From 26914fa9b02e54d5cfd2ada6ddab2a705ca38b3c Mon Sep 17 00:00:00 2001 From: OV2 Date: Fri, 7 Oct 2016 21:39:32 +0200 Subject: [PATCH] Win32: update xbrz --- filter/xbrz-config.h | 19 +- filter/xbrz.cpp | 747 +++++++++++++++++++------------------------ filter/xbrz.h | 9 +- win32/render.cpp | 22 ++ win32/wsnes9x.cpp | 1 + win32/wsnes9x.h | 2 + 6 files changed, 359 insertions(+), 441 deletions(-) diff --git a/filter/xbrz-config.h b/filter/xbrz-config.h index ae11f7a0..824ed5a2 100644 --- a/filter/xbrz-config.h +++ b/filter/xbrz-config.h @@ -1,6 +1,6 @@ // **************************************************************************** // * This file is part of the HqMAME project. It is distributed under * -// * GNU General Public License: http://www.gnu.org/licenses/gpl.html * +// * GNU General Public License: http://www.gnu.org/licenses/gpl-3.0 * // * Copyright (C) Zenju (zenju AT gmx DOT de) - All Rights Reserved * // * * // * Additionally and as a special exception, the author gives permission * @@ -22,18 +22,11 @@ namespace xbrz { struct ScalerCfg { - ScalerCfg() : - luminanceWeight_(1), - equalColorTolerance_(30), - dominantDirectionThreshold(3.6), - steepDirectionThreshold(2.2), - newTestAttribute_(0) {} - - double luminanceWeight_; - double equalColorTolerance_; - double dominantDirectionThreshold; - double steepDirectionThreshold; - double newTestAttribute_; //unused; test new parameters + double luminanceWeight = 1; + double equalColorTolerance = 30; + double dominantDirectionThreshold = 3.6; + double steepDirectionThreshold = 2.2; + double newTestAttribute = 0; //unused; test new parameters }; } diff --git a/filter/xbrz.cpp b/filter/xbrz.cpp index 9ab43618..752009fd 100644 --- a/filter/xbrz.cpp +++ b/filter/xbrz.cpp @@ -1,6 +1,6 @@ // **************************************************************************** // * This file is part of the HqMAME project. It is distributed under * -// * GNU General Public License: http://www.gnu.org/licenses/gpl.html * +// * GNU General Public License: http://www.gnu.org/licenses/gpl-3.0 * // * Copyright (C) Zenju (zenju AT gmx DOT de) - All Rights Reserved * // * * // * Additionally and as a special exception, the author gives permission * @@ -16,48 +16,55 @@ #include "xbrz.h" #include #include -#ifdef unix - #include -#endif #include - namespace { template inline unsigned char getByte(uint32_t val) { return static_cast((val >> (8 * N)) & 0xff); } -inline unsigned char getAlpha(uint32_t val) { return getByte<3>(val); } -inline unsigned char getRed (uint32_t val) { return getByte<2>(val); } -inline unsigned char getGreen(uint32_t val) { return getByte<1>(val); } -inline unsigned char getBlue (uint32_t val) { return getByte<0>(val); } +inline unsigned char getAlpha(uint32_t pix) { return getByte<3>(pix); } +inline unsigned char getRed (uint32_t pix) { return getByte<2>(pix); } +inline unsigned char getGreen(uint32_t pix) { return getByte<1>(pix); } +inline unsigned char getBlue (uint32_t pix) { return getByte<0>(pix); } -template inline -T abs(T value) -{ - //static_assert(std::is_signed::value, "abs() requires signed types"); - return value < 0 ? -value : value; -} +inline uint32_t makePixel( unsigned char r, unsigned char g, unsigned char b) { return (r << 16) | (g << 8) | b; } +inline uint32_t makePixel(unsigned char a, unsigned char r, unsigned char g, unsigned char b) { return (a << 24) | (r << 16) | (g << 8) | b; } -const uint32_t redMask = 0xff0000; -const uint32_t greenMask = 0x00ff00; -const uint32_t blueMask = 0x0000ff; template inline -void alphaBlend(uint32_t& dst, uint32_t col) //blend color over destination with opacity M / N +uint32_t gradientRGB(uint32_t pixFront, uint32_t pixBack) //blend front color with opacity M / N over opaque background: http://en.wikipedia.org/wiki/Alpha_compositing#Alpha_blending { - //static_assert(0 < M && M < N && N <= 256, "possible overflow of (col & byte1Mask) * M + (dst & byte1Mask) * (N - M)"); + static_assert(0 < M && M < N && N <= 1000, ""); - const uint32_t byte1Mask = 0x000000ff; - const uint32_t byte2Mask = 0x0000ff00; - const uint32_t byte3Mask = 0x00ff0000; - const uint32_t byte4Mask = 0xff000000; + auto calcColor = [](unsigned char colFront, unsigned char colBack) -> unsigned char { return (colFront * M + colBack * (N - M)) / N; }; - dst = (byte1Mask & (((col & byte1Mask) * M + (dst & byte1Mask) * (N - M)) / N)) | // - (byte2Mask & (((col & byte2Mask) * M + (dst & byte2Mask) * (N - M)) / N)) | //this works because next higher 8 bits are free - (byte3Mask & (((col & byte3Mask) * M + (dst & byte3Mask) * (N - M)) / N)) | // - (byte4Mask & (((((col & byte4Mask) >> 8) * M + ((dst & byte4Mask) >> 8) * (N - M)) / N) << 8)); //next 8 bits are not free, so shift - //the last row operating on a potential alpha channel costs only ~1% perf => negligible! + return makePixel(calcColor(getRed (pixFront), getRed (pixBack)), + calcColor(getGreen(pixFront), getGreen(pixBack)), + calcColor(getBlue (pixFront), getBlue (pixBack))); +} + + +template inline +uint32_t gradientARGB(uint32_t pixFront, uint32_t pixBack) //find