Software multithread

src/CMakeLists.txt

@@ -8,8 +8,6 @@ target_sources(SRB2SDL2 PRIVATE comptime.c md5.c config.h.in)
 
 set(SRB2_ASM_SOURCES vid_copy.s)
 
-set(SRB2_NASM_SOURCES tmap_mmx.nas tmap.nas)
-
 ### Configuration
 set(SRB2_CONFIG_HAVE_PNG ON CACHE BOOL
 	"Enable PNG support. Depends on zlib, so will be disabled if you don't enable that too.")
@@ -213,8 +211,7 @@ if(${SRB2_CONFIG_USEASM})
 	target_compile_definitions(SRB2SDL2 PRIVATE -DUSEASM)
 	set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -msse3 -mfpmath=sse")
 
-	target_sources(SRB2SDL2 PRIVATE ${SRB2_ASM_SOURCES}
+	target_sources(SRB2SDL2 PRIVATE ${SRB2_ASM_SOURCES})
-		${SRB2_NASM_SOURCES})
 else()
 	set(SRB2_USEASM OFF)
 	target_compile_definitions(SRB2SDL2 PRIVATE -DNONX86 -DNORUSEASM)

src/Makefile

@@ -210,7 +210,7 @@ sources+=\
 	$(call List,blua/Sourcefile)\
 
 depends:=$(basename $(filter %.c %.s,$(sources)))
-objects:=$(basename $(filter %.c %.s %.nas,$(sources)))
+objects:=$(basename $(filter %.c %.s,$(sources)))
 
 depends:=$(depends:%=$(depdir)/%.d)
 

src/Makefile.d/features.mk

@@ -20,7 +20,6 @@ endif
 
 ifndef NOASM
 ifndef NONX86
-sources+=tmap.nas tmap_mmx.nas
 opts+=-DUSEASM
 endif
 endif

src/d_main.c

@@ -511,8 +511,10 @@ static void D_Display(void)
 						M_Memcpy(ylookup, ylookup1, viewheight*sizeof (ylookup[0]));
 					}
 				}
+				ps_rendercalltime = I_GetPreciseTime() - ps_rendercalltime;
 
 				// Image postprocessing effect
+				ps_postprocesstime = I_GetPreciseTime();
 				if (rendermode == render_soft)
 				{
 					if (!splitscreen)
@@ -523,7 +525,7 @@ static void D_Display(void)
 					if (postimgtype2)
 						V_DoPostProcessor(1, postimgtype2, postimgparam2);
 				}
-				ps_rendercalltime = I_GetPreciseTime() - ps_rendercalltime;
+				ps_postprocesstime = I_GetPreciseTime() - ps_postprocesstime;
 			}
 
 			if (lastdraw)

src/doomdef.h

@@ -537,6 +537,9 @@ INT32 I_GetKey(void);
 // Max gamepad/joysticks that can be detected/used.
 #define MAX_JOYSTICKS 4
 
+// Software multithreading
+#define MAX_RENDER_THREADS 8
+
 #ifndef M_PIl
 #define M_PIl 3.1415926535897932384626433832795029L
 #endif
@@ -608,13 +611,6 @@ extern const char *compdate, *comptime, *comprevision, *compbranch;
 /// Experimental attempts at preventing MF_PAPERCOLLISION objects from getting stuck in walls.
 //#define PAPER_COLLISIONCORRECTION
 
-/// FINALLY some real clipping that doesn't make walls dissappear AND speeds the game up
-/// (that was the original comment from SRB2CB, sadly it is a lie and actually slows game down)
-/// on the bright side it fixes some weird issues with translucent walls
-/// \note	SRB2CB port.
-///      	SRB2CB itself ported this from PrBoom+
-#define NEWCLIP
-
 /// OpenGL shaders
 #define GL_SHADERS
 

src/hardware/hw_main.h

@@ -35,7 +35,6 @@ void HWR_ClearSkyDome(void);
 void HWR_BuildSkyDome(void);
 void HWR_DrawViewBorder(INT32 clearlines);
 void HWR_DrawFlatFill(INT32 x, INT32 y, INT32 w, INT32 h, lumpnum_t flatlumpnum);
-void HWR_InitTextureMapping(void);
 void HWR_SetViewSize(void);
 void HWR_DrawPatch(patch_t *gpatch, INT32 x, INT32 y, INT32 option);
 void HWR_DrawStretchyFixedPatch(patch_t *gpatch, fixed_t x, fixed_t y, fixed_t pscale, fixed_t vscale, INT32 option, const UINT8 *colormap);

src/i_threads.h

@@ -15,6 +15,17 @@
 #define I_THREADS_H
 
 typedef void (*I_thread_fn)(void *userdata);
+typedef void * I_thread_handle;
+
+#ifdef HAVE_SDL
+#include <SDL.h>
+#define ATOMIC_TYPE SDL_atomic_t
+#else
+#define ATOMIC_TYPE UINT32
+#endif
+
+typedef ATOMIC_TYPE I_Atomicval_t;
+typedef I_Atomicval_t * I_Atomicptr_t;
 
 typedef void * I_mutex;
 typedef void * I_cond;
@@ -22,7 +33,7 @@ typedef void * I_cond;
 void            I_start_threads (void);
 void            I_stop_threads  (void);
 
-void      I_spawn_thread (const char *name, I_thread_fn, void *userdata);
+I_thread_handle I_spawn_thread (const char *name, I_thread_fn, void *userdata);
 
 /* check in your thread whether to return early */
 int             I_thread_is_stopped (void);
@@ -35,5 +46,8 @@ void      I_hold_cond       (I_cond *, I_mutex);
 void            I_wake_one_cond   (I_cond *);
 void            I_wake_all_cond   (I_cond *);
 
+INT32           I_atomic_load     (I_Atomicptr_t atomic);
+INT32           I_atomic_exchange (I_Atomicptr_t atomic, INT32 val);
+
 #endif/*I_THREADS_H*/
 #endif/*HAVE_THREADS*/

src/m_perfstats.c

@@ -100,7 +100,30 @@ enum {
 	PERF_COUNT,
 };
 
-static void M_DrawPerfString(perfstatcol_t *col, int type)
+static void M_DrawPerfString(const char *string, INT32 lores_x, INT32 hires_x, INT32 flags)
+{
+	const boolean hires = M_HighResolution();
+
+	INT32 draw_flags = V_MONOSPACE | flags;
+
+	if (hires)
+		draw_flags |= V_ALLOWLOWERCASE;
+
+	if (hires)
+	{
+		V_DrawFillConsoleMap(hires_x, draw_row, V_SmallStringWidth(string, draw_flags), 5, 0);
+		V_DrawSmallString(hires_x, draw_row, draw_flags, string);
+		draw_row += 5;
+	}
+	else
+	{
+		V_DrawFillConsoleMap(lores_x, draw_row, V_ThinStringWidth(string, draw_flags), 8, 0);
+		V_DrawThinString(lores_x, draw_row, draw_flags, string);
+		draw_row += 8;
+	}
+}
+
+static void M_DrawPerfMeasurement(perfstatcol_t *col, int type)
 {
 	const boolean hires = M_HighResolution();
 
@@ -122,15 +145,19 @@ static void M_DrawPerfString(perfstatcol_t *col, int type)
 
 		if (hires)
 		{
-			V_DrawSmallString(col->hires_x, draw_row, draw_flags,
+			const char *string = va("%s %d", row->hires_label, value);
-					va("%s %d", row->hires_label, value));
+
+			V_DrawFillConsoleMap(col->hires_x, draw_row, V_SmallStringWidth(string, draw_flags), 5, 0);
+			V_DrawSmallString(col->hires_x, draw_row, draw_flags, string);
 
 			draw_row += 5;
 		}
 		else
 		{
-			V_DrawThinString(col->lores_x, draw_row, draw_flags,
+			const char *string = va("%s %d", row->lores_label, value);
-					va("%s %d", row->lores_label, value));
+
+			V_DrawFillConsoleMap(col->lores_x, draw_row, V_SmallStringWidth(string, draw_flags), 8, 0);
+			V_DrawThinString(col->lores_x, draw_row, draw_flags, string);
 
 			draw_row += 8;
 		}
@@ -139,12 +166,12 @@ static void M_DrawPerfString(perfstatcol_t *col, int type)
 
 static void M_DrawPerfTiming(perfstatcol_t *col)
 {
-	M_DrawPerfString(col, PERF_TIME);
+	M_DrawPerfMeasurement(col, PERF_TIME);
 }
 
 static void M_DrawPerfCount(perfstatcol_t *col)
 {
-	M_DrawPerfString(col, PERF_COUNT);
+	M_DrawPerfMeasurement(col, PERF_COUNT);
 }
 
 static void M_DrawRenderStats(void)
@@ -161,7 +188,7 @@ static void M_DrawRenderStats(void)
 	};
 
 	perfstatrow_t rendercalltime_row[] = {
-		{"drwtime", "3d rendering:  ", &ps_rendercalltime},
+		{"drwtime", "3D rendering:  ", &ps_rendercalltime},
 		{0}
 	};
 
@@ -176,16 +203,6 @@ static void M_DrawRenderStats(void)
 		{0}
 	};
 
-	perfstatrow_t softwaretime_row[] = {
-		{"bsptime", "RenderBSPNode: ", &ps_bsptime},
-		{"sprclip", "R_ClipSprites: ", &ps_sw_spritecliptime},
-		{"portals", "Portals+Skybox:", &ps_sw_portaltime},
-		{"planes ", "R_DrawPlanes:  ", &ps_sw_planetime},
-		{"masked ", "R_DrawMasked:  ", &ps_sw_maskedtime},
-		{"other  ", "Other:         ", &extrarendertime},
-		{0}
-	};
-
 	perfstatrow_t uiswaptime_row[] = {
 		{"ui     ", "UI render:     ", &ps_uitime},
 		{"finupdt", "I_FinishUpdate:", &ps_swaptime},
@@ -197,14 +214,6 @@ static void M_DrawRenderStats(void)
 		{0}
 	};
 
-	perfstatrow_t rendercalls_row[] = {
-		{"bspcall", "BSP calls:   ", &ps_numbspcalls},
-		{"sprites", "Sprites:     ", &ps_numsprites},
-		{"drwnode", "Drawnodes:   ", &ps_numdrawnodes},
-		{"plyobjs", "Polyobjects: ", &ps_numpolyobjects},
-		{0}
-	};
-
 	perfstatrow_t batchtime_row[] = {
 		{"batsort", "Batch sort:  ", &ps_hw_batchsorttime},
 		{"batdraw", "Batch render:", &ps_hw_batchdrawtime},
@@ -230,13 +239,10 @@ static void M_DrawRenderStats(void)
 	perfstatcol_t rendercalltime_col =  {20,  20, V_YELLOWMAP, rendercalltime_row};
 
 	perfstatcol_t     opengltime_col =  {24,  24, V_YELLOWMAP,     opengltime_row};
-	perfstatcol_t   softwaretime_col =  {24,  24, V_YELLOWMAP,   softwaretime_row};
 
 	perfstatcol_t     uiswaptime_col =  {20,  20, V_YELLOWMAP,     uiswaptime_row};
 	perfstatcol_t        tictime_col =  {20,  20, V_GRAYMAP,          tictime_row};
 
