Addition of the scale feature, menu control and option added.

optional code which can be selected by the define SCALE

Makefile

@@ -13,14 +13,16 @@ objects = src/bonus/bonus.o \
 	src/menu/gamemenu.o src/menu/mainmenu.o src/menu/menu.o \
 	src/menu/menuutil.o src/menu/setupmenu.o \
 	src/player/bird.o src/player/player.o src/player/playerframe.o \
-	src/main.o src/movable.o src/planet.o src/scene.o src/util.o
+	src/main.o src/movable.o src/planet.o src/scene.o src/util.o \
+	src/scale2x/getopt.o src/scale2x/pixel.o src/scale2x/scale2x.o \
+	src/scale2x/scale3x.o src/scale2x/scalebit.o src/scale2x/simple2x.o
 
 
 OpenJazz: $(objects)
 	cc -Wall -o OpenJazz -lSDL -lstdc++ $(objects)
 
 %.o: %.cpp
-	cc -Wall -DUSE_SOCKETS -Isrc -O2 -c $< -o $@
+	cc -Wall -DUSE_SOCKETS -DSCALE -Isrc -O2 -c $< -o $@
 
 clean:
 	rm -f OpenJazz $(objects)

Makefile.arm

@@ -37,12 +37,14 @@ OBJS =	src/bonus/bonus.o \
 	src/menu/gamemenu.o src/menu/mainmenu.o src/menu/menu.o \
 	src/menu/menuutil.o src/menu/setupmenu.o \
 	src/player/bird.o src/player/player.o src/player/playerframe.o \
-	src/main.o src/movable.o src/planet.o src/scene.o src/util.o
+	src/main.o src/movable.o src/planet.o src/scene.o src/util.o \
+	src/scale2x/getopt.o src/scale2x/pixel.o src/scale2x/scale2x.o \
+	src/scale2x/scale3x.o src/scale2x/scalebit.o src/scale2x/simple2x.o
 
 CXX = $(PREFIX)/bin/$(TARGET)g++
 
 ifeq ($(BUILDTYPE),PANDORA)
-CXXFLAGS += 
+CXXFLAGS += -DSCALE
 LDFLAGS  += -lts
 else
 ifeq ($(BUILDTYPE),WIZ)
@@ -55,13 +57,13 @@ OBJS += src/platforms/wiz.o
 CXXFLAGS += -DGP2X -DFULLSCREEN_ONLY
 LDFLAGS  += -static -lpthread
 else # default linux
-CXXFLAGS += 
+CXXFLAGS += -DSCALE
 LDFLAGS  +=
 endif
 endif
 endif
 
-CXXFLAGS += -g -Wall -O2 -fsigned-char -DUSE_SOCKETS -I$(PREFIX)/include -I$(PREFIX)/include/SDL -Isrc
+CXXFLAGS += -g -Wall -O3 -fsigned-char -DUSE_SOCKETS -I$(PREFIX)/include -I$(PREFIX)/include/SDL -Isrc
 LDFLAGS  += -L$(PREFIX)/lib -lSDL
 
 # Uncomment the following two lines for music (requires libmodplug)

Makefile.debian

@@ -13,10 +13,12 @@ objects = src/bonus/bonus.o \
 	src/menu/gamemenu.o src/menu/mainmenu.o src/menu/menu.o \
 	src/menu/menuutil.o src/menu/setupmenu.o \
 	src/player/bird.o src/player/player.o src/player/playerframe.o \
-	src/main.o src/movable.o src/planet.o src/scene.o src/util.o
+	src/main.o src/movable.o src/planet.o src/scene.o src/util.o \
+	src/scale2x/getopt.o src/scale2x/pixel.o src/scale2x/scale2x.o \
+	src/scale2x/scale3x.o src/scale2x/scalebit.o src/scale2x/simple2x.o
 
 
-CXXFLAGS += -g -Wall -O2 -DUSE_SOCKETS
+CXXFLAGS += -g -Wall -O2 -DUSE_SOCKETS -DSCALE
 
 # Uncomment the following two lines for music (requires libmodplug)
 #CXXFLAGS += -DUSE_MODPLUG `pkg-config --cflags libmodplug`

src/io/gfx/video.cpp

@@ -42,6 +42,8 @@ SDL_Surface * createSurface (unsigned char * pixels, int width, int height) {
 	// Set the surface's palette
 	SDL_SetPalette(ret, SDL_LOGPAL, logicalPalette, 0, 256);
 
+	if (pixels != NULL )
+	{
 		// Upload pixel data to the surface
 		if (SDL_MUSTLOCK(ret)) SDL_LockSurface(ret);
 
@@ -53,6 +55,7 @@ SDL_Surface * createSurface (unsigned char * pixels, int width, int height) {
 
 		// Free redundant pixel data
 		delete[] pixels;
+	}
 
 	return ret;
 
@@ -65,13 +68,29 @@ void createFullscreen () {
 	screen = SDL_SetVideoMode(320, 240, 8,
 		SDL_FULLSCREEN | SDL_SWSURFACE | SDL_HWPALETTE);
 #else
+
+#ifdef SCALE	
+	screen_scaled = SDL_SetVideoMode(screenW*scalar, screenH*scalar, 8,
+		SDL_FULLSCREEN | SDL_DOUBLEBUF | SDL_HWSURFACE | SDL_HWPALETTE);
+
+	if (screen)
+		SDL_FreeSurface(screen);
+
+	screen = createSurface( NULL, screenW, screenH );
+
+	SDL_SetPalette(screen_scaled, SDL_LOGPAL, logicalPalette, 0, 256);
+	SDL_SetPalette(screen_scaled, SDL_PHYSPAL, currentPalette, 0, 256);
+#else
 	screen = SDL_SetVideoMode(screenW, screenH, 8,
 		SDL_FULLSCREEN | SDL_DOUBLEBUF | SDL_HWSURFACE | SDL_HWPALETTE);
-#endif
-	SDL_ShowCursor(SDL_DISABLE);
 		
 	SDL_SetPalette(screen, SDL_LOGPAL, logicalPalette, 0, 256);
 	SDL_SetPalette(screen, SDL_PHYSPAL, currentPalette, 0, 256);		
+#endif
+#endif
+	SDL_ShowCursor(SDL_DISABLE);
+
+
 
 	/* A real 8-bit display is quite likely if the user has the right video
 	card, the right video drivers, the right version of DirectX/whatever, and
@@ -91,12 +110,24 @@ void createFullscreen () {
 #ifndef FULLSCREEN_ONLY
 void createWindow () {
   
+#ifdef SCALE
+	screen_scaled = SDL_SetVideoMode(screenW*scalar, screenH*scalar, 8,
+		SDL_RESIZABLE | SDL_DOUBLEBUF | SDL_HWSURFACE);
+
+	if (screen)
+		SDL_FreeSurface(screen);
+
+	screen = createSurface( NULL, screenW, screenH );
+	
+	SDL_SetPalette(screen_scaled, SDL_LOGPAL, logicalPalette, 0, 256);
+	SDL_SetPalette(screen_scaled, SDL_PHYSPAL, currentPalette, 0, 256);
+#else
 	screen = SDL_SetVideoMode(screenW, screenH, 8,
 		SDL_RESIZABLE | SDL_DOUBLEBUF | SDL_HWSURFACE | SDL_HWPALETTE);
 		
 	SDL_SetPalette(screen, SDL_LOGPAL, logicalPalette, 0, 256);
 	SDL_SetPalette(screen, SDL_PHYSPAL, currentPalette, 0, 256);
-
+#endif
 	SDL_ShowCursor(SDL_ENABLE);
 
 	/* Assume that in windowed mode the palette is being emulated.
@@ -114,9 +145,14 @@ void usePalette (SDL_Color *palette) {
 
 	// Make palette changes invisible until the next draw. Hopefully.
 	clearScreen(SDL_MapRGB(screen->format, 0, 0, 0));
+#ifndef SCALE
 	SDL_Flip(screen);
+#endif
 
 	SDL_SetPalette(screen, SDL_PHYSPAL, palette, 0, 256);
+#ifdef SCALE
+	SDL_SetPalette(screen_scaled, SDL_PHYSPAL, palette, 0, 256);
+#endif
 	currentPalette = palette;
 
 	return;
@@ -160,5 +196,3 @@ void drawRect (int x, int y, int width, int height, int index) {
 	return;
 
 }
-
-

src/io/gfx/video.h

@@ -39,6 +39,10 @@
 
 EXTERN SDL_Surface *screen;
 EXTERN int          viewW, viewH, screenW, screenH;
+EXTERN int          scalar;
+#ifdef SCALE
+EXTERN SDL_Surface *screen_scaled;
+#endif
 #ifndef FULLSCREEN_ONLY
 EXTERN bool         fullscreen;
 #endif

src/main.cpp

@@ -57,6 +57,10 @@ extern int volume_direction;
 extern char KOpenJazzPath[256];
 #endif
 
+#ifdef SCALE
+#include "scale2x/scalebit.h"
+#endif
+
 int loadMain (int argc, char *argv[]) {
 
 	File *file;
@@ -191,6 +195,13 @@ int loadMain (int argc, char *argv[]) {
 		// Read video settings
 		screenW = file->loadShort();
 		screenH = file->loadShort();
+#ifdef SCALE
+		scalar  = file->loadChar();
+#else
+		scalar = 1;
+		file->loadChar();
+#endif
+
 #ifdef FULLSCREEN_ONLY
 		file->loadChar();
 #else
@@ -255,8 +266,11 @@ int loadMain (int argc, char *argv[]) {
 	else createWindow();
 #endif
 
+#ifdef SCALE
+	if (!screen_scaled) {
+#else
 	if (!screen) {
-
+#endif
 		logError("Could not set video mode", SDL_GetError());
 
 		delete characterName;
@@ -279,6 +293,9 @@ int loadMain (int argc, char *argv[]) {
  			logicalPalette[count].b = count;
 
 	restorePalette(screen);
+#ifdef SCALE
+	restorePalette(screen_scaled);
+#endif
 
 	firstPE = NULL;
 
@@ -465,6 +482,9 @@ void freeMain () {
 	SDL_FreeSurface(panelAmmo[3]);
 	SDL_FreeSurface(panelAmmo[4]);
 
+#ifdef SCALE
+	SDL_FreeSurface(screen);
+#endif
 
 	closeAudio();
 
@@ -489,6 +509,12 @@ void freeMain () {
 		// Write video settings
 		file->storeShort(screenW);
 		file->storeShort(screenH);
+#ifdef SCALE
+		file->storeChar(scalar);
+#else
+		file->storeChar(1);
+#endif
+		
 #ifdef FULLSCREEN_ONLY
 		file->storeChar(1);
 #else
@@ -544,9 +570,27 @@ int loop (int type) {
 	SDL_Event event;
 	int prevTicks, ret;
 
