, INVALPHA(d), d, INVALPHA(s))); - } - } + comp_func_XOR_template<Rgba64Operations>(dest, src, length, const_alpha); } struct QFullCoverage { @@ -827,84 +1030,67 @@ static inline uint mix_alpha_rgb64(uint da, uint sa) Dca' = Sca.Da + Dca.Sa + Sca.(1 - Da) + Dca.(1 - Sa) = Sca + Dca */ -template <typename T> -Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_solid_Plus_impl(uint *dest, int length, uint color, const T &coverage) +template<class Ops> +inline static void comp_func_solid_Plus_template(typename Ops::Type *dest, int length, typename Ops::Type color, uint const_alpha) { - uint s = color; - - for (int i = 0; i < length; ++i) { - uint d = dest[i]; - d = comp_func_Plus_one_pixel(d, s); - coverage.store(&dest[i], d); + auto c = Ops::convert(color); + if (const_alpha == 255) { + for (int i = 0; i < length; ++i) { + auto d = Ops::load(&dest[i]); + d = Ops::plus(d, c); + Ops::store(&dest[i], d); + } + } else { + uint ia = 255 - const_alpha; + for (int i = 0; i < length; ++i) { + auto d = Ops::load(&dest[i]); + d = Ops::interpolate8bit(Ops::plus(d, c), const_alpha, d, ia); + Ops::store(&dest[i], d); + } } } -void QT_FASTCALL comp_func_solid_Plus(uint *dest, int length, uint color, uint const_alpha) +template<class Ops> +inline static void comp_func_Plus_template(typename Ops::Type *Q_DECL_RESTRICT dest, + const typename Ops::Type *Q_DECL_RESTRICT src, + int length, uint const_alpha) { - if (const_alpha == 255) - comp_func_solid_Plus_impl(dest, length, color, QFullCoverage()); - else - comp_func_solid_Plus_impl(dest, length, color, QPartialCoverage(const_alpha)); -} - -void QT_FASTCALL comp_func_solid_Plus_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) -{ - auto b = CONVERT(color); if (const_alpha == 255) { for (int i = 0; i < length; ++i) { - auto a = LOAD(&dest[i]); - a = addWithSaturation(a, b); - STORE(&dest[i], a); + auto d = Ops::load(&dest[i]); + auto s = Ops::load(&src[i]); + d = Ops::plus(d, s); + Ops::store(&dest[i], d); } } else { + uint ia = 255 - const_alpha; for (int i = 0; i < length; ++i) { - auto a = LOAD(&dest[i]); - auto d = addWithSaturation(a, b); - a = interpolate255(d, const_alpha, a, 255 - const_alpha); - STORE(&dest[i], a); + auto d = Ops::load(&dest[i]); + auto s = Ops::load(&src[i]); + d = Ops::interpolate8bit(Ops::plus(d, s), const_alpha, d, ia); + Ops::store(&dest[i], d); } } } -template <typename T> -Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Plus_impl(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, const T &coverage) +void QT_FASTCALL comp_func_solid_Plus(uint *dest, int length, uint color, uint const_alpha) { - for (int i = 0; i < length; ++i) { - uint d = dest[i]; - uint s = src[i]; - - d = comp_func_Plus_one_pixel(d, s); + comp_func_solid_Plus_template<Argb32Operations>(dest, length, color, const_alpha); +} - coverage.store(&dest[i], d); - } +void QT_FASTCALL comp_func_solid_Plus_rgb64(QRgba64 *dest, int length, QRgba64 color, uint const_alpha) +{ + comp_func_solid_Plus_template<Rgba64Operations>(dest, length, color, const_alpha); } void QT_FASTCALL comp_func_Plus(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) - comp_func_Plus_impl(dest, src, length, QFullCoverage()); - else - comp_func_Plus_impl(dest, src, length, QPartialCoverage(const_alpha)); + comp_func_Plus_template<Argb32Operations>(dest, src, length, const_alpha); } void QT_FASTCALL comp_func_Plus_rgb64(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha) { - if (const_alpha == 255) { - for (int i = 0; i < length; ++i) { - auto a = LOAD(&dest[i]); - auto b = LOAD(&src[i]); - a = addWithSaturation(a, b); - STORE(&dest[i], a); - } - } else { - for (int i = 0; i < length; ++i) { - auto a = LOAD(&dest[i]); - auto b = LOAD(&src[i]); - auto d = addWithSaturation(a, b); - a = interpolate255(d, const_alpha, a, 255 - const_alpha); - STORE(&dest[i], a); - } - } + comp_func_Plus_template<Rgba64Operations>(dest, src, length, const_alpha); } /* diff --git a/src/gui/painting/qcosmeticstroker_p.h b/src/gui/painting/qcosmeticstroker_p.h index 68f4e00cdc..082ddee30f 100644 --- a/src/gui/painting/qcosmeticstroker_p.h +++ b/src/gui/painting/qcosmeticstroker_p.h @@ -85,6 +85,7 @@ public: // used to avoid drop outs or duplicated points enum Direction { + NoDirection = 0, TopToBottom = 0x1, BottomToTop = 0x2, LeftToRight = 0x4, @@ -104,7 +105,7 @@ public: patternOffset(0), legacyRounding(false), current_span(0), - lastDir(LeftToRight), + lastDir(NoDirection), lastAxisAligned(false) { setup(); } diff --git a/src/gui/painting/qdrawhelper.cpp b/src/gui/painting/qdrawhelper.cpp index 0f2d163840..bbeb9fd9ea 100644 --- a/src/gui/painting/qdrawhelper.cpp +++ b/src/gui/painting/qdrawhelper.cpp @@ -1,6 +1,6 @@ /**************************************************************************** ** -** Copyright (C) 2016 The Qt Company Ltd. +** Copyright (C) 2018 The Qt Company Ltd. ** Copyright (C) 2016 Intel Corporation. ** Contact: https://www.qt.io/licensing/ ** @@ -73,9 +73,6 @@ enum { half_point = 1 << 15 }; -// must be multiple of 4 for easier SIMD implementations -static const int buffer_size = 2048; - template<QImage::Format> Q_DECL_CONSTEXPR uint redWidth(); template<QImage::Format> Q_DECL_CONSTEXPR uint redShift(); template<QImage::Format> Q_DECL_CONSTEXPR uint greenWidth(); @@ -94,6 +91,10 @@ template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_ARGB4444_Premultiplied> template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_ARGB8555_Premultiplied>() { return 5; } template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_ARGB8565_Premultiplied>() { return 5; } template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_ARGB6666_Premultiplied>() { return 6; } +template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_RGBX8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_RGBA8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint redWidth<QImage::Format_RGBA8888_Premultiplied>() { return 8; } + template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGB16>() { return 11; } template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGB444>() { return 8; } template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGB555>() { return 10; } @@ -103,6 +104,15 @@ template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_ARGB4444_Premultiplied> template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_ARGB8555_Premultiplied>() { return 18; } template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_ARGB8565_Premultiplied>() { return 19; } template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_ARGB6666_Premultiplied>() { return 12; } +#if Q_BYTE_ORDER == Q_BIG_ENDIAN +template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGBX8888>() { return 24; } +template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGBA8888>() { return 24; } +template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGBA8888_Premultiplied>() { return 24; } +#else +template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGBX8888>() { return 0; } +template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGBA8888>() { return 0; } +template<> Q_DECL_CONSTEXPR uint redShift<QImage::Format_RGBA8888_Premultiplied>() { return 0; } +#endif template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_RGB16>() { return 6; } template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_RGB444>() { return 4; } template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_RGB555>() { return 5; } @@ -112,6 +122,10 @@ template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_ARGB4444_Premultiplie template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_ARGB8555_Premultiplied>() { return 5; } template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_ARGB8565_Premultiplied>() { return 6; } template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_ARGB6666_Premultiplied>() { return 6; } +template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_RGBX8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_RGBA8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint greenWidth<QImage::Format_RGBA8888_Premultiplied>() { return 8; } + template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGB16>() { return 5; } template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGB444>() { return 4; } template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGB555>() { return 5; } @@ -121,6 +135,15 @@ template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_ARGB4444_Premultiplie template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_ARGB8555_Premultiplied>() { return 13; } template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_ARGB8565_Premultiplied>() { return 13; } template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_ARGB6666_Premultiplied>() { return 6; } +#if Q_BYTE_ORDER == Q_BIG_ENDIAN +template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGBX8888>() { return 16; } +template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGBA8888>() { return 16; } +template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGBA8888_Premultiplied>() { return 16; } +#else +template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGBX8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGBA8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint greenShift<QImage::Format_RGBA8888_Premultiplied>() { return 8; } +#endif template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_RGB16>() { return 5; } template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_RGB444>() { return 4; } template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_RGB555>() { return 5; } @@ -130,6 +153,10 @@ template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_ARGB4444_Premultiplied template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_ARGB8555_Premultiplied>() { return 5; } template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_ARGB8565_Premultiplied>() { return 5; } template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_ARGB6666_Premultiplied>() { return 6; } +template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_RGBX8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_RGBA8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint blueWidth<QImage::Format_RGBA8888_Premultiplied>() { return 8; } + template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGB16>() { return 0; } template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGB444>() { return 0; } template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGB555>() { return 0; } @@ -139,6 +166,15 @@ template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_ARGB4444_Premultiplied template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_ARGB8555_Premultiplied>() { return 8; } template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_ARGB8565_Premultiplied>() { return 8; } template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_ARGB6666_Premultiplied>() { return 0; } +#if Q_BYTE_ORDER == Q_BIG_ENDIAN +template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGBX8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGBA8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGBA8888_Premultiplied>() { return 8; } +#else +template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGBX8888>() { return 16; } +template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGBA8888>() { return 16; } +template<> Q_DECL_CONSTEXPR uint blueShift<QImage::Format_RGBA8888_Premultiplied>() { return 16; } +#endif template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_RGB16>() { return 0; } template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_RGB444>() { return 0; } template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_RGB555>() { return 0; } @@ -148,6 +184,10 @@ template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_ARGB4444_Premultiplie template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_ARGB8555_Premultiplied>() { return 8; } template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_ARGB8565_Premultiplied>() { return 8; } template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_ARGB6666_Premultiplied>() { return 6; } +template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_RGBX8888>() { return 0; } +template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_RGBA8888>() { return 8; } +template<> Q_DECL_CONSTEXPR uint alphaWidth<QImage::Format_RGBA8888_Premultiplied>() { return 8; } + template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGB16>() { return 0; } template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGB444>() { return 0; } template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGB555>() { return 0; } @@ -157,138 +197,200 @@ template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_ARGB4444_Premultiplie template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_ARGB8555_Premultiplied>() { return 0; } template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_ARGB8565_Premultiplied>() { return 0; } template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_ARGB6666_Premultiplied>() { return 18; } +#if Q_BYTE_ORDER == Q_BIG_ENDIAN +template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGBX8888>() { return 0; } +template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGBA8888>() { return 0; } +template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGBA8888_Premultiplied>() { return 0; } +#else +template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGBX8888>() { return 24; } +template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGBA8888>() { return 24; } +template<> Q_DECL_CONSTEXPR uint alphaShift<QImage::Format_RGBA8888_Premultiplied>() { return 24; } +#endif -template<QImage::Format> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel(); -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB16>() { return QPixelLayout::BPP16; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB444>() { return QPixelLayout::BPP16; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB555>() { return QPixelLayout::BPP16; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB666>() { return QPixelLayout::BPP24; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB888>() { return QPixelLayout::BPP24; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB4444_Premultiplied>() { return QPixelLayout::BPP16; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB8555_Premultiplied>() { return QPixelLayout::BPP24; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB8565_Premultiplied>() { return QPixelLayout::BPP24; } -template<> Q_DECL_CONSTEXPR QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB6666_Premultiplied>() { return QPixelLayout::BPP24; } +template<QImage::Format> constexpr QPixelLayout::BPP bitsPerPixel(); +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB16>() { return QPixelLayout::BPP16; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB444>() { return QPixelLayout::BPP16; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB555>() { return QPixelLayout::BPP16; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB666>() { return QPixelLayout::BPP24; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGB888>() { return QPixelLayout::BPP24; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB4444_Premultiplied>() { return QPixelLayout::BPP16; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB8555_Premultiplied>() { return QPixelLayout::BPP24; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB8565_Premultiplied>() { return QPixelLayout::BPP24; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_ARGB6666_Premultiplied>() { return QPixelLayout::BPP24; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGBX8888>() { return QPixelLayout::BPP32; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGBA8888>() { return QPixelLayout::BPP32; } +template<> constexpr QPixelLayout::BPP bitsPerPixel<QImage::Format_RGBA8888_Premultiplied>() { return QPixelLayout::BPP32; } -template<QImage::Format Format> -static const uint *QT_FASTCALL convertToRGB32(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +typedef const uint *(QT_FASTCALL *FetchPixelsFunc)(uint *buffer, const uchar *src, int index, int count); + +template <QPixelLayout::BPP bpp> static +uint QT_FASTCALL fetchPixel(const uchar *, int) { - auto conversion = [](uint s) { - // MSVC needs these constexpr defined in here otherwise it will create a capture. - Q_CONSTEXPR uint redMask = ((1 << redWidth<Format>()) - 1); - Q_CONSTEXPR uint greenMask = ((1 << greenWidth<Format>()) - 1); - Q_CONSTEXPR uint blueMask = ((1 << blueWidth<Format>()) - 1); + Q_UNREACHABLE(); + return 0; +} - Q_CONSTEXPR uchar redLeftShift = 8 - redWidth<Format>(); - Q_CONSTEXPR uchar greenLeftShift = 8 - greenWidth<Format>(); - Q_CONSTEXPR uchar blueLeftShift = 8 - blueWidth<Format>(); +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP1LSB>(const uchar *src, int index) +{ + return (src[index >> 3] >> (index & 7)) & 1; +} - Q_CONSTEXPR uchar redRightShift = 2 * redWidth<Format>() - 8; - Q_CONSTEXPR uchar greenRightShift = 2 * greenWidth<Format>() - 8; - Q_CONSTEXPR uchar blueRightShift = 2 * blueWidth<Format>() - 8; +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP1MSB>(const uchar *src, int index) +{ + return (src[index >> 3] >> (~index & 7)) & 1; +} - uint red = (s >> redShift<Format>()) & redMask; - uint green = (s >> greenShift<Format>()) & greenMask; - uint blue = (s >> blueShift<Format>()) & blueMask; +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP8>(const uchar *src, int index) +{ + return src[index]; +} - red = ((red << redLeftShift) | (red >> redRightShift)) << 16; - green = ((green << greenLeftShift) | (green >> greenRightShift)) << 8; - blue = (blue << blueLeftShift) | (blue >> blueRightShift); - return 0xff000000 | red | green | blue; - }; +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP16>(const uchar *src, int index) +{ + return reinterpret_cast<const quint16 *>(src)[index]; +} - UNALIASED_CONVERSION_LOOP(buffer, src, count, conversion); - return buffer; +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP24>(const uchar *src, int index) +{ + return reinterpret_cast<const quint24 *>(src)[index]; } -template<QImage::Format Format> -static const QRgba64 *QT_FASTCALL convertToRGB64(QRgba64 *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP32>(const uchar *src, int index) { - Q_CONSTEXPR uint redMask = ((1 << redWidth<Format>()) - 1); - Q_CONSTEXPR uint greenMask = ((1 << greenWidth<Format>()) - 1); - Q_CONSTEXPR uint blueMask = ((1 << blueWidth<Format>()) - 1); + return reinterpret_cast<const uint *>(src)[index]; +} - Q_CONSTEXPR uchar redLeftShift = 8 - redWidth<Format>(); - Q_CONSTEXPR uchar greenLeftShift = 8 - greenWidth<Format>(); - Q_CONSTEXPR uchar blueLeftShift = 8 - blueWidth<Format>(); +template <> +inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP64>(const uchar *src, int index) +{ + // We have to do the conversion in fetch to fit into a 32bit uint + QRgba64 c = reinterpret_cast<const QRgba64 *>(src)[index]; + return c.toArgb32(); +} - Q_CONSTEXPR uchar redRightShift = 2 * redWidth<Format>() - 8; - Q_CONSTEXPR uchar greenRightShift = 2 * greenWidth<Format>() - 8; - Q_CONSTEXPR uchar blueRightShift = 2 * blueWidth<Format>() - 8; +template <QPixelLayout::BPP bpp> +static quint64 QT_FASTCALL fetchPixel64(const uchar *src, int index) +{ + Q_STATIC_ASSERT(bpp != QPixelLayout::BPP64); + return fetchPixel<bpp>(src, index); +} - for (int i = 0; i < count; ++i) { - uint red = (src[i] >> redShift<Format>()) & redMask; - uint green = (src[i] >> greenShift<Format>()) & greenMask; - uint blue = (src[i] >> blueShift<Format>()) & blueMask; +template <QPixelLayout::BPP width> static +void QT_FASTCALL storePixel(uchar *dest, int index, uint pixel); - red = ((red << redLeftShift) | (red >> redRightShift)); - green = ((green << greenLeftShift) | (green >> greenRightShift)); - blue = (blue << blueLeftShift) | (blue >> blueRightShift); - buffer[i] = QRgba64::fromRgba(red, green, blue, 255); - } +template <> +inline void QT_FASTCALL storePixel<QPixelLayout::BPP16>(uchar *dest, int index, uint pixel) +{ + reinterpret_cast<quint16 *>(dest)[index] = quint16(pixel); +} - return buffer; +template <> +inline void QT_FASTCALL storePixel<QPixelLayout::BPP24>(uchar *dest, int index, uint pixel) +{ + reinterpret_cast<quint24 *>(dest)[index] = quint24(pixel); } +typedef uint (QT_FASTCALL *FetchPixelFunc)(const uchar *src, int index); + +static const FetchPixelFunc qFetchPixel[QPixelLayout::BPPCount] = { + 0, // BPPNone + fetchPixel<QPixelLayout::BPP1MSB>, // BPP1MSB + fetchPixel<QPixelLayout::BPP1LSB>, // BPP1LSB + fetchPixel<QPixelLayout::BPP8>, // BPP8 + fetchPixel<QPixelLayout::BPP16>, // BPP16 + fetchPixel<QPixelLayout::BPP24>, // BPP24 + fetchPixel<QPixelLayout::BPP32>, // BPP32 + fetchPixel<QPixelLayout::BPP64> // BPP64 +}; + template<QImage::Format Format> -static const uint *QT_FASTCALL convertARGBPMToARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static Q_ALWAYS_INLINE uint convertPixelToRGB32(uint s) { - Q_CONSTEXPR uint alphaMask = ((1 << alphaWidth<Format>()) - 1); Q_CONSTEXPR uint redMask = ((1 << redWidth<Format>()) - 1); Q_CONSTEXPR uint greenMask = ((1 << greenWidth<Format>()) - 1); Q_CONSTEXPR uint blueMask = ((1 << blueWidth<Format>()) - 1); - Q_CONSTEXPR uchar alphaLeftShift = 8 - alphaWidth<Format>(); Q_CONSTEXPR uchar redLeftShift = 8 - redWidth<Format>(); Q_CONSTEXPR uchar greenLeftShift = 8 - greenWidth<Format>(); Q_CONSTEXPR uchar blueLeftShift = 8 - blueWidth<Format>(); - Q_CONSTEXPR uchar alphaRightShift = 2 * alphaWidth<Format>() - 8; Q_CONSTEXPR uchar redRightShift = 2 * redWidth<Format>() - 8; Q_CONSTEXPR uchar greenRightShift = 2 * greenWidth<Format>() - 8; Q_CONSTEXPR uchar blueRightShift = 2 * blueWidth<Format>() - 8; - Q_CONSTEXPR bool mustMin = (alphaWidth<Format>() != redWidth<Format>()) || - (alphaWidth<Format>() != greenWidth<Format>()) || - (alphaWidth<Format>() != blueWidth<Format>()); + uint red = (s >> redShift<Format>()) & redMask; + uint green = (s >> greenShift<Format>()) & greenMask; + uint blue = (s >> blueShift<Format>()) & blueMask; - if (mustMin) { - for (int i = 0; i < count; ++i) { - uint alpha = (src[i] >> alphaShift<Format>()) & alphaMask; - uint red = (src[i] >> redShift<Format>()) & redMask; - uint green = (src[i] >> greenShift<Format>()) & greenMask; - uint blue = (src[i] >> blueShift<Format>()) & blueMask; - - alpha = (alpha << alphaLeftShift) | (alpha >> alphaRightShift); - red = qMin(alpha, (red << redLeftShift) | (red >> redRightShift)); - green = qMin(alpha, (green << greenLeftShift) | (green >> greenRightShift)); - blue = qMin(alpha, (blue << blueLeftShift) | (blue >> blueRightShift)); - buffer[i] = (alpha << 24) | (red << 16) | (green << 8) | blue; - } - } else { - for (int i = 0; i < count; ++i) { - uint alpha = (src[i] >> alphaShift<Format>()) & alphaMask; - uint red = (src[i] >> redShift<Format>()) & redMask; - uint green = (src[i] >> greenShift<Format>()) & greenMask; - uint blue = (src[i] >> blueShift<Format>()) & blueMask; - - alpha = ((alpha << alphaLeftShift) | (alpha >> alphaRightShift)) << 24; - red = ((red << redLeftShift) | (red >> redRightShift)) << 16; - green = ((green << greenLeftShift) | (green >> greenRightShift)) << 8; - blue = (blue << blueLeftShift) | (blue >> blueRightShift); - buffer[i] = alpha | red | green | blue; - } + red = ((red << redLeftShift) | (red >> redRightShift)) << 16; + green = ((green << greenLeftShift) | (green >> greenRightShift)) << 8; + blue = (blue << blueLeftShift) | (blue >> blueRightShift); + return 0xff000000 | red | green | blue; +} + +template<QImage::Format Format> +static void QT_FASTCALL convertToRGB32(uint *buffer, int count, const QVector<QRgb> *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToRGB32<Format>(buffer[i]); +} + +#if defined(__SSE2__) && !defined(__SSSE3__) && QT_COMPILER_SUPPORTS_SSSE3 +extern const uint * QT_FASTCALL fetchPixelsBPP24_ssse3(uint *dest, const uchar*src, int index, int count); +#endif + +template<QImage::Format Format> +static const uint *QT_FASTCALL fetchRGBToRGB32(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + constexpr QPixelLayout::BPP BPP = bitsPerPixel<Format>(); +#if defined(__SSE2__) && !defined(__SSSE3__) && QT_COMPILER_SUPPORTS_SSSE3 + if (BPP == QPixelLayout::BPP24 && qCpuHasFeature(SSSE3)) { + // With SSE2 can convertToRGB32 be vectorized, but it takes SSSE3 + // to vectorize the deforested version below. + fetchPixelsBPP24_ssse3(buffer, src, index, count); + convertToRGB32<Format>(buffer, count, nullptr); + return buffer; } +#endif + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToRGB32<Format>(fetchPixel<BPP>(src, index + i)); + return buffer; +} +template<QImage::Format Format> +static Q_ALWAYS_INLINE QRgba64 convertPixelToRGB64(uint s) +{ + return QRgba64::fromArgb32(convertPixelToRGB32<Format>(s)); +} + +template<QImage::Format Format> +static const QRgba64 *QT_FASTCALL convertToRGB64(QRgba64 *buffer, const uint *src, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToRGB64<Format>(src[i]); return buffer; } template<QImage::Format Format> -static const QRgba64 *QT_FASTCALL convertARGBPMToARGB64PM(QRgba64 *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static const QRgba64 *QT_FASTCALL fetchRGBToRGB64(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToRGB64<Format>(fetchPixel<bitsPerPixel<Format>()>(src, index + i)); + return buffer; +} + +template<QImage::Format Format> +static Q_ALWAYS_INLINE uint convertPixelToARGB32PM(uint s) { Q_CONSTEXPR uint alphaMask = ((1 << alphaWidth<Format>()) - 1); Q_CONSTEXPR uint redMask = ((1 << redWidth<Format>()) - 1); @@ -309,62 +411,101 @@ static const QRgba64 *QT_FASTCALL convertARGBPMToARGB64PM(QRgba64 *buffer, const (alphaWidth<Format>() != greenWidth<Format>()) || (alphaWidth<Format>() != blueWidth<Format>()); + uint alpha = (s >> alphaShift<Format>()) & alphaMask; + uint red = (s >> redShift<Format>()) & redMask; + uint green = (s >> greenShift<Format>()) & greenMask; + uint blue = (s >> blueShift<Format>()) & blueMask; + + alpha = (alpha << alphaLeftShift) | (alpha >> alphaRightShift); + red = (red << redLeftShift) | (red >> redRightShift); + green = (green << greenLeftShift) | (green >> greenRightShift); + blue = (blue << blueLeftShift) | (blue >> blueRightShift); + if (mustMin) { - for (int i = 0; i < count; ++i) { - uint alpha = (src[i] >> alphaShift<Format>()) & alphaMask; - uint red = (src[i] >> redShift<Format>()) & redMask; - uint green = (src[i] >> greenShift<Format>()) & greenMask; - uint blue = (src[i] >> blueShift<Format>()) & blueMask; - - alpha = (alpha << alphaLeftShift) | (alpha >> alphaRightShift); - red = qMin(alpha, (red << redLeftShift) | (red >> redRightShift)); - green = qMin(alpha, (green << greenLeftShift) | (green >> greenRightShift)); - blue = qMin(alpha, (blue << blueLeftShift) | (blue >> blueRightShift)); - buffer[i] = QRgba64::fromRgba(red, green, blue, alpha); - } - } else { - for (int i = 0; i < count; ++i) { - uint alpha = (src[i] >> alphaShift<Format>()) & alphaMask; - uint red = (src[i] >> redShift<Format>()) & redMask; - uint green = (src[i] >> greenShift<Format>()) & greenMask; - uint blue = (src[i] >> blueShift<Format>()) & blueMask; - - alpha = (alpha << alphaLeftShift) | (alpha >> alphaRightShift); - red = (red << redLeftShift) | (red >> redRightShift); - green = (green << greenLeftShift) | (green >> greenRightShift); - blue = (blue << blueLeftShift) | (blue >> blueRightShift); - buffer[i] = QRgba64::fromRgba(red, green, blue, alpha); - } + red = qMin(alpha, red); + green = qMin(alpha, green); + blue = qMin(alpha, blue); } + return (alpha << 24) | (red << 16) | (green << 8) | blue; +} + +template<QImage::Format Format> +static void QT_FASTCALL convertARGBPMToARGB32PM(uint *buffer, int count, const QVector<QRgb> *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToARGB32PM<Format>(buffer[i]); +} + +template<QImage::Format Format> +static const uint *QT_FASTCALL fetchARGBPMToARGB32PM(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + constexpr QPixelLayout::BPP BPP = bitsPerPixel<Format>(); +#if defined(__SSE2__) && !defined(__SSSE3__) && QT_COMPILER_SUPPORTS_SSSE3 + if (BPP == QPixelLayout::BPP24 && qCpuHasFeature(SSSE3)) { + // With SSE2 can convertToRGB32 be vectorized, but it takes SSSE3 + // to vectorize the deforested version below. + fetchPixelsBPP24_ssse3(buffer, src, index, count); + convertARGBPMToARGB32PM<Format>(buffer, count, nullptr); + return buffer; + } +#endif + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToARGB32PM<Format>(fetchPixel<BPP>(src, index + i)); + return buffer; +} + +template<QImage::Format Format> +static Q_ALWAYS_INLINE QRgba64 convertPixelToRGBA64PM(uint s) +{ + return QRgba64::fromArgb32(convertPixelToARGB32PM<Format>(s)); +} + +template<QImage::Format Format> +static const QRgba64 *QT_FASTCALL convertARGBPMToRGBA64PM(QRgba64 *buffer, const uint *src, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToRGB64<Format>(src[i]); + return buffer; +} + +template<QImage::Format Format> +static const QRgba64 *QT_FASTCALL fetchARGBPMToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + constexpr QPixelLayout::BPP bpp = bitsPerPixel<Format>(); + for (int i = 0; i < count; ++i) + buffer[i] = convertPixelToRGBA64PM<Format>(fetchPixel<bpp>(src, index + i)); return buffer; } template<QImage::Format Format, bool fromRGB> -static const uint *QT_FASTCALL convertRGBFromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *dither) +static void QT_FASTCALL storeRGBFromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *dither) { Q_CONSTEXPR uchar rWidth = redWidth<Format>(); Q_CONSTEXPR uchar gWidth = greenWidth<Format>(); Q_CONSTEXPR uchar bWidth = blueWidth<Format>(); + constexpr QPixelLayout::BPP BPP = bitsPerPixel<Format>(); // RGB32 -> RGB888 is not a precision loss. if (!dither || (rWidth == 8 && gWidth == 8 && bWidth == 8)) { - auto conversion = [](uint s) { - const uint c = fromRGB ? s : qUnpremultiply(s); - Q_CONSTEXPR uint rMask = (1 << redWidth<Format>()) - 1; - Q_CONSTEXPR uint gMask = (1 << greenWidth<Format>()) - 1; - Q_CONSTEXPR uint bMask = (1 << blueWidth<Format>()) - 1; - Q_CONSTEXPR uchar rRightShift = 24 - redWidth<Format>(); - Q_CONSTEXPR uchar gRightShift = 16 - greenWidth<Format>(); - Q_CONSTEXPR uchar bRightShift = 8 - blueWidth<Format>(); + Q_CONSTEXPR uint rMask = (1 << redWidth<Format>()) - 1; + Q_CONSTEXPR uint gMask = (1 << greenWidth<Format>()) - 1; + Q_CONSTEXPR uint bMask = (1 << blueWidth<Format>()) - 1; + Q_CONSTEXPR uchar rRightShift = 24 - redWidth<Format>(); + Q_CONSTEXPR uchar gRightShift = 16 - greenWidth<Format>(); + Q_CONSTEXPR uchar bRightShift = 8 - blueWidth<Format>(); + for (int i = 0; i < count; ++i) { + const uint c = fromRGB ? src[i] : qUnpremultiply(src[i]); const uint r = ((c >> rRightShift) & rMask) << redShift<Format>(); const uint g = ((c >> gRightShift) & gMask) << greenShift<Format>(); const uint b = ((c >> bRightShift) & bMask) << blueShift<Format>(); - return r | g | b; + storePixel<BPP>(dest, index + i, r | g | b); }; - UNALIASED_CONVERSION_LOOP(buffer, src, count, conversion); } else { // We do ordered dither by using a rounding conversion, but instead of // adding half of input precision, we add the adjusted result from the @@ -384,38 +525,39 @@ static const uint *QT_FASTCALL convertRGBFromARGB32PM(uint *buffer, const uint * r = (r + ((dr - r) >> rWidth) + 1) >> (8 - rWidth); g = (g + ((dg - g) >> gWidth) + 1) >> (8 - gWidth); b = (b + ((db - b) >> bWidth) + 1) >> (8 - bWidth); - buffer[i] = (r << redShift<Format>()) - | (g << greenShift<Format>()) - | (b << blueShift<Format>()); + const uint s = (r << redShift<Format>()) + | (g << greenShift<Format>()) + | (b << blueShift<Format>()); + storePixel<BPP>(dest, index + i, s); } } - return buffer; } template<QImage::Format Format, bool fromRGB> -static const uint *QT_FASTCALL convertARGBPMFromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *dither) +static void QT_FASTCALL storeARGBPMFromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *dither) { + constexpr QPixelLayout::BPP BPP = bitsPerPixel<Format>(); if (!dither) { - auto conversion = [](uint c) { - Q_CONSTEXPR uint aMask = (1 << alphaWidth<Format>()) - 1; - Q_CONSTEXPR uint rMask = (1 << redWidth<Format>()) - 1; - Q_CONSTEXPR uint gMask = (1 << greenWidth<Format>()) - 1; - Q_CONSTEXPR uint bMask = (1 << blueWidth<Format>()) - 1; - - Q_CONSTEXPR uchar aRightShift = 32 - alphaWidth<Format>(); - Q_CONSTEXPR uchar rRightShift = 24 - redWidth<Format>(); - Q_CONSTEXPR uchar gRightShift = 16 - greenWidth<Format>(); - Q_CONSTEXPR uchar bRightShift = 8 - blueWidth<Format>(); - - Q_CONSTEXPR uint aOpaque = aMask << alphaShift<Format>(); + Q_CONSTEXPR uint aMask = (1 << alphaWidth<Format>()) - 1; + Q_CONSTEXPR uint rMask = (1 << redWidth<Format>()) - 1; + Q_CONSTEXPR uint gMask = (1 << greenWidth<Format>()) - 1; + Q_CONSTEXPR uint bMask = (1 << blueWidth<Format>()) - 1; + + Q_CONSTEXPR uchar aRightShift = 32 - alphaWidth<Format>(); + Q_CONSTEXPR uchar rRightShift = 24 - redWidth<Format>(); + Q_CONSTEXPR uchar gRightShift = 16 - greenWidth<Format>(); + Q_CONSTEXPR uchar bRightShift = 8 - blueWidth<Format>(); + + Q_CONSTEXPR uint aOpaque = aMask << alphaShift<Format>(); + for (int i = 0; i < count; ++i) { + const uint c = src[i]; const uint a = fromRGB ? aOpaque : (((c >> aRightShift) & aMask) << alphaShift<Format>()); const uint r = ((c >> rRightShift) & rMask) << redShift<Format>(); const uint g = ((c >> gRightShift) & gMask) << greenShift<Format>(); const uint b = ((c >> bRightShift) & bMask) << blueShift<Format>(); - return a | r | g | b; + storePixel<BPP>(dest, index + i, a | r | g | b); }; - UNALIASED_CONVERSION_LOOP(buffer, src, count, conversion); } else { Q_CONSTEXPR uchar aWidth = alphaWidth<Format>(); Q_CONSTEXPR uchar rWidth = redWidth<Format>(); @@ -441,59 +583,147 @@ static const uint *QT_FASTCALL convertARGBPMFromARGB32PM(uint *buffer, const uin r = (r + ((dr - r) >> rWidth) + 1) >> (8 - rWidth); g = (g + ((dg - g) >> gWidth) + 1) >> (8 - gWidth); b = (b + ((db - b) >> bWidth) + 1) >> (8 - bWidth); - buffer[i] = (a << alphaShift<Format>()) - | (r << redShift<Format>()) - | (g << greenShift<Format>()) - | (b << blueShift<Format>()); + uint s = (a << alphaShift<Format>()) + | (r << redShift<Format>()) + | (g << greenShift<Format>()) + | (b << blueShift<Format>()); + storePixel<BPP>(dest, index + i, s); } } - return buffer; } +template<QImage::Format Format> +static void QT_FASTCALL rbSwap(uchar *dst, const uchar *src, int count) +{ + Q_CONSTEXPR uchar aWidth = alphaWidth<Format>(); + Q_CONSTEXPR uchar aShift = alphaShift<Format>(); + Q_CONSTEXPR uchar rWidth = redWidth<Format>(); + Q_CONSTEXPR uchar rShift = redShift<Format>(); + Q_CONSTEXPR uchar gWidth = greenWidth<Format>(); + Q_CONSTEXPR uchar gShift = greenShift<Format>(); + Q_CONSTEXPR uchar bWidth = blueWidth<Format>(); + Q_CONSTEXPR uchar bShift = blueShift<Format>(); #ifdef Q_COMPILER_CONSTEXPR + Q_STATIC_ASSERT(rWidth == bWidth); +#endif + Q_CONSTEXPR uint redBlueMask = (1 << rWidth) - 1; + Q_CONSTEXPR uint alphaGreenMask = (((1 << aWidth) - 1) << aShift) + | (((1 << gWidth) - 1) << gShift); + constexpr QPixelLayout::BPP bpp = bitsPerPixel<Format>(); + + for (int i = 0; i < count; ++i) { + const uint c = fetchPixel<bpp>(src, i); + const uint r = (c >> rShift) & redBlueMask; + const uint b = (c >> bShift) & redBlueMask; + const uint t = (c & alphaGreenMask) + | (r << bShift) + | (b << rShift); + storePixel<bpp>(dst, i, t); + } +} + +static void QT_FASTCALL rbSwap_rgb32(uchar *d, const uchar *s, int count) +{ + const uint *src = reinterpret_cast<const uint *>(s); + uint *dest = reinterpret_cast<uint *>(d); + for (int i = 0; i < count; ++i) { + const uint c = src[i]; + const uint ag = c & 0xff00ff00; + const uint rb = c & 0x00ff00ff; + dest[i] = ag | (rb << 16) | (rb >> 16); + } +} + +#if Q_BYTE_ORDER == Q_LITTLE_ENDIAN +template<> +void QT_FASTCALL rbSwap<QImage::Format_RGBA8888>(uchar *d, const uchar *s, int count) +{ + return rbSwap_rgb32(d, s, count); +} +#else +template<> +void QT_FASTCALL rbSwap<QImage::Format_RGBA8888>(uchar *d, const uchar *s, int count) +{ + const uint *src = reinterpret_cast<const uint *>(s); + uint *dest = reinterpret_cast<uint *>(d); + for (int i = 0; i < count; ++i) { + const uint c = src[i]; + const uint rb = c & 0xff00ff00; + const uint ga = c & 0x00ff00ff; + dest[i] = ga | (rb << 16) | (rb >> 16); + } +} +#endif + +static void QT_FASTCALL rbSwap_rgb30(uchar *d, const uchar *s, int count) +{ + const uint *src = reinterpret_cast<const uint *>(s); + uint *dest = reinterpret_cast<uint *>(d); + for (int i = 0; i < count; ++i) + dest[i] = qRgbSwapRgb30(src[i]); +} template<QImage::Format Format> Q_DECL_CONSTEXPR static inline QPixelLayout pixelLayoutRGB() { return QPixelLayout{ - uchar(redWidth<Format>()), uchar(redShift<Format>()), - uchar(greenWidth<Format>()), uchar(greenShift<Format>()), - uchar(blueWidth<Format>()), uchar(blueShift<Format>()), - 0, 0, - false, bitsPerPixel<Format>(), + false, + false, + bitsPerPixel<Format>(), + rbSwap<Format>, convertToRGB32<Format>, - convertRGBFromARGB32PM<Format, false>, - convertRGBFromARGB32PM<Format, true>, - convertToRGB64<Format> + convertToRGB64<Format>, + fetchRGBToRGB32<Format>, + fetchRGBToRGB64<Format>, + storeRGBFromARGB32PM<Format, false>, + storeRGBFromARGB32PM<Format, true> }; } template<QImage::Format