SWAR implementations of the image drawing functions, plus some tweaks to the vanilla versions for speed.

2 files changed, 390 insertions, 23 deletions

diff --git a/draw/archport.c b/draw/archport.c
index 93e50fe9..7b70e84f 100644
--- a/draw/archport.c
+++ b/draw/archport.c
@@ -197,7 +197,7 @@ text_w4i1o4_32bit(byte * restrict argb,
 				ca += ca>>7;
 				ca = (ca*alpha)>>8;
 				if (ca == 0)
-				 continue;
+					continue;
 				drb = dag & MASK;
 				dag = (dag<<8) & MASK;
 				crb = rb - (drb>>8);
@@ -226,7 +226,7 @@ text_w4i1o4_32bit(byte * restrict argb,
 				dag = *dst32++;
 				ca += ca>>7;
 				if (ca == 0)
-				 continue;
+					continue;
 				drb = dag & MASK;
 				dag = (dag<<8) & MASK;
 				crb = rb - (drb>>8);
@@ -244,6 +244,320 @@ text_w4i1o4_32bit(byte * restrict argb,
 	}
 }
 
+static void
+img_4o4_32bit(byte * restrict src, byte cov, int len, byte * restrict dst,
+	fz_pixmap *image, int u, int v, int fa, int fb)
+{
+	unsigned int *dst32 = (unsigned int *)(void *)dst;
+	unsigned int *samples = (unsigned int *)(void *)image->samples;
+	int w = image->w;
+	int h = image->h-1;
+
+	while (len--)
+	{
+		unsigned int a, a1, d, d1;
+		int sa;
+		cov += *src; *src = 0; src++;
+		/* (a,a1) = sampleargb(samples, w, h, u, v, argb); */
+		{
+			int ui, ui1, vi, vi1, ud, vd;
+			unsigned int b, b1, c, c1;
+			ui1 = 1;
+			ui = u >> 16;
+			if (ui < 0)
+			{
+				ui = 0;
+				ui1 = 0;
+			}
+			else if (ui >= w-1)
+			{
+				ui = w-1;
+				ui1 = 0;
+			}
+			vi1 = w;
+			vi = v >> 16;
+			if (vi < 0)
+			{
+				vi = 0;
+				vi1 = 0;
+			}
+			else if (vi >= h)
+			{
+				vi = h;
+				vi1 = 0;
+			}
+			ui += vi*w;
+			a = samples[ui];
+			b = samples[ui + ui1];
+			c = samples[ui + vi1];
+			d = samples[ui + ui1 + vi1];
+			ud = (u>>8) & 0xFF;
+			vd = (v>>8) & 0xFF;
+			ud = FZ_EXPAND(ud);
+			vd = FZ_EXPAND(vd);
+			/* (a,a1) = blend(a,b,ud) */
+			a1 = a & MASK;
+			a = (a<<8) & MASK;
+			b1 = (b>>8) & ~MASK;
+			b = b & ~MASK;
+			a = ((b -(a >>8)) * ud + a ) & MASK;
+			a1 = ((b1-(a1>>8)) * ud + a1) & MASK;
+			/* (c,c1) = blend(c,d,ud) */
+			c1 = c & MASK;
+			c = (c<<8) & MASK;
+			d1 = (d>>8) & ~MASK;
+			d = d & ~MASK;
+			c = ((d -(c >>8)) * ud + c ) & MASK;
+			c1 = ((d1-(c1>>8)) * ud + c1) & MASK;
+			/* (a,a1) = blend((a,a1),(c,c1),vd) */
+			a = (((c >>8)-(a >>8)) * vd + a ) & MASK;
+			a1 = (((c1>>8)-(a1>>8)) * vd + a1) & MASK;
+		}
+		sa = (a>>8) & 0xFF;
+		sa = FZ_COMBINE(FZ_EXPAND(sa), FZ_EXPAND(cov));
+		a |= 0xFF00;
+		d = *dst32++;
+		d1 = d & MASK;
+		d = (d<<8) & MASK;
+		a = (((a >>8)-(d >>8)) * sa + d ) & MASK;
+		a1 = (((a1>>8)-(d1>>8)) * sa + d1) & MASK;
+		dst32[-1] = (a>>8) | a1;
+		u += fa;