-
 	// Show everything that has been drawn so far
+#ifdef SCALE
+	if (scalar>1) {
+		int depth = screen_scaled->format->BytesPerPixel;
+	  
+		scale(scalar, screen_scaled->pixels, screen->w*depth*scalar, screen->pixels, screen->w*depth,
+			depth, screen->w, screen->h );
+			
+		//Simple2x((unsigned char*)screen->pixels, screen->w*depth, 0, (unsigned char*)screen_scaled->pixels, screen->w*depth*2,
+		//	 screen->w, screen->h );
+	}
+	else
+	{
+		SDL_BlitSurface(screen, NULL, screen_scaled, NULL);
+	}
+	    
+	SDL_Flip(screen_scaled);
+#else
 	SDL_Flip(screen);
+#endif
+	
 
 	prevTicks = globalTicks;
 	globalTicks = SDL_GetTicks();
@@ -605,9 +649,20 @@ int loop (int type) {
 
 				screenW = event.resize.w;
 				screenH = event.resize.h;
+#ifdef SCALE
+				screen_scaled = SDL_SetVideoMode(screenW*scalar, screenH*scalar, 8,
+					SDL_RESIZABLE | SDL_DOUBLEBUF | SDL_HWSURFACE |
+					SDL_HWPALETTE);
+					
+				if (screen)
+					SDL_FreeSurface(screen);
+				  
+				screen = SDL_CreateRGBSurface(SDL_HWSURFACE, screenW, screenH, 8, 0, 0, 0, 0);
+#else
 				screen = SDL_SetVideoMode(screenW, screenH, 8,
 					SDL_RESIZABLE | SDL_DOUBLEBUF | SDL_HWSURFACE |
 					SDL_HWPALETTE);
+#endif
 
 				// The absence of a break statement is intentional
 
@@ -615,6 +670,10 @@ int loop (int type) {
 
 				SDL_SetPalette(screen, SDL_LOGPAL, logicalPalette, 0, 256);
 				SDL_SetPalette(screen, SDL_PHYSPAL, currentPalette, 0, 256);
+#ifdef SCALE
+				SDL_SetPalette(screen_scaled, SDL_LOGPAL, logicalPalette, 0, 256);
+				SDL_SetPalette(screen_scaled, SDL_PHYSPAL, currentPalette, 0, 256);
+#endif
 
 				break;
 #endif
@@ -647,6 +706,10 @@ int loop (int type) {
 
 			SDL_SetPalette(screen, SDL_PHYSPAL, shownPalette, 0, 256);
 			
+#ifdef SCALE
+			SDL_SetPalette(screen_scaled, SDL_PHYSPAL, shownPalette, 0, 256);
+#endif
+
 		} else {
 
 			firstPE->apply(shownPalette, true, globalTicks - prevTicks);

src/menu/setupmenu.cpp

@@ -265,6 +265,19 @@ int Menu::setupResolution () {
 	SDL_Rect **resolutions;
 	int dimension, count, maxW, maxH;
 
+#ifdef SCALE
+	int minS = 1;
+	int maxS = 4;
+	#define minD 0
+	#define maxD 2
+	
+	if ( scalar < minS || scalar > maxS )
+		scalar = 1;
+#else
+	#define minD 0
+	#define maxD 1
+#endif
+
 	dimension = 0;
 
 #ifndef FULLSCREEN_ONLY
@@ -276,6 +289,7 @@ int Menu::setupResolution () {
 		resolutions = SDL_ListModes(NULL,
 			SDL_FULLSCREEN | SDL_DOUBLEBUF | SDL_HWSURFACE | SDL_HWPALETTE);
 
+
 #if defined(WIZ) || defined(GP2X)
 	maxW = 320;
 	maxH = 240;
@@ -318,44 +332,84 @@ int Menu::setupResolution () {
 		drawRect(0, screenH - 32, 32, 32, 79);
 
 
+		// X 1
 		fontmn2->showString("x", (screenW >> 2) + 40, screenH >> 1);
 		
+		// X 2
+		fontmn2->showString("x", (screenW >> 2) + 112, screenH >> 1);
+
+		// Width
 		if (dimension == 0) fontmn2->mapPalette(240, 8, 114, 16);
 		
 		fontmn2->showNumber(screenW, (screenW >> 2) + 32, screenH >> 1);
 
-		if (dimension == 0) fontmn2->restorePalette();
+		// Height
+		if (dimension == 0 || dimension == 2) fontmn2->restorePalette();
 		else fontmn2->mapPalette(240, 8, 114, 16);
 		
 		fontmn2->showNumber(screenH, (screenW >> 2) + 104, screenH >> 1);
 		
+		// Scalar
+		if (dimension == 0 || dimension == 1) fontmn2->restorePalette();
+		else fontmn2->mapPalette(240, 8, 114, 16);
+		
+		fontmn2->showNumber(scalar, (screenW >> 2) + 150, screenH >> 1);
+
 		if (dimension != 0) fontmn2->restorePalette();
 
 
 		count = 0;
 
-		if (controls.release(C_LEFT)) dimension = !dimension;
+		if (controls.release(C_LEFT)) {
+		  
+			dimension--;
+
+			if(dimension<minD)
+				dimension = minD;
+		}
+
+		if (controls.release(C_RIGHT)) {
+		  
+			dimension++;
 
-		if (controls.release(C_RIGHT)) dimension = !dimension;
+			if(dimension>maxD)
+				dimension = maxD;
+		}
 
 		if (controls.release(C_UP)) {
 
-			if ((dimension == 0) && (screenW < maxW)) {
+			if ((dimension == 0) && (screenW*scalar < maxW)) {
+
+				while (screenW >= widthOptions[count] && (screenW*scalar)<maxW) count++;
 
-				while (screenW >= widthOptions[count]) count++;
+				if ((widthOptions[count]*scalar) > maxW)
+					count--;
 
 				screenW = widthOptions[count];
 
 			}
 
-			if ((dimension == 1) && (screenH < maxH)) {
+			if ((dimension == 1) && (screenH*scalar < maxH)) {
 
 				while (screenH >= heightOptions[count]) count++;
 
+				if ((heightOptions[count]*scalar) > maxH)
+					count--;
+
 				screenH = heightOptions[count];
 
 			}
+#ifdef SCALE
+			if ((dimension == 2) && (scalar < maxS)) {
 			  
+				if ( screenH*(scalar+1) <= maxH && screenW*(scalar+1) <= maxW )
+				{
+					scalar++;
+					count = 1;
+				}
+
+			}
+#endif
 		}
 
 		if (controls.release(C_DOWN)) {
@@ -382,6 +436,15 @@ int Menu::setupResolution () {
 
 			}
 
+#ifdef SCALE
+			if ((dimension == 2) && (scalar > minS)) {
+			  
+				scalar--;
+				count = 1;
+
+			}
+#endif
+
 		}
 
 		// Check for a resolution change

src/scale2x/getopt.cpp

+/*
+ * This file is part of the Advance project.
+ *
+ * Copyright (C) 2002 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details. 
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#if !HAVE_GETOPT
+
+/* This source is extracted from the DJGPP LIBC library */
+
+#define unconst(var, type) ((type)(var))
+
+/* Copyright (C) 1995 DJ Delorie, see COPYING.DJ for details */
+#include <string.h>
+#include <stdio.h>
+
+int opterr = 1, optind = 1, optopt = 0;
+char *optarg = 0;
+
+#define BADCH (int)'?'
+#define EMSG ""
+
+int getopt(int nargc, char *const nargv[], char *ostr)
+{
+  static const char *place = EMSG; /* option letter processing */
+  char *oli; /* option letter list index */
+  char *p;
+
+  if (!*place)
+  {
+    if (optind >= nargc || *(place = nargv[optind]) != '-')
+    {
+      place = EMSG;
+      return(EOF);
+    }
+    if (place[1] && *++place == '-')
+    {
+      ++optind;
+      place = EMSG;
+      return(EOF);
+    }
+  }
+
+  if ((optopt = (int)*place++) == (int)':'
+      || !(oli = strchr(ostr, optopt)))
+  {
+    /*
+     * if the user didn't specify '-' as an option,
+     * assume it means EOF.
+     */
+    if (optopt == (int)'-')
+      return EOF;
+    if (!*place)
+      ++optind;
+    if (opterr)
+    {
+      if (!(p = strrchr(*nargv, '/')))
+        p = *nargv;
+      else
+        ++p;
+      fprintf(stderr, "%s: illegal option -- %c\n", p, optopt);
+    }
+    return BADCH;
+  }
+  if (*++oli != ':')
+  { /* don't need argument */
+    optarg = NULL;
+    if (!*place)
+      ++optind;
+  }
+  else
+  { /* need an argument */
+    if (*place) /* no white space */
+      optarg = unconst(place, char *);
+    else if (nargc <= ++optind)
+    { /* no arg */
+      place = EMSG;
+      if (!(p = strrchr(*nargv, '/')))
+        p = *nargv;
+      else
+        ++p;
+      if (opterr)
+        fprintf(stderr, "%s: option requires an argument -- %c\n", p, optopt);
+      return BADCH;
+    }
+    else /* white space */
+      optarg = nargv[optind];
+    place = EMSG;
+    ++optind;
+  }
+  return optopt; /* dump back option letter */
+}
+
+#endif
+

src/scale2x/pixel.cpp

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2003 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include "pixel.h"
+
+pixel_t pixel_get(int x, int y, const unsigned char* pix, unsigned slice, unsigned pixel, unsigned dx, unsigned dy, int opt_tes)
+{
+	const unsigned char* p;
+	unsigned i;
+	pixel_t v;
+
+	if (opt_tes) {
+		if (x < 0)
+			x += dx;
+		if (x >= dx)
+			x -= dx;
+		if (y < 0)
+			y += dy;
+		if (y >= dy)
+			y -= dy;
+	} else {
+		if (x < 0)
+			x = 0;
+		if (x >= dx)
+			x = dx - 1;
+		if (y < 0)
+			y = 0;
+		if (y >= dy)
+			y = dy - 1;
+	}
+
+	p = pix + (y * slice) + (x * pixel);
+
+	v = 0;
+	for(i=0;i<pixel;++i)
+		v |= ((pixel_t)p[i]) << (i*8);
+
+	return v;
+}
+
+void pixel_put(int x, int y, unsigned char* pix, unsigned slice, unsigned pixel, unsigned dx, unsigned dy, pixel_t v)
+{
+	unsigned char* p;
+	unsigned i;
+
+	p = pix + (y * slice) + (x * pixel);
+
+	for(i=0;i<pixel;++i) {
+		p[i] = v >> (i*8);
+	}
+}

src/scale2x/pixel.h

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2003 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef __PIXEL_H
+#define __PIXEL_H
+
+typedef unsigned long long pixel_t;
+
+pixel_t pixel_get(int x, int y, const unsigned char* pix, unsigned slice, unsigned pixel, unsigned dx, unsigned dy, int opt_tes);
+void pixel_put(int x, int y, unsigned char* pix, unsigned slice, unsigned pixel, unsigned dx, unsigned dy, pixel_t v);
+
+#endif
+

src/scale2x/portable.h

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2001, 2002, 2003 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details. 
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef __PORTABLE_H
+#define __PORTABLE_H
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+/* ------------------------------------------------------------------------ */
+/* getopt */
+
+#if HAVE_GETOPT_H
+#include <getopt.h>
+#endif
+
+#if HAVE_UNISTD_H
+#include <unistd.h>
+#endif
+
+#if !HAVE_GETOPT
+int getopt(int argc, char * const *argv, const char *options);
+extern char *optarg;
+extern int optind, opterr, optopt;
+#endif
+
+#endif
+