-	perfstatcol_t    rendercalls_col =  {90, 115, V_BLUEMAP,      rendercalls_row};
-
 	perfstatcol_t      batchtime_col =  {90, 115, V_REDMAP,         batchtime_row};
 
 	perfstatcol_t     batchcount_col = {155, 200, V_PURPLEMAP,     batchcount_row};
@@ -256,12 +262,13 @@ static void M_DrawRenderStats(void)
 		M_DrawPerfTiming(&rendercalltime_col);
 
 		// Remember to update this calculation when adding more 3d rendering stats!
-		extrarendertime = ps_rendercalltime - ps_bsptime;
+		extrarendertime = ps_rendercalltime;
 
 #ifdef HWRENDER
 		if (rendermode == render_opengl)
 		{
-			extrarendertime -=
+			extrarendertime =
+				ps_bsptime[0] +
 				ps_hw_skyboxtime +
 				ps_hw_nodesorttime +
 				ps_hw_nodedrawtime +
@@ -280,14 +287,33 @@ static void M_DrawRenderStats(void)
 		else
 #endif
 		{
-			extrarendertime -=
+			INT32 i = 0;
-				ps_sw_spritecliptime +
-				ps_sw_portaltime +
-				ps_sw_planetime +
-				ps_sw_maskedtime;
 
+			perfstatrow_t other_row[] = {
+				{"postprc", "Postprocessing:", &ps_postprocesstime},
+				{0}
+			};
+
+			perfstatcol_t other_col = {24, 24, V_YELLOWMAP, other_row};
+
+			for (i = 0; i < numusablerendercontexts; i++) {
+				perfstatrow_t softwaretime_row[] = {
+					{"bsptime", "RenderBSPNode: ", &ps_bsptime[i]},
+					{"sprclip", "R_ClipSprites: ", &ps_sw_spritecliptime[i]},
+					{"portals", "Portals+Skybox:", &ps_sw_portaltime[i]},
+					{"planes ", "R_DrawPlanes:  ", &ps_sw_planetime[i]},
+					{"masked ", "R_DrawMasked:  ", &ps_sw_maskedtime[i]},
+					{0}
+				};
+
+				perfstatcol_t softwaretime_col = {28, 28, V_YELLOWMAP, softwaretime_row};
+
+				M_DrawPerfString(va("Thread %d:", i+1), 24, 24, V_REDMAP);
 				M_DrawPerfTiming(&softwaretime_col);
 			}
+
+			M_DrawPerfTiming(&other_col);
+		}
 	}
 
 	M_DrawPerfTiming(&uiswaptime_col);
@@ -298,7 +324,40 @@ static void M_DrawRenderStats(void)
 	if (rendering)
 	{
 		draw_row = 10;	
+
+#ifdef HWRENDER
+		if (rendermode == render_opengl) {
+			perfstatrow_t rendercalls_row[] = {
+				{"bspcall", "BSP calls:   ", &ps_numbspcalls},
+				{"sprites", "Sprites:     ", &ps_numsprites},
+				{"drwnode", "Drawnodes:   ", &ps_numdrawnodes},
+				{"plyobjs", "Polyobjects: ", &ps_numpolyobjects},
+				{0}
+			};
+
+			perfstatcol_t rendercalls_col = {90, 115, V_BLUEMAP, rendercalls_row};
+
 			M_DrawPerfCount(&rendercalls_col);
+		} else
+#endif
+		{
+			INT32 i;
+
+			for (i = 0; i < numusablerendercontexts; i++) {
+				perfstatrow_t rendercalls_row[] = {
+					{"bspcall", "BSP calls:   ", &ps_numbspcalls[i]},
+					{"sprites", "Sprites:     ", &ps_numsprites[i]},
+					{"drwnode", "Drawnodes:   ", &ps_numdrawnodes[i]},
+					{"plyobjs", "Polyobjects: ", &ps_numpolyobjects[i]},
+					{0}
+				};
+
+				perfstatcol_t rendercalls_col = {94, 120, V_BLUEMAP, rendercalls_row};
+
+				M_DrawPerfString(va("Thread %d:", i+1), 90, 115, V_REDMAP);
+				M_DrawPerfCount(&rendercalls_col);
+			}
+		}
 
 #ifdef HWRENDER
 		if (rendermode == render_opengl && cv_glbatching.value)

src/r_bsp.c

@@ -23,50 +23,37 @@
 #include "z_zone.h" // Check R_Prep3DFloors
 #include "taglist.h"
 
-seg_t *curline;
+#ifdef HAVE_THREADS
-side_t *sidedef;
+static I_mutex R_BSP_mutex;
-line_t *linedef;
+#  define Lock_state()    I_lock_mutex(&R_BSP_mutex)
-sector_t *frontsector;
+#  define Unlock_state() I_unlock_mutex(R_BSP_mutex)
-sector_t *backsector;
+#else
-
+#  define Lock_state()
-// very ugly realloc() of drawsegs at run-time, I upped it to 512
+#  define Unlock_state()
-// instead of 256.. and someone managed to send me a level with
+#endif
-// 896 drawsegs! So too bad here's a limit removal a-la-Boom
-drawseg_t *curdrawsegs = NULL; /**< This is used to handle multiple lists for masked drawsegs. */
-drawseg_t *drawsegs = NULL;
-drawseg_t *ds_p = NULL;
-
-// indicates doors closed wrt automap bugfix:
-INT32 doorclosed;
 
 //
 // R_ClearDrawSegs
 //
-void R_ClearDrawSegs(void)
+void R_ClearDrawSegs(bspcontext_t *context)
 {
-	ds_p = drawsegs;
+	context->ds_p = context->drawsegs;
 }
 
-// Fix from boom.
-#define MAXSEGS (MAXVIDWIDTH/2+1)
-
-// newend is one past the last valid seg
-static cliprange_t *newend;
-static cliprange_t solidsegs[MAXSEGS];
-
 //
 // R_ClipSolidWallSegment
 // Does handle solid walls,
 //  e.g. single sided LineDefs (middle texture)
 //  that entirely block the view.
 //
-static void R_ClipSolidWallSegment(INT32 first, INT32 last)
+static void R_ClipSolidWallSegment(rendercontext_t *context, wallcontext_t *wallcontext, INT32 first, INT32 last)
 {
+	bspcontext_t *bspcontext = &context->bspcontext;
 	cliprange_t *next;
 	cliprange_t *start;
 
 	// Find the first range that touches the range (adjacent pixels are touching).
-	start = solidsegs;
+	start = bspcontext->solidsegs;
 	while (start->last < first - 1)
 		start++;
 
@@ -75,11 +62,11 @@ static void R_ClipSolidWallSegment(INT32 first, INT32 last)
 		if (last < start->first - 1)
 		{
 			// Post is entirely visible (above start), so insert a new clippost.
-			R_StoreWallRange(first, last);
+			R_StoreWallRange(context, wallcontext, first, last);
-			next = newend;
+			next = bspcontext->newend;
-			newend++;
+			bspcontext->newend++;
 			// NO MORE CRASHING!
-			if (newend - solidsegs > MAXSEGS)
+			if (bspcontext->newend - bspcontext->solidsegs > MAXSEGS)
 				I_Error("R_ClipSolidWallSegment: Solid Segs overflow!\n");
 
 			while (next != start)
@@ -93,7 +80,7 @@ static void R_ClipSolidWallSegment(INT32 first, INT32 last)
 		}
 
 		// There is a fragment above *start.
-		R_StoreWallRange(first, start->first - 1);
+		R_StoreWallRange(context, wallcontext, first, start->first - 1);
 		// Now adjust the clip size.
 		start->first = first;
 	}
@@ -106,7 +93,7 @@ static void R_ClipSolidWallSegment(INT32 first, INT32 last)
 	while (last >= (next+1)->first - 1)
 	{
 		// There is a fragment between two posts.
-		R_StoreWallRange(next->last + 1, (next+1)->first - 1);
+		R_StoreWallRange(context, wallcontext, next->last + 1, (next+1)->first - 1);
 		next++;
 
 		if (last <= next->last)
@@ -119,7 +106,7 @@ static void R_ClipSolidWallSegment(INT32 first, INT32 last)
 	}
 
 	// There is a fragment after *next.
-	R_StoreWallRange(next->last + 1, last);
+	R_StoreWallRange(context, wallcontext, next->last + 1, last);
 	// Adjust the clip size.
 	start->last = last;
 
@@ -128,13 +115,13 @@ crunch:
 	if (next == start)
 		return; // Post just extended past the bottom of one post.
 
-	while (next++ != newend)
+	while (next++ != bspcontext->newend)
 		*++start = *next; // Remove a post.
 
-	newend = start + 1;
+	bspcontext->newend = start + 1;
 
 	// NO MORE CRASHING!
-	if (newend - solidsegs > MAXSEGS)
+	if (bspcontext->newend - bspcontext->solidsegs > MAXSEGS)
 		I_Error("R_ClipSolidWallSegment: Solid Segs overflow!\n");
 }
 
@@ -143,13 +130,13 @@ crunch:
 // Clips the given range of columns, but does not include it in the clip list.
 // Does handle windows, e.g. LineDefs with upper and lower texture.
 //
-static inline void R_ClipPassWallSegment(INT32 first, INT32 last)
+static inline void R_ClipPassWallSegment(rendercontext_t *context, wallcontext_t *wallcontext, INT32 first, INT32 last)
 {
 	cliprange_t *start;
 
 	// Find the first range that touches the range
 	//  (adjacent pixels are touching).
-	start = solidsegs;
+	start = context->bspcontext.solidsegs;
 	while (start->last < first - 1)
 		start++;
 
@@ -158,12 +145,12 @@ static inline void R_ClipPassWallSegment(INT32 first, INT32 last)
 		if (last < start->first - 1)
 		{
 			// Post is entirely visible (above start).
-			R_StoreWallRange(first, last);
+			R_StoreWallRange(context, wallcontext, first, last);
 			return;
 		}
 
 		// There is a fragment above *start.
-		R_StoreWallRange(first, start->first - 1);
+		R_StoreWallRange(context, wallcontext, first, start->first - 1);
 	}
 
 	// Bottom contained in start?
@@ -173,7 +160,7 @@ static inline void R_ClipPassWallSegment(INT32 first, INT32 last)
 	while (last >= (start+1)->first - 1)
 	{
 		// There is a fragment between two posts.
-		R_StoreWallRange(start->last + 1, (start+1)->first - 1);
+		R_StoreWallRange(context, wallcontext, start->last + 1, (start+1)->first - 1);
 		start++;
 
 		if (last <= start->last)
@@ -181,27 +168,19 @@ static inline void R_ClipPassWallSegment(INT32 first, INT32 last)
 	}
 
 	// There is a fragment after *next.
-	R_StoreWallRange(start->last + 1, last);
+	R_StoreWallRange(context, wallcontext, start->last + 1, last);
 }
 
 //
 // R_ClearClipSegs
 //
-void R_ClearClipSegs(void)
+void R_ClearClipSegs(bspcontext_t *context, INT32 start, INT32 end)
 {
-	solidsegs[0].first = -0x7fffffff;
+	context->solidsegs[0].first = -0x7fffffff;
-	solidsegs[0].last = -1;
+	context->solidsegs[0].last = start-1;
-	solidsegs[1].first = viewwidth;
+	context->solidsegs[1].first = end;
-	solidsegs[1].last = 0x7fffffff;
+	context->solidsegs[1].last = 0x7fffffff;
-	newend = solidsegs + 2;
+	context->newend = context->solidsegs + 2;
-}
-void R_PortalClearClipSegs(INT32 start, INT32 end)
-{
-	solidsegs[0].first = -0x7fffffff;
-	solidsegs[0].last = start-1;
-	solidsegs[1].first = end;
-	solidsegs[1].last = 0x7fffffff;
-	newend = solidsegs + 2;
 }
 
