- Added an assembly version of rt_shaded4cols, since that's the main decal

drawing function. The most improvement came from being able to turn some
  constant variables into immediate values with self-modifying code, but I
  also managed to reorder it to make it a little faster. It's about 9% faster
  than VC++'s code and 19% faster than GCC's code. That's not a huge
  improvement (for VC++), but at least it's measurable.
- Removed the solid fill "optimization" from rt_shaded4cols(), because in my
  testing, it didn't help any and in fact, hurt just a little bit.
- In the name of simplification, all the rt_tlate* drawers were changed to do
  the translation in one step and the drawing in another. This lets me call
  the untranslated drawer to do the real drawing instead of mostly duplicating
  them. Performance wise, there is practically no difference from before.


SVN r771 (trunk)

docs/rh-log.txt

@@ -1,3 +1,17 @@
+February 26, 2008
+- Added an assembly version of rt_shaded4cols, since that's the main decal
+  drawing function. The most improvement came from being able to turn some
+  constant variables into immediate values with self-modifying code, but I
+  also managed to reorder it to make it a little faster. It's about 9% faster
+  than VC++'s code and 19% faster than GCC's code. That's not a huge
+  improvement (for VC++), but at least it's measurable.
+- Removed the solid fill "optimization" from rt_shaded4cols(), because in my
+  testing, it didn't help any and in fact, hurt just a little bit.
+- In the name of simplification, all the rt_tlate* drawers were changed to do
+  the translation in one step and the drawing in another. This lets me call
+  the untranslated drawer to do the real drawing instead of mostly duplicating
+  them. Performance wise, there is practically no difference from before.
+
 February 25, 2008 (Changes by Graf Zahl)
 - Fixed: The DECORATE expression evaluator's random function could produce
   incorrect results for ranges > 255. Changed so that FRandom's default

src/r_draw.cpp

@@ -80,6 +80,7 @@ int				detailyshift;		// [RH] Y shift for vertical detail level
 extern "C" void STACK_ARGS DoubleHoriz_MMX (int height, int width, BYTE *dest, int pitch);
 extern "C" void STACK_ARGS DoubleHorizVert_MMX (int height, int width, BYTE *dest, int pitch);
 extern "C" void STACK_ARGS DoubleVert_ASM (int height, int width, BYTE *dest, int pitch);
+extern "C" void R_SetupShadedCol();
 #endif
 
 // [RH] Pointers to the different column drawers.
@@ -94,7 +95,7 @@ void (*R_DrawSpan)(void);
 void (*R_DrawSpanMasked)(void);
 void (*R_DrawSpanTranslucent)(void);
 void (*R_DrawSpanMaskedTranslucent)(void);
-void (*rt_map4cols)(int,int,int);
+void (STACK_ARGS *rt_map4cols)(int,int,int);
 
 //
 // R_DrawColumn
@@ -2270,6 +2271,7 @@ ESPSResult R_SetPatchStyle (FRenderStyle style, fixed_t alpha, int translation,
 		{
 			dc_colormap += fixedlightlev;
 		}
+		R_SetupShadedCol();
 		return r_columnmethod ? DoDraw1 : DoDraw0;
 	}
 

src/r_draw.h

@@ -106,47 +106,59 @@ void R_DrawMaskedColumnHoriz (const BYTE *column, const FTexture::Span *spans);
 void R_InitColumnDrawers ();
 
 // [RH] Moves data from the temporary buffer to the screen.
+extern "C"
+{
 void rt_copy1col_c (int hx, int sx, int yl, int yh);
-void rt_copy4cols_c (int sx, int yl, int yh);
+void STACK_ARGS rt_copy4cols_c (int sx, int yl, int yh);
-void rt_map1col_c (int hx, int sx, int yl, int yh);
+
-void rt_map4cols_c (int sx, int yl, int yh);
-void rt_add1col (int hx, int sx, int yl, int yh);
-void rt_add4cols (int sx, int yl, int yh);
-void rt_tlate1col (int hx, int sx, int yl, int yh);
-void rt_tlate4cols (int sx, int yl, int yh);
-void rt_tlateadd1col (int hx, int sx, int yl, int yh);
-void rt_tlateadd4cols (int sx, int yl, int yh);
 void rt_shaded1col (int hx, int sx, int yl, int yh);
-void rt_shaded4cols (int sx, int yl, int yh);
+void STACK_ARGS rt_shaded4cols_c (int sx, int yl, int yh);
+void STACK_ARGS rt_shaded4cols_asm (int sx, int yl, int yh);
+
+void rt_map1col_c (int hx, int sx, int yl, int yh);
+void rt_add1col (int hx, int sx, int yl, int yh);
 void rt_addclamp1col (int hx, int sx, int yl, int yh);
-void rt_addclamp4cols (int sx, int yl, int yh);
-void rt_tlateaddclamp1col (int hx, int sx, int yl, int yh);
-void rt_tlateaddclamp4cols (int sx, int yl, int yh);
 void rt_subclamp1col (int hx, int sx, int yl, int yh);
-void rt_subclamp4cols (int sx, int yl, int yh);
-void rt_tlatesubclamp1col (int hx, int sx, int yl, int yh);
-void rt_tlatesubclamp4cols (int sx, int yl, int yh);
 void rt_revsubclamp1col (int hx, int sx, int yl, int yh);
-void rt_revsubclamp4cols (int sx, int yl, int yh);
+
+void rt_tlate1col (int hx, int sx, int yl, int yh);
+void rt_tlateadd1col (int hx, int sx, int yl, int yh);
+void rt_tlateaddclamp1col (int hx, int sx, int yl, int yh);
+void rt_tlatesubclamp1col (int hx, int sx, int yl, int yh);
 void rt_tlaterevsubclamp1col (int hx, int sx, int yl, int yh);
-void rt_tlaterevsubclamp4cols (int sx, int yl, int yh);
 