intermediate color between two colors with alpha channels (=> NO alpha blending!!!) +{ + static_assert(0 < M && M < N && N <= 1000, ""); + + const unsigned int weightFront = getAlpha(pixFront) * M; + const unsigned int weightBack = getAlpha(pixBack) * (N - M); + const unsigned int weightSum = weightFront + weightBack; + if (weightSum == 0) + return 0; + + auto calcColor = [=](unsigned char colFront, unsigned char colBack) + { + return static_cast((colFront * weightFront + colBack * weightBack) / weightSum); + }; + + return makePixel(static_cast(weightSum / N), + calcColor(getRed (pixFront), getRed (pixBack)), + calcColor(getGreen(pixFront), getGreen(pixBack)), + calcColor(getBlue (pixFront), getBlue (pixBack))); } @@ -73,15 +80,6 @@ void alphaBlend(uint32_t& dst, uint32_t col) //blend color over destination with // -//inline -//uint32_t alphaBlend2(uint32_t pix1, uint32_t pix2, double alpha) -//{ -// return (redMask & static_cast((pix1 & redMask ) * alpha + (pix2 & redMask ) * (1 - alpha))) | -// (greenMask & static_cast((pix1 & greenMask) * alpha + (pix2 & greenMask) * (1 - alpha))) | -// (blueMask & static_cast((pix1 & blueMask ) * alpha + (pix2 & blueMask ) * (1 - alpha))); -//} - - uint32_t* byteAdvance( uint32_t* ptr, int bytes) { return reinterpret_cast< uint32_t*>(reinterpret_cast< char*>(ptr) + bytes); } const uint32_t* byteAdvance(const uint32_t* ptr, int bytes) { return reinterpret_cast(reinterpret_cast(ptr) + bytes); } @@ -164,199 +162,6 @@ template inline T square(T value) { return value * value; } -/* -inline -void rgbtoLuv(uint32_t c, double& L, double& u, double& v) -{ - //http://www.easyrgb.com/index.php?X=MATH&H=02#text2 - double r = getRed (c) / 255.0; - double g = getGreen(c) / 255.0; - double b = getBlue (c) / 255.0; - - if ( r > 0.04045 ) - r = std::pow(( ( r + 0.055 ) / 1.055 ) , 2.4); - else - r /= 12.92; - if ( g > 0.04045 ) - g = std::pow(( ( g + 0.055 ) / 1.055 ) , 2.4); - else - g /= 12.92; - if ( b > 0.04045 ) - b = std::pow(( ( b + 0.055 ) / 1.055 ) , 2.4); - else - b /= 12.92; - - r *= 100; - g *= 100; - b *= 100; - - double x = 0.4124564 * r + 0.3575761 * g + 0.1804375 * b; - double y = 0.2126729 * r + 0.7151522 * g + 0.0721750 * b; - double z = 0.0193339 * r + 0.1191920 * g + 0.9503041 * b; - //--------------------- - double var_U = 4 * x / ( x + 15 * y + 3 * z ); - double var_V = 9 * y / ( x + 15 * y + 3 * z ); - double var_Y = y / 100; - - if ( var_Y > 0.008856 ) var_Y = std::pow(var_Y , 1.0/3 ); - else var_Y = 7.787 * var_Y + 16.0 / 116; - - const double ref_X = 95.047; //Observer= 2°, Illuminant= D65 - const double ref_Y = 100.000; - const double ref_Z = 108.883; - - const double ref_U = ( 4 * ref_X ) / ( ref_X + ( 15 * ref_Y ) + ( 3 * ref_Z ) ); - const double ref_V = ( 9 * ref_Y ) / ( ref_X + ( 15 * ref_Y ) + ( 3 * ref_Z ) ); - - L = ( 116 * var_Y ) - 16; - u = 13 * L * ( var_U - ref_U ); - v = 13 * L * ( var_V - ref_V ); -} -*/ - -inline -void rgbtoLab(uint32_t c, unsigned char& L, signed char& A, signed char& B) -{ - //code: http://www.easyrgb.com/index.php?X=MATH - //test: http://www.workwithcolor.com/color-converter-01.htm - //------RGB to XYZ------ - double r = getRed (c) / 255.0; - double g = getGreen(c) / 255.0; - double b = getBlue (c) / 255.0; - - r = r > 0.04045 ? std::pow(( r + 0.055 ) / 1.055, 2.4) : r / 12.92; - r = g > 0.04045 ? std::pow(( g + 0.055 ) / 1.055, 2.4) : g / 12.92; - r = b > 0.04045 ? std::pow(( b + 0.055 ) / 1.055, 2.4) : b / 12.92; - - r *= 100; - g *= 100; - b *= 100; - - double x = 0.4124564 * r + 0.3575761 * g + 0.1804375 * b; - double y = 0.2126729 * r + 0.7151522 * g + 0.0721750 * b; - double z = 0.0193339 * r + 0.1191920 * g + 0.9503041 * b; - //------XYZ to Lab------ - const double refX = 95.047; // - const double refY = 100.000; //Observer= 2°, Illuminant= D65 - const double refZ = 108.883; // - double var_X = x / refX; - double var_Y = y / refY; - double var_Z = z / refZ; - - var_X = var_X > 0.008856 ? std::pow(var_X, 1.0 / 3) : 7.787 * var_X + 4.0 / 29; - var_Y = var_Y > 0.008856 ? std::pow(var_Y, 1.0 / 3) : 7.787 * var_Y + 4.0 / 29; - var_Z = var_Z > 0.008856 ? std::pow(var_Z, 1.0 / 3) : 7.787 * var_Z + 4.0 / 29; - - L = static_cast(116 * var_Y - 16); - A = static_cast< signed char>(500 * (var_X - var_Y)); - B = static_cast< signed char>(200 * (var_Y - var_Z)); -}; - - -inline -double distLAB(uint32_t pix1, uint32_t pix2) -{ - unsigned char L1 = 0; //[0, 100] - signed char a1 = 0; //[-128, 127] - signed char b1 = 0; //[-128, 127] - rgbtoLab(pix1, L1, a1, b1); - - unsigned char L2 = 0; - signed char a2 = 0; - signed char b2 = 0; - rgbtoLab(pix2, L2, a2, b2); - - //----------------------------- - //http://www.easyrgb.com/index.php?X=DELT - - //Delta E/CIE76 - return std::sqrt(square(1.0 * L1 - L2) + - square(1.0 * a1 - a2) + - square(1.0 * b1 - b2)); -} - - -/* -inline -void rgbtoHsl(uint32_t c, double& h, double& s, double& l) -{ - //http://www.easyrgb.com/index.php?X=MATH&H=18#text18 - const int r = getRed (c); - const int g = getGreen(c); - const int b = getBlue (c); - - const int varMin = numeric::min(r, g, b); - const