src/scale2x/scale2x.cpp

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2001, 2002, 2003, 2004 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/*
+ * This file contains a C and MMX implementation of the Scale2x effect.
+ *
+ * You can find an high level description of the effect at :
+ *
+ * http://scale2x.sourceforge.net/
+ *
+ * Alternatively at the previous license terms, you are allowed to use this
+ * code in your program with these conditions:
+ * - the program is not used in commercial activities.
+ * - the whole source code of the program is released with the binary.
+ * - derivative works of the program are allowed.
+ */
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "scale2x.h"
+
+#include <assert.h>
+
+/***************************************************************************/
+/* Scale2x C implementation */
+
+/**
+ * Define the macro USE_SCALE_RANDOMWRITE to enable
+ * an optimized version which writes memory in random order.
+ * This version is a little faster if you write in system memory.
+ * But it's a lot slower if you write in video memory.
+ * So, enable it only if you are sure to never write directly in video memory.
+ */
+/* #define USE_SCALE_RANDOMWRITE */
+
+static inline void scale2x_8_def_whole(scale2x_uint8* restrict dst0, scale2x_uint8* restrict dst1, const scale2x_uint8* restrict src0, const scale2x_uint8* restrict src1, const scale2x_uint8* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst0[0] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst0[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+		dst1[0] = src1[0] == src2[0] ? src2[0] : src1[0];
+		dst1[1] = src1[1] == src2[0] ? src2[0] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst0 += 2;
+	dst1 += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst0[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+			dst0[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+			dst1[0] = src1[-1] == src2[0] ? src2[0] : src1[0];
+			dst1[1] = src1[1] == src2[0] ? src2[0] : src1[0];
+		} else {
+			dst0[0] = src1[0];
+			dst0[1] = src1[0];
+			dst1[0] = src1[0];
+			dst1[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst0 += 2;
+		dst1 += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst0[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+		dst0[1] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst1[0] = src1[-1] == src2[0] ? src2[0] : src1[0];
+		dst1[1] = src1[0] == src2[0] ? src2[0] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+	}
+}
+
+static inline void scale2x_8_def_border(scale2x_uint8* restrict dst, const scale2x_uint8* restrict src0, const scale2x_uint8* restrict src1, const scale2x_uint8* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+			dst[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+		dst[1] = src1[0] == src0[0] ? src0[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+}
+
+static inline void scale2x_8_def_center(scale2x_uint8* restrict dst, const scale2x_uint8* restrict src0, const scale2x_uint8* restrict src1, const scale2x_uint8* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0];
+		dst[1] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst[1] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst[1] = src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+}
+
+static inline void scale2x_16_def_whole(scale2x_uint16* restrict dst0, scale2x_uint16* restrict dst1, const scale2x_uint16* restrict src0, const scale2x_uint16* restrict src1, const scale2x_uint16* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst0[0] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst0[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+		dst1[0] = src1[0] == src2[0] ? src2[0] : src1[0];
+		dst1[1] = src1[1] == src2[0] ? src2[0] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst0 += 2;
+	dst1 += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst0[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+			dst0[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+			dst1[0] = src1[-1] == src2[0] ? src2[0] : src1[0];
+			dst1[1] = src1[1] == src2[0] ? src2[0] : src1[0];
+		} else {
+			dst0[0] = src1[0];
+			dst0[1] = src1[0];
+			dst1[0] = src1[0];
+			dst1[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst0 += 2;
+		dst1 += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst0[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+		dst0[1] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst1[0] = src1[-1] == src2[0] ? src2[0] : src1[0];
+		dst1[1] = src1[0] == src2[0] ? src2[0] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+	}
+}
+
+static inline void scale2x_16_def_border(scale2x_uint16* restrict dst, const scale2x_uint16* restrict src0, const scale2x_uint16* restrict src1, const scale2x_uint16* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+			dst[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+		dst[1] = src1[0] == src0[0] ? src0[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+}
+
+static inline void scale2x_16_def_center(scale2x_uint16* restrict dst, const scale2x_uint16* restrict src0, const scale2x_uint16* restrict src1, const scale2x_uint16* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0];
+		dst[1] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst[1] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst[1] = src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+}
+
+static inline void scale2x_32_def_whole(scale2x_uint32* restrict dst0, scale2x_uint32* restrict dst1, const scale2x_uint32* restrict src0, const scale2x_uint32* restrict src1, const scale2x_uint32* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst0[0] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst0[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+		dst1[0] = src1[0] == src2[0] ? src2[0] : src1[0];
+		dst1[1] = src1[1] == src2[0] ? src2[0] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst0 += 2;
+	dst1 += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst0[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+			dst0[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+			dst1[0] = src1[-1] == src2[0] ? src2[0] : src1[0];
+			dst1[1] = src1[1] == src2[0] ? src2[0] : src1[0];
+		} else {
+			dst0[0] = src1[0];
+			dst0[1] = src1[0];
+			dst1[0] = src1[0];
+			dst1[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst0 += 2;
+		dst1 += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst0[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+		dst0[1] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst1[0] = src1[-1] == src2[0] ? src2[0] : src1[0];
+		dst1[1] = src1[0] == src2[0] ? src2[0] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+	}
+}
+
+static inline void scale2x_32_def_border(scale2x_uint32* restrict dst, const scale2x_uint32* restrict src0, const scale2x_uint32* restrict src1, const scale2x_uint32* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0] == src0[0] ? src0[0] : src1[0];
+		dst[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+			dst[1] = src1[1] == src0[0] ? src0[0] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = src1[-1] == src0[0] ? src0[0] : src1[0];
+		dst[1] = src1[0] == src0[0] ? src0[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+}
+
+static inline void scale2x_32_def_center(scale2x_uint32* restrict dst, const scale2x_uint32* restrict src0, const scale2x_uint32* restrict src1, const scale2x_uint32* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0];
+		dst[1] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 2;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst[1] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 2;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst[1] = src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+	}
+}
+
+/**
+ * Scale by a factor of 2 a row of pixels of 8 bits.
+ * The function is implemented in C.
+ * The pixels over the left and right borders are assumed of the same color of
+ * the pixels on the border.
+ * Note that the implementation is optimized to write data sequentially to
+ * maximize the bandwidth on video memory.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows.
+ * It must be at least 2.
+ * \param dst0 First destination row, double length in pixels.
+ * \param dst1 Second destination row, double length in pixels.
+ */
+void scale2x_8_def(scale2x_uint8* dst0, scale2x_uint8* dst1, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_8_def_whole(dst0, dst1, src0, src1, src2, count);
+#else
+	scale2x_8_def_border(dst0, src0, src1, src2, count);
+	scale2x_8_def_border(dst1, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2 a row of pixels of 16 bits.
+ * This function operates like scale2x_8_def() but for 16 bits pixels.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows.
+ * It must be at least 2.
+ * \param dst0 First destination row, double length in pixels.
+ * \param dst1 Second destination row, double length in pixels.
+ */
+void scale2x_16_def(scale2x_uint16* dst0, scale2x_uint16* dst1, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_16_def_whole(dst0, dst1, src0, src1, src2, count);
+#else
+	scale2x_16_def_border(dst0, src0, src1, src2, count);
+	scale2x_16_def_border(dst1, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2 a row of pixels of 32 bits.
+ * This function operates like scale2x_8_def() but for 32 bits pixels.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows.
+ * It must be at least 2.
+ * \param dst0 First destination row, double length in pixels.
+ * \param dst1 Second destination row, double length in pixels.
+ */
+void scale2x_32_def(scale2x_uint32* dst0, scale2x_uint32* dst1, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_32_def_whole(dst0, dst1, src0, src1, src2, count);
+#else
+	scale2x_32_def_border(dst0, src0, src1, src2, count);
+	scale2x_32_def_border(dst1, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2x3 a row of pixels of 8 bits.
+ * \note Like scale2x_8_def();
+ */
+void scale2x3_8_def(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_8_def_whole(dst0, dst2, src0, src1, src2, count);
+	scale2x_8_def_center(dst1, src0, src1, src2, count);
+#else
+	scale2x_8_def_border(dst0, src0, src1, src2, count);
+	scale2x_8_def_center(dst1, src0, src1, src2, count);
+	scale2x_8_def_border(dst2, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2x3 a row of pixels of 16 bits.
+ * \note Like scale2x_16_def();
+ */
+void scale2x3_16_def(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_16_def_whole(dst0, dst2, src0, src1, src2, count);
+	scale2x_16_def_center(dst1, src0, src1, src2, count);
+#else
+	scale2x_16_def_border(dst0, src0, src1, src2, count);
+	scale2x_16_def_center(dst1, src0, src1, src2, count);
+	scale2x_16_def_border(dst2, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2x3 a row of pixels of 32 bits.
+ * \note Like scale2x_32_def();
+ */
+void scale2x3_32_def(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_32_def_whole(dst0, dst2, src0, src1, src2, count);
+	scale2x_32_def_center(dst1, src0, src1, src2, count);
+#else
+	scale2x_32_def_border(dst0, src0, src1, src2, count);
+	scale2x_32_def_center(dst1, src0, src1, src2, count);
+	scale2x_32_def_border(dst2, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2x4 a row of pixels of 8 bits.
+ * \note Like scale2x_8_def();
+ */
+void scale2x4_8_def(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, scale2x_uint8* dst3, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_8_def_whole(dst0, dst3, src0, src1, src2, count);
+	scale2x_8_def_center(dst1, src0, src1, src2, count);
+	scale2x_8_def_center(dst2, src0, src1, src2, count);
+#else
+	scale2x_8_def_border(dst0, src0, src1, src2, count);
+	scale2x_8_def_center(dst1, src0, src1, src2, count);
+	scale2x_8_def_center(dst2, src0, src1, src2, count);
+	scale2x_8_def_border(dst3, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2x4 a row of pixels of 16 bits.
+ * \note Like scale2x_16_def();
+ */
+void scale2x4_16_def(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, scale2x_uint16* dst3, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_16_def_whole(dst0, dst3, src0, src1, src2, count);
+	scale2x_16_def_center(dst1, src0, src1, src2, count);
+	scale2x_16_def_center(dst2, src0, src1, src2, count);
+#else
+	scale2x_16_def_border(dst0, src0, src1, src2, count);
+	scale2x_16_def_center(dst1, src0, src1, src2, count);
+	scale2x_16_def_center(dst2, src0, src1, src2, count);
+	scale2x_16_def_border(dst3, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 2x4 a row of pixels of 32 bits.
+ * \note Like scale2x_32_def();
+ */
+void scale2x4_32_def(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, scale2x_uint32* dst3, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale2x_32_def_whole(dst0, dst3, src0, src1, src2, count);
+	scale2x_32_def_center(dst1, src0, src1, src2, count);
+	scale2x_32_def_center(dst2, src0, src1, src2, count);
+#else
+	scale2x_32_def_border(dst0, src0, src1, src2, count);
+	scale2x_32_def_center(dst1, src0, src1, src2, count);
+	scale2x_32_def_center(dst2, src0, src1, src2, count);
+	scale2x_32_def_border(dst3, src2, src1, src0, count);
+#endif
+}
+
+/***************************************************************************/
+/* Scale2x MMX implementation */
+
+#if defined(__GNUC__) && defined(__i386__)
+
+/*
+ * Apply the Scale2x effect at a single row.
+ * This function must be called only by the other scale2x functions.
+ *
+ * Considering the pixel map :
+ *
+ *      ABC (src0)
+ *      DEF (src1)
+ *      GHI (src2)
+ *
+ * this functions compute 2 new pixels in substitution of the source pixel E
+ * like this map :
+ *
+ *      ab (dst)
+ *
+ * with these variables :
+ *
+ *      &current -> E
+ *      &current_left -> D
+ *      &current_right -> F
+ *      &current_upper -> B
+ *      &current_lower -> H
+ *
+ *      %0 -> current_upper
+ *      %1 -> current
+ *      %2 -> current_lower
+ *      %3 -> dst
+ *      %4 -> counter
+ *
+ *      %mm0 -> *current_left
+ *      %mm1 -> *current_next
+ *      %mm2 -> tmp0
+ *      %mm3 -> tmp1
+ *      %mm4 -> tmp2
+ *      %mm5 -> tmp3
+ *      %mm6 -> *current_upper
+ *      %mm7 -> *current
+ */
+static inline void scale2x_8_mmx_border(scale2x_uint8* dst, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+	assert(count >= 16);
+	assert(count % 8 == 0);
+
+	/* always do the first and last run */
+	count -= 2*8;
+
+	__asm__ __volatile__(
+/* first run */
+		/* set the current, current_pre, current_next registers */
+		"movq 0(%1), %%mm0\n"