-extern "C" void rt_copy1col_asm (int hx, int sx, int yl, int yh);
+void STACK_ARGS rt_map4cols_c (int sx, int yl, int yh);
-extern "C" void rt_copy4cols_asm (int sx, int yl, int yh);
+void STACK_ARGS rt_add4cols (int sx, int yl, int yh);
-extern "C" void rt_map1col_asm (int hx, int sx, int yl, int yh);
+void STACK_ARGS rt_addclamp4cols (int sx, int yl, int yh);
-extern "C" void rt_map4cols_asm1 (int sx, int yl, int yh);
+void STACK_ARGS rt_subclamp4cols (int sx, int yl, int yh);
-extern "C" void rt_map4cols_asm2 (int sx, int yl, int yh);
+void STACK_ARGS rt_revsubclamp4cols (int sx, int yl, int yh);
 
-extern void (*rt_map4cols)(int sx, int yl, int yh);
+void STACK_ARGS rt_tlate4cols (int sx, int yl, int yh);
+void STACK_ARGS rt_tlateadd4cols (int sx, int yl, int yh);
+void STACK_ARGS rt_tlateaddclamp4cols (int sx, int yl, int yh);
+void STACK_ARGS rt_tlatesubclamp4cols (int sx, int yl, int yh);
+void STACK_ARGS rt_tlaterevsubclamp4cols (int sx, int yl, int yh);
+
+void rt_copy1col_asm (int hx, int sx, int yl, int yh);
+void rt_map1col_asm (int hx, int sx, int yl, int yh);
+
+void STACK_ARGS rt_copy4cols_asm (int sx, int yl, int yh);
+void STACK_ARGS rt_map4cols_asm1 (int sx, int yl, int yh);
+void STACK_ARGS rt_map4cols_asm2 (int sx, int yl, int yh);
+}
+
+extern void (STACK_ARGS *rt_map4cols)(int sx, int yl, int yh);
 
 #ifdef USEASM
 #define rt_copy1col		rt_copy1col_asm
 #define rt_copy4cols	rt_copy4cols_asm
 #define rt_map1col		rt_map1col_asm
+#define rt_shaded4cols	rt_shaded4cols_asm
 #else
 #define rt_copy1col		rt_copy1col_c
 #define rt_copy4cols	rt_copy4cols_c
 #define rt_map1col		rt_map1col_c
+#define rt_shaded4cols	rt_shaded4cols_c
 #endif
 
 void rt_draw4cols (int sx);

src/r_drawt.cpp

@@ -102,7 +102,7 @@ void rt_copy1col_c (int hx, int sx, int yl, int yh)
 }
 
 // Copies all four spans to the screen starting at sx.
-void rt_copy4cols_c (int sx, int yl, int yh)
+void STACK_ARGS rt_copy4cols_c (int sx, int yl, int yh)
 {
 	int *source;
 	int *dest;
@@ -170,7 +170,7 @@ void rt_map1col_c (int hx, int sx, int yl, int yh)
 }
 
 // Maps all four spans to the screen starting at sx.
-void rt_map4cols_c (int sx, int yl, int yh)
+void STACK_ARGS rt_map4cols_c (int sx, int yl, int yh)
 {
 	BYTE *colormap;
 	BYTE *source;
@@ -214,63 +214,106 @@ void rt_map4cols_c (int sx, int yl, int yh)
 }
 #endif	/* !USEASM */
 