Format> Q_DECL_CONSTEXPR static inline QPixelLayout pixelLayoutARGBPM() { return QPixelLayout{ - uchar(redWidth<Format>()), uchar(redShift<Format>()), - uchar(greenWidth<Format>()), uchar(greenShift<Format>()), - uchar(blueWidth<Format>()), uchar(blueShift<Format>()), - uchar(alphaWidth<Format>()), uchar(alphaShift<Format>()), - true, bitsPerPixel<Format>(), + true, + true, + bitsPerPixel<Format>(), + rbSwap<Format>, convertARGBPMToARGB32PM<Format>, - convertARGBPMFromARGB32PM<Format, false>, - convertARGBPMFromARGB32PM<Format, true>, - convertARGBPMToARGB64PM<Format> + convertARGBPMToRGBA64PM<Format>, + fetchARGBPMToARGB32PM<Format>, + fetchARGBPMToRGBA64PM<Format>, + storeARGBPMFromARGB32PM<Format, false>, + storeARGBPMFromARGB32PM<Format, true> }; } -#endif - -// To convert in place, let 'dest' and 'src' be the same. -static const uint *QT_FASTCALL convertIndexedToARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *clut, QDitherInfo *) +static void QT_FASTCALL convertIndexedToARGB32PM(uint *buffer, int count, const QVector<QRgb> *clut) { for (int i = 0; i < count; ++i) - buffer[i] = qPremultiply(clut->at(src[i])); + buffer[i] = qPremultiply(clut->at(buffer[i])); +} + +template<QPixelLayout::BPP BPP> +static const uint *QT_FASTCALL fetchIndexedToARGB32PM(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) { + const uint s = fetchPixel<BPP>(src, index + i); + buffer[i] = qPremultiply(clut->at(s)); + } + return buffer; +} + +template<QPixelLayout::BPP BPP> +static const QRgba64 *QT_FASTCALL fetchIndexedToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) { + const uint s = fetchPixel<BPP>(src, index + i); + buffer[i] = QRgba64::fromArgb32(clut->at(s)).premultiplied(); + } return buffer; } -static const QRgba64 *QT_FASTCALL convertIndexedToARGB64PM(QRgba64 *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL convertIndexedToRGBA64PM(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *clut, QDitherInfo *) { for (int i = 0; i < count; ++i) @@ -501,44 +731,77 @@ static const QRgba64 *QT_FASTCALL convertIndexedToARGB64PM(QRgba64 *buffer, cons return buffer; } -static const uint *QT_FASTCALL convertPassThrough(uint *, const uint *src, int, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL convertPassThrough(uint *, int, const QVector<QRgb> *) { - return src; } -static const uint *QT_FASTCALL convertARGB32ToARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static const uint *QT_FASTCALL fetchPassThrough(uint *, const uchar *src, int index, int, + const QVector<QRgb> *, QDitherInfo *) { - return qt_convertARGB32ToARGB32PM(buffer, src, count); + return reinterpret_cast<const uint *>(src) + index; } -static const uint *QT_FASTCALL convertRGBA8888PMToARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static const QRgba64 *QT_FASTCALL fetchPassThrough64(QRgba64 *, const uchar *src, int index, int, + const QVector<QRgb> *, QDitherInfo *) { - UNALIASED_CONVERSION_LOOP(buffer, src, count, RGBA2ARGB); - return buffer; + return reinterpret_cast<const QRgba64 *>(src) + index; } -static const uint *QT_FASTCALL convertRGBA8888ToARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL storePassThrough(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - return qt_convertRGBA8888ToARGB32PM(buffer, src, count); + uint *d = reinterpret_cast<uint *>(dest) + index; + if (d != src) + memcpy(d, src, count * sizeof(uint)); } -static const uint *QT_FASTCALL convertAlpha8ToRGB32(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL convertARGB32ToARGB32PM(uint *buffer, int count, const QVector<QRgb> *) +{ + qt_convertARGB32ToARGB32PM(buffer, buffer, count); +} + +static const uint *QT_FASTCALL fetchARGB32ToARGB32PM(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + return qt_convertARGB32ToARGB32PM(buffer, reinterpret_cast<const uint *>(src) + index, count); +} + +static void QT_FASTCALL convertRGBA8888PMToARGB32PM(uint *buffer, int count, const QVector<QRgb> *) { for (int i = 0; i < count; ++i) - buffer[i] = qRgba(0, 0, 0, src[i]); + buffer[i] = RGBA2ARGB(buffer[i]); +} + +static const uint *QT_FASTCALL fetchRGBA8888PMToARGB32PM(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + const uint *s = reinterpret_cast<const uint *>(src) + index; + UNALIASED_CONVERSION_LOOP(buffer, s, count, RGBA2ARGB); return buffer; } -static const uint *QT_FASTCALL convertGrayscale8ToRGB32(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL convertRGBA8888ToARGB32PM(uint *buffer, int count, const QVector<QRgb> *) +{ + qt_convertRGBA8888ToARGB32PM(buffer, buffer, count); +} + +static const uint *QT_FASTCALL fetchRGBA8888ToARGB32PM(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + return qt_convertRGBA8888ToARGB32PM(buffer, reinterpret_cast<const uint *>(src) + index, count); +} + +static void QT_FASTCALL convertAlpha8ToRGB32(uint *buffer, int count, const QVector<QRgb> *) { for (int i = 0; i < count; ++i) - buffer[i] = qRgb(src[i], src[i], src[i]); + buffer[i] = qRgba(0, 0, 0, buffer[i]); +} + +static const uint *QT_FASTCALL fetchAlpha8ToRGB32(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = qRgba(0, 0, 0, src[index + i]); return buffer; } @@ -549,6 +812,31 @@ static const QRgba64 *QT_FASTCALL convertAlpha8ToRGB64(QRgba64 *buffer, const ui buffer[i] = QRgba64::fromRgba(0, 0, 0, src[i]); return buffer; } +static const QRgba64 *QT_FASTCALL fetchAlpha8ToRGB64(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) + buffer[i] = QRgba64::fromRgba(0, 0, 0, src[index + i]); + return buffer; +} + +static void QT_FASTCALL convertGrayscale8ToRGB32(uint *buffer, int count, const QVector<QRgb> *) +{ + for (int i = 0; i < count; ++i) { + const uint s = buffer[i]; + buffer[i] = qRgb(s, s, s); + } +} + +static const uint *QT_FASTCALL fetchGrayscale8ToRGB32(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + for (int i = 0; i < count; ++i) { + const uint s = src[index + i]; + buffer[i] = qRgb(s, s, s); + } + return buffer; +} static const QRgba64 *QT_FASTCALL convertGrayscale8ToRGB64(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *) @@ -558,24 +846,33 @@ static const QRgba64 *QT_FASTCALL convertGrayscale8ToRGB64(QRgba64 *buffer, cons return buffer; } -static const uint *QT_FASTCALL convertARGB32FromARGB32PM(uint *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL fetchGrayscale8ToRGB64(QRgba64 *buffer, const uchar *src, int index, int count, const QVector<QRgb> *, QDitherInfo *) { - for (int i = 0; i < count; ++i) - buffer[i] = qUnpremultiply(src[i]); + for (int i = 0; i < count; ++i) { + const uint s = src[index + i]; + buffer[i] = QRgba64::fromRgba(s, s, s, 255); + } return buffer; } -static const uint *QT_FASTCALL convertRGBA8888PMFromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL storeARGB32FromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - UNALIASED_CONVERSION_LOOP(buffer, src, count, ARGB2RGBA); - return buffer; + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, [](uint c) { return qUnpremultiply(c); }); +} + +static void QT_FASTCALL storeRGBA8888PMFromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, ARGB2RGBA); } #ifdef __SSE2__ template<bool RGBA, bool maskAlpha> -static inline void qConvertARGB32PMToARGB64PM_sse2(QRgba64 *buffer, const uint *src, int count) +static inline void qConvertARGB32PMToRGBA64PM_sse2(QRgba64 *buffer, const uint *src, int count) { if (count <= 0) return; @@ -618,6 +915,66 @@ static inline void qConvertARGB32PMToARGB64PM_sse2(QRgba64 *buffer, const uint * *buffer++ = QRgba64::fromArgb32(s); } } + +template<QtPixelOrder PixelOrder> +static inline void qConvertRGBA64PMToA2RGB30PM_sse2(uint *dest, const QRgba64 *buffer, int count) +{ + const __m128i gmask = _mm_set1_epi32(0x000ffc00); + const __m128i cmask = _mm_set1_epi32(0x000003ff); + int i = 0; + __m128i vr, vg, vb, va; + for (; i < count && uintptr_t(buffer) & 0xF; ++i) { + *dest++ = qConvertRgb64ToRgb30<PixelOrder>(*buffer++); + } + + for (; i < count-15; i += 16) { + // Repremultiplying is really expensive and hard to do in SIMD without AVX2, + // so we try to avoid it by checking if it is needed 16 samples at a time. + __m128i vOr = _mm_set1_epi32(0); + __m128i vAnd = _mm_set1_epi32(0xffffffff); + for (int j = 0; j < 16; j += 2) { + __m128i vs = _mm_load_si128((const __m128i*)(buffer + j)); + vOr = _mm_or_si128(vOr, vs); + vAnd = _mm_and_si128(vAnd, vs); + } + const quint16 orAlpha = ((uint)_mm_extract_epi16(vOr, 3)) | ((uint)_mm_extract_epi16(vOr, 7)); + const quint16 andAlpha = ((uint)_mm_extract_epi16(vAnd, 3)) & ((uint)_mm_extract_epi16(vAnd, 7)); + + if (andAlpha == 0xffff) { + for (int j = 0; j < 16; j += 2) { + __m128i vs = _mm_load_si128((const __m128i*)buffer); + buffer += 2; + vr = _mm_srli_epi64(vs, 6); + vg = _mm_srli_epi64(vs, 16 + 6 - 10); + vb = _mm_srli_epi64(vs, 32 + 6); + vr = _mm_and_si128(vr, cmask); + vg = _mm_and_si128(vg, gmask); + vb = _mm_and_si128(vb, cmask); + va = _mm_srli_epi64(vs, 48 + 14); + if (PixelOrder == PixelOrderRGB) + vr = _mm_slli_epi32(vr, 20); + else + vb = _mm_slli_epi32(vb, 20); + va = _mm_slli_epi32(va, 30); + __m128i vd = _mm_or_si128(_mm_or_si128(vr, vg), _mm_or_si128(vb, va)); + vd = _mm_shuffle_epi32(vd, _MM_SHUFFLE(3, 1, 2, 0)); + _mm_storel_epi64((__m128i*)dest, vd); + dest += 2; + } + } else if (orAlpha == 0) { + for (int j = 0; j < 16; ++j) { + *dest++ = 0; + buffer++; + } + } else { + for (int j = 0; j < 16; ++j) + *dest++ = qConvertRgb64ToRgb30<PixelOrder>(*buffer++); + } + } + + SIMD_EPILOGUE(i, count, 15) + *dest++ = qConvertRgb64ToRgb30<PixelOrder>(*buffer++); +} #elif defined(__ARM_NEON__) template<bool RGBA, bool maskAlpha> static inline void qConvertARGB32PMToRGBA64PM_neon(QRgba64 *buffer, const uint *src, int count) @@ -671,7 +1028,7 @@ static const QRgba64 *QT_FASTCALL convertRGB32ToRGB64(QRgba64 *buffer, const uin const QVector<QRgb> *, QDitherInfo *) { #ifdef __SSE2__ - qConvertARGB32PMToARGB64PM_sse2<false, true>(buffer, src, count); + qConvertARGB32PMToRGBA64PM_sse2<false, true>(buffer, src, count); #elif defined(__ARM_NEON__) qConvertARGB32PMToRGBA64PM_neon<false, true>(buffer, src, count); #else @@ -681,11 +1038,17 @@ static const QRgba64 *QT_FASTCALL convertRGB32ToRGB64(QRgba64 *buffer, const uin return buffer; } -static const QRgba64 *QT_FASTCALL convertARGB32ToARGB64PM(QRgba64 *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL fetchRGB32ToRGB64(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + return convertRGB32ToRGB64(buffer, reinterpret_cast<const uint *>(src) + index, count, nullptr, nullptr); +} + +static const QRgba64 *QT_FASTCALL convertARGB32ToRGBA64PM(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *) { #ifdef __SSE2__ - qConvertARGB32PMToARGB64PM_sse2<false, false>(buffer, src, count); + qConvertARGB32PMToRGBA64PM_sse2<false, false>(buffer, src, count); for (int i = 0; i < count; ++i) buffer[i] = buffer[i].premultiplied(); #elif defined(__ARM_NEON__) @@ -699,11 +1062,17 @@ static const QRgba64 *QT_FASTCALL convertARGB32ToARGB64PM(QRgba64 *buffer, const return buffer; } -static const QRgba64 *QT_FASTCALL convertARGB32PMToARGB64PM(QRgba64 *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL fetchARGB32ToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + return convertARGB32ToRGBA64PM(buffer, reinterpret_cast<const uint *>(src) + index, count, nullptr, nullptr); +} + +static const QRgba64 *QT_FASTCALL convertARGB32PMToRGBA64PM(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *) { #ifdef __SSE2__ - qConvertARGB32PMToARGB64PM_sse2<false, false>(buffer, src, count); + qConvertARGB32PMToRGBA64PM_sse2<false, false>(buffer, src, count); #elif defined(__ARM_NEON__) qConvertARGB32PMToRGBA64PM_neon<false, false>(buffer, src, count); #else @@ -713,11 +1082,36 @@ static const QRgba64 *QT_FASTCALL convertARGB32PMToARGB64PM(QRgba64 *buffer, con return buffer; } -static const QRgba64 *QT_FASTCALL convertRGBA8888ToARGB64PM(QRgba64 *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL fetchARGB32PMToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + return convertARGB32PMToRGBA64PM(buffer, reinterpret_cast<const uint *>(src) + index, count, nullptr, nullptr); +} + +static void convertRGBA64ToRGBA64PM(QRgba64 *buffer, int count) +{ + for (int i = 0; i < count; ++i) + buffer[i] = buffer[i].premultiplied(); +} + +static void convertRGBA64PMToRGBA64PM(QRgba64 *, int) +{ +} + +static const QRgba64 *QT_FASTCALL fetchRGBA64ToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + const QRgba64 *s = reinterpret_cast<const QRgba64 *>(src) + index; + for (int i = 0; i < count; ++i) + buffer[i] = QRgba64::fromRgba64(s[i]).premultiplied(); + return buffer; +} + +static const QRgba64 *QT_FASTCALL convertRGBA8888ToRGBA64PM(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *) { #ifdef __SSE2__ - qConvertARGB32PMToARGB64PM_sse2<true, false>(buffer, src, count); + qConvertARGB32PMToRGBA64PM_sse2<true, false>(buffer, src, count); for (int i = 0; i < count; ++i) buffer[i] = buffer[i].premultiplied(); #elif defined(__ARM_NEON__) @@ -731,11 +1125,17 @@ static const QRgba64 *QT_FASTCALL convertRGBA8888ToARGB64PM(QRgba64 *buffer, con return buffer; } -static const QRgba64 *QT_FASTCALL convertRGBA8888PMToARGB64PM(QRgba64 *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL fetchRGBA8888ToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + return convertRGBA8888ToRGBA64PM(buffer, reinterpret_cast<const uint *>(src) + index, count, nullptr, nullptr); +} + +static const QRgba64 *QT_FASTCALL convertRGBA8888PMToRGBA64PM(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *) { #ifdef __SSE2__ - qConvertARGB32PMToARGB64PM_sse2<true, false>(buffer, src, count); + qConvertARGB32PMToRGBA64PM_sse2<true, false>(buffer, src, count); #elif defined(__ARM_NEON__) qConvertARGB32PMToRGBA64PM_neon<true, false>(buffer, src, count); #else @@ -745,33 +1145,45 @@ static const QRgba64 *QT_FASTCALL convertRGBA8888PMToARGB64PM(QRgba64 *buffer, c return buffer; } -static const uint *QT_FASTCALL convertRGBA8888FromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static const QRgba64 *QT_FASTCALL fetchRGBA8888PMToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - for (int i = 0; i < count; ++i) - buffer[i] = ARGB2RGBA(qUnpremultiply(src[i])); - return buffer; + return convertRGBA8888PMToRGBA64PM(buffer, reinterpret_cast<const uint *>(src) + index, count, nullptr, nullptr); } -static const uint *QT_FASTCALL convertRGBXFromRGB32(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL storeRGBA8888FromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - UNALIASED_CONVERSION_LOOP(buffer, src, count, [](uint c) { return ARGB2RGBA(0xff000000 | c); }); - return buffer; + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, [](uint c) { return ARGB2RGBA(qUnpremultiply(c)); }); } -static const uint *QT_FASTCALL convertRGBXFromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL storeRGBXFromRGB32(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, [](uint c) { return ARGB2RGBA(0xff000000 | c); }); +} + +static void QT_FASTCALL storeRGBXFromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, [](uint c) { return ARGB2RGBA(0xff000000 | qUnpremultiply(c)); }); +} + +template<QtPixelOrder PixelOrder> +static void QT_FASTCALL convertA2RGB30PMToARGB32PM(uint *buffer, int count, const QVector<QRgb> *) { for (int i = 0; i < count; ++i) - buffer[i] = ARGB2RGBA(0xff000000 | qUnpremultiply(src[i])); - return buffer; + buffer[i] = qConvertA2rgb30ToArgb32<PixelOrder>(buffer[i]); } template<QtPixelOrder PixelOrder> -static const uint *QT_FASTCALL convertA2RGB30PMToARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *dither) +static const uint *QT_FASTCALL fetchA2RGB30PMToARGB32PM(uint *buffer, const uchar *s, int index, int count, + const QVector<QRgb> *, QDitherInfo *dither) { + const uint *src = reinterpret_cast<const uint *>(s) + index; if (!dither) { UNALIASED_CONVERSION_LOOP(buffer, src, count, qConvertA2rgb30ToArgb32<PixelOrder>); } else { @@ -796,7 +1208,7 @@ static const uint *QT_FASTCALL convertA2RGB30PMToARGB32PM(uint *buffer, const ui #ifdef __SSE2__ template<QtPixelOrder PixelOrder> -static inline void qConvertA2RGB30PMToARGB64PM_sse2(QRgba64 *buffer, const uint *src, int count) +static inline void qConvertA2RGB30PMToRGBA64PM_sse2(QRgba64 *buffer, const uint *src, int count) { if (count <= 0) return; @@ -839,11 +1251,11 @@ static inline void qConvertA2RGB30PMToARGB64PM_sse2(QRgba64 *buffer, const uint #endif template<QtPixelOrder PixelOrder> -static const QRgba64 *QT_FASTCALL convertA2RGB30PMToARGB64PM(QRgba64 *buffer, const uint *src, int count, +static const QRgba64 *QT_FASTCALL convertA2RGB30PMToRGBA64PM(QRgba64 *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *) { #ifdef __SSE2__ - qConvertA2RGB30PMToARGB64PM_sse2<PixelOrder>(buffer, src, count); + qConvertA2RGB30PMToRGBA64PM_sse2<PixelOrder>(buffer, src, count); #else for (int i = 0; i < count; ++i) buffer[i] = qConvertA2rgb30ToRgb64<PixelOrder>(src[i]); @@ -852,290 +1264,334 @@ static const QRgba64 *QT_FASTCALL convertA2RGB30PMToARGB64PM(QRgba64 *buffer, co } template<QtPixelOrder PixelOrder> -static const uint *QT_FASTCALL convertA2RGB30PMFromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static const QRgba64 *QT_FASTCALL fetchA2RGB30PMToRGBA64PM(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - UNALIASED_CONVERSION_LOOP(buffer, src, count, qConvertArgb32ToA2rgb30<PixelOrder>); - return buffer; + return convertA2RGB30PMToRGBA64PM<PixelOrder>(buffer, reinterpret_cast<const uint *>(src) + index, count, nullptr, nullptr); } template<QtPixelOrder PixelOrder> -static const uint *QT_FASTCALL convertRGB30FromRGB32(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL storeA2RGB30PMFromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - for (int i = 0; i < count; ++i) - buffer[i] = qConvertRgb32ToRgb30<PixelOrder>(src[i]); - return buffer; + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, qConvertArgb32ToA2rgb30<PixelOrder>); } template<QtPixelOrder PixelOrder> -static const uint *QT_FASTCALL convertRGB30FromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static void QT_FASTCALL storeRGB30FromRGB32(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - UNALIASED_CONVERSION_LOOP(buffer, src, count, qConvertRgb32ToRgb30<PixelOrder>); - return buffer; + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, qConvertRgb32ToRgb30<PixelOrder>); } -static const uint *QT_FASTCALL convertAlpha8FromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +template<QtPixelOrder PixelOrder> +static void QT_FASTCALL storeRGB30FromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - for (int i = 0; i < count; ++i) - buffer[i] = qAlpha(src[i]); - return buffer; + uint *d = reinterpret_cast<uint *>(dest) + index; + UNALIASED_CONVERSION_LOOP(d, src, count, qConvertRgb32ToRgb30<PixelOrder>); } -static const uint *QT_FASTCALL convertGrayscale8FromRGB32(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +template<bool RGBA> +void qt_convertRGBA64ToARGB32(uint *dst, const QRgba64 *src, int count) { - for (int i = 0; i < count; ++i) - buffer[i] = qGray(src[i]); - return buffer; + int i = 0; +#ifdef __SSE2__ + if (((uintptr_t)dst & 0x7) && count > 0) { + uint s = (*src++).toArgb32(); + if (RGBA) + s = ARGB2RGBA(s); + *dst++ = s; + i++; + } + const __m128i vhalf = _mm_set1_epi32(0x80); + const __m128i vzero = _mm_setzero_si128(); + for (; i < count-1; i += 2) { + __m128i vs = _mm_loadu_si128((const __m128i*)src); + src += 2; + if (!RGBA) { + vs = _mm_shufflelo_epi16(vs, _MM_SHUFFLE(3, 0, 1, 2)); + vs = _mm_shufflehi_epi16(vs, _MM_SHUFFLE(3, 0, 1, 2)); + } + __m128i v1 = _mm_unpacklo_epi16(vs, vzero); + __m128i v2 = _mm_unpackhi_epi16(vs, vzero); + v1 = _mm_add_epi32(v1, vhalf); + v2 = _mm_add_epi32(v2, vhalf); + v1 = _mm_sub_epi32(v1, _mm_srli_epi32(v1, 8)); + v2 = _mm_sub_epi32(v2, _mm_srli_epi32(v2, 8)); + v1 = _mm_srli_epi32(v1, 8); + v2 = _mm_srli_epi32(v2, 8); + v1 = _mm_packs_epi32(v1, v2); + v1 = _mm_packus_epi16(v1, vzero); + _mm_storel_epi64((__m128i*)(dst), v1); + dst += 2; + } +#endif + for (; i < count; i++) { + uint s = (*src++).toArgb32(); + if (RGBA) + s = ARGB2RGBA(s); + *dst++ = s; + } } +template void qt_convertRGBA64ToARGB32<false>(uint *dst, const QRgba64 *src, int count); +template void qt_convertRGBA64ToARGB32<true>(uint *dst, const QRgba64 *src, int count); -static const uint *QT_FASTCALL convertGrayscale8FromARGB32PM(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) + +static void QT_FASTCALL storeAlpha8FromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { for (int i = 0; i < count; ++i) - buffer[i] = qGray(qUnpremultiply(src[i])); - return buffer; + dest[index + i] = qAlpha(src[i]); } -template <QPixelLayout::BPP bpp> static -uint QT_FASTCALL fetchPixel(const uchar *, int) +static void QT_FASTCALL storeGrayscale8FromRGB32(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - Q_UNREACHABLE(); - return 0; + for (int i = 0; i < count; ++i) + dest[index + i] = qGray(src[i]); } -template <> -inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP1LSB>(const uchar *src, int index) +static void QT_FASTCALL storeGrayscale8FromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - return (src[index >> 3] >> (index & 7)) & 1; + for (int i = 0; i < count; ++i) + dest[index + i] = qGray(qUnpremultiply(src[i])); } -template <> -inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP1MSB>(const uchar *src, int index) +static const uint *QT_FASTCALL fetchRGB64ToRGB32(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - return (src[index >> 3] >> (~index & 7)) & 1; + const QRgba64 *s = reinterpret_cast<const QRgba64 *>(src) + index; + for (int i = 0; i < count; ++i) + buffer[i] = toArgb32(s[i]); + return buffer; } -template <> -inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP8>(const uchar *src, int index) +static void QT_FASTCALL storeRGB64FromRGB32(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - return src[index]; + QRgba64 *d = reinterpret_cast<QRgba64 *>(dest) + index; + for (int i = 0; i < count; ++i) + d[i] = QRgba64::fromArgb32(src[i]); } -template <> -inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP16>(const uchar *src, int index) +static const uint *QT_FASTCALL fetchRGBA64ToARGB32PM(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - return reinterpret_cast<const quint16 *>(src)[index]; + const QRgba64 *s = reinterpret_cast<const QRgba64 *>(src) + index; + for (int i = 0; i < count; ++i) + buffer[i] = toArgb32(s[i].premultiplied()); + return buffer; } -template <> -inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP24>(const uchar *src, int index) +static void QT_FASTCALL storeRGBA64FromARGB32PM(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - return reinterpret_cast<const quint24 *>(src)[index]; + QRgba64 *d = reinterpret_cast<QRgba64 *>(dest) + index; + for (int i = 0; i < count; ++i) + d[i] = QRgba64::fromArgb32(src[i]).unpremultiplied(); } -template <> -inline uint QT_FASTCALL fetchPixel<QPixelLayout::BPP32>(const uchar *src, int index) -{ - return reinterpret_cast<const uint *>(src)[index]; -} +// Note: +// convertToArgb32() assumes that no color channel is less than 4 bits. +// storeRGBFromARGB32PM() assumes that no color channel is more than 8 bits. +// QImage::rgbSwapped() assumes that the red and blue color channels have the same number of bits. +QPixelLayout qPixelLayouts[QImage::NImageFormats] = { + { false, false, QPixelLayout::BPPNone, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr }, // Format_Invalid + { false, false, QPixelLayout::BPP1MSB, nullptr, + convertIndexedToARGB32PM, convertIndexedToRGBA64PM, + fetchIndexedToARGB32PM<QPixelLayout::BPP1MSB>, fetchIndexedToRGBA64PM<QPixelLayout::BPP1MSB>, + nullptr, nullptr }, // Format_Mono + { false, false, QPixelLayout::BPP1LSB, nullptr, + convertIndexedToARGB32PM, convertIndexedToRGBA64PM, + fetchIndexedToARGB32PM<QPixelLayout::BPP1LSB>, fetchIndexedToRGBA64PM<QPixelLayout::BPP1LSB>, + nullptr, nullptr }, // Format_MonoLSB + { false, false, QPixelLayout::BPP8, nullptr, + convertIndexedToARGB32PM, convertIndexedToRGBA64PM, + fetchIndexedToARGB32PM<QPixelLayout::BPP8>, fetchIndexedToRGBA64PM<QPixelLayout::BPP8>, + nullptr, nullptr }, // Format_Indexed8 + // Technically using convertPassThrough to convert from ARGB32PM to RGB32 is wrong, + // but everywhere this generic conversion would be wrong is currently overloaded. + { false, false, QPixelLayout::BPP32, rbSwap_rgb32, convertPassThrough, + convertRGB32ToRGB64, fetchPassThrough, fetchRGB32ToRGB64, storePassThrough, storePassThrough }, // Format_RGB32 + { true, false, QPixelLayout::BPP32, rbSwap_rgb32, convertARGB32ToARGB32PM, + convertARGB32ToRGBA64PM, fetchARGB32ToARGB32PM, fetchARGB32ToRGBA64PM, storeARGB32FromARGB32PM, storePassThrough }, // Format_ARGB32 + { true, true, QPixelLayout::BPP32, rbSwap_rgb32, convertPassThrough, + convertARGB32PMToRGBA64PM, fetchPassThrough, fetchARGB32PMToRGBA64PM, storePassThrough, storePassThrough }, // Format_ARGB32_Premultiplied + pixelLayoutRGB<QImage::Format_RGB16>(), + pixelLayoutARGBPM<QImage::Format_ARGB8565_Premultiplied>(), + pixelLayoutRGB<QImage::Format_RGB666>(), + pixelLayoutARGBPM<QImage::Format_ARGB6666_Premultiplied>(), + pixelLayoutRGB<QImage::Format_RGB555>(), + pixelLayoutARGBPM<QImage::Format_ARGB8555_Premultiplied>(), + pixelLayoutRGB<QImage::Format_RGB888>(), + pixelLayoutRGB<QImage::Format_RGB444>(), + pixelLayoutARGBPM<QImage::Format_ARGB4444_Premultiplied>(), + { false, false, QPixelLayout::BPP32, rbSwap<QImage::Format_RGBA8888>, convertRGBA8888PMToARGB32PM, + convertRGBA8888PMToRGBA64PM, fetchRGBA8888PMToARGB32PM, fetchRGBA8888PMToRGBA64PM, storeRGBXFromARGB32PM, storeRGBXFromRGB32 }, // Format_RGBX8888 + { true, false, QPixelLayout::BPP32, rbSwap<QImage::Format_RGBA8888>, convertRGBA8888ToARGB32PM, + convertRGBA8888ToRGBA64PM, fetchRGBA8888ToARGB32PM, fetchRGBA8888ToRGBA64PM, storeRGBA8888FromARGB32PM, storeRGBXFromRGB32 }, // Format_RGBA8888 + { true, true, QPixelLayout::BPP32, rbSwap<QImage::Format_RGBA8888>, convertRGBA8888PMToARGB32PM, + convertRGBA8888PMToRGBA64PM, fetchRGBA8888PMToARGB32PM, fetchRGBA8888PMToRGBA64PM, storeRGBA8888PMFromARGB32PM, storeRGBXFromRGB32 }, // Format_RGBA8888_Premultiplied + { false, false, QPixelLayout::BPP32, rbSwap_rgb30, + convertA2RGB30PMToARGB32PM<PixelOrderBGR>, + convertA2RGB30PMToRGBA64PM<PixelOrderBGR>, + fetchA2RGB30PMToARGB32PM<PixelOrderBGR>, + fetchA2RGB30PMToRGBA64PM<PixelOrderBGR>, + storeRGB30FromARGB32PM<PixelOrderBGR>, + storeRGB30FromRGB32<PixelOrderBGR> + }, // Format_BGR30 + { true, true, QPixelLayout::BPP32, rbSwap_rgb30, + convertA2RGB30PMToARGB32PM<PixelOrderBGR>, + convertA2RGB30PMToRGBA64PM<PixelOrderBGR>, + fetchA2RGB30PMToARGB32PM<PixelOrderBGR>, + fetchA2RGB30PMToRGBA64PM<PixelOrderBGR>, + storeA2RGB30PMFromARGB32PM<PixelOrderBGR>, + storeRGB30FromRGB32<PixelOrderBGR> + }, // Format_A2BGR30_Premultiplied + { false, false, QPixelLayout::BPP32, rbSwap_rgb30, + convertA2RGB30PMToARGB32PM<PixelOrderRGB>, + convertA2RGB30PMToRGBA64PM<PixelOrderRGB>, + fetchA2RGB30PMToARGB32PM<PixelOrderRGB>, + fetchA2RGB30PMToRGBA64PM<PixelOrderRGB>, + storeRGB30FromARGB32PM<PixelOrderRGB>, + storeRGB30FromRGB32<PixelOrderRGB> + }, // Format_RGB30 + { true, true, QPixelLayout::BPP32, rbSwap_rgb30, + convertA2RGB30PMToARGB32PM<PixelOrderRGB>, + convertA2RGB30PMToRGBA64PM<PixelOrderRGB>, + fetchA2RGB30PMToARGB32PM<PixelOrderRGB>, + fetchA2RGB30PMToRGBA64PM<PixelOrderRGB>, + storeA2RGB30PMFromARGB32PM<PixelOrderRGB>, + storeRGB30FromRGB32<PixelOrderRGB> + }, // Format_A2RGB30_Premultiplied + { true, true, QPixelLayout::BPP8, nullptr, + convertAlpha8ToRGB32, convertAlpha8ToRGB64, + fetchAlpha8ToRGB32, fetchAlpha8ToRGB64, + storeAlpha8FromARGB32PM, nullptr }, // Format_Alpha8 + { false, false, QPixelLayout::BPP8, nullptr, + convertGrayscale8ToRGB32, convertGrayscale8ToRGB64, + fetchGrayscale8ToRGB32, fetchGrayscale8ToRGB64, + storeGrayscale8FromARGB32PM, storeGrayscale8FromRGB32 }, // Format_Grayscale8 + { false, false, QPixelLayout::BPP64, nullptr, + convertPassThrough, nullptr, + fetchRGB64ToRGB32, fetchPassThrough64, + storeRGB64FromRGB32, storeRGB64FromRGB32 }, // Format_RGBX64 + { true, false, QPixelLayout::BPP64, nullptr, + convertARGB32ToARGB32PM, nullptr, + fetchRGBA64ToARGB32PM, fetchRGBA64ToRGBA64PM, + storeRGBA64FromARGB32PM, storeRGB64FromRGB32 }, // Format_RGBA64 + { true, true, QPixelLayout::BPP64, nullptr, + convertPassThrough, nullptr, + fetchRGB64ToRGB32, fetchPassThrough64, + storeRGB64FromRGB32, storeRGB64FromRGB32 } // Format_RGBA64_Premultiplied +}; -template <QPixelLayout::BPP bpp> -inline const uint *QT_FASTCALL fetchPixels(uint *buffer, const uchar *src, int index, int count) -{ - for (int i = 0; i < count; ++i) - buffer[i] = fetchPixel<bpp>(src, index + i); - return buffer; -} +Q_STATIC_ASSERT(sizeof(qPixelLayouts) / sizeof(*qPixelLayouts) == QImage::NImageFormats); -template <> -inline const uint *QT_FASTCALL fetchPixels<QPixelLayout::BPP32>(uint *, const uchar *src, int index, int) +static void QT_FASTCALL convertFromRgb64(uint *dest, const QRgba64 *src, int length) { - return reinterpret_cast<const uint *>(src) + index; + for (int i = 0; i < length; ++i) { + dest[i] = toArgb32(src[i]); + } } -template <QPixelLayout::BPP width> static -void QT_FASTCALL storePixel(uchar *dest, int index, uint pixel); - -template <> -inline void QT_FASTCALL storePixel<QPixelLayout::BPP1LSB>(uchar *dest, int index, uint pixel) +template<QImage::Format format> +static void QT_FASTCALL storeGenericFromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *dither) { - if (pixel) - dest[index >> 3] |= 1 << (index & 7); - else - dest[index >> 3] &= ~(1 << (index & 7)); + uint buffer[BufferSize]; + convertFromRgb64(buffer, src, count); + qPixelLayouts[format].storeFromARGB32PM(dest, buffer, index, count, clut, dither); } -template <> -inline void QT_FASTCALL storePixel<QPixelLayout::BPP1MSB>(uchar *dest, int index, uint pixel) +static void QT_FASTCALL storeARGB32FromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - if (pixel) - dest[index >> 3] |= 1 << (~index & 7); - else - dest[index >> 3] &= ~(1 << (~index & 7)); + uint *d = (uint*)dest + index; + for (int i = 0; i < count; ++i) + d[i] = toArgb32(src[i].unpremultiplied()); } -template <> -inline void QT_FASTCALL storePixel<QPixelLayout::BPP8>(uchar *dest, int index, uint pixel) +static void QT_FASTCALL storeRGBA8888FromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - dest[index] = uchar(pixel); + uint *d = (uint*)dest + index; + for (int i = 0; i < count; ++i) + d[i] = toRgba8888(src[i].unpremultiplied()); } -template <> -inline void QT_FASTCALL storePixel<QPixelLayout::BPP16>(uchar *dest, int index, uint pixel) +template<QtPixelOrder PixelOrder> +static void QT_FASTCALL storeRGB30FromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - reinterpret_cast<quint16 *>(dest)[index] = quint16(pixel); + uint *d = (uint*)dest + index; +#ifdef __SSE2__ + qConvertRGBA64PMToA2RGB30PM_sse2<PixelOrder>(d, src, count); +#else + for (int i = 0; i < count; ++i) + d[i] = qConvertRgb64ToRgb30<PixelOrder>(src[i]); +#endif } -template <> -inline void QT_FASTCALL storePixel<QPixelLayout::BPP24>(uchar *dest, int index, uint pixel) +static void QT_FASTCALL storeRGBX64FromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - reinterpret_cast<quint24 *>(dest)[index] = quint24(pixel); + QRgba64 *d = reinterpret_cast<QRgba64*>(dest) + index; + for (int i = 0; i < count; ++i) { + d[i] = src[i].unpremultiplied(); + d[i].setAlpha(65535); + } } -template <QPixelLayout::BPP width> -inline void QT_FASTCALL storePixels(uchar *dest, const uint *src, int index, int count) +static void QT_FASTCALL storeRGBA64FromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { + QRgba64 *d = reinterpret_cast<QRgba64*>(dest) + index; for (int i = 0; i < count; ++i) - storePixel<width>(dest, index + i, src[i]); + d[i] = src[i].unpremultiplied(); } -template <> -inline void QT_FASTCALL storePixels<QPixelLayout::BPP32>(uchar *dest, const uint *src, int index, int count) +static void QT_FASTCALL storeRGBA64PMFromRGBA64PM(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - memcpy(reinterpret_cast<uint *>(dest) + index, src, count * sizeof(uint)); -} - -// Note: -// convertToArgb32() assumes that no color channel is less than 4 bits. -// convertFromArgb32() assumes that no color channel is more than 8 bits. -// QImage::rgbSwapped() assumes that the red and blue color channels have the same number of bits. -QPixelLayout qPixelLayouts[QImage::NImageFormats] = { - { 0, 0, 0, 0, 0, 0, 0, 0, false, QPixelLayout::BPPNone, 0, 0, 0, 0 }, // Format_Invalid - { 0, 0, 0, 0, 0, 0, 0, 0, false, QPixelLayout::BPP1MSB, convertIndexedToARGB32PM, 0, 0, convertIndexedToARGB64PM }, // Format_Mono - { 0, 0, 0, 0, 0, 0, 0, 0, false, QPixelLayout::BPP1LSB, convertIndexedToARGB32PM, 0, 0, convertIndexedToARGB64PM }, // Format_MonoLSB - { 0, 0, 0, 0, 0, 0, 0, 0, false, QPixelLayout::BPP8, convertIndexedToARGB32PM, 0, 0, convertIndexedToARGB64PM }, // Format_Indexed8 - // Technically using convertPassThrough to convert from ARGB32PM to RGB32 is wrong, - // but everywhere this generic conversion would be wrong is currently overloaded. - { 8, 16, 8, 8, 8, 0, 0, 0, false, QPixelLayout::BPP32, convertPassThrough, convertPassThrough, convertPassThrough, convertRGB32ToRGB64 }, // Format_RGB32 - { 8, 16, 8, 8, 8, 0, 8, 24, false, QPixelLayout::BPP32, convertARGB32ToARGB32PM, convertARGB32FromARGB32PM, convertPassThrough, convertARGB32ToARGB64PM }, // Format_ARGB32 - { 8, 16, 8, 8, 8, 0, 8, 24, true, QPixelLayout::BPP32, convertPassThrough, convertPassThrough, convertPassThrough, convertARGB32PMToARGB64PM }, // Format_ARGB32_Premultiplied -#ifdef Q_COMPILER_CONSTEXPR - pixelLayoutRGB<QImage::Format_RGB16>(), - pixelLayoutARGBPM<QImage::Format_ARGB8565_Premultiplied>(), - pixelLayoutRGB<QImage::Format_RGB666>(), - pixelLayoutARGBPM<QImage::Format_ARGB6666_Premultiplied>(), - pixelLayoutRGB<QImage::Format_RGB555>(), - pixelLayoutARGBPM<QImage::Format_ARGB8555_Premultiplied>(), - pixelLayoutRGB<QImage::Format_RGB888>(), - pixelLayoutRGB<QImage::Format_RGB444>(), - pixelLayoutARGBPM<QImage::Format_ARGB4444_Premultiplied>(), -#else - { 5, 11, 6, 5, 5, 0, 0, 0, false, QPixelLayout::BPP16, - convertToRGB32<QImage::Format_RGB16>, - convertRGBFromARGB32PM<QImage::Format_RGB16, false>, - convertRGBFromARGB32PM<QImage::Format_RGB16, true>, - convertToRGB64<QImage::Format_RGB16>, - }, - { 5, 19, 6, 13, 5, 8, 8, 0, true, QPixelLayout::BPP24, - convertARGBPMToARGB32PM<QImage::Format_ARGB8565_Premultiplied>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB8565_Premultiplied, false>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB8565_Premultiplied, true>, - convertARGBPMToARGB64PM<QImage::Format_ARGB8565_Premultiplied>, - }, - { 6, 12, 6, 6, 6, 0, 0, 0, false, QPixelLayout::BPP24, - convertToRGB32<QImage::Format_RGB666>, - convertRGBFromARGB32PM<QImage::Format_RGB666, false>, - convertRGBFromARGB32PM<QImage::Format_RGB666, true>, - convertToRGB64<QImage::Format_RGB666>, - }, - { 6, 12, 6, 6, 6, 0, 6, 18, true, QPixelLayout::BPP24, - convertARGBPMToARGB32PM<QImage::Format_ARGB6666_Premultiplied>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB6666_Premultiplied, false>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB6666_Premultiplied, true>, - convertARGBPMToARGB64PM<QImage::Format_ARGB6666_Premultiplied>, - }, - { 5, 10, 5, 5, 5, 0, 0, 0, false, QPixelLayout::BPP16, - convertToRGB32<QImage::Format_RGB555>, - convertRGBFromARGB32PM<QImage::Format_RGB555, false>, - convertRGBFromARGB32PM<QImage::Format_RGB555, true>, - convertToRGB64<QImage::Format_RGB555>, - }, - { 5, 18, 5, 13, 5, 8, 8, 0, true, QPixelLayout::BPP24, - convertARGBPMToARGB32PM<QImage::Format_ARGB8555_Premultiplied>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB8555_Premultiplied, false>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB8555_Premultiplied, true>, - convertARGBPMToARGB64PM<QImage::Format_ARGB8555_Premultiplied>, - }, - { 8, 16, 8, 8, 8, 0, 0, 0, false, QPixelLayout::BPP24, - convertToRGB32<QImage::Format_RGB888>, - convertRGBFromARGB32PM<QImage::Format_RGB888, false>, - convertRGBFromARGB32PM<QImage::Format_RGB888, true>, - convertToRGB64<QImage::Format_RGB888>, - }, - { 4, 8, 4, 4, 4, 0, 0, 0, false, QPixelLayout::BPP16, - convertToRGB32<QImage::Format_RGB444>, - convertRGBFromARGB32PM<QImage::Format_RGB444, false>, - convertRGBFromARGB32PM<QImage::Format_RGB444, true>, - convertToRGB64<QImage::Format_RGB444>, - }, - { 4, 8, 4, 4, 4, 0, 4, 12, true, QPixelLayout::BPP16, - convertARGBPMToARGB32PM<QImage::Format_ARGB4444_Premultiplied>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB4444_Premultiplied, false>, - convertARGBPMFromARGB32PM<QImage::Format_ARGB4444_Premultiplied, true>, - convertARGBPMToARGB64PM<QImage::Format_ARGB4444_Premultiplied>, - }, -#endif -#if Q_BYTE_ORDER == Q_BIG_ENDIAN - { 8, 24, 8, 16, 8, 8, 0, 0, false, QPixelLayout::BPP32, convertRGBA8888PMToARGB32PM, convertRGBXFromARGB32PM, convertRGBXFromRGB32, convertRGBA8888PMToARGB64PM }, // Format_RGBX8888 - { 8, 24, 8, 16, 8, 8, 8, 0, false, QPixelLayout::BPP32, convertRGBA8888ToARGB32PM, convertRGBA8888FromARGB32PM, convertRGBXFromRGB32, convertRGBA8888ToARGB64PM }, // Format_RGBA8888 - { 8, 24, 8, 16, 