 
@@ -212,17 +191,17 @@ void R_PortalClearClipSegs(INT32 start, INT32 end)
 //
 // It assumes that Doom has already ruled out a door being closed because
 // of front-back closure (e.g. front floor is taller than back ceiling).
-static INT32 R_DoorClosed(void)
+static INT32 R_DoorClosed(bspcontext_t *bspcontext)
 {
 	return
 
 	// if door is closed because back is shut:
-	backsector->ceilingheight <= backsector->floorheight
+	bspcontext->backsector->ceilingheight <= bspcontext->backsector->floorheight
 
 	// preserve a kind of transparent door/lift special effect:
-	&& (backsector->ceilingheight >= frontsector->ceilingheight || curline->sidedef->toptexture)
+	&& (bspcontext->backsector->ceilingheight >= bspcontext->frontsector->ceilingheight || bspcontext->curline->sidedef->toptexture)
 
-	&& (backsector->floorheight <= frontsector->floorheight || curline->sidedef->bottomtexture);
+	&& (bspcontext->backsector->floorheight <= bspcontext->frontsector->floorheight || bspcontext->curline->sidedef->bottomtexture);
 }
 
 //
@@ -233,8 +212,8 @@ static INT32 R_DoorClosed(void)
 //
 // Similar for ceiling, only reflected.
 //
-sector_t *R_FakeFlat(sector_t *sec, sector_t *tempsec, INT32 *floorlightlevel,
+sector_t *R_FakeFlat(viewcontext_t *viewcontext, sector_t *sec, sector_t *tempsec,
-	INT32 *ceilinglightlevel, boolean back)
+	INT32 *floorlightlevel, INT32 *ceilinglightlevel, boolean back)
 {
 	if (floorlightlevel)
 		*floorlightlevel = sec->floorlightsec == -1 ?
@@ -250,19 +229,19 @@ sector_t *R_FakeFlat(sector_t *sec, sector_t *tempsec, INT32 *floorlightlevel,
 	if (!sec->extra_colormap && sec->heightsec != -1)
 	{
 		const sector_t *s = &sectors[sec->heightsec];
-		mobj_t *viewmobj = viewplayer->mo;
+		mobj_t *viewmobj = viewcontext->player->mo;
 		INT32 heightsec;
 		boolean underwater;
 
-		if (splitscreen && viewplayer == &players[secondarydisplayplayer] && camera2.chase)
+		if (splitscreen && viewcontext->player == &players[secondarydisplayplayer] && camera2.chase)
 			heightsec = R_PointInSubsector(camera2.x, camera2.y)->sector->heightsec;
-		else if (camera.chase && viewplayer == &players[displayplayer])
+		else if (camera.chase && viewcontext->player == &players[displayplayer])
 			heightsec = R_PointInSubsector(camera.x, camera.y)->sector->heightsec;
 		else if (viewmobj)
 			heightsec = R_PointInSubsector(viewmobj->x, viewmobj->y)->sector->heightsec;
 		else
 			return sec;
-		underwater = heightsec != -1 && viewz <= sectors[heightsec].floorheight;
+		underwater = heightsec != -1 && viewcontext->z <= sectors[heightsec].floorheight;
 
 		// Replace sector being drawn, with a copy to be hacked
 		*tempsec = *sec;
@@ -272,7 +251,7 @@ sector_t *R_FakeFlat(sector_t *sec, sector_t *tempsec, INT32 *floorlightlevel,
 		tempsec->ceilingheight = s->ceilingheight;
 
 		if ((underwater && (tempsec->  floorheight = sec->floorheight,
-			tempsec->ceilingheight = s->floorheight - 1, !back)) || viewz <= s->floorheight)
+			tempsec->ceilingheight = s->floorheight - 1, !back)) || viewcontext->z <= s->floorheight)
 		{ // head-below-floor hack
 			tempsec->floorpic = s->floorpic;
 			tempsec->floor_xoffs = s->floor_xoffs;
@@ -308,7 +287,7 @@ sector_t *R_FakeFlat(sector_t *sec, sector_t *tempsec, INT32 *floorlightlevel,
 				*ceilinglightlevel = s->ceilinglightsec == -1 ? s->lightlevel
 					: sectors[s->ceilinglightsec].lightlevel;
 		}
-		else if (heightsec != -1 && viewz >= sectors[heightsec].ceilingheight
+		else if (heightsec != -1 && viewcontext->z >= sectors[heightsec].ceilingheight
 			&& sec->ceilingheight > s->ceilingheight)
 		{ // Above-ceiling hack
 			tempsec->ceilingheight = s->ceilingheight;
@@ -382,22 +361,25 @@ boolean R_IsEmptyLine(seg_t *line, sector_t *front, sector_t *back)
 // R_AddLine
 // Clips the given segment and adds any visible pieces to the line list.
 //
-static void R_AddLine(seg_t *line)
+static void R_AddLine(rendercontext_t *rendercontext, seg_t *line)
 {
 	INT32 x1, x2;
 	angle_t angle1, angle2, span, tspan;
-	static sector_t tempsec;
 	boolean bothceilingssky = false, bothfloorssky = false;
 
-	portalline = false;
+	bspcontext_t *bspcontext = &rendercontext->bspcontext;
+	viewcontext_t *viewcontext = &rendercontext->viewcontext;
+	wallcontext_t wallcontext;
+
+	bspcontext->portalline = false;
 
 	if (line->polyseg && !(line->polyseg->flags & POF_RENDERSIDES))
 		return;
 
 	// big room fix
-	angle1 = R_PointToAngleEx(viewx, viewy, line->v1->x, line->v1->y);
+	angle1 = R_PointToAngleEx(viewcontext->x, viewcontext->y, line->v1->x, line->v1->y);
-	angle2 = R_PointToAngleEx(viewx, viewy, line->v2->x, line->v2->y);
+	angle2 = R_PointToAngleEx(viewcontext->x, viewcontext->y, line->v2->x, line->v2->y);
-	curline = line;
+	bspcontext->curline = line;
 
 	// Clip to view edges.
 	span = angle1 - angle2;
@@ -407,9 +389,9 @@ static void R_AddLine(seg_t *line)
 		return;
 
 	// Global angle needed by segcalc.
-	rw_angle1 = angle1;
+	wallcontext.angle1 = angle1;
-	angle1 -= viewangle;
+	angle1 -= viewcontext->angle;
-	angle2 -= viewangle;
+	angle2 -= viewcontext->angle;
 
 	tspan = angle1 + clipangle;
 	if (tspan > doubleclipangle)
@@ -440,18 +422,21 @@ static void R_AddLine(seg_t *line)
 	x1 = viewangletox[angle1];
 	x2 = viewangletox[angle2];
 
+	x1 = max(rendercontext->begincolumn, x1);
+	x2 = min(rendercontext->endcolumn, x2);
+
 	// Does not cross a pixel?
 	if (x1 >= x2)       // killough 1/31/98 -- change == to >= for robustness
 		return;
 
-	backsector = line->backsector;
+	bspcontext->backsector = line->backsector;
 
 	// Portal line
 	if (line->linedef->special == 40 && line->side == 0)
 	{
 		// Render portal if recursiveness limit hasn't been reached.
 		// Otherwise, render the wall normally.
-		if (portalrender < cv_maxportals.value)
+		if (bspcontext->portalrender < cv_maxportals.value)
 		{
 			size_t p;
 			mtag_t tag = Tag_FGet(&line->linedef->tags);
@@ -464,38 +449,40 @@ static void R_AddLine(seg_t *line)
 				if ((tag != Tag_FGet(&lines[li2].tags)) || (lines[li1].special != lines[li2].special) || (li1 == li2))
 					continue;
 
-				Portal_Add2Lines(li1, li2, x1, x2);
+				Portal_Add2Lines(rendercontext, li1, li2, x1, x2);
 				goto clipsolid;
 			}
 		}
 	}
 
 	// Single sided line?
-	if (!backsector)
+	if (!bspcontext->backsector)
 		goto clipsolid;
 
-	backsector = R_FakeFlat(backsector, &tempsec, NULL, NULL, true);
+	bspcontext->backsector = R_FakeFlat(viewcontext, bspcontext->backsector, &bspcontext->btempsec, NULL, NULL, true);
 
-	doorclosed = 0;
+	// indicates doors closed wrt automap bugfix:
+	bspcontext->doorclosed = 0;
 
-	if (backsector->ceilingpic == skyflatnum && frontsector->ceilingpic == skyflatnum)
+	if (bspcontext->backsector->ceilingpic == skyflatnum && bspcontext->frontsector->ceilingpic == skyflatnum)
 		bothceilingssky = true;
-	if (backsector->floorpic == skyflatnum && frontsector->floorpic == skyflatnum)
+	if (bspcontext->backsector->floorpic == skyflatnum && bspcontext->frontsector->floorpic == skyflatnum)
 		bothfloorssky = true;
 
 	if (bothceilingssky && bothfloorssky) // everything's sky? let's save us a bit of time then
 	{
 		if (!line->polyseg &&
 			!line->sidedef->midtexture
-			&& ((!frontsector->ffloors && !backsector->ffloors)
+			&& ((!bspcontext->frontsector->ffloors && !bspcontext->backsector->ffloors)
-				|| Tag_Compare(&frontsector->tags, &backsector->tags)))
+				|| Tag_Compare(&bspcontext->frontsector->tags, &bspcontext->backsector->tags)))
 			return; // line is empty, don't even bother
 
 		goto clippass; // treat like wide open window instead
 	}
 
 	// Closed door.
-	if (frontsector->f_slope || frontsector->c_slope || backsector->f_slope || backsector->c_slope)
+	if (bspcontext->frontsector->f_slope || bspcontext->frontsector->c_slope
+	|| bspcontext->backsector->f_slope || bspcontext->backsector->c_slope)
 	{
 		fixed_t frontf1,frontf2, frontc1, frontc2; // front floor/ceiling ends
 		fixed_t backf1, backf2, backc1, backc2; // back floor ceiling ends
@@ -503,10 +490,10 @@ static void R_AddLine(seg_t *line)
 		end1 = P_GetZAt(slope, line->v1->x, line->v1->y, normalheight); \
 		end2 = P_GetZAt(slope, line->v2->x, line->v2->y, normalheight);
 
-		SLOPEPARAMS(frontsector->f_slope, frontf1, frontf2, frontsector->  floorheight)
+		SLOPEPARAMS(bspcontext->frontsector->f_slope, frontf1, frontf2, bspcontext->frontsector->  floorheight)
-		SLOPEPARAMS(frontsector->c_slope, frontc1, frontc2, frontsector->ceilingheight)
+		SLOPEPARAMS(bspcontext->frontsector->c_slope, frontc1, frontc2, bspcontext->frontsector->ceilingheight)
-		SLOPEPARAMS( backsector->f_slope,  backf1,  backf2,  backsector->  floorheight)
+		SLOPEPARAMS( bspcontext->backsector->f_slope,  backf1,  backf2,  bspcontext->backsector->  floorheight)
-		SLOPEPARAMS( backsector->c_slope,  backc1,  backc2,  backsector->ceilingheight)
+		SLOPEPARAMS( bspcontext->backsector->c_slope,  backc1,  backc2,  bspcontext->backsector->ceilingheight)
 #undef SLOPEPARAMS
 		// if both ceilings are skies, consider it always "open"
 		// same for floors
@@ -519,11 +506,11 @@ static void R_AddLine(seg_t *line)
 			}
 