+		v += fb;
+	}
+}
+
+static void
+img_w4i1o4_32bit(byte *argb, byte * restrict src, byte cov, int len,
+	byte * restrict dst, fz_pixmap *image, int u, int v, int fa, int fb)
+{
+	byte *samples = image->samples;
+	int w = image->w;
+	int h = image->h-1;
+	int alpha = FZ_EXPAND(argb[0]);
+	unsigned int rb = argb[1] | (argb[3] << 16);
+	unsigned int ag = 255 | (argb[2] << 16);
+	unsigned int *dst32 = (unsigned int *)(void *)dst;
+
+	if (alpha == 0)
+		return;
+	if (alpha != 256)
+	{
+		while (len--)
+		{
+			unsigned int ca, drb, dag, crb, cag;
+			unsigned int a, b;
+			cov += *src; *src = 0; src++;
+			dag = *dst32++;
+			ca = FZ_COMBINE(FZ_EXPAND(cov), alpha);
+			if (ca != 0)
+			{
+				int ui, ui1, vi, vi1, ud, vd;
+				/* a = samplemask(samples, w, h, u, v); */
+				ui1 = 1;
+				ui = u >> 16;
+				if (ui < 0)
+				{
+					ui = 0;
+					ui1 = 0;
+				}
+				else if (ui >= w-1)
+				{
+					ui = w-1;
+					ui1 = 0;
+				}
+				vi1 = w;
+				vi = v >> 16;
+				if (vi < 0)
+				{
+				 	vi = 0;
+					vi1 = 0;
+				}
+				else if (vi >= h)
+				{
+					vi = h;
+					vi1 = 0;
+				}
+				ui += vi*w;
+				a = samples[ui];
+				b = samples[ui + ui1];
+				a |= samples[ui + vi1]<<16;
+				b |= samples[ui + ui1 + vi1]<<16;
+				ud = (u>>8) & 0xFF;
+				vd = (v>>8) & 0xFF;
+				ud = FZ_EXPAND(ud);
+				vd = FZ_EXPAND(vd);
+				/* a = blend(a,b,ud) */
+				a = ((b-a) * ud + (a<<8)) & MASK;
+				/* a = blend(a,a>>16,vd) */
+				a = (((a>>24)-(a>>8)) * vd + a);
+				a = (a>>8) & 0xFF;
+				ca = FZ_COMBINE(ca, FZ_EXPAND(a));
+			}
+			if (ca != 0)
+			{
+				drb = dag & MASK;
+				dag = (dag<<8) & MASK;
+				crb = rb - (drb>>8);
+				cag = ag - (dag>>8);
+				drb += crb * ca;
+				dag += cag * ca;
+				drb &= MASK;
+				dag &= MASK;
+				dag = drb | (dag>>8);
+				dst32[-1] = dag;
+			}
+			u += fa;
+			v += fb;
+		}
+	}
+	else
+	{
+		while (len--)
+		{
+			unsigned int ca, drb, dag, crb, cag;
+			unsigned int a, b;
+			cov += *src; *src = 0; src++;
+			dag = *dst32++;
+			if (cov != 0)
+			{
+				int ui, ui1, vi, vi1, ud, vd;
+				/* a = samplemask(samples, w, h, u, v); */
+				ui1 = 1;
+				ui = u >> 16;
+				if (ui < 0)
+				{
+					ui = 0;
+					ui1 = 0;
+				}
+				else if (ui >= w-1)
+				{
+					ui = w-1;
+					ui1 = 0;
+				}
+				vi1 = w;
+				vi = v >> 16;
+				if (vi < 0)
+				{
+				 	vi = 0;
+					vi1 = 0;
+				}
+				else if (vi >= h)
+				{
+					vi = h;
+					vi1 = 0;
+				}
+				ui += vi*w;
+				a = samples[ui];
+				b = samples[ui + ui1];
+				a |= samples[ui + vi1]<<16;
+				b |= samples[ui + ui1 + vi1]<<16;
+				ud = (u>>8) & 0xFF;
+				vd = (v>>8) & 0xFF;
+				ud = FZ_EXPAND(ud);
+				vd = FZ_EXPAND(vd);