+		"movq 0(%1), %%mm7\n"
+		"movq 8(%1), %%mm1\n"
+		"psllq $56, %%mm0\n"
+		"psllq $56, %%mm1\n"
+		"psrlq $56, %%mm0\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $8, %%mm2\n"
+		"psrlq $8, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqb %%mm6, %%mm2\n"
+		"pcmpeqb %%mm6, %%mm4\n"
+		"pcmpeqb (%2), %%mm3\n"
+		"pcmpeqb (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqb %%mm1, %%mm2\n"
+		"pcmpeqb %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpcklbw %%mm4, %%mm2\n"
+		"punpckhbw %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		/* next */
+		"addl $8, %0\n"
+		"addl $8, %1\n"
+		"addl $8, %2\n"
+		"addl $16, %3\n"
+
+/* central runs */
+		"shrl $3, %4\n"
+		"jz 1f\n"
+
+		"0:\n"
+
+		/* set the current, current_pre, current_next registers */
+		"movq -8(%1), %%mm0\n"
+		"movq (%1), %%mm7\n"
+		"movq 8(%1), %%mm1\n"
+		"psrlq $56, %%mm0\n"
+		"psllq $56, %%mm1\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $8, %%mm2\n"
+		"psrlq $8, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqb %%mm6, %%mm2\n"
+		"pcmpeqb %%mm6, %%mm4\n"
+		"pcmpeqb (%2), %%mm3\n"
+		"pcmpeqb (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqb %%mm1, %%mm2\n"
+		"pcmpeqb %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpcklbw %%mm4, %%mm2\n"
+		"punpckhbw %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		/* next */
+		"addl $8, %0\n"
+		"addl $8, %1\n"
+		"addl $8, %2\n"
+		"addl $16, %3\n"
+
+		"decl %4\n"
+		"jnz 0b\n"
+		"1:\n"
+
+/* final run */
+		/* set the current, current_pre, current_next registers */
+		"movq (%1), %%mm1\n"
+		"movq (%1), %%mm7\n"
+		"movq -8(%1), %%mm0\n"
+		"psrlq $56, %%mm1\n"
+		"psrlq $56, %%mm0\n"
+		"psllq $56, %%mm1\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $8, %%mm2\n"
+		"psrlq $8, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqb %%mm6, %%mm2\n"
+		"pcmpeqb %%mm6, %%mm4\n"
+		"pcmpeqb (%2), %%mm3\n"
+		"pcmpeqb (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqb %%mm1, %%mm2\n"
+		"pcmpeqb %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpcklbw %%mm4, %%mm2\n"
+		"punpckhbw %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		: "+r" (src0), "+r" (src1), "+r" (src2), "+r" (dst), "+r" (count)
+		:
+		: "cc"
+	);
+}
+
+static inline void scale2x_16_mmx_border(scale2x_uint16* dst, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+	assert(count >= 8);
+	assert(count % 4 == 0);
+
+	/* always do the first and last run */
+	count -= 2*4;
+
+	__asm__ __volatile__(
+/* first run */
+		/* set the current, current_pre, current_next registers */
+		"movq 0(%1), %%mm0\n"
+		"movq 0(%1), %%mm7\n"
+		"movq 8(%1), %%mm1\n"
+		"psllq $48, %%mm0\n"
+		"psllq $48, %%mm1\n"
+		"psrlq $48, %%mm0\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $16, %%mm2\n"
+		"psrlq $16, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqw %%mm6, %%mm2\n"
+		"pcmpeqw %%mm6, %%mm4\n"
+		"pcmpeqw (%2), %%mm3\n"
+		"pcmpeqw (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqw %%mm1, %%mm2\n"
+		"pcmpeqw %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpcklwd %%mm4, %%mm2\n"
+		"punpckhwd %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		/* next */
+		"addl $8, %0\n"
+		"addl $8, %1\n"
+		"addl $8, %2\n"
+		"addl $16, %3\n"
+
+/* central runs */
+		"shrl $2, %4\n"
+		"jz 1f\n"
+
+		"0:\n"
+
+		/* set the current, current_pre, current_next registers */
+		"movq -8(%1), %%mm0\n"
+		"movq (%1), %%mm7\n"
+		"movq 8(%1), %%mm1\n"
+		"psrlq $48, %%mm0\n"
+		"psllq $48, %%mm1\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $16, %%mm2\n"
+		"psrlq $16, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqw %%mm6, %%mm2\n"
+		"pcmpeqw %%mm6, %%mm4\n"
+		"pcmpeqw (%2), %%mm3\n"
+		"pcmpeqw (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqw %%mm1, %%mm2\n"
+		"pcmpeqw %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpcklwd %%mm4, %%mm2\n"
+		"punpckhwd %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		/* next */
+		"addl $8, %0\n"
+		"addl $8, %1\n"
+		"addl $8, %2\n"
+		"addl $16, %3\n"
+
+		"decl %4\n"
+		"jnz 0b\n"
+		"1:\n"
+
+/* final run */
+		/* set the current, current_pre, current_next registers */
+		"movq (%1), %%mm1\n"
+		"movq (%1), %%mm7\n"
+		"movq -8(%1), %%mm0\n"
+		"psrlq $48, %%mm1\n"
+		"psrlq $48, %%mm0\n"
+		"psllq $48, %%mm1\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $16, %%mm2\n"
+		"psrlq $16, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqw %%mm6, %%mm2\n"
+		"pcmpeqw %%mm6, %%mm4\n"
+		"pcmpeqw (%2), %%mm3\n"
+		"pcmpeqw (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqw %%mm1, %%mm2\n"
+		"pcmpeqw %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpcklwd %%mm4, %%mm2\n"
+		"punpckhwd %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		: "+r" (src0), "+r" (src1), "+r" (src2), "+r" (dst), "+r" (count)
+		:
+		: "cc"
+	);
+}
+
+static inline void scale2x_32_mmx_border(scale2x_uint32* dst, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+	assert(count >= 4);
+	assert(count % 2 == 0);
+
+	/* always do the first and last run */
+	count -= 2*2;
+
+	__asm__ __volatile__(
+/* first run */
+		/* set the current, current_pre, current_next registers */
+		"movq 0(%1), %%mm0\n"
+		"movq 0(%1), %%mm7\n"
+		"movq 8(%1), %%mm1\n"
+		"psllq $32, %%mm0\n"
+		"psllq $32, %%mm1\n"
+		"psrlq $32, %%mm0\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $32, %%mm2\n"
+		"psrlq $32, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqd %%mm6, %%mm2\n"
+		"pcmpeqd %%mm6, %%mm4\n"
+		"pcmpeqd (%2), %%mm3\n"
+		"pcmpeqd (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqd %%mm1, %%mm2\n"
+		"pcmpeqd %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpckldq %%mm4, %%mm2\n"
+		"punpckhdq %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		/* next */
+		"addl $8, %0\n"
+		"addl $8, %1\n"
+		"addl $8, %2\n"
+		"addl $16, %3\n"
+
+/* central runs */
+		"shrl $1, %4\n"
+		"jz 1f\n"
+
+		"0:\n"
+
+		/* set the current, current_pre, current_next registers */
+		"movq -8(%1), %%mm0\n"
+		"movq (%1), %%mm7\n"
+		"movq 8(%1), %%mm1\n"
+		"psrlq $32, %%mm0\n"
+		"psllq $32, %%mm1\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $32, %%mm2\n"
+		"psrlq $32, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqd %%mm6, %%mm2\n"
+		"pcmpeqd %%mm6, %%mm4\n"
+		"pcmpeqd (%2), %%mm3\n"
+		"pcmpeqd (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqd %%mm1, %%mm2\n"
+		"pcmpeqd %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpckldq %%mm4, %%mm2\n"
+		"punpckhdq %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		/* next */
+		"addl $8, %0\n"
+		"addl $8, %1\n"
+		"addl $8, %2\n"
+		"addl $16, %3\n"
+
+		"decl %4\n"
+		"jnz 0b\n"
+		"1:\n"
+
+/* final run */
+		/* set the current, current_pre, current_next registers */
+		"movq (%1), %%mm1\n"
+		"movq (%1), %%mm7\n"
+		"movq -8(%1), %%mm0\n"
+		"psrlq $32, %%mm1\n"
+		"psrlq $32, %%mm0\n"
+		"psllq $32, %%mm1\n"
+		"movq %%mm7, %%mm2\n"
+		"movq %%mm7, %%mm3\n"
+		"psllq $32, %%mm2\n"
+		"psrlq $32, %%mm3\n"
+		"por %%mm2, %%mm0\n"
+		"por %%mm3, %%mm1\n"
+
+		/* current_upper */
+		"movq (%0), %%mm6\n"
+
+		/* compute the upper-left pixel for dst on %%mm2 */
+		/* compute the upper-right pixel for dst on %%mm4 */
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"movq %%mm0, %%mm3\n"
+		"movq %%mm1, %%mm5\n"
+		"pcmpeqd %%mm6, %%mm2\n"
+		"pcmpeqd %%mm6, %%mm4\n"
+		"pcmpeqd (%2), %%mm3\n"
+		"pcmpeqd (%2), %%mm5\n"
+		"pandn %%mm2, %%mm3\n"
+		"pandn %%mm4, %%mm5\n"
+		"movq %%mm0, %%mm2\n"
+		"movq %%mm1, %%mm4\n"
+		"pcmpeqd %%mm1, %%mm2\n"
+		"pcmpeqd %%mm0, %%mm4\n"
+		"pandn %%mm3, %%mm2\n"
+		"pandn %%mm5, %%mm4\n"
+		"movq %%mm2, %%mm3\n"
+		"movq %%mm4, %%mm5\n"
+		"pand %%mm6, %%mm2\n"
+		"pand %%mm6, %%mm4\n"
+		"pandn %%mm7, %%mm3\n"
+		"pandn %%mm7, %%mm5\n"
+		"por %%mm3, %%mm2\n"
+		"por %%mm5, %%mm4\n"
+
+		/* set *dst */
+		"movq %%mm2, %%mm3\n"
+		"punpckldq %%mm4, %%mm2\n"
+		"punpckhdq %%mm4, %%mm3\n"
+		"movq %%mm2, (%3)\n"
+		"movq %%mm3, 8(%3)\n"
+
+		: "+r" (src0), "+r" (src1), "+r" (src2), "+r" (dst), "+r" (count)
+		:
+		: "cc"
+	);
+}
+
+/**
+ * Scale by a factor of 2 a row of pixels of 8 bits.
+ * This is a very fast MMX implementation.
+ * The implementation uses a combination of cmp/and/not operations to
+ * completly remove the need of conditional jumps. This trick give the
+ * major speed improvement.
+ * Also, using the 8 bytes MMX registers more than one pixel are computed
+ * at the same time.
+ * Before calling this function you must ensure that the currenct CPU supports
+ * the MMX instruction set. After calling it you must be sure to call the EMMS
+ * instruction before any floating-point operation.
+ * The pixels over the left and right borders are assumed of the same color of
+ * the pixels on the border.
+ * Note that the implementation is optimized to write data sequentially to
+ * maximize the bandwidth on video memory.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows. It must
+ * be at least 16 and a multiple of 8.
+ * \param dst0 First destination row, double length in pixels.
+ * \param dst1 Second destination row, double length in pixels.
+ */
+void scale2x_8_mmx(scale2x_uint8* dst0, scale2x_uint8* dst1, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+	if (count % 8 != 0 || count < 16) {
+		scale2x_8_def(dst0, dst1, src0, src1, src2, count);
+	} else {
+		scale2x_8_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_8_mmx_border(dst1, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2 a row of pixels of 16 bits.
+ * This function operates like scale2x_8_mmx() but for 16 bits pixels.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows. It must
+ * be at least 8 and a multiple of 4.
+ * \param dst0 First destination row, double length in pixels.
+ * \param dst1 Second destination row, double length in pixels.
+ */
+void scale2x_16_mmx(scale2x_uint16* dst0, scale2x_uint16* dst1, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+	if (count % 4 != 0 || count < 8) {
+		scale2x_16_def(dst0, dst1, src0, src1, src2, count);
+	} else {
+		scale2x_16_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_16_mmx_border(dst1, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2 a row of pixels of 32 bits.
+ * This function operates like scale2x_8_mmx() but for 32 bits pixels.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows. It must
+ * be at least 4 and a multiple of 2.
+ * \param dst0 First destination row, double length in pixels.
+ * \param dst1 Second destination row, double length in pixels.
+ */
+void scale2x_32_mmx(scale2x_uint32* dst0, scale2x_uint32* dst1, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+	if (count % 2 != 0 || count < 4) {
+		scale2x_32_def(dst0, dst1, src0, src1, src2, count);
+	} else {
+		scale2x_32_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_32_mmx_border(dst1, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2x3 a row of pixels of 8 bits.
+ * This function operates like scale2x_8_mmx() but with an expansion
+ * factor of 2x3 instead of 2x2.
+ */
+void scale2x3_8_mmx(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+	if (count % 8 != 0 || count < 16) {
+		scale2x3_8_def(dst0, dst1, dst2, src0, src1, src2, count);
+	} else {
+		scale2x_8_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_8_def_center(dst1, src0, src1, src2, count);
+		scale2x_8_mmx_border(dst2, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2x3 a row of pixels of 16 bits.
+ * This function operates like scale2x_16_mmx() but with an expansion
+ * factor of 2x3 instead of 2x2.
+ */
+void scale2x3_16_mmx(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+	if (count % 4 != 0 || count < 8) {
+		scale2x3_16_def(dst0, dst1, dst2, src0, src1, src2, count);
+	} else {
+		scale2x_16_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_16_def_center(dst1, src0, src1, src2, count);
+		scale2x_16_mmx_border(dst2, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2x3 a row of pixels of 32 bits.
+ * This function operates like scale2x_32_mmx() but with an expansion
+ * factor of 2x3 instead of 2x2.
+ */
+void scale2x3_32_mmx(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+	if (count % 2 != 0 || count < 4) {
+		scale2x3_32_def(dst0, dst1, dst2, src0, src1, src2, count);
+	} else {
+		scale2x_32_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_32_def_center(dst1, src0, src1, src2, count);
+		scale2x_32_mmx_border(dst2, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2x4 a row of pixels of 8 bits.
+ * This function operates like scale2x_8_mmx() but with an expansion
+ * factor of 2x4 instead of 2x2.
+ */
+void scale2x4_8_mmx(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, scale2x_uint8* dst3, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count)
+{
+	if (count % 8 != 0 || count < 16) {
+		scale2x4_8_def(dst0, dst1, dst2, dst3, src0, src1, src2, count);
+	} else {
+		scale2x_8_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_8_def_center(dst1, src0, src1, src2, count);
+		scale2x_8_def_center(dst2, src0, src1, src2, count);
+		scale2x_8_mmx_border(dst3, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2x4 a row of pixels of 16 bits.
+ * This function operates like scale2x_16_mmx() but with an expansion
+ * factor of 2x4 instead of 2x2.
+ */
+void scale2x4_16_mmx(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, scale2x_uint16* dst3, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count)
+{
+	if (count % 4 != 0 || count < 8) {
+		scale2x4_16_def(dst0, dst1, dst2, dst3, src0, src1, src2, count);
+	} else {
+		scale2x_16_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_16_def_center(dst1, src0, src1, src2, count);
+		scale2x_16_def_center(dst2, src0, src1, src2, count);
+		scale2x_16_mmx_border(dst3, src2, src1, src0, count);
+	}
+}
+
+/**
+ * Scale by a factor of 2x4 a row of pixels of 32 bits.
+ * This function operates like scale2x_32_mmx() but with an expansion
+ * factor of 2x4 instead of 2x2.
+ */
+void scale2x4_32_mmx(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, scale2x_uint32* dst3, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count)
+{
+	if (count % 2 != 0 || count < 4) {
+		scale2x4_32_def(dst0, dst1, dst2, dst3, src0, src1, src2, count);
+	} else {
+		scale2x_32_mmx_border(dst0, src0, src1, src2, count);
+		scale2x_32_def_center(dst1, src0, src1, src2, count);
+		scale2x_32_def_center(dst2, src0, src1, src2, count);
+		scale2x_32_mmx_border(dst3, src2, src1, src0, count);
+	}
+}
+
+#endif
+