8, 8, 8, 0, true, QPixelLayout::BPP32, convertRGBA8888PMToARGB32PM, convertRGBA8888PMFromARGB32PM, convertRGBXFromRGB32, convertRGBA8888PMToARGB64PM}, // Format_RGBA8888_Premultiplied -#else - { 8, 0, 8, 8, 8, 16, 0, 24, false, QPixelLayout::BPP32, convertRGBA8888PMToARGB32PM, convertRGBXFromARGB32PM, convertRGBXFromRGB32, convertRGBA8888PMToARGB64PM }, // Format_RGBX8888 - { 8, 0, 8, 8, 8, 16, 8, 24, false, QPixelLayout::BPP32, convertRGBA8888ToARGB32PM, convertRGBA8888FromARGB32PM, convertRGBXFromRGB32, convertRGBA8888ToARGB64PM }, // Format_RGBA8888 (ABGR32) - { 8, 0, 8, 8, 8, 16, 8, 24, true, QPixelLayout::BPP32, convertRGBA8888PMToARGB32PM, convertRGBA8888PMFromARGB32PM, convertRGBXFromRGB32, convertRGBA8888PMToARGB64PM }, // Format_RGBA8888_Premultiplied -#endif - { 10, 20, 10, 10, 10, 0, 0, 30, false, QPixelLayout::BPP32, convertA2RGB30PMToARGB32PM<PixelOrderBGR>, convertRGB30FromARGB32PM<PixelOrderBGR>, convertRGB30FromRGB32<PixelOrderBGR>, convertA2RGB30PMToARGB64PM<PixelOrderBGR> }, // Format_BGR30 - { 10, 20, 10, 10, 10, 0, 2, 30, true, QPixelLayout::BPP32, convertA2RGB30PMToARGB32PM<PixelOrderBGR>, convertA2RGB30PMFromARGB32PM<PixelOrderBGR>, convertRGB30FromRGB32<PixelOrderBGR>, convertA2RGB30PMToARGB64PM<PixelOrderBGR> }, // Format_A2BGR30_Premultiplied - { 10, 0, 10, 10, 10, 20, 0, 30, false, QPixelLayout::BPP32, convertA2RGB30PMToARGB32PM<PixelOrderRGB>, convertRGB30FromARGB32PM<PixelOrderRGB>, convertRGB30FromRGB32<PixelOrderRGB>, convertA2RGB30PMToARGB64PM<PixelOrderRGB> }, // Format_RGB30 - { 10, 0, 10, 10, 10, 20, 2, 30, true, QPixelLayout::BPP32, convertA2RGB30PMToARGB32PM<PixelOrderRGB>, convertA2RGB30PMFromARGB32PM<PixelOrderRGB>, convertRGB30FromRGB32<PixelOrderRGB>, convertA2RGB30PMToARGB64PM<PixelOrderRGB> }, // Format_A2RGB30_Premultiplied - { 0, 0, 0, 0, 0, 0, 8, 0, true, QPixelLayout::BPP8, convertAlpha8ToRGB32, convertAlpha8FromARGB32PM, 0, convertAlpha8ToRGB64 }, // Format_Alpha8 - { 0, 0, 0, 0, 0, 0, 0, 0, false, QPixelLayout::BPP8, convertGrayscale8ToRGB32, convertGrayscale8FromARGB32PM, convertGrayscale8FromRGB32, convertGrayscale8ToRGB64 } // Format_Grayscale8 -}; - -const FetchPixelsFunc qFetchPixels[QPixelLayout::BPPCount] = { - 0, // BPPNone - fetchPixels<QPixelLayout::BPP1MSB>, // BPP1MSB - fetchPixels<QPixelLayout::BPP1LSB>, // BPP1LSB - fetchPixels<QPixelLayout::BPP8>, // BPP8 - fetchPixels<QPixelLayout::BPP16>, // BPP16 - fetchPixels<QPixelLayout::BPP24>, // BPP24 - fetchPixels<QPixelLayout::BPP32> // BPP32 -}; - -StorePixelsFunc qStorePixels[QPixelLayout::BPPCount] = { - 0, // BPPNone - storePixels<QPixelLayout::BPP1MSB>, // BPP1MSB - storePixels<QPixelLayout::BPP1LSB>, // BPP1LSB - storePixels<QPixelLayout::BPP8>, // BPP8 - storePixels<QPixelLayout::BPP16>, // BPP16 - storePixels<QPixelLayout::BPP24>, // BPP24 - storePixels<QPixelLayout::BPP32> // BPP32 -}; - -typedef uint (QT_FASTCALL *FetchPixelFunc)(const uchar *src, int index); - -static const FetchPixelFunc qFetchPixel[QPixelLayout::BPPCount] = { - 0, // BPPNone - fetchPixel<QPixelLayout::BPP1MSB>, // BPP1MSB - fetchPixel<QPixelLayout::BPP1LSB>, // BPP1LSB - fetchPixel<QPixelLayout::BPP8>, // BPP8 - fetchPixel<QPixelLayout::BPP16>, // BPP16 - fetchPixel<QPixelLayout::BPP24>, // BPP24 - fetchPixel<QPixelLayout::BPP32> // BPP32 + QRgba64 *d = reinterpret_cast<QRgba64*>(dest) + index; + if (d != src) + memcpy(d, src, count * sizeof(QRgba64)); +} + +ConvertAndStorePixelsFunc64 qStoreFromRGBA64PM[QImage::NImageFormats] = { + nullptr, + nullptr, + nullptr, + nullptr, + storeGenericFromRGBA64PM<QImage::Format_RGB32>, + storeARGB32FromRGBA64PM, + storeGenericFromRGBA64PM<QImage::Format_ARGB32_Premultiplied>, + storeGenericFromRGBA64PM<QImage::Format_RGB16>, + storeGenericFromRGBA64PM<QImage::Format_ARGB8565_Premultiplied>, + storeGenericFromRGBA64PM<QImage::Format_RGB666>, + storeGenericFromRGBA64PM<QImage::Format_ARGB6666_Premultiplied>, + storeGenericFromRGBA64PM<QImage::Format_RGB555>, + storeGenericFromRGBA64PM<QImage::Format_ARGB8555_Premultiplied>, + storeGenericFromRGBA64PM<QImage::Format_RGB888>, + storeGenericFromRGBA64PM<QImage::Format_RGB444>, + storeGenericFromRGBA64PM<QImage::Format_ARGB4444_Premultiplied>, + storeGenericFromRGBA64PM<QImage::Format_RGBX8888>, + storeRGBA8888FromRGBA64PM, + storeGenericFromRGBA64PM<QImage::Format_RGBA8888_Premultiplied>, + storeRGB30FromRGBA64PM<PixelOrderBGR>, + storeRGB30FromRGBA64PM<PixelOrderBGR>, + storeRGB30FromRGBA64PM<PixelOrderRGB>, + storeRGB30FromRGBA64PM<PixelOrderRGB>, + storeGenericFromRGBA64PM<QImage::Format_Alpha8>, + storeGenericFromRGBA64PM<QImage::Format_Grayscale8>, + storeRGBX64FromRGBA64PM, + storeRGBA64FromRGBA64PM, + storeRGBA64PMFromRGBA64PM }; /* @@ -1185,23 +1641,23 @@ static uint * QT_FASTCALL destFetchRGB16(uint *buffer, QRasterBuffer *rasterBuff static uint *QT_FASTCALL destFetch(uint *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length) { const QPixelLayout *layout = &qPixelLayouts[rasterBuffer->format]; - const uint *ptr = qFetchPixels[layout->bpp](buffer, rasterBuffer->scanLine(y), x, length); - return const_cast<uint *>(layout->convertToARGB32PM(buffer, ptr, length, 0, 0)); + return const_cast<uint *>(layout->fetchToARGB32PM(buffer, rasterBuffer->scanLine(y), x, length, nullptr, nullptr)); +} + +static uint *QT_FASTCALL destFetchUndefined(uint *buffer, QRasterBuffer *, int, int, int) +{ + return buffer; } static QRgba64 *QT_FASTCALL destFetch64(QRgba64 *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length) { const QPixelLayout *layout = &qPixelLayouts[rasterBuffer->format]; - uint buffer32[buffer_size]; - const uint *ptr = qFetchPixels[layout->bpp](buffer32, rasterBuffer->scanLine(y), x, length); - return const_cast<QRgba64 *>(layout->convertToARGB64PM(buffer, ptr, length, 0, 0)); + return const_cast<QRgba64 *>(layout->fetchToRGBA64PM(buffer, rasterBuffer->scanLine(y), x, length, nullptr, nullptr)); } -static QRgba64 *QT_FASTCALL destFetch64uint32(QRgba64 *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length) +static QRgba64 * QT_FASTCALL destFetchRGB64(QRgba64 *, QRasterBuffer *rasterBuffer, int x, int y, int) { - const QPixelLayout *layout = &qPixelLayouts[rasterBuffer->format]; - const uint *src = ((const uint *)rasterBuffer->scanLine(y)) + x; - return const_cast<QRgba64 *>(layout->convertToARGB64PM(buffer, src, length, 0, 0)); + return (QRgba64 *)rasterBuffer->scanLine(y) + x; } static QRgba64 * QT_FASTCALL destFetch64Undefined(QRgba64 *buffer, QRasterBuffer *, int, int, int) @@ -1236,6 +1692,9 @@ static DestFetchProc destFetchProc[QImage::NImageFormats] = destFetch, // Format_A2RGB30_Premultiplied destFetch, // Format_Alpha8 destFetch, // Format_Grayscale8 + destFetch, // Format_RGBX64 + destFetch, // Format_RGBA64 + destFetch, // Format_RGBA64_Premultiplied }; static DestFetchProc64 destFetchProc64[QImage::NImageFormats] = @@ -1244,9 +1703,9 @@ static DestFetchProc64 destFetchProc64[QImage::NImageFormats] = 0, // Format_Mono, 0, // Format_MonoLSB 0, // Format_Indexed8 - destFetch64uint32, // Format_RGB32 - destFetch64uint32, // Format_ARGB32, - destFetch64uint32, // Format_ARGB32_Premultiplied + destFetch64, // Format_RGB32 + destFetch64, // Format_ARGB32, + destFetch64, // Format_ARGB32_Premultiplied destFetch64, // Format_RGB16 destFetch64, // Format_ARGB8565_Premultiplied destFetch64, // Format_RGB666 @@ -1256,15 +1715,18 @@ static DestFetchProc64 destFetchProc64[QImage::NImageFormats] = destFetch64, // Format_RGB888 destFetch64, // Format_RGB444 destFetch64, // Format_ARGB4444_Premultiplied - destFetch64uint32, // Format_RGBX8888 - destFetch64uint32, // Format_RGBA8888 - destFetch64uint32, // Format_RGBA8888_Premultiplied - destFetch64uint32, // Format_BGR30 - destFetch64uint32, // Format_A2BGR30_Premultiplied - destFetch64uint32, // Format_RGB30 - destFetch64uint32, // Format_A2RGB30_Premultiplied + destFetch64, // Format_RGBX8888 + destFetch64, // Format_RGBA8888 + destFetch64, // Format_RGBA8888_Premultiplied + destFetch64, // Format_BGR30 + destFetch64, // Format_A2BGR30_Premultiplied + destFetch64, // Format_RGB30 + destFetch64, // Format_A2RGB30_Premultiplied destFetch64, // Format_Alpha8 destFetch64, // Format_Grayscale8 + destFetchRGB64, // Format_RGBX64 + destFetch64, // Format_RGBA64 + destFetchRGB64, // Format_RGBA64_Premultiplied }; /* @@ -1365,143 +1827,29 @@ static void QT_FASTCALL destStoreRGB16(QRasterBuffer *rasterBuffer, int x, int y static void QT_FASTCALL destStore(QRasterBuffer *rasterBuffer, int x, int y, const uint *buffer, int length) { - uint buf[buffer_size]; const QPixelLayout *layout = &qPixelLayouts[rasterBuffer->format]; - StorePixelsFunc store = qStorePixels[layout->bpp]; + ConvertAndStorePixelsFunc store = layout->storeFromARGB32PM; + if (!layout->premultiplied && !layout->hasAlphaChannel) + store = layout->storeFromRGB32; uchar *dest = rasterBuffer->scanLine(y); - while (length) { - int l = qMin(length, buffer_size); - const uint *ptr = 0; - if (!layout->premultiplied && !layout->alphaWidth) - ptr = layout->convertFromRGB32(buf, buffer, l, 0, 0); - else - ptr = layout->convertFromARGB32PM(buf, buffer, l, 0, 0); - store(dest, ptr, x, l); - length -= l; - buffer += l; - x += l; - } -} - -static void QT_FASTCALL convertFromRgb64(uint *dest, const QRgba64 *src, int length) -{ - for (int i = 0; i < length; ++i) { - dest[i] = toArgb32(src[i]); - } + store(dest, buffer, x, length, nullptr, nullptr); } static void QT_FASTCALL destStore64(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) { - uint buf[buffer_size]; - const QPixelLayout *layout = &qPixelLayouts[rasterBuffer->format]; - StorePixelsFunc store = qStorePixels[layout->bpp]; + auto store = qStoreFromRGBA64PM[rasterBuffer->format]; uchar *dest = rasterBuffer->scanLine(y); - while (length) { - int l = qMin(length, buffer_size); - const uint *ptr = 0; - convertFromRgb64(buf, buffer, l); - if (!layout->premultiplied && !layout->alphaWidth) - ptr = layout->convertFromRGB32(buf, buf, l, 0, 0); - else - ptr = layout->convertFromARGB32PM(buf, buf, l, 0, 0); - store(dest, ptr, x, l); - length -= l; - buffer += l; - x += l; - } + store(dest, buffer, x, length, nullptr, nullptr); } -#ifdef __SSE2__ -template<QtPixelOrder PixelOrder> -static inline void qConvertARGB64PMToA2RGB30PM_sse2(uint *dest, const QRgba64 *buffer, int count) +static void QT_FASTCALL destStore64RGBA64(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) { - const __m128i gmask = _mm_set1_epi32(0x000ffc00); - const __m128i cmask = _mm_set1_epi32(0x000003ff); - int i = 0; - __m128i vr, vg, vb, va; - for (; i < count && uintptr_t(buffer) & 0xF; ++i) { - *dest++ = qConvertRgb64ToRgb30<PixelOrder>(*buffer++); - } - - for (; i < count-15; i += 16) { - // Repremultiplying is really expensive and hard to do in SIMD without AVX2, - // so we try to avoid it by checking if it is needed 16 samples at a time. - __m128i vOr = _mm_set1_epi32(0); - __m128i vAnd = _mm_set1_epi32(0xffffffff); - for (int j = 0; j < 16; j += 2) { - __m128i vs = _mm_load_si128((const __m128i*)(buffer + j)); - vOr = _mm_or_si128(vOr, vs); - vAnd = _mm_and_si128(vAnd, vs); - } - const quint16 orAlpha = ((uint)_mm_extract_epi16(vOr, 3)) | ((uint)_mm_extract_epi16(vOr, 7)); - const quint16 andAlpha = ((uint)_mm_extract_epi16(vAnd, 3)) & ((uint)_mm_extract_epi16(vAnd, 7)); - - if (andAlpha == 0xffff) { - for (int j = 0; j < 16; j += 2) { - __m128i vs = _mm_load_si128((const __m128i*)buffer); - buffer += 2; - vr = _mm_srli_epi64(vs, 6); - vg = _mm_srli_epi64(vs, 16 + 6 - 10); - vb = _mm_srli_epi64(vs, 32 + 6); - vr = _mm_and_si128(vr, cmask); - vg = _mm_and_si128(vg, gmask); - vb = _mm_and_si128(vb, cmask); - va = _mm_srli_epi64(vs, 48 + 14); - if (PixelOrder == PixelOrderRGB) - vr = _mm_slli_epi32(vr, 20); - else - vb = _mm_slli_epi32(vb, 20); - va = _mm_slli_epi32(va, 30); - __m128i vd = _mm_or_si128(_mm_or_si128(vr, vg), _mm_or_si128(vb, va)); - vd = _mm_shuffle_epi32(vd, _MM_SHUFFLE(3, 1, 2, 0)); - _mm_storel_epi64((__m128i*)dest, vd); - dest += 2; - } - } else if (orAlpha == 0) { - for (int j = 0; j < 16; ++j) { - *dest++ = 0; - buffer++; - } - } else { - for (int j = 0; j < 16; ++j) - *dest++ = qConvertRgb64ToRgb30<PixelOrder>(*buffer++); - } - } - - SIMD_EPILOGUE(i, count, 15) - *dest++ = qConvertRgb64ToRgb30<PixelOrder>(*buffer++); -} -#endif - -static void QT_FASTCALL destStore64ARGB32(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) -{ - uint *dest = (uint*)rasterBuffer->scanLine(y) + x; + QRgba64 *dest = reinterpret_cast<QRgba64*>(rasterBuffer->scanLine(y)) + x; for (int i = 0; i < length; ++i) { - dest[i] = toArgb32(buffer[i].unpremultiplied()); + dest[i] = buffer[i].unpremultiplied(); } } -static void QT_FASTCALL destStore64RGBA8888(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) -{ - uint *dest = (uint*)rasterBuffer->scanLine(y) + x; - for (int i = 0; i < length; ++i) { - dest[i] = toRgba8888(buffer[i].unpremultiplied()); - } -} - -template<QtPixelOrder PixelOrder> -static void QT_FASTCALL destStore64RGB30(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) -{ - uint *dest = (uint*)rasterBuffer->scanLine(y) + x; -#ifdef __SSE2__ - qConvertARGB64PMToA2RGB30PM_sse2<PixelOrder>(dest, buffer, length); -#else - for (int i = 0; i < length; ++i) { - dest[i] = qConvertRgb64ToRgb30<PixelOrder>(buffer[i]); - } -#endif -} - static DestStoreProc destStoreProc[QImage::NImageFormats] = { 0, // Format_Invalid @@ -1529,16 +1877,19 @@ static DestStoreProc destStoreProc[QImage::NImageFormats] = destStore, // Format_A2RGB30_Premultiplied destStore, // Format_Alpha8 destStore, // Format_Grayscale8 + destStore, // Format_RGBX64 + destStore, // Format_RGBA64 + destStore, // Format_RGBA64_Premultiplied }; static DestStoreProc64 destStoreProc64[QImage::NImageFormats] = { 0, // Format_Invalid - destStore64, // Format_Mono, - destStore64, // Format_MonoLSB + 0, // Format_Mono, + 0, // Format_MonoLSB 0, // Format_Indexed8 destStore64, // Format_RGB32 - destStore64ARGB32, // Format_ARGB32, + destStore64, // Format_ARGB32, destStore64, // Format_ARGB32_Premultiplied destStore64, // Format_RGB16 destStore64, // Format_ARGB8565_Premultiplied @@ -1550,14 +1901,17 @@ static DestStoreProc64 destStoreProc64[QImage::NImageFormats] = destStore64, // Format_RGB444 destStore64, // Format_ARGB4444_Premultiplied destStore64, // Format_RGBX8888 - destStore64RGBA8888, // Format_RGBA8888 + destStore64, // Format_RGBA8888 destStore64, // Format_RGBA8888_Premultiplied - destStore64RGB30<PixelOrderBGR>, // Format_BGR30 - destStore64RGB30<PixelOrderBGR>, // Format_A2BGR30_Premultiplied - destStore64RGB30<PixelOrderRGB>, // Format_RGB30 - destStore64RGB30<PixelOrderRGB>, // Format_A2RGB30_Premultiplied + destStore64, // Format_BGR30 + destStore64, // Format_A2BGR30_Premultiplied + destStore64, // Format_RGB30 + destStore64, // Format_A2RGB30_Premultiplied destStore64, // Format_Alpha8 destStore64, // Format_Grayscale8 + 0, // Format_RGBX64 + destStore64RGBA64, // Format_RGBA64 + 0 // Format_RGBA64_Premultiplied }; /* @@ -1589,15 +1943,14 @@ static const uint *QT_FASTCALL fetchUntransformed(uint *buffer, const Operator * const QSpanData *data, int y, int x, int length) { const QPixelLayout *layout = &qPixelLayouts[data->texture.format]; - const uint *ptr = qFetchPixels[layout->bpp](buffer, data->texture.scanLine(y), x, length); - return layout->convertToARGB32PM(buffer, ptr, length, data->texture.colorTable, 0); + return layout->fetchToARGB32PM(buffer, data->texture.scanLine(y), x, length, data->texture.colorTable, nullptr); } static const uint *QT_FASTCALL fetchUntransformedARGB32PM(uint *, const Operator *, const QSpanData *data, int y, int x, int) { const uchar *scanLine = data->texture.scanLine(y); - return ((const uint *)scanLine) + x; + return reinterpret_cast<const uint *>(scanLine) + x; } static const uint *QT_FASTCALL fetchUntransformedRGB16(uint *buffer, const Operator *, @@ -1614,39 +1967,47 @@ static const QRgba64 *QT_FASTCALL fetchUntransformed64(QRgba64 *buffer, const Op const QSpanData *data, int y, int x, int length) { const QPixelLayout *layout = &qPixelLayouts[data->texture.format]; - if (layout->bpp != QPixelLayout::BPP32) { - uint buffer32[buffer_size]; - const uint *ptr = qFetchPixels[layout->bpp](buffer32, data->texture.scanLine(y), x, length); - return layout->convertToARGB64PM(buffer, ptr, length, data->texture.colorTable, 0); + return layout->fetchToRGBA64PM(buffer, data->texture.scanLine(y), x, length, data->texture.colorTable, nullptr); +} + +static const QRgba64 *QT_FASTCALL fetchUntransformedRGBA64PM(QRgba64 *, const Operator *, + const QSpanData *data, int y, int x, int) +{ + const uchar *scanLine = data->texture.scanLine(y); + return reinterpret_cast<const QRgba64 *>(scanLine) + x; +} + +template<TextureBlendType blendType> +inline void fetchTransformed_pixelBounds(int max, int l1, int l2, int &v) +{ + Q_STATIC_ASSERT(blendType == BlendTransformed || blendType == BlendTransformedTiled); + if (blendType == BlendTransformedTiled) { + if (v < 0 || v >= max) { + v %= max; + if (v < 0) v += max; + } } else { - const uint *src = (const uint *)data->texture.scanLine(y) + x; - return layout->convertToARGB64PM(buffer, src, length, data->texture.colorTable, 0); + v = qBound(l1, v, l2); } } -template<TextureBlendType blendType, QPixelLayout::BPP bpp> -static const uint *QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, const QSpanData *data, - int y, int x, int length) +template<TextureBlendType blendType, QPixelLayout::BPP bpp, typename T> +static void QT_FASTCALL fetchTransformed_fetcher(T *buffer, const QSpanData *data, + int y, int x, int length) { Q_STATIC_ASSERT(blendType == BlendTransformed || blendType == BlendTransformedTiled); - const int image_width = data->texture.width; - const int image_height = data->texture.height; - const int image_x1 = data->texture.x1; - const int image_y1 = data->texture.y1; - const int image_x2 = data->texture.x2 - 1; - const int image_y2 = data->texture.y2 - 1; + const QTextureData &image = data->texture; const qreal cx = x + qreal(0.5); const qreal cy = y + qreal(0.5); + constexpr bool useFetch = (bpp < QPixelLayout::BPP32) && sizeof(T) == sizeof(uint); const QPixelLayout *layout = &qPixelLayouts[data->texture.format]; - if (bpp != QPixelLayout::BPPNone) // Like this to not ICE on GCC 5.3.1 + if (!useFetch) Q_ASSERT(layout->bpp == bpp); // When templated 'fetch' should be inlined at compile time: const FetchPixelFunc fetch = (bpp == QPixelLayout::BPPNone) ? qFetchPixel[layout->bpp] : FetchPixelFunc(fetchPixel<bpp>); - uint *const end = buffer + length; - uint *b = buffer; if (data->fast_matrix) { // The increment pr x in the scanline int fdx = (int)(data->m11 * fixed_scale); @@ -1657,24 +2018,105 @@ static const uint *QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, int fy = int((data->m22 * cy + data->m12 * cx + data->dy) * fixed_scale); - while (b < end) { - int px = fx >> 16; - int py = fy >> 16; - - if (blendType == BlendTransformedTiled) { - px %= image_width; - py %= image_height; - if (px < 0) px += image_width; - if (py < 0) py += image_height; - } else { - px = qBound(image_x1, px, image_x2); - py = qBound(image_y1, py, image_y2); + if (fdy == 0) { // simple scale, no rotation or shear + int py = (fy >> 16); + fetchTransformed_pixelBounds<blendType>(image.height, image.y1, image.y2 - 1, py); + const uchar *src = image.scanLine(py); + + int i = 0; + if (blendType == BlendTransformed) { + int fastLen = length; + if (fdx > 0) + fastLen = qMin(fastLen, int((qint64(image.x2 - 1) * fixed_scale - fx) / fdx)); + else if (fdx < 0) + fastLen = qMin(fastLen, int((qint64(image.x1) * fixed_scale - fx) / fdx)); + + for (; i < fastLen; ++i) { + int x1 = (fx >> 16); + int x2 = x1; + fetchTransformed_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, x1); + if (x1 == x2) + break; + if (useFetch) + buffer[i] = fetch(src, x1); + else + buffer[i] = reinterpret_cast<const T*>(src)[x1]; + fx += fdx; + } + + for (; i < fastLen; ++i) { + int px = (fx >> 16); + if (useFetch) + buffer[i] = fetch(src, px); + else + buffer[i] = reinterpret_cast<const T*>(src)[px]; + fx += fdx; + } } - *b = fetch(data->texture.scanLine(py), px); - fx += fdx; - fy += fdy; - ++b; + for (; i < length; ++i) { + int px = (fx >> 16); + fetchTransformed_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, px); + if (useFetch) + buffer[i] = fetch(src, px); + else + buffer[i] = reinterpret_cast<const T*>(src)[px]; + fx += fdx; + } + } else { // rotation or shear + int i = 0; + if (blendType == BlendTransformed) { + int fastLen = length; + if (fdx > 0) + fastLen = qMin(fastLen, int((qint64(image.x2 - 1) * fixed_scale - fx) / fdx)); + else if (fdx < 0) + fastLen = qMin(fastLen, int((qint64(image.x1) * fixed_scale - fx) / fdx)); + if (fdy > 0) + fastLen = qMin(fastLen, int((qint64(image.y2 - 1) * fixed_scale - fy) / fdy)); + else if (fdy < 0) + fastLen = qMin(fastLen, int((qint64(image.y1) * fixed_scale - fy) / fdy)); + + for (; i < fastLen; ++i) { + int x1 = (fx >> 16); + int y1 = (fy >> 16); + int x2 = x1; + int y2 = y1; + fetchTransformed_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, x1); + fetchTransformed_pixelBounds<blendType>(image.height, image.y1, image.y2 - 1, y1); + if (x1 == x2 && y1 == y2) + break; + if (useFetch) + buffer[i] = fetch(image.scanLine(y1), x1); + else + buffer[i] = reinterpret_cast<const T*>(image.scanLine(y1))[x1]; + fx += fdx; + fy += fdy; + } + + for (; i < fastLen; ++i) { + int px = (fx >> 16); + int py = (fy >> 16); + if (useFetch) + buffer[i] = fetch(image.scanLine(py), px); + else + buffer[i] = reinterpret_cast<const T*>(image.scanLine(py))[px]; + fx += fdx; + fy += fdy; + } + } + + for (; i < length; ++i) { + int px = (fx >> 16); + int py = (fy >> 16); + fetchTransformed_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, px); + fetchTransformed_pixelBounds<blendType>(image.height, image.y1, image.y2 - 1, py); + if (useFetch) + buffer[i] = fetch(image.scanLine(py), px); + else + buffer[i] = reinterpret_cast<const T*>(image.scanLine(py))[px]; + fx += fdx; + fy += fdy; + } } } else { const qreal fdx = data->m11; @@ -1685,23 +2127,21 @@ static const uint *QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, qreal fy = data->m22 * cy + data->m12 * cx + data->dy; qreal fw = data->m23 * cy + data->m13 * cx + data->m33; + T *const end = buffer + length; + T *b = buffer; 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); - - if (blendType == BlendTransformedTiled) { - px %= image_width; - py %= image_height; - if (px < 0) px += image_width; - if (py < 0) py += image_height; - } else { - px = qBound(image_x1, px, image_x2); - py = qBound(image_y1, py, image_y2); - } - *b = fetch(data->texture.scanLine(py), px); + int px = qFloor(tx); + int py = qFloor(ty); + + fetchTransformed_pixelBounds<blendType>(image.height, image.y1, image.y2 - 1, py); + fetchTransformed_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, px); + if (useFetch) + *b = fetch(image.scanLine(py), px); + else + *b = reinterpret_cast<const T*>(image.scanLine(py))[px]; fx += fdx; fy += fdy; @@ -1713,115 +2153,37 @@ static const uint *QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, ++b; } } - return layout->convertToARGB32PM(buffer, buffer, length, data->texture.colorTable, 0); +} + +template<TextureBlendType blendType, QPixelLayout::BPP bpp> +static const uint *QT_FASTCALL fetchTransformed(uint *buffer, const Operator *, const QSpanData *data, + int y, int x, int length) +{ + Q_STATIC_ASSERT(blendType == BlendTransformed || blendType == BlendTransformedTiled); + const QPixelLayout *layout = &qPixelLayouts[data->texture.format]; + fetchTransformed_fetcher<blendType, bpp, uint>(buffer, data, y, x, length); + layout->convertToARGB32PM(buffer, length, data->texture.colorTable); + return buffer; } template<TextureBlendType blendType> /* either BlendTransformed or BlendTransformedTiled */ static const QRgba64 *QT_FASTCALL fetchTransformed64(QRgba64 *buffer, const Operator *, const QSpanData *data, int y, int x, int length) { - const int image_width = data->texture.width; - const int image_height = data->texture.height; - const int image_x1 = data->texture.x1; - const int image_y1 = data->texture.y1; - const int image_x2 = data->texture.x2 - 1; - const int image_y2 = data->texture.y2 - 1; - - const qreal cx = x + qreal(0.5); - const qreal cy = y + qreal(0.5); - const QPixelLayout *layout = &qPixelLayouts[data->texture.format]; - FetchPixelFunc fetch = qFetchPixel[layout->bpp]; - const QVector<QRgb> *clut = data->texture.colorTable; - - uint buffer32[buffer_size]; - QRgba64 *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); - - int i = 0, j = 0; - while (i < length) { - if (j == buffer_size) { - layout->convertToARGB64PM(b, buffer32, buffer_size, clut, 0); - b += buffer_size; - j = 0; - } - int px = fx >> 16; - int py = fy >> 16; - - if (blendType == BlendTransformedTiled) { - px %= image_width; - py %= image_height; - if (px < 0) px += image_width; - if (py < 0) py += image_height; - } else { - px = qBound(image_x1, px, image_x2); - py = qBound(image_y1, py, image_y2); - } - buffer32[j] = fetch(data->texture.scanLine(py), px); - - fx += fdx; - fy += fdy; - ++i; ++j; - } - if (j > 0) { - layout->convertToARGB64PM(b, buffer32, j, clut, 0); - b += j; - } - } 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; - - int i = 0, j = 0; - while (i < length) { - if (j == buffer_size) { - layout->convertToARGB64PM(b, buffer32, buffer_size, clut, 0); - b += buffer_size; - j = 0; - } - 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); - - if (blendType == BlendTransformedTiled) { - px %= image_width; - py %= image_height; - if (px < 0) px += image_width; - if (py < 0) py += image_height; - } else { - px = qBound(image_x1, px, image_x2); - py = qBound(image_y1, py, image_y2); - } - buffer32[j] = fetch(data->texture.scanLine(py), px); - - fx += fdx; - fy += fdy; - fw += fdw; - //force increment to avoid /0 - if (!fw) { - fw += fdw; - } - ++i; ++j; - } - if (j > 0) { - layout->convertToARGB64PM(b, buffer32, j, clut, 0); - b += j; - } + if (layout->bpp != QPixelLayout::BPP64) { + uint buffer32[BufferSize]; + Q_ASSERT(length <= BufferSize); + if (layout->bpp == QPixelLayout::BPP32) + fetchTransformed_fetcher<blendType, QPixelLayout::BPP32, uint>(buffer32, data, y, x, length); + else + fetchTransformed_fetcher<blendType, QPixelLayout::BPPNone, uint>(buffer32, data, y, x, length); + return layout->convertToRGBA64PM(buffer, buffer32, length, data->texture.colorTable, nullptr); } + + fetchTransformed_fetcher<blendType, QPixelLayout::BPP64, QRgba64>(buffer, data, y, x, length); + if (data->texture.format == QImage::Format_RGBA64) + convertRGBA64ToRGBA64PM(buffer, length); return buffer; } @@ -1915,43 +2277,6 @@ static inline uint interpolate_4_pixels_16(uint tl, uint tr, uint bl, uint br, u } #endif -#if defined(__SSE2__) -static inline QRgba64 interpolate_4_pixels_rgb64(QRgba64 t[], QRgba64 b[], uint distx, uint disty) -{ - __m128i vt = _mm_loadu_si128((const __m128i*)t); - if (disty) { - __m128i vb = _mm_loadu_si128((const __m128i*)b); - vt = _mm_mulhi_epu16(vt, _mm_set1_epi16(0x10000 - disty)); - vb = _mm_mulhi_epu16(vb, _mm_set1_epi16(disty)); - vt = _mm_add_epi16(vt, vb); - } - if (distx) { - const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0)); - const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(0x10000 - distx), _MM_SHUFFLE(0, 0, 0, 0)); - vt = _mm_mulhi_epu16(vt, _mm_unpacklo_epi64(vidistx, vdistx)); - vt = _mm_add_epi16(vt, _mm_srli_si128(vt, 8)); - } -#ifdef Q_PROCESSOR_X86_64 - return QRgba64::fromRgba64(_mm_cvtsi128_si64(vt)); -#else - QRgba64 out; - _mm_storel_epi64((__m128i*)&out, vt); - return out; -#endif -} -#else -static inline QRgba64 interpolate_4_pixels_rgb64(QRgba64 t[], QRgba64 b[], uint distx, uint disty) -{ - const uint dx = distx>>8; - const uint dy = disty>>8; - const uint idx = 256 - dx; - const uint idy = 256 - dy; - QRgba64 xtop = interpolate256(t[0], idx, t[1], dx); - QRgba64 xbot = interpolate256(b[0], idx, b[1], dx); - return interpolate256(xtop, idy, xbot, dy); -} -#endif - template<TextureBlendType blendType> void fetchTransformedBilinear_pixelBounds(int max, int l1, int l2, int &v1, int &v2); @@ -1982,7 +2307,7 @@ inline void fetchTransformedBilinear_pixelBounds<BlendTransformedBilinear>(int, } enum FastTransformTypes { - SimpleUpscaleTransform, + SimpleScaleTransform, UpscaleTransform, DownscaleTransform, RotateTransform, @@ -1990,11 +2315,38 @@ enum FastTransformTypes { NFastTransformTypes }; +// Completes the partial interpolation stored in IntermediateBuffer. +// by performing the x-axis interpolation and joining the RB and AG buffers. +static void QT_FASTCALL intermediate_adder(uint *b, uint *end, const IntermediateBuffer &intermediate, int offset, int &fx, int fdx) +{ +#if defined(QT_COMPILER_SUPPORTS_AVX2) + extern void QT_FASTCALL intermediate_adder_avx2(uint *b, uint *end, const IntermediateBuffer &intermediate, int offset, int &fx, int fdx); + if (qCpuHasFeature(AVX2)) + return intermediate_adder_avx2(b, end, intermediate, offset, fx, fdx); +#endif + + // Switch to intermediate buffer coordinates + fx -= offset * fixed_scale; + + while (b < end) { + const int x = (fx >> 16); + + const uint distx = (fx & 0x0000ffff) >> 8; + const uint idistx = 256 - distx; + const uint rb = (intermediate.buffer_rb[x] * idistx + intermediate.buffer_rb[x + 1] * distx) & 0xff00ff00; + const uint ag = (intermediate.buffer_ag[x] * idistx + intermediate.buffer_ag[x + 1] * distx) & 0xff00ff00; + *b = (rb >> 8) | ag; + b++; + fx += fdx; + } + fx += offset * fixed_scale; +} + typedef void (QT_FASTCALL *BilinearFastTransformHelper)(uint *b, uint *end, const QTextureData &image, int &fx, int &fy, int fdx, int fdy); template<TextureBlendType blendType> -static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper(uint *b, uint *end, const QTextureData &image, - int &fx, int &fy, int fdx, int /*fdy*/) +static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_scale_helper(uint *b, uint *end, const QTextureData &image, + int &fx, int &fy, int fdx, int /*fdy*/) { int y1 = (fy >> 16); int y2; @@ -2011,16 +2363,12 @@ static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper(u const int offset = (fx + adjust) >> 16; int x = offset; - // The idea is first to do the interpolation between the row s1 and the row s2 - // into an intermediate buffer, then we interpolate between two pixel of this buffer. - - // intermediate_buffer[0] is a buffer of red-blue component of the pixel, in the form 0x00RR00BB - // intermediate_buffer[1] is the alpha-green component of the pixel, in the form 0x00AA00GG - // +1 for the last pixel to interpolate with, and +1 for rounding errors. - quint32 intermediate_buffer[2][buffer_size + 2]; - // count is the size used in the intermediate_buffer. + IntermediateBuffer intermediate; + // count is the size used in the intermediate.buffer. int count = (qint64(length) * qAbs(fdx) + fixed_scale - 1) / fixed_scale + 2; - Q_ASSERT(count <= buffer_size + 2); //length is supposed to be <= buffer_size and data->m11 < 1 in this case + // length is supposed to be <= BufferSize either because data->m11 < 1 or + // data->m11 < 2, and any larger buffers split + Q_ASSERT(count <= BufferSize + 2); int f = 0; int lim = count; if (blendType == BlendTransformedBilinearTiled) { @@ -2035,8 +2383,8 @@ static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper(u quint32 rb = (((t & 0xff00ff) * idisty + (b & 0xff00ff) * disty) >> 8) & 0xff00ff; quint32 ag = ((((t>>8) & 0xff00ff) * idisty + ((b>>8) & 0xff00ff) * disty) >> 8) & 0xff00ff; do { - intermediate_buffer[0][f] = rb; - intermediate_buffer[1][f] = ag; + intermediate.buffer_rb[f] = rb; + intermediate.buffer_ag[f] = ag; f++; x++; } while (x < image.x1 && f < lim); @@ -2069,10 +2417,10 @@ static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper(u // Add the values, and shift to only keep 8 significant bits per colors __m128i rAG =_mm_add_epi16(topAG, bottomAG); rAG = _mm_srli_epi16(rAG, 8); - _mm_storeu_si128((__m128i*)(&intermediate_buffer[1][f]), rAG); + _mm_storeu_si128((__m128i*)(&intermediate.buffer_ag[f]), rAG); __m128i rRB =_mm_add_epi16(topRB, bottomRB); rRB = _mm_srli_epi16(rRB, 8); - _mm_storeu_si128((__m128i*)(&intermediate_buffer[0][f]), rRB); + _mm_storeu_si128((__m128i*)(&intermediate.buffer_rb[f]), rRB); } #elif defined(__ARM_NEON__) const int16x8_t disty_ = vdupq_n_s16(disty); @@ -2099,10 +2447,10 @@ static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper(u // Add the values, and shift to only keep 8 significant bits per colors int16x8_t rAG = vaddq_s16(topAG, bottomAG); rAG = vreinterpretq_s16_u16(vshrq_n_u16(vreinterpretq_u16_s16(rAG), 8)); - vst1q_s16((int16_t*)(&intermediate_buffer[1][f]), rAG); + vst1q_s16((int16_t*)(&intermediate.buffer_ag[f]), rAG); int16x8_t rRB = vaddq_s16(topRB, bottomRB); rRB = vreinterpretq_s16_u16(vshrq_n_u16(vreinterpretq_u16_s16(rRB), 8)); - vst1q_s16((int16_t*)(&intermediate_buffer[0][f]), rRB); + vst1q_s16((int16_t*)(&intermediate.buffer_rb[f]), rRB); } #endif } @@ -2116,28 +2464,13 @@ static void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper(u uint t = s1[x]; uint b = s2[x]; - intermediate_buffer[0][f] = (((t & 0xff00ff) * idisty + (b & 0xff00ff) * disty) >> 8) & 0xff00ff; - intermediate_buffer[1][f] = ((((t>>8) & 0xff00ff) * idisty + ((b>>8) & 0xff00ff) * disty) >> 8) & 0xff00ff; + intermediate.buffer_rb[f] = (((t & 0xff00ff) * idisty + (b & 0xff00ff) * disty) >> 8) & 0xff00ff; + intermediate.buffer_ag[f] = ((((t>>8) & 0xff00ff) * idisty + ((b>>8) & 0xff00ff) * disty) >> 8) & 0xff00ff; x++; } - // Now interpolate the values from the intermediate_buffer to get the final result. - fx -= offset * fixed_scale; // Switch to intermediate buffer coordinates - - while (b < end) { - int x1 = (fx >> 16); - int x2 = x1 + 1; - Q_ASSERT(x1 >= 0); - Q_ASSERT(x2 < count); - - int distx = (fx & 0x0000ffff) >> 8; - int idistx = 256 - distx; - int rb = ((intermediate_buffer[0][x1] * idistx + intermediate_buffer[0][x2] * distx) >> 8) & 0xff00ff; - int ag = (intermediate_buffer[1][x1] * idistx + intermediate_buffer[1][x2] * distx) & 0xff00ff00; - *b = rb | ag; - b++; - fx += fdx; - } + // Now interpolate the values from the intermediate.buffer to get the final result. + intermediate_adder(b, end, intermediate, offset, fx, fdx); } template<TextureBlendType blendType> @@ -2603,14 +2936,14 @@ static void QT_FASTCALL fetchTransformedBilinearARGB32PM_fast_rotate_helper(uint static BilinearFastTransformHelper bilinearFastTransformHelperARGB32PM[2][NFastTransformTypes] = { { - fetchTransformedBilinearARGB32PM_simple_upscale_helper<BlendTransformedBilinear>, + fetchTransformedBilinearARGB32PM_simple_scale_helper<BlendTransformedBilinear>, fetchTransformedBilinearARGB32PM_upscale_helper<BlendTransformedBilinear>, fetchTransformedBilinearARGB32PM_downscale_helper<BlendTransformedBilinear>, fetchTransformedBilinearARGB32PM_rotate_helper<BlendTransformedBilinear>, fetchTransformedBilinearARGB32PM_fast_rotate_helper<BlendTransformedBilinear> }, { - fetchTransformedBilinearARGB32PM_simple_upscale_helper<BlendTransformedBilinearTiled>, + fetchTransformedBilinearARGB32PM_simple_scale_helper<BlendTransformedBilinearTiled>, fetchTransformedBilinearARGB32PM_upscale_helper<BlendTransformedBilinearTiled>, fetchTransformedBilinearARGB32PM_downscale_helper<BlendTransformedBilinearTiled>, fetchTransformedBilinearARGB32PM_rotate_helper<BlendTransformedBilinearTiled>, @@ -2645,7 +2978,13 @@ static const uint * QT_FASTCALL fetchTransformedBilinearARGB32PM(uint *buffer, c if (fdy == 0) { // simple scale, no rotation or shear if (qAbs(fdx) <= fixed_scale) { // simple scale up on X - bilinearFastTransformHelperARGB32PM[tiled][SimpleUpscaleTransform](b, end, data->texture, fx, fy, fdx, fdy); + bilinearFastTransformHelperARGB32PM[tiled][SimpleScaleTransform](b, end, data->texture, fx, fy, fdx, fdy); + } else if (qAbs(fdx) <= 2 * fixed_scale) { + // simple scale down on X, less than 2x + const int mid = (length * 2 < BufferSize) ? length : ((length + 1) / 2); + bilinearFastTransformHelperARGB32PM[tiled][SimpleScaleTransform](buffer, buffer + mid, data->texture, fx, fy, fdx, fdy); + if (mid != length) + bilinearFastTransformHelperARGB32PM[tiled][SimpleScaleTransform](buffer + mid, buffer + length, data->texture, fx, fy, fdx, fdy); } else if (qAbs(data->m22) < qreal(1./8.)) { // scale up more than 8x (on Y) bilinearFastTransformHelperARGB32PM[tiled][UpscaleTransform](b, end, data->texture, fx, fy, fdx, fdy); @@ -2713,16 +3052,13 @@ static const uint * QT_FASTCALL fetchTransformedBilinearARGB32PM(uint *buffer, c return buffer; } -template<TextureBlendType blendType, QPixelLayout::BPP bpp> -static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, uint *end, const QTextureData &image, - int &fx, int &fy, int fdx, int /*fdy*/) +template<TextureBlendType blendType> +static void QT_FASTCALL fetchTransformedBilinear_simple_scale_helper(uint *b, uint *end, const QTextureData &image, + int &fx, int &fy, int fdx, int /*fdy*/) { const QPixelLayout *layout = &qPixelLayouts[image.format]; const QVector<QRgb> *clut = image.colorTable; - Q_ASSERT(bpp == QPixelLayout::BPPNone || bpp == layout->bpp); - // When templated 'fetch' should be inlined at compile time: - const FetchPixelsFunc fetch = (bpp == QPixelLayout::BPPNone) ? qFetchPixels[layout->bpp] : fetchPixels<bpp>; - const ConvertFunc convertToARGB32PM = layout->convertToARGB32PM; + const FetchAndConvertPixelsFunc fetch = layout->fetchToARGB32PM; int y1 = (fy >> 16); int y2; @@ -2739,16 +3075,14 @@ static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, const int offset = (fx + adjust) >> 16; int x = offset; - // The idea is first to do the interpolation between the row s1 and the row s2 - // into an intermediate buffer, then we interpolate between two pixel of this buffer. - // +1 for the last pixel to interpolate with, and +1 for rounding errors. - uint buf1[buffer_size + 2]; - uint buf2[buffer_size + 2]; + IntermediateBuffer intermediate; + uint *buf1 = intermediate.buffer_rb; + uint *buf2 = intermediate.buffer_ag; const uint *ptr1; const uint *ptr2; int count = (qint64(length) * qAbs(fdx) + fixed_scale - 1) / fixed_scale + 2; - Q_ASSERT(count <= buffer_size + 2); //length is supposed to be <= buffer_size and data->m11 < 1 in this case + Q_ASSERT(count <= BufferSize + 2); if (blendType == BlendTransformedBilinearTiled) { x %= image.width; @@ -2757,10 +3091,8 @@ static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, int len1 = qMin(count, image.width - x); int len2 = qMin(x, count - len1); - ptr1 = fetch(buf1, s1, x, len1); - ptr1 = convertToARGB32PM(buf1, ptr1, len1, clut, 0); - ptr2 = fetch(buf2, s2, x, len1); - ptr2 = convertToARGB32PM(buf2, ptr2, len1, clut, 0); + ptr1 = fetch(buf1, s1, x, len1, clut, nullptr); + ptr2 = fetch(buf2, s2, x, len1, clut, nullptr); for (int i = 0; i < len1; ++i) { uint t = ptr1[i]; uint b = ptr2[i]; @@ -2769,10 +3101,8 @@ static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, } if (len2) { - ptr1 = fetch(buf1 + len1, s1, 0, len2); - ptr1 = convertToARGB32PM(buf1 + len1, ptr1, len2, clut, 0); - ptr2 = fetch(buf2 + len1, s2, 0, len2); - ptr2 = convertToARGB32PM(buf2 + len1, ptr2, len2, clut, 0); + ptr1 = fetch(buf1 + len1, s1, 0, len2, clut, nullptr); + ptr2 = fetch(buf2 + len1, s2, 0, len2, clut, nullptr); for (int i = 0; i < len2; ++i) { uint t = ptr1[i]; uint b = ptr2[i]; @@ -2780,6 +3110,7 @@ static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, buf2[i + len1] = ((((t >> 8) & 0xff00ff) * idisty + ((b >> 8) & 0xff00ff) * disty) >> 8) & 0xff00ff; } } + // Generate the rest by repeatedly repeating the previous set of pixels for (int i = image.width; i < count; ++i) { buf1[i] = buf1[i - image.width]; buf2[i] = buf2[i - image.width]; @@ -2790,10 +3121,8 @@ static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, int len = qMax(1, end - start); int leading = start - x; - ptr1 = fetch(buf1 + leading, s1, start, len); - ptr1 = convertToARGB32PM(buf1 + leading, ptr1, len, clut, 0); - ptr2 = fetch(buf2 + leading, s2, start, len); - ptr2 = convertToARGB32PM(buf2 + leading, ptr2, len, clut, 0); + ptr1 = fetch(buf1 + leading, s1, start, len, clut, nullptr); + ptr2 = fetch(buf2 + leading, s2, start, len, clut, nullptr); for (int i = 0; i < len; ++i) { uint t = ptr1[i]; @@ -2812,35 +3141,21 @@ static void QT_FASTCALL fetchTransformedBilinear_simple_upscale_helper(uint *b, } } - // Now interpolate the values from the intermediate_buffer to get the final result. - fx -= offset * fixed_scale; // Switch to intermediate buffer coordinates - - while (b < end) { - int x1 = (fx >> 16); - int x2 = x1 + 1; - Q_ASSERT(x1 >= 0); - Q_ASSERT(x2 < count); - - int distx = (fx & 0x0000ffff) >> 8; - int idistx = 256 - distx; - int rb = ((buf1[x1] * idistx + buf1[x2] * distx) >> 8) & 0xff00ff; - int ag = (buf2[x1] * idistx + buf2[x2] * distx) & 0xff00ff00; - *b++ = rb | ag; - fx += fdx; - } + // Now interpolate the values from the intermediate.buffer to get the final result. + intermediate_adder(b, end, intermediate, offset, fx, fdx); } -typedef void (QT_FASTCALL *BilinearFastTransformFetcher)(uint *buf1, uint *buf2, const int len, const QTextureData &image, - int fx, int fy, const int fdx, const int fdy); - -template<TextureBlendType blendType, QPixelLayout::BPP bpp> -static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, const int len, const QTextureData &image, +template<TextureBlendType blendType, QPixelLayout::BPP bpp, typename T> +static void QT_FASTCALL fetchTransformedBilinear_fetcher(T *buf1, T *buf2, const int len, const QTextureData &image, int fx, int fy, const int fdx, const int fdy) { const QPixelLayout &layout = qPixelLayouts[image.format]; - Q_ASSERT(bpp == QPixelLayout::BPPNone || bpp == layout.bpp); - // When templated 'fetch1' should be inlined at compile time: + constexpr bool useFetch = (bpp < QPixelLayout::BPP32); + if (useFetch) + Q_ASSERT(sizeof(T) == sizeof(uint)); + else + Q_ASSERT(layout.bpp == bpp); const FetchPixelFunc fetch1 = (bpp == QPixelLayout::BPPNone) ? qFetchPixel[layout.bpp] : fetchPixel<bpp>; if (fdy == 0) { int y1 = (fy >> 16); @@ -2857,8 +3172,13 @@ static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, fetchTransformedBilinear_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, x1, x2); if (x1 != x2) break; - buf1[i * 2 + 0] = buf1[i * 2 + 1] = fetch1(s1, x1); - buf2[i * 2 + 0] = buf2[i * 2 + 1] = fetch1(s2, x1); + if (useFetch) { + buf1[i * 2 + 0] = buf1[i * 2 + 1] = fetch1(s1, x1); + buf2[i * 2 + 0] = buf2[i * 2 + 1] = fetch1(s2, x1); + } else { + buf1[i * 2 + 0] = buf1[i * 2 + 1] = reinterpret_cast<const T *>(s1)[x1]; + buf2[i * 2 + 0] = buf2[i * 2 + 1] = reinterpret_cast<const T *>(s2)[x1]; + } fx += fdx; } int fastLen = len; @@ -2869,10 +3189,17 @@ static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, for (; i < fastLen; ++i) { int x = (fx >> 16); - buf1[i * 2 + 0] = fetch1(s1, x); - buf1[i * 2 + 1] = fetch1(s1, x + 1); - buf2[i * 2 + 0] = fetch1(s2, x); - buf2[i * 2 + 1] = fetch1(s2, x + 1); + if (useFetch) { + buf1[i * 2 + 0] = fetch1(s1, x); + buf1[i * 2 + 1] = fetch1(s1, x + 1); + buf2[i * 2 + 0] = fetch1(s2, x); + buf2[i * 2 + 1] = fetch1(s2, x + 1); + } else { + buf1[i * 2 + 0] = reinterpret_cast<const T *>(s1)[x]; + buf1[i * 2 + 1] = reinterpret_cast<const T *>(s1)[x + 1]; + buf2[i * 2 + 0] = reinterpret_cast<const T *>(s2)[x]; + buf2[i * 2 + 1] = reinterpret_cast<const T *>(s2)[x + 1]; + } fx += fdx; } } @@ -2881,10 +3208,17 @@ static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, int x1 = (fx >> 16); int x2; fetchTransformedBilinear_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, x1, x2); - buf1[i * 2 + 0] = fetch1(s1, x1); - buf1[i * 2 + 1] = fetch1(s1, x2); - buf2[i * 2 + 0] = fetch1(s2, x1); - buf2[i * 2 + 1] = fetch1(s2, x2); + if (useFetch) { + buf1[i * 2 + 0] = fetch1(s1, x1); + buf1[i * 2 + 1] = fetch1(s1, x2); + buf2[i * 2 + 0] = fetch1(s2, x1); + buf2[i * 2 + 1] = fetch1(s2, x2); + } else { + buf1[i * 2 + 0] = reinterpret_cast<const T *>(s1)[x1]; + buf1[i * 2 + 1] = reinterpret_cast<const T *>(s1)[x2]; + buf2[i * 2 + 0] = reinterpret_cast<const T *>(s2)[x1]; + buf2[i * 2 + 1] = reinterpret_cast<const T *>(s2)[x2]; + } fx += fdx; } } else { @@ -2901,10 +3235,17 @@ static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, break; const uchar *s1 = image.scanLine(y1); const uchar *s2 = image.scanLine(y2); - buf1[i * 2 + 0] = fetch1(s1, x1); - buf1[i * 2 + 1] = fetch1(s1, x2); - buf2[i * 2 + 0] = fetch1(s2, x1); - buf2[i * 2 + 1] = fetch1(s2, x2); + if (useFetch) { + buf1[i * 2 + 0] = fetch1(s1, x1); + buf1[i * 2 + 1] = fetch1(s1, x2); + buf2[i * 2 + 0] = fetch1(s2, x1); + buf2[i * 2 + 1] = fetch1(s2, x2); + } else { + buf1[i * 2 + 0] = reinterpret_cast<const T *>(s1)[x1]; + buf1[i * 2 + 1] = reinterpret_cast<const T *>(s1)[x2]; + buf2[i * 2 + 0] = reinterpret_cast<const T *>(s2)[x1]; + buf2[i * 2 + 1] = reinterpret_cast<const T *>(s2)[x2]; + } fx += fdx; fy += fdy; } @@ -2923,10 +3264,17 @@ static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, int y = (fy >> 16); const uchar *s1 = image.scanLine(y); const uchar *s2 = s1 + image.bytesPerLine; - buf1[i * 2 + 0] = fetch1(s1, x); - buf1[i * 2 + 1] = fetch1(s1, x + 1); - buf2[i * 2 + 0] = fetch1(s2, x); - buf2[i * 2 + 1] = fetch1(s2, x + 1); + if (useFetch) { + buf1[i * 2 + 0] = fetch1(s1, x); + buf1[i * 2 + 1] = fetch1(s1, x + 1); + buf2[i * 2 + 0] = fetch1(s2, x); + buf2[i * 2 + 1] = fetch1(s2, x + 1); + } else { + buf1[i * 2 + 0] = reinterpret_cast<const T *>(s1)[x]; + buf1[i * 2 + 1] = reinterpret_cast<const T *>(s1)[x + 1]; + buf2[i * 2 + 0] = reinterpret_cast<const T *>(s2)[x]; + buf2[i * 2 + 1] = reinterpret_cast<const T *>(s2)[x + 1]; + } fx += fdx; fy += fdy; } @@ -2942,10 +3290,17 @@ static void QT_FASTCALL fetchTransformedBilinear_fetcher(uint *buf1, uint *buf2, const uchar *s1 = image.scanLine(y1); const uchar *s2 = image.scanLine(y2); - buf1[i * 2 + 0] = fetch1(s1, x1); - buf1[i * 2 + 1] = fetch1(s1, x2); - buf2[i * 2 + 0] = fetch1(s2, x1); - buf2[i * 2 + 1] = fetch1(s2, x2); + if (useFetch) { + buf1[i * 2 + 0] = fetch1(s1, x1); + buf1[i * 2 + 1] = fetch1(s1, x2); + buf2[i * 2 + 0] = fetch1(s2, x1); + buf2[i * 2 + 1] = fetch1(s2, x2); + } else { + buf1[i * 2 + 0] = reinterpret_cast<const T *>(s1)[x1]; + buf1[i * 2 + 1] = reinterpret_cast<const T *>(s1)[x2]; + buf2[i * 2 + 0] = reinterpret_cast<const T *>(s2)[x1]; + buf2[i * 2 + 1] = reinterpret_cast<const T *>(s2)[x2]; + } fx += fdx; fy += fdy; } @@ -2977,18 +3332,23 @@ static const uint *QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Oper if (fdy == 0) { // simple scale, no rotation or shear if (qAbs(fdx) <= fixed_scale) { // scale up on X - fetchTransformedBilinear_simple_upscale_helper<blendType, bpp>(buffer, buffer + length, data->texture, fx, fy, fdx, fdy); + fetchTransformedBilinear_simple_scale_helper<blendType>(buffer, buffer + length, data->texture, fx, fy, fdx, fdy); + } else if (qAbs(fdx) <= 2 * fixed_scale) { // scale down on X less than 2x + const int mid = (length * 2 < BufferSize) ? length : ((length + 1) / 2); + fetchTransformedBilinear_simple_scale_helper<blendType>(buffer, buffer + mid, data->texture, fx, fy, fdx, fdy); + if (mid != length) + fetchTransformedBilinear_simple_scale_helper<blendType>(buffer + mid, buffer + length, data->texture, fx, fy, fdx, fdy); } else { - const BilinearFastTransformFetcher fetcher = fetchTransformedBilinear_fetcher<blendType,bpp>; + const auto fetcher = fetchTransformedBilinear_fetcher<blendType,bpp,uint>; - uint buf1[buffer_size]; - uint buf2[buffer_size]; + uint buf1[BufferSize]; + uint buf2[BufferSize]; uint *b = buffer; while (length) { - int len = qMin(length, buffer_size / 2); + int len = qMin(length, BufferSize / 2); fetcher(buf1, buf2, len, data->texture, fx, fy, fdx, 0); - layout->convertToARGB32PM(buf1, buf1, len * 2, clut, 0); - layout->convertToARGB32PM(buf2, buf2, len * 2, clut, 0); + layout->convertToARGB32PM(buf1, len * 2, clut); + layout->convertToARGB32PM(buf2, len * 2, clut); if (hasFastInterpolate4() || qAbs(data->m22) < qreal(1./8.)) { // scale up more than 8x (on Y) int disty = (fy & 0x0000ffff) >> 8; @@ -3014,16 +3374,16 @@ static const uint *QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Oper } } } else { // rotation or shear - const BilinearFastTransformFetcher fetcher = fetchTransformedBilinear_fetcher<blendType,bpp>; + const auto fetcher = fetchTransformedBilinear_fetcher<blendType,bpp,uint>; - uint buf1[buffer_size]; - uint buf2[buffer_size]; + uint buf1[BufferSize]; + uint buf2[BufferSize]; uint *b = buffer; while (length) { - int len = qMin(length, buffer_size / 2); + int len = qMin(length, BufferSize / 2); fetcher(buf1, buf2, len, data->texture, fx, fy, fdx, fdy); - layout->convertToARGB32PM(buf1, buf1, len * 2, clut, 0); - layout->convertToARGB32PM(buf2, buf2, len * 2, clut, 0); + layout->convertToARGB32PM(buf1, len * 2, clut); + layout->convertToARGB32PM(buf2, len * 2, clut); if (hasFastInterpolate4() || qAbs(data->m11) < qreal(1./8.) || qAbs(data->m22) < qreal(1./8.)) { // If we are zooming more than 8 times, we use 8bit precision for the position. @@ -3070,15 +3430,15 @@ static const uint *QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Oper qreal fy = data->m22 * cy + data->m12 * cx + data->dy; qreal fw = data->m23 * cy + data->m13 * cx + data->m33; - uint buf1[buffer_size]; - uint buf2[buffer_size]; + uint buf1[BufferSize]; + uint buf2[BufferSize]; uint *b = buffer; - int distxs[buffer_size / 2]; - int distys[buffer_size / 2]; + int distxs[BufferSize / 2]; + int distys[BufferSize / 2]; while (length) { - int len = qMin(length, buffer_size / 2); + int len = qMin(length, BufferSize / 2); for (int i = 0; i < len; ++i) { const qreal iw = fw == 0 ? 1 : 1 / fw; const qreal px = fx * iw - qreal(0.5); @@ -3110,8 +3470,8 @@ static const uint *QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Oper fw += fdw; } - layout->convertToARGB32PM(buf1, buf1, len * 2, clut, 0); - layout->convertToARGB32PM(buf2, buf2, len * 2, clut, 0); + layout->convertToARGB32PM(buf1, len * 2, clut); + layout->convertToARGB32PM(buf2, len * 2, clut); for (int i = 0; i < len; ++i) { int distx = distxs[i]; @@ -3128,19 +3488,27 @@ static const uint *QT_FASTCALL fetchTransformedBilinear(uint *buffer, const Oper } template<TextureBlendType blendType> -static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, const Operator *, - const QSpanData *data, int y, int x, int length) +static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64_uint32(QRgba64 *buffer, const QSpanData *data, + int y, int x, int length) { - const QPixelLayout *layout = &qPixelLayouts[data->texture.format]; + const QTextureData &texture = data->texture; + const QPixelLayout *layout = &qPixelLayouts[texture.format]; const QVector<QRgb> *clut = data->texture.colorTable; const qreal cx = x + qreal(0.5); const qreal cy = y + qreal(0.5); + uint sbuf1[BufferSize]; + uint sbuf2[BufferSize]; + QRgba64 buf1[BufferSize]; + QRgba64 buf2[BufferSize]; + QRgba64 *end = buffer + length; + QRgba64 *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); + const int fdx = (int)(data->m11 * fixed_scale); + const 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); @@ -3148,20 +3516,14 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co fx -= half_point; fy -= half_point; - const BilinearFastTransformFetcher fetcher = + const auto fetcher = (layout->bpp == QPixelLayout::BPP32) - ? fetchTransformedBilinear_fetcher<blendType, QPixelLayout::BPP32> - : fetchTransformedBilinear_fetcher<blendType, QPixelLayout::BPPNone>; + ? fetchTransformedBilinear_fetcher<blendType, QPixelLayout::BPP32, uint> + : fetchTransformedBilinear_fetcher<blendType, QPixelLayout::BPPNone, uint>; if (fdy == 0) { //simple scale, no rotation - - uint sbuf1[buffer_size]; - uint sbuf2[buffer_size]; - quint64 buf1[buffer_size]; - quint64 buf2[buffer_size]; - QRgba64 *b = buffer; while (length) { - int len = qMin(length, buffer_size / 2); + int len = qMin(length, BufferSize / 2); int disty = (fy & 0x0000ffff); #if defined(__SSE2__) const __m128i vdy = _mm_set1_epi16(disty); @@ -3169,9 +3531,9 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co #endif fetcher(sbuf1, sbuf2, len, data->texture, fx, fy, fdx, fdy); - layout->convertToARGB64PM((QRgba64 *)buf1, sbuf1, len * 2, clut, 0); + layout->convertToRGBA64PM(buf1, sbuf1, len * 2, clut, 0); if (disty) - layout->convertToARGB64PM((QRgba64 *)buf2, sbuf2, len * 2, clut, 0); + layout->convertToRGBA64PM(buf2, sbuf2, len * 2, clut, 0); for (int i = 0; i < len; ++i) { int distx = (fx & 0x0000ffff); @@ -3191,33 +3553,26 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co } _mm_storel_epi64((__m128i*)(b+i), vt); #else - b[i] = interpolate_4_pixels_rgb64((QRgba64 *)buf1 + i*2, (QRgba64 *)buf2 + i*2, distx, disty); + b[i] = interpolate_4_pixels_rgb64(buf1 + i*2, buf2 + i*2, distx, disty); #endif fx += fdx; } length -= len; b += len; } - } else { //rotation - uint sbuf1[buffer_size]; - uint sbuf2[buffer_size]; - quint64 buf1[buffer_size]; - quint64 buf2[buffer_size]; - QRgba64 *end = buffer + length; - QRgba64 *b = buffer; - + } else { // rotation or shear while (b < end) { - int len = qMin(length, buffer_size / 2); + int len = qMin(length, BufferSize / 2); fetcher(sbuf1, sbuf2, len, data->texture, fx, fy, fdx, fdy); - layout->convertToARGB64PM((QRgba64 *)buf1, sbuf1, len * 2, clut, 0); - layout->convertToARGB64PM((QRgba64 *)buf2, sbuf2, len * 2, clut, 0); + layout->convertToRGBA64PM(buf1, sbuf1, len * 2, clut, 0); + layout->convertToRGBA64PM(buf2, sbuf2, len * 2, clut, 0); for (int i = 0; i < len; ++i) { int distx = (fx & 0x0000ffff); int disty = (fy & 0x0000ffff); - b[i] = interpolate_4_pixels_rgb64((QRgba64 *)buf1 + i*2, (QRgba64 *)buf2 + i*2, distx, disty); + b[i] = interpolate_4_pixels_rgb64(buf1 + i*2, buf2 + i*2, distx, disty); fx += fdx; fy += fdy; } @@ -3226,7 +3581,7 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co b += len; } } - } else { + } else { // !(data->fast_matrix) const QTextureData &image = data->texture; const qreal fdx = data->m11; @@ -3238,17 +3593,12 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co qreal fw = data->m23 * cy + data->m13 * cx + data->m33; FetchPixelFunc fetch = qFetchPixel[layout->bpp]; - uint sbuf1[buffer_size]; - uint sbuf2[buffer_size]; - quint64 buf1[buffer_size]; - quint64 buf2[buffer_size]; - QRgba64 *b = buffer; - int distxs[buffer_size / 2]; - int distys[buffer_size / 2]; + int distxs[BufferSize / 2]; + int distys[BufferSize / 2]; - while (length) { - int len = qMin(length, buffer_size / 2); + while (b < end) { + int len = qMin(length, BufferSize / 2); for (int i = 0; i < len; ++i) { const qreal iw = fw == 0 ? 1 : 1 / fw; const qreal px = fx * iw - qreal(0.5); @@ -3265,21 +3615,165 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co fetchTransformedBilinear_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, x1, x2); fetchTransformedBilinear_pixelBounds<blendType>(image.height, image.y1, image.y2 - 1, y1, y2); - const uchar *s1 = data->texture.scanLine(y1); - const uchar *s2 = data->texture.scanLine(y2); + const uchar *s1 = texture.scanLine(y1); + const uchar *s2 = texture.scanLine(y2); - if (layout->bpp == QPixelLayout::BPP32) { - sbuf1[i * 2 + 0] = ((const uint*)s1)[x1]; - sbuf1[i * 2 + 1] = ((const uint*)s1)[x2]; - sbuf2[i * 2 + 0] = ((const uint*)s2)[x1]; - sbuf2[i * 2 + 1] = ((const uint*)s2)[x2]; + sbuf1[i * 2 + 0] = fetch(s1, x1); + sbuf1[i * 2 + 1] = fetch(s1, x2); + sbuf2[i * 2 + 0] = fetch(s2, x1); + sbuf2[i * 2 + 1] = fetch(s2, x2); - } else { - sbuf1[i * 2 + 0] = fetch(s1, x1); - sbuf1[i * 2 + 1] = fetch(s1, x2); - sbuf2[i * 2 + 0] = fetch(s2, x1); - sbuf2[i * 2 + 1] = fetch(s2, x2); + fx += fdx; + fy += fdy; + fw += fdw; + //force increment to avoid /0 + if (!fw) + fw += fdw; + } + + layout->convertToRGBA64PM(buf1, sbuf1, len * 2, clut, 0); + layout->convertToRGBA64PM(buf2, sbuf2, len * 2, clut, 0); + + for (int i = 0; i < len; ++i) { + int distx = distxs[i]; + int disty = distys[i]; + b[i] = interpolate_4_pixels_rgb64(buf1 + i*2, buf2 + i*2, distx, disty); + } + + length -= len; + b += len; + } + } + return buffer; +} + +template<TextureBlendType blendType> +static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64_uint64(QRgba64 *buffer, const QSpanData *data, + int y, int x, int length) +{ + const QTextureData &texture = data->texture; + Q_ASSERT(qPixelLayouts[texture.format].bpp == QPixelLayout::BPP64); + const auto convert = (data->texture.format == QImage::Format_RGBA64) ? convertRGBA64ToRGBA64PM : convertRGBA64PMToRGBA64PM; + + const qreal cx = x + qreal(0.5); + const qreal cy = y + qreal(0.5); + + QRgba64 buf1[BufferSize]; + QRgba64 buf2[BufferSize]; + QRgba64 *end = buffer + length; + QRgba64 *b = buffer; + + if (data->fast_matrix) { + // The increment pr x in the scanline + const int fdx = (int)(data->m11 * fixed_scale); + const 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; + const auto fetcher = fetchTransformedBilinear_fetcher<blendType, QPixelLayout::BPP64, QRgba64>; + + if (fdy == 0) { //simple scale, no rotation + while (length) { + int len = qMin(length, BufferSize / 2); + int disty = (fy & 0x0000ffff); +#if defined(__SSE2__) + const __m128i vdy = _mm_set1_epi16(disty); + const __m128i vidy = _mm_set1_epi16(0x10000 - disty); +#endif + fetcher(buf1, buf2, len, data->texture, fx, fy, fdx, fdy); + + convert(buf1, len * 2); + if (disty) + convert(buf2, len * 2); + + for (int i = 0; i < len; ++i) { + int distx = (fx & 0x0000ffff); +#if defined(__SSE2__) + __m128i vt = _mm_loadu_si128((const __m128i*)(buf1 + i*2)); + if (disty) { + __m128i vb = _mm_loadu_si128((const __m128i*)(buf2 + i*2)); + vt = _mm_mulhi_epu16(vt, vidy); + vb = _mm_mulhi_epu16(vb, vdy); + vt = _mm_add_epi16(vt, vb); + } + if (distx) { + const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0)); + const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(0x10000 - distx), _MM_SHUFFLE(0, 0, 0, 0)); + vt = _mm_mulhi_epu16(vt, _mm_unpacklo_epi64(vidistx, vdistx)); + vt = _mm_add_epi16(vt, _mm_srli_si128(vt, 8)); + } + _mm_storel_epi64((__m128i*)(b+i), vt); +#else + b[i] = interpolate_4_pixels_rgb64(buf1 + i*2, buf2 + i*2, distx, disty); +#endif + fx += fdx; } + length -= len; + b += len; + } + } else { // rotation or shear + while (b < end) { + int len = qMin(length, BufferSize / 2); + + fetcher(buf1, buf2, len, data->texture, fx, fy, fdx, fdy); + + convert(buf1, len * 2); + convert(buf2, len * 2); + + for (int i = 0; i < len; ++i) { + int distx = (fx & 0x0000ffff); + int disty = (fy & 0x0000ffff); + b[i] = interpolate_4_pixels_rgb64(buf1 + i*2, buf2 + i*2, distx, disty); + fx += fdx; + fy += fdy; + } + + length -= len; + b += len; + } + } + } else { // !(data->fast_matrix) + const QTextureData &image = data->texture; + + 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; + + int distxs[BufferSize / 2]; + int distys[BufferSize / 2]; + + while (b < end) { + int len = qMin(length, BufferSize / 2); + for (int i = 0; i < len; ++i) { + const qreal iw = fw == 0 ? 1 : 1 / fw; + const qreal px = fx * iw - qreal(0.5); + const qreal py = fy * iw - qreal(0.5); + + int x1 = int(px) - (px < 0); + int x2; + int y1 = int(py) - (py < 0); + int y2; + + distxs[i] = int((px - x1) * (1<<16)); + distys[i] = int((py - y1) * (1<<16)); + + fetchTransformedBilinear_pixelBounds<blendType>(image.width, image.x1, image.x2 - 1, x1, x2); + fetchTransformedBilinear_pixelBounds<blendType>(image.height, image.y1, image.y2 - 1, y1, y2); + + const uchar *s1 = texture.scanLine(y1); + const uchar *s2 = texture.scanLine(y2); + + buf1[i * 2 + 0] = reinterpret_cast<const QRgba64 *>(s1)[x1]; + buf1[i * 2 + 1] = reinterpret_cast<const QRgba64 *>(s1)[x2]; + buf2[i * 2 + 0] = reinterpret_cast<const QRgba64 *>(s2)[x1]; + buf2[i * 2 + 1] = reinterpret_cast<const QRgba64 *>(s2)[x2]; fx += fdx; fy += fdy; @@ -3289,23 +3783,31 @@ static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, co fw += fdw; } - layout->convertToARGB64PM((QRgba64 *)buf1, sbuf1, len * 2, clut, 0); - layout->convertToARGB64PM((QRgba64 *)buf2, sbuf2, len * 2, clut, 0); + convert(buf1, len * 2); + convert(buf2, len * 2); for (int i = 0; i < len; ++i) { int distx = distxs[i]; int disty = distys[i]; - b[i] = interpolate_4_pixels_rgb64((QRgba64 *)buf1 + i*2, (QRgba64 *)buf2 + i*2, distx, disty); + b[i] = interpolate_4_pixels_rgb64(buf1 + i*2, buf2 + i*2, distx, disty); } length -= len; b += len; } } - return buffer; } +template<TextureBlendType blendType> +static const QRgba64 *QT_FASTCALL fetchTransformedBilinear64(QRgba64 *buffer, const Operator *, + const QSpanData *data, int y, int x, int length) +{ + if (qPixelLayouts[data->texture.format].bpp == QPixelLayout::BPP64) + return fetchTransformedBilinear64_uint64<blendType>(buffer, data, y, x, length); + return fetchTransformedBilinear64_uint32<blendType>(buffer, data, y, x, length); +} + // FetchUntransformed can have more specialized methods added depending on SIMD features. static SourceFetchProc sourceFetchUntransformed[QImage::NImageFormats] = { 0, // Invalid @@ -3333,6 +3835,9 @@ static SourceFetchProc sourceFetchUntransformed[QImage::NImageFormats] = { fetchUntransformed, // Format_A2RGB30_Premultiplied fetchUntransformed, // Alpha8 fetchUntransformed, // Grayscale8 + fetchUntransformed, // RGBX64 + fetchUntransformed, // RGBA64 + fetchUntransformed, // RGBA64_Premultiplied }; static const SourceFetchProc sourceFetchGeneric[NBlendTypes] = { @@ -3353,6 +3858,15 @@ static SourceFetchProc sourceFetchARGB32PM[NBlendTypes] = { fetchTransformedBilinearARGB32PM<BlendTransformedBilinearTiled> // BilinearTiled }; +static SourceFetchProc sourceFetchAny16[NBlendTypes] = { + fetchUntransformed, // Untransformed + fetchUntransformed, // Tiled + fetchTransformed<BlendTransformed, QPixelLayout::BPP16>, // Transformed + fetchTransformed<BlendTransformedTiled, QPixelLayout::BPP16>, // TransformedTiled + fetchTransformedBilinear<BlendTransformedBilinear, QPixelLayout::BPP16>, // TransformedBilinear + fetchTransformedBilinear<BlendTransformedBilinearTiled, QPixelLayout::BPP16> // TransformedBilinearTiled +}; + static SourceFetchProc sourceFetchAny32[NBlendTypes] = { fetchUntransformed, // Untransformed fetchUntransformed, // Tiled @@ -3371,17 +3885,35 @@ static const SourceFetchProc64 sourceFetchGeneric64[NBlendTypes] = { fetchTransformedBilinear64<BlendTransformedBilinearTiled> // BilinearTiled }; +static const SourceFetchProc64 sourceFetchRGBA64PM[NBlendTypes] = { + fetchUntransformedRGBA64PM, // Untransformed + fetchUntransformedRGBA64PM, // Tiled + fetchTransformed64<BlendTransformed>, // Transformed + fetchTransformed64<BlendTransformedTiled>, // TransformedTiled + fetchTransformedBilinear64<BlendTransformedBilinear>, // Bilinear + fetchTransformedBilinear64<BlendTransformedBilinearTiled> // BilinearTiled +}; + static inline SourceFetchProc getSourceFetch(TextureBlendType blendType, QImage::Format format) { if (format == QImage::Format_RGB32 || format == QImage::Format_ARGB32_Premultiplied) return sourceFetchARGB32PM[blendType]; if (blendType == BlendUntransformed || blendType == BlendTiled) return sourceFetchUntransformed[format]; + if (qPixelLayouts[format].bpp == QPixelLayout::BPP16) + return sourceFetchAny16[blendType]; if (qPixelLayouts[format].bpp == QPixelLayout::BPP32) return sourceFetchAny32[blendType]; return sourceFetchGeneric[blendType]; } +static inline SourceFetchProc64 getSourceFetch64(TextureBlendType blendType, QImage::Format format) +{ + if (format == QImage::Format_RGBX64 || format == QImage::Format_RGBA64_Premultiplied) + return sourceFetchRGBA64PM[blendType]; + return sourceFetchGeneric64[blendType]; +} + #define FIXPT_BITS 8 #define FIXPT_SIZE (1<<FIXPT_BITS) @@ -3724,7 +4256,7 @@ static inline Operator getOperator(const QSpanData *data, const QSpan *spans, in case QSpanData::Texture: solidSource = !data->texture.hasAlpha; op.srcFetch = getSourceFetch(getBlendType(data), data->texture.format); - op.srcFetch64 = sourceFetchGeneric64[getBlendType(data)]; + op.srcFetch64 = getSourceFetch64(getBlendType(data), data->texture.format);; break; default: Q_UNREACHABLE(); @@ -3752,8 +4284,9 @@ static inline Operator getOperator(const QSpanData *data, const QSpan *spans, in // If all spans are opaque we do not need to fetch dest. // But don't clear passthrough destFetch as they are just as fast and save destStore. if (op.destFetch != destFetchARGB32P) - op.destFetch = 0; - op.destFetch64 = destFetch64Undefined; + op.destFetch = destFetchUndefined; + if (op.destFetch64 != destFetchRGB64) + op.destFetch64 = destFetch64Undefined; } } @@ -3778,7 +4311,7 @@ static void blend_color_generic(int count, const QSpan *spans, void *userData) { QSpanData *data = reinterpret_cast<QSpanData *>(userData); - uint buffer[buffer_size]; + uint buffer[BufferSize]; Operator op = getOperator(data, spans, count); const uint color = data->solid.color.toArgb32(); @@ -3786,8 +4319,8 @@ void blend_color_generic(int count, const QSpan *spans, void *userData) int x = spans->x; int length = spans->len; while (length) { - int l = qMin(buffer_size, length); - uint *dest = op.destFetch ? op.destFetch(buffer, data->rasterBuffer, x, spans->y, l) : buffer; + int l = qMin(BufferSize, length); + uint *dest = op.destFetch(buffer, data->rasterBuffer, x, spans->y, l); op.funcSolid(dest, l, color, spans->coverage); if (op.destStore) op.destStore(data->rasterBuffer, x, spans->y, dest, l); @@ -3802,7 +4335,7 @@ static void blend_color_argb(int count, const QSpan *spans, void *userData) { QSpanData *data = reinterpret_cast<QSpanData *>(userData); - Operator op = getOperator(data, spans, count); + const Operator op = getOperator(data, spans, count); const uint color = data->solid.color.toArgb32(); if (op.mode == QPainter::CompositionMode_Source) { @@ -3838,17 +4371,31 @@ void blend_color_generic_rgb64(int count, const QSpan *spans, void *userData) return blend_color_generic(count, spans, userData); } - quint64 buffer[buffer_size]; + quint64 buffer[BufferSize]; const QRgba64 color = data->solid.color; + bool solidFill = data->rasterBuffer->compositionMode == QPainter::CompositionMode_Source + || (data->rasterBuffer->compositionMode == QPainter::CompositionMode_SourceOver && color.isOpaque()); + bool isBpp32 = qPixelLayouts[data->rasterBuffer->format].bpp == QPixelLayout::BPP32; while (count--) { int x = spans->x; int length = spans->len; + if (solidFill && isBpp32 && spans->coverage == 255) { + // If dest doesn't matter we don't need to bother with blending or converting all the identical pixels + if (length > 0) { + op.destStore64(data->rasterBuffer, x, spans->y, &color, 1); + uint *dest = (uint*)data->rasterBuffer->scanLine(spans->y) + x; + qt_memfill32(dest + 1, dest[0], length - 1); + length = 0; + } + } + while (length) { - int l = qMin(buffer_size, length); + int l = qMin(BufferSize, length); QRgba64 *dest = op.destFetch64((QRgba64 *)buffer, data->rasterBuffer, x, spans->y, l); op.funcSolid64(dest, l, color, spans->coverage); - op.destStore64(data->rasterBuffer, x, spans->y, dest, l); + if (op.destStore64) + op.destStore64(data->rasterBuffer, x, spans->y, dest, l); length -= l; x += l; } @@ -3950,7 +4497,7 @@ void handleSpans(int count, const QSpan *spans, const QSpanData *data, T &handle int length = right - x; while (length) { - int l = qMin(buffer_size, length); + int l = qMin(BufferSize, length); length -= l; int process_length = l; @@ -3997,8 +4544,8 @@ struct QBlendBase BlendType *dest; - BlendType buffer[buffer_size]; - BlendType src_buffer[buffer_size]; + BlendType buffer[BufferSize]; + BlendType src_buffer[BufferSize]; }; class BlendSrcGeneric : public QBlendBase<uint> @@ -4011,7 +4558,7 @@ public: const uint *fetch(int x, int y, int len) { - dest = op.destFetch ? op.destFetch(buffer, data->rasterBuffer, x, y, len) : buffer; + dest = op.destFetch(buffer, data->rasterBuffer, x, y, len); return op.srcFetch(src_buffer, &op, data, y, x, len); } @@ -4037,7 +4584,7 @@ public: bool isSupported() const { - return op.func64 && op.destFetch64 && op.destStore64; + return op.func64 && op.destFetch64; } const quint64 *fetch(int x, int y, int len) @@ -4053,7 +4600,8 @@ public: void store(int x, int y, int len) { - op.destStore64(data->rasterBuffer, x, y, (QRgba64 *)dest, len); + if (op.destStore64) + op.destStore64(data->rasterBuffer, x, y, (QRgba64 *)dest, len); } }; @@ -4082,8 +4630,8 @@ static void blend_untransformed_generic(int count, const QSpan *spans, void *use { QSpanData *data = reinterpret_cast<QSpanData *>(userData); - uint buffer[buffer_size]; - uint src_buffer[buffer_size]; + uint buffer[BufferSize]; + uint src_buffer[BufferSize]; Operator op = getOperator(data, spans, count); const int image_width = data->texture.width; @@ -4107,9 +4655,9 @@ static void blend_untransformed_generic(int count, const QSpan *spans, void *use if (length > 0) { const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; while (length) { - int l = qMin(buffer_size, length); + int l = qMin(BufferSize, length); const uint *src = op.srcFetch(src_buffer, &op, data, sy, sx, l); - uint *dest = op.destFetch ? op.destFetch(buffer, data->rasterBuffer, x, spans->y, l) : buffer; + uint *dest = op.destFetch(buffer, data->rasterBuffer, x, spans->y, l); op.func(dest, src, l, coverage); if (op.destStore) op.destStore(data->rasterBuffer, x, spans->y, dest, l); @@ -4132,8 +4680,8 @@ static void blend_untransformed_generic_rgb64(int count, const QSpan *spans, voi qCDebug(lcQtGuiDrawHelper, "blend_untransformed_generic_rgb64: unsupported 64-bit blend attempted, falling back to 32-bit"); return blend_untransformed_generic(count, spans, userData); } - quint64 buffer[buffer_size]; - quint64 src_buffer[buffer_size]; + quint64 buffer[BufferSize]; + quint64 src_buffer[BufferSize]; const int image_width = data->texture.width; const int image_height = data->texture.height; @@ -4156,11 +4704,12 @@ static void blend_untransformed_generic_rgb64(int count, const QSpan *spans, voi if (length > 0) { const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; while (length) { - int l = qMin(buffer_size, length); + int l = qMin(BufferSize, length); const QRgba64 *src = op.srcFetch64((QRgba64 *)src_buffer, &op, data, sy, sx, l); QRgba64 *dest = op.destFetch64((QRgba64 *)buffer, data->rasterBuffer, x, spans->y, l); op.func64(dest, src, l, coverage); - op.destStore64(data->rasterBuffer, x, spans->y, dest, l); + if (op.destStore64) + op.destStore64(data->rasterBuffer, x, spans->y, dest, l); x += l; sx += l; length -= l; @@ -4320,8 +4869,8 @@ static void blend_tiled_generic(int count, const QSpan *spans, void *userData) { QSpanData *data = reinterpret_cast<QSpanData *>(userData); - uint buffer[buffer_size]; - uint src_buffer[buffer_size]; + uint buffer[BufferSize]; + uint src_buffer[BufferSize]; Operator op = getOperator(data, spans, count); const int image_width = data->texture.width; @@ -4347,10 +4896,10 @@ static void blend_tiled_generic(int count, const QSpan *spans, void *userData) const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; while (length) { int l = qMin(image_width - sx, length); - if (buffer_size < l) - l = buffer_size; + if (BufferSize < l) + l = BufferSize; const uint *src = op.srcFetch(src_buffer, &op, data, sy, sx, l); - uint *dest = op.destFetch ? op.destFetch(buffer, data->rasterBuffer, x, spans->y, l) : buffer; + uint *dest = op.destFetch(buffer, data->rasterBuffer, x, spans->y, l); op.func(dest, src, l, coverage); if (op.destStore) op.destStore(data->rasterBuffer, x, spans->y, dest, l); @@ -4373,8 +4922,8 @@ static void blend_tiled_generic_rgb64(int count, const QSpan *spans, void *userD qCDebug(lcQtGuiDrawHelper, "blend_tiled_generic_rgb64: unsupported 64-bit blend attempted, falling back to 32-bit"); return blend_tiled_generic(count, spans, userData); } - quint64 buffer[buffer_size]; - quint64 src_buffer[buffer_size]; + quint64 buffer[BufferSize]; + quint64 src_buffer[BufferSize]; const int image_width = data->texture.width; const int image_height = data->texture.height; @@ -4386,6 +4935,50 @@ static void blend_tiled_generic_rgb64(int count, const QSpan *spans, void *userD if (yoff < 0) yoff += image_height; + bool isBpp32 = qPixelLayouts[data->rasterBuffer->format].bpp == QPixelLayout::BPP32; + if (op.destFetch64 == destFetch64Undefined && image_width <= BufferSize && isBpp32) { + // If destination isn't blended into the result, we can do the tiling directly on destination pixels. + while (count--) { + int x = spans->x; + int y = spans->y; + int length = spans->len; + int sx = (xoff + spans->x) % image_width; + int sy = (spans->y + yoff) % image_height; + if (sx < 0) + sx += image_width; + if (sy < 0) + sy += image_height; + + int sl = qMin(image_width, length); + if (sx > 0 && sl > 0) { + int l = qMin(image_width - sx, sl); + const QRgba64 *src = op.srcFetch64((QRgba64 *)src_buffer, &op, data, sy, sx, l); + op.destStore64(data->rasterBuffer, x, y, src, l); + x += l; + sx += l; + sl -= l; + if (sx >= image_width) + sx = 0; + } + if (sl > 0) { + Q_ASSERT(sx == 0); + const QRgba64 *src = op.srcFetch64((QRgba64 *)src_buffer, &op, data, sy, sx, sl); + op.destStore64(data->rasterBuffer, x, y, src, sl); + x += sl; + sx += sl; + sl -= sl; + if (sx >= image_width) + sx = 0; + } + uint *dest = (uint*)data->rasterBuffer->scanLine(y) + x - image_width; + for (int i = image_width; i < length; ++i) { + dest[i] = dest[i - image_width]; + } + ++spans; + } + return; + } + while (count--) { int x = spans->x; int length = spans->len; @@ -4399,12 +4992,13 @@ static void blend_tiled_generic_rgb64(int count, const QSpan *spans, void *userD const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; while (length) { int l = qMin(image_width - sx, length); - if (buffer_size < l) - l = buffer_size; + if (BufferSize < l) + l = BufferSize; const QRgba64 *src = op.srcFetch64((QRgba64 *)src_buffer, &op, data, sy, sx, l); QRgba64 *dest = op.destFetch64((QRgba64 *)buffer, data->rasterBuffer, x, spans->y, l); op.func64(dest, src, l, coverage); - op.destStore64(data->rasterBuffer, x, spans->y, dest, l); + if (op.destStore64) + op.destStore64(data->rasterBuffer, x, spans->y, dest, l); x += l; sx += l; length -= l; @@ -4449,8 +5043,8 @@ static void blend_tiled_argb(int count, const QSpan *spans, void *userData) const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; while (length) { int l = qMin(image_width - sx, length); - if (buffer_size < l) - l = buffer_size; + if (BufferSize < l) + l = BufferSize; const uint *src = (const uint *)data->texture.scanLine(sy) + sx; uint *dest = ((uint *)data->rasterBuffer->scanLine(spans->y)) + x; op.func(dest, src, l, coverage); @@ -4509,8 +5103,8 @@ static void blend_tiled_rgb565(int count, const QSpan *spans, void *userData) int tx = x; while (length) { int l = qMin(image_width - sx, length); - if (buffer_size < l) - l = buffer_size; + if (BufferSize < l) + l = BufferSize; quint16 *dest = ((quint16 *)data->rasterBuffer->scanLine(spans->y)) + tx; const quint16 *src = (const quint16 *)data->texture.scanLine(sy) + sx; memcpy(dest, src, l * sizeof(quint16)); @@ -4545,8 +5139,8 @@ static void blend_tiled_rgb565(int count, const QSpan *spans, void *userData) if (alpha > 0) { while (length) { int l = qMin(image_width - sx, length); - if (buffer_size < l) - l = buffer_size; + if (BufferSize < l) + l = BufferSize; quint16 *dest = ((quint16 *)data->rasterBuffer->scanLine(spans->y)) + x; const quint16 *src = (const quint16 *)data->texture.scanLine(sy) + sx; blend_sourceOver_rgb16_rgb16(dest, src, l, alpha, ialpha); @@ -4562,723 +5156,12 @@ static void blend_tiled_rgb565(int count, const QSpan *spans, void *userData) } } -static void blend_transformed_bilinear_rgb565(int count, const QSpan *spans, void *userData) -{ - QSpanData *data = reinterpret_cast<QSpanData*>(userData); - QPainter::CompositionMode mode = data->rasterBuffer->compositionMode; - - if (data->texture.format != QImage::Format_RGB16 - || (mode != QPainter::CompositionMode_SourceOver - && mode != QPainter::CompositionMode_Source)) - { - blend_src_generic(count, spans, userData); - return; - } - - quint16 buffer[buffer_size]; - - const int src_minx = data->texture.x1; - const int src_miny = data->texture.y1; - const int src_maxx = data->texture.x2 - 1; - const int src_maxy = data->texture.y2 - 1; - - if (data->fast_matrix) { - // The increment pr x in the scanline - const int fdx = (int)(data->m11 * fixed_scale); - const int fdy = (int)(data->m12 * fixed_scale); - - while (count--) { - const quint8 coverage = (data->texture.const_alpha * spans->coverage) >> 8; - const quint8 alpha = (coverage + 1) >> 3; - const quint8 ialpha = 0x20 - alpha; - if (alpha == 0) { - ++spans; - continue; - } - - quint16 *dest = (quint16 *)data->rasterBuffer->scanLine(spans->y) + spans->x; - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - int x = int((data->m21 * cy - + data->m11 * cx + data->dx) * fixed_scale) - half_point; - int y = int((data->m22 * cy - + data->m12 * cx + data->dy) * fixed_scale) - half_point; - int length = spans->len; - - while (length) { - int l; - quint16 *b; - if (ialpha == 0) { - l = length; - b = dest; - } else { - l = qMin(length, buffer_size); - b = buffer; - } - const quint16 *end = b + l; - - while (b < end) { - int x1 = (x >> 16); - int x2; - int y1 = (y >> 16); - int y2; - - fetchTransformedBilinear_pixelBounds<BlendTransformedBilinear>(0, src_minx, src_maxx, x1, x2); - fetchTransformedBilinear_pixelBounds<BlendTransformedBilinear>(0, src_miny, src_maxy, y1, y2); - - const quint16 *src1 = (const quint16*)data->texture.scanLine(y1); - const quint16 *src2 = (const quint16*)data->texture.scanLine(y2); - quint16 tl = src1[x1]; - const quint16 tr = src1[x2]; - quint16 bl = src2[x1]; - const quint16 br = src2[x2]; - - const uint distxsl8 = x & 0xff00; - const uint distysl8 = y & 0xff00; - const uint distx = distxsl8 >> 8; - const uint disty = distysl8 >> 8; - const uint distxy = distx * disty; - - const uint tlw = 0x10000 - distxsl8 - distysl8 + distxy; // (256 - distx) * (256 - disty) - const uint trw = distxsl8 - distxy; // distx * (256 - disty) - const uint blw = distysl8 - distxy; // (256 - distx) * disty - const uint brw = distxy; // distx * disty - uint red = ((tl & 0xf800) * tlw + (tr & 0xf800) * trw - + (bl & 0xf800) * blw + (br & 0xf800) * brw) & 0xf8000000; - uint green = ((tl & 0x07e0) * tlw + (tr & 0x07e0) * trw - + (bl & 0x07e0) * blw + (br & 0x07e0) * brw) & 0x07e00000; - uint blue = ((tl & 0x001f) * tlw + (tr & 0x001f) * trw - + (bl & 0x001f) * blw + (br & 0x001f) * brw); - *b = quint16((red | green | blue) >> 16); - - ++b; - x += fdx; - y += fdy; - } - - if (ialpha != 0) - blend_sourceOver_rgb16_rgb16(dest, buffer, l, alpha, ialpha); - - dest += l; - length -= l; - } - ++spans; - } - } else { - const qreal fdx = data->m11; - const qreal fdy = data->m12; - const qreal fdw = data->m13; - - while (count--) { - const quint8 coverage = (data->texture.const_alpha * spans->coverage) >> 8; - const quint8 alpha = (coverage + 1) >> 3; - const quint8 ialpha = 0x20 - alpha; - if (alpha == 0) { - ++spans; - continue; - } - - quint16 *dest = (quint16 *)data->rasterBuffer->scanLine(spans->y) + spans->x; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - qreal x = data->m21 * cy + data->m11 * cx + data->dx; - qreal y = data->m22 * cy + data->m12 * cx + data->dy; - qreal w = data->m23 * cy + data->m13 * cx + data->m33; - - int length = spans->len; - while (length) { - int l; - quint16 *b; - if (ialpha == 0) { - l = length; - b = dest; - } else { - l = qMin(length, buffer_size); - b = buffer; - } - const quint16 *end = b + l; - - while (b < end) { - const qreal iw = w == 0 ? 