int varMax = numeric::max(r, g, b); - const int delMax = varMax - varMin; - - l = (varMax + varMin) / 2.0 / 255.0; - - if (delMax == 0) //gray, no chroma... - { - h = 0; - s = 0; - } - else - { - s = l < 0.5 ? - delMax / (1.0 * varMax + varMin) : - delMax / (2.0 * 255 - varMax - varMin); - - double delR = ((varMax - r) / 6.0 + delMax / 2.0) / delMax; - double delG = ((varMax - g) / 6.0 + delMax / 2.0) / delMax; - double delB = ((varMax - b) / 6.0 + delMax / 2.0) / delMax; - - if (r == varMax) - h = delB - delG; - else if (g == varMax) - h = 1 / 3.0 + delR - delB; - else if (b == varMax) - h = 2 / 3.0 + delG - delR; - - if (h < 0) - h += 1; - if (h > 1) - h -= 1; - } -} - -inline -double distHSL(uint32_t pix1, uint32_t pix2, double lightningWeight) -{ - double h1 = 0; - double s1 = 0; - double l1 = 0; - rgbtoHsl(pix1, h1, s1, l1); - double h2 = 0; - double s2 = 0; - double l2 = 0; - rgbtoHsl(pix2, h2, s2, l2); - - //HSL is in cylindric coordinatates where L represents height, S radius, H angle, - //however we interpret the cylinder as a bi-conic solid with top/bottom radius 0, middle radius 1 - assert(0 <= h1 && h1 <= 1); - assert(0 <= h2 && h2 <= 1); - - double r1 = l1 < 0.5 ? - l1 * 2 : - 2 - l1 * 2; - - double x1 = r1 * s1 * std::cos(h1 * 2 * numeric::pi); - double y1 = r1 * s1 * std::sin(h1 * 2 * numeric::pi); - double z1 = l1; - - double r2 = l2 < 0.5 ? - l2 * 2 : - 2 - l2 * 2; - - double x2 = r2 * s2 * std::cos(h2 * 2 * numeric::pi); - double y2 = r2 * s2 * std::sin(h2 * 2 * numeric::pi); - double z2 = l2; - - return 255 * std::sqrt(square(x1 - x2) + square(y1 - y2) + square(lightningWeight * (z1 - z2))); -} -*/ - inline double distRGB(uint32_t pix1, uint32_t pix2) @@ -370,19 +175,6 @@ double distRGB(uint32_t pix1, uint32_t pix2) } -inline -double distNonLinearRGB(uint32_t pix1, uint32_t pix2) -{ - //non-linear rgb: http://www.compuphase.com/cmetric.htm - const double r_diff = static_cast(getRed (pix1)) - getRed (pix2); - const double g_diff = static_cast(getGreen(pix1)) - getGreen(pix2); - const double b_diff = static_cast(getBlue (pix1)) - getBlue (pix2); - - const double r_avg = (static_cast(getRed(pix1)) + getRed(pix2)) / 2; - return std::sqrt((2 + r_avg / 255) * square(r_diff) + 4 * square(g_diff) + (2 + (255 - r_avg) / 255) * square(b_diff)); -} - - inline double distYCbCr(uint32_t pix1, uint32_t pix2, double lumaWeight) { @@ -413,6 +205,16 @@ double distYCbCr(uint32_t pix1, uint32_t pix2, double lumaWeight) struct DistYCbCrBuffer //30% perf boost compared to distYCbCr()! { public: + static double dist(uint32_t pix1, uint32_t pix2) + { +#if defined _MSC_VER && _MSC_VER < 1900 +#error function scope static initialization is not yet thread-safe! +#endif + static const DistYCbCrBuffer inst; + return inst.distImpl(pix1, pix2); + } + +private: DistYCbCrBuffer() : buffer(256 * 256 * 256) { for (uint32_t i = 0; i < 256 * 256 * 256; ++i) //startup time: 114 ms on Intel Core i5 (four cores) @@ -436,7 +238,7 @@ public: } } - double dist(uint32_t pix1, uint32_t pix2) const + double distImpl(uint32_t pix1, uint32_t pix2) const { //if (pix1 == pix2) -> 8% perf degradation! // return 0; @@ -452,44 +254,8 @@ public: (( b_diff + 255) / 2)]; } -private: - std::vector buffer; //consumes 64 MB memory; using double is 2% faster, but takes 128 MB -} distYCbCrBuffer; - - -inline -double distYUV(uint32_t pix1, uint32_t pix2, double luminanceWeight) -{ - //perf: it's not worthwhile to buffer the YUV-conversion, the direct code is faster by ~ 6% - //since RGB -> YUV conversion is essentially a matrix multiplication, we can calculate the RGB diff before the conversion (distributive property) - const double r_diff = static_cast(getRed (pix1)) - getRed (pix2); - const double g_diff = static_cast(getGreen(pix1)) - getGreen(pix2); - const double b_diff = static_cast(getBlue (pix1)) - getBlue (pix2); - - //http://en.wikipedia.org/wiki/YUV#Conversion_to.2Ffrom_RGB - const double w_b = 0.114; - const double w_r = 0.299; - const double w_g = 1 - w_r - w_b; - - const double u_max = 0.436; - const double v_max = 0.615; - - const double scale_u = u_max / (1 - w_b); - const double scale_v = v_max / (1 - w_r); - - double y = w_r * r_diff + w_g * g_diff + w_b * b_diff;//value range: 255 * [-1, 1] - double u = scale_u * (b_diff - y); //value range: 255 * 2 * u_max * [-1, 1] - double v = scale_v * (r_diff - y); //value range: 255 * 2 * v_max * [-1, 1] - -#ifndef NDEBUG - const double eps = 0.5; -#endif - assert(abs(y) <= 255 + eps); - assert(abs(u) <= 255 * 2 * u_max + eps); - assert(abs(v) <= 255 * 2 * v_max + eps); - - return std::sqrt(square(luminanceWeight * y) + square(u) + square(v)); -} + std::vector buffer; //consumes 64 MB memory; using double is only 2% faster, but takes 128 MB +}; enum BlendType @@ -517,7 +283,6 @@ struct Kernel_4x4 //kernel for preprocessing step /**/m, n, o, p; }; -#define cdist(pix1, pix2) ColorDistance::dist((pix1), (pix2), cfg.luminanceWeight_) /* input kernel area naming convention: ----------------- @@ -542,11 +307,11 @@ BlendResult preProcessCorners(const Kernel_4x4& ker, const