src/scale2x/scale2x.h

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2001, 2002, 2003, 2004 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef __SCALE2X_H
+#define __SCALE2X_H
+
+#define restrict
+
+typedef unsigned char scale2x_uint8;
+typedef unsigned short scale2x_uint16;
+typedef unsigned scale2x_uint32;
+
+void scale2x_8_def(scale2x_uint8* dst0, scale2x_uint8* dst1, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count);
+void scale2x_16_def(scale2x_uint16* dst0, scale2x_uint16* dst1, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count);
+void scale2x_32_def(scale2x_uint32* dst0, scale2x_uint32* dst1, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count);
+
+void scale2x3_8_def(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count);
+void scale2x3_16_def(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count);
+void scale2x3_32_def(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count);
+
+void scale2x4_8_def(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, scale2x_uint8* dst3, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count);
+void scale2x4_16_def(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, scale2x_uint16* dst3, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count);
+void scale2x4_32_def(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, scale2x_uint32* dst3, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count);
+
+#if defined(__GNUC__) && defined(__i386__)
+
+void scale2x_8_mmx(scale2x_uint8* dst0, scale2x_uint8* dst1, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count);
+void scale2x_16_mmx(scale2x_uint16* dst0, scale2x_uint16* dst1, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count);
+void scale2x_32_mmx(scale2x_uint32* dst0, scale2x_uint32* dst1, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count);
+
+void scale2x3_8_mmx(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count);
+void scale2x3_16_mmx(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count);
+void scale2x3_32_mmx(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count);
+
+void scale2x4_8_mmx(scale2x_uint8* dst0, scale2x_uint8* dst1, scale2x_uint8* dst2, scale2x_uint8* dst3, const scale2x_uint8* src0, const scale2x_uint8* src1, const scale2x_uint8* src2, unsigned count);
+void scale2x4_16_mmx(scale2x_uint16* dst0, scale2x_uint16* dst1, scale2x_uint16* dst2, scale2x_uint16* dst3, const scale2x_uint16* src0, const scale2x_uint16* src1, const scale2x_uint16* src2, unsigned count);
+void scale2x4_32_mmx(scale2x_uint32* dst0, scale2x_uint32* dst1, scale2x_uint32* dst2, scale2x_uint32* dst3, const scale2x_uint32* src0, const scale2x_uint32* src1, const scale2x_uint32* src2, unsigned count);
+
+/**
+ * End the use of the MMX instructions.
+ * This function must be called before using any floating-point operations.
+ */
+static inline void scale2x_mmx_emms(void)
+{
+	__asm__ __volatile__ (
+		"emms"
+	);
+}
+
+#endif
+
+#endif
+