+void rt_Translate1col(const BYTE *translation, int hx, int yl, int yh)
+{
+	int count = yh - yl + 1;
+	BYTE *source = &dc_temp[yl*4 + hx];
+
+	// Things we do to hit the compiler's optimizer with a clue bat:
+	// 1. Parallelism is explicitly spelled out by using a separate
+	//    C instruction for each assembly instruction. GCC lets me
+	//    have four temporaries, but VC++ spills to the stack with
+	//    more than two. Two is probably optimal, anyway.
+	// 2. The results of the translation lookups are explicitly
+	//    stored in byte-sized variables. This causes the VC++ code
+	//    to use byte mov instructions in most cases; for apparently
+	//    random reasons, it will use movzx for some places. GCC
+	//    ignores this and uses movzx always.
+
+	// Do 8 rows at a time.
+	for (int count8 = count >> 3; count8; --count8)
+	{
+		int c0, c1;
+		BYTE b0, b1;
+
+		c0 = source[0];			c1 = source[4];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[0] = b0;			source[4] = b1;
+
+		c0 = source[8];			c1 = source[12];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[8] = b0;			source[12] = b1;
+
+		c0 = source[16];		c1 = source[20];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[16] = b0;		source[20] = b1;
+
+		c0 = source[24];		c1 = source[28];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[28] = b0;		source[28] = b1;
+
+		source += 32;
+	}
+	// Finish by doing 1 row at a time.
+	for (count &= 7; count; --count, source++)
+	{
+		source[0] = translation[source[0]];
+	}
+}
+
+void rt_Translate4cols(const BYTE *translation, int yl, int yh)
+{
+	int count = yh - yl + 1;
+	BYTE *source = &dc_temp[yl*4];
+	int c0, c1;
+	BYTE b0, b1;
+
+	// Do 2 rows at a time.
+	for (int count8 = count >> 1; count8; --count8)
+	{
+		c0 = source[0];			c1 = source[1];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[0] = b0;			source[1] = b1;
+
+		c0 = source[2];			c1 = source[3];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[2] = b0;			source[3] = b1;
+
+		c0 = source[4];			c1 = source[5];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[4] = b0;			source[5] = b1;
+
+		c0 = source[6];			c1 = source[7];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[6] = b0;			source[7] = b1;
+
+		source += 8;
+	}
+	// Do the final row if count was odd.
+	if (count & 1)
+	{
+		c0 = source[0];			c1 = source[1];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[0] = b0;			source[1] = b1;
+
+		c0 = source[2];			c1 = source[3];
+		b0 = translation[c0];	b1 = translation[c1];
+		source[2] = b0;			source[3] = b1;
+	}
+}
+
 // Translates one span at hx to the screen at sx.
 void rt_tlate1col (int hx, int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate1col(dc_translation, hx, yl, yh);
-	BYTE *colormap;
+	rt_map1col(hx, sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	translation = dc_translation;
-	colormap = dc_colormap;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4 + hx];
-	pitch = dc_pitch;
-
-	do {
-		*dest = colormap[translation[*source]];
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Translates all four spans to the screen starting at sx.
-void rt_tlate4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_tlate4cols (int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate4cols(dc_translation, yl, yh);
-	BYTE *colormap;
+	rt_map4cols(sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	translation = dc_translation;
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	colormap = dc_colormap;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4];
-	pitch = dc_pitch;
-	
-	do {
-		dest[0] = colormap[translation[source[0]]];
-		dest[1] = colormap[translation[source[1]]];
-		dest[2] = colormap[translation[source[2]]];
-		dest[3] = colormap[translation[source[3]]];
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Adds one span at hx to the screen at sx without clamping.
@@ -308,7 +351,7 @@ void rt_add1col (int hx, int sx, int yl, int yh)
 }
 