 			// Check for automap fix. Store in doorclosed for r_segs.c
-			doorclosed = (backc1 <= backf1 && backc2 <= backf2
+			bspcontext->doorclosed = (backc1 <= backf1 && backc2 <= backf2
-			&& ((backc1 >= frontc1 && backc2 >= frontc2) || curline->sidedef->toptexture)
+			&& ((backc1 >= frontc1 && backc2 >= frontc2) || bspcontext->curline->sidedef->toptexture)
-			&& ((backf1 <= frontf1 && backf2 >= frontf2) || curline->sidedef->bottomtexture));
+			&& ((backf1 <= frontf1 && backf2 >= frontf2) || bspcontext->curline->sidedef->bottomtexture));
 
-			if (doorclosed)
+			if (bspcontext->doorclosed)
 				goto clipsolid;
 		}
 
@@ -541,40 +528,39 @@ static void R_AddLine(seg_t *line)
 		// same for floors
 		if (!bothceilingssky && !bothfloorssky)
 		{
-			if (backsector->ceilingheight <= frontsector->floorheight
+			if (bspcontext->backsector->ceilingheight <= bspcontext->frontsector->floorheight
-				|| backsector->floorheight >= frontsector->ceilingheight)
+				|| bspcontext->backsector->floorheight >= bspcontext->frontsector->ceilingheight)
 			{
 				goto clipsolid;
 			}
 
 			// Check for automap fix. Store in doorclosed for r_segs.c
-			doorclosed = R_DoorClosed();
+			bspcontext->doorclosed = R_DoorClosed(bspcontext);
-			if (doorclosed)
+			if (bspcontext->doorclosed)
 				goto clipsolid;
 		}
 
 		// Window.
 		if (!bothceilingssky) // ceilings are always the "same" when sky
-			if (backsector->ceilingheight != frontsector->ceilingheight)
+			if (bspcontext->backsector->ceilingheight != bspcontext->frontsector->ceilingheight)
 				goto clippass;
 		if (!bothfloorssky)	// floors are always the "same" when sky
-			if (backsector->floorheight != frontsector->floorheight)
+			if (bspcontext->backsector->floorheight != bspcontext->frontsector->floorheight)
 				goto clippass;
 	}
 
 	// Reject empty lines used for triggers and special events.
 	// Identical floor and ceiling on both sides, identical light levels on both sides,
 	// and no middle texture.
-
+	if (R_IsEmptyLine(line, bspcontext->frontsector, bspcontext->backsector))
-	if (R_IsEmptyLine(line, frontsector, backsector))
 		return;
 
 clippass:
-	R_ClipPassWallSegment(x1, x2 - 1);
+	R_ClipPassWallSegment(rendercontext, &wallcontext, x1, x2 - 1);
 	return;
 
 clipsolid:
-	R_ClipSolidWallSegment(x1, x2 - 1);
+	R_ClipSolidWallSegment(rendercontext, &wallcontext, x1, x2 - 1);
 }
 
 //
@@ -602,22 +588,24 @@ INT32 checkcoord[12][4] =
 	{2, 1, 3, 0}
 };
 
-static boolean R_CheckBBox(const fixed_t *bspcoord)
+static boolean R_CheckBBox(rendercontext_t *context, const fixed_t *bspcoord)
 {
 	angle_t angle1, angle2;
 	INT32 sx1, sx2, boxpos;
 	const INT32* check;
 	cliprange_t *start;
+	viewcontext_t *viewcontext = &context->viewcontext;
 
 	// Find the corners of the box that define the edges from current viewpoint.
-	if ((boxpos = (viewx <= bspcoord[BOXLEFT] ? 0 : viewx < bspcoord[BOXRIGHT] ? 1 : 2) + (viewy >= bspcoord[BOXTOP] ? 0 : viewy > bspcoord[BOXBOTTOM] ? 4 : 8)) == 5)
+	if ((boxpos = (viewcontext->x <= bspcoord[BOXLEFT] ? 0 : viewcontext->x < bspcoord[BOXRIGHT] ? 1 : 2)
+	+ (viewcontext->y >= bspcoord[BOXTOP] ? 0 : viewcontext->y > bspcoord[BOXBOTTOM] ? 4 : 8)) == 5)
 		return true;
 
 	check = checkcoord[boxpos];
 
 	// big room fix
-	angle1 = R_PointToAngleEx(viewx, viewy, bspcoord[check[0]], bspcoord[check[1]]) - viewangle;
+	angle1 = R_PointToAngleEx(viewcontext->x, viewcontext->y, bspcoord[check[0]], bspcoord[check[1]]) - viewcontext->angle;
-	angle2 = R_PointToAngleEx(viewx, viewy, bspcoord[check[2]], bspcoord[check[3]]) - viewangle;
+	angle2 = R_PointToAngleEx(viewcontext->x, viewcontext->y, bspcoord[check[2]], bspcoord[check[3]]) - viewcontext->angle;
 
 	if ((signed)angle1 < (signed)angle2)
 	{
@@ -641,7 +629,7 @@ static boolean R_CheckBBox(const fixed_t *bspcoord)
 	// Does not cross a pixel.
 	if (sx1 >= sx2) return false;
 
-	start = solidsegs;
+	start = context->bspcontext.solidsegs;
 	while (start->last < sx2)
 		start++;
 
@@ -789,7 +777,7 @@ static int R_PolysegCompare(const void *p1, const void *p2)
 // haleyjd 02/19/06
 // Adds all segs in all polyobjects in the given subsector.
 //
-static void R_AddPolyObjects(subsector_t *sub)
+static void R_AddPolyObjects(rendercontext_t *context, subsector_t *sub)
 {
 	polyobj_t *po = sub->polyList;
 	size_t i, j;
@@ -804,7 +792,7 @@ static void R_AddPolyObjects(subsector_t *sub)
 	}
 
 	// for render stats
-	ps_numpolyobjects += numpolys;
+	ps_numpolyobjects[context->num] += numpolys;
 
 	// sort polyobjects
 	R_SortPolyObjects(sub);
@@ -814,7 +802,7 @@ static void R_AddPolyObjects(subsector_t *sub)
 	{
 		qsort(po_ptrs[i]->segs, po_ptrs[i]->segCount, sizeof(seg_t *), R_PolysegCompare);
 		for (j = 0; j < po_ptrs[i]->segCount; ++j)
-			R_AddLine(po_ptrs[i]->segs[j]);
+			R_AddLine(context, po_ptrs[i]->segs[j]);
 	}
 }
 
@@ -825,17 +813,19 @@ static void R_AddPolyObjects(subsector_t *sub)
 // Draw one or more line segments.
 //
 
-drawseg_t *firstseg;
+static void R_Subsector(rendercontext_t *context, size_t num)
-
-static void R_Subsector(size_t num)
 {
 	INT32 count, floorlightlevel, ceilinglightlevel, light;
 	seg_t *line;
 	subsector_t *sub;
-	static sector_t tempsec; // Deep water hack
 	extracolormap_t *floorcolormap;
 	extracolormap_t *ceilingcolormap;
 	fixed_t floorcenterz, ceilingcenterz;
+	sector_t *frontsector;
+
+	bspcontext_t *bspcontext = &context->bspcontext;
+	planecontext_t *planecontext = &context->planecontext;
+	viewcontext_t *viewcontext = &context->viewcontext;
 
 #ifdef RANGECHECK
 	if (num >= numsubsectors)
@@ -847,12 +837,11 @@ static void R_Subsector(size_t num)
 		return;
 
 	sub = &subsectors[num];
-	frontsector = sub->sector;
 	count = sub->numlines;
 	line = &segs[sub->firstline];
 
 	// Deep water/fake ceiling effect.
-	frontsector = R_FakeFlat(frontsector, &tempsec, &floorlightlevel, &ceilinglightlevel, false);
+	frontsector = R_FakeFlat(viewcontext, sub->sector, &bspcontext->ftempsec, &floorlightlevel, &ceilinglightlevel, false);
 
 	floorcolormap = ceilingcolormap = frontsector->extra_colormap;
 
@@ -866,9 +855,12 @@ static void R_Subsector(size_t num)
 		{
 			frontsector->numlights = sub->sector->numlights = 0;
 			R_Prep3DFloors(frontsector);
+
+			Lock_state();
 			sub->sector->lightlist = frontsector->lightlist;
 			sub->sector->numlights = frontsector->numlights;
 			sub->sector->moved = frontsector->moved = false;
+			Unlock_state();
 		}
 
 		light = R_GetPlaneLight(frontsector, floorcenterz, false);
@@ -883,106 +875,109 @@ static void R_Subsector(size_t num)
 
 	sub->sector->extra_colormap = frontsector->extra_colormap;
 
-	if (P_GetSectorFloorZAt(frontsector, viewx, viewy) < viewz
+	if (P_GetSectorFloorZAt(frontsector, viewcontext->x, viewcontext->y) < viewcontext->z
 		|| frontsector->floorpic == skyflatnum
 		|| (frontsector->heightsec != -1 && sectors[frontsector->heightsec].ceilingpic == skyflatnum))
 	{
-		floorplane = R_FindPlane(frontsector->floorheight, frontsector->floorpic, floorlightlevel,
+		planecontext->floorplane = R_FindPlane(planecontext, viewcontext, frontsector->floorheight, frontsector->floorpic, floorlightlevel,
 			frontsector->floor_xoffs, frontsector->floor_yoffs, frontsector->floorpic_angle, floorcolormap, NULL, NULL, frontsector->f_slope);
 	}
 	else
-		floorplane = NULL;
+		planecontext->floorplane = NULL;
 
-	if (P_GetSectorCeilingZAt(frontsector, viewx, viewy) > viewz
+	if (P_GetSectorCeilingZAt(frontsector, viewcontext->x, viewcontext->y) > viewcontext->z
 		|| frontsector->ceilingpic == skyflatnum
 		|| (frontsector->heightsec != -1 && sectors[frontsector->heightsec].floorpic == skyflatnum))
 	{
-		ceilingplane = R_FindPlane(frontsector->ceilingheight, frontsector->ceilingpic,
+		planecontext->ceilingplane = R_FindPlane(planecontext, viewcontext, frontsector->ceilingheight, frontsector->ceilingpic,
 			ceilinglightlevel, frontsector->ceiling_xoffs, frontsector->ceiling_yoffs, frontsector->ceilingpic_angle,
 			ceilingcolormap, NULL, NULL, frontsector->c_slope);
 	}
 	else
-		ceilingplane = NULL;
+		planecontext->ceilingplane = NULL;
+
+	planecontext->numffloors = 0;
+	planecontext->ffloor[0].slope = NULL;
+	planecontext->ffloor[0].plane = NULL;
+	planecontext->ffloor[0].polyobj = NULL;
 
-	numffloors = 0;
-	ffloor[numffloors].slope = NULL;
-	ffloor[numffloors].plane = NULL;
-	ffloor[numffloors].polyobj = NULL;
 	if (frontsector->ffloors)
 	{
 		ffloor_t *rover;
 		fixed_t heightcheck, planecenterz;
 