src/scale2x/scale3x.cpp

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2001, 2002, 2003, 2004 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/*
+ * This file contains a C and MMX implementation of the Scale2x effect.
+ *
+ * You can find an high level description of the effect at :
+ *
+ * http://scale2x.sourceforge.net/
+ *
+ * Alternatively at the previous license terms, you are allowed to use this
+ * code in your program with these conditions:
+ * - the program is not used in commercial activities.
+ * - the whole source code of the program is released with the binary.
+ * - derivative works of the program are allowed.
+ */
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "scale3x.h"
+
+#include <assert.h>
+
+/***************************************************************************/
+/* Scale3x C implementation */
+
+/**
+ * Define the macro USE_SCALE_RANDOMWRITE to enable
+ * an optimized version which writes memory in random order.
+ * This version is a little faster if you write in system memory.
+ * But it's a lot slower if you write in video memory.
+ * So, enable it only if you are sure to never write directly in video memory.
+ */
+/* #define USE_SCALE_RANDOMWRITE */
+
+static inline void scale3x_8_def_whole(scale3x_uint8* restrict dst0, scale3x_uint8* restrict dst1, scale3x_uint8* restrict dst2, const scale3x_uint8* restrict src0, const scale3x_uint8* restrict src1, const scale3x_uint8* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst0[0] = src1[0];
+		dst0[1] = (src1[0] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[0]) ? src0[0] : src1[0];
+		dst0[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+		dst1[0] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = (src1[0] == src2[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src2[0]) ? src2[0] : src1[0];
+		dst2[2] = src1[1] == src2[0] ? src1[1] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = src1[0];
+		dst2[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst0 += 3;
+	dst1 += 3;
+	dst2 += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst0[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+			dst0[1] = (src1[-1] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+			dst0[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+			dst1[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst1[1] = src1[0];
+			dst1[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+			dst2[0] = src1[-1] == src2[0] ? src1[-1] : src1[0];
+			dst2[1] = (src1[-1] == src2[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src2[-1]) ? src2[0] : src1[0];
+			dst2[2] = src1[1] == src2[0] ? src1[1] : src1[0];
+		} else {
+			dst0[0] = src1[0];
+			dst0[1] = src1[0];
+			dst0[2] = src1[0];
+			dst1[0] = src1[0];
+			dst1[1] = src1[0];
+			dst1[2] = src1[0];
+			dst2[0] = src1[0];
+			dst2[1] = src1[0];
+			dst2[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst0 += 3;
+		dst1 += 3;
+		dst2 += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst0[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+		dst0[1] = (src1[-1] == src0[0] && src1[0] != src0[0]) || (src1[0] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst2[0] = src1[-1] == src2[0] ? src1[-1] : src1[0];
+		dst2[1] = (src1[-1] == src2[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src2[-1]) ? src2[0] : src1[0];
+		dst2[2] = src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = src1[0];
+		dst2[2] = src1[0];
+	}
+}
+
+static inline void scale3x_8_def_border(scale3x_uint8* restrict dst, const scale3x_uint8* restrict src0, const scale3x_uint8* restrict src1, const scale3x_uint8* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0];
+		dst[1] = (src1[0] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[0]) ? src0[0] : src1[0];
+		dst[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+			dst[1] = (src1[-1] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+			dst[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+			dst[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+		dst[1] = (src1[-1] == src0[0] && src1[0] != src0[0]) || (src1[0] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+		dst[2] = src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+}
+
+static inline void scale3x_8_def_center(scale3x_uint8* restrict dst, const scale3x_uint8* restrict src0, const scale3x_uint8* restrict src1, const scale3x_uint8* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst[1] = src1[0];
+		dst[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst[1] = src1[0];
+			dst[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+			dst[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst[1] = src1[0];
+		dst[2] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+}
+
+static inline void scale3x_16_def_whole(scale3x_uint16* restrict dst0, scale3x_uint16* restrict dst1, scale3x_uint16* restrict dst2, const scale3x_uint16* restrict src0, const scale3x_uint16* restrict src1, const scale3x_uint16* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst0[0] = src1[0];
+		dst0[1] = (src1[0] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[0]) ? src0[0] : src1[0];
+		dst0[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+		dst1[0] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = (src1[0] == src2[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src2[0]) ? src2[0] : src1[0];
+		dst2[2] = src1[1] == src2[0] ? src1[1] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = src1[0];
+		dst2[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst0 += 3;
+	dst1 += 3;
+	dst2 += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst0[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+			dst0[1] = (src1[-1] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+			dst0[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+			dst1[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst1[1] = src1[0];
+			dst1[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+			dst2[0] = src1[-1] == src2[0] ? src1[-1] : src1[0];
+			dst2[1] = (src1[-1] == src2[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src2[-1]) ? src2[0] : src1[0];
+			dst2[2] = src1[1] == src2[0] ? src1[1] : src1[0];
+		} else {
+			dst0[0] = src1[0];
+			dst0[1] = src1[0];
+			dst0[2] = src1[0];
+			dst1[0] = src1[0];
+			dst1[1] = src1[0];
+			dst1[2] = src1[0];
+			dst2[0] = src1[0];
+			dst2[1] = src1[0];
+			dst2[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst0 += 3;
+		dst1 += 3;
+		dst2 += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst0[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+		dst0[1] = (src1[-1] == src0[0] && src1[0] != src0[0]) || (src1[0] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst2[0] = src1[-1] == src2[0] ? src1[-1] : src1[0];
+		dst2[1] = (src1[-1] == src2[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src2[-1]) ? src2[0] : src1[0];
+		dst2[2] = src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = src1[0];
+		dst2[2] = src1[0];
+	}
+}
+
+static inline void scale3x_16_def_border(scale3x_uint16* restrict dst, const scale3x_uint16* restrict src0, const scale3x_uint16* restrict src1, const scale3x_uint16* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0];
+		dst[1] = (src1[0] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[0]) ? src0[0] : src1[0];
+		dst[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+			dst[1] = (src1[-1] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+			dst[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+			dst[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+		dst[1] = (src1[-1] == src0[0] && src1[0] != src0[0]) || (src1[0] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+		dst[2] = src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+}
+
+static inline void scale3x_16_def_center(scale3x_uint16* restrict dst, const scale3x_uint16* restrict src0, const scale3x_uint16* restrict src1, const scale3x_uint16* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst[1] = src1[0];
+		dst[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst[1] = src1[0];
+			dst[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+			dst[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst[1] = src1[0];
+		dst[2] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+}
+
+static inline void scale3x_32_def_whole(scale3x_uint32* restrict dst0, scale3x_uint32* restrict dst1, scale3x_uint32* restrict dst2, const scale3x_uint32* restrict src0, const scale3x_uint32* restrict src1, const scale3x_uint32* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst0[0] = src1[0];
+		dst0[1] = (src1[0] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[0]) ? src0[0] : src1[0];
+		dst0[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+		dst1[0] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = (src1[0] == src2[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src2[0]) ? src2[0] : src1[0];
+		dst2[2] = src1[1] == src2[0] ? src1[1] : src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = src1[0];
+		dst2[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst0 += 3;
+	dst1 += 3;
+	dst2 += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst0[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+			dst0[1] = (src1[-1] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+			dst0[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+			dst1[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst1[1] = src1[0];
+			dst1[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+			dst2[0] = src1[-1] == src2[0] ? src1[-1] : src1[0];
+			dst2[1] = (src1[-1] == src2[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src2[-1]) ? src2[0] : src1[0];
+			dst2[2] = src1[1] == src2[0] ? src1[1] : src1[0];
+		} else {
+			dst0[0] = src1[0];
+			dst0[1] = src1[0];
+			dst0[2] = src1[0];
+			dst1[0] = src1[0];
+			dst1[1] = src1[0];
+			dst1[2] = src1[0];
+			dst2[0] = src1[0];
+			dst2[1] = src1[0];
+			dst2[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst0 += 3;
+		dst1 += 3;
+		dst2 += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst0[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+		dst0[1] = (src1[-1] == src0[0] && src1[0] != src0[0]) || (src1[0] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst2[0] = src1[-1] == src2[0] ? src1[-1] : src1[0];
+		dst2[1] = (src1[-1] == src2[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src2[-1]) ? src2[0] : src1[0];
+		dst2[2] = src1[0];
+	} else {
+		dst0[0] = src1[0];
+		dst0[1] = src1[0];
+		dst0[2] = src1[0];
+		dst1[0] = src1[0];
+		dst1[1] = src1[0];
+		dst1[2] = src1[0];
+		dst2[0] = src1[0];
+		dst2[1] = src1[0];
+		dst2[2] = src1[0];
+	}
+}
+
+static inline void scale3x_32_def_border(scale3x_uint32* restrict dst, const scale3x_uint32* restrict src0, const scale3x_uint32* restrict src1, const scale3x_uint32* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = src1[0];
+		dst[1] = (src1[0] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[0]) ? src0[0] : src1[0];
+		dst[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+			dst[1] = (src1[-1] == src0[0] && src1[0] != src0[1]) || (src1[1] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+			dst[2] = src1[1] == src0[0] ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+			dst[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = src1[-1] == src0[0] ? src1[-1] : src1[0];
+		dst[1] = (src1[-1] == src0[0] && src1[0] != src0[0]) || (src1[0] == src0[0] && src1[0] != src0[-1]) ? src0[0] : src1[0];
+		dst[2] = src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+}
+
+static inline void scale3x_32_def_center(scale3x_uint32* restrict dst, const scale3x_uint32* restrict src0, const scale3x_uint32* restrict src1, const scale3x_uint32* restrict src2, unsigned count)
+{
+	assert(count >= 2);
+
+	/* first pixel */
+	if (src0[0] != src2[0] && src1[0] != src1[1]) {
+		dst[0] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+		dst[1] = src1[0];
+		dst[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+	++src0;
+	++src1;
+	++src2;
+	dst += 3;
+
+	/* central pixels */
+	count -= 2;
+	while (count) {
+		if (src0[0] != src2[0] && src1[-1] != src1[1]) {
+			dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+			dst[1] = src1[0];
+			dst[2] = (src1[1] == src0[0] && src1[0] != src2[1]) || (src1[1] == src2[0] && src1[0] != src0[1]) ? src1[1] : src1[0];
+		} else {
+			dst[0] = src1[0];
+			dst[1] = src1[0];
+			dst[2] = src1[0];
+		}
+
+		++src0;
+		++src1;
+		++src2;
+		dst += 3;
+		--count;
+	}
+
+	/* last pixel */
+	if (src0[0] != src2[0] && src1[-1] != src1[0]) {
+		dst[0] = (src1[-1] == src0[0] && src1[0] != src2[-1]) || (src1[-1] == src2[0] && src1[0] != src0[-1]) ? src1[-1] : src1[0];
+		dst[1] = src1[0];
+		dst[2] = (src1[0] == src0[0] && src1[0] != src2[0]) || (src1[0] == src2[0] && src1[0] != src0[0]) ? src1[0] : src1[0];
+	} else {
+		dst[0] = src1[0];
+		dst[1] = src1[0];
+		dst[2] = src1[0];
+	}
+}
+
+/**
+ * Scale by a factor of 3 a row of pixels of 8 bits.
+ * The function is implemented in C.
+ * The pixels over the left and right borders are assumed of the same color of
+ * the pixels on the border.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows.
+ * It must be at least 2.
+ * \param dst0 First destination row, triple length in pixels.
+ * \param dst1 Second destination row, triple length in pixels.
+ * \param dst2 Third destination row, triple length in pixels.
+ */
+void scale3x_8_def(scale3x_uint8* dst0, scale3x_uint8* dst1, scale3x_uint8* dst2, const scale3x_uint8* src0, const scale3x_uint8* src1, const scale3x_uint8* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale3x_8_def_whole(dst0, dst1, dst2, src0, src1, src2, count);
+#else
+	scale3x_8_def_border(dst0, src0, src1, src2, count);
+	scale3x_8_def_center(dst1, src0, src1, src2, count);
+	scale3x_8_def_border(dst2, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 3 a row of pixels of 16 bits.
+ * This function operates like scale3x_8_def() but for 16 bits pixels.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows.
+ * It must be at least 2.
+ * \param dst0 First destination row, triple length in pixels.
+ * \param dst1 Second destination row, triple length in pixels.
+ * \param dst2 Third destination row, triple length in pixels.
+ */
+void scale3x_16_def(scale3x_uint16* dst0, scale3x_uint16* dst1, scale3x_uint16* dst2, const scale3x_uint16* src0, const scale3x_uint16* src1, const scale3x_uint16* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale3x_16_def_whole(dst0, dst1, dst2, src0, src1, src2, count);
+#else
+	scale3x_16_def_border(dst0, src0, src1, src2, count);
+	scale3x_16_def_center(dst1, src0, src1, src2, count);
+	scale3x_16_def_border(dst2, src2, src1, src0, count);
+#endif
+}
+
+/**
+ * Scale by a factor of 3 a row of pixels of 32 bits.
+ * This function operates like scale3x_8_def() but for 32 bits pixels.
+ * \param src0 Pointer at the first pixel of the previous row.
+ * \param src1 Pointer at the first pixel of the current row.
+ * \param src2 Pointer at the first pixel of the next row.
+ * \param count Length in pixels of the src0, src1 and src2 rows.
+ * It must be at least 2.
+ * \param dst0 First destination row, triple length in pixels.
+ * \param dst1 Second destination row, triple length in pixels.
+ * \param dst2 Third destination row, triple length in pixels.
+ */
+void scale3x_32_def(scale3x_uint32* dst0, scale3x_uint32* dst1, scale3x_uint32* dst2, const scale3x_uint32* src0, const scale3x_uint32* src1, const scale3x_uint32* src2, unsigned count)
+{
+#ifdef USE_SCALE_RANDOMWRITE
+	scale3x_32_def_whole(dst0, dst1, dst2, src0, src1, src2, count);
+#else
+	scale3x_32_def_border(dst0, src0, src1, src2, count);
+	scale3x_32_def_center(dst1, src0, src1, src2, count);
+	scale3x_32_def_border(dst2, src2, src1, src0, count);
+#endif
+}
+