xbrz::ScalerCfg& cfg) ker.g == ker.k)) return result; - //auto dist = [&](uint32_t pix1, uint32_t pix2) { return ColorDistance::dist(pix1, pix2, cfg.luminanceWeight_); }; + auto dist = [&](uint32_t pix1, uint32_t pix2) { return ColorDistance::dist(pix1, pix2, cfg.luminanceWeight); }; const int weight = 4; - double jg = cdist(ker.i, ker.f) + cdist(ker.f, ker.c) + cdist(ker.n, ker.k) + cdist(ker.k, ker.h) + weight * cdist(ker.j, ker.g); - double fk = cdist(ker.e, ker.j) + cdist(ker.j, ker.o) + cdist(ker.b, ker.g) + cdist(ker.g, ker.l) + weight * cdist(ker.f, ker.k); + double jg = dist(ker.i, ker.f) + dist(ker.f, ker.c) + dist(ker.n, ker.k) + dist(ker.k, ker.h) + weight * dist(ker.j, ker.g); + double fk = dist(ker.e, ker.j) + dist(ker.j, ker.o) + dist(ker.b, ker.g) + dist(ker.g, ker.l) + weight * dist(ker.f, ker.k); if (jg < fk) //test sample: 70% of values max(jg, fk) / min(jg, fk) are between 1.1 and 3.7 with median being 1.8 { @@ -625,11 +390,10 @@ template <> inline unsigned char rotateBlendInfo(unsigned char b) { ret #ifndef NDEBUG int debugPixelX = -1; - int debugPixelY = 84; - bool breakIntoDebugger = false; + int debugPixelY = 12; + __declspec(thread) bool breakIntoDebugger = false; #endif -#define eq(pix1, pix2) (ColorDistance::dist((pix1), (pix2), cfg.luminanceWeight_) < cfg.equalColorTolerance_) /* input kernel area naming convention: @@ -658,42 +422,44 @@ void blendPixel(const Kernel_3x3& ker, #define h get_h(ker) #define i get_i(ker) -#ifdef WIN32 -# ifndef NDEBUG +#ifndef NDEBUG if (breakIntoDebugger) __debugbreak(); //__asm int 3; -# endif #endif const unsigned char blend = rotateBlendInfo(blendInfo); if (getBottomR(blend) >= BLEND_NORMAL) { - //auto eq = [&](uint32_t pix1, uint32_t pix2) { return ColorDistance::dist(pix1, pix2, cfg.luminanceWeight_) < cfg.equalColorTolerance_; }; - //auto dist = [&](uint32_t pix1, uint32_t pix2) { return ColorDistance::dist(pix1, pix2, cfg.luminanceWeight_); }; + auto eq = [&](uint32_t pix1, uint32_t pix2) { return ColorDistance::dist(pix1, pix2, cfg.luminanceWeight) < cfg.equalColorTolerance; }; + auto dist = [&](uint32_t pix1, uint32_t pix2) { return ColorDistance::dist(pix1, pix2, cfg.luminanceWeight); }; - bool doLineBlend = true; - if (getBottomR(blend) >= BLEND_DOMINANT) - doLineBlend = true; - //make sure there is no second blending in an adjacent rotation for this pixel: handles insular pixels, mario eyes - else if (getTopR(blend) != BLEND_NONE && !eq(e, g)) //but support double-blending for 90° corners - doLineBlend = false; - else if(getBottomL(blend) != BLEND_NONE && !eq(e, c)) - doLineBlend = false; - //no full blending for L-shapes; blend corner only (handles "mario mushroom eyes") - else if (!eq(e, i) && eq(g, h) && eq(h , i) && eq(i, f) && eq(f, c)) - doLineBlend = false; - else - doLineBlend = true; + const bool doLineBlend = [&]() -> bool + { + if (getBottomR(blend) >= BLEND_DOMINANT) + return true; - const uint32_t px = cdist(e, f) <= cdist(e, h) ? f : h; //choose most similar color + //make sure there is no second blending in an adjacent rotation for this pixel: handles insular pixels, mario eyes + if (getTopR(blend) != BLEND_NONE && !eq(e, g)) //but support double-blending for 90° corners + return false; + if (getBottomL(blend) != BLEND_NONE && !eq(e, c)) + return false; + + //no full blending for L-shapes; blend corner only (handles "mario mushroom eyes") + if (!eq(e, i) && eq(g, h) && eq(h , i) && eq(i, f) && eq(f, c)) + return false; + + return true; + }(); + + const uint32_t px = dist(e, f) <= dist(e, h) ? f : h; //choose most similar color OutputMatrix out(target, trgWidth); if (doLineBlend) { - const double fg = cdist(f, g); //test sample: 70% of values max(fg, hc) / min(fg, hc) are between 1.1 and 3.7 with median being 1.9 - const double hc = cdist(h, c); // + const double fg = dist(f, g); //test sample: 70% of values max(fg, hc) / min(fg, hc) are between 1.1 and 3.7 with median being 1.9 + const double hc = dist(h, c); // const bool haveShallowLine = cfg.steepDirectionThreshold * fg <= hc && e != g && d != g; const bool haveSteepLine = cfg.steepDirectionThreshold * hc <= fg && e != c && b != c; @@ -744,7 +510,7 @@ void scaleImage(const uint32_t* src, uint32_t* trg, int srcWidth, int srcHeight, const int bufferSize = srcWidth; unsigned char* preProcBuffer = reinterpret_cast(trg + yLast * Scaler::scale * trgWidth) - bufferSize; std::fill(preProcBuffer, preProcBuffer + bufferSize, 0); - //static_assert(BLEND_NONE == 0, ""); + static_assert(BLEND_NONE == 0, ""); //initialize preprocessing buffer for first row of current stripe: detect upper left and right corner blending //this cannot be optimized for adjacent processing stripes; we must not allow for a memory race condition! @@ -900,112 +666,127 @@ void scaleImage(const uint32_t* src, uint32_t* trg, int srcWidth, int srcHeight, //------------------------------------------------------------------------------------ -struct Scaler2x +template +struct Scaler2x : public ColorGradient { static const int scale = 2; + template //bring template function into scope for GCC + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) { ColorGradient::template alphaGrad(pixBack, pixFront); } + + template static void blendLineShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); } template static void blendLineSteep(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<0, scale - 1>(), col); - alphaBlend<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); } template static void blendLineSteepAndShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<1, 0>(), col); - alphaBlend<1, 4>(out.template ref<0, 1>(), col); - alphaBlend<5, 6>(out.template ref<1, 1>(), col); //[!] fixes 7/8 used in xBR + alphaGrad<1, 4>(out.template ref<1, 0>(), col); + alphaGrad<1, 4>(out.template ref<0, 1>(), col); + alphaGrad<5, 6>(out.template ref<1, 1>(), col); //[!] fixes 7/8 used in xBR } template static void blendLineDiagonal(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 2>(out.template ref<1, 1>(), col); + alphaGrad<1, 2>(out.template ref<1, 1>(), col); } template static void blendCorner(uint32_t col, OutputMatrix& out) { //model a round corner - alphaBlend<21, 100>(out.template ref<1, 1>(), col); //exact: 1 - pi/4 = 0.2146018366 + alphaGrad<21, 100>(out.template ref<1, 1>(), col); //exact: 1 - pi/4 = 0.2146018366 } }; -struct Scaler3x +template +struct Scaler3x : public ColorGradient { static const int scale = 3; + template //bring template function into scope for GCC + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) { ColorGradient::template alphaGrad(pixBack, pixFront); } + + template static void blendLineShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); out.template ref() = col; } template static void blendLineSteep(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<0, scale - 1>(), col); - alphaBlend<1, 4>(out.template ref<2, scale - 2>(), col); + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<2, scale - 2>(), col); - alphaBlend<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); out.template ref<2, scale - 1>() = col; } template static void blendLineSteepAndShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<2, 0>(), col); - alphaBlend<1, 4>(out.template ref<0, 2>(), col); - alphaBlend<3, 4>(out.template ref<2, 1>(), col); - alphaBlend<3, 4>(out.template ref<1, 2>(), col); + alphaGrad<1, 4>(out.template ref<2, 0>(), col); + alphaGrad<1, 4>(out.template ref<0, 2>(), col); + alphaGrad<3, 4>(out.template ref<2, 1>(), col); + alphaGrad<3, 4>(out.template ref<1, 2>(), col); out.template ref<2, 2>() = col; } template static void blendLineDiagonal(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 8>(out.template ref<1, 2>(), col); - alphaBlend<1, 8>(out.template ref<2, 1>(), col); - alphaBlend<7, 8>(out.template ref<2, 2>(), col); + alphaGrad<1, 8>(out.template ref<1, 2>(), col); //conflict with other rotations for this odd scale + alphaGrad<1, 8>(out.template ref<2, 1>(), col); + alphaGrad<7, 8>(out.template ref<2, 2>(), col); // } template static void blendCorner(uint32_t col, OutputMatrix& out) { //model a round corner - alphaBlend<45, 100>(out.template ref<2, 2>(), col); //exact: 0.4545939598 - //alphaBlend<14, 1000>(out.template ref<2, 1>(), col); //0.01413008627 -> negligible - //alphaBlend<14, 1000>(out.template ref<1, 2>(), col); //0.01413008627 + alphaGrad<45, 100>(out.template ref<2, 2>(), col); //exact: 0.4545939598 + //alphaGrad<7, 256>(out.template ref<2, 1>(), col); //0.02826017254 -> negligible + avoid conflicts with other rotations for this odd scale + //alphaGrad<7, 256>(out.template ref<1, 2>(), col); //0.02826017254 } }; -struct Scaler4x +template +struct Scaler4x : public ColorGradient { static const int scale = 4; + template //bring template function into scope for GCC + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) { ColorGradient::template alphaGrad(pixBack, pixFront); } + + template static void blendLineShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); out.template ref() = col; out.template ref() = col; @@ -1014,11 +795,11 @@ struct Scaler4x template static void blendLineSteep(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<0, scale - 1>(), col); - alphaBlend<1, 4>(out.template ref<2, scale - 2>(), col); + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<2, scale - 2>(), col); - alphaBlend<3, 4>(out.template ref<1, scale - 1>(), col); - alphaBlend<3, 4>(out.template ref<3, scale - 2>(), col); + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<3, 4>(out.template ref<3, scale - 2>(), col); out.template ref<2, scale - 1>() = col; out.template ref<3, scale - 1>() = col; @@ -1027,19 +808,23 @@ struct Scaler4x template static void blendLineSteepAndShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<3, 4>(out.template ref<3, 1>(), col); - alphaBlend<3, 4>(out.template ref<1, 3>(), col); - alphaBlend<1, 4>(out.template ref<3, 0>(), col); - alphaBlend<1, 4>(out.template ref<0, 3>(), col); - alphaBlend<1, 3>(out.template ref<2, 2>(), col); //[!] fixes 1/4 used in xBR - out.template ref<3, 3>() = out.template ref<3, 2>() = out.template ref<2, 3>() = col; + alphaGrad<3, 4>(out.template ref<3, 1>(), col); + alphaGrad<3, 