1 : 1 / w; - const qreal px = x * iw - qreal(0.5); - const qreal py = y * iw - qreal(0.5); - - int x1 = int(px) - (px < 0); - int x2; - int y1 = int(py) - (py < 0); - int y2; - - fetchTransformedBilinear_pixelBounds<BlendTransformedBilinear>(0, src_minx, src_maxx, x1, x2); - fetchTransformedBilinear_pixelBounds<BlendTransformedBilinear>(0, src_miny, src_maxy, y1, y2); - - const quint16 *src1 = (const quint16 *)data->texture.scanLine(y1); - const quint16 *src2 = (const quint16 *)data->texture.scanLine(y2); - quint16 tl = src1[x1]; - const quint16 tr = src1[x2]; - quint16 bl = src2[x1]; - const quint16 br = src2[x2]; - - const uint distx = uint((px - x1) * 256); - const uint disty = uint((py - y1) * 256); - const uint distxsl8 = distx << 8; - const uint distysl8 = disty << 8; - const uint distxy = distx * disty; - - const uint tlw = 0x10000 - distxsl8 - distysl8 + distxy; // (256 - distx) * (256 - disty) - const uint trw = distxsl8 - distxy; // distx * (256 - disty) - const uint blw = distysl8 - distxy; // (256 - distx) * disty - const uint brw = distxy; // distx * disty - uint red = ((tl & 0xf800) * tlw + (tr & 0xf800) * trw - + (bl & 0xf800) * blw + (br & 0xf800) * brw) & 0xf8000000; - uint green = ((tl & 0x07e0) * tlw + (tr & 0x07e0) * trw - + (bl & 0x07e0) * blw + (br & 0x07e0) * brw) & 0x07e00000; - uint blue = ((tl & 0x001f) * tlw + (tr & 0x001f) * trw - + (bl & 0x001f) * blw + (br & 0x001f) * brw); - *b = quint16((red | green | blue) >> 16); - - ++b; - x += fdx; - y += fdy; - w += fdw; - } - - if (ialpha != 0) - blend_sourceOver_rgb16_rgb16(dest, buffer, l, alpha, ialpha); - - dest += l; - length -= l; - } - ++spans; - } - } -} - -static 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(count, spans, userData); - return; - } - - CompositionFunction func = functionForMode[data->rasterBuffer->compositionMode]; - uint buffer[buffer_size]; - quint32 mask = (data->texture.format == QImage::Format_RGB32) ? 0xff000000 : 0; - - const int image_x1 = data->texture.x1; - const int image_y1 = data->texture.y1; - const int image_x2 = data->texture.x2 - 1; - const int image_y2 = data->texture.y2 - 1; - - 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); - - while (count--) { - void *t = data->rasterBuffer->scanLine(spans->y); - - uint *target = ((uint *)t) + spans->x; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - int x = int((data->m21 * cy - + data->m11 * cx + data->dx) * fixed_scale); - int y = int((data->m22 * cy - + data->m12 * cx + data->dy) * fixed_scale); - - int length = spans->len; - const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; - while (length) { - int l = qMin(length, buffer_size); - const uint *end = buffer + l; - uint *b = buffer; - while (b < end) { - int px = qBound(image_x1, x >> 16, image_x2); - int py = qBound(image_y1, y >> 16, image_y2); - *b = reinterpret_cast<const uint *>(data->texture.scanLine(py))[px] | mask; - - x += fdx; - y += fdy; - ++b; - } - func(target, buffer, l, coverage); - target += l; - length -= l; - } - ++spans; - } - } else { - const qreal fdx = data->m11; - const qreal fdy = data->m12; - const qreal fdw = data->m13; - while (count--) { - void *t = data->rasterBuffer->scanLine(spans->y); - - uint *target = ((uint *)t) + spans->x; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - qreal x = data->m21 * cy + data->m11 * cx + data->dx; - qreal y = data->m22 * cy + data->m12 * cx + data->dy; - qreal w = data->m23 * cy + data->m13 * cx + data->m33; - - int length = spans->len; - const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; - while (length) { - int l = qMin(length, buffer_size); - const uint *end = buffer + l; - uint *b = buffer; - while (b < end) { - const qreal iw = w == 0 ? 1 : 1 / w; - const qreal tx = x * iw; - const qreal ty = y * iw; - const int px = qBound(image_x1, int(tx) - (tx < 0), image_x2); - const int py = qBound(image_y1, int(ty) - (ty < 0), image_y2); - - *b = reinterpret_cast<const uint *>(data->texture.scanLine(py))[px] | mask; - x += fdx; - y += fdy; - w += fdw; - - ++b; - } - func(target, buffer, l, coverage); - target += l; - length -= l; - } - ++spans; - } - } -} - -static void blend_transformed_rgb565(int count, const QSpan *spans, void *userData) -{ - QSpanData *data = reinterpret_cast<QSpanData*>(userData); - QPainter::CompositionMode mode = data->rasterBuffer->compositionMode; - - if (data->texture.format != QImage::Format_RGB16 - || (mode != QPainter::CompositionMode_SourceOver - && mode != QPainter::CompositionMode_Source)) - { - blend_src_generic(count, spans, userData); - return; - } - - quint16 buffer[buffer_size]; - const int image_x1 = data->texture.x1; - const int image_y1 = data->texture.y1; - const int image_x2 = data->texture.x2 - 1; - const int image_y2 = data->texture.y2 - 1; - - if (data->fast_matrix) { - // The increment pr x in the scanline - const int fdx = (int)(data->m11 * fixed_scale); - const int fdy = (int)(data->m12 * fixed_scale); - - while (count--) { - const quint8 coverage = (data->texture.const_alpha * spans->coverage) >> 8; - const quint8 alpha = (coverage + 1) >> 3; - const quint8 ialpha = 0x20 - alpha; - if (alpha == 0) { - ++spans; - continue; - } - - quint16 *dest = (quint16 *)data->rasterBuffer->scanLine(spans->y) + spans->x; - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - int x = int((data->m21 * cy - + data->m11 * cx + data->dx) * fixed_scale); - int y = int((data->m22 * cy - + data->m12 * cx + data->dy) * fixed_scale); - int length = spans->len; - - while (length) { - int l; - quint16 *b; - if (ialpha == 0) { - l = length; - b = dest; - } else { - l = qMin(length, buffer_size); - b = buffer; - } - const quint16 *end = b + l; - - while (b < end) { - const int px = qBound(image_x1, x >> 16, image_x2); - const int py = qBound(image_y1, y >> 16, image_y2); - - *b = ((const quint16 *)data->texture.scanLine(py))[px]; - ++b; - - x += fdx; - y += fdy; - } - - if (ialpha != 0) - blend_sourceOver_rgb16_rgb16(dest, buffer, l, alpha, ialpha); - - dest += l; - length -= l; - } - ++spans; - } - } else { - const qreal fdx = data->m11; - const qreal fdy = data->m12; - const qreal fdw = data->m13; - - while (count--) { - const quint8 coverage = (data->texture.const_alpha * spans->coverage) >> 8; - const quint8 alpha = (coverage + 1) >> 3; - const quint8 ialpha = 0x20 - alpha; - if (alpha == 0) { - ++spans; - continue; - } - - quint16 *dest = (quint16 *)data->rasterBuffer->scanLine(spans->y) + spans->x; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - qreal x = data->m21 * cy + data->m11 * cx + data->dx; - qreal y = data->m22 * cy + data->m12 * cx + data->dy; - qreal w = data->m23 * cy + data->m13 * cx + data->m33; - - int length = spans->len; - while (length) { - int l; - quint16 *b; - if (ialpha == 0) { - l = length; - b = dest; - } else { - l = qMin(length, buffer_size); - b = buffer; - } - const quint16 *end = b + l; - - while (b < end) { - const qreal iw = w == 0 ? 1 : 1 / w; - const qreal tx = x * iw; - const qreal ty = y * iw; - - const int px = qBound(image_x1, int(tx) - (tx < 0), image_x2); - const int py = qBound(image_y1, int(ty) - (ty < 0), image_y2); - - *b = ((const quint16 *)data->texture.scanLine(py))[px]; - ++b; - - x += fdx; - y += fdy; - w += fdw; - } - - if (ialpha != 0) - blend_sourceOver_rgb16_rgb16(dest, buffer, l, alpha, ialpha); - - dest += l; - length -= l; - } - ++spans; - } - } -} - -static 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(count, spans, userData); - return; - } - - CompositionFunction func = functionForMode[data->rasterBuffer->compositionMode]; - uint buffer[buffer_size]; - - int image_width = data->texture.width; - int image_height = data->texture.height; - const qsizetype scanline_offset = data->texture.bytesPerLine / 4; - - 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); - - while (count--) { - void *t = data->rasterBuffer->scanLine(spans->y); - - uint *target = ((uint *)t) + spans->x; - const uint *image_bits = (const uint *)data->texture.imageData; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - int x = int((data->m21 * cy - + data->m11 * cx + data->dx) * fixed_scale); - int y = int((data->m22 * cy - + data->m12 * cx + data->dy) * fixed_scale); - - const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; - int length = spans->len; - while (length) { - int l = qMin(length, buffer_size); - const uint *end = buffer + l; - uint *b = buffer; - int px16 = x % (image_width << 16); - int py16 = y % (image_height << 16); - int px_delta = fdx % (image_width << 16); - int py_delta = fdy % (image_height << 16); - while (b < end) { - if (px16 < 0) px16 += image_width << 16; - if (py16 < 0) py16 += image_height << 16; - int px = px16 >> 16; - int py = py16 >> 16; - int y_offset = py * scanline_offset; - - Q_ASSERT(px >= 0 && px < image_width); - Q_ASSERT(py >= 0 && py < image_height); - - *b = image_bits[y_offset + px]; - x += fdx; - y += fdy; - px16 += px_delta; - if (px16 >= image_width << 16) - px16 -= image_width << 16; - py16 += py_delta; - if (py16 >= image_height << 16) - py16 -= image_height << 16; - ++b; - } - func(target, buffer, l, coverage); - target += l; - length -= l; - } - ++spans; - } - } else { - const qreal fdx = data->m11; - const qreal fdy = data->m12; - const qreal fdw = data->m13; - while (count--) { - void *t = data->rasterBuffer->scanLine(spans->y); - - uint *target = ((uint *)t) + spans->x; - const uint *image_bits = (const uint *)data->texture.imageData; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - qreal x = data->m21 * cy + data->m11 * cx + data->dx; - qreal y = data->m22 * cy + data->m12 * cx + data->dy; - qreal w = data->m23 * cy + data->m13 * cx + data->m33; - - const int coverage = (spans->coverage * data->texture.const_alpha) >> 8; - int length = spans->len; - while (length) { - int l = qMin(length, buffer_size); - const uint *end = buffer + l; - uint *b = buffer; - while (b < end) { - const qreal iw = w == 0 ? 1 : 1 / w; - const qreal tx = x * iw; - const qreal ty = y * 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; - int y_offset = py * scanline_offset; - - Q_ASSERT(px >= 0 && px < image_width); - Q_ASSERT(py >= 0 && py < image_height); - - *b = image_bits[y_offset + px]; - x += fdx; - y += fdy; - w += fdw; - //force increment to avoid /0 - if (!w) { - w += fdw; - } - ++b; - } - func(target, buffer, l, coverage); - target += l; - length -= l; - } - ++spans; - } - } -} - -static void blend_transformed_tiled_rgb565(int count, const QSpan *spans, void *userData) -{ - QSpanData *data = reinterpret_cast<QSpanData*>(userData); - QPainter::CompositionMode mode = data->rasterBuffer->compositionMode; - - if (data->texture.format != QImage::Format_RGB16 - || (mode != QPainter::CompositionMode_SourceOver - && mode != QPainter::CompositionMode_Source)) - { - blend_src_generic(count, spans, userData); - return; - } - - quint16 buffer[buffer_size]; - const int image_width = data->texture.width; - const int image_height = data->texture.height; - - if (data->fast_matrix) { - // The increment pr x in the scanline - const int fdx = (int)(data->m11 * fixed_scale); - const int fdy = (int)(data->m12 * fixed_scale); - - while (count--) { - const quint8 coverage = (data->texture.const_alpha * spans->coverage) >> 8; - const quint8 alpha = (coverage + 1) >> 3; - const quint8 ialpha = 0x20 - alpha; - if (alpha == 0) { - ++spans; - continue; - } - - quint16 *dest = (quint16 *)data->rasterBuffer->scanLine(spans->y) + spans->x; - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - int x = int((data->m21 * cy - + data->m11 * cx + data->dx) * fixed_scale); - int y = int((data->m22 * cy - + data->m12 * cx + data->dy) * fixed_scale); - int length = spans->len; - - while (length) { - int l; - quint16 *b; - if (ialpha == 0) { - l = length; - b = dest; - } else { - l = qMin(length, buffer_size); - b = buffer; - } - const quint16 *end = b + l; - - while (b < end) { - int px = (x >> 16) % image_width; - int py = (y >> 16) % image_height; - - if (px < 0) - px += image_width; - if (py < 0) - py += image_height; - - *b = ((const quint16 *)data->texture.scanLine(py))[px]; - ++b; - - x += fdx; - y += fdy; - } - - if (ialpha != 0) - blend_sourceOver_rgb16_rgb16(dest, buffer, l, alpha, ialpha); - - dest += l; - length -= l; - } - ++spans; - } - } else { - const qreal fdx = data->m11; - const qreal fdy = data->m12; - const qreal fdw = data->m13; - - while (count--) { - const quint8 coverage = (data->texture.const_alpha * spans->coverage) >> 8; - const quint8 alpha = (coverage + 1) >> 3; - const quint8 ialpha = 0x20 - alpha; - if (alpha == 0) { - ++spans; - continue; - } - - quint16 *dest = (quint16 *)data->rasterBuffer->scanLine(spans->y) + spans->x; - - const qreal cx = spans->x + qreal(0.5); - const qreal cy = spans->y + qreal(0.5); - - qreal x = data->m21 * cy + data->m11 * cx + data->dx; - qreal y = data->m22 * cy + data->m12 * cx + data->dy; - qreal w = data->m23 * cy + data->m13 * cx + data->m33; - - int length = spans->len; - while (length) { - int l; - quint16 *b; - if (ialpha == 0) { - l = length; - b = dest; - } else { - l = qMin(length, buffer_size); - b = buffer; - } - const quint16 *end = b + l; - - while (b < end) { - const qreal iw = w == 0 ? 1 : 1 / w; - const qreal tx = x * iw; - const qreal ty = y * 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; - - *b = ((const quint16 *)data->texture.scanLine(py))[px]; - ++b; - - x += fdx; - y += fdy; - w += fdw; - // force increment to avoid /0 - if (!w) - w += fdw; - } - - if (ialpha != 0) - blend_sourceOver_rgb16_rgb16(dest, buffer, l, alpha, ialpha); - - dest += l; - length -= l; - } - ++spans; - } - } -} - - /* Image formats here are target formats */ static const ProcessSpans processTextureSpansARGB32PM[NBlendTypes] = { blend_untransformed_argb, // Untransformed blend_tiled_argb, // Tiled - blend_transformed_argb, // Transformed - blend_transformed_tiled_argb, // TransformedTiled + blend_src_generic, // Transformed + blend_src_generic, // TransformedTiled blend_src_generic, // TransformedBilinear blend_src_generic // TransformedBilinearTiled }; @@ -5286,9 +5169,9 @@ static const ProcessSpans processTextureSpansARGB32PM[NBlendTypes] = { static const ProcessSpans processTextureSpansRGB16[NBlendTypes] = { blend_untransformed_rgb565, // Untransformed blend_tiled_rgb565, // Tiled - blend_transformed_rgb565, // Transformed - blend_transformed_tiled_rgb565, // TransformedTiled - blend_transformed_bilinear_rgb565, // TransformedBilinear + blend_src_generic, // Transformed + blend_src_generic, // TransformedTiled + blend_src_generic, // TransformedBilinear blend_src_generic // TransformedBilinearTiled }; @@ -5330,6 +5213,9 @@ void qBlendTexture(int count, const QSpan *spans, void *userData) case QImage::Format_A2BGR30_Premultiplied: case QImage::Format_RGB30: case QImage::Format_A2RGB30_Premultiplied: + case QImage::Format_RGBX64: + case QImage::Format_RGBA64: + case QImage::Format_RGBA64_Premultiplied: proc = processTextureSpansGeneric64[blendType]; break; case QImage::Format_Invalid: @@ -5586,7 +5472,7 @@ static void qt_alphamapblit_generic(QRasterBuffer *rasterBuffer, srcColor = colorProfile->toLinear(srcColor.unpremultiplied()).premultiplied(); } - quint64 buffer[buffer_size]; + quint64 buffer[BufferSize]; const DestFetchProc64 destFetch64 = destFetchProc64[rasterBuffer->format]; const DestStoreProc64 destStore64 = destStoreProc64[rasterBuffer->format]; @@ -5595,13 +5481,14 @@ static void qt_alphamapblit_generic(QRasterBuffer *rasterBuffer, int i = x; int length = mapWidth; while (length > 0) { - int l = qMin(buffer_size, length); + int l = qMin(BufferSize, length); QRgba64 *dest = destFetch64((QRgba64*)buffer, rasterBuffer, i, y + ly, l); for (int j=0; j < l; ++j) { const int coverage = map[j + (i - x)]; alphamapblend_generic(coverage, dest, j, srcColor, color, colorProfile); } - destStore64(rasterBuffer, i, y + ly, dest, l); + if (destStore64) + destStore64(rasterBuffer, i, y + ly, dest, l); length -= l; i += l; } @@ -5624,14 +5511,15 @@ static void qt_alphamapblit_generic(QRasterBuffer *rasterBuffer, int end = qMin<int>(x + mapWidth, clip.x + clip.len); if (end <= start) continue; - Q_ASSERT(end - start <= buffer_size); + Q_ASSERT(end - start <= BufferSize); QRgba64 *dest = destFetch64((QRgba64*)buffer, rasterBuffer, start, clip.y, end - start); for (int xp=start; xp<end; ++xp) { const int coverage = map[xp - x]; alphamapblend_generic(coverage, dest, xp - start, srcColor, color, colorProfile); } - destStore64(rasterBuffer, start, clip.y, dest, end - start); + if (destStore64) + destStore64(rasterBuffer, start, clip.y, dest, end - start); } // for (i -> line.count) map += mapStride; } // for (yp -> bottom) @@ -5903,7 +5791,7 @@ static void qt_alphargbblit_generic(QRasterBuffer *rasterBuffer, srcColor = colorProfile->toLinear(srcColor.unpremultiplied()).premultiplied(); } - quint64 buffer[buffer_size]; + quint64 buffer[BufferSize]; const DestFetchProc64 destFetch64 = destFetchProc64[rasterBuffer->format]; const DestStoreProc64 destStore64 = destStoreProc64[rasterBuffer->format]; @@ -5912,13 +5800,14 @@ static void qt_alphargbblit_generic(QRasterBuffer *rasterBuffer, int i = x; int length = mapWidth; while (length > 0) { - int l = qMin(buffer_size, length); + int l = qMin(BufferSize, length); QRgba64 *dest = destFetch64((QRgba64*)buffer, rasterBuffer, i, y + ly, l); for (int j=0; j < l; ++j) { const uint coverage = src[j + (i - x)]; alphargbblend_generic(coverage, dest, j, srcColor, color, colorProfile); } - destStore64(rasterBuffer, i, y + ly, dest, l); + if (destStore64) + destStore64(rasterBuffer, i, y + ly, dest, l); length -= l; i += l; } @@ -5941,14 +5830,15 @@ static void qt_alphargbblit_generic(QRasterBuffer *rasterBuffer, int end = qMin<int>(x + mapWidth, clip.x + clip.len); if (end <= start) continue; - Q_ASSERT(end - start <= buffer_size); + Q_ASSERT(end - start <= BufferSize); QRgba64 *dest = destFetch64((QRgba64*)buffer, rasterBuffer, start, clip.y, end - start); for (int xp=start; xp<end; ++xp) { const uint coverage = src[xp - x]; alphargbblend_generic(coverage, dest, xp - start, srcColor, color, colorProfile); } - destStore64(rasterBuffer, start, clip.y, dest, end - start); + if (destStore64) + destStore64(rasterBuffer, start, clip.y, dest, end - start); } // for (i -> line.count) src += srcStride; } // for (yp -> bottom) @@ -6031,8 +5921,24 @@ static void qt_rectfill_quint16(QRasterBuffer *rasterBuffer, int x, int y, int width, int height, const QRgba64 &color) { + const QPixelLayout &layout = qPixelLayouts[rasterBuffer->format]; + quint32 c32 = color.toArgb32(); + quint16 c16; + layout.storeFromARGB32PM(reinterpret_cast<uchar *>(&c16), &c32, 0, 1, nullptr, nullptr); qt_rectfill<quint16>(reinterpret_cast<quint16 *>(rasterBuffer->buffer()), - color.toRgb16(), x, y, width, height, rasterBuffer->bytesPerLine()); + c16, x, y, width, height, rasterBuffer->bytesPerLine()); +} + +static void qt_rectfill_quint24(QRasterBuffer *rasterBuffer, + int x, int y, int width, int height, + const QRgba64 &color) +{ + const QPixelLayout &layout = qPixelLayouts[rasterBuffer->format]; + quint32 c32 = color.toArgb32(); + quint24 c24; + layout.storeFromARGB32PM(reinterpret_cast<uchar *>(&c24), &c32, 0, 1, nullptr, nullptr); + qt_rectfill<quint24>(reinterpret_cast<quint24 *>(rasterBuffer->buffer()), + c24, x, y, width, height, rasterBuffer->bytesPerLine()); } static void qt_rectfill_nonpremul_argb32(QRasterBuffer *rasterBuffer, @@ -6084,6 +5990,17 @@ static void qt_rectfill_gray(QRasterBuffer *rasterBuffer, qGray(color.toArgb32()), x, y, width, height, rasterBuffer->bytesPerLine()); } +static void qt_rectfill_quint64(QRasterBuffer *rasterBuffer, + int x, int y, int width, int height, + const QRgba64 &color) +{ + const auto store = qStoreFromRGBA64PM[rasterBuffer->format]; + quint64 c64; + store(reinterpret_cast<uchar *>(&c64), &color, 0, 1, nullptr, nullptr); + qt_rectfill<quint64>(reinterpret_cast<quint64 *>(rasterBuffer->buffer()), + c64, x, y, width, height, rasterBuffer->bytesPerLine()); +} + // Map table for destination image format. Contains function pointers // for blends of various types unto the destination @@ -6152,7 +6069,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint24 }, // Format_RGB666 { @@ -6161,7 +6078,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint24 }, // Format_ARGB6666_Premultiplied { @@ -6170,7 +6087,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint24 }, // Format_RGB555 { @@ -6179,7 +6096,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint16 }, // Format_ARGB8555_Premultiplied { @@ -6188,7 +6105,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint24 }, // Format_RGB888 { @@ -6197,7 +6114,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint24 }, // Format_RGB444 { @@ -6206,7 +6123,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint16 }, // Format_ARGB4444_Premultiplied { @@ -6215,7 +6132,7 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = 0, qt_alphamapblit_generic, qt_alphargbblit_generic, - 0 + qt_rectfill_quint16 }, // Format_RGBX8888 { @@ -6298,6 +6215,33 @@ DrawHelper qDrawHelper[QImage::NImageFormats] = qt_alphargbblit_generic, qt_rectfill_gray }, + // Format_RGBX64 + { + blend_color_generic_rgb64, + blend_src_generic_rgb64, + 0, + qt_alphamapblit_generic, + qt_alphargbblit_generic, + qt_rectfill_quint64 + }, + // Format_RGBA64 + { + blend_color_generic_rgb64, + blend_src_generic_rgb64, + 0, + qt_alphamapblit_generic, + qt_alphargbblit_generic, + qt_rectfill_quint64 + }, + // Format_RGBA64_Premultiplied + { + blend_color_generic_rgb64, + blend_src_generic_rgb64, + 0, + qt_alphamapblit_generic, + qt_alphargbblit_generic, + qt_rectfill_quint64 + }, }; #if defined(Q_CC_MSVC) && !defined(_MIPS_) @@ -6370,7 +6314,7 @@ void qt_memfill32(quint32 *dest, quint32 color, int count) #endif #ifdef QT_COMPILER_SUPPORTS_SSE4_1 -template<QtPixelOrder> const uint *QT_FASTCALL convertA2RGB30PMFromARGB32PM_sse4(uint *buffer, const uint *src, int count, const QVector<QRgb> *, QDitherInfo *); +template<QtPixelOrder> void QT_FASTCALL storeA2RGB30PMFromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, const QVector<QRgb> *, QDitherInfo *); #endif extern void qInitBlendFunctions(); @@ -6439,42 +6383,54 @@ static void qInitDrawhelperFunctions() int w, int h, int const_alpha); - extern void QT_FASTCALL storePixelsBPP24_ssse3(uchar *dest, const uint *src, int index, int count); extern const uint * QT_FASTCALL qt_fetchUntransformed_888_ssse3(uint *buffer, const Operator *, const QSpanData *data, int y, int x, int length); 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; qBlendFunctions[QImage::Format_RGBX8888][QImage::Format_RGBA8888_Premultiplied] = qt_blend_argb32_on_argb32_ssse3; qBlendFunctions[QImage::Format_RGBA8888_Premultiplied][QImage::Format_RGBA8888_Premultiplied] = qt_blend_argb32_on_argb32_ssse3; - qStorePixels[QPixelLayout::BPP24] = storePixelsBPP24_ssse3; sourceFetchUntransformed[QImage::Format_RGB888] = qt_fetchUntransformed_888_ssse3; } #endif // SSSE3 #if defined(QT_COMPILER_SUPPORTS_SSE4_1) if (qCpuHasFeature(SSE4_1)) { - extern const uint *QT_FASTCALL convertARGB32ToARGB32PM_sse4(uint *buffer, const uint *src, int count, + extern void QT_FASTCALL convertARGB32ToARGB32PM_sse4(uint *buffer, int count, const QVector<QRgb> *); + extern void QT_FASTCALL convertRGBA8888ToARGB32PM_sse4(uint *buffer, int count, const QVector<QRgb> *); + extern const uint *QT_FASTCALL fetchARGB32ToARGB32PM_sse4(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + extern const uint *QT_FASTCALL fetchRGBA8888ToARGB32PM_sse4(uint *buffer, const uchar *src, int index, int count, const QVector<QRgb> *, QDitherInfo *); - extern const uint *QT_FASTCALL convertRGBA8888ToARGB32PM_sse4(uint *buffer, const uint *src, int count, + extern void QT_FASTCALL storeARGB32FromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, const QVector<QRgb> *, QDitherInfo *); - extern const uint *QT_FASTCALL convertARGB32FromARGB32PM_sse4(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *); - extern const uint *QT_FASTCALL convertRGBA8888FromARGB32PM_sse4(uint *buffer, const uint *src, int count, + extern void QT_FASTCALL storeRGBA8888FromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, const QVector<QRgb> *, QDitherInfo *); - extern const uint *QT_FASTCALL convertRGBXFromARGB32PM_sse4(uint *buffer, const uint *src, int count, + extern void QT_FASTCALL storeRGBXFromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, const QVector<QRgb> *, QDitherInfo *); + extern void QT_FASTCALL storeARGB32FromRGBA64PM_sse4(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + extern void QT_FASTCALL storeRGBA8888FromRGBA64PM_sse4(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + extern void QT_FASTCALL destStore64ARGB32_sse4(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length); + extern void QT_FASTCALL destStore64RGBA8888_sse4(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length); + qPixelLayouts[QImage::Format_ARGB32].fetchToARGB32PM = fetchARGB32ToARGB32PM_sse4; qPixelLayouts[QImage::Format_ARGB32].convertToARGB32PM = convertARGB32ToARGB32PM_sse4; + qPixelLayouts[QImage::Format_RGBA8888].fetchToARGB32PM = fetchRGBA8888ToARGB32PM_sse4; qPixelLayouts[QImage::Format_RGBA8888].convertToARGB32PM = convertRGBA8888ToARGB32PM_sse4; - qPixelLayouts[QImage::Format_ARGB32].convertFromARGB32PM = convertARGB32FromARGB32PM_sse4; - qPixelLayouts[QImage::Format_RGBA8888].convertFromARGB32PM = convertRGBA8888FromARGB32PM_sse4; - qPixelLayouts[QImage::Format_RGBX8888].convertFromARGB32PM = convertRGBXFromARGB32PM_sse4; - qPixelLayouts[QImage::Format_A2BGR30_Premultiplied].convertFromARGB32PM = convertA2RGB30PMFromARGB32PM_sse4<PixelOrderBGR>; - qPixelLayouts[QImage::Format_A2RGB30_Premultiplied].convertFromARGB32PM = convertA2RGB30PMFromARGB32PM_sse4<PixelOrderRGB>; + qPixelLayouts[QImage::Format_ARGB32].storeFromARGB32PM = storeARGB32FromARGB32PM_sse4; + qPixelLayouts[QImage::Format_RGBA8888].storeFromARGB32PM = storeRGBA8888FromARGB32PM_sse4; + qPixelLayouts[QImage::Format_RGBX8888].storeFromARGB32PM = storeRGBXFromARGB32PM_sse4; + qPixelLayouts[QImage::Format_A2BGR30_Premultiplied].storeFromARGB32PM = storeA2RGB30PMFromARGB32PM_sse4<PixelOrderBGR>; + qPixelLayouts[QImage::Format_A2RGB30_Premultiplied].storeFromARGB32PM = storeA2RGB30PMFromARGB32PM_sse4<PixelOrderRGB>; + qStoreFromRGBA64PM[QImage::Format_ARGB32] = storeARGB32FromRGBA64PM_sse4; + qStoreFromRGBA64PM[QImage::Format_RGBA8888] = storeRGBA8888FromRGBA64PM_sse4; + destStoreProc64[QImage::Format_ARGB32] = destStore64ARGB32_sse4; + destStoreProc64[QImage::Format_RGBA8888] = destStore64RGBA8888_sse4; } #endif #if defined(QT_COMPILER_SUPPORTS_AVX2) - if (qCpuHasFeature(AVX2)) { + if (qCpuHasFeature(ArchHaswell)) { extern void qt_blend_rgb32_on_rgb32_avx2(uchar *destPixels, int dbpl, const uchar *srcPixels, int sbpl, int w, int h, int const_alpha); @@ -6504,14 +6460,14 @@ static void qInitDrawhelperFunctions() qt_functionForModeSolid_C[QPainter::CompositionMode_SourceOver] = comp_func_solid_SourceOver_avx2; qt_functionForModeSolid64_C[QPainter::CompositionMode_SourceOver] = comp_func_solid_SourceOver_rgb64_avx2; - extern void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uint *b, uint *end, const QTextureData &image, - int &fx, int &fy, int fdx, int /*fdy*/); + extern void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_scale_helper_avx2(uint *b, uint *end, const QTextureData &image, + int &fx, int &fy, int fdx, int /*fdy*/); extern void QT_FASTCALL fetchTransformedBilinearARGB32PM_downscale_helper_avx2(uint *b, uint *end, const QTextureData &image, int &fx, int &fy, int fdx, int /*fdy*/); extern void QT_FASTCALL fetchTransformedBilinearARGB32PM_fast_rotate_helper_avx2(uint *b, uint *end, const QTextureData &image, int &fx, int &fy, int fdx, int fdy); - bilinearFastTransformHelperARGB32PM[0][SimpleUpscaleTransform] = fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2; + bilinearFastTransformHelperARGB32PM[0][SimpleScaleTransform] = fetchTransformedBilinearARGB32PM_simple_scale_helper_avx2; bilinearFastTransformHelperARGB32PM[0][DownscaleTransform] = fetchTransformedBilinearARGB32PM_downscale_helper_avx2; bilinearFastTransformHelperARGB32PM[0][FastRotateTransform] = fetchTransformedBilinearARGB32PM_fast_rotate_helper_avx2; } @@ -6543,12 +6499,25 @@ static void qInitDrawhelperFunctions() sourceFetchUntransformed[QImage::Format_RGB888] = qt_fetchUntransformed_888_neon; #if Q_BYTE_ORDER == Q_LITTLE_ENDIAN - extern const uint *QT_FASTCALL convertARGB32ToARGB32PM_neon(uint *buffer, const uint *src, int count, + extern void QT_FASTCALL convertARGB32ToARGB32PM_neon(uint *buffer, int count, const QVector<QRgb> *); + extern void QT_FASTCALL convertRGBA8888ToARGB32PM_neon(uint *buffer, int count, const QVector<QRgb> *); + extern const uint *QT_FASTCALL fetchARGB32ToARGB32PM_neon(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + extern const uint *QT_FASTCALL fetchRGBA8888ToARGB32PM_neon(uint *buffer, const uchar *src, int index, int count, const QVector<QRgb> *, QDitherInfo *); - extern const uint *QT_FASTCALL convertRGBA8888ToARGB32PM_neon(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *); + extern void QT_FASTCALL storeARGB32FromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + extern void QT_FASTCALL storeRGBA8888FromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + extern void QT_FASTCALL storeRGBXFromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); + qPixelLayouts[QImage::Format_ARGB32].fetchToARGB32PM = fetchARGB32ToARGB32PM_neon; qPixelLayouts[QImage::Format_ARGB32].convertToARGB32PM = convertARGB32ToARGB32PM_neon; + qPixelLayouts[QImage::Format_ARGB32].storeFromARGB32PM = storeARGB32FromARGB32PM_neon; + qPixelLayouts[QImage::Format_RGBA8888].fetchToARGB32PM = fetchRGBA8888ToARGB32PM_neon; qPixelLayouts[QImage::Format_RGBA8888].convertToARGB32PM = convertRGBA8888ToARGB32PM_neon; + qPixelLayouts[QImage::Format_RGBA8888].storeFromARGB32PM = storeRGBA8888FromARGB32PM_neon; + qPixelLayouts[QImage::Format_RGBX8888].storeFromARGB32PM = storeRGBXFromARGB32PM_neon; #endif #if defined(ENABLE_PIXMAN_DRAWHELPERS) diff --git a/src/gui/painting/qdrawhelper_avx2.cpp b/src/gui/painting/qdrawhelper_avx2.cpp index 3a70524a9d..ec6643deed 100644 --- a/src/gui/painting/qdrawhelper_avx2.cpp +++ b/src/gui/painting/qdrawhelper_avx2.cpp @@ -45,8 +45,6 @@ QT_BEGIN_NAMESPACE -static Q_CONSTEXPR int BufferSize = 2048; - enum { FixedScale = 1 << 16, HalfPoint = 1 << 15 @@ -576,8 +574,10 @@ inline void fetchTransformedBilinear_pixelBounds(int, int l1, int l2, int &v1, i Q_ASSERT(v2 >= l1 && v2 <= l2); } -void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uint *b, uint *end, const QTextureData &image, - int &fx, int &fy, int fdx, int /*fdy*/) +void QT_FASTCALL intermediate_adder_avx2(uint *b, uint *end, const IntermediateBuffer &intermediate, int offset, int &fx, int fdx); + +void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_scale_helper_avx2(uint *b, uint *end, const QTextureData &image, + int &fx, int &fy, int fdx, int /*fdy*/) { int y1 = (fy >> 16); int y2; @@ -594,16 +594,12 @@ void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uin const int offset = (fx + adjust) >> 16; int x = offset; - // The idea is first to do the interpolation between the row s1 and the row s2 - // into an intermediate buffer, then we interpolate between two pixel of this buffer. - - // intermediate_buffer[0] is a buffer of red-blue component of the pixel, in the form 0x00RR00BB - // intermediate_buffer[1] is the alpha-green component of the pixel, in the form 0x00AA00GG - // +1 for the last pixel to interpolate with, and +1 for rounding errors. - quint32 intermediate_buffer[2][BufferSize + 2]; + IntermediateBuffer intermediate; // count is the size used in the intermediate_buffer. int count = (qint64(length) * qAbs(fdx) + FixedScale - 1) / FixedScale + 2; - Q_ASSERT(count <= BufferSize + 2); //length is supposed to be <= buffer_size and data->m11 < 1 in this case + // length is supposed to be <= BufferSize either because data->m11 < 1 or + // data->m11 < 2, and any larger buffers split + Q_ASSERT(count <= BufferSize + 2); int f = 0; int lim = qMin(count, image.x2 - x); if (x < image.x1) { @@ -613,8 +609,8 @@ void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uin quint32 rb = (((t & 0xff00ff) * idisty + (b & 