 // Adds all four spans to the screen starting at sx without clamping.
-void rt_add4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_add4cols (int sx, int yl, int yh)
 {
 	BYTE *colormap;
 	BYTE *source;
@@ -366,95 +409,15 @@ void rt_add4cols (int sx, int yl, int yh)
 // Translates and adds one span at hx to the screen at sx without clamping.
 void rt_tlateadd1col (int hx, int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate1col(dc_translation, hx, yl, yh);
-	BYTE *colormap;
+	rt_add1col(hx, sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	translation = dc_translation;
-	colormap = dc_colormap;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4 + hx];
-	pitch = dc_pitch;
-
-	do {
-		DWORD fg = colormap[translation[*source]];
-		DWORD bg = *dest;
-
-		fg = fg2rgb[fg];
-		bg = bg2rgb[bg];
-		fg = (fg+bg) | 0x1f07c1f;
-		*dest = RGB32k[0][0][fg & (fg>>15)];
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Translates and adds all four spans to the screen starting at sx without clamping.
-void rt_tlateadd4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_tlateadd4cols (int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate4cols(dc_translation, yl, yh);
-	BYTE *colormap;
+	rt_add4cols(sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	translation = dc_translation;
-	colormap = dc_colormap;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4];
-	pitch = dc_pitch;
-	
-	do {
-		DWORD fg = colormap[translation[source[0]]];
-		DWORD bg = dest[0];
-		fg = fg2rgb[fg];
-		bg = bg2rgb[bg];
-		fg = (fg+bg) | 0x1f07c1f;
-		dest[0] = RGB32k[0][0][fg & (fg>>15)];
-
-		fg = colormap[translation[source[1]]];
-		bg = dest[1];
-		fg = fg2rgb[fg];
-		bg = bg2rgb[bg];
-		fg = (fg+bg) | 0x1f07c1f;
-		dest[1] = RGB32k[0][0][fg & (fg>>15)];
-
-
-		fg = colormap[translation[source[2]]];
-		bg = dest[2];
-		fg = fg2rgb[fg];
-		bg = bg2rgb[bg];
-		fg = (fg+bg) | 0x1f07c1f;
-		dest[2] = RGB32k[0][0][fg & (fg>>15)];
-
-		fg = colormap[translation[source[3]]];
-		bg = dest[3];
-		fg = fg2rgb[fg];
-		bg = bg2rgb[bg];
-		fg = (fg+bg) | 0x1f07c1f;
-		dest[3] = RGB32k[0][0][fg & (fg>>15)];
-
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Shades one span at hx to the screen at sx.
@@ -489,9 +452,8 @@ void rt_shaded1col (int hx, int sx, int yl, int yh)
 }
 
 // Shades all four spans to the screen starting at sx.
-void rt_shaded4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_shaded4cols_c (int sx, int yl, int yh)
 {
-	BYTE fill;
 	DWORD *fgstart;
 	BYTE *colormap;
 	BYTE *source;
@@ -509,66 +471,73 @@ void rt_shaded4cols (int sx, int yl, int yh)
 	dest = ylookup[yl] + sx + dc_destorg;
 	source = &dc_temp[yl*4];
 	pitch = dc_pitch;
-	{
-		DWORD val = fgstart[64<<8] | 0x1f07c1f;
-		fill = RGB32k[0][0][val & (val>>15)];
-	}
 
+	// 107.1, 108.4, 118.2/117.7, 119.4
 	do {
-		DWORD val = colormap[source[0]];
+		DWORD val;
-		DWORD fg;
+		
-		if (val < 64)
+		val = colormap[source[0]];
-		{
+		val = (Col2RGB8[64-val][dest[0]] + fgstart[val<<8]) | 0x1f07c1f;
-			fg = fgstart[val<<8];
-			val = (Col2RGB8[64-val][dest[0]] + fg) | 0x1f07c1f;
 		dest[0] = RGB32k[0][0][val & (val>>15)];
-		}
-		else
-		{
-			dest[0] = fill;
-		}
 
 		val = colormap[source[1]];
-		if (val < 64)
+		val = (Col2RGB8[64-val][dest[1]] + fgstart[val<<8]) | 0x1f07c1f;
-		{
-			fg = fgstart[val<<8];
-			val = (Col2RGB8[64-val][dest[1]] + fg) | 0x1f07c1f;
 		dest[1] = RGB32k[0][0][val & (val>>15)];
-		}
-		else
-		{
-			dest[1] = fill;
-		}
 
 		val = colormap[source[2]];
-		if (val < 64)
+		val = (Col2RGB8[64-val][dest[2]] + fgstart[val<<8]) | 0x1f07c1f;
-		{
-			fg = fgstart[val<<8];
-			val = (Col2RGB8[64-val][dest[2]] + fg) | 0x1f07c1f;
 		dest[2] = RGB32k[0][0][val & (val>>15)];
-		}
-		else
-		{
-			dest[2] = fill;
-		}
 
 		val = colormap[source[3]];
-		if (val < 64)
+		val = (Col2RGB8[64-val][dest[3]] + fgstart[val<<8]) | 0x1f07c1f;
-		{
-			fg = fgstart[val<<8];
-			val = (Col2RGB8[64-val][dest[3]] + fg) | 0x1f07c1f;
 		dest[3] = RGB32k[0][0][val & (val>>15)];
-		}
-		else
-		{
-			dest[3] = fill;
-		}
 