4>(out.template ref<1, 3>(), col); + alphaGrad<1, 4>(out.template ref<3, 0>(), col); + alphaGrad<1, 4>(out.template ref<0, 3>(), col); + + alphaGrad<1, 3>(out.template ref<2, 2>(), col); //[!] fixes 1/4 used in xBR + + out.template ref<3, 3>() = col; + out.template ref<3, 2>() = col; + out.template ref<2, 3>() = col; } template static void blendLineDiagonal(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 2>(out.template ref(), col); - alphaBlend<1, 2>(out.template ref(), col); + alphaGrad<1, 2>(out.template ref(), col); + alphaGrad<1, 2>(out.template ref(), col); out.template ref() = col; } @@ -1047,26 +832,31 @@ struct Scaler4x static void blendCorner(uint32_t col, OutputMatrix& out) { //model a round corner - alphaBlend<68, 100>(out.template ref<3, 3>(), col); //exact: 0.6848532563 - alphaBlend< 9, 100>(out.template ref<3, 2>(), col); //0.08677704501 - alphaBlend< 9, 100>(out.template ref<2, 3>(), col); //0.08677704501 + alphaGrad<68, 100>(out.template ref<3, 3>(), col); //exact: 0.6848532563 + alphaGrad< 9, 100>(out.template ref<3, 2>(), col); //0.08677704501 + alphaGrad< 9, 100>(out.template ref<2, 3>(), col); //0.08677704501 } }; -struct Scaler5x +template +struct Scaler5x : public ColorGradient { static const int scale = 5; + template //bring template function into scope for GCC + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) { ColorGradient::template alphaGrad(pixBack, pixFront); } + + template static void blendLineShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); out.template ref() = col; out.template ref() = col; @@ -1077,12 +867,12 @@ struct Scaler5x template static void blendLineSteep(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<0, scale - 1>(), col); - alphaBlend<1, 4>(out.template ref<2, scale - 2>(), col); - alphaBlend<1, 4>(out.template ref<4, scale - 3>(), col); + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<2, scale - 2>(), col); + alphaGrad<1, 4>(out.template ref<4, scale - 3>(), col); - alphaBlend<3, 4>(out.template ref<1, scale - 1>(), col); - alphaBlend<3, 4>(out.template ref<3, scale - 2>(), col); + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<3, 4>(out.template ref<3, scale - 2>(), col); out.template ref<2, scale - 1>() = col; out.template ref<3, scale - 1>() = col; @@ -1093,34 +883,33 @@ struct Scaler5x template static void blendLineSteepAndShallow(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 4>(out.template ref<0, scale - 1>(), col); - alphaBlend<1, 4>(out.template ref<2, scale - 2>(), col); - alphaBlend<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<2, scale - 2>(), col); + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<1, 4>(out.template ref(), col); - alphaBlend<3, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + + alphaGrad<2, 3>(out.template ref<3, 3>(), col); out.template ref<2, scale - 1>() = col; out.template ref<3, scale - 1>() = col; + out.template ref<4, scale - 1>() = col; out.template ref() = col; out.template ref() = col; - - out.template ref<4, scale - 1>() = col; - - alphaBlend<2, 3>(out.template ref<3, 3>(), col); } template static void blendLineDiagonal(uint32_t col, OutputMatrix& out) { - alphaBlend<1, 8>(out.template ref(), col); - alphaBlend<1, 8>(out.template ref(), col); - alphaBlend<1, 8>(out.template ref(), col); + alphaGrad<1, 8>(out.template ref(), col); //conflict with other rotations for this odd scale + alphaGrad<1, 8>(out.template ref(), col); + alphaGrad<1, 8>(out.template ref(), col); // - alphaBlend<7, 8>(out.template ref<4, 3>(), col); - alphaBlend<7, 8>(out.template ref<3, 4>(), col); + alphaGrad<7, 8>(out.template ref<4, 3>(), col); + alphaGrad<7, 8>(out.template ref<3, 4>(), col); out.template ref<4, 4>() = col; } @@ -1129,11 +918,110 @@ struct Scaler5x static void blendCorner(uint32_t col, OutputMatrix& out) { //model a round corner - alphaBlend<86, 100>(out.template ref<4, 4>(), col); //exact: 0.8631434088 - alphaBlend<23, 100>(out.template ref<4, 3>(), col); //0.2306749731 - alphaBlend<23, 100>(out.template ref<3, 4>(), col); //0.2306749731 - //alphaBlend<8, 1000>(out.template ref<4, 2>(), col); //0.008384061834 -> negligible - //alphaBlend<8, 1000>(out.template ref<2, 4>(), col); //0.008384061834 + alphaGrad<86, 100>(out.template ref<4, 4>(), col); //exact: 0.8631434088 + alphaGrad<23, 100>(out.template ref<4, 3>(), col); //0.2306749731 + alphaGrad<23, 100>(out.template ref<3, 4>(), col); //0.2306749731 + //alphaGrad<1, 64>(out.template ref<4, 2>(), col); //0.01676812367 -> negligible + avoid conflicts with other rotations for this odd scale + //alphaGrad<1, 64>(out.template ref<2, 4>(), col); //0.01676812367 + } +}; + + +template +struct Scaler6x : public ColorGradient +{ + static const int scale = 6; + + template //bring template function into scope for GCC + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) { ColorGradient::template alphaGrad(pixBack, pixFront); } + + + template + static void blendLineShallow(uint32_t col, OutputMatrix& out) + { + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + + alphaGrad<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + + out.template ref() = col; + out.template ref() = col; + out.template ref() = col; + out.template ref() = col; + + out.template ref() = col; + out.template ref() = col; + } + + template + static void blendLineSteep(uint32_t col, OutputMatrix& out) + { + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<2, scale - 2>(), col); + alphaGrad<1, 4>(out.template ref<4, scale - 3>(), col); + + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<3, 4>(out.template ref<3, scale - 2>(), col); + alphaGrad<3, 4>(out.template ref<5, scale - 3>(), col); + + out.template ref<2, scale - 1>() = col; + out.template ref<3, scale - 1>() = col; + out.template ref<4, scale - 1>() = col; + out.template ref<5, scale - 1>() = col; + + out.template ref<4, scale - 2>() = col; + out.template ref<5, scale - 2>() = col; + } + + template + static void blendLineSteepAndShallow(uint32_t col, OutputMatrix& out) + { + alphaGrad<1, 4>(out.template ref<0, scale - 1>(), col); + alphaGrad<1, 4>(out.template ref<2, scale - 2>(), col); + alphaGrad<3, 4>(out.template ref<1, scale - 1>(), col); + alphaGrad<3, 4>(out.template ref<3, scale - 2>(), col); + + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<1, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + alphaGrad<3, 4>(out.template ref(), col); + + out.template ref<2, scale - 1>() = col; + out.template ref<3, scale - 1>() = col; + out.template ref<4, scale - 1>() = col; + out.template ref<5, scale - 1>() = col; + + out.template ref<4, scale - 2>() = col; + out.template ref<5, scale - 2>() = col; + + out.template ref() = col; + out.template ref() = col; + } + + template + static void blendLineDiagonal(uint32_t col, OutputMatrix& out) + { + alphaGrad<1, 2>(out.template ref(), col); + alphaGrad<1, 2>(out.template ref(), col); + alphaGrad<1, 2>(out.template ref(), col); + + out.template ref() = col; + out.template ref() = col; + out.template ref() = col; + } + + template + static void blendCorner(uint32_t col, OutputMatrix& out) + { + //model a round corner + alphaGrad<97, 100>(out.template ref<5, 5>(), col); //exact: 0.9711013910 + alphaGrad<42, 100>(out.template ref<4, 5>(), col); //0.4236372243 + alphaGrad<42, 100>(out.template ref<5, 4>(), col); //0.4236372243 + alphaGrad< 6, 100>(out.template ref<5, 3>(), col); //0.05652034508 + alphaGrad< 6, 100>(out.template ref<3, 5>(), col); //0.05652034508 } }; @@ -1143,7 +1031,7 @@ struct ColorDistanceRGB { static double dist(uint32_t pix1, uint32_t pix2, double luminanceWeight) { - return distYCbCrBuffer.dist(pix1, pix2); + return DistYCbCrBuffer::dist(pix1, pix2); //if (pix1 == pix2) //about 4% perf boost // return 0; @@ -1162,14 +1050,37 @@ struct ColorDistanceARGB 1. if a1 = a2, distance should be: a1 * distYCbCr() 2. if a1 = 0, distance should be: a2 * distYCbCr(black, white) = a2 * 255 - 3. if a1 = 1, distance should be: 255 * (1 - a2) + a2 * distYCbCr() + 3. if a1 = 1, ??? maybe: 255 * (1 - a2) + a2 * distYCbCr() */ - return std::min(a1, a2) * distYCbCrBuffer.dist(pix1, pix2) + 255 * abs(a1 - a2); + //return std::min(a1, a2) * DistYCbCrBuffer::dist(pix1, pix2) + 255 * abs(a1 - a2); + //=> following code is 15% faster: + const double d = DistYCbCrBuffer::dist(pix1, pix2); + if (a1 < a2) + return a1 * d + 255 * (a2 - a1); + else + return a2 * d + 255 * (a1 - a2); - //if (pix1 == pix2) - // return 0; - //return std::min(a1, a2) * distYCbCr(pix1, pix2, luminanceWeight) + 255 * abs(a1 - a2); + //alternative? return std::sqrt(a1 * a2 * square(DistYCbCrBuffer::dist(pix1, pix2)) + square(255 * (a1 - a2))); + } +}; + + +struct ColorGradientRGB +{ + template + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) + { + pixBack = gradientRGB(pixFront, pixBack); + } +}; + +struct ColorGradientARGB +{ + template + static void alphaGrad(uint32_t& pixBack, uint32_t pixFront) + { + pixBack = gradientARGB(pixFront, pixBack); } }; } @@ -1179,39 +1090,35 @@ void xbrz::scale(size_t factor, const uint32_t* src, uint32_t* trg, int srcWidth { switch (colFmt) { -#ifdef WIN32 - case ColorFormat::ARGB:// not Standard C++. -#else - case ARGB: -#endif + case ColorFormat::ARGB: switch (factor) { case 2: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceARGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); case 3: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceARGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); case 4: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceARGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); case 5: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceARGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + case 6: + return scaleImage, ColorDistanceARGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); } break; -#ifdef WIN32 - case ColorFormat::RGB:// not Standard C++. -#else - case RGB: -#endif + case ColorFormat::RGB: switch (factor) { case 2: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceRGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); case 3: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceRGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); case 