0xff00ff) * disty) >> 8) & 0xff00ff; quint32 ag = ((((t>>8) & 0xff00ff) * idisty + ((b>>8) & 0xff00ff) * disty) >> 8) & 0xff00ff; do { - intermediate_buffer[0][f] = rb; - intermediate_buffer[1][f] = ag; + intermediate.buffer_rb[f] = rb; + intermediate.buffer_ag[f] = ag; f++; x++; } while (x < image.x1 && f < lim); @@ -644,10 +640,10 @@ void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uin // Add the values, and shift to only keep 8 significant bits per colors __m256i rAG =_mm256_add_epi16(topAG, bottomAG); rAG = _mm256_srli_epi16(rAG, 8); - _mm256_storeu_si256((__m256i*)(&intermediate_buffer[1][f]), rAG); + _mm256_storeu_si256((__m256i*)(&intermediate.buffer_ag[f]), rAG); __m256i rRB =_mm256_add_epi16(topRB, bottomRB); rRB = _mm256_srli_epi16(rRB, 8); - _mm256_storeu_si256((__m256i*)(&intermediate_buffer[0][f]), rRB); + _mm256_storeu_si256((__m256i*)(&intermediate.buffer_rb[f]), rRB); } for (; f < count; f++) { // Same as above but without simd @@ -656,30 +652,37 @@ void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uin uint t = s1[x]; uint b = s2[x]; - intermediate_buffer[0][f] = (((t & 0xff00ff) * idisty + (b & 0xff00ff) * disty) >> 8) & 0xff00ff; - intermediate_buffer[1][f] = ((((t>>8) & 0xff00ff) * idisty + ((b>>8) & 0xff00ff) * disty) >> 8) & 0xff00ff; + intermediate.buffer_rb[f] = (((t & 0xff00ff) * idisty + (b & 0xff00ff) * disty) >> 8) & 0xff00ff; + intermediate.buffer_ag[f] = ((((t>>8) & 0xff00ff) * idisty + ((b>>8) & 0xff00ff) * disty) >> 8) & 0xff00ff; x++; } + // Now interpolate the values from the intermediate_buffer to get the final result. + intermediate_adder_avx2(b, end, intermediate, offset, fx, fdx); +} + +void QT_FASTCALL intermediate_adder_avx2(uint *b, uint *end, const IntermediateBuffer &intermediate, int offset, int &fx, int fdx) +{ fx -= offset * FixedScale; const __m128i v_fdx = _mm_set1_epi32(fdx * 4); const __m128i v_blend = _mm_set1_epi32(0x00800080); + const __m128i vdx_shuffle = _mm_set_epi8(char(0x80), 13, char(0x80), 13, char(0x80), 9, char(0x80), 9, + char(0x80), 5, char(0x80), 5, char(0x80), 1, char(0x80), 1); __m128i v_fx = _mm_setr_epi32(fx, fx + fdx, fx + fdx + fdx, fx + fdx + fdx + fdx); while (b < end - 3) { const __m128i offset = _mm_srli_epi32(v_fx, 16); - __m256i vrb = _mm256_i32gather_epi64((const long long *)intermediate_buffer[0], offset, 4); - __m256i vag = _mm256_i32gather_epi64((const long long *)intermediate_buffer[1], offset, 4); + __m256i vrb = _mm256_i32gather_epi64((const long long *)intermediate.buffer_rb, offset, 4); + __m256i vag = _mm256_i32gather_epi64((const long long *)intermediate.buffer_ag, offset, 4); - __m128i vdx = _mm_and_si128(v_fx, _mm_set1_epi32(0x0000ffff)); - vdx = _mm_srli_epi16(vdx, 8); - __m128i vidx = _mm_sub_epi32(_mm_set1_epi32(256), vdx); + __m128i vdx = _mm_shuffle_epi8(v_fx, vdx_shuffle); + __m128i vidx = _mm_sub_epi16(_mm_set1_epi16(256), vdx); __m256i vmulx = _mm256_castsi128_si256(_mm_unpacklo_epi32(vidx, vdx)); vmulx = _mm256_inserti128_si256(vmulx, _mm_unpackhi_epi32(vidx, vdx), 1); - vrb = _mm256_mullo_epi32(vrb, vmulx); - vag = _mm256_mullo_epi32(vag, vmulx); + vrb = _mm256_mullo_epi16(vrb, vmulx); + vag = _mm256_mullo_epi16(vag, vmulx); __m256i vrbag = _mm256_hadd_epi32(vrb, vag); vrbag = _mm256_permute4x64_epi64(vrbag, _MM_SHUFFLE(3, 1, 2, 0)); @@ -691,21 +694,21 @@ void QT_FASTCALL fetchTransformedBilinearARGB32PM_simple_upscale_helper_avx2(uin _mm_storeu_si128((__m128i*)b, _mm_blendv_epi8(ag, rb, v_blend)); b += 4; - fx += 4 * fdx; v_fx = _mm_add_epi32(v_fx, v_fdx); } + fx = _mm_cvtsi128_si32(v_fx); while (b < end) { - int x = (fx >> 16); - - uint distx = (fx & 0x0000ffff) >> 8; - uint idistx = 256 - distx; + const int x = (fx >> 16); - uint rb = ((intermediate_buffer[0][x] * idistx + intermediate_buffer[0][x + 1] * distx) >> 8) & 0xff00ff; - uint ag = (intermediate_buffer[1][x] * idistx + intermediate_buffer[1][x + 1] * distx) & 0xff00ff00; - *b = rb | ag; + const uint distx = (fx & 0x0000ffff) >> 8; + const uint idistx = 256 - distx; + const uint rb = (intermediate.buffer_rb[x] * idistx + intermediate.buffer_rb[x + 1] * distx) & 0xff00ff00; + const uint ag = (intermediate.buffer_ag[x] * idistx + intermediate.buffer_ag[x + 1] * distx) & 0xff00ff00; + *b = (rb >> 8) | ag; b++; fx += fdx; } + fx += offset * FixedScale; } void QT_FASTCALL fetchTransformedBilinearARGB32PM_downscale_helper_avx2(uint *b, uint *end, const QTextureData &image, diff --git a/src/gui/painting/qdrawhelper_neon.cpp b/src/gui/painting/qdrawhelper_neon.cpp index e126f4b670..629dfe2358 100644 --- a/src/gui/painting/qdrawhelper_neon.cpp +++ b/src/gui/painting/qdrawhelper_neon.cpp @@ -1081,15 +1081,28 @@ const uint * QT_FASTCALL qt_fetchUntransformed_888_neon(uint *buffer, const Oper } #if Q_BYTE_ORDER == Q_LITTLE_ENDIAN -template<bool RGBA> -static inline void convertARGBToARGB32PM_neon(uint *buffer, const uint *src, int count) +static inline uint32x4_t vrgba2argb(uint32x4_t srcVector) { - int i = 0; #if defined(Q_PROCESSOR_ARM_64) const uint8x16_t rgbaMask = { 2, 1, 0, 3, 6, 5, 4, 7, 10, 9, 8, 11, 14, 13, 12, 15}; #else const uint8x8_t rgbaMask = { 2, 1, 0, 3, 6, 5, 4, 7 }; #endif +#if defined(Q_PROCESSOR_ARM_64) + srcVector = vreinterpretq_u32_u8(vqtbl1q_u8(vreinterpretq_u8_u32(srcVector), rgbaMask)); +#else + // no vqtbl1q_u8, so use two vtbl1_u8 + const uint8x8_t low = vtbl1_u8(vreinterpret_u8_u32(vget_low_u32(srcVector)), rgbaMask); + const uint8x8_t high = vtbl1_u8(vreinterpret_u8_u32(vget_high_u32(srcVector)), rgbaMask); + srcVector = vcombine_u32(vreinterpret_u32_u8(low), vreinterpret_u32_u8(high)); +#endif + return srcVector; +} + +template<bool RGBA> +static inline void convertARGBToARGB32PM_neon(uint *buffer, const uint *src, int count) +{ + int i = 0; const uint8x8_t shuffleMask = { 3, 3, 3, 3, 7, 7, 7, 7}; const uint32x4_t blendMask = vdupq_n_u32(0xff000000); @@ -1105,16 +1118,8 @@ static inline void convertARGBToARGB32PM_neon(uint *buffer, const uint *src, int #endif if (alphaSum) { if (alphaSum != 255 * 4) { - if (RGBA) { -#if defined(Q_PROCESSOR_ARM_64) - srcVector = vreinterpretq_u32_u8(vqtbl1q_u8(vreinterpretq_u8_u32(srcVector), rgbaMask)); -#else - // no vqtbl1q_u8 - const uint8x8_t low = vtbl1_u8(vreinterpret_u8_u32(vget_low_u32(srcVector)), rgbaMask); - const uint8x8_t high = vtbl1_u8(vreinterpret_u8_u32(vget_high_u32(srcVector)), rgbaMask); - srcVector = vcombine_u32(vreinterpret_u32_u8(low), vreinterpret_u32_u8(high)); -#endif - } + if (RGBA) + srcVector = vrgba2argb(srcVector); const uint8x8_t s1 = vreinterpret_u8_u32(vget_low_u32(srcVector)); const uint8x8_t s2 = vreinterpret_u8_u32(vget_high_u32(srcVector)); const uint8x8_t alpha1 = vtbl1_u8(s1, shuffleMask); @@ -1128,19 +1133,10 @@ static inline void convertARGBToARGB32PM_neon(uint *buffer, const uint *src, int const uint32x4_t d = vbslq_u32(blendMask, srcVector, vreinterpretq_u32_u8(vcombine_u8(d1, d2))); vst1q_u32(buffer + i, d); } else { - if (RGBA) { -#if defined(Q_PROCESSOR_ARM_64) - srcVector = vreinterpretq_u32_u8(vqtbl1q_u8(vreinterpretq_u8_u32(srcVector), rgbaMask)); -#else - // no vqtbl1q_u8 - const uint8x8_t low = vtbl1_u8(vreinterpret_u8_u32(vget_low_u32(srcVector)), rgbaMask); - const uint8x8_t high = vtbl1_u8(vreinterpret_u8_u32(vget_high_u32(srcVector)), rgbaMask); - srcVector = vcombine_u32(vreinterpret_u32_u8(low), vreinterpret_u32_u8(high)); -#endif - vst1q_u32(buffer + i, srcVector); - } else if (buffer != src) { + if (RGBA) + vst1q_u32(buffer + i, vrgba2argb(srcVector)); + else if (buffer != src) vst1q_u32(buffer + i, srcVector); - } } } else { vst1q_u32(buffer + i, vdupq_n_u32(0)); @@ -1153,20 +1149,143 @@ static inline void convertARGBToARGB32PM_neon(uint *buffer, const uint *src, int } } -const uint *QT_FASTCALL convertARGB32ToARGB32PM_neon(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static inline float32x4_t reciprocal_mul_ps(float32x4_t a, float mul) +{ + float32x4_t ia = vrecpeq_f32(a); // estimate 1/a + ia = vmulq_f32(vrecpsq_f32(a, ia), vmulq_n_f32(ia, mul)); // estimate improvement step * mul + return ia; +} + +template<bool RGBA, bool RGBx> +static inline void convertARGBFromARGB32PM_neon(uint *buffer, const uint *src, int count) +{ + int i = 0; + const uint32x4_t alphaMask = vdupq_n_u32(0xff000000); + + for (; i < count - 3; i += 4) { + uint32x4_t srcVector = vld1q_u32(src + i); + uint32x4_t alphaVector = vshrq_n_u32(srcVector, 24); +#if defined(Q_PROCESSOR_ARM_64) + uint32_t alphaSum = vaddvq_u32(alphaVector); +#else + // no vaddvq_u32 + uint32x2_t tmp = vpadd_u32(vget_low_u32(alphaVector), vget_high_u32(alphaVector)); + uint32_t alphaSum = vget_lane_u32(vpadd_u32(tmp, tmp), 0); +#endif + if (alphaSum) { + if (alphaSum != 255 * 4) { + if (RGBA) + srcVector = vrgba2argb(srcVector); + const float32x4_t a = vcvtq_f32_u32(alphaVector); + const float32x4_t ia = reciprocal_mul_ps(a, 255.0f); + // Convert 4x(4xU8) to 4x(4xF32) + uint16x8_t tmp1 = vmovl_u8(vget_low_u8(vreinterpretq_u8_u32(srcVector))); + uint16x8_t tmp3 = vmovl_u8(vget_high_u8(vreinterpretq_u8_u32(srcVector))); + float32x4_t src1 = vcvtq_f32_u32(vmovl_u16(vget_low_u16(tmp1))); + float32x4_t src2 = vcvtq_f32_u32(vmovl_u16(vget_high_u16(tmp1))); + float32x4_t src3 = vcvtq_f32_u32(vmovl_u16(vget_low_u16(tmp3))); + float32x4_t src4 = vcvtq_f32_u32(vmovl_u16(vget_high_u16(tmp3))); + src1 = vmulq_lane_f32(src1, vget_low_f32(ia), 0); + src2 = vmulq_lane_f32(src2, vget_low_f32(ia), 1); + src3 = vmulq_lane_f32(src3, vget_high_f32(ia), 0); + src4 = vmulq_lane_f32(src4, vget_high_f32(ia), 1); + // Convert 4x(4xF32) back to 4x(4xU8) (over a 8.1 fixed point format to get rounding) + tmp1 = vcombine_u16(vrshrn_n_u32(vcvtq_n_u32_f32(src1, 1), 1), + vrshrn_n_u32(vcvtq_n_u32_f32(src2, 1), 1)); + tmp3 = vcombine_u16(vrshrn_n_u32(vcvtq_n_u32_f32(src3, 1), 1), + vrshrn_n_u32(vcvtq_n_u32_f32(src4, 1), 1)); + uint32x4_t dstVector = vreinterpretq_u32_u8(vcombine_u8(vmovn_u16(tmp1), vmovn_u16(tmp3))); + // Overwrite any undefined results from alpha==0 with zeros: +#if defined(Q_PROCESSOR_ARM_64) + uint32x4_t srcVectorAlphaMask = vceqzq_u32(alphaVector); +#else + uint32x4_t srcVectorAlphaMask = vceqq_u32(alphaVector, vdupq_n_u32(0)); +#endif + dstVector = vbicq_u32(dstVector, srcVectorAlphaMask); + // Restore or mask alpha values: + if (RGBx) + dstVector = vorrq_u32(alphaMask, dstVector); + else + dstVector = vbslq_u32(alphaMask, srcVector, dstVector); + vst1q_u32(&buffer[i], dstVector); + } else { + // 4xAlpha==255, no change except if we are doing RGBA->ARGB: + if (RGBA) + vst1q_u32(&buffer[i], vrgba2argb(srcVector)); + else if (buffer != src) + vst1q_u32(&buffer[i], srcVector); + } + } else { + // 4xAlpha==0, always zero, except if output is RGBx: + if (RGBx) + vst1q_u32(&buffer[i], alphaMask); + else + vst1q_u32(&buffer[i], vdupq_n_u32(0)); + } + } + + SIMD_EPILOGUE(i, count, 3) { + uint v = qUnpremultiply(src[i]); + if (RGBx) + v = 0xff000000 | v; + if (RGBA) + v = ARGB2RGBA(v); + buffer[i] = v; + } +} + +void QT_FASTCALL convertARGB32ToARGB32PM_neon(uint *buffer, int count, const QVector<QRgb> *) +{ + convertARGBToARGB32PM_neon<false>(buffer, buffer, count); +} + +void QT_FASTCALL convertRGBA8888ToARGB32PM_neon(uint *buffer, int count, const QVector<QRgb> *) +{ + convertARGBToARGB32PM_neon<true>(buffer, buffer, count); +} + +const uint *QT_FASTCALL fetchARGB32ToARGB32PM_neon(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - convertARGBToARGB32PM_neon<false>(buffer, src, count); + convertARGBToARGB32PM_neon<false>(buffer, reinterpret_cast<const uint *>(src) + index, count); return buffer; } -const uint *QT_FASTCALL convertRGBA8888ToARGB32PM_neon(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +const uint *QT_FASTCALL fetchRGBA8888ToARGB32PM_neon(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { - convertARGBToARGB32PM_neon<true>(buffer, src, count); + convertARGBToARGB32PM_neon<true>(buffer, reinterpret_cast<const uint *>(src) + index, count); return buffer; } +void QT_FASTCALL storeRGB32FromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_neon<false,true>(d, src, count); +} + +void QT_FASTCALL storeARGB32FromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_neon<false,false>(d, src, count); +} + +void QT_FASTCALL storeRGBA8888FromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_neon<true,false>(d, src, count); +} + +void QT_FASTCALL storeRGBXFromARGB32PM_neon(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_neon<true,true>(d, src, count); +} + #endif // Q_BYTE_ORDER == Q_LITTLE_ENDIAN QT_END_NAMESPACE diff --git a/src/gui/painting/qdrawhelper_p.h b/src/gui/painting/qdrawhelper_p.h index 6f3c92ca64..fb08261205 100644 --- a/src/gui/painting/qdrawhelper_p.h +++ b/src/gui/painting/qdrawhelper_p.h @@ -747,6 +747,77 @@ static constexpr inline bool hasFastInterpolate4() { return false; } #endif +static inline QRgba64 multiplyAlpha256(QRgba64 rgba64, uint alpha256) +{ + return QRgba64::fromRgba64((rgba64.red() * alpha256) >> 8, + (rgba64.green() * alpha256) >> 8, + (rgba64.blue() * alpha256) >> 8, + (rgba64.alpha() * alpha256) >> 8); +} +static inline QRgba64 interpolate256(QRgba64 x, uint alpha1, QRgba64 y, uint alpha2) +{ + return QRgba64::fromRgba64(multiplyAlpha256(x, alpha1) + multiplyAlpha256(y, alpha2)); +} + +#ifdef __SSE2__ +static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty) +{ + __m128i vt = _mm_loadu_si128((const __m128i*)t); + if (disty) { + __m128i vb = _mm_loadu_si128((const __m128i*)b); + vt = _mm_mulhi_epu16(vt, _mm_set1_epi16(0x10000 - disty)); + vb = _mm_mulhi_epu16(vb, _mm_set1_epi16(disty)); + vt = _mm_add_epi16(vt, vb); + } + if (distx) { + const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0)); + const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(0x10000 - distx), _MM_SHUFFLE(0, 0, 0, 0)); + vt = _mm_mulhi_epu16(vt, _mm_unpacklo_epi64(vidistx, vdistx)); + vt = _mm_add_epi16(vt, _mm_srli_si128(vt, 8)); + } +#ifdef Q_PROCESSOR_X86_64 + return QRgba64::fromRgba64(_mm_cvtsi128_si64(vt)); +#else + QRgba64 out; + _mm_storel_epi64((__m128i*)&out, vt); + return out; +#endif // Q_PROCESSOR_X86_64 +} +#elif defined(__ARM_NEON__) +static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty) +{ + uint64x1x2_t vt = vld2_u64(reinterpret_cast<const uint64_t *>(t)); + if (disty) { + uint64x1x2_t vb = vld2_u64(reinterpret_cast<const uint64_t *>(b)); + uint32x4_t vt0 = vmull_n_u16(vreinterpret_u16_u64(vt.val[0]), 0x10000 - disty); + uint32x4_t vt1 = vmull_n_u16(vreinterpret_u16_u64(vt.val[1]), 0x10000 - disty); + vt0 = vmlal_n_u16(vt0, vreinterpret_u16_u64(vb.val[0]), disty); + vt1 = vmlal_n_u16(vt1, vreinterpret_u16_u64(vb.val[1]), disty); + vt.val[0] = vreinterpret_u64_u16(vshrn_n_u32(vt0, 16)); + vt.val[1] = vreinterpret_u64_u16(vshrn_n_u32(vt1, 16)); + } + if (distx) { + uint32x4_t vt0 = vmull_n_u16(vreinterpret_u16_u64(vt.val[0]), 0x10000 - distx); + vt0 = vmlal_n_u16(vt0, vreinterpret_u16_u64(vt.val[1]), distx); + vt.val[0] = vreinterpret_u64_u16(vshrn_n_u32(vt0, 16)); + } + QRgba64 out; + vst1_u64(reinterpret_cast<uint64_t *>(&out), vt.val[0]); + return out; +} +#else +static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty) +{ + const uint dx = distx>>8; + const uint dy = disty>>8; + const uint idx = 256 - dx; + const uint idy = 256 - dy; + QRgba64 xtop = interpolate256(t[0], idx, t[1], dx); + QRgba64 xbot = interpolate256(b[0], idx, b[1], dx); + return interpolate256(xtop, idy, xbot, dy); +} +#endif // __SSE2__ + #if Q_BYTE_ORDER == Q_BIG_ENDIAN static Q_ALWAYS_INLINE quint32 RGBA2ARGB(quint32 x) { quint32 rgb = x >> 8; @@ -798,6 +869,7 @@ static Q_ALWAYS_INLINE uint qAlphaRgb30(uint c) } struct quint24 { + quint24() = default; quint24(uint value); operator uint() const; uchar data[3]; @@ -1142,6 +1214,8 @@ static Q_ALWAYS_INLINE const uint *qt_convertRGBA8888ToARGB32PM(uint *buffer, co return buffer; } +template<bool RGBA> void qt_convertRGBA64ToARGB32(uint *dst, const QRgba64 *src, int count); + const uint qt_bayer_matrix[16][16] = { { 0x1, 0xc0, 0x30, 0xf0, 0xc, 0xcc, 0x3c, 0xfc, 0x3, 0xc3, 0x33, 0xf3, 0xf, 0xcf, 0x3f, 0xff}, @@ -1205,15 +1279,43 @@ inline uint comp_func_Plus_one_pixel(uint d, const uint s) #undef MIX #undef AMIX +// must be multiple of 4 for easier SIMD implementations +static Q_CONSTEXPR int BufferSize = 2048; + +// A buffer of intermediate results used by simple bilinear scaling. +struct IntermediateBuffer +{ + // The idea is first to do the interpolation between the row s1 and the row s2 + // into this intermediate buffer, then later interpolate between two pixel of this buffer. + // + // buffer_rb is a buffer of red-blue component of the pixel, in the form 0x00RR00BB + // buffer_ag is the alpha-green component of the pixel, in the form 0x00AA00GG + // +1 for the last pixel to interpolate with, and +1 for rounding errors. + quint32 buffer_rb[BufferSize+2]; + quint32 buffer_ag[BufferSize+2]; +}; + struct QDitherInfo { int x; int y; }; -typedef const uint *(QT_FASTCALL *ConvertFunc)(uint *buffer, const uint *src, int count, - const QVector<QRgb> *clut, QDitherInfo *dither); -typedef const QRgba64 *(QT_FASTCALL *ConvertFunc64)(QRgba64 *buffer, const uint *src, int count, - const QVector<QRgb> *clut, QDitherInfo *dither); +typedef const uint *(QT_FASTCALL *FetchAndConvertPixelsFunc)(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *dither); +typedef void (QT_FASTCALL *ConvertAndStorePixelsFunc)(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *dither); + +typedef const QRgba64 *(QT_FASTCALL *FetchAndConvertPixelsFunc64)(QRgba64 *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *dither); +typedef void (QT_FASTCALL *ConvertAndStorePixelsFunc64)(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *clut, QDitherInfo *dither); + +typedef void (QT_FASTCALL *ConvertFunc)(uint *buffer, int count, const QVector<QRgb> *clut); +typedef void (QT_FASTCALL *Convert64Func)(quint64 *buffer, int count, const QVector<QRgb> *clut); +typedef const QRgba64 *(QT_FASTCALL *ConvertTo64Func)(QRgba64 *buffer, const uint *src, int count, + const QVector<QRgb> *clut, QDitherInfo *dither); +typedef void (QT_FASTCALL *RbSwapFunc)(uchar *dst, const uchar *src, int count); + struct QPixelLayout { @@ -1226,36 +1328,28 @@ struct QPixelLayout BPP16, BPP24, BPP32, + BPP64, BPPCount }; - // All numbers in bits. - uchar redWidth; - uchar redShift; - uchar greenWidth; - uchar greenShift; - uchar blueWidth; - uchar blueShift; - uchar alphaWidth; - uchar alphaShift; + bool hasAlphaChannel; bool premultiplied; BPP bpp; + RbSwapFunc rbSwap; ConvertFunc convertToARGB32PM; - ConvertFunc convertFromARGB32PM; - ConvertFunc convertFromRGB32; - ConvertFunc64 convertToARGB64PM; + ConvertTo64Func convertToRGBA64PM; + FetchAndConvertPixelsFunc fetchToARGB32PM; + FetchAndConvertPixelsFunc64 fetchToRGBA64PM; + ConvertAndStorePixelsFunc storeFromARGB32PM; + ConvertAndStorePixelsFunc storeFromRGB32; }; -typedef const uint *(QT_FASTCALL *FetchPixelsFunc)(uint *buffer, const uchar *src, int index, int count); -typedef void (QT_FASTCALL *StorePixelsFunc)(uchar *dest, const uint *src, int index, int count); +extern ConvertAndStorePixelsFunc64 qStoreFromRGBA64PM[QImage::NImageFormats]; extern QPixelLayout qPixelLayouts[QImage::NImageFormats]; -extern const FetchPixelsFunc qFetchPixels[QPixelLayout::BPPCount]; -extern StorePixelsFunc qStorePixels[QPixelLayout::BPPCount]; extern MemRotateFunc qMemRotateFunctions[QPixelLayout::BPPCount][3]; - QT_END_NAMESPACE #endif // QDRAWHELPER_P_H diff --git a/src/gui/painting/qdrawhelper_sse4.cpp b/src/gui/painting/qdrawhelper_sse4.cpp index 14bfaabf09..e3cc1dd43e 100644 --- a/src/gui/painting/qdrawhelper_sse4.cpp +++ b/src/gui/painting/qdrawhelper_sse4.cpp @@ -39,13 +39,14 @@ #include <private/qdrawhelper_p.h> #include <private/qdrawingprimitive_sse2_p.h> +#include <private/qpaintengine_raster_p.h> #if defined(QT_COMPILER_SUPPORTS_SSE4_1) QT_BEGIN_NAMESPACE template<bool RGBA> -static inline void convertARGBToARGB32PM_sse4(uint *buffer, const uint *src, int count) +static void convertARGBToARGB32PM_sse4(uint *buffer, const uint *src, int count) { int i = 0; const __m128i alphaMask = _mm_set1_epi32(0xff000000); @@ -83,7 +84,7 @@ static inline void convertARGBToARGB32PM_sse4(uint *buffer, const uint *src, int _mm_storeu_si128((__m128i *)&buffer[i], srcVector); } } else { - _mm_storeu_si128((__m128i *)&buffer[i], _mm_setzero_si128()); + _mm_storeu_si128((__m128i *)&buffer[i], zero); } } @@ -93,59 +94,264 @@ static inline void convertARGBToARGB32PM_sse4(uint *buffer, const uint *src, int } } -const uint *QT_FASTCALL convertARGB32ToARGB32PM_sse4(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +static inline __m128 reciprocal_mul_ps(__m128 a, float mul) { - convertARGBToARGB32PM_sse4<false>(buffer, src, count); - return buffer; + __m128 ia = _mm_rcp_ps(a); // Approximate 1/a + // Improve precision of ia using Newton-Raphson + ia = _mm_sub_ps(_mm_add_ps(ia, ia), _mm_mul_ps(ia, _mm_mul_ps(ia, a))); + ia = _mm_mul_ps(ia, _mm_set1_ps(mul)); + return ia; } -const uint *QT_FASTCALL convertRGBA8888ToARGB32PM_sse4(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +template<bool RGBA, bool RGBx> +static inline void convertARGBFromARGB32PM_sse4(uint *buffer, const uint *src, int count) { - convertARGBToARGB32PM_sse4<true>(buffer, src, count); - return buffer; + int i = 0; + const __m128i alphaMask = _mm_set1_epi32(0xff000000); + const __m128i rgbaMask = _mm_setr_epi8(2, 1, 0, 3, 6, 5, 4, 7, 10, 9, 8, 11, 14, 13, 12, 15); + const __m128i zero = _mm_setzero_si128(); + + for (; i < count - 3; i += 4) { + __m128i srcVector = _mm_loadu_si128((const __m128i *)&src[i]); + if (!_mm_testz_si128(srcVector, alphaMask)) { + if (!_mm_testc_si128(srcVector, alphaMask)) { + __m128i srcVectorAlpha = _mm_srli_epi32(srcVector, 24); + if (RGBA) + srcVector = _mm_shuffle_epi8(srcVector, rgbaMask); + const __m128 a = _mm_cvtepi32_ps(srcVectorAlpha); + const __m128 ia = reciprocal_mul_ps(a, 255.0f); + __m128i src1 = _mm_unpacklo_epi8(srcVector, zero); + __m128i src3 = _mm_unpackhi_epi8(srcVector, zero); + __m128i src2 = _mm_unpackhi_epi16(src1, zero); + __m128i src4 = _mm_unpackhi_epi16(src3, zero); + src1 = _mm_unpacklo_epi16(src1, zero); + src3 = _mm_unpacklo_epi16(src3, zero); + __m128 ia1 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(0, 0, 0, 0)); + __m128 ia2 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(1, 1, 1, 1)); + __m128 ia3 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 ia4 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(3, 3, 3, 3)); + src1 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src1), ia1)); + src2 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src2), ia2)); + src3 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src3), ia3)); + src4 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src4), ia4)); + src1 = _mm_packus_epi32(src1, src2); + src3 = _mm_packus_epi32(src3, src4); + src1 = _mm_packus_epi16(src1, src3); + // Handle potential alpha == 0 values: + __m128i srcVectorAlphaMask = _mm_cmpeq_epi32(srcVectorAlpha, zero); + src1 = _mm_andnot_si128(srcVectorAlphaMask, src1); + // Fixup alpha values: + if (RGBx) + srcVector = _mm_or_si128(src1, alphaMask); + else + srcVector = _mm_blendv_epi8(src1, srcVector, alphaMask); + _mm_storeu_si128((__m128i *)&buffer[i], srcVector); + } else { + if (RGBA) + _mm_storeu_si128((__m128i *)&buffer[i], _mm_shuffle_epi8(srcVector, rgbaMask)); + else if (buffer != src) + _mm_storeu_si128((__m128i *)&buffer[i], srcVector); + } + } else { + if (RGBx) + _mm_storeu_si128((__m128i *)&buffer[i], alphaMask); + else + _mm_storeu_si128((__m128i *)&buffer[i], zero); + } + } + + SIMD_EPILOGUE(i, count, 3) { + uint v = qUnpremultiply_sse4(src[i]); + if (RGBx) + v = 0xff000000 | v; + if (RGBA) + v = ARGB2RGBA(v); + buffer[i] = v; + } } -const uint *QT_FASTCALL convertARGB32FromARGB32PM_sse4(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +template<bool RGBA> +static inline void convertARGBFromRGBA64PM_sse4(uint *buffer, const QRgba64 *src, int count) { - for (int i = 0; i < count; ++i) - buffer[i] = qUnpremultiply_sse4(src[i]); - return buffer; + int i = 0; + const __m128i alphaMask = _mm_set1_epi64x(Q_UINT64_C(0xffff) << 48); + const __m128i alphaMask32 = _mm_set1_epi32(0xff000000); + const __m128i rgbaMask = _mm_setr_epi8(2, 1, 0, 3, 6, 5, 4, 7, 10, 9, 8, 11, 14, 13, 12, 15); + const __m128i zero = _mm_setzero_si128(); + + for (; i < count - 3; i += 4) { + __m128i srcVector1 = _mm_loadu_si128((const __m128i *)&src[i]); + __m128i srcVector2 = _mm_loadu_si128((const __m128i *)&src[i + 2]); + bool transparent1 = _mm_testz_si128(srcVector1, alphaMask); + bool opaque1 = _mm_testc_si128(srcVector1, alphaMask); + bool transparent2 = _mm_testz_si128(srcVector2, alphaMask); + bool opaque2 = _mm_testc_si128(srcVector2, alphaMask); + + if (!(transparent1 && transparent2)) { + if (!(opaque1 && opaque2)) { + __m128i srcVector1Alpha = _mm_srli_epi64(srcVector1, 48); + __m128i srcVector2Alpha = _mm_srli_epi64(srcVector2, 48); + __m128i srcVectorAlpha = _mm_packus_epi32(srcVector1Alpha, srcVector2Alpha); + const __m128 a = _mm_cvtepi32_ps(srcVectorAlpha); + // Convert srcVectorAlpha to final 8-bit alpha channel + srcVectorAlpha = _mm_add_epi32(srcVectorAlpha, _mm_set1_epi32(128)); + srcVectorAlpha = _mm_sub_epi32(srcVectorAlpha, _mm_srli_epi32(srcVectorAlpha, 8)); + srcVectorAlpha = _mm_srli_epi32(srcVectorAlpha, 8); + srcVectorAlpha = _mm_slli_epi32(srcVectorAlpha, 24); + const __m128 ia = reciprocal_mul_ps(a, 255.0f); + __m128i src1 = _mm_unpacklo_epi16(srcVector1, zero); + __m128i src2 = _mm_unpackhi_epi16(srcVector1, zero); + __m128i src3 = _mm_unpacklo_epi16(srcVector2, zero); + __m128i src4 = _mm_unpackhi_epi16(srcVector2, zero); + __m128 ia1 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(0, 0, 0, 0)); + __m128 ia2 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(1, 1, 1, 1)); + __m128 ia3 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(2, 2, 2, 2)); + __m128 ia4 = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(3, 3, 3, 3)); + src1 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src1), ia1)); + src2 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src2), ia2)); + src3 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src3), ia3)); + src4 = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(src4), ia4)); + src1 = _mm_packus_epi32(src1, src2); + src3 = _mm_packus_epi32(src3, src4); + // Handle potential alpha == 0 values: + __m128i srcVector1AlphaMask = _mm_cmpeq_epi64(srcVector1Alpha, zero); + __m128i srcVector2AlphaMask = _mm_cmpeq_epi64(srcVector2Alpha, zero); + src1 = _mm_andnot_si128(srcVector1AlphaMask, src1); + src3 = _mm_andnot_si128(srcVector2AlphaMask, src3); + src1 = _mm_packus_epi16(src1, src3); + // Fixup alpha values: + src1 = _mm_blendv_epi8(src1, srcVectorAlpha, alphaMask32); + // Fix RGB order + if (!RGBA) + src1 = _mm_shuffle_epi8(src1, rgbaMask); + _mm_storeu_si128((__m128i *)&buffer[i], src1); + } else { + __m128i src1 = _mm_unpacklo_epi16(srcVector1, zero); + __m128i src2 = _mm_unpackhi_epi16(srcVector1, zero); + __m128i src3 = _mm_unpacklo_epi16(srcVector2, zero); + __m128i src4 = _mm_unpackhi_epi16(srcVector2, zero); + src1 = _mm_add_epi32(src1, _mm_set1_epi32(128)); + src2 = _mm_add_epi32(src2, _mm_set1_epi32(128)); + src3 = _mm_add_epi32(src3, _mm_set1_epi32(128)); + src4 = _mm_add_epi32(src4, _mm_set1_epi32(128)); + src1 = _mm_sub_epi32(src1, _mm_srli_epi32(src1, 8)); + src2 = _mm_sub_epi32(src2, _mm_srli_epi32(src2, 8)); + src3 = _mm_sub_epi32(src3, _mm_srli_epi32(src3, 8)); + src4 = _mm_sub_epi32(src4, _mm_srli_epi32(src4, 8)); + src1 = _mm_srli_epi32(src1, 8); + src2 = _mm_srli_epi32(src2, 8); + src3 = _mm_srli_epi32(src3, 8); + src4 = _mm_srli_epi32(src4, 8); + src1 = _mm_packus_epi32(src1, src2); + src3 = _mm_packus_epi32(src3, src4); + src1 = _mm_packus_epi16(src1, src3); + if (!RGBA) + src1 = _mm_shuffle_epi8(src1, rgbaMask); + _mm_storeu_si128((__m128i *)&buffer[i], src1); + } + } else { + _mm_storeu_si128((__m128i *)&buffer[i], zero); + } + } + + SIMD_EPILOGUE(i, count, 3) { + buffer[i] = qConvertRgba64ToRgb32_sse4<RGBA ? PixelOrderRGB : PixelOrderBGR>(src[i]); + } } -const uint *QT_FASTCALL convertRGBA8888FromARGB32PM_sse4(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +void QT_FASTCALL convertARGB32ToARGB32PM_sse4(uint *buffer, int count, const QVector<QRgb> *) { - for (int i = 0; i < count; ++i) - buffer[i] = ARGB2RGBA(qUnpremultiply_sse4(src[i])); + convertARGBToARGB32PM_sse4<false>(buffer, buffer, count); +} + +void QT_FASTCALL convertRGBA8888ToARGB32PM_sse4(uint *buffer, int count, const QVector<QRgb> *) +{ + convertARGBToARGB32PM_sse4<true>(buffer, buffer, count); +} + +const uint *QT_FASTCALL fetchARGB32ToARGB32PM_sse4(uint *buffer, const uchar *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + convertARGBToARGB32PM_sse4<false>(buffer, reinterpret_cast<const uint *>(src) + index, count); return buffer; } -const uint *QT_FASTCALL convertRGBXFromARGB32PM_sse4(uint *buffer, const uint *src, int count, +const uint *QT_FASTCALL fetchRGBA8888ToARGB32PM_sse4(uint *buffer, const uchar *src, int index, int count, const QVector<QRgb> *, QDitherInfo *) { - for (int i = 0; i < count; ++i) - buffer[i] = ARGB2RGBA(0xff000000 | qUnpremultiply_sse4(src[i])); + convertARGBToARGB32PM_sse4<true>(buffer, reinterpret_cast<const uint *>(src) + index, count); return buffer; } +void QT_FASTCALL storeRGB32FromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_sse4<false,true>(d, src, count); +} + +void QT_FASTCALL storeARGB32FromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_sse4<false,false>(d, src, count); +} + +void QT_FASTCALL storeRGBA8888FromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_sse4<true,false>(d, src, count); +} + +void QT_FASTCALL storeRGBXFromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = reinterpret_cast<uint *>(dest) + index; + convertARGBFromARGB32PM_sse4<true,true>(d, src, count); +} + template<QtPixelOrder PixelOrder> -const uint *QT_FASTCALL convertA2RGB30PMFromARGB32PM_sse4(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *) +void QT_FASTCALL storeA2RGB30PMFromARGB32PM_sse4(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) { + uint *d = reinterpret_cast<uint *>(dest) + index; for (int i = 0; i < count; ++i) - buffer[i] = qConvertArgb32ToA2rgb30_sse4<PixelOrder>(src[i]); - return buffer; + d[i] = qConvertArgb32ToA2rgb30_sse4<PixelOrder>(src[i]); +} + +void QT_FASTCALL destStore64ARGB32_sse4(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) +{ + uint *dest = (uint*)rasterBuffer->scanLine(y) + x; + convertARGBFromRGBA64PM_sse4<false>(dest, buffer, length); +} + +void QT_FASTCALL destStore64RGBA8888_sse4(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length) +{ + uint *dest = (uint*)rasterBuffer->scanLine(y) + x; + convertARGBFromRGBA64PM_sse4<true>(dest, buffer, length); +} + +void QT_FASTCALL storeARGB32FromRGBA64PM_sse4(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = (uint*)dest + index; + convertARGBFromRGBA64PM_sse4<false>(d, src, count); +} + +void QT_FASTCALL storeRGBA8888FromRGBA64PM_sse4(uchar *dest, const QRgba64 *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *) +{ + uint *d = (uint*)dest + index; + convertARGBFromRGBA64PM_sse4<true>(d, src, count); } template -const uint *QT_FASTCALL convertA2RGB30PMFromARGB32PM_sse4<PixelOrderBGR>(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *); +void QT_FASTCALL storeA2RGB30PMFromARGB32PM_sse4<PixelOrderBGR>(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); template -const uint *QT_FASTCALL convertA2RGB30PMFromARGB32PM_sse4<PixelOrderRGB>(uint *buffer, const uint *src, int count, - const QVector<QRgb> *, QDitherInfo *); +void QT_FASTCALL storeA2RGB30PMFromARGB32PM_sse4<PixelOrderRGB>(uchar *dest, const uint *src, int index, int count, + const QVector<QRgb> *, QDitherInfo *); QT_END_NAMESPACE diff --git a/src/gui/painting/qdrawhelper_ssse3.cpp b/src/gui/painting/qdrawhelper_ssse3.cpp index 45ecc8b422..42d760d5cc 100644 --- a/src/gui/painting/qdrawhelper_ssse3.cpp +++ b/src/gui/painting/qdrawhelper_ssse3.cpp @@ -167,61 +167,12 @@ void qt_blend_argb32_on_argb32_ssse3(uchar *destPixels, int dbpl, } } -static inline void store_uint24_ssse3(uchar *dst, const uint *src, int len) +const uint *QT_FASTCALL fetchPixelsBPP24_ssse3(uint *buffer, const uchar *src, int index, int count) { - int i = 0; - - quint24 *dst24 = reinterpret_cast<quint24*>(dst); - // Align dst on 16 bytes - for (; i < len && (reinterpret_cast<quintptr>(dst24) & 0xf); ++i) - *dst24++ = quint24(*src++); - - // Shuffle masks for first and second half of every output, all outputs are aligned so the shuffled ends are not used. - const __m128i shuffleMask1 = _mm_setr_epi8(char(0x80), char(0x80), char(0x80), char(0x80), 2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12); - const __m128i shuffleMask2 = _mm_setr_epi8(2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12, char(0x80), char(0x80), char(0x80), char(0x80)); - - const __m128i *inVectorPtr = (const __m128i *)src; - __m128i *dstVectorPtr = (__m128i *)dst24; - - for (; i < (len - 15); i += 16) { - // Load four vectors, store three. - // Create each output vector by combining two shuffled input vectors. - __m128i srcVector1 = _mm_loadu_si128(inVectorPtr); - ++inVectorPtr; - __m128i srcVector2 = _mm_loadu_si128(inVectorPtr); - ++inVectorPtr; - __m128i outputVector1 = _mm_shuffle_epi8(srcVector1, shuffleMask1); - __m128i outputVector2 = _mm_shuffle_epi8(srcVector2, shuffleMask2); - __m128i outputVector = _mm_alignr_epi8(outputVector2, outputVector1, 4); - _mm_store_si128(dstVectorPtr, outputVector); - ++dstVectorPtr; - - srcVector1 = _mm_loadu_si128(inVectorPtr); - ++inVectorPtr; - outputVector1 = _mm_shuffle_epi8(srcVector2, shuffleMask1); - outputVector2 = _mm_shuffle_epi8(srcVector1, shuffleMask2); - outputVector = _mm_alignr_epi8(outputVector2, outputVector1, 8); - _mm_store_si128(dstVectorPtr, outputVector); - ++dstVectorPtr; - - srcVector2 = _mm_loadu_si128(inVectorPtr); - ++inVectorPtr; - outputVector1 = _mm_shuffle_epi8(srcVector1, shuffleMask1); - outputVector2 = _mm_shuffle_epi8(srcVector2, shuffleMask2); - outputVector = _mm_alignr_epi8(outputVector2, outputVector1, 12); - _mm_store_si128(dstVectorPtr, outputVector); - ++dstVectorPtr; - } - dst24 = reinterpret_cast<quint24*>(dstVectorPtr); - src = reinterpret_cast<const uint*>(inVectorPtr); - - SIMD_EPILOGUE(i, len, 15) - *dst24++ = quint24(*src++); -} - -void QT_FASTCALL storePixelsBPP24_ssse3(uchar *dest, const uint *src, int index, int count) -{ - store_uint24_ssse3(dest + index * 3, src, count); + const quint24 *s = reinterpret_cast<const quint24 *>(src); + for (int i = 0; i < count; ++i) + buffer[i] = s[index + i]; + return buffer; } extern void QT_FASTCALL qt_convert_rgb888_to_rgb32_ssse3(quint32 *dst, const uchar *src, int len); diff --git a/src/gui/painting/qdrawingprimitive_sse2_p.h b/src/gui/painting/qdrawingprimitive_sse2_p.h index 93e4b9f572..b237ea1611 100644 --- a/src/gui/painting/qdrawingprimitive_sse2_p.h +++ b/src/gui/painting/qdrawingprimitive_sse2_p.h @@ -43,6 +43,7 @@ #include <QtGui/private/qtguiglobal_p.h> #include <private/qsimd_p.h> #include "qdrawhelper_p.h" +#include "qrgba64_p.h" #ifdef __SSE2__ @@ -230,21 +231,31 @@ QT_END_NAMESPACE QT_BEGIN_NAMESPACE #if QT_COMPILER_SUPPORTS_HERE(SSE4_1) +QT_FUNCTION_TARGET(SSE2) +Q_ALWAYS_INLINE void reciprocal_mul_ss(__m128 &ia, const __m128 a, float mul) +{ + ia = _mm_rcp_ss(a); // Approximate 1/a + // Improve precision of ia using Newton-Raphson + ia = _mm_sub_ss(_mm_add_ss(ia, ia), _mm_mul_ss(ia, _mm_mul_ss(ia, a))); + ia = _mm_mul_ss(ia, _mm_set_ss(mul)); + ia = _mm_shuffle_ps(ia, ia, _MM_SHUFFLE(0,0,0,0)); +} + QT_FUNCTION_TARGET(SSE4_1) inline QRgb qUnpremultiply_sse4(QRgb p) { const uint alpha = qAlpha(p); - if (alpha == 255 || alpha == 0) + if (alpha == 255) return p; - const uint invAlpha = qt_inv_premul_factor[alpha]; - const __m128i via = _mm_set1_epi32(invAlpha); - const __m128i vr = _mm_set1_epi32(0x8000); + if (alpha == 0) + return 0; + const __m128 va = _mm_set1_ps(alpha); + __m128 via; + reciprocal_mul_ss(via, va, 255.0f); // Approximate 1/a __m128i vl = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(p)); - vl = _mm_mullo_epi32(vl, via); - vl = _mm_add_epi32(vl, vr); - vl = _mm_srai_epi32(vl, 16); - vl = _mm_insert_epi32(vl, alpha, 3); + vl = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(vl), via)); vl = _mm_packus_epi32(vl, vl); + vl = _mm_insert_epi16(vl, alpha, 3); vl = _mm_packus_epi16(vl, vl); return _mm_cvtsi128_si32(vl); } @@ -258,21 +269,14 @@ inline uint qConvertArgb32ToA2rgb30_sse4(QRgb p) return qConvertRgb32ToRgb30<PixelOrder>(p); if (alpha == 0) return 0; - Q_CONSTEXPR uint mult = 255 / (255 >> 6); - const uint invAlpha = qt_inv_premul_factor[alpha]; + Q_CONSTEXPR float mult = 1023.0f / (255 >> 6); const uint newalpha = (alpha >> 6); - const __m128i via = _mm_set1_epi32(invAlpha); - const __m128i vna = _mm_set1_epi32(mult * newalpha); - const __m128i vr1 = _mm_set1_epi32(0x1000); - const __m128i vr2 = _mm_set1_epi32(0x80); - __m128i vl = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(p)); - vl = _mm_mullo_epi32(vl, via); - vl = _mm_add_epi32(vl, vr1); - vl = _mm_srli_epi32(vl, 14); - vl = _mm_mullo_epi32(vl, vna); - vl = _mm_add_epi32(vl, _mm_srli_epi32(vl, 8)); - vl = _mm_add_epi32(vl, vr2); - vl = _mm_srli_epi32(vl, 