 		source += 4;
 		dest += pitch;
 	} while (--count);
 }
 
+#if 0
+static DWORD t_fgstart[1];
+static BYTE t_colormap[1];
+
+void STACK_ARGS rt_shaded4cols_t (int sx, int yl, int yh)
+{
+	BYTE *source;
+	BYTE *dest;
+	int count;
+
+	count = yh-yl;
+	if (count < 0)
+		return;
+	count++;
+
+	dest = ylookup[yl] + sx + dc_destorg;
+	source = &dc_temp[yl*4];
+
+	do {
+		DWORD val, val2;
+		
+		val = t_colormap[source[0]];
+		val2 = t_colormap[source[1]];
+		val = (Col2RGB8[64-val][dest[0]] + t_fgstart[val<<8]) | 0x1f07c1f;
+		val2 = (Col2RGB8[64-val2][dest[1]] + t_fgstart[val2<<8]) | 0x1f07c1f;
+		dest[0] = RGB32k[0][0][val & (val>>15)];
+		dest[1] = RGB32k[0][0][val2 & (val2>>15)];
+
+		val = t_colormap[source[2]];
+		val2 = t_colormap[source[3]];
+		val = (Col2RGB8[64-val][dest[2]] + t_fgstart[val<<8]) | 0x1f07c1f;
+		val2 = (Col2RGB8[64-val2][dest[3]] + t_fgstart[val2<<8]) | 0x1f07c1f;
+		dest[2] = RGB32k[0][0][val & (val>>15)];
+		dest[3] = RGB32k[0][0][val2 & (val2>>15)];
+
+		source += 4;
+		dest += 320;
+	} while (--count);
+}
+#endif
+
 // Adds one span at hx to the screen at sx with clamping.
 void rt_addclamp1col (int hx, int sx, int yl, int yh)
 {
@@ -606,7 +575,7 @@ void rt_addclamp1col (int hx, int sx, int yl, int yh)
 }
 
 // Adds all four spans to the screen starting at sx with clamping.
-void rt_addclamp4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_addclamp4cols (int sx, int yl, int yh)
 {
 	BYTE *colormap;
 	BYTE *source;
@@ -672,105 +641,15 @@ void rt_addclamp4cols (int sx, int yl, int yh)
 // Translates and adds one span at hx to the screen at sx with clamping.
 void rt_tlateaddclamp1col (int hx, int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate1col(dc_translation, hx, yl, yh);
-	BYTE *colormap;
+	rt_addclamp1col(hx, sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4 + hx];
-	pitch = dc_pitch;
-	colormap = dc_colormap;
-	translation = dc_translation;
-
-	do {
-		DWORD a = fg2rgb[colormap[translation[*source]]] + bg2rgb[*dest];
-		DWORD b = a;
-
-		a |= 0x01f07c1f;
-		b &= 0x40100400;
-		a &= 0x3fffffff;
-		b = b - (b >> 5);
-		a |= b;
-		*dest = RGB32k[0][0][(a>>15) & a];
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Translates and adds all four spans to the screen starting at sx with clamping.
-void rt_tlateaddclamp4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_tlateaddclamp4cols (int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate4cols(dc_translation, yl, yh);
-	BYTE *colormap;
+	rt_addclamp4cols(sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4];
-	pitch = dc_pitch;
-	colormap = dc_colormap;
-	translation = dc_translation;
-
-	do {
-		DWORD a = fg2rgb[colormap[translation[source[0]]]] + bg2rgb[dest[0]];
-		DWORD b = a;
-
-		a |= 0x01f07c1f;
-		b &= 0x40100400;
-		a &= 0x3fffffff;
-		b = b - (b >> 5);
-		a |= b;
-		dest[0] = RGB32k[0][0][(a>>15) & a];
-
-		a = fg2rgb[colormap[translation[source[1]]]] + bg2rgb[dest[1]];
-		b = a;
-		a |= 0x01f07c1f;
-		b &= 0x40100400;
-		a &= 0x3fffffff;
-		b = b - (b >> 5);
-		a |= b;
-		dest[1] = RGB32k[0][0][(a>>15) & a];
-
-		a = fg2rgb[colormap[translation[source[2]]]] + bg2rgb[dest[2]];
-		b = a;
-		a |= 0x01f07c1f;
-		b &= 0x40100400;
-		a &= 0x3fffffff;
-		b = b - (b >> 5);
-		a |= b;
-		dest[2] = RGB32k[0][0][(a>>15) & a];
-
-		a = fg2rgb[colormap[translation[source[3]]]] + bg2rgb[dest[3]];
-		b = a;
-		a |= 0x01f07c1f;
-		b &= 0x40100400;
-		a &= 0x3fffffff;
-		b = b - (b >> 5);
-		a |= b;
-		dest[3] = RGB32k[0][0][(a>>15) & a];
-
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Subtracts one span at hx to the screen at sx with clamping.
@@ -809,7 +688,7 @@ void rt_subclamp1col (int hx, int sx, int yl, int yh)
 }
 