+				/* a = blend(a,b,ud) */
+				a = ((b-a) * ud + (a<<8)) & MASK;
+				/* a = blend(a,a>>16,vd) */
+				a = (((a>>24)-(a>>8)) * vd + a);
+				a = (a>>8) & 0xFF;
+				ca = FZ_COMBINE(FZ_EXPAND(cov),FZ_EXPAND(a));
+				if (ca != 0)
+				{
+					if (ca == 256)
+					{
+						dag = (rb<<8) | ag;
+					}
+					else
+					{
+						drb = dag & MASK;
+						dag = (dag<<8) & MASK;
+						crb = rb - (drb>>8);
+						cag = ag - (dag>>8);
+						drb += crb * ca;
+						dag += cag * ca;
+						drb &= MASK;
+						dag &= MASK;
+						dag = drb | (dag>>8);
+					}
+					dst32[-1] = dag;
+				}
+			}
+			u += fa;
+			v += fb;
+		}
+	}
+}
+
+static void
+img_1o1_32bit(byte * restrict src, byte cov, int len, byte * restrict dst,
+	fz_pixmap *image, int u, int v, int fa, int fb)
+{
+	byte *samples = image->samples;
+	int w = image->w;
+	int h = image->h-1;
+
+	while (len--)
+	{
+		unsigned int a, b;
+		cov += *src; *src = 0; src++;
+		if (cov != 0)
+		{
+			int ui, ui1, vi, vi1, ud, vd;
+			/* sa = samplemask(samples, w, h, u, v); */
+			ui1 = 1;
+			ui = u >> 16;
+			if (ui < 0)
+			{
+				ui = 0;
+				ui1 = 0;
+			}
+			else if (ui >= w-1)
+			{
+				ui = w-1;
+				ui1 = 0;
+			}
+			vi1 = w;
+			vi = v >> 16;
+			if (vi < 0)
+			{
+			 	vi = 0;
+				vi1 = 0;
+			}
+			else if (vi >= h)
+			{
+				vi = h;
+				vi1 = 0;
+			}
+			ui += vi*w;
+			a = samples[ui];
+			b = samples[ui + ui1];
+			a |= samples[ui + vi1]<<16;
+			b |= samples[ui + ui1 + vi1]<<16;
+			ud = (u>>8) & 0xFF;
+			vd = (v>>8) & 0xFF;
+			ud = FZ_EXPAND(ud);
+			vd = FZ_EXPAND(vd);
+			/* a = blend(a,b,ud) */
+			a = ((b-a) * ud + (a<<8)) & MASK;
+			/* a = blend(a,a>>16,vd) */
+			a = (((a>>24)-(a>>8)) * vd + a);
+			a = (a>>8) & 0xFF;
+			a = FZ_COMBINE(FZ_EXPAND(a), FZ_EXPAND(cov));
+			if (a != 0)
+			{
+				if (a == 256)
+					dst[0] = 255;
+				else
+					dst[0] = FZ_BLEND(255, dst[0], a);
+			}
+		}
+		dst++;
+		u += fa;
+		v += fb;
+	}
+}
+
 void fz_accelerate(void)
 {
 	if (sizeof(int) == 4 && sizeof(unsigned int) == 4)
@@ -252,6 +566,9 @@ void fz_accelerate(void)
 		fz_duff_4i1o4 = duff_4i1o4_32bit;
 		fz_path_w4i1o4 = path_w4i1o4_32bit;
 		fz_text_w4i1o4 = text_w4i1o4_32bit;
+		fz_img_4o4 = img_4o4_32bit;
+		fz_img_w4i1o4 = img_w4i1o4_32bit;
+		fz_img_1o1 = img_1o1_32bit;
 	}
 
 	if (sizeof(int) == 8)
diff --git a/draw/imagedraw.c b/draw/imagedraw.c
index 556724bb..9f6d93bf 100644
--- a/draw/imagedraw.c
+++ b/draw/imagedraw.c
@@ -108,13 +108,20 @@ img_1o1(byte * restrict src, byte cov, int len, byte * restrict dst,
 	byte *samples = image->samples;
 	int w = image->w;
 	int h = image->h;
+
 	while (len--)
 	{
 		int sa;
 		cov += *src; *src = 0; src++;