-		for (rover = frontsector->ffloors; rover && numffloors < MAXFFLOORS; rover = rover->next)
+		for (rover = frontsector->ffloors; rover && planecontext->numffloors < MAXFFLOORS; rover = rover->next)
 		{
 			if (!(rover->flags & FF_EXISTS) || !(rover->flags & FF_RENDERPLANES))
 				continue;
 
 			if (frontsector->cullheight)
 			{
-				if (R_DoCulling(frontsector->cullheight, viewsector->cullheight, viewz, *rover->bottomheight, *rover->topheight))
+				if (R_DoCulling(frontsector->cullheight, viewcontext->sector->cullheight, viewcontext->z, *rover->bottomheight, *rover->topheight))
 				{
 					rover->norender = leveltime;
 					continue;
 				}
 			}
 
-			ffloor[numffloors].plane = NULL;
+			planecontext->ffloor[planecontext->numffloors].plane = NULL;
-			ffloor[numffloors].polyobj = NULL;
+			planecontext->ffloor[planecontext->numffloors].polyobj = NULL;
 
-			heightcheck = P_GetFFloorBottomZAt(rover, viewx, viewy);
+			heightcheck = P_GetFFloorBottomZAt(rover, viewcontext->x, viewcontext->y);
 
 			planecenterz = P_GetFFloorBottomZAt(rover, frontsector->soundorg.x, frontsector->soundorg.y);
 			if (planecenterz <= ceilingcenterz
 				&& planecenterz >= floorcenterz
-				&& ((viewz < heightcheck && (rover->flags & FF_BOTHPLANES || !(rover->flags & FF_INVERTPLANES)))
+				&& ((viewcontext->z < heightcheck && (rover->flags & FF_BOTHPLANES || !(rover->flags & FF_INVERTPLANES)))
-				|| (viewz > heightcheck && (rover->flags & FF_BOTHPLANES || rover->flags & FF_INVERTPLANES))))
+				|| (viewcontext->z > heightcheck && (rover->flags & FF_BOTHPLANES || rover->flags & FF_INVERTPLANES))))
 			{
 				light = R_GetPlaneLight(frontsector, planecenterz,
-					viewz < heightcheck);
+					viewcontext->z < heightcheck);
 
-				ffloor[numffloors].plane = R_FindPlane(*rover->bottomheight, *rover->bottompic,
+				planecontext->ffloor[planecontext->numffloors].plane = R_FindPlane(planecontext, viewcontext,
+					*rover->bottomheight, *rover->bottompic,
 					*frontsector->lightlist[light].lightlevel, *rover->bottomxoffs,
 					*rover->bottomyoffs, *rover->bottomangle, *frontsector->lightlist[light].extra_colormap, rover, NULL, *rover->b_slope);
 
-				ffloor[numffloors].slope = *rover->b_slope;
+				planecontext->ffloor[planecontext->numffloors].slope = *rover->b_slope;
 
 				// Tell the renderer this sector has slopes in it.
-				if (ffloor[numffloors].slope)
+				if (planecontext->ffloor[planecontext->numffloors].slope)
 					frontsector->hasslope = true;
 
-				ffloor[numffloors].height = heightcheck;
+				planecontext->ffloor[planecontext->numffloors].height = heightcheck;
-				ffloor[numffloors].ffloor = rover;
+				planecontext->ffloor[planecontext->numffloors].ffloor = rover;
-				numffloors++;
+				planecontext->numffloors++;
 			}
-			if (numffloors >= MAXFFLOORS)
+			if (planecontext->numffloors >= MAXFFLOORS)
 				break;
-			ffloor[numffloors].plane = NULL;
+			planecontext->ffloor[planecontext->numffloors].plane = NULL;
-			ffloor[numffloors].polyobj = NULL;
+			planecontext->ffloor[planecontext->numffloors].polyobj = NULL;
 
-			heightcheck = P_GetFFloorTopZAt(rover, viewx, viewy);
+			heightcheck = P_GetFFloorTopZAt(rover, viewcontext->x, viewcontext->y);
 
 			planecenterz = P_GetFFloorTopZAt(rover, frontsector->soundorg.x, frontsector->soundorg.y);
 			if (planecenterz >= floorcenterz
 				&& planecenterz <= ceilingcenterz
-				&& ((viewz > heightcheck && (rover->flags & FF_BOTHPLANES || !(rover->flags & FF_INVERTPLANES)))
+				&& ((viewcontext->z > heightcheck && (rover->flags & FF_BOTHPLANES || !(rover->flags & FF_INVERTPLANES)))
-				|| (viewz < heightcheck && (rover->flags & FF_BOTHPLANES || rover->flags & FF_INVERTPLANES))))
+				|| (viewcontext->z < heightcheck && (rover->flags & FF_BOTHPLANES || rover->flags & FF_INVERTPLANES))))
 			{
-				light = R_GetPlaneLight(frontsector, planecenterz, viewz < heightcheck);
+				light = R_GetPlaneLight(frontsector, planecenterz, viewcontext->z < heightcheck);
 
-				ffloor[numffloors].plane = R_FindPlane(*rover->topheight, *rover->toppic,
+				planecontext->ffloor[planecontext->numffloors].plane = R_FindPlane(planecontext, viewcontext,
+					*rover->topheight, *rover->toppic,
 					*frontsector->lightlist[light].lightlevel, *rover->topxoffs, *rover->topyoffs, *rover->topangle,
 					*frontsector->lightlist[light].extra_colormap, rover, NULL, *rover->t_slope);
 
-				ffloor[numffloors].slope = *rover->t_slope;
+				planecontext->ffloor[planecontext->numffloors].slope = *rover->t_slope;
 
 				// Tell the renderer this sector has slopes in it.
-				if (ffloor[numffloors].slope)
+				if (planecontext->ffloor[planecontext->numffloors].slope)
 					frontsector->hasslope = true;
 
-				ffloor[numffloors].height = heightcheck;
+				planecontext->ffloor[planecontext->numffloors].height = heightcheck;
-				ffloor[numffloors].ffloor = rover;
+				planecontext->ffloor[planecontext->numffloors].ffloor = rover;
-				numffloors++;
+				planecontext->numffloors++;
 			}
 		}
 	}
@@ -995,7 +990,7 @@ static void R_Subsector(size_t num)
 
 		while (po)
 		{
-			if (numffloors >= MAXFFLOORS)
+			if (planecontext->numffloors >= MAXFFLOORS)
 				break;
 
 			if (!(po->flags & POF_RENDERPLANES)) // Don't draw planes
@@ -1005,54 +1000,58 @@ static void R_Subsector(size_t num)
 			}
 
 			polysec = po->lines[0]->backsector;
-			ffloor[numffloors].plane = NULL;
+			planecontext->ffloor[planecontext->numffloors].plane = NULL;
 
 			if (polysec->floorheight <= ceilingcenterz
 				&& polysec->floorheight >= floorcenterz
-				&& (viewz < polysec->floorheight))
+				&& (viewcontext->z < polysec->floorheight))
 			{
-				light = R_GetPlaneLight(frontsector, polysec->floorheight, viewz < polysec->floorheight);
+				light = R_GetPlaneLight(frontsector, polysec->floorheight, viewcontext->z < polysec->floorheight);
-				ffloor[numffloors].plane = R_FindPlane(polysec->floorheight, polysec->floorpic,
+				planecontext->ffloor[planecontext->numffloors].plane = R_FindPlane(planecontext, viewcontext,
+					polysec->floorheight, polysec->floorpic,
 					(light == -1 ? frontsector->lightlevel : *frontsector->lightlist[light].lightlevel), polysec->floor_xoffs, polysec->floor_yoffs,
 					polysec->floorpic_angle-po->angle,
 					(light == -1 ? frontsector->extra_colormap : *frontsector->lightlist[light].extra_colormap), NULL, po,
 					NULL); // will ffloors be slopable eventually?
 
-				ffloor[numffloors].height = polysec->floorheight;
+				planecontext->ffloor[planecontext->numffloors].height = polysec->floorheight;
-				ffloor[numffloors].polyobj = po;
+				planecontext->ffloor[planecontext->numffloors].polyobj = po;
-				ffloor[numffloors].slope = NULL;
+				planecontext->ffloor[planecontext->numffloors].slope = NULL;
-				//ffloor[numffloors].ffloor = rover;
+				//planecontext->ffloor[planecontext->numffloors].ffloor = rover;
-				po->visplane = ffloor[numffloors].plane;
+				po->visplane = planecontext->ffloor[planecontext->numffloors].plane;
-				numffloors++;
+				planecontext->numffloors++;
 			}
 
-			if (numffloors >= MAXFFLOORS)
+			if (planecontext->numffloors >= MAXFFLOORS)
 				break;
 
-			ffloor[numffloors].plane = NULL;
+			planecontext->ffloor[planecontext->numffloors].plane = NULL;
 
 			if (polysec->ceilingheight >= floorcenterz
 				&& polysec->ceilingheight <= ceilingcenterz
-				&& (viewz > polysec->ceilingheight))
+				&& (viewcontext->z > polysec->ceilingheight))
 			{
-				light = R_GetPlaneLight(frontsector, polysec->floorheight, viewz < polysec->floorheight);
+				light = R_GetPlaneLight(frontsector, polysec->floorheight, viewcontext->z < polysec->floorheight);
-				ffloor[numffloors].plane = R_FindPlane(polysec->ceilingheight, polysec->ceilingpic,
+				planecontext->ffloor[planecontext->numffloors].plane = R_FindPlane(planecontext, viewcontext,
+					polysec->ceilingheight, polysec->ceilingpic,
 					(light == -1 ? frontsector->lightlevel : *frontsector->lightlist[light].lightlevel), polysec->ceiling_xoffs, polysec->ceiling_yoffs, polysec->ceilingpic_angle-po->angle,
 					(light == -1 ? frontsector->extra_colormap : *frontsector->lightlist[light].extra_colormap), NULL, po,
 					NULL); // will ffloors be slopable eventually?
 