8); + const __m128 va = _mm_set1_ps(alpha); + __m128 via; + reciprocal_mul_ss(via, va, mult * newalpha); + __m128i vl = _mm_cvtsi32_si128(p); + vl = _mm_cvtepu8_epi32(vl); + vl = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(vl), via)); vl = _mm_packus_epi32(vl, vl); uint rgb30 = (newalpha << 30); rgb30 |= ((uint)_mm_extract_epi16(vl, 1)) << 10; @@ -285,6 +289,27 @@ inline uint qConvertArgb32ToA2rgb30_sse4(QRgb p) } return rgb30; } + +template<enum QtPixelOrder PixelOrder> +QT_FUNCTION_TARGET(SSE4_1) +inline uint qConvertRgba64ToRgb32_sse4(QRgba64 p) +{ + if (p.isTransparent()) + return 0; + __m128i vl = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(&p)); + if (!p.isOpaque()) { + const __m128 va = _mm_set1_ps(p.alpha()); + __m128 via; + reciprocal_mul_ss(via, va, 65535.0f); + vl = _mm_unpacklo_epi16(vl, _mm_setzero_si128()); + vl = _mm_cvtps_epi32(_mm_mul_ps(_mm_cvtepi32_ps(vl) , via)); + vl = _mm_packus_epi32(vl, vl); + vl = _mm_insert_epi16(vl, p.alpha(), 3); + } + if (PixelOrder == PixelOrderBGR) + vl = _mm_shufflelo_epi16(vl, _MM_SHUFFLE(3, 0, 1, 2)); + return toArgb32(vl); +} #endif QT_END_NAMESPACE diff --git a/src/gui/painting/qemulationpaintengine.cpp b/src/gui/painting/qemulationpaintengine.cpp index e6686e3721..0c0df0fb13 100644 --- a/src/gui/painting/qemulationpaintengine.cpp +++ b/src/gui/painting/qemulationpaintengine.cpp @@ -74,10 +74,11 @@ QPainterState *QEmulationPaintEngine::createState(QPainterState *orig) const static inline void combineXForm(QBrush *brush, const QRectF &r) { - QTransform t = brush->transform(); - t.translate(r.x(), r.y()); - t.scale(r.width(), r.height()); - brush->setTransform(t); + QTransform t(r.width(), 0, 0, r.height(), r.x(), r.y()); + if (brush->gradient() && brush->gradient()->coordinateMode() != QGradient::ObjectMode) + brush->setTransform(t * brush->transform()); // compat mode + else + brush->setTransform(brush->transform() * t); } void QEmulationPaintEngine::fill(const QVectorPath &path, const QBrush &brush) @@ -96,11 +97,19 @@ void QEmulationPaintEngine::fill(const QVectorPath &path, const QBrush &brush) if (coMode > QGradient::LogicalMode) { QBrush copy = brush; const QPaintDevice *d = real_engine->painter()->device(); - QRectF r = (coMode == QGradient::ObjectBoundingMode) ? path.controlPointRect() : QRectF(0, 0, d->width(), d->height()); + QRectF r = (coMode == QGradient::StretchToDeviceMode) ? QRectF(0, 0, d->width(), d->height()) : path.controlPointRect(); combineXForm(©, r); real_engine->fill(path, copy); return; } + } else if (style == Qt::TexturePattern) { + qreal dpr = qHasPixmapTexture(brush) ? brush.texture().devicePixelRatioF() : brush.textureImage().devicePixelRatioF(); + if (!qFuzzyCompare(dpr, 1.0)) { + QBrush copy = brush; + combineXForm(©, QRectF(0, 0, 1.0/dpr, 1.0/dpr)); + real_engine->fill(path, copy); + return; + } } real_engine->fill(path, brush); @@ -124,7 +133,7 @@ void QEmulationPaintEngine::stroke(const QVectorPath &path, const QPen &pen) QGradient::CoordinateMode coMode = brush.gradient()->coordinateMode(); if (coMode > QGradient::LogicalMode) { const QPaintDevice *d = real_engine->painter()->device(); - QRectF r = (coMode == QGradient::ObjectBoundingMode) ? path.controlPointRect() : QRectF(0, 0, d->width(), d->height()); + QRectF r = (coMode == QGradient::StretchToDeviceMode) ? QRectF(0, 0, d->width(), d->height()) : path.controlPointRect(); combineXForm(&brush, r); copy.setBrush(brush); real_engine->stroke(path, copy); @@ -166,9 +175,9 @@ void QEmulationPaintEngine::drawTextItem(const QPointF &p, const QTextItem &text QBrush copy = s->pen.brush(); const QPaintDevice *d = real_engine->painter()->device(); const QTextItemInt &ti = static_cast<const QTextItemInt &>(textItem); - QRectF r = (g.coordinateMode() == QGradient::ObjectBoundingMode) ? - QRectF(p.x(), p.y() - ti.ascent.toReal(), ti.width.toReal(), (ti.ascent + ti.descent + 1).toReal()) : - QRectF(0, 0, d->width(), d->height()); + QRectF r = (g.coordinateMode() == QGradient::StretchToDeviceMode) ? + QRectF(0, 0, d->width(), d->height()) : + QRectF(p.x(), p.y() - ti.ascent.toReal(), ti.width.toReal(), (ti.ascent + ti.descent + 1).toReal()); combineXForm(©, r); g.setCoordinateMode(QGradient::LogicalMode); QBrush brush(g); diff --git a/src/gui/painting/qimagescale.cpp b/src/gui/painting/qimagescale.cpp index 22787b91fe..8a5274bd37 100644 --- a/src/gui/painting/qimagescale.cpp +++ b/src/gui/painting/qimagescale.cpp @@ -41,6 +41,7 @@ #include "qimage.h" #include "qcolor.h" +#include "qrgba64_p.h" QT_BEGIN_NAMESPACE @@ -85,7 +86,7 @@ QT_BEGIN_NAMESPACE * #ifdef'ed code, and removal of unneeded border calculation code. * Later the code has been refactored, an SSE4.1 optimizated path have been * added instead of the removed MMX assembler, and scaling of clipped area - * removed. + * removed, and an RGBA64 version written * * Imlib2 is (C) Carsten Haitzler and various contributors. The MMX code * is by Willem Monsuwe <willem@stack.nl>. All other modifications are @@ -94,12 +95,11 @@ QT_BEGIN_NAMESPACE namespace QImageScale { - const unsigned int** qimageCalcYPoints(const unsigned int *src, int sw, int sh, int dh); - int* qimageCalcXPoints(int sw, int dw); - int* qimageCalcApoints(int s, int d, int up); - QImageScaleInfo* qimageFreeScaleInfo(QImageScaleInfo *isi); - QImageScaleInfo *qimageCalcScaleInfo(const QImage &img, int sw, int sh, - int dw, int dh, char aa); + static const unsigned int** qimageCalcYPoints(const unsigned int *src, int sw, int sh, int dh); + static int* qimageCalcXPoints(int sw, int dw); + static int* qimageCalcApoints(int s, int d, int up); + static QImageScaleInfo* qimageFreeScaleInfo(QImageScaleInfo *isi); + static QImageScaleInfo *qimageCalcScaleInfo(const QImage &img, int sw, int sh, int dw, int dh, char aa); } using namespace QImageScale; @@ -108,8 +108,8 @@ using namespace QImageScale; // Code ported from Imlib... // -const unsigned int** QImageScale::qimageCalcYPoints(const unsigned int *src, - int sw, int sh, int dh) +static const unsigned int** QImageScale::qimageCalcYPoints(const unsigned int *src, + int sw, int sh, int dh) { const unsigned int **p; int j = 0, rv = 0; @@ -138,7 +138,7 @@ const unsigned int** QImageScale::qimageCalcYPoints(const unsigned int *src, return(p); } -int* QImageScale::qimageCalcXPoints(int sw, int dw) +static int* QImageScale::qimageCalcXPoints(int sw, int dw) { int *p, j = 0, rv = 0; qint64 val, inc; @@ -167,7 +167,7 @@ int* QImageScale::qimageCalcXPoints(int sw, int dw) return p; } -int* QImageScale::qimageCalcApoints(int s, int d, int up) +static int* QImageScale::qimageCalcApoints(int s, int d, int up) { int *p, j = 0, rv = 0; @@ -214,7 +214,7 @@ int* QImageScale::qimageCalcApoints(int s, int d, int up) return p; } -QImageScaleInfo* QImageScale::qimageFreeScaleInfo(QImageScaleInfo *isi) +static QImageScaleInfo* QImageScale::qimageFreeScaleInfo(QImageScaleInfo *isi) { if (isi) { delete[] isi->xpoints; @@ -226,9 +226,9 @@ QImageScaleInfo* QImageScale::qimageFreeScaleInfo(QImageScaleInfo *isi) return 0; } -QImageScaleInfo* QImageScale::qimageCalcScaleInfo(const QImage &img, - int sw, int sh, - int dw, int dh, char aa) +static QImageScaleInfo* QImageScale::qimageCalcScaleInfo(const QImage &img, + int sw, int sh, + int dw, int dh, char aa) { QImageScaleInfo *isi; int scw, sch; @@ -239,7 +239,6 @@ QImageScaleInfo* QImageScale::qimageCalcScaleInfo(const QImage &img, isi = new QImageScaleInfo; if (!isi) return 0; - memset(isi, 0, sizeof(QImageScaleInfo)); isi->xup_yup = (qAbs(dw) >= sw) + ((qAbs(dh) >= sh) << 1); @@ -333,7 +332,7 @@ static void qt_qimageScaleAARGBA_up_xy(QImageScaleInfo *isi, unsigned int *dest, } } -/* scale by area sampling */ +/* scale by area sampling - with alpha */ static void qt_qimageScaleAARGBA(QImageScaleInfo *isi, unsigned int *dest, int dw, int dh, int dow, int sow) { @@ -529,6 +528,207 @@ static void qt_qimageScaleAARGBA_down_xy(QImageScaleInfo *isi, unsigned int *des } } +static void qt_qimageScaleRgba64_up_x_down_y(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow); + +static void qt_qimageScaleRgba64_down_x_up_y(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow); + +static void qt_qimageScaleRgba64_down_xy(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow); + +static void qt_qimageScaleRgba64_up_xy(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow) +{ + const QRgba64 **ypoints = (const QRgba64 **)isi->ypoints; + int *xpoints = isi->xpoints; + int *xapoints = isi->xapoints; + int *yapoints = isi->yapoints; + + for (int y = 0; y < dh; y++) { + const QRgba64 *sptr = ypoints[y]; + QRgba64 *dptr = dest + (y * dow); + const int yap = yapoints[y]; + if (yap > 0) { + for (int x = 0; x < dw; x++) { + const QRgba64 *pix = sptr + xpoints[x]; + const int xap = xapoints[x]; + if (xap > 0) + *dptr = interpolate_4_pixels_rgb64(pix, pix + sow, xap * 256, yap * 256); + else + *dptr = interpolate256(pix[0], 256 - yap, pix[sow], yap); + dptr++; + } + } else { + for (int x = 0; x < dw; x++) { + const QRgba64 *pix = sptr + xpoints[x]; + const int xap = xapoints[x]; + if (xap > 0) + *dptr = interpolate256(pix[0], 256 - xap, pix[1], xap); + else + *dptr = pix[0]; + dptr++; + } + } + } +} + +void qt_qimageScaleRgba64(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow) +{ + if (isi->xup_yup == 3) + qt_qimageScaleRgba64_up_xy(isi, dest, dw, dh, dow, sow); + else if (isi->xup_yup == 1) + qt_qimageScaleRgba64_up_x_down_y(isi, dest, dw, dh, dow, sow); + else if (isi->xup_yup == 2) + qt_qimageScaleRgba64_down_x_up_y(isi, dest, dw, dh, dow, sow); + else + qt_qimageScaleRgba64_down_xy(isi, dest, dw, dh, dow, sow); +} + +inline static void qt_qimageScaleRgba64_helper(const QRgba64 *pix, int xyap, int Cxy, int step, qint64 &r, qint64 &g, qint64 &b, qint64 &a) +{ + r = pix->red() * xyap; + g = pix->green() * xyap; + b = pix->blue() * xyap; + a = pix->alpha() * xyap; + int j; + for (j = (1 << 14) - xyap; j > Cxy; j -= Cxy ){ + pix += step; + r += pix->red() * Cxy; + g += pix->green() * Cxy; + b += pix->blue() * Cxy; + a += pix->alpha() * Cxy; + } + pix += step; + r += pix->red() * j; + g += pix->green() * j; + b += pix->blue() * j; + a += pix->alpha() * j; +} + +static void qt_qimageScaleRgba64_up_x_down_y(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow) +{ + const QRgba64 **ypoints = (const QRgba64 **)isi->ypoints; + int *xpoints = isi->xpoints; + int *xapoints = isi->xapoints; + int *yapoints = isi->yapoints; + + for (int y = 0; y < dh; y++) { + int Cy = (yapoints[y]) >> 16; + int yap = (yapoints[y]) & 0xffff; + + QRgba64 *dptr = dest + (y * dow); + for (int x = 0; x < dw; x++) { + const QRgba64 *sptr = ypoints[y] + xpoints[x]; + qint64 r, g, b, a; + qt_qimageScaleRgba64_helper(sptr, yap, Cy, sow, r, g, b, a); + + int xap = xapoints[x]; + if (xap > 0) { + qint64 rr, gg, bb, aa; + qt_qimageScaleRgba64_helper(sptr + 1, yap, Cy, sow, rr, gg, bb, aa); + + r = r * (256 - xap); + g = g * (256 - xap); + b = b * (256 - xap); + a = a * (256 - xap); + r = (r + (rr * xap)) >> 8; + g = (g + (gg * xap)) >> 8; + b = (b + (bb * xap)) >> 8; + a = (a + (aa * xap)) >> 8; + } + *dptr++ = qRgba64(r >> 14, g >> 14, b >> 14, a >> 14); + } + } +} + +static void qt_qimageScaleRgba64_down_x_up_y(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow) +{ + const QRgba64 **ypoints = (const QRgba64 **)isi->ypoints; + int *xpoints = isi->xpoints; + int *xapoints = isi->xapoints; + int *yapoints = isi->yapoints; + + for (int y = 0; y < dh; y++) { + QRgba64 *dptr = dest + (y * dow); + for (int x = 0; x < dw; x++) { + int Cx = xapoints[x] >> 16; + int xap = xapoints[x] & 0xffff; + + const QRgba64 *sptr = ypoints[y] + xpoints[x]; + qint64 r, g, b, a; + qt_qimageScaleRgba64_helper(sptr, xap, Cx, 1, r, g, b, a); + + int yap = yapoints[y]; + if (yap > 0) { + qint64 rr, gg, bb, aa; + qt_qimageScaleRgba64_helper(sptr + sow, xap, Cx, 1, rr, gg, bb, aa); + + r = r * (256 - yap); + g = g * (256 - yap); + b = b * (256 - yap); + a = a * (256 - yap); + r = (r + (rr * yap)) >> 8; + g = (g + (gg * yap)) >> 8; + b = (b + (bb * yap)) >> 8; + a = (a + (aa * yap)) >> 8; + } + *dptr = qRgba64(r >> 14, g >> 14, b >> 14, a >> 14); + dptr++; + } + } +} + +static void qt_qimageScaleRgba64_down_xy(QImageScaleInfo *isi, QRgba64 *dest, + int dw, int dh, int dow, int sow) +{ + const QRgba64 **ypoints = (const QRgba64 **)isi->ypoints; + int *xpoints = isi->xpoints; + int *xapoints = isi->xapoints; + int *yapoints = isi->yapoints; + + for (int y = 0; y < dh; y++) { + int Cy = (yapoints[y]) >> 16; + int yap = (yapoints[y]) & 0xffff; + + QRgba64 *dptr = dest + (y * dow); + for (int x = 0; x < dw; x++) { + int Cx = xapoints[x] >> 16; + int xap = xapoints[x] & 0xffff; + + const QRgba64 *sptr = ypoints[y] + xpoints[x]; + qint64 rx, gx, bx, ax; + qt_qimageScaleRgba64_helper(sptr, xap, Cx, 1, rx, gx, bx, ax); + + qint64 r = rx * yap; + qint64 g = gx * yap; + qint64 b = bx * yap; + qint64 a = ax * yap; + int j; + for (j = (1 << 14) - yap; j > Cy; j -= Cy) { + sptr += sow; + qt_qimageScaleRgba64_helper(sptr, xap, Cx, 1, rx, gx, bx, ax); + r += rx * Cy; + g += gx * Cy; + b += bx * Cy; + a += ax * Cy; + } + sptr += sow; + qt_qimageScaleRgba64_helper(sptr, xap, Cx, 1, rx, gx, bx, ax); + r += rx * j; + g += gx * j; + b += bx * j; + a += ax * j; + + *dptr = qRgba64(r >> 28, g >> 28, b >> 28, a >> 28); + dptr++; + } + } +} + static void qt_qimageScaleAARGB_up_x_down_y(QImageScaleInfo *isi, unsigned int *dest, int dw, int dh, int dow, int sow); @@ -745,7 +945,10 @@ QImage qSmoothScaleImage(const QImage &src, int dw, int dh) return QImage(); } - if (src.hasAlphaChannel()) + if (src.depth() > 32) + qt_qimageScaleRgba64(scaleinfo, (QRgba64 *)buffer.scanLine(0), + dw, dh, dw, src.bytesPerLine() / 8); + else if (src.hasAlphaChannel()) qt_qimageScaleAARGBA(scaleinfo, (unsigned int *)buffer.scanLine(0), dw, dh, dw, src.bytesPerLine() / 4); else diff --git a/src/gui/painting/qimagescale_p.h b/src/gui/painting/qimagescale_p.h index 415623a575..244d681718 100644 --- a/src/gui/painting/qimagescale_p.h +++ b/src/gui/painting/qimagescale_p.h @@ -61,10 +61,11 @@ QImage qSmoothScaleImage(const QImage &img, int w, int h); namespace QImageScale { struct QImageScaleInfo { - int *xpoints; - const unsigned int **ypoints; - int *xapoints, *yapoints; - int xup_yup; + int *xpoints{nullptr}; + const unsigned int **ypoints{nullptr}; + int *xapoints{nullptr}; + int *yapoints{nullptr}; + int xup_yup{0}; }; } diff --git a/src/gui/painting/qmemrotate.cpp b/src/gui/painting/qmemrotate.cpp index 25aa6a3122..43aeff3268 100644 --- a/src/gui/painting/qmemrotate.cpp +++ b/src/gui/painting/qmemrotate.cpp @@ -239,6 +239,12 @@ inline void qt_memrotate90_template<quint32>(const quint32 *src, int w, int h, i qt_memrotate90_tiled_unpacked(src, w, h, sstride, dest, dstride); } +template <> +inline void qt_memrotate90_template<quint64>(const quint64 *src, int w, int h, int sstride, quint64 *dest, int dstride) +{ + qt_memrotate90_tiled_unpacked(src, w, h, sstride, dest, dstride); +} + template <class T> Q_STATIC_TEMPLATE_FUNCTION inline void qt_memrotate180_template(const T *src, int w, int h, int sstride, T *dest, int dstride) @@ -275,6 +281,12 @@ inline void qt_memrotate270_template<quint32>(const quint32 *src, int w, int h, qt_memrotate270_tiled_unpacked(src, w, h, sstride, dest, dstride); } +template <> +inline void qt_memrotate270_template<quint64>(const quint64 *src, int w, int h, int sstride, quint64 *dest, int dstride) +{ + qt_memrotate270_tiled_unpacked(src, w, h, sstride, dest, dstride); +} + #define QT_IMPL_MEMROTATE(type) \ Q_GUI_EXPORT void qt_memrotate90(const type *src, int w, int h, int sstride, \ type *dest, int dstride) \ @@ -309,9 +321,7 @@ Q_GUI_EXPORT void qt_memrotate270(const type *src, int w, int h, int sstride, \ qt_memrotate270_tiled_unpacked(src, w, h, sstride, dest, dstride); \ } - - - +QT_IMPL_MEMROTATE(quint64) QT_IMPL_MEMROTATE(quint32) QT_IMPL_MEMROTATE(quint16) QT_IMPL_MEMROTATE(quint24) @@ -377,6 +387,22 @@ void qt_memrotate270_32(const uchar *srcPixels, int w, int h, int sbpl, uchar *d qt_memrotate270((const uint *)srcPixels, w, h, sbpl, (uint *)destPixels, dbpl); } + +void qt_memrotate90_64(const uchar *srcPixels, int w, int h, int sbpl, uchar *destPixels, int dbpl) +{ + qt_memrotate90((const quint64 *)srcPixels, w, h, sbpl, (quint64 *)destPixels, dbpl); +} + +void qt_memrotate180_64(const uchar *srcPixels, int w, int h, int sbpl, uchar *destPixels, int dbpl) +{ + qt_memrotate180((const quint64 *)srcPixels, w, h, sbpl, (quint64 *)destPixels, dbpl); +} + +void qt_memrotate270_64(const uchar *srcPixels, int w, int h, int sbpl, uchar *destPixels, int dbpl) +{ + qt_memrotate270((const quint64 *)srcPixels, w, h, sbpl, (quint64 *)destPixels, dbpl); +} + MemRotateFunc qMemRotateFunctions[QPixelLayout::BPPCount][3] = // 90, 180, 270 { @@ -387,6 +413,7 @@ MemRotateFunc qMemRotateFunctions[QPixelLayout::BPPCount][3] = { qt_memrotate90_16, qt_memrotate180_16, qt_memrotate270_16 }, // BPP16, { qt_memrotate90_24, qt_memrotate180_24, qt_memrotate270_24 }, // BPP24 { qt_memrotate90_32, qt_memrotate180_32, qt_memrotate270_32 }, // BPP32 + { qt_memrotate90_64, qt_memrotate180_64, qt_memrotate270_64 }, // BPP64 }; QT_END_NAMESPACE diff --git a/src/gui/painting/qmemrotate_p.h b/src/gui/painting/qmemrotate_p.h index 9bc3fd1010..677fd57af9 100644 --- a/src/gui/painting/qmemrotate_p.h +++ b/src/gui/painting/qmemrotate_p.h @@ -65,6 +65,7 @@ QT_DECL_MEMROTATE(quint32); QT_DECL_MEMROTATE(quint16); QT_DECL_MEMROTATE(quint24); QT_DECL_MEMROTATE(quint8); +QT_DECL_MEMROTATE(quint64); #undef QT_DECL_MEMROTATE diff --git a/src/gui/painting/qpagedpaintdevice.cpp b/src/gui/painting/qpagedpaintdevice.cpp index 1c7d6471b6..72b2834470 100644 --- a/src/gui/painting/qpagedpaintdevice.cpp +++ b/src/gui/painting/qpagedpaintdevice.cpp @@ -42,6 +42,41 @@ QT_BEGIN_NAMESPACE +class QDummyPagedPaintDevicePrivate : public QPagedPaintDevicePrivate +{ + bool setPageLayout(const QPageLayout &newPageLayout) override + { + m_pageLayout = newPageLayout; + return m_pageLayout.isEquivalentTo(newPageLayout); + } + + bool setPageSize(const QPageSize &pageSize) override + { + m_pageLayout.setPageSize(pageSize); + return m_pageLayout.pageSize().isEquivalentTo(pageSize); + } + + bool setPageOrientation(QPageLayout::Orientation orientation) override + { + m_pageLayout.setOrientation(orientation); + return m_pageLayout.orientation() == orientation; + } + + bool setPageMargins(const QMarginsF &margins, QPageLayout::Unit units) override + { + m_pageLayout.setUnits(units); + m_pageLayout.setMargins(margins); + return m_pageLayout.margins() == margins && m_pageLayout.units() == units; + } + + QPageLayout pageLayout() const override + { + return m_pageLayout; + } + + QPageLayout m_pageLayout; +}; + QPagedPaintDevicePrivate::~QPagedPaintDevicePrivate() { } @@ -61,9 +96,11 @@ QPagedPaintDevicePrivate::~QPagedPaintDevicePrivate() /*! Constructs a new paged paint device. + + \deprecated */ QPagedPaintDevice::QPagedPaintDevice() - : d(new QPagedPaintDevicePrivate) + : d(new QDummyPagedPaintDevicePrivate) { } @@ -253,7 +290,8 @@ QPagedPaintDevicePrivate *QPagedPaintDevice::dd() \value PdfVersion_A1b A PDF/A-1b compatible document is produced. - \since 5.10 + \value PdfVersion_1_6 A PDF 1.6 compatible document is produced. + This value was added in Qt 5.12. */ /*! @@ -263,7 +301,7 @@ QPagedPaintDevicePrivate *QPagedPaintDevice::dd() */ void QPagedPaintDevice::setPageSize(PageSize size) { - d->m_pageLayout.setPageSize(QPageSize(QPageSize::PageSizeId(size))); + d->setPageSize(QPageSize(QPageSize::PageSizeId(size))); } /*! @@ -271,7 +309,7 @@ void QPagedPaintDevice::setPageSize(PageSize size) */ QPagedPaintDevice::PageSize QPagedPaintDevice::pageSize() const { - return PageSize(d->m_pageLayout.pageSize().id()); + return PageSize(d->pageLayout().pageSize().id()); } /*! @@ -282,7 +320,7 @@ QPagedPaintDevice::PageSize QPagedPaintDevice::pageSize() const */ void QPagedPaintDevice::setPageSizeMM(const QSizeF &size) { - d->m_pageLayout.setPageSize(QPageSize(size, QPageSize::Millimeter)); + d->setPageSize(QPageSize(size, QPageSize::Millimeter)); } /*! @@ -290,7 +328,7 @@ void QPagedPaintDevice::setPageSizeMM(const QSizeF &size) */ QSizeF QPagedPaintDevice::pageSizeMM() const { - return d->m_pageLayout.pageSize().size(QPageSize::Millimeter); + return d->pageLayout().pageSize().size(QPageSize::Millimeter); } /*! @@ -305,8 +343,7 @@ QSizeF QPagedPaintDevice::pageSizeMM() const */ void QPagedPaintDevice::setMargins(const Margins &margins) { - d->m_pageLayout.setUnits(QPageLayout::Millimeter); - d->m_pageLayout.setMargins(QMarginsF(margins.left, margins.top, margins.right, margins.bottom)); + d->setPageMargins(QMarginsF(margins.left, margins.top, margins.right, margins.bottom), QPageLayout::Millimeter); } /*! @@ -318,7 +355,7 @@ void QPagedPaintDevice::setMargins(const Margins &margins) */ QPagedPaintDevice::Margins QPagedPaintDevice::margins() const { - QMarginsF margins = d->m_pageLayout.margins(QPageLayout::Millimeter); + QMarginsF margins = d->pageLayout().margins(QPageLayout::Millimeter); Margins result; result.left = margins.left(); result.top = margins.top(); @@ -413,7 +450,7 @@ bool QPagedPaintDevice::setPageOrientation(QPageLayout::Orientation orientation) bool QPagedPaintDevice::setPageMargins(const QMarginsF &margins) { - return d->setPageMargins(margins); + return setPageMargins(margins, pageLayout().units()); } /*! @@ -458,23 +495,30 @@ QPageLayout QPagedPaintDevice::pageLayout() const /*! \internal + \deprecated + Returns the internal device page layout. */ QPageLayout QPagedPaintDevice::devicePageLayout() const { - return d->m_pageLayout; + qWarning("QPagedPaintDevice::devicePageLayout() is deprecated, just use QPagedPaintDevice::pageLayout()"); + return d->pageLayout(); } /*! \internal + \deprecated + Returns the internal device page layout. */ QPageLayout &QPagedPaintDevice::devicePageLayout() { - return d->m_pageLayout; + qWarning("QPagedPaintDevice::devicePageLayout() is deprecated, you shouldn't be using this at all."); + static QPageLayout dummy; + return dummy; } QT_END_NAMESPACE diff --git a/src/gui/painting/qpagedpaintdevice.h b/src/gui/painting/qpagedpaintdevice.h index 66dd6fa8cf..1c37c17fa3 100644 --- a/src/gui/painting/qpagedpaintdevice.h +++ b/src/gui/painting/qpagedpaintdevice.h @@ -55,7 +55,7 @@ class QPagedPaintDevicePrivate; class Q_GUI_EXPORT QPagedPaintDevice : public QPaintDevice { public: - QPagedPaintDevice(); + QT_DEPRECATED QPagedPaintDevice(); ~QPagedPaintDevice(); virtual bool newPage() = 0; @@ -213,7 +213,7 @@ public: Envelope10 = Comm10E }; - enum PdfVersion { PdfVersion_1_4, PdfVersion_A1b }; + enum PdfVersion { PdfVersion_1_4, PdfVersion_A1b, PdfVersion_1_6 }; // ### Qt6 Make these virtual bool setPageLayout(const QPageLayout &pageLayout); @@ -243,8 +243,8 @@ public: protected: QPagedPaintDevice(QPagedPaintDevicePrivate *dd); QPagedPaintDevicePrivate *dd(); - QPageLayout devicePageLayout() const; - QPageLayout &devicePageLayout(); + QT_DEPRECATED QPageLayout devicePageLayout() const; + QT_DEPRECATED QPageLayout &devicePageLayout(); friend class QPagedPaintDevicePrivate; QPagedPaintDevicePrivate *d; }; diff --git a/src/gui/painting/qpagedpaintdevice_p.h b/src/gui/painting/qpagedpaintdevice_p.h index a993ea4cac..3a43bd7828 100644 --- a/src/gui/painting/qpagedpaintdevice_p.h +++ b/src/gui/painting/qpagedpaintdevice_p.h @@ -60,8 +60,7 @@ class Q_GUI_EXPORT QPagedPaintDevicePrivate { public: QPagedPaintDevicePrivate() - : m_pageLayout(QPageSize(QPageSize::A4), QPageLayout::Portrait, QMarginsF(0, 0, 0, 0)), - fromPage(0), + : fromPage(0), toPage(0), pageOrderAscending(true), printSelectionOnly(false) @@ -70,46 +69,19 @@ public: virtual ~QPagedPaintDevicePrivate(); - // ### Qt6 Remove these and make public class methods virtual - virtual bool setPageLayout(const QPageLayout &newPageLayout) - { - m_pageLayout = newPageLayout; - return m_pageLayout.isEquivalentTo(newPageLayout);; - } - virtual bool setPageSize(const QPageSize &pageSize) - { - m_pageLayout.setPageSize(pageSize); - return m_pageLayout.pageSize().isEquivalentTo(pageSize); - } + virtual bool setPageLayout(const QPageLayout &newPageLayout) = 0; - virtual bool setPageOrientation(QPageLayout::Orientation orientation) - { - m_pageLayout.setOrientation(orientation); - return m_pageLayout.orientation() == orientation; - } + virtual bool setPageSize(const QPageSize &pageSize) = 0; - virtual bool setPageMargins(const QMarginsF &margins) - { - return setPageMargins(margins, m_pageLayout.units()); - } + virtual bool setPageOrientation(QPageLayout::Orientation orientation) = 0; - virtual bool setPageMargins(const QMarginsF &margins, QPageLayout::Unit units) - { - m_pageLayout.setUnits(units); - m_pageLayout.setMargins(margins); - return m_pageLayout.margins() == margins && m_pageLayout.units() == units; - } + virtual bool setPageMargins(const QMarginsF &margins, QPageLayout::Unit units) = 0; - virtual QPageLayout pageLayout() const - { - return m_pageLayout; - } + virtual QPageLayout pageLayout() const = 0; static inline QPagedPaintDevicePrivate *get(QPagedPaintDevice *pd) { return pd->d; } - QPageLayout m_pageLayout; - // These are currently required to keep QPrinter functionality working in QTextDocument::print() int fromPage; int toPage; diff --git a/src/gui/painting/qpaintengine_raster.cpp b/src/gui/painting/qpaintengine_raster.cpp index cade334ea6..7caaf3a8fa 100644 --- a/src/gui/painting/qpaintengine_raster.cpp +++ b/src/gui/painting/qpaintengine_raster.cpp @@ -2590,10 +2590,13 @@ void QRasterPaintEngine::drawTiledPixmap(const QRectF &r, const QPixmap &pixmap, if (image.depth() == 1) image = d->rasterBuffer->colorizeBitmap(image, s->pen.color()); - if (s->matrix.type() > QTransform::TxTranslate) { + const qreal pixmapDevicePixelRatio = pixmap.devicePixelRatio(); + if (s->matrix.type() > QTransform::TxTranslate || pixmapDevicePixelRatio > qreal(1.0)) { QTransform copy = s->matrix; copy.translate(r.x(), r.y()); copy.translate(-sr.x(), -sr.y()); + const qreal inverseDpr = qreal(1.0) / pixmapDevicePixelRatio; + copy.scale(inverseDpr, inverseDpr); d->image_filler_xform.clip = d->clip(); d->image_filler_xform.initTexture(&image, s->intOpacity, QTextureData::Tiled); if (!d->image_filler_xform.blend) diff --git a/src/gui/painting/qpaintengineex.cpp b/src/gui/painting/qpaintengineex.cpp index 0643a7cbb6..9f07af92e4 100644 --- a/src/gui/painting/qpaintengineex.cpp +++ b/src/gui/painting/qpaintengineex.cpp @@ -948,6 +948,8 @@ void QPaintEngineEx::drawTiledPixmap(const QRectF &r, const QPixmap &pixmap, con { QBrush brush(state()->pen.color(), pixmap); QTransform xform = QTransform::fromTranslate(r.x() - s.x(), r.y() - s.y()); + if (!qFuzzyCompare(pixmap.devicePixelRatioF(), 1.0)) + xform.scale(1.0/pixmap.devicePixelRatioF(), 1.0/pixmap.devicePixelRatioF()); brush.setTransform(xform); qreal pts[] = { r.x(), r.y(), diff --git a/src/gui/painting/qpainter.cpp b/src/gui/painting/qpainter.cpp index b992e8f55d..b70b29e54e 100644 --- a/src/gui/painting/qpainter.cpp +++ b/src/gui/painting/qpainter.cpp @@ -190,6 +190,13 @@ void QPainterPrivate::checkEmulation() if (pg && pg->coordinateMode() > QGradient::LogicalMode) doEmulation = true; + if (state->brush.style() == Qt::TexturePattern) { + if (qHasPixmapTexture(state->brush)) + doEmulation |= !qFuzzyCompare(state->brush.texture().devicePixelRatioF(), 1.0); + else + doEmulation |= !qFuzzyCompare(state->brush.textureImage().devicePixelRatioF(), 1.0); + } + if (doEmulation && extended->flags() & QPaintEngineEx::DoNotEmulate) return; @@ -523,7 +530,10 @@ static inline QBrush stretchGradientToUserSpace(const QBrush &brush, const QRect g.setCoordinateMode(QGradient::LogicalMode); QBrush b(g); - b.setTransform(gradientToUser * b.transform()); + if (brush.gradient()->coordinateMode() == QGradient::ObjectMode) + b.setTransform(b.transform() * gradientToUser); + else + b.setTransform(gradientToUser * b.transform()); return b; } @@ -562,7 +572,7 @@ void QPainterPrivate::drawStretchedGradient(const QPainterPath &path, DrawOperat } else { needsFill = true; - if (brushMode == QGradient::ObjectBoundingMode) { + if (brushMode == QGradient::ObjectBoundingMode || brushMode == QGradient::ObjectMode) { Q_ASSERT(engine->hasFeature(QPaintEngine::PatternTransform)); boundingRect = path.boundingRect(); q->setBrush(stretchGradientToUserSpace(brush, boundingRect)); @@ -606,11 +616,11 @@ void QPainterPrivate::drawStretchedGradient(const QPainterPath &path, DrawOperat changedBrush = true; } - if (penMode == QGradient::ObjectBoundingMode) { + if (penMode == QGradient::ObjectBoundingMode || penMode == QGradient::ObjectMode) { Q_ASSERT(engine->hasFeature(QPaintEngine::PatternTransform)); // avoid computing the bounding rect twice - if (!needsFill || brushMode != QGradient::ObjectBoundingMode) + if (!needsFill || (brushMode != QGradient::ObjectBoundingMode && brushMode != QGradient::ObjectMode)) boundingRect = path.boundingRect(); QPen p = pen; @@ -842,8 +852,8 @@ void QPainterPrivate::updateEmulationSpecifier(QPainterState *s) gradientStretch |= (brushMode == QGradient::StretchToDeviceMode); gradientStretch |= (penMode == QGradient::StretchToDeviceMode); - objectBoundingMode |= (brushMode == QGradient::ObjectBoundingMode); - objectBoundingMode |= (penMode == QGradient::ObjectBoundingMode); + objectBoundingMode |= (brushMode == QGradient::ObjectBoundingMode || brushMode == QGradient::ObjectMode); + objectBoundingMode |= (penMode == QGradient::ObjectBoundingMode || penMode == QGradient::ObjectMode); } if (gradientStretch) s->emulationSpecifier |= QGradient_StretchToDevice; @@ -1287,7 +1297,7 @@ void QPainterPrivate::updateState(QPainterState *newState) itself and its bounding rectangle: The bounding rect contains pixels with alpha == 0 (i.e the pixels surrounding the primitive). These pixels will overwrite the other image's pixels, - affectively clearing those, while the primitive only overwrites + effectively clearing those, while the primitive only overwrites its own area. \table 100% @@ -1377,7 +1387,7 @@ void QPainterPrivate::updateState(QPainterState *newState) clip. \li Composition Modes \c QPainter::CompositionMode_Source and - QPainter::CompositionMode_SourceOver + QPainter::CompositionMode_SourceOver. \li Rounded rectangle filling using solid color and two-color linear gradients fills. @@ -6657,6 +6667,16 @@ QRectF QPainter::boundingRect(const QRectF &r, const QString &text, const QTextO potentially much more efficient depending on the underlying window system. + drawTiledPixmap() will produce the same visual tiling pattern on + high-dpi displays (with devicePixelRatio > 1), compared to normal- + dpi displays. Set the devicePixelRatio on the \a pixmap to control + the tile size. For example, setting it to 2 halves the tile width + and height (on both 1x and 2x displays), and produces high-resolution + output on 2x displays. + + The \a position offset is always in the painter coordinate system, + indepentent of display devicePixelRatio. + \sa drawPixmap() */ void QPainter::drawTiledPixmap(const QRectF &r, const QPixmap &pixmap, const QPointF &sp) @@ -6840,7 +6860,8 @@ static inline bool needsResolving(const QBrush &brush) Qt::BrushStyle s = brush.style(); return ((s == Qt::LinearGradientPattern || s == Qt::RadialGradientPattern || s == Qt::ConicalGradientPattern) && - brush.gradient()->coordinateMode() == QGradient::ObjectBoundingMode); + (brush.gradient()->coordinateMode() == QGradient::ObjectBoundingMode || + brush.gradient()->coordinateMode() == QGradient::ObjectMode)); } /*! @@ -7068,6 +7089,37 @@ void QPainter::fillRect(const QRectF &r, const QColor &color) */ /*! + \fn void QPainter::fillRect(int x, int y, int width, int height, QGradient::Preset preset) + + \overload + + Fills the rectangle beginning at (\a{x}, \a{y}) with the given \a + width and \a height, using the given gradient \a preset. + + \since 5.12 +*/ + +/*! + \fn void QPainter::fillRect(const QRect &rectangle, QGradient::Preset preset); + + \overload + + Fills the given \a rectangle with the specified gradient \a preset. + + \since 5.12 +*/ + +/*! + \fn void QPainter::fillRect(const QRectF &rectangle, QGradient::Preset preset); + + \overload + + Fills the given \a rectangle with the specified gradient \a preset. + + \since 5.12 +*/ + +/*! Sets the given render \a hint on the painter if \a on is true; otherwise clears the render hint. @@ -8207,6 +8259,7 @@ void QPainter::setTransform(const QTransform &transform, bool combine ) } /*! + Alias for worldTransform(). Returns the world transformation matrix. \sa worldTransform() diff --git a/src/gui/painting/qpainter.h b/src/gui/painting/qpainter.h index 12a9c720a8..482f5fb63d 100644 --- a/src/gui/painting/qpainter.h +++ b/src/gui/painting/qpainter.h @@ -448,6 +448,10 @@ public: inline void fillRect(const QRect &r, Qt::BrushStyle style); inline void fillRect(const QRectF &r, Qt::BrushStyle style); + inline void fillRect(int x, int y, int w, int h, QGradient::Preset preset); + inline void fillRect(const QRect &r, QGradient::Preset preset); + inline void fillRect(const QRectF &r, QGradient::Preset preset); + void eraseRect(const QRectF &); inline void eraseRect(int x, int y, int w, int h); inline void eraseRect(const QRect &); @@ -746,6 +750,20 @@ inline void QPainter::fillRect(const QRectF &r, Qt::BrushStyle style) fillRect(r, QBrush(style)); } +inline void QPainter::fillRect(int x, int y, int w, int h, QGradient::Preset p) +{ + fillRect(QRect(x, y, w, h), QGradient(p)); +} + +inline void QPainter::fillRect(const QRect &r, QGradient::Preset p) +{ + fillRect(r, QGradient(p)); +} + +inline void QPainter::fillRect(const QRectF &r, QGradient::Preset p) +{ + fillRect(r, QGradient(p)); +} inline void QPainter::setBrushOrigin(int x, int y) { diff --git a/src/gui/painting/qpainterpath.cpp b/src/gui/painting/qpainterpath.cpp index 2574a00838..c5ccf0003d 100644 --- a/src/gui/painting/qpainterpath.cpp +++ b/src/gui/painting/qpainterpath.cpp @@ -2476,7 +2476,7 @@ QDataStream &operator>>(QDataStream &s, QPainterPath &p) s >> p.d_func()->cStart; int fillRule; s >> fillRule; - Q_ASSERT(fillRule == Qt::OddEvenFill || Qt::WindingFill); + Q_ASSERT(fillRule == Qt::OddEvenFill || fillRule == Qt::WindingFill); p.d_func()->fillRule = Qt::FillRule(fillRule); p.d_func()->dirtyBounds = true; p.d_func()->dirtyControlBounds = true; diff --git a/src/gui/painting/qpdf.cpp b/src/gui/painting/qpdf.cpp index 8a0020bd2d..e69726b617 100644 --- a/src/gui/painting/qpdf.cpp +++ b/src/gui/painting/qpdf.cpp @@ -953,7 +953,18 @@ void QPdfEngine::drawPixmap (const QRectF &rectangle, const QPixmap &pixmap, con if (object < 0) return; - *d->currentPage << "q\n/GSa gs\n"; + *d->currentPage << "q\n"; + + if ((d->pdfVersion != QPdfEngine::Version_A1b) && (d->opacity != 1.0)) { + int stateObject = d->addConstantAlphaObject(qRound(255 * d->opacity), qRound(255 * d->opacity)); + if (stateObject) + *d->currentPage << "/GState" << stateObject << "gs\n"; + else + *d->currentPage << "/GSa gs\n"; + } else { + *d->currentPage << "/GSa gs\n"; + } + *d->currentPage << QPdf::generateMatrix(QTransform(rectangle.width() / sr.width(), 0, 0, rectangle.height() / sr.height(), rectangle.x(), rectangle.y()) * (d->simplePen ? QTransform() : d->stroker.matrix)); @@ -981,7 +992,18 @@ void QPdfEngine::drawImage(const QRectF &rectangle, const QImage &image, const Q if (object < 0) return; - *d->currentPage << "q\n/GSa gs\n"; + *d->currentPage << "q\n"; + + if ((d->pdfVersion != QPdfEngine::Version_A1b) && (d->opacity != 1.0)) { + int stateObject = d->addConstantAlphaObject(qRound(255 * d->opacity), qRound(255 * d->opacity)); + if (stateObject) + *d->currentPage << "/GState" << stateObject << "gs\n"; + else + *d->currentPage << "/GSa gs\n"; + } else { + *d->currentPage << "/GSa gs\n"; + } + *d->currentPage << QPdf::generateMatrix(QTransform(rectangle.width() / sr.width(), 0, 0, rectangle.height() / sr.height(), rectangle.x(), rectangle.y()) * (d->simplePen ? QTransform() : d->stroker.matrix)); @@ -1114,8 +1136,9 @@ void QPdfEngine::updateState(const QPaintEngineState &state) d->hasPen = d->pen.style() != Qt::NoPen; d->stroker.setPen(d->pen, state.renderHints()); QBrush penBrush = d->pen.brush(); + bool cosmeticPen = qt_pen_is_cosmetic(d->pen, state.renderHints()); bool oldSimple = d->simplePen; - d->simplePen = (d->hasPen && (penBrush.style() == Qt::SolidPattern) && penBrush.isOpaque() && d->opacity == 1.0); + d->simplePen = (d->hasPen && !cosmeticPen && (penBrush.style() == Qt::SolidPattern) && penBrush.isOpaque() && d->opacity == 1.0); if (oldSimple != d->simplePen) flags |= DirtyTransform; } else if (flags & DirtyHints) { @@ -1546,7 +1569,14 @@ void QPdfEnginePrivate::writeHeader() { addXrefEntry(0,false); - xprintf("%%PDF-1.4\n"); + static const QHash<QPdfEngine::PdfVersion, const char *> mapping { + {QPdfEngine::Version_1_4, "1.4"}, + {QPdfEngine::Version_A1b, "1.4"}, + {QPdfEngine::Version_1_6, "1.6"} + }; + const char *verStr = mapping.value(pdfVersion, "1.4"); + + xprintf("%%PDF-%s\n", verStr); xprintf("%%\303\242\303\243\n"); writeInfo(); @@ -1879,6 +1909,19 @@ void QPdfEnginePrivate::embedFont(QFontSubset *font) } } +qreal QPdfEnginePrivate::calcUserUnit() const +{ + // PDF standards < 1.6 support max 200x200in pages (no UserUnit) + if (pdfVersion < QPdfEngine::Version_1_6) + return 1.0; + + const int maxLen = qMax(currentPage->pageSize.width(), currentPage->pageSize.height()); + if (maxLen <= 14400) + return 1.0; // for pages up to 200x200in (14400x14400 units) use default scaling + + // for larger pages, rescale units so we can have up to 381x381km + return qMin(maxLen / 14400.0, 75000.0); +} void QPdfEnginePrivate::writeFonts() { @@ -1901,6 +1944,8 @@ void QPdfEnginePrivate::writePage() uint resources = requestObject(); uint annots = requestObject(); + qreal userUnit = calcUserUnit(); + addXrefEntry(pages.constLast()); xprintf("<<\n" "/Type /Page\n" @@ -1908,12 +1953,16 @@ void QPdfEnginePrivate::writePage() "/Contents %d 0 R\n" "/Resources %d 0 R\n" "/Annots %d 0 R\n" - "/MediaBox [0 0 %d %d]\n" - ">>\n" - "endobj\n", + "/MediaBox [0 0 %f %f]\n", pageRoot, pageStream, resources, annots, // make sure we use the pagesize from when we started the page, since the user may have changed it - currentPage->pageSize.width(), currentPage->pageSize.height()); + currentPage->pageSize.width() / userUnit, currentPage->pageSize.height() / userUnit); + + if (pdfVersion >= QPdfEngine::Version_1_6) + xprintf("/UserUnit %f\n", userUnit); + + xprintf(">>\n" + "endobj\n"); addXrefEntry(resources); xprintf("<<\n" @@ -2982,8 +3031,9 @@ void QPdfEnginePrivate::drawTextItem(const QPointF &p, const QTextItemInt &ti) QTransform QPdfEnginePrivate::pageMatrix() const { - qreal scale = 72./resolution; - QTransform tmp(scale, 0.0, 0.0, -scale, 0.0, m_pageLayout.fullRectPoints().height()); + qreal userUnit = calcUserUnit(); + qreal scale = 72. / userUnit / resolution; + QTransform tmp(scale, 0.0, 0.0, -scale, 0.0, m_pageLayout.fullRectPoints().height() / userUnit); if (m_pageLayout.mode() != QPageLayout::FullPageMode) { QRect r = m_pageLayout.paintRectPixels(resolution); tmp.translate(r.left(), r.top()); diff --git a/src/gui/painting/qpdf_p.h b/src/gui/painting/qpdf_p.h index 5a909f2ede..e2526de67d 100644 --- a/src/gui/painting/qpdf_p.h +++ b/src/gui/painting/qpdf_p.h @@ -171,7 +171,8 @@ public: enum PdfVersion { Version_1_4, - Version_A1b + Version_A1b, + Version_1_6 }; QPdfEngine(); @@ -300,6 +301,7 @@ private: void writePageRoot(); void writeFonts(); void embedFont(QFontSubset *font); + qreal calcUserUnit() const; QVector<int> xrefPositions; QDataStream* stream; diff --git a/src/gui/painting/qpdfwriter.cpp b/src/gui/painting/qpdfwriter.cpp index 2f24c7efcb..6c85d65538 100644 --- a/src/gui/painting/qpdfwriter.cpp +++ b/src/gui/painting/qpdfwriter.cpp @@ -83,41 +83,28 @@ public: { // Try to set the paint engine page layout pd->engine->setPageLayout(newPageLayout); - // Set QPagedPaintDevice layout to match the current paint engine layout - m_pageLayout = pd->engine->pageLayout(); - return m_pageLayout.isEquivalentTo(newPageLayout); + return pageLayout().isEquivalentTo(newPageLayout); } bool setPageSize(const QPageSize &pageSize) override { // Try to set the paint engine page size pd->engine->setPageSize(pageSize); - // Set QPagedPaintDevice layout to match the current paint engine layout - m_pageLayout = pd->engine->pageLayout(); - return m_pageLayout.pageSize().isEquivalentTo(pageSize); + return pageLayout().pageSize().isEquivalentTo(pageSize); } bool setPageOrientation(QPageLayout::Orientation orientation) override { // Set the print engine value pd->engine->setPageOrientation(orientation); - // Set QPagedPaintDevice layout to match the current paint engine layout - m_pageLayout = pd->engine->pageLayout(); - return m_pageLayout.orientation() == orientation; - } - - bool setPageMargins(const QMarginsF &margins) override - { - return setPageMargins(margins, pageLayout().units()); + return pageLayout().orientation() == orientation; } bool setPageMargins(const QMarginsF &margins, QPageLayout::Unit units) override { // Try to set engine margins pd->engine->setPageMargins(margins, units); - // Set QPagedPaintDevice layout to match the current paint engine layout - m_pageLayout = pd->engine->pageLayout(); - return m_pageLayout.margins() == margins && m_pageLayout.units() == units; + return pageLayout().margins() == margins && pageLayout().units() == units; } QPageLayout pageLayout() const override @@ -150,9 +137,6 @@ QPdfWriter::QPdfWriter(const QString &filename) Q_D(QPdfWriter); d->engine->setOutputFilename(filename); - - // Set QPagedPaintDevice layout to match the current paint engine layout - devicePageLayout() = d->engine->pageLayout(); } /*! @@ -165,9 +149,6 @@ QPdfWriter::QPdfWriter(QIODevice *device) Q_D(QPdfWriter); d->engine->d_func()->outDevice = device; - - // Set QPagedPaintDevice layout to match the current paint engine layout - devicePageLayout() = d->engine->pageLayout(); } /*! @@ -189,11 +170,17 @@ void QPdfWriter::setPdfVersion(PdfVersion version) { Q_D(QPdfWriter); + static const QHash<QPdfWriter::PdfVersion, QPdfEngine::PdfVersion> engineMapping { + {QPdfWriter::PdfVersion_1_4, QPdfEngine::Version_1_4}, + {QPdfWriter::PdfVersion_A1b, QPdfEngine::Version_A1b}, + {QPdfWriter::PdfVersion_1_6, QPdfEngine::Version_1_6} + }; + if (d->pdfVersion == version) return; d->pdfVersion = version; - d->engine->setPdfVersion(d->pdfVersion == QPdfWriter::PdfVersion_1_4 ? QPdfEngine::Version_1_4 : QPdfEngine::Version_A1b); + d->engine->setPdfVersion(engineMapping.value(version, QPdfEngine::Version_1_4)); } /*! diff --git a/src/gui/painting/qplatformbackingstore.cpp b/src/gui/painting/qplatformbackingstore.cpp index e05efca9c8..3081a4b1b6 100644 --- a/src/gui/painting/qplatformbackingstore.cpp +++ b/src/gui/painting/qplatformbackingstore.cpp @@ -80,6 +80,8 @@ QT_BEGIN_NAMESPACE +Q_LOGGING_CATEGORY(lcQpaBackingStore, "qt.qpa.backingstore", QtWarningMsg); + class QPlatformBackingStorePrivate { public: @@ -331,16 +333,19 @@ void QPlatformBackingStore::composeAndFlush(QWindow *window, const QRegion ®i d_ptr->context->setScreen(d_ptr->window->screen()); d_ptr->context->setShareContext(qt_window_private(d_ptr->window)->shareContext()); if (!d_ptr->context->create()) { - qWarning("composeAndFlush: QOpenGLContext creation failed"); + qCWarning(lcQpaBackingStore, "composeAndFlush: QOpenGLContext creation failed"); return; } } if (!d_ptr->context->makeCurrent(window)) { - qWarning("composeAndFlush: makeCurrent() failed"); + qCWarning(lcQpaBackingStore, "composeAndFlush: makeCurrent() failed"); return; } + qCDebug(lcQpaBackingStore) << "Composing and flushing" << region << "of" << window + << "at offset" << offset << "with" << textures->count() << "texture(s) in" << textures; + QWindowPrivate::get(window)->lastComposeTime.start(); QOpenGLFunctions *funcs = d_ptr->context->functions(); diff --git a/src/gui/painting/qplatformbackingstore.h b/src/gui/painting/qplatformbackingstore.h index 381c564079..de5ba964dc 100644 --- a/src/gui/painting/qplatformbackingstore.h +++ b/src/gui/painting/qplatformbackingstore.h @@ -50,6 +50,7 @@ // #include <QtGui/qtguiglobal.h> +#include <QtCore/qloggingcategory.h> #include <QtCore/qrect.h> #include <QtCore/qobject.h> @@ -59,6 +60,7 @@ QT_BEGIN_NAMESPACE +Q_GUI_EXPORT Q_DECLARE_LOGGING_CATEGORY(lcQpaBackingStore) class QRegion; class QRect; diff --git a/src/gui/painting/qrgba64.h b/src/gui/painting/qrgba64.h index 2c8f8fa8c4..0e5344cacb 100644 --- a/src/gui/painting/qrgba64.h +++ b/src/gui/painting/qrgba64.h @@ -127,6 +127,10 @@ public: Q_DECL_RELAXED_CONSTEXPR QRgba64 premultiplied() const { + if (isOpaque()) + return *this; + if (isTransparent()) + return QRgba64::fromRgba64(0); const quint32 a = alpha(); const quint16 r = div_65535(red() * a); const quint16 g = div_65535(green() * a); diff --git a/src/gui/painting/qrgba64.qdoc b/src/gui/painting/qrgba64.qdoc index 26c78a5dcd..3ee8a355e6 100644 --- a/src/gui/painting/qrgba64.qdoc +++ b/src/gui/painting/qrgba64.qdoc @@ -35,7 +35,7 @@ QRgba64 is a 64-bit data-structure containing four 16-bit color channels: Red, green, blue and alpha. - QRgba64 can be used a replacement for QRgb when higher precision is needed. In particular a + QRgba64 can be used as a replacement for QRgb when higher precision is needed. In particular a premultiplied QRgba64 can operate on unpremultiplied QRgb without loss of precision except for alpha 0. diff --git a/src/gui/painting/qrgba64_p.h b/src/gui/painting/qrgba64_p.h index adceda2210..b7e4d4d905 100644 --- a/src/gui/painting/qrgba64_p.h +++ b/src/gui/painting/qrgba64_p.h @@ -51,10 +51,11 @@ // We mean it. // +#include "qrgba64.h" +#include "qdrawhelper_p.h" + +#include <QtCore/private/qsimd_p.h> #include <QtGui/private/qtguiglobal_p.h> -#include <QtGui/qrgba64.h> -#include <QtGui/private/qdrawhelper_p.h> -#include <private/qsimd_p.h> QT_BEGIN_NAMESPACE @@ -63,14 +64,6 @@ inline QRgba64 combineAlpha256(QRgba64 rgba64, uint alpha256) return QRgba64::fromRgba64(rgba64.red(), rgba64.green(), rgba64.blue(), (rgba64.alpha() * alpha256) >> 8); } -inline QRgba64 multiplyAlpha256(QRgba64 rgba64, uint alpha256) -{ - return QRgba64::fromRgba64((rgba64.red() * alpha256) >> 8, - (rgba64.green() * alpha256) >> 8, - (rgba64.blue() * alpha256) >> 8, - (rgba64.alpha() * alpha256) >> 8); -} - inline QRgba64 multiplyAlpha65535(QRgba64 rgba64, uint alpha65535) { return QRgba64::fromRgba64(qt_div_65535(rgba64.red() * alpha65535), @@ -125,11 +118,6 @@ inline T multiplyAlpha255(T rgba64, uint alpha255) #endif } -inline QRgba64 interpolate256(QRgba64 x, uint alpha1, QRgba64 y, uint alpha2) -{ - return QRgba64::fromRgba64(multiplyAlpha256(x, alpha1) + multiplyAlpha256(y, alpha2)); -} - inline QRgba64 interpolate255(QRgba64 x, uint alpha1, QRgba64 y, uint alpha2) { return QRgba64::fromRgba64(multiplyAlpha255(x, alpha1) + multiplyAlpha255(y, alpha2)); @@ -159,7 +147,7 @@ Q_ALWAYS_INLINE __m128i interpolate65535(__m128i x, uint alpha1, __m128i y, uint { return _mm_add_epi32(multiplyAlpha65535(x, alpha1), multiplyAlpha65535(y, alpha2)); } -// alpha2 below is const-ref because otherwise MSVC2013 complains that it can't 16-byte align the argument. +// alpha2 below is const-ref because otherwise MSVC2015 complains that it can't 16-byte align the argument. Q_ALWAYS_INLINE __m128i interpolate65535(__m128i x, __m128i alpha1, __m128i y, const __m128i &alpha2) { return _mm_add_epi32(multiplyAlpha65535(x, alpha1), multiplyAlpha65535(y, alpha2)); @@ -185,7 +173,8 @@ inline QRgba64 addWithSaturation(QRgba64 a, QRgba64 b) qMin(a.alpha() + b.alpha(), 65535)); } -#if defined __SSE2__ +#if QT_COMPILER_SUPPORTS_HERE(SSE2) +QT_FUNCTION_TARGET(SSE2) Q_ALWAYS_INLINE uint toArgb32(__m128i v) { v = _mm_unpacklo_epi16(v, _mm_setzero_si128()); @@ -239,20 +228,6 @@ inline uint toRgba8888(QRgba64 rgba64) #endif } -#if defined(__SSE2__) -Q_ALWAYS_INLINE __m128i addWithSaturation(__m128i a, __m128i b) -{ - return _mm_adds_epu16(a, b); -} -#endif - -#if defined(__ARM_NEON__) -Q_ALWAYS_INLINE uint16x4_t addWithSaturation(uint16x4_t a, uint16x4_t b) -{ - return vqmovn_u32(vaddl_u16(a, b)); -} -#endif - inline QRgba64 rgbBlend(QRgba64 d, QRgba64 s, uint rgbAlpha) { QRgba64 blend; diff --git a/src/gui/painting/qt_attribution.json b/src/gui/painting/qt_attribution.json index 7eb7db4876..9d3debc1b9 100644 --- a/src/gui/painting/qt_attribution.json +++ b/src/gui/painting/qt_attribution.json @@ -27,5 +27,19 @@ "Copyright": "Copyright (C) 2004, 2005 Daniel M. Duley. (C) Carsten Haitzler and various contributors. (C) Willem Monsuwe <willem@stack.nl>" + }, + { + "Id": "webgradients", + "Name": "WebGradients", + "QDocModule": "qtgui", + "QtUsage": "Used in Qt GUI to provide presets for QGradient.", + "Files": "webgradients.css", + + "Description": "WebGradients is a free collection of 180 linear gradients.", + "Homepage": "https://webgradients.com/", + "License": "MIT License", + "LicenseId": "MIT", + "LicenseFile": "WEBGRADIENTS_LICENSE.txt", + "Copyright": "Copyright (c) 2017 itmeo" } ] diff --git a/src/gui/painting/qtriangulator.cpp b/src/gui/painting/qtriangulator.cpp index 6d57eba123..9be3eeaffd 100644 --- a/src/gui/painting/qtriangulator.cpp +++ b/src/gui/painting/qtriangulator.cpp @@ -472,7 +472,7 @@ class QInt64Set { public: inline QInt64Set(int capacity = 64); - inline ~QInt64Set() {if (m_array) delete[] m_array;} + inline ~QInt64Set() {delete[] m_array;} inline bool isValid() const {return m_array;} void insert(quint64 key); bool contains(quint64 key) const; @@ -493,10 +493,7 @@ inline QInt64Set::QInt64Set(int capacity) { m_capacity = primeForCount(capacity); m_array = new quint64[m_capacity]; - if (m_array) - clear(); - else - m_capacity = 0; + clear(); } bool QInt64Set::rehash(int capacity) @@ -506,28 +503,19 @@ bool QInt64Set::rehash(int capacity) m_capacity = capacity; m_array = new quint64[m_capacity]; - if (m_array) { - clear(); - if (oldArray) { - for (int i = 0; i < oldCapacity; ++i) { - if (oldArray[i] != UNUSED) - insert(oldArray[i]); - } - delete[] oldArray; - } - return true; - } else { - m_capacity = oldCapacity; - m_array = oldArray; - return false; + clear(); + for (int i = 0; i < oldCapacity; ++i) { + if (oldArray[i] != UNUSED) + insert(oldArray[i]); } + delete[] oldArray; + return true; } void QInt64Set::insert(quint64 key) { if (m_count > 3 * m_capacity / 4) rehash(primeForCount(2 * m_capacity)); - Q_ASSERT_X(m_array, "QInt64Hash<T>::insert", "Hash set not allocated."); int index = int(key % m_capacity); for (int i = 0; i < m_capacity; ++i) { index += i; @@ -546,7 +534,6 @@ void QInt64Set::insert(quint64 key) bool QInt64Set::contains(quint64 key) const { - Q_ASSERT_X(m_array, "QInt64Hash<T>::contains", "Hash set not allocated."); int index = int(key % m_capacity); for (int i = 0; i < m_capacity; ++i) { index += i; @@ -562,7 +549,6 @@ bool QInt64Set::contains(quint64 key) const inline void QInt64Set::clear() { - Q_ASSERT_X(m_array, "QInt64Hash<T>::clear", "Hash set not allocated."); for (int i = 0; i < m_capacity; ++i) m_array[i] = UNUSED; m_count = 0; diff --git a/src/gui/painting/webgradients.binaryjson b/src/gui/painting/webgradients.binaryjson Binary files differnew file mode 100644 index 0000000000..75edd487be --- /dev/null +++ b/src/gui/painting/webgradients.binaryjson diff --git a/src/gui/painting/webgradients.css b/src/gui/painting/webgradients.css new file mode 100644 index 0000000000..870866bb46 --- /dev/null +++ b/src/gui/painting/webgradients.css @@ -0,0 +1,909 @@ +/*001 Warm Flame*/ +.warm_flame{ + background-image: linear-gradient(45deg, #ff9a9e 0%, #fad0c4 99%, #fad0c4 100%); +} + +/*002 Night Fade*/ +.night_fade{ + background-image: linear-gradient(to top, #a18cd1 0%, #fbc2eb 100%); +} + +/*003 Spring Warmth*/ +.spring_warmth{ + background-image: linear-gradient(to top, #fad0c4 0%, #fad0c4 1%, #ffd1ff 100%); +} + +/*004 Juicy Peach*/ +.juicy_peach{ + background-image: linear-gradient(to right, #ffecd2 0%, #fcb69f 100%); +} + +/*005 Young Passion*/ +.young_passion{ + background-image: linear-gradient(to right, #ff8177 0%, #ff867a 0%, #ff8c7f 21%, #f99185 52%, #cf556c 78%, #b12a5b 100%); +} + +/*006 Lady Lips*/ +.lady_lips{ + background-image: linear-gradient(to top, #ff9a9e 0%, #fecfef 99%, #fecfef 100%); +} + +/*007 Sunny Morning*/ +.sunny_morning{ + background-image: linear-gradient(120deg, #f6d365 0%, #fda085 100%); +} + +/*008 Rainy Ashville*/ +.rainy_ashville{ + background-image: linear-gradient(to top, #fbc2eb 0%, #a6c1ee 100%); +} + +/*009 Frozen Dreams*/ +.frozen_dreams{ + background-image: linear-gradient(to top, #fdcbf1 0%, #fdcbf1 1%, #e6dee9 100%); +} + +/*010 Winter Neva*/ +.winter_neva{ + background-image: linear-gradient(120deg, #a1c4fd 0%, #c2e9fb 100%); +} + +/*011 Dusty Grass*/ +.dusty_grass{ + background-image: linear-gradient(120deg, #d4fc79 0%, #96e6a1 100%); +} + +/*012 Tempting Azure*/ +.tempting_azure{ + background-image: linear-gradient(120deg, #84fab0 0%, #8fd3f4 100%); +} + +/*013 Heavy Rain*/ +.heavy_rain{ + background-image: linear-gradient(to top, #cfd9df 0%, #e2ebf0 100%); +} + +/*014 Amy Crisp*/ +.amy_crisp{ + background-image: linear-gradient(120deg, #a6c0fe 0%, #f68084 100%); +} + +/*015 Mean Fruit*/ +.mean_fruit{ + background-image: linear-gradient(120deg, #fccb90 0%, #d57eeb 100%); +} + +/*016 Deep Blue*/ +.deep_blue{ + background-image: linear-gradient(120deg, #e0c3fc 0%, #8ec5fc 100%); +} + +/*017 Ripe Malinka*/ +.ripe_malinka{ + background-image: linear-gradient(120deg, #f093fb 0%, #f5576c 100%); +} + +/*018 Cloudy Knoxville*/ +.cloudy_knoxville{ + background-image: linear-gradient(120deg, #fdfbfb 0%, #ebedee 100%); +} + +/*019 Malibu Beach*/ +.malibu_beach{ + background-image: linear-gradient(to right, #4facfe 0%, #00f2fe 100%); +} + +/*020 New Life*/ +.new_life{ + background-image: linear-gradient(to right, #43e97b 0%, #38f9d7 100%); +} + +/*021 True Sunset*/ +.true_sunset{ + background-image: linear-gradient(to right, #fa709a 0%, #fee140 100%); +} + +/*022 Morpheus Den*/ +.morpheus_den{ + background-image: linear-gradient(to top, #30cfd0 0%, #330867 100%); +} + +/*023 Rare Wind*/ +.rare_wind{ + background-image: linear-gradient(to top, #a8edea 0%, #fed6e3 100%); +} + +/*024 Near Moon*/ +.near_moon{ + background-image: linear-gradient(to top, #5ee7df 0%, #b490ca 100%); +} + +/*025 Wild Apple*/ +.wild_apple{ + background-image: linear-gradient(to top, #d299c2 0%, #fef9d7 100%); +} + +/*026 Saint Petersburg*/ +.saint_petersburg{ + background-image: linear-gradient(135deg, #f5f7fa 0%, #c3cfe2 100%); +} + +/*027 Arielle's Smile*/ +.arielles_smile{ + background-image: radial-gradient(circle 248px at center, #16d9e3 0%, #30c7ec 47%, #46aef7 100%); +} + +/*028 Plum Plate*/ +.plum_plate{ + background-image: linear-gradient(135deg, #667eea 0%, #764ba2 100%); +} + +/*029 Everlasting Sky*/ +.everlasting_sky{ + background-image: linear-gradient(135deg, #fdfcfb 0%, #e2d1c3 100%); +} + +/*030 Happy Fisher*/ +.happy_fisher{ + background-image: linear-gradient(120deg, #89f7fe 0%, #66a6ff 100%); +} + +/*031 Blessing*/ +.blessing{ + background-image: linear-gradient(to top, #fddb92 0%, #d1fdff 100%); +} + +/*032 Sharpeye Eagle*/ +.sharpeye_eagle{ + background-image: linear-gradient(to top, #9890e3 0%, #b1f4cf 100%); +} + +/*033 Ladoga Bottom*/ +.ladoga_bottom{ + background-image: linear-gradient(to top, #ebc0fd 0%, #d9ded8 100%); +} + +/*034 Lemon Gate*/ +.lemon_gate{ + background-image: linear-gradient(to top, #96fbc4 0%, #f9f586 100%); +} + +/*035 Itmeo Branding*/ +.itmeo_branding{ + background-image: linear-gradient(180deg, #2af598 0%, #009efd 100%); +} + +/*036 Zeus Miracle*/ +.zeus_miracle{ + background-image: linear-gradient(to top, #cd9cf2 0%, #f6f3ff 100%); +} + +/*037 Old Hat*/ +.old_hat{ + background-image: linear-gradient(to right, #e4afcb 0%, #b8cbb8 0%, #b8cbb8 0%, #e2c58b 30%, #c2ce9c 64%, #7edbdc 100%); +} + +/*038 Star Wine*/ +.star_wine{ + background-image: linear-gradient(to right, #b8cbb8 0%, #b8cbb8 0%, #b465da 0%, #cf6cc9 33%, #ee609c 66%, #ee609c 100%); +} + +/*039 Deep Blue*/ +.deep_blue{ + background-image: linear-gradient(to right, #6a11cb 0%, #2575fc 100%); +} + +/*040 Coup de Grace*/ +.coup_de_grace{ + background: #DCD9D4 linear-gradient(to bottom, rgba(255, 255, 255, 0.50) 0%, rgba(0, 0, 0, 0.50) 100%), radial-gradient(at 50% 0%, rgba(255, 255, 255, 0.10) 0%, rgba(0, 0, 0, 0.50) 50%); + background-blend-mode: soft-light,screen; +} + +/*041 Happy Acid*/ +.happy_acid{ + background-image: linear-gradient(to top, #37ecba 0%, #72afd3 100%); +} + +/*042 Awesome Pine*/ +.awesome_pine{ + background-image: linear-gradient(to top, #ebbba7 0%, #cfc7f8 100%); +} + +/*043 New York*/ +.new_york{ + background-image: linear-gradient(to top, #fff1eb 0%, #ace0f9 100%); +} + +/*044 Shy Rainbow*/ +.shy_rainbow{ + background-image: linear-gradient(to right, #eea2a2 0%, #bbc1bf 19%, #57c6e1 42%, #b49fda 79%, #7ac5d8 100%); +} + +/*045 Loon Crest*/ +.loon_crest{ + background: linear-gradient(to bottom, rgba(255,255,255,0.15) 0%, rgba(0,0,0,0.15) 100%), radial-gradient(at top center, rgba(255,255,255,0.40) 0%, rgba(0,0,0,0.40) 120%) #989898; + background-blend-mode: multiply,multiply; +} + +/*046 Mixed Hopes*/ +.mixed_hopes{ + background-image: linear-gradient(to top, #c471f5 0%, #fa71cd 100%); +} + +/*047 Fly High*/ +.fly_high{ + background-image: linear-gradient(to top, #48c6ef 0%, #6f86d6 100%); +} + +/*048 Strong Bliss*/ +.strong_bliss{ + background-image: linear-gradient(to right, #f78ca0 0%, #f9748f 19%, #fd868c 60%, #fe9a8b 100%); +} + +/*049 Fresh Milk*/ +.fresh_milk{ + background-image: linear-gradient(to top, #feada6 0%, #f5efef 100%); +} + +/*050 Snow Again*/ +.snow_again{ + background-image: linear-gradient(to top, #e6e9f0 0%, #eef1f5 100%); +} + +/*051 February Ink*/ +.february_ink{ + background-image: linear-gradient(to top, #accbee 0%, #e7f0fd 100%); +} + +/*052 Kind Steel*/ +.kind_steel{ + background-image: linear-gradient(-20deg, #e9defa 0%, #fbfcdb 100%); +} + +/*053 Soft Grass*/ +.soft_grass{ + background-image: linear-gradient(to top, #c1dfc4 0%, #deecdd 100%); +} + +/*054 Grown Early*/ +.grown_early{ + background-image: linear-gradient(to top, #0ba360 0%, #3cba92 100%); +} + +/*055 Sharp Blues*/ +.sharp_blues{ + background-image: linear-gradient(to top, #00c6fb 0%, #005bea 100%); +} + +/*056 Shady Water*/ +.shady_water{ + background-image: linear-gradient(to right, #74ebd5 0%, #9face6 100%); +} + +/*057 Dirty Beauty*/ +.dirty_beauty{ + background-image: linear-gradient(to top, #6a85b6 0%, #bac8e0 100%); +} + +/*058 Great Whale*/ +.great_whale{ + background-image: linear-gradient(to top, #a3bded 0%, #6991c7 100%); +} + +/*059 Teen Notebook*/ +.teen_notebook{ + background-image: linear-gradient(to top, #9795f0 0%, #fbc8d4 100%); +} + +/*060 Polite Rumors*/ +.polite_rumors{ + background-image: linear-gradient(to top, #a7a6cb 0%, #8989ba 52%, #8989ba 100%); +} + +/*061 Sweet Period*/ +.sweet_period{ + background-image: linear-gradient(to top, #3f51b1 0%, #5a55ae 13%, #7b5fac 25%, #8f6aae 38%, #a86aa4 50%, #cc6b8e 62%, #f18271 75%, #f3a469 87%, #f7c978 100%); +} + +/*062 Wide Matrix*/ +.wide_matrix{ + background-image: linear-gradient(to top, #fcc5e4 0%, #fda34b 15%, #ff7882 35%, #c8699e 52%, #7046aa 71%, #0c1db8 87%, #020f75 100%); +} + +/*063 Soft Cherish*/ +.soft_cherish{ + background-image: linear-gradient(to top, #dbdcd7 0%, #dddcd7 24%, #e2c9cc 30%, #e7627d 46%, #b8235a 59%, #801357 71%, #3d1635 84%, #1c1a27 100%); +} + +/*064 Red Salvation*/ +.red_salvation{ + background-image: linear-gradient(to top, #f43b47 0%, #453a94 100%); +} + +/*065 Burning Spring*/ +.burning_spring{ + background-image: linear-gradient(to top, #4fb576 0%, #44c489 30%, #28a9ae 46%, #28a2b7 59%, #4c7788 71%, #6c4f63 86%, #432c39 100%); +} + +/*066 Night Party*/ +.night_party{ + background-image: linear-gradient(to top, #0250c5 0%, #d43f8d 100%); +} + +/*067 Sky Glider*/ +.sky_glider{ + background-image: linear-gradient(to top, #88d3ce 0%, #6e45e2 100%); +} + +/*068 Heaven Peach*/ +.heaven_peach{ + background-image: linear-gradient(to top, #d9afd9 0%, #97d9e1 100%); +} + +/*069 Purple Division*/ +.purple_division{ + background-image: linear-gradient(to top, #7028e4 0%, #e5b2ca 100%); +} + +/*070 Aqua Splash*/ +.aqua_splash{ + background-image: linear-gradient(15deg, #13547a 0%, #80d0c7 100%); +} + +/*071 Above Clouds*/ +.above_clouds{ + background-image: linear-gradient(to left, #BDBBBE 0%, #9D9EA3 100%), radial-gradient(88% 271%, rgba(255, 255, 255, 0.25) 0%, rgba(254, 254, 254, 0.25) 1%, rgba(0, 0, 0, 0.25) 100%), radial-gradient(50% 100%, rgba(255, 255, 255, 0.30) 0%, rgba(0, 0, 0, 0.30) 100%); + background-blend-mode: normal, lighten, soft-light; +} + +/*072 Spiky Naga*/ +.spiky_naga{ + background-image: linear-gradient(to top, #505285 0%, #585e92 12%, #65689f 25%, #7474b0 37%, #7e7ebb 50%, #8389c7 62%, #9795d4 75%, #a2a1dc 87%, #b5aee4 100%); +} + +/*073 Love Kiss*/ +.love_kiss{ + background-image: linear-gradient(to top, #ff0844 0%, #ffb199 100%); +} + +/*074 Sharp Glass*/ +.sharp_glass{ + background: #C9CCD3 linear-gradient(-180deg, rgba(255, 255, 255, 0.50) 0%, rgba(0, 0, 0, 0.50) 100%); + background-blend-mode: lighten; +} + +/*075 Clean Mirror*/ +.clean_mirror{ + background-image: linear-gradient(45deg, #93a5cf 0%, #e4efe9 100%); +} + +/*076 Premium Dark*/ +.premium_dark{ + background-image: linear-gradient(to right, #434343 0%, black 100%); +} + +/*077 Cold Evening*/ +.cold_evening{ + background-image: linear-gradient(to top, #0c3483 0%, #a2b6df 100%, #6b8cce 100%, #a2b6df 100%); +} + +/*078 Cochiti Lake*/ +.cochiti_lake{ + background-image: linear-gradient(45deg, #93a5cf 0%, #e4efe9 100%); +} + +/*079 Summer Games*/ +.summer_games{ + background-image: linear-gradient(to right, #92fe9d 0%, #00c9ff 100%); +} + +/*080 Passionate Bed*/ +.passionate_bed{ + background-image: linear-gradient(to right, #ff758c 0%, #ff7eb3 100%); +} + +/*081 Mountain Rock*/ +.mountain_rock{ + background-image: linear-gradient(to right, #868f96 0%, #596164 100%); +} + +/*082 Desert Hump*/ +.desert_hump{ + background-image: linear-gradient(to top, #c79081 0%, #dfa579 100%); +} + +/*083 Jungle Day*/ +.jungle_day{ + background-image: linear-gradient(45deg, #8baaaa 0%, #ae8b9c 100%); +} + +/*084 Phoenix Start*/ +.phoenix_start{ + background-image: linear-gradient(to right, #f83600 0%, #f9d423 100%); +} + +/*085 October Silence*/ +.october_silence{ + background-image: linear-gradient(-20deg, #b721ff 0%, #21d4fd 100%); +} + +/*086 Faraway River*/ +.faraway_river{ + background-image: linear-gradient(-20deg, #6e45e2 0%, #88d3ce 100%); +} + +/*087 Alchemist Lab*/ +.alchemist_lab{ + background-image: linear-gradient(-20deg, #d558c8 0%, #24d292 100%); +} + +/*088 Over Sun*/ +.over_sun{ + background-image: linear-gradient(60deg, #abecd6 0%, #fbed96 100%); +} + +/*089 Premium White*/ +.premium_white{ + background-image: linear-gradient(to top, #d5d4d0 0%, #d5d4d0 1%, #eeeeec 31%, #efeeec 75%, #e9e9e7 100%); +} + +/*090 Mars Party*/ +.mars_party{ + background-image: linear-gradient(to top, #5f72bd 0%, #9b23ea 100%); +} + +/*091 Eternal Constance*/ +.eternal_constance{ + background-image: linear-gradient(to top, #09203f 0%, #537895 100%); +} + +/*092 Japan Blush*/ +.japan_blush{ + background-image: linear-gradient(-20deg, #ddd6f3 0%, #faaca8 100%, #faaca8 100%); +} + +/*093 Smiling Rain*/ +.smiling_rain{ + background-image: linear-gradient(-20deg, #dcb0ed 0%, #99c99c 100%); +} + +/*094 Cloudy Apple*/ +.cloudy_apple{ + background-image: linear-gradient(to top, #f3e7e9 0%, #e3eeff 99%, #e3eeff 100%); +} + +/*095 Big Mango*/ +.big_mango{ + background-image: linear-gradient(to top, #c71d6f 0%, #d09693 100%); +} + +/*096 Healthy Water*/ +.healthy_water{ + background-image: linear-gradient(60deg, #96deda 0%, #50c9c3 100%); +} + +/*097 Amour Amour*/ +.amour_amour{ + background-image: linear-gradient(to top, #f77062 0%, #fe5196 100%); +} + +/*098 Risky Concrete*/ +.risky_concrete{ + background-image: linear-gradient(to top, #c4c5c7 0%, #dcdddf 52%, #ebebeb 100%); +} + +/*099 Strong Stick*/ +.strong_stick{ + background-image: linear-gradient(to right, #a8caba 0%, #5d4157 100%); +} + +/*100 Vicious Stance*/ +.vicious_stance{ + background-image: linear-gradient(60deg, #29323c 0%, #485563 100%); +} + +/*101 Palo Alto*/ +.palo_alto{ + background-image: linear-gradient(-60deg, #16a085 0%, #f4d03f 100%); +} + +/*102 Happy Memories*/ +.happy_memories{ + background-image: linear-gradient(-60deg, #ff5858 0%, #f09819 100%); +} + +/*103 Midnight Bloom*/ +.midnight_bloom{ + background-image: linear-gradient(-20deg, #2b5876 0%, #4e4376 100%); +} + +/*104 Crystalline*/ +.crystalline{ + background-image: linear-gradient(-20deg, #00cdac 0%, #8ddad5 100%); +} + +/*105 Raccoon Back*/ +.raccoon_back{ + background: linear-gradient(-180deg, #BCC5CE 0%, #929EAD 98%), radial-gradient(at top left, rgba(255,255,255,0.30) 0%, rgba(0,0,0,0.30) 100%); + background-blend-mode: screen; +} + +/*106 Party Bliss*/ +.party_bliss{ + background-image: linear-gradient(to top, #4481eb 0%, #04befe 100%); +} + +/*107 Confident Cloud*/ +.confident_cloud{ + background-image: linear-gradient(to top, #dad4ec 0%, #dad4ec 1%, #f3e7e9 100%); +} + +/*108 Le Cocktail*/ +.le_cocktail{ + background-image: linear-gradient(45deg, #874da2 0%, #c43a30 100%); +} + +/*109 River City*/ +.river_city{ + background-image: linear-gradient(to top, #4481eb 0%, #04befe 100%); +} + +/*110 Frozen Berry*/ +.frozen_berry{ + background-image: linear-gradient(to top, #e8198b 0%, #c7eafd 100%); +} + +/*111 Elegance*/ +.elegance{ + background-image: radial-gradient(73% 147%, #EADFDF 59%, #ECE2DF 100%), radial-gradient(91% 146%, rgba(255,255,255,0.50) 47%, rgba(0,0,0,0.50) 100%); + background-blend-mode: screen; +} + +/*112 Child Care*/ +.child_care{ + background-image: linear-gradient(-20deg, #f794a4 0%, #fdd6bd 100%); +} + +/*113 Flying Lemon*/ +.flying_lemon{ + background-image: linear-gradient(60deg, #64b3f4 0%, #c2e59c 100%); +} + +/*114 New Retrowave*/ +.new_retrowave{ + background-image: linear-gradient(to top, #3b41c5 0%, #a981bb 49%, #ffc8a9 100%); +} + +/*115 Hidden Jaguar*/ +.hidden_jaguar{ + background-image: linear-gradient(to top, #0fd850 0%, #f9f047 100%); +} + +/*116 Above The Sky*/ +.above_the_sky{ + background-image: linear-gradient(to top, lightgrey 0%, lightgrey 1%, #e0e0e0 26%, #efefef 48%, #d9d9d9 75%, #bcbcbc 100%); +} + +/*117 Nega*/ +.nega{ + background-image: linear-gradient(45deg, #ee9ca7 0%, #ffdde1 100%); +} + +/*118 Dense Water*/ +.dense_water{ + background-image: linear-gradient(to right, #3ab5b0 0%, #3d99be 31%, #56317a 100%); +} + +/*119 Chemic Aqua*/ +.chemic_aqua{ + background: #CDDCDC radial-gradient(at 50% 100%, rgba(255, 255, 255, 0.50) 0%, rgba(0, 0, 0, 0.50) 100%), linear-gradient(to bottom, rgba(255, 255, 255, 0.25) 0%, rgba(0, 0, 0, 0.25) 100%); + background-blend-mode: screen, overlay; +} + +/*120 Seashore*/ +.seashore{ + background-image: linear-gradient(to top, #209cff 0%, #68e0cf 100%); +} + +/*121 Marble Wall*/ +.marble_wall{ + background-image: linear-gradient(to top, #bdc2e8 0%, #bdc2e8 1%, #e6dee9 100%); +} + +/*122 Cheerful Caramel*/ +.cheerful_caramel{ + background-image: linear-gradient(to top, #e6b980 0%, #eacda3 100%); +} + +/*123 Night Sky*/ +.night_sky{ + background-image: linear-gradient(to top, #1e3c72 0%, #1e3c72 1%, #2a5298 100%); +} + +/*124 Magic Lake*/ +.magic_lake{ + background-image: linear-gradient(to top, #d5dee7 0%, #ffafbd 0%, #c9ffbf 100%); +} + +/*125 Young Grass*/ +.young_grass{ + background-image: linear-gradient(to top, #9be15d 0%, #00e3ae 100%); +} + +/*126 Colorful Peach*/ +.colorful_peach{ + background-image: linear-gradient(to right, #ed6ea0 0%, #ec8c69 100%); +} + +/*127 Gentle Care*/ +.gentle_care{ + background-image: linear-gradient(to right, #ffc3a0 0%, #ffafbd 100%); +} + +/*128 Plum Bath*/ +.plum_bath{ + background-image: linear-gradient(to top, #cc208e 0%, #6713d2 100%); +} + +/*129 Happy Unicorn*/ +.happy_unicorn{ + background-image: linear-gradient(to top, #b3ffab 0%, #12fff7 100%); +} + +/*130 Full Metal*/ +.full_metal{ + background: linear-gradient(to bottom, #D5DEE7 0%, #E8EBF2 50%, #E2E7ED 100%), linear-gradient(to bottom, rgba(0,0,0,0.02) 50%, rgba(255,255,255,0.02) 61%, rgba(0,0,0,0.02) 73%), linear-gradient(33deg, rgba(255,255,255,0.20) 0%, rgba(0,0,0,0.20) 100%); + background-blend-mode: normal,color-burn; +} + +/*131 African Field*/ +.african_field{ + background-image: linear-gradient(to top, #65bd60 0%, #5ac1a8 25%, #3ec6ed 50%, #b7ddb7 75%, #fef381 100%); +} + +/*132 Solid Stone*/ +.solid_stone{ + background-image: linear-gradient(to right, #243949 0%, #517fa4 100%); +} + +/*133 Orange Juice*/ +.orange_juice{ + background-image: linear-gradient(-20deg, #fc6076 0%, #ff9a44 100%); +} + +/*134 Glass Water*/ +.glass_water{ + background-image: linear-gradient(to top, #dfe9f3 0%, white 100%); +} + +/*135 Slick Carbon*/ +.slick_carbon{ + background: linear-gradient(to bottom, #323232 0%, #3F3F3F 40%, #1C1C1C 150%), linear-gradient(to top, rgba(255,255,255,0.40) 0%, rgba(0,0,0,0.25) 200%); + background-blend-mode: multiply; +} + +/*136 North Miracle*/ +.north_miracle{ + background-image: linear-gradient(to right, #00dbde 0%, #fc00ff 100%); +} + +/*137 Fruit Blend*/ +.fruit_blend{ + background-image: linear-gradient(to right, #f9d423 0%, #ff4e50 100%); +} + +/*138 Millennium Pine*/ +.millennium_pine{ + background-image: linear-gradient(to top, #50cc7f 0%, #f5d100 100%); +} + +/*139 High Flight*/ +.high_flight{ + background-image: linear-gradient(to right, #0acffe 0%, #495aff 100%); +} + +/*140 Mole Hall*/ +.mole_hall{ + background-image: linear-gradient(-20deg, #616161 0%, #9bc5c3 100%); +} + +/*141 Earl Gray*/ +.earl_gray{ + background: #E4E4E1 radial-gradient(at top center, rgba(255, 255, 255, 0.03) 0%, rgba(0, 0, 0, 0.03) 100%), linear-gradient(to top, rgba(255, 255, 255, 0.1) 0%, rgba(143, 152, 157, 0.60) 100%); + background-blend-mode: normal, multiply; +} + +/*142 Space Shift*/ +.space_shift{ + background-image: linear-gradient(60deg, #3d3393 0%, #2b76b9 37%, #2cacd1 65%, #35eb93 100%); +} + +/*143 Forest Inei*/ +.forest_inei{ + background-image: linear-gradient(to top, #df89b5 0%, #bfd9fe 100%); +} + +/*144 Royal Garden*/ +.royal_garden{ + background-image: linear-gradient(to right, #ed6ea0 0%, #ec8c69 100%); +} + +/*145 Rich Metal*/ +.rich_metal{ + background-image: linear-gradient(to right, #d7d2cc 0%, #304352 100%); +} + +/*146 Juicy Cake*/ +.juicy_cake{ + background-image: linear-gradient(to top, #e14fad 0%, #f9d423 100%); +} + +/*147 Smart Indigo*/ +.smart_indigo{ + background-image: linear-gradient(to top, #b224ef 0%, #7579ff 100%); +} + +/*148 Sand Strike*/ +.sand_strike{ + background-image: linear-gradient(to right, #c1c161 0%, #c1c161 0%, #d4d4b1 100%); +} + +/*149 Norse Beauty*/ +.norse_beauty{ + background-image: linear-gradient(to right, #ec77ab 0%, #7873f5 100%); +} + +/*150 Aqua Guidance*/ +.aqua_guidance{ + background-image: linear-gradient(to top, #007adf 0%, #00ecbc 100%); +} + +/*151 Sun Veggie*/ +.sun_veggie{ + background-image: linear-gradient(-225deg, #20E2D7 0%, #F9FEA5 100%); +} + +/*152 Sea Lord*/ +.sea_lord{ + background-image: linear-gradient(-225deg, #2CD8D5 0%, #C5C1FF 56%, #FFBAC3 100%); +} + +/*153 Black Sea*/ +.black_sea{ + background-image: linear-gradient(-225deg, #2CD8D5 0%, #6B8DD6 48%, #8E37D7 100%); +} + +/*154 Grass Shampoo*/ +.grass_shampoo{ + background-image: linear-gradient(-225deg, #DFFFCD 0%, #90F9C4 48%, #39F3BB 100%); +} + +/*155 Landing Aircraft*/ +.landing_aircraft{ + background-image: linear-gradient(-225deg, #5D9FFF 0%, #B8DCFF 48%, #6BBBFF 100%); +} + +/*156 Witch Dance*/ +.witch_dance{ + background-image: linear-gradient(-225deg, #A8BFFF 0%, #884D80 100%); +} + +/*157 Sleepless Night*/ +.sleepless_night{ + background-image: linear-gradient(-225deg, #5271C4 0%, #B19FFF 48%, #ECA1FE 100%); +} + +/*158 Angel Care*/ +.angel_care{ + background-image: linear-gradient(-225deg, #FFE29F 0%, #FFA99F 48%, #FF719A 100%); +} + +/*159 Crystal River*/ +.crystal_river{ + background-image: linear-gradient(-225deg, #22E1FF 0%, #1D8FE1 48%, #625EB1 100%); +} + +/*160 Soft Lipstick*/ +.soft_lipstick{ + background-image: linear-gradient(-225deg, #B6CEE8 0%, #F578DC 100%); +} + +/*161 Salt Mountain*/ +.salt_mountain{ + background-image: linear-gradient(-225deg, #FFFEFF 0%, #D7FFFE 100%); +} + +/*162 Perfect White*/ +.perfect_white{ + background-image: linear-gradient(-225deg, #E3FDF5 0%, #FFE6FA 100%); +} + +/*163 Fresh Oasis*/ +.fresh_oasis{ + background-image: linear-gradient(-225deg, #7DE2FC 0%, #B9B6E5 100%); +} + +/*164 Strict November*/ +.strict_november{ + background-image: linear-gradient(-225deg, #CBBACC 0%, #2580B3 100%); +} + +/*165 Morning Salad*/ +.morning_salad{ + background-image: linear-gradient(-225deg, #B7F8DB 0%, #50A7C2 100%); +} + +/*166 Deep Relief*/ +.deep_relief{ + background-image: linear-gradient(-225deg, #7085B6 0%, #87A7D9 50%, #DEF3F8 100%); +} + +/*167 Sea Strike*/ +.sea_strike{ + background-image: linear-gradient(-225deg, #77FFD2 0%, #6297DB 48%, #1EECFF 100%); +} + +/*168 Night Call*/ +.night_call{ + background-image: linear-gradient(-225deg, #AC32E4 0%, #7918F2 48%, #4801FF 100%); +} + +/*169 Supreme Sky*/ +.supreme_sky{ + background-image: linear-gradient(-225deg, #D4FFEC 0%, #57F2CC 48%, #4596FB 100%); +} + +/*170 Light Blue*/ +.light_blue{ + background-image: linear-gradient(-225deg, #9EFBD3 0%, #57E9F2 48%, #45D4FB 100%); +} + +/*171 Mind Crawl*/ +.mind_crawl{ + background-image: linear-gradient(-225deg, #473B7B 0%, #3584A7 51%, #30D2BE 100%); +} + +/*172 Lily Meadow*/ +.lily_meadow{ + background-image: linear-gradient(-225deg, #65379B 0%, #886AEA 53%, #6457C6 100%); +} + +/*173 Sugar Lollipop*/ +.sugar_lollipop{ + background-image: linear-gradient(-225deg, #A445B2 0%, #D41872 52%, #FF0066 100%); +} + +/*174 Sweet Dessert*/ +.sweet_dessert{ + background-image: linear-gradient(-225deg, #7742B2 0%, #F180FF 52%, #FD8BD9 100%); +} + +/*175 Magic Ray*/ +.magic_ray{ + background-image: linear-gradient(-225deg, #FF3CAC 0%, #562B7C 52%, #2B86C5 100%); +} + +/*176 Teen Party*/ +.teen_party{ + background-image: linear-gradient(-225deg, #FF057C 0%, #8D0B93 50%, #321575 100%); +} + +/*177 Frozen Heat*/ +.frozen_heat{ + background-image: linear-gradient(-225deg, #FF057C 0%, #7C64D5 48%, #4CC3FF 100%); +} + +/*178 Gagarin View*/ +.gagarin_view{ + background-image: linear-gradient(-225deg, #69EACB 0%, #EACCF8 48%, #6654F1 100%); +} + +/*179 Fabled Sunset*/ +.fabled_sunset{ + background-image: linear-gradient(-225deg, #231557 0%, #44107A 29%, #FF1361 67%, #FFF800 100%); +} + +/*180 Perfect Blue*/ +.perfect_blue{ + background-image: linear-gradient(-225deg, #3D4E81 0%, #5753C9 48%, #6E7FF3 100%); +} |