-		sa = samplemask(samples, w, h, u, v);
-		sa = FZ_COMBINE(FZ_EXPAND(sa), FZ_EXPAND(cov));
-		dst[0] = FZ_BLEND(255, dst[0], sa);
+		if (cov != 0)
+		{
+			sa = samplemask(samples, w, h, u, v);
+			sa = FZ_COMBINE(FZ_EXPAND(sa), FZ_EXPAND(cov));
+			if (sa != 0)
+			{
+				dst[0] = FZ_BLEND(255, dst[0], sa);
+			}
+		}
 		dst++;
 		u += fa;
 		v += fb;
@@ -129,16 +136,23 @@ img_4o4(byte * restrict src, byte cov, int len, byte * restrict dst,
 	int w = image->w;
 	int h = image->h;
 	byte argb[4];
+
 	while (len--)
 	{
 		int sa;
 		cov += *src; *src = 0; src++;
-		sampleargb(samples, w, h, u, v, argb);
-		sa = FZ_COMBINE(FZ_EXPAND(argb[0]), FZ_EXPAND(cov));
-		dst[0] = FZ_BLEND(255, dst[0], sa);
-		dst[1] = FZ_BLEND(argb[1], dst[1], sa);
-		dst[2] = FZ_BLEND(argb[2], dst[2], sa);
-		dst[3] = FZ_BLEND(argb[3], dst[3], sa);
+		if (cov != 0)
+		{
+			sampleargb(samples, w, h, u, v, argb);
+			sa = FZ_COMBINE(FZ_EXPAND(argb[0]), FZ_EXPAND(cov));
+			if (sa != 0)
+			{
+				dst[0] = FZ_BLEND(255, dst[0], sa);
+				dst[1] = FZ_BLEND(argb[1], dst[1], sa);
+				dst[2] = FZ_BLEND(argb[2], dst[2], sa);
+				dst[3] = FZ_BLEND(argb[3], dst[3], sa);
+			}
+		}
 		dst += 4;
 		u += fa;
 		v += fb;
@@ -156,19 +170,55 @@ img_w4i1o4(byte *argb, byte * restrict src, byte cov, int len, byte * restrict d
 	byte r = argb[1];
 	byte g = argb[2];
 	byte b = argb[3];
-	while (len--)
+
+	if (alpha == 0)
+		return;
+	if (alpha != 256)
 	{
-		int ca;
-		cov += *src; *src = 0; src++;
-		ca = samplemask(samples, w, h, u, v);
-		ca = FZ_COMBINE(FZ_EXPAND(ca), alpha);
-		dst[0] = FZ_BLEND(255, dst[0], ca);
-		dst[1] = FZ_BLEND(r, dst[1], ca);
-		dst[2] = FZ_BLEND(g, dst[2], ca);
-		dst[3] = FZ_BLEND(b, dst[3], ca);
-		dst += 4;
-		u += fa;
-		v += fb;
+		while (len--)
+		{
+			int ca;
+			cov += *src; *src = 0; src++;
+			if (cov != 0)
+			{
+				ca = samplemask(samples, w, h, u, v);
+				ca =FZ_COMBINE(FZ_EXPAND(cov),FZ_EXPAND(ca));
+				ca = FZ_COMBINE(ca, alpha);
+				if (ca != 0)
+				{
+					dst[0] = FZ_BLEND(255, dst[0], ca);
+					dst[1] = FZ_BLEND(r, dst[1], ca);
+					dst[2] = FZ_BLEND(g, dst[2], ca);
+					dst[3] = FZ_BLEND(b, dst[3], ca);
+				}
+			}
+			dst += 4;
+			u += fa;
+			v += fb;
+		}
+	}
+	else
+	{
+		while (len--)
+		{
+			int ca;
+			cov += *src; *src = 0; src++;
+			if (cov != 0)
+			{
+				ca = samplemask(samples, w, h, u, v);
+				ca =FZ_COMBINE(FZ_EXPAND(cov),FZ_EXPAND(ca));
+				if (ca != 0)
+				{
+					dst[0] = FZ_BLEND(255, dst[0], ca);
+					dst[1] = FZ_BLEND(r, dst[1], ca);
+					dst[2] = FZ_BLEND(g, dst[2], ca);
+					dst[3] = FZ_BLEND(b, dst[3], ca);
+				}
+			}
+			dst += 4;
+			u += fa;
+			v += fb;
+		}
 	}
 }