-				ffloor[numffloors].polyobj = po;
+				planecontext->ffloor[planecontext->numffloors].polyobj = po;
-				ffloor[numffloors].height = polysec->ceilingheight;
+				planecontext->ffloor[planecontext->numffloors].height = polysec->ceilingheight;
-				ffloor[numffloors].slope = NULL;
+				planecontext->ffloor[planecontext->numffloors].slope = NULL;
-				//ffloor[numffloors].ffloor = rover;
+				//planecontext->ffloor[planecontext->numffloors].ffloor = rover;
-				po->visplane = ffloor[numffloors].plane;
+				po->visplane = planecontext->ffloor[planecontext->numffloors].plane;
-				numffloors++;
+				planecontext->numffloors++;
 			}
 
 			po = (polyobj_t *)(po->link.next);
 		}
 	}
 
+	bspcontext->frontsector = frontsector;
+
 	// killough 9/18/98: Fix underwater slowdown, by passing real sector
 	// instead of fake one. Improve sprite lighting by basing sprite
 	// lightlevels on floor & ceiling lightlevels in the surrounding area.
@@ -1065,21 +1064,20 @@ static void R_Subsector(size_t num)
 	// Either you must pass the fake sector and handle validcount here, on the
 	// real sector, or you must account for the lighting in some other way,
 	// like passing it as an argument.
-	R_AddSprites(sub->sector, (floorlightlevel+ceilinglightlevel)/2);
+	R_AddSprites(context, sub->sector, (floorlightlevel+ceilinglightlevel)/2);
 
-	firstseg = NULL;
+	bspcontext->firstseg = NULL;
 
 	// haleyjd 02/19/06: draw polyobjects before static lines
 	if (sub->polyList)
-		R_AddPolyObjects(sub);
+		R_AddPolyObjects(context, sub);
 
 	while (count--)
 	{
-//		CONS_Debug(DBG_GAMELOGIC, "Adding normal line %d...(%d)\n", line->linedef-lines, leveltime);
 		if (!line->glseg && !line->polyseg) // ignore segs that belong to polyobjects
-			R_AddLine(line);
+			R_AddLine(context, line);
 		line++;
-		curline = NULL; /* cph 2001/11/18 - must clear curline now we're done with it, so stuff doesn't try using it for other things */
+		bspcontext->curline = NULL; /* cph 2001/11/18 - must clear curline now we're done with it, so stuff doesn't try using it for other things */
 	}
 }
 
@@ -1234,37 +1232,37 @@ INT32 R_GetPlaneLight(sector_t *sector, fixed_t planeheight, boolean underside)
 //
 // killough 5/2/98: reformatted, removed tail recursion
 
-void R_RenderBSPNode(INT32 bspnum)
+void R_RenderBSPNode(rendercontext_t *context, INT32 bspnum)
 {
 	node_t *bsp;
 	INT32 side;
 
-	ps_numbspcalls++;
+	ps_numbspcalls[context->num]++;
 
 	while (!(bspnum & NF_SUBSECTOR))  // Found a subsector?
 	{
 		bsp = &nodes[bspnum];
 
 		// Decide which side the view point is on.
-		side = R_PointOnSide(viewx, viewy, bsp);
+		side = R_PointOnSide(context->viewcontext.x, context->viewcontext.y, bsp);
+
 		// Recursively divide front space.
-		R_RenderBSPNode(bsp->children[side]);
+		R_RenderBSPNode(context, bsp->children[side]);
 
 		// Possibly divide back space.
-
+		if (!R_CheckBBox(context, bsp->bbox[side^1]))
-		if (!R_CheckBBox(bsp->bbox[side^1]))
 			return;
 
 		bspnum = bsp->children[side^1];
 	}
 
 	// PORTAL CULLING
-	if (portalcullsector) {
+	if (context->bspcontext.portalcullsector) {
 		sector_t *sect = subsectors[bspnum & ~NF_SUBSECTOR].sector;
-		if (sect != portalcullsector)
+		if (sect != context->bspcontext.portalcullsector)
 			return;
-		portalcullsector = NULL;
+		context->bspcontext.portalcullsector = NULL;
 	}
 
-	R_Subsector(bspnum == -1 ? 0 : bspnum & ~NF_SUBSECTOR);
+	R_Subsector(context, bspnum == -1 ? 0 : bspnum & ~NF_SUBSECTOR);
 }

src/r_bsp.h

@@ -18,27 +18,16 @@
 #pragma interface
 #endif
 
-extern seg_t *curline;
-extern side_t *sidedef;
-extern line_t *linedef;
-extern sector_t *frontsector;
-extern sector_t *backsector;
-extern boolean portalline; // is curline a portal seg?
-
-// drawsegs are allocated on the fly... see r_segs.c
-
 extern INT32 checkcoord[12][4];
 
-extern drawseg_t *curdrawsegs;
+struct bspcontext_s;
-extern drawseg_t *drawsegs;
+struct rendercontext_s;
-extern drawseg_t *ds_p;
+struct viewcontext_s;
-extern INT32 doorclosed;
 
 // BSP?
-void R_ClearClipSegs(void);
+void R_ClearClipSegs(struct bspcontext_s *context, INT32 start, INT32 end);
-void R_PortalClearClipSegs(INT32 start, INT32 end);
+void R_ClearDrawSegs(struct bspcontext_s *context);
-void R_ClearDrawSegs(void);
+void R_RenderBSPNode(struct rendercontext_s *context, INT32 bspnum);
-void R_RenderBSPNode(INT32 bspnum);
 
 void R_SortPolyObjects(subsector_t *sub);
 
@@ -46,8 +35,8 @@ extern size_t numpolys;        // number of polyobjects in current subsector
 extern size_t num_po_ptrs;     // number of polyobject pointers allocated
 extern polyobj_t **po_ptrs; // temp ptr array to sort polyobject pointers
 
-sector_t *R_FakeFlat(sector_t *sec, sector_t *tempsec, INT32 *floorlightlevel,
+sector_t *R_FakeFlat(struct viewcontext_s *viewcontext, sector_t *sec, sector_t *tempsec,
-	INT32 *ceilinglightlevel, boolean back);
+	INT32 *floorlightlevel, INT32 *ceilinglightlevel, boolean back);
 boolean R_IsEmptyLine(seg_t *line, sector_t *front, sector_t *back);
 
 INT32 R_GetPlaneLight(sector_t *sector, fixed_t planeheight, boolean underside);

src/r_context.h

@@ -0,0 +1,187 @@
+// SONIC ROBO BLAST 2
+//-----------------------------------------------------------------------------
+// Copyright (C) 1993-1996 by id Software, Inc.
+// Copyright (C) 2020 by Ethan Watson.
+//
+// This program is free software distributed under the
+// terms of the GNU General Public License, version 2.
+// See the 'LICENSE' file for more details.
+//-----------------------------------------------------------------------------
+/// \file  r_context.h
+/// \brief Render context
+
+#ifndef __R_CONTEXT__
+#define __R_CONTEXT__
+
+#include "r_defs.h"
+#include "r_bsp.h"
+#include "r_plane.h"
+#include "r_splats.h"
+#include "r_things.h"
+
+//
+// Render context is required for threaded rendering.
+// So everything "global" goes in here. Everything.
+//
+
+typedef struct viewcontext_s
+{
+	fixed_t x, y, z;
+	angle_t angle, aimingangle;
+	fixed_t cos, sin;
+	sector_t *sector;
+	player_t *player;
+	mobj_t *mobj;
+} viewcontext_t;
+
+typedef struct bspcontext_s
+{
+    seg_t *curline;
+    side_t *sidedef;
+    line_t *linedef;
+    sector_t *frontsector;
+    sector_t *backsector;
+
+    // very ugly realloc() of drawsegs at run-time, I upped it to 512
+    // instead of 256.. and someone managed to send me a level with
+    // 896 drawsegs! So too bad here's a limit removal a-la-Boom
+    drawseg_t *curdrawsegs; /**< This is used to handle multiple lists for masked drawsegs. */
+    drawseg_t *drawsegs;
+    drawseg_t *ds_p;
+    drawseg_t *firstseg;
+    size_t maxdrawsegs;
+
+    // newend is one past the last valid seg
+    cliprange_t *newend;
+    cliprange_t solidsegs[MAXSEGS];
+
+    sector_t ftempsec;
+    sector_t btempsec;
+
+    UINT8 doorclosed;
+
+    // Linked list for portals.
+    struct portal_s *portal_base, *portal_cap;
+
+    // When rendering a portal, it establishes the depth of the current BSP traversal.
+    UINT8 portalrender;
+
+    boolean portalline;
+    line_t *portalclipline;
+    sector_t *portalcullsector;
+    INT32 portalclipstart, portalclipend;
+} bspcontext_t;
+
+typedef struct planecontext_s
+{
+	//SoM: 3/23/2000: Use Boom visplane hashing.
+	visplane_t *visplanes[MAXVISPLANES];
+	visplane_t *freetail;
+	visplane_t **freehead;
+
+	visplane_t *floorplane;
+	visplane_t *ceilingplane;
+	visplane_t *currentplane;
+
+	planemgr_t ffloor[MAXFFLOORS];
+	INT32 numffloors;
+
+	//SoM: 3/23/2000: Use boom opening limit removal
+	size_t maxopenings;
+	INT16 *openings, *lastopening; /// \todo free leak
+
+	//
+	// Clip values are the solid pixel bounding the range.
+	//  floorclip starts out SCREENHEIGHT
+	//  ceilingclip starts out -1
+	//
+	INT16 floorclip[MAXVIDWIDTH], ceilingclip[MAXVIDWIDTH];
+	fixed_t frontscale[MAXVIDWIDTH];
+
+	//
+	// spanstart holds the start of a plane span
+	// initialized to 0 at start
+	//
+	INT32 spanstart[MAXVIDHEIGHT];
+
+	//
+	// texture mapping
+	//
+	lighttable_t **zlight;
+	fixed_t planeheight;
+
+	fixed_t cachedheight[MAXVIDHEIGHT];
+	fixed_t cacheddistance[MAXVIDHEIGHT];
+	fixed_t cachedxstep[MAXVIDHEIGHT];
+	fixed_t cachedystep[MAXVIDHEIGHT];
+
+	fixed_t xoffs, yoffs;
+
+	struct
+	{
+		INT32 offset;
+		fixed_t xfrac, yfrac;
+		boolean active;
+	} ripple;
+
+	struct rastery_s *prastertab;
+	struct rastery_s rastertab[MAXVIDHEIGHT];
+} planecontext_t;
+
+typedef struct spritecontext_s
+{
+	UINT32 visspritecount;
+	UINT32 clippedvissprites;
+	vissprite_t *visspritechunks[MAXVISSPRITES >> VISSPRITECHUNKBITS];
+	vissprite_t overflowsprite;
+
+	lighttable_t **spritelights;
+
+	boolean *sectorvisited;
+
+	drawnode_t nodebankhead;
+	vissprite_t vsprsortedhead;
+} spritecontext_t;
+
+typedef struct wallcontext_s
+{
+	angle_t normalangle;
+	// angle to line origin
+	angle_t angle1;
+	fixed_t distance;
+
+	//
+	// regular wall
+	//
+	angle_t centerangle;
+	fixed_t offset;
+	fixed_t offset2; // for splats
+	fixed_t scale, scalestep;
+	fixed_t midtexturemid, toptexturemid, bottomtexturemid;
+	fixed_t toptextureslide, midtextureslide, bottomtextureslide; // Defines how to adjust Y offsets along the wall for slopes
+	fixed_t midtextureback, midtexturebackslide; // Values for masked midtexture height calculation
+
+	INT32 *silhouette;
+	fixed_t *tsilheight;
+	fixed_t *bsilheight;
+} wallcontext_t;
+
+typedef struct rendercontext_s
+{
+	// Setup
+	vbuffer_t buffer;
+	INT32 num;
+
+	INT32 begincolumn;
+	INT32 endcolumn;
+
+	viewcontext_t viewcontext;
+	bspcontext_t bspcontext;
+	planecontext_t planecontext;
+	spritecontext_t spritecontext;
+
+	colcontext_t colcontext;
+	spancontext_t spancontext;
+} rendercontext_t;
+
+#endif

src/r_data.c

@@ -1240,8 +1240,8 @@ void R_PrecacheLevel(void)
 		return;
 
 	// do not flush the memory, Z_Malloc twice with same user will cause error in Z_CheckHeap()
-	if (rendermode != render_soft)
+	// if (rendermode != render_soft)
-		return;
+	// 	return;
 
 	// Precache flats.
 	flatmemory = P_PrecacheLevelFlats();

src/r_defs.h

@@ -22,7 +22,7 @@
 // SECTORS do store MObjs anyway.
 #include "p_mobj.h"
 