 // Subtracts all four spans to the screen starting at sx with clamping.
-void rt_subclamp4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_subclamp4cols (int sx, int yl, int yh)
 {
 	BYTE *colormap;
 	BYTE *source;
@@ -871,100 +750,15 @@ void rt_subclamp4cols (int sx, int yl, int yh)
 // Translates and subtracts one span at hx to the screen at sx with clamping.
 void rt_tlatesubclamp1col (int hx, int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate1col(dc_translation, hx, yl, yh);
-	BYTE *colormap;
+	rt_subclamp1col(hx, sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4 + hx];
-	pitch = dc_pitch;
-	colormap = dc_colormap;
-	translation = dc_translation;
-
-	do {
-		DWORD a = (fg2rgb[colormap[translation[*source]]] | 0x40100400) - bg2rgb[*dest];
-		DWORD b = a;
-
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		*dest = RGB32k[0][0][(a>>15) & a];
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Translates and subtracts all four spans to the screen starting at sx with clamping.
-void rt_tlatesubclamp4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_tlatesubclamp4cols (int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate4cols(dc_translation, yl, yh);
-	BYTE *colormap;
+	rt_subclamp4cols(sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4];
-	pitch = dc_pitch;
-	colormap = dc_colormap;
-	translation = dc_translation;
-
-	do {
-		DWORD a = (fg2rgb[colormap[translation[source[0]]]] | 0x40100400) - bg2rgb[dest[0]];
-		DWORD b = a;
-
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[0] = RGB32k[0][0][(a>>15) & a];
-
-		a = (fg2rgb[colormap[translation[source[1]]]] | 0x40100400) - bg2rgb[dest[1]];
-		b = a;
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[1] = RGB32k[0][0][(a>>15) & a];
-
-		a = (fg2rgb[colormap[translation[source[2]]]] | 0x40100400) - bg2rgb[dest[2]];
-		b = a;
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[2] = RGB32k[0][0][(a>>15) & a];
-
-		a = (fg2rgb[colormap[translation[source[3]]]] | 0x40100400) - bg2rgb[dest[3]];
-		b = a;
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[3] = RGB32k[0][0][(a>>15) & a];
-
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Subtracts one span at hx from the screen at sx with clamping.
@@ -1003,7 +797,7 @@ void rt_revsubclamp1col (int hx, int sx, int yl, int yh)
 }
 
 // Subtracts all four spans from the screen starting at sx with clamping.
-void rt_revsubclamp4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_revsubclamp4cols (int sx, int yl, int yh)
 {
 	BYTE *colormap;
 	BYTE *source;
@@ -1065,104 +859,19 @@ void rt_revsubclamp4cols (int sx, int yl, int yh)
 // Translates and subtracts one span at hx from the screen at sx with clamping.
 void rt_tlaterevsubclamp1col (int hx, int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate1col(dc_translation, hx, yl, yh);
-	BYTE *colormap;
+	rt_revsubclamp1col(hx, sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4 + hx];
-	pitch = dc_pitch;
-	colormap = dc_colormap;
-	translation = dc_translation;
-
-	do {
-		DWORD a = (bg2rgb[dest[0]] | 0x40100400) - fg2rgb[colormap[translation[source[0]]]];
-		DWORD b = a;
-
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		*dest = RGB32k[0][0][(a>>15) & a];
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Translates and subtracts all four spans from the screen starting at sx with clamping.
-void rt_tlaterevsubclamp4cols (int sx, int yl, int yh)
+void STACK_ARGS rt_tlaterevsubclamp4cols (int sx, int yl, int yh)
 {
-	BYTE *translation;
+	rt_Translate4cols(dc_translation, yl, yh);
-	BYTE *colormap;
+	rt_revsubclamp4cols(sx, yl, yh);
-	BYTE *source;
-	BYTE *dest;
-	int count;
-	int pitch;
-
-	count = yh-yl;
-	if (count < 0)
-		return;
-	count++;
-
-	DWORD *fg2rgb = dc_srcblend;
-	DWORD *bg2rgb = dc_destblend;
-	dest = ylookup[yl] + sx + dc_destorg;
-	source = &dc_temp[yl*4];
-	pitch = dc_pitch;
-	colormap = dc_colormap;
-	translation = dc_translation;
-
-	do {
-		DWORD a = (bg2rgb[dest[0]] | 0x40100400) - fg2rgb[colormap[translation[source[0]]]];
-		DWORD b = a;
-
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[0] = RGB32k[0][0][(a>>15) & a];
-
-		a = (bg2rgb[dest[1]] | 0x40100400) - fg2rgb[colormap[translation[source[1]]]];
-		b = a;
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[1] = RGB32k[0][0][(a>>15) & a];
-
-		a = (bg2rgb[dest[2]] | 0x40100400) - fg2rgb[colormap[translation[source[2]]]];
-		b = a;
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[2] = RGB32k[0][0][(a>>15) & a];
-
-		a = (bg2rgb[dest[3]] | 0x40100400) - fg2rgb[colormap[translation[source[3]]]];
-		b = a;
-		b &= 0x40100400;
-		b = b - (b >> 5);
-		a &= b;
-		a |= 0x01f07c1f;
-		dest[3] = RGB32k[0][0][(a>>15) & a];
-
-		source += 4;
-		dest += pitch;
-	} while (--count);
 }
 