4: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceRGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); case 5: - return scaleImage(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + return scaleImage, ColorDistanceRGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); + case 6: + return scaleImage, ColorDistanceRGB>(src, trg, srcWidth, srcHeight, cfg, yFirst, yLast); } break; } @@ -1223,18 +1130,10 @@ bool xbrz::equalColorTest(uint32_t col1, uint32_t col2, ColorFormat colFmt, doub { switch (colFmt) { -#ifdef WIN32 - case ColorFormat::ARGB: // not Standard C++. -#else - case ARGB: -#endif + case ColorFormat::ARGB: return ColorDistanceARGB::dist(col1, col2, luminanceWeight) < equalColorTolerance; -#ifdef WIN32 - case ColorFormat::RGB:// not Standard C++. -#else - case RGB: -#endif + case ColorFormat::RGB: return ColorDistanceRGB::dist(col1, col2, luminanceWeight) < equalColorTolerance; } assert(false); diff --git a/filter/xbrz.h b/filter/xbrz.h index afb5828b..fbc19650 100644 --- a/filter/xbrz.h +++ b/filter/xbrz.h @@ -36,12 +36,13 @@ http://board.byuu.org/viewtopic.php?f=10&t=2248 - support multithreading - support 64-bit architectures - support processing image slices +- support scaling up to 6xBRZ */ -enum ColorFormat //from high bits -> low bits, 8 bit per channel +enum class ColorFormat //from high bits -> low bits, 8 bit per channel { - ARGB, //including alpha channel, BGRA byte order on little-endian machines RGB, //8 bit for each red, green, blue, upper 8 bits unused + ARGB, //including alpha channel, BGRA byte order on little-endian machines }; /* @@ -53,9 +54,9 @@ enum ColorFormat //from high bits -> low bits, 8 bit per channel in the target image data if you are using multiple threads for processing each enlarged slice! THREAD-SAFETY: - parts of the same image may be scaled by multiple threads as long as the [yFirst, yLast) ranges do not overlap! - - there is a minor inefficiency for the first row of a slice, so avoid processing single rows only; suggestion: process 6 rows at least + - there is a minor inefficiency for the first row of a slice, so avoid processing single rows only; suggestion: process 8-16 rows at least */ -void scale(size_t factor, //valid range: 2 - 5 +void scale(size_t factor, //valid range: 2 - 6 const uint32_t* src, uint32_t* trg, int srcWidth, int srcHeight, ColorFormat colFmt, const ScalerCfg& cfg = ScalerCfg(), diff --git a/win32/render.cpp b/win32/render.cpp index f3b9fe49..dbcb587e 100644 --- a/win32/render.cpp +++ b/win32/render.cpp @@ -219,6 +219,8 @@ void RenderHQ4X (SSurface Src, SSurface Dst, RECT *rect); void Render2xBRZ(SSurface Src, SSurface Dst, RECT* rect); void Render3xBRZ(SSurface Src, SSurface Dst, RECT* rect); void Render4xBRZ(SSurface Src, SSurface Dst, RECT* rect); +void Render5xBRZ(SSurface Src, SSurface Dst, RECT* rect); +void Render6xBRZ(SSurface Src, SSurface Dst, RECT* rect); void RenderEPXA (SSurface Src, SSurface Dst, RECT *); void RenderEPXB (SSurface Src, SSurface Dst, RECT *); void RenderEPXC (SSurface Src, SSurface Dst, RECT *); @@ -319,6 +321,8 @@ TRenderMethod FilterToMethod(RenderFilter filterID) case FILTER_SIMPLE4X: return RenderSimple4X; case FILTER_HQ4X: return RenderHQ4X; case FILTER_4XBRZ: return Render4xBRZ; + case FILTER_5XBRZ: return Render5xBRZ; + case FILTER_6XBRZ: return Render6xBRZ; } } @@ -357,6 +361,8 @@ const char* GetFilterName(RenderFilter filterID) case FILTER_SIMPLE4X: return "Simple 4X"; case FILTER_HQ4X: return "hq4x"; case FILTER_4XBRZ: return "4xBRZ"; + case FILTER_5XBRZ: return "5xBRZ"; + case FILTER_6XBRZ: return "6xBRZ"; } } @@ -388,6 +394,10 @@ int GetFilterScale(RenderFilter filterID) case FILTER_HQ4X: case FILTER_4XBRZ: return 4; + case FILTER_5XBRZ: + return 5; + case FILTER_6XBRZ: + return 6; } } @@ -409,6 +419,8 @@ bool GetFilterHiResSupport(RenderFilter filterID) case FILTER_2XBRZ: case FILTER_3XBRZ: case FILTER_4XBRZ: + case FILTER_5XBRZ: + case FILTER_6XBRZ: return true; default: @@ -2746,6 +2758,16 @@ void Render4xBRZ(SSurface Src, SSurface Dst, RECT* rect) RenderxBRZ(Src, Dst, rect, 4); } +void Render5xBRZ(SSurface Src, SSurface Dst, RECT* rect) +{ + RenderxBRZ(Src, Dst, rect, 5); +} + +void Render6xBRZ(SSurface Src, SSurface Dst, RECT* rect) +{ + RenderxBRZ(Src, Dst, rect, 6); +} + void RenderxBRZ(SSurface Src, SSurface Dst, RECT* rect, int scalingFactor) { xbrz_thread_data::scalingFactor = scalingFactor; diff --git a/win32/wsnes9x.cpp b/win32/wsnes9x.cpp index d2f46bd9..c5424d17 100644 --- a/win32/wsnes9x.cpp +++ b/win32/wsnes9x.cpp @@ -7541,6 +7541,7 @@ checkUpdateFilterBox2: default: case 2: filter = FILTER_SIMPLE2X; break; case 3: filter = FILTER_SIMPLE3X; break; case 4: filter = FILTER_SIMPLE4X; break; + case 5: case 6: filter = FILTER_SIMPLE4X; break; } strcpy(textOriginal, GetFilterName(filter)); } diff --git a/win32/wsnes9x.h b/win32/wsnes9x.h index 3ff8f01e..7b514f7c 100644 --- a/win32/wsnes9x.h +++ b/win32/wsnes9x.h @@ -271,6 +271,8 @@ enum RenderFilter{ FILTER_2XBRZ, FILTER_3XBRZ, FILTER_4XBRZ, + FILTER_5XBRZ, + FILTER_6XBRZ, NUM_FILTERS };