-#include "screen.h" // MAXVIDWIDTH, MAXVIDHEIGHT
+#include "screen.h"
 
 #ifdef HWRENDER
 #include "m_aatree.h"
@@ -41,6 +41,9 @@ typedef struct
 	INT32 last;
 } cliprange_t;
 
+// Fix from boom.
+#define MAXSEGS (MAXVIDWIDTH/2+1)
+
 // Silhouette, needed for clipping segs (mainly) and sprites representing things.
 #define SIL_NONE   0
 #define SIL_BOTTOM 1
@@ -56,6 +59,15 @@ typedef UINT8 lighttable_t;
 #define CMF_FADEFULLBRIGHTSPRITES  1
 #define CMF_FOG 4
 
+typedef struct vbuffer_s
+{
+	UINT8 *screens[5];
+} vbuffer_t;
+
+typedef struct {
+	float x, y, z;
+} floatv3_t;
+
 // ExtraColormap type. Use for extra_colormaps from now on.
 typedef struct extracolormap_s
 {

src/r_draw.c

@@ -62,14 +62,6 @@ UINT8 *topleft;
 //                      COLUMN DRAWING CODE STUFF
 // =========================================================================
 
-lighttable_t *dc_colormap;
-INT32 dc_x = 0, dc_yl = 0, dc_yh = 0;
-
-fixed_t dc_iscale, dc_texturemid;
-UINT8 dc_hires; // under MSVC boolean is a byte, while on other systems, it a bit,
-               // soo lets make it a byte on all system for the ASM code
-UINT8 *dc_source;
-
 // -----------------------
 // translucency stuff here
 // -----------------------
@@ -78,48 +70,11 @@ UINT8 *dc_source;
 UINT8 *transtables; // translucency tables
 UINT8 *blendtables[NUMBLENDMAPS];
 
-/**	\brief R_DrawTransColumn uses this
-*/
-UINT8 *dc_transmap; // one of the translucency tables
-
-// ----------------------
-// translation stuff here
-// ----------------------
-
-
-/**	\brief R_DrawTranslatedColumn uses this
-*/
-UINT8 *dc_translation;
-
-struct r_lightlist_s *dc_lightlist = NULL;
-INT32 dc_numlights = 0, dc_maxlights, dc_texheight;
-
 // =========================================================================
 //                      SPAN DRAWING CODE STUFF
 // =========================================================================
 
-INT32 ds_y, ds_x1, ds_x2;
+float focallengthf;
-lighttable_t *ds_colormap;
-lighttable_t *ds_translation; // Lactozilla: Sprite splat drawer
-
-fixed_t ds_xfrac, ds_yfrac, ds_xstep, ds_ystep;
-INT32 ds_waterofs, ds_bgofs;
-
-UINT16 ds_flatwidth, ds_flatheight;
-boolean ds_powersoftwo;
-
-UINT8 *ds_source; // points to the start of a flat
-UINT8 *ds_transmap; // one of the translucency tables
-
-// Vectors for Software's tilted slope drawers
-floatv3_t *ds_su, *ds_sv, *ds_sz;
-floatv3_t *ds_sup, *ds_svp, *ds_szp;
-float focallengthf, zeroheight;
-
-/**	\brief Variable flat sizes
-*/
-
-UINT32 nflatxshift, nflatyshift, nflatshiftup, nflatmask;
 
 // =========================================================================
 //                   TRANSLATION COLORMAP CODE
@@ -572,6 +527,15 @@ static void R_GenerateTranslationColormap(UINT8 *dest_colormap, INT32 skinnum, U
 		dest_colormap[starttranscolor + i] = (UINT8)skincolors[color].ramp[i];
 }
 
+#ifdef HAVE_THREADS
+static I_mutex r_colormap_mutex;
+
+#  define Lock_state()    I_lock_mutex(&r_colormap_mutex)
+#  define Unlock_state() I_unlock_mutex(r_colormap_mutex)
+#else/*HAVE_THREADS*/
+#  define Lock_state()
+#  define Unlock_state()
+#endif/*HAVE_THREADS*/
 
 /**	\brief	Retrieves a translation colormap from the cache.
 
@@ -589,13 +553,16 @@ UINT8* R_GetTranslationColormap(INT32 skinnum, skincolornum_t color, UINT8 flags
 
 	if (flags & GTC_CACHE)
 	{
+		Lock_state();
 		// Allocate table for skin if necessary
 		if (!translationtablecache[skintableindex])
 			translationtablecache[skintableindex] = Z_Calloc(MAXSKINCOLORS * sizeof(UINT8**), PU_STATIC, NULL);
+		Unlock_state();
 
 		// Get colormap
 		ret = translationtablecache[skintableindex][color];
 
+		Lock_state();
 		// Rebuild the cache if necessary
 		if (skincolor_modified[color])
 		{
@@ -605,18 +572,21 @@ UINT8* R_GetTranslationColormap(INT32 skinnum, skincolornum_t color, UINT8 flags
 
 			skincolor_modified[color] = false;
 		}
+		Unlock_state();
 	}
 	else ret = NULL;
 
 	// Generate the colormap if necessary
 	if (!ret)
 	{
+		Lock_state();
 		ret = Z_MallocAlign(NUM_PALETTE_ENTRIES, (flags & GTC_CACHE) ? PU_LEVEL : PU_STATIC, NULL, 8);
 		R_GenerateTranslationColormap(ret, skinnum, color);
 
 		// Cache the colormap if desired
 		if (flags & GTC_CACHE)
 			translationtablecache[skintableindex][color] = ret;
+		Unlock_state();
 	}
 
 	return ret;
@@ -887,13 +857,37 @@ void R_DrawViewBorder(void)
 //                   INCLUDE 8bpp DRAWING CODE HERE
 // ==========================================================================
 
+// R_CalcTiltedLighting
+// Exactly what it says on the tin. I wish I wasn't too lazy to explain things properly.
+static void R_CalcTiltedLighting(spancontext_t *ds, fixed_t start, fixed_t end)
+{
+	// ZDoom uses a different lighting setup to us, and I couldn't figure out how to adapt their version
+	// of this function. Here's my own.
+	INT32 left = ds->x1, right = ds->x2;
+	fixed_t step = (end-start)/(ds->x2-ds->x1+1);
+	INT32 i;
+
+	// I wanna do some optimizing by checking for out-of-range segments on either side to fill in all at once,
+	// but I'm too bad at coding to not crash the game trying to do that. I guess this is fast enough for now...
+
+	for (i = left; i <= right; i++) {
+		ds->tiltlighting[i] = (start += step) >> FRACBITS;
+		if (ds->tiltlighting[i] < 0)
+			ds->tiltlighting[i] = 0;
+		else if (ds->tiltlighting[i] >= MAXLIGHTSCALE)
+			ds->tiltlighting[i] = MAXLIGHTSCALE-1;
+	}
+}
+
+#define PLANELIGHTFLOAT (BASEVIDWIDTH * BASEVIDWIDTH / vid.width / ds->zeroheight / 21.0f * FIXED_TO_FLOAT(fovtan))
+
+// Lighting is simple. It's just linear interpolation from start to end
+#define CALC_TILTED_LIGHTING { \
+	float planelightfloat = PLANELIGHTFLOAT; \
+	float lightend = (iz + ds->szp->x*width) * planelightfloat; \
+	float lightstart = iz * planelightfloat; \
+	R_CalcTiltedLighting(ds, FLOAT_TO_FIXED(lightstart), FLOAT_TO_FIXED(lightend)); \
+}
+
 #include "r_draw8.c"
 #include "r_draw8_npo2.c"
-
-// ==========================================================================
-//                   INCLUDE 16bpp DRAWING CODE HERE
-// ==========================================================================
-
-#ifdef HIGHCOLOR
-#include "r_draw16.c"
-#endif

src/r_draw.h

@@ -25,61 +25,7 @@ extern UINT8 *ylookup2[MAXVIDHEIGHT*4];
 extern INT32 columnofs[MAXVIDWIDTH*4];
 extern UINT8 *topleft;
 
-// -------------------------
+extern float focallengthf;
-// COLUMN DRAWING CODE STUFF
-// -------------------------
-
-extern lighttable_t *dc_colormap;
-extern INT32 dc_x, dc_yl, dc_yh;
-extern fixed_t dc_iscale, dc_texturemid;
-extern UINT8 dc_hires;
-
-extern UINT8 *dc_source; // first pixel in a column
-
-// translucency stuff here
-extern UINT8 *dc_transmap;
-
-// translation stuff here
-
-extern UINT8 *dc_translation;
-
-extern struct r_lightlist_s *dc_lightlist;
-extern INT32 dc_numlights, dc_maxlights;
-
-//Fix TUTIFRUTI
-extern INT32 dc_texheight;
-
-// -----------------------
-// SPAN DRAWING CODE STUFF
-// -----------------------
-
-extern INT32 ds_y, ds_x1, ds_x2;
-extern lighttable_t *ds_colormap;
-extern lighttable_t *ds_translation;
-
-extern fixed_t ds_xfrac, ds_yfrac, ds_xstep, ds_ystep;
-extern INT32 ds_waterofs, ds_bgofs;
-
-extern UINT16 ds_flatwidth, ds_flatheight;
-extern boolean ds_powersoftwo;
-
-extern UINT8 *ds_source;
-extern UINT8 *ds_transmap;
-
-typedef struct {
-	float x, y, z;
-} floatv3_t;
-
-// Vectors for Software's tilted slope drawers
-extern floatv3_t *ds_su, *ds_sv, *ds_sz;
-extern floatv3_t *ds_sup, *ds_svp, *ds_szp;
-extern float focallengthf, zeroheight;
-
-// Variable flat sizes
-extern UINT32 nflatxshift;
-extern UINT32 nflatyshift;
-extern UINT32 nflatshiftup;
-extern UINT32 nflatmask;
 
 /// \brief Top border
 #define BRDR_T 0
@@ -163,85 +109,180 @@ void R_DrawViewBorder(void);
 