 // Copies all spans in all four columns to the screen starting at sx.
-// sx should be longword-aligned.
+// sx should be dword-aligned.
 void rt_draw4cols (int sx)
 {
 	int x, bad;

src/r_main.cpp

@@ -205,7 +205,7 @@ void (*spanfunc) (void);
 void (*hcolfunc_pre) (void);
 void (*hcolfunc_post1) (int hx, int sx, int yl, int yh);
 void (*hcolfunc_post2) (int hx, int sx, int yl, int yh);
-void (*hcolfunc_post4) (int sx, int yl, int yh);
+void (STACK_ARGS *hcolfunc_post4) (int sx, int yl, int yh);
 
 cycle_t WallCycles, PlaneCycles, MaskedCycles, WallScanCycles;
 

src/r_main.h

@@ -147,7 +147,7 @@ extern void 			(*spanfunc) (void);
 extern void (*hcolfunc_pre) (void);
 extern void (*hcolfunc_post1) (int hx, int sx, int yl, int yh);
 extern void (*hcolfunc_post2) (int hx, int sx, int yl, int yh);
-extern void (*hcolfunc_post4) (int sx, int yl, int yh);
+extern void (STACK_ARGS *hcolfunc_post4) (int sx, int yl, int yh);
 
 
 //

src/tmap.nas

@@ -55,6 +55,7 @@ EXTERN CPU
 
 EXTERN dc_pitch
 EXTERN dc_colormap
+EXTERN dc_color
 EXTERN dc_iscale
 EXTERN dc_texturefrac
 EXTERN dc_source
@@ -68,6 +69,9 @@ EXTERN dc_destorg
 EXTERN dc_ctspan
 EXTERN dc_temp
 
+EXTERN Col2RGB8
+EXTERN RGB32k
+
 EXTERN ds_xstep
 EXTERN ds_ystep
 EXTERN ds_colormap
@@ -94,6 +98,7 @@ EXTERN _CPU
 
 EXTERN _dc_pitch
 EXTERN _dc_colormap
+EXTERN _dc_color
 EXTERN _dc_iscale
 EXTERN _dc_texturefrac
 EXTERN _dc_source
@@ -107,6 +112,9 @@ EXTERN _dc_destorg
 EXTERN _dc_ctspan
 EXTERN _dc_temp
 
+EXTERN _Col2RGB8
+EXTERN _RGB32k
+
 EXTERN _ds_xstep
 EXTERN _ds_ystep
 EXTERN _ds_colormap
@@ -131,6 +139,7 @@ GLOBAL _ds_curcolormap
 
 %define dc_pitch	_dc_pitch
 %define dc_colormap	_dc_colormap
+%define dc_color	_dc_color
 %define dc_iscale	_dc_iscale
 %define dc_texturefrac	_dc_texturefrac
 %define dc_source	_dc_source
@@ -141,6 +150,9 @@ GLOBAL _ds_curcolormap
 %define dc_dest		_dc_dest
 %define dc_destorg	_dc_destorg
 
+%define Col2RGB8	_Col2RGB8
+%define RGB32k		_RGB32k
+
 %define dc_ctspan	_dc_ctspan
 %define dc_temp		_dc_temp
 