src/scale2x/scale3x.h

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2001, 2002, 2003, 2004 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef __SCALE3X_H
+#define __SCALE3X_H
+
+#define restrict
+
+typedef unsigned char scale3x_uint8;
+typedef unsigned short scale3x_uint16;
+typedef unsigned scale3x_uint32;
+
+void scale3x_8_def(scale3x_uint8* dst0, scale3x_uint8* dst1, scale3x_uint8* dst2, const scale3x_uint8* src0, const scale3x_uint8* src1, const scale3x_uint8* src2, unsigned count);
+void scale3x_16_def(scale3x_uint16* dst0, scale3x_uint16* dst1, scale3x_uint16* dst2, const scale3x_uint16* src0, const scale3x_uint16* src1, const scale3x_uint16* src2, unsigned count);
+void scale3x_32_def(scale3x_uint32* dst0, scale3x_uint32* dst1, scale3x_uint32* dst2, const scale3x_uint32* src0, const scale3x_uint32* src1, const scale3x_uint32* src2, unsigned count);
+
+#endif
+

src/scale2x/scalebit.cpp

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2003, 2004 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/*
+ * This file contains an example implementation of the Scale effect
+ * applyed to a generic bitmap.
+ *
+ * You can find an high level description of the effect at :
+ *
+ * http://scale2x.sourceforge.net/
+ *
+ * Alternatively at the previous license terms, you are allowed to use this
+ * code in your program with these conditions:
+ * - the program is not used in commercial activities.
+ * - the whole source code of the program is released with the binary.
+ * - derivative works of the program are allowed.
+ */
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "scale2x.h"
+#include "scale3x.h"
+
+#if HAVE_ALLOCA_H
+#include <alloca.h>
+#endif
+
+#include <assert.h>
+#include <stdlib.h>
+
+#define SSDST(bits, num) (scale2x_uint##bits *)dst##num
+#define SSSRC(bits, num) (const scale2x_uint##bits *)src##num
+
+/**
+ * Apply the Scale2x effect on a group of rows. Used internally.
+ */
+static inline void stage_scale2x(void* dst0, void* dst1, const void* src0, const void* src1, const void* src2, unsigned pixel, unsigned pixel_per_row)
+{
+	switch (pixel) {
+#if defined(__GNUC__) && defined(__i386__)
+		case 1 : scale2x_8_mmx(SSDST(8,0), SSDST(8,1), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale2x_16_mmx(SSDST(16,0), SSDST(16,1), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale2x_32_mmx(SSDST(32,0), SSDST(32,1), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+#else
+		case 1 : scale2x_8_def(SSDST(8,0), SSDST(8,1), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale2x_16_def(SSDST(16,0), SSDST(16,1), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale2x_32_def(SSDST(32,0), SSDST(32,1), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+#endif
+	}
+}
+
+/**
+ * Apply the Scale2x3 effect on a group of rows. Used internally.
+ */
+static inline void stage_scale2x3(void* dst0, void* dst1, void* dst2, const void* src0, const void* src1, const void* src2, unsigned pixel, unsigned pixel_per_row)
+{
+	switch (pixel) {
+#if defined(__GNUC__) && defined(__i386__)
+		case 1 : scale2x3_8_mmx(SSDST(8,0), SSDST(8,1), SSDST(8,2), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale2x3_16_mmx(SSDST(16,0), SSDST(16,1), SSDST(16,2), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale2x3_32_mmx(SSDST(32,0), SSDST(32,1), SSDST(32,2), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+#else
+		case 1 : scale2x3_8_def(SSDST(8,0), SSDST(8,1), SSDST(8,2), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale2x3_16_def(SSDST(16,0), SSDST(16,1), SSDST(16,2), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale2x3_32_def(SSDST(32,0), SSDST(32,1), SSDST(32,2), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+#endif
+	}
+}
+
+/**
+ * Apply the Scale2x4 effect on a group of rows. Used internally.
+ */
+static inline void stage_scale2x4(void* dst0, void* dst1, void* dst2, void* dst3, const void* src0, const void* src1, const void* src2, unsigned pixel, unsigned pixel_per_row)
+{
+	switch (pixel) {
+#if defined(__GNUC__) && defined(__i386__)
+		case 1 : scale2x4_8_mmx(SSDST(8,0), SSDST(8,1), SSDST(8,2), SSDST(8,3), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale2x4_16_mmx(SSDST(16,0), SSDST(16,1), SSDST(16,2), SSDST(16,3), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale2x4_32_mmx(SSDST(32,0), SSDST(32,1), SSDST(32,2), SSDST(32,3), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+#else
+		case 1 : scale2x4_8_def(SSDST(8,0), SSDST(8,1), SSDST(8,2), SSDST(8,3), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale2x4_16_def(SSDST(16,0), SSDST(16,1), SSDST(16,2), SSDST(16,3), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale2x4_32_def(SSDST(32,0), SSDST(32,1), SSDST(32,2), SSDST(32,3), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+#endif
+	}
+}
+
+/**
+ * Apply the Scale3x effect on a group of rows. Used internally.
+ */
+static inline void stage_scale3x(void* dst0, void* dst1, void* dst2, const void* src0, const void* src1, const void* src2, unsigned pixel, unsigned pixel_per_row)
+{
+	switch (pixel) {
+		case 1 : scale3x_8_def(SSDST(8,0), SSDST(8,1), SSDST(8,2), SSSRC(8,0), SSSRC(8,1), SSSRC(8,2), pixel_per_row); break;
+		case 2 : scale3x_16_def(SSDST(16,0), SSDST(16,1), SSDST(16,2), SSSRC(16,0), SSSRC(16,1), SSSRC(16,2), pixel_per_row); break;
+		case 4 : scale3x_32_def(SSDST(32,0), SSDST(32,1), SSDST(32,2), SSSRC(32,0), SSSRC(32,1), SSSRC(32,2), pixel_per_row); break;
+	}
+}
+
+/**
+ * Apply the Scale4x effect on a group of rows. Used internally.
+ */
+static inline void stage_scale4x(void* dst0, void* dst1, void* dst2, void* dst3, const void* src0, const void* src1, const void* src2, const void* src3, unsigned pixel, unsigned pixel_per_row)
+{
+	stage_scale2x(dst0, dst1, src0, src1, src2, pixel, 2 * pixel_per_row);
+	stage_scale2x(dst2, dst3, src1, src2, src3, pixel, 2 * pixel_per_row);
+}
+
+#define SCDST(i) (dst+(i)*dst_slice)
+#define SCSRC(i) (src+(i)*src_slice)
+#define SCMID(i) (mid[(i)])
+
+/**
+ * Apply the Scale2x effect on a bitmap.
+ * The destination bitmap is filled with the scaled version of the source bitmap.
+ * The source bitmap isn't modified.
+ * The destination bitmap must be manually allocated before calling the function,
+ * note that the resulting size is exactly 2x2 times the size of the source bitmap.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+static void scale2x(void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	unsigned char* dst = (unsigned char*)void_dst;
+	const unsigned char* src = (const unsigned char*)void_src;
+	unsigned count;
+
+	assert(height >= 2);
+
+	count = height;
+
+	stage_scale2x(SCDST(0), SCDST(1), SCSRC(0), SCSRC(0), SCSRC(1), pixel, width);
+
+	dst = SCDST(2);
+
+	count -= 2;
+	while (count) {
+		stage_scale2x(SCDST(0), SCDST(1), SCSRC(0), SCSRC(1), SCSRC(2), pixel, width);
+
+		dst = SCDST(2);
+		src = SCSRC(1);
+
+		--count;
+	}
+
+	stage_scale2x(SCDST(0), SCDST(1), SCSRC(0), SCSRC(1), SCSRC(1), pixel, width);
+
+#if defined(__GNUC__) && defined(__i386__)
+	scale2x_mmx_emms();
+#endif
+}
+
+/**
+ * Apply the Scale2x3 effect on a bitmap.
+ * The destination bitmap is filled with the scaled version of the source bitmap.
+ * The source bitmap isn't modified.
+ * The destination bitmap must be manually allocated before calling the function,
+ * note that the resulting size is exactly 2x3 times the size of the source bitmap.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+static void scale2x3(void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	unsigned char* dst = (unsigned char*)void_dst;
+	const unsigned char* src = (const unsigned char*)void_src;
+	unsigned count;
+
+	assert(height >= 2);
+
+	count = height;
+
+	stage_scale2x3(SCDST(0), SCDST(1), SCDST(2), SCSRC(0), SCSRC(0), SCSRC(1), pixel, width);
+
+	dst = SCDST(3);
+
+	count -= 2;
+	while (count) {
+		stage_scale2x3(SCDST(0), SCDST(1), SCDST(2), SCSRC(0), SCSRC(1), SCSRC(2), pixel, width);
+
+		dst = SCDST(3);
+		src = SCSRC(1);
+
+		--count;
+	}
+
+	stage_scale2x3(SCDST(0), SCDST(1), SCDST(2), SCSRC(0), SCSRC(1), SCSRC(1), pixel, width);
+
+#if defined(__GNUC__) && defined(__i386__)
+	scale2x_mmx_emms();
+#endif
+}
+
+/**
+ * Apply the Scale2x4 effect on a bitmap.
+ * The destination bitmap is filled with the scaled version of the source bitmap.
+ * The source bitmap isn't modified.
+ * The destination bitmap must be manually allocated before calling the function,
+ * note that the resulting size is exactly 2x4 times the size of the source bitmap.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+static void scale2x4(void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	unsigned char* dst = (unsigned char*)void_dst;
+	const unsigned char* src = (const unsigned char*)void_src;
+	unsigned count;
+
+	assert(height >= 2);
+
+	count = height;
+
+	stage_scale2x4(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCSRC(0), SCSRC(0), SCSRC(1), pixel, width);
+
+	dst = SCDST(4);
+
+	count -= 2;
+	while (count) {
+		stage_scale2x4(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCSRC(0), SCSRC(1), SCSRC(2), pixel, width);
+
+		dst = SCDST(4);
+		src = SCSRC(1);
+
+		--count;
+	}
+
+	stage_scale2x4(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCSRC(0), SCSRC(1), SCSRC(1), pixel, width);
+
+#if defined(__GNUC__) && defined(__i386__)
+	scale2x_mmx_emms();
+#endif
+}
+
+/**
+ * Apply the Scale3x effect on a bitmap.
+ * The destination bitmap is filled with the scaled version of the source bitmap.
+ * The source bitmap isn't modified.
+ * The destination bitmap must be manually allocated before calling the function,
+ * note that the resulting size is exactly 3x3 times the size of the source bitmap.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+static void scale3x(void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	unsigned char* dst = (unsigned char*)void_dst;
+	const unsigned char* src = (const unsigned char*)void_src;
+	unsigned count;
+
+	assert(height >= 2);
+
+	count = height;
+
+	stage_scale3x(SCDST(0), SCDST(1), SCDST(2), SCSRC(0), SCSRC(0), SCSRC(1), pixel, width);
+
+	dst = SCDST(3);
+
+	count -= 2;
+	while (count) {
+		stage_scale3x(SCDST(0), SCDST(1), SCDST(2), SCSRC(0), SCSRC(1), SCSRC(2), pixel, width);
+
+		dst = SCDST(3);
+		src = SCSRC(1);
+
+		--count;
+	}
+
+	stage_scale3x(SCDST(0), SCDST(1), SCDST(2), SCSRC(0), SCSRC(1), SCSRC(1), pixel, width);
+}
+
+/**
+ * Apply the Scale4x effect on a bitmap.
+ * The destination bitmap is filled with the scaled version of the source bitmap.
+ * The source bitmap isn't modified.
+ * The destination bitmap must be manually allocated before calling the function,
+ * note that the resulting size is exactly 4x4 times the size of the source bitmap.
+ * \note This function requires also a small buffer bitmap used internally to store
+ * intermediate results. This bitmap must have at least an horizontal size in bytes of 2*width*pixel,
+ * and a vertical size of 6 rows. The memory of this buffer must not be allocated
+ * in video memory because it's also read and not only written. Generally
+ * a heap (malloc) or a stack (alloca) buffer is the best choice.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_mid Pointer at the first pixel of the buffer bitmap.
+ * \param mid_slice Size in bytes of a buffer bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+static void scale4x_buf(void* void_dst, unsigned dst_slice, void* void_mid, unsigned mid_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	unsigned char* dst = (unsigned char*)void_dst;
+	const unsigned char* src = (const unsigned char*)void_src;
+	unsigned count;
+	unsigned char* mid[6];
+
+	assert(height >= 4);
+
+	count = height;
+
+	/* set the 6 buffer pointers */
+	mid[0] = (unsigned char*)void_mid;
+	mid[1] = mid[0] + mid_slice;
+	mid[2] = mid[1] + mid_slice;
+	mid[3] = mid[2] + mid_slice;
+	mid[4] = mid[3] + mid_slice;
+	mid[5] = mid[4] + mid_slice;
+
+	stage_scale2x(SCMID(-2+6), SCMID(-1+6), SCSRC(0), SCSRC(0), SCSRC(1), pixel, width);
+	stage_scale2x(SCMID(0), SCMID(1), SCSRC(0), SCSRC(1), SCSRC(2), pixel, width);
+	stage_scale2x(SCMID(2), SCMID(3), SCSRC(1), SCSRC(2), SCSRC(3), pixel, width);
+	stage_scale4x(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCMID(-2+6), SCMID(-2+6), SCMID(-1+6), SCMID(0), pixel, width);
+
+	dst = SCDST(4);
+
+	stage_scale4x(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCMID(-1+6), SCMID(0), SCMID(1), SCMID(2), pixel, width);
+
+	dst = SCDST(4);
+
+	count -= 4;
+	while (count) {
+		unsigned char* tmp;
+
+		stage_scale2x(SCMID(4), SCMID(5), SCSRC(2), SCSRC(3), SCSRC(4), pixel, width);
+		stage_scale4x(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCMID(1), SCMID(2), SCMID(3), SCMID(4), pixel, width);
+
+		dst = SCDST(4);
+		src = SCSRC(1);
+
+		tmp = SCMID(0); /* shift by 2 position */
+		SCMID(0) = SCMID(2);
+		SCMID(2) = SCMID(4);
+		SCMID(4) = tmp;
+		tmp = SCMID(1);
+		SCMID(1) = SCMID(3);
+		SCMID(3) = SCMID(5);
+		SCMID(5) = tmp;
+
+		--count;
+	}
+
+	stage_scale2x(SCMID(4), SCMID(5), SCSRC(2), SCSRC(3), SCSRC(3), pixel, width);
+	stage_scale4x(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCMID(1), SCMID(2), SCMID(3), SCMID(4), pixel, width);
+
+	dst = SCDST(4);
+
+	stage_scale4x(SCDST(0), SCDST(1), SCDST(2), SCDST(3), SCMID(3), SCMID(4), SCMID(5), SCMID(5), pixel, width);
+
+#if defined(__GNUC__) && defined(__i386__)
+	scale2x_mmx_emms();
+#endif
+}
+
+/**
+ * Apply the Scale4x effect on a bitmap.
+ * The destination bitmap is filled with the scaled version of the source bitmap.
+ * The source bitmap isn't modified.
+ * The destination bitmap must be manually allocated before calling the function,
+ * note that the resulting size is exactly 4x4 times the size of the source bitmap.
+ * \note This function operates like ::scale4x_buf() but the intermediate buffer is
+ * automatically allocated in the stack.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+static void scale4x(void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	unsigned mid_slice;
+	void* mid;
+
+	mid_slice = 2 * pixel * width; /* required space for 1 row buffer */
+
+	mid_slice = (mid_slice + 0x7) & ~0x7; /* align to 8 bytes */
+
+#if HAVE_ALLOCA
+	mid = alloca(6 * mid_slice); /* allocate space for 6 row buffers */
+
+	assert(mid != 0); /* alloca should never fails */
+#else
+	mid = malloc(6 * mid_slice); /* allocate space for 6 row buffers */
+
+	if (!mid)
+		return;
+#endif
+
+	scale4x_buf(void_dst, dst_slice, mid, mid_slice, void_src, src_slice, pixel, width, height);
+
+#if !HAVE_ALLOCA
+	free(mid);
+#endif
+}
+
+/**
+ * Check if the scale implementation is applicable at the given arguments.
+ * \param scale Scale factor. 2, 203 (fox 2x3), 204 (for 2x4), 3 or 4.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ * \return
+ *   - -1 on precondition violated.
+ *   - 0 on success.
+ */
+int scale_precondition(unsigned scale, unsigned pixel, unsigned width, unsigned height)
+{
+	if (pixel != 1 && pixel != 2 && pixel != 4)
+		return -1;
+
+	switch (scale) {
+	case 202 :
+	case 203 :
+	case 204 :
+	case 2 :
+	case 303 :
+	case 3 :
+		if (height < 2)
+			return -1;
+		break;
+	case 404 :
+	case 4 :
+		if (height < 4)
+			return -1;
+		break;
+	default:
+		return -1;
+	}
+
+	if (width < 2)
+		return -1;
+
+	return 0;
+}
+
+/**
+ * Apply the Scale effect on a bitmap.
+ * This function is simply a common interface for ::scale2x(), ::scale3x() and ::scale4x().
+ * \param scale Scale factor. 2, 203 (fox 2x3), 204 (for 2x4), 3 or 4.
+ * \param void_dst Pointer at the first pixel of the destination bitmap.
+ * \param dst_slice Size in bytes of a destination bitmap row.
+ * \param void_src Pointer at the first pixel of the source bitmap.
+ * \param src_slice Size in bytes of a source bitmap row.
+ * \param pixel Bytes per pixel of the source and destination bitmap.
+ * \param width Horizontal size in pixels of the source bitmap.
+ * \param height Vertical size in pixels of the source bitmap.
+ */
+void scale(unsigned scale, void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height)
+{
+	switch (scale) {
+	case 202 :
+	case 2 :
+		scale2x(void_dst, dst_slice, void_src, src_slice, pixel, width, height);
+		break;
+	case 203 :
+		scale2x3(void_dst, dst_slice, void_src, src_slice, pixel, width, height);
+		break;
+	case 204 :
+		scale2x4(void_dst, dst_slice, void_src, src_slice, pixel, width, height);
+		break;
+	case 303 :
+	case 3 :
+		scale3x(void_dst, dst_slice, void_src, src_slice, pixel, width, height);
+		break;
+	case 404 :
+	case 4 :
+		scale4x(void_dst, dst_slice, void_src, src_slice, pixel, width, height);
+		break;
+	}
+}
+