@@ -1476,10 +1488,121 @@ _rt_map4cols_asm2:
 	pop	ebx
 	ret	4
 
+	align 16
+
+GLOBAL rt_shaded4cols_asm
+GLOBAL _rt_shaded4cols_asm
+
+rt_shaded4cols_asm:
+_rt_shaded4cols_asm:
+		mov		ecx,[esp+8]
+		push	ebp
+		mov		ebp,[esp+16]
+		sub		ebp,ecx
+		js		near s4nil
+		mov		eax,[ylookup+ecx*4]
+		add		eax,[dc_destorg]				; eax = destination
+		push	ebx
+		push	esi
+		inc		ebp								; ebp = count
+		add		eax,[esp+16]
+		push	edi
+		lea		esi,[dc_temp+ecx*4]				; esi = source
+
+		align	16
+
+s4loop:	movzx	edx,byte [esi]
+		movzx	ecx,byte [esi+1]
+s4cm1:	movzx	edx,byte [SPACEFILLER4+edx]		; colormap
+s4cm2:	movzx	edi,byte [SPACEFILLER4+ecx]		; colormap
+		shl		edx,8
+		movzx	ebx,byte [eax]
+		shl		edi,8
+		movzx	ecx,byte [eax+1]
+		sub		ebx,edx
+		sub		ecx,edi
+		mov		ebx,[Col2RGB8+0x10000+ebx*4]
+		mov		ecx,[Col2RGB8+0x10000+ecx*4]
+s4fg1:	add		ebx,[SPACEFILLER4+edx*4]
+s4fg2:	add		ecx,[SPACEFILLER4+edi*4]
+		or		ebx,0x1f07c1f
+		or		ecx,0x1f07c1f
+		mov		edx,ebx
+		shr		ebx,15
+		mov		edi,ecx
+		shr		ecx,15
+		and		edx,ebx
+		and		ecx,edi
+		mov		bl,[RGB32k+edx]
+		movzx	edx,byte [esi+2]
+		mov		bh,[RGB32k+ecx]
+		movzx	ecx,byte [esi+3]
+		mov		[eax],bl
+		mov		[eax+1],bh
+
+s4cm3:	movzx	edx,byte [SPACEFILLER4+edx]		; colormap
+s4cm4:	movzx	edi,byte [SPACEFILLER4+ecx]		; colormap
+		shl		edx,8
+		movzx	ebx,byte [eax+2]
+		shl		edi,8
+		movzx	ecx,byte [eax+3]
+		sub		ebx,edx
+		sub		ecx,edi
+		mov		ebx,[Col2RGB8+0x10000+ebx*4]
+		mov		ecx,[Col2RGB8+0x10000+ecx*4]
+s4fg3:	add		ebx,[SPACEFILLER4+edx*4]
+s4fg4:	add		ecx,[SPACEFILLER4+edi*4]
+		or		ebx,0x1f07c1f
+		or		ecx,0x1f07c1f
+		mov		edx,ebx
+		shr		ebx,15
+		mov		edi,ecx
+		shr		ecx,15
+		and		edx,ebx
+		and		ecx,edi
+s4p:	add		eax,320							; pitch
+		add		esi,4
+		mov		bl,[RGB32k+edx]
+		mov		bh,[RGB32k+ecx]
+s4p2:	mov		[eax-320+2],bl
+s4p3:	mov		[eax-320+3],bh
+		dec		ebp
+		jne		s4loop
+
+		pop		edi
+		pop		esi
+		pop		ebx
+s4nil:	pop		ebp
+		ret
+
+		align 16
+
 ;************************
 
 	SECTION .text
 
+GLOBAL	R_SetupShadedCol
+GLOBAL	_R_SetupShadedCol
+GLOBAL	@R_SetupShadedCol@0
+
+# Patch the values of dc_colormap and dc_color into the shaded column drawer.
+
+R_SetupShadedCol:
+_R_SetupShadedCol:
+@R_SetupShadedCol@0:
+		mov		eax,[dc_colormap]
+		mov		[s4cm1+3],eax
+		mov		[s4cm2+3],eax
+		mov		[s4cm3+3],eax
+		mov		[s4cm4+3],eax
+		mov		eax,[dc_color]
+		lea		eax,[Col2RGB8+eax*4]
+		mov		[s4fg1+3],eax
+		mov		[s4fg2+3],eax
+		mov		[s4fg3+3],eax
+		mov		[s4fg4+3],eax
+		ret
+
 EXTERN setvlinebpl_
 EXTERN setpitch3
 
@@ -1494,6 +1617,14 @@ _ASM_PatchPitch:
 		mov		[rdcp1+2],eax
 		mov		[rdcp2+2],eax
 		mov		[rdcp3+2],eax
+		mov		[s4p+1],eax
+		mov		ecx,eax
+		neg		ecx
+		inc		ecx
+		inc		ecx
+		mov		[s4p2+2],ecx
+		inc		ecx
+		mov		[s4p3+2],ecx
 		call	setpitch3
 		jmp		setvlinebpl_