src/scale2x/scalebit.h

+/*
+ * This file is part of the Scale2x project.
+ *
+ * Copyright (C) 2003 Andrea Mazzoleni
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/*
+ * This file contains an example implementation of the Scale effect
+ * applyed to a generic bitmap.
+ *
+ * You can find an high level description of the effect at :
+ *
+ * http://scale2x.sourceforge.net/
+ *
+ * Alternatively at the previous license terms, you are allowed to use this
+ * code in your program with these conditions:
+ * - the program is not used in commercial activities.
+ * - the whole source code of the program is released with the binary.
+ * - derivative works of the program are allowed.
+ */
+
+#ifndef __SCALEBIT_H
+#define __SCALEBIT_H
+
+int scale_precondition(unsigned scale, unsigned pixel, unsigned width, unsigned height);
+void scale(unsigned scale, void* void_dst, unsigned dst_slice, const void* void_src, unsigned src_slice, unsigned pixel, unsigned width, unsigned height);
+
+void Simple2x(unsigned char *srcPtr, unsigned int srcPitch, unsigned char *deltaPtr, unsigned char *dstPtr, unsigned int dstPitch, int width, int height);
+
+#endif
+

src/scale2x/simple2x.cpp

+// VisualBoyAdvance - Nintendo Gameboy/GameboyAdvance (TM) emulator.
+// Copyright (C) 1999-2003 Forgotten
+// Copyright (C) 2004 Forgotten and the VBA development team
+
+// This program is free software; you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation; either version 2, or(at your option)
+// any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program; if not, write to the Free Software Foundation,
+// Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+
+/* 
+ * Code adapted To openjazz by Pickle (from Openbor adapted by SX
+ * simple2x.c - Trying to scale 2x.
+ *
+ *
+ */
+
+#include "scalebit.h"
+
+void Simple2x(unsigned char *srcPtr, unsigned int srcPitch, unsigned char *deltaPtr, unsigned char *dstPtr, unsigned int dstPitch, int width, int height)
+{
+    unsigned char *nextLine, *finish;
+  
+	nextLine = dstPtr + dstPitch;
+  
+	do 
+	{
+		unsigned char *bP = (unsigned char *) srcPtr;
+		unsigned char *dP = (unsigned char *) dstPtr;
+		unsigned char *nL = (unsigned char *) nextLine;
+		unsigned char currentPixel;
+    
+		finish = (unsigned char *) bP + ((width+2) << 1);
+		currentPixel = *bP++;
+    
+		do 
+		{
+#ifdef BIG_ENDIAN
+			unsigned char color = currentPixel >> 16;
+#else
+			unsigned char color = currentPixel & 0xffff;
+#endif
+
+			color = color | (color << 16);
+
+			*(dP) = color;
+			*(nL) = color;
+
+//#ifdef BIG_ENDIAN
+//			color = currentPixel & 0xffff;
+//#else
+//			color = currentPixel >> 16;
+//#endif
+			color = color| (color << 16);      
+			*(dP + 1) = color;
+			*(nL + 1) = color;
+      
+			currentPixel = *bP++;
+      
+			dP += 2;
+			nL += 2;
+		}
+		while ((unsigned char *) bP < finish);
+    
+		srcPtr += srcPitch;
+		dstPtr += dstPitch << 1;
+		nextLine += dstPitch << 1;
+	}
+	while (--height);
+}
+
+void Simple2x32(unsigned char *srcPtr, unsigned int srcPitch, unsigned char *deltaPtr, unsigned char *dstPtr, unsigned int dstPitch, int width, int height)
+{
+    unsigned char *nextLine, *finish;
+  
+	nextLine = dstPtr + dstPitch;
+  
+	do 
+	{
+		unsigned int *bP = (unsigned int *) srcPtr;
+	    unsigned int *dP = (unsigned int *) dstPtr;
+		unsigned int *nL = (unsigned int *) nextLine;
+		unsigned int currentPixel;
+    
+		finish = (unsigned char *) bP + ((width+1) << 2);
+		currentPixel = *bP++;
+    
+		do 
+		{
+			unsigned int color = currentPixel;
+
+		    *(dP) = color;
+			*(dP+1) = color;
+			*(nL) = color;
+			*(nL + 1) = color;
+      
+			currentPixel = *bP++;
+	      
+			dP += 2;
+			nL += 2;
+		}
+		while ((unsigned char *) bP < finish);
+    
+		srcPtr += srcPitch;
+		dstPtr += dstPitch << 1;
+		nextLine += dstPitch << 1;
+	}
+	while (--height);
+}