 #define TRANSPARENTPIXEL 255
 
+typedef struct colcontext_s
+{
+	void (*func) (struct colcontext_s *);
+	vbuffer_t *dest;
+
+	INT32 x, yl, yh;
+	fixed_t iscale, texturemid;
+	lighttable_t *colormap;
+
+	UINT8 *source; // first pixel in a column
+
+	// translucency stuff here
+	UINT8 *transmap;
+
+	// translation stuff here
+	UINT8 *translation;
+
+	struct r_lightlist_s *lightlist;
+	INT32 numlights, maxlights;
+
+	//Fix TUTIFRUTI
+	INT32 texheight;
+
+	// vars for R_DrawMaskedColumn
+	// Rum and Raisin put these on the sprite context.
+	// I put them on the column context because it felt
+	// more appropriate (for [REDACTED] anyway)
+	INT16 *mfloorclip;
+	INT16 *mceilingclip;
+
+	fixed_t spryscale, sprtopscreen, sprbotscreen;
+	fixed_t windowtop, windowbottom;
+
+	// column->length : for flipped column function pointers and multi-patch on 2sided wall = texture->height
+	INT32 lengthcol;
+} colcontext_t;
+
+typedef struct spancontext_s
+{
+	void (*func) (struct spancontext_s *);
+	vbuffer_t *dest;
+
+	INT32 y, x1, x2;
+	lighttable_t *colormap;
+	lighttable_t *translation;
+
+	fixed_t xfrac, yfrac, xstep, ystep;
+	INT32 waterofs, bgofs;
+
+	UINT16 flatwidth, flatheight;
+	boolean powersoftwo;
+
+	UINT8 *source;
+	UINT8 *transmap;
+
+	// Vectors for Software's tilted slope drawers
+	floatv3_t *su, *sv, *sz;
+	floatv3_t *sup, *svp, *szp;
+	floatv3_t slope_origin, slope_u, slope_v;
+	float zeroheight;
+
+	// Variable flat sizes
+	UINT32 nflatxshift;
+	UINT32 nflatyshift;
+	UINT32 nflatshiftup;
+	UINT32 nflatmask;
+
+	lighttable_t **zlight;
+	INT32 tiltlighting[MAXVIDWIDTH];
+} spancontext_t;
+
+typedef void (*colfunc_t) (colcontext_t *);
+typedef void (*spanfunc_t) (spancontext_t *);
+
+// ---------------------------------------------
+// color mode dependent drawer function pointers
+// ---------------------------------------------
+
+#define BASEDRAWFUNC 0
+
+enum
+{
+	COLDRAWFUNC_BASE = BASEDRAWFUNC,
+	COLDRAWFUNC_FUZZY,
+	COLDRAWFUNC_TRANS,
+	COLDRAWFUNC_SHADE,
+	COLDRAWFUNC_SHADOWED,
+	COLDRAWFUNC_TRANSTRANS,
+	COLDRAWFUNC_TWOSMULTIPATCH,
+	COLDRAWFUNC_TWOSMULTIPATCHTRANS,
+	COLDRAWFUNC_FOG,
+
+	COLDRAWFUNC_MAX
+};
+
+extern colfunc_t colfuncs[COLDRAWFUNC_MAX];
+
+enum
+{
+	SPANDRAWFUNC_BASE = BASEDRAWFUNC,
+	SPANDRAWFUNC_TRANS,
+	SPANDRAWFUNC_TILTED,
+	SPANDRAWFUNC_TILTEDTRANS,
+
+	SPANDRAWFUNC_SPLAT,
+	SPANDRAWFUNC_TRANSSPLAT,
+	SPANDRAWFUNC_TILTEDSPLAT,
+
+	SPANDRAWFUNC_SPRITE,
+	SPANDRAWFUNC_TRANSSPRITE,
+	SPANDRAWFUNC_TILTEDSPRITE,
+	SPANDRAWFUNC_TILTEDTRANSSPRITE,
+
+	SPANDRAWFUNC_WATER,
+	SPANDRAWFUNC_TILTEDWATER,
+
+	SPANDRAWFUNC_FOG,
+
+	SPANDRAWFUNC_MAX
+};
+
+extern spanfunc_t spanfuncs[SPANDRAWFUNC_MAX];
+extern spanfunc_t spanfuncs_npo2[SPANDRAWFUNC_MAX];
+
 // -----------------
 // 8bpp DRAWING CODE
 // -----------------
 
-void R_DrawColumn_8(void);
+void R_DrawColumn_8(colcontext_t *dc);
-void R_DrawShadeColumn_8(void);
+void R_DrawShadeColumn_8(colcontext_t *dc);
-void R_DrawTranslucentColumn_8(void);
+void R_DrawTranslucentColumn_8(colcontext_t *dc);
-void R_DrawTranslatedColumn_8(void);
+void R_DrawTranslatedColumn_8(colcontext_t *dc);
-void R_DrawTranslatedTranslucentColumn_8(void);
+void R_DrawTranslatedTranslucentColumn_8(colcontext_t *dc);
-void R_Draw2sMultiPatchColumn_8(void);
+void R_Draw2sMultiPatchColumn_8(colcontext_t *dc);
-void R_Draw2sMultiPatchTranslucentColumn_8(void);
+void R_Draw2sMultiPatchTranslucentColumn_8(colcontext_t *dc);
-void R_DrawFogColumn_8(void);
+void R_DrawFogColumn_8(colcontext_t *dc);
-void R_DrawColumnShadowed_8(void);
+void R_DrawColumnShadowed_8(colcontext_t *dc);
 
-#define PLANELIGHTFLOAT (BASEVIDWIDTH * BASEVIDWIDTH / vid.width / zeroheight / 21.0f * FIXED_TO_FLOAT(fovtan))
+void R_DrawSpan_8(spancontext_t *ds);
+void R_DrawTranslucentSpan_8(spancontext_t *ds);
+void R_DrawTiltedSpan_8(spancontext_t *ds);
+void R_DrawTiltedTranslucentSpan_8(spancontext_t *ds);
 
-void R_DrawSpan_8(void);
+void R_DrawSplat_8(spancontext_t *ds);
-void R_DrawTranslucentSpan_8(void);
+void R_DrawTranslucentSplat_8(spancontext_t *ds);
-void R_DrawTiltedSpan_8(void);
+void R_DrawTiltedSplat_8(spancontext_t *ds);
-void R_DrawTiltedTranslucentSpan_8(void);
 
-void R_DrawSplat_8(void);
+void R_DrawFloorSprite_8(spancontext_t *ds);
-void R_DrawTranslucentSplat_8(void);
+void R_DrawTranslucentFloorSprite_8(spancontext_t *ds);
-void R_DrawTiltedSplat_8(void);
+void R_DrawTiltedFloorSprite_8(spancontext_t *ds);
+void R_DrawTiltedTranslucentFloorSprite_8(spancontext_t *ds);
 
-void R_DrawFloorSprite_8(void);
+void R_DrawTranslucentWaterSpan_8(spancontext_t *ds);
-void R_DrawTranslucentFloorSprite_8(void);
+void R_DrawTiltedTranslucentWaterSpan_8(spancontext_t *ds);
-void R_DrawTiltedFloorSprite_8(void);
-void R_DrawTiltedTranslucentFloorSprite_8(void);
 
-void R_CalcTiltedLighting(fixed_t start, fixed_t end);
+void R_DrawFogSpan_8(spancontext_t *ds);
-extern INT32 tiltlighting[MAXVIDWIDTH];
-
-void R_DrawTranslucentWaterSpan_8(void);
-void R_DrawTiltedTranslucentWaterSpan_8(void);
-
-void R_DrawFogSpan_8(void);
 
 // Lactozilla: Non-powers-of-two
-void R_DrawSpan_NPO2_8(void);
+void R_DrawSpan_NPO2_8(spancontext_t *ds);
-void R_DrawTranslucentSpan_NPO2_8(void);
+void R_DrawTranslucentSpan_NPO2_8(spancontext_t *ds);
-void R_DrawTiltedSpan_NPO2_8(void);
+void R_DrawTiltedSpan_NPO2_8(spancontext_t *ds);
-void R_DrawTiltedTranslucentSpan_NPO2_8(void);
+void R_DrawTiltedTranslucentSpan_NPO2_8(spancontext_t *ds);
 
-void R_DrawSplat_NPO2_8(void);
+void R_DrawSplat_NPO2_8(spancontext_t *ds);
-void R_DrawTranslucentSplat_NPO2_8(void);
+void R_DrawTranslucentSplat_NPO2_8(spancontext_t *ds);
-void R_DrawTiltedSplat_NPO2_8(void);
+void R_DrawTiltedSplat_NPO2_8(spancontext_t *ds);
 
-void R_DrawFloorSprite_NPO2_8(void);
+void R_DrawFloorSprite_NPO2_8(spancontext_t *ds);
-void R_DrawTranslucentFloorSprite_NPO2_8(void);
+void R_DrawTranslucentFloorSprite_NPO2_8(spancontext_t *ds);
-void R_DrawTiltedFloorSprite_NPO2_8(void);
+void R_DrawTiltedFloorSprite_NPO2_8(spancontext_t *ds);
-void R_DrawTiltedTranslucentFloorSprite_NPO2_8(void);
+void R_DrawTiltedTranslucentFloorSprite_NPO2_8(spancontext_t *ds);
 
-void R_DrawTranslucentWaterSpan_NPO2_8(void);
+void R_DrawTranslucentWaterSpan_NPO2_8(spancontext_t *ds);
-void R_DrawTiltedTranslucentWaterSpan_NPO2_8(void);
+void R_DrawTiltedTranslucentWaterSpan_NPO2_8(spancontext_t *ds);
-
-#ifdef USEASM
-void ASMCALL R_DrawColumn_8_ASM(void);
-void ASMCALL R_DrawShadeColumn_8_ASM(void);
-void ASMCALL R_DrawTranslucentColumn_8_ASM(void);
-void ASMCALL R_Draw2sMultiPatchColumn_8_ASM(void);
-
-void ASMCALL R_DrawColumn_8_MMX(void);
-
-void ASMCALL R_Draw2sMultiPatchColumn_8_MMX(void);
-void ASMCALL R_DrawSpan_8_MMX(void);
-#endif
-
-// ------------------
-// 16bpp DRAWING CODE
-// ------------------
-
-#ifdef HIGHCOLOR
-void R_DrawColumn_16(void);
-void R_DrawWallColumn_16(void);
-void R_DrawTranslucentColumn_16(void);
-void R_DrawTranslatedColumn_16(void);
-void R_DrawSpan_16(void);
-#endif
 
 // =========================================================================
 #endif  // __R_DRAW__

src/r_draw16.c

@@ -1,214 +0,0 @@
-// SONIC ROBO BLAST 2
-//-----------------------------------------------------------------------------
-// Copyright (C) 1998-2000 by DooM Legacy Team.
-// Copyright (C) 1999-2021 by Sonic Team Junior.
-//
-// This program is free software distributed under the
-// terms of the GNU General Public License, version 2.
-// See the 'LICENSE' file for more details.
-//-----------------------------------------------------------------------------
-/// \file  r_draw16.c
-/// \brief 16bpp (HIGHCOLOR) span/column drawer functions
-/// \note  no includes because this is included as part of r_draw.c
-
-// ==========================================================================
-// COLUMNS
-// ==========================================================================
-
-/// \brief kick out the upper bit of each component (we're in 5 : 5 : 5)
-#define HIMASK1 0x7bde
-
-/**	\brief The R_DrawColumn_16 function
-	standard upto 128high posts column drawer
-*/
-void R_DrawColumn_16(void)
-{
-	INT32 count;
-	INT16 *dest;
-	fixed_t frac, fracstep;
-
-	count = dc_yh - dc_yl + 1;
-
-	// Zero length, column does not exceed a pixel.
-	if (count <= 0)
-		return;
-
-#ifdef RANGECHECK
-	if (dc_x >= vid.width || dc_yl < 0 || dc_yh >= vid.height)
-		I_Error("R_DrawColumn_16: %d to %d at %d", dc_yl, dc_yh, dc_x);
-#endif
-
-	// Framebuffer destination address.
-	// Use ylookup LUT to avoid multiply with ScreenWidth.
-	// Use columnofs LUT for subwindows?
-	dest = (INT16 *)(void *)(ylookup[dc_yl] + columnofs[dc_x]);
-
-	// Determine scaling, which is the only mapping to be done.
-	fracstep = dc_iscale;
-	frac = dc_texturemid + (dc_yl - centery)*fracstep;
-
-	// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
-	// This is as fast as it gets.
-
-	do
-	{
-		// Re-map color indices from wall texture column using a lighting/special effects LUT.
-		*dest = hicolormaps[((INT16 *)(void *)dc_source)[(frac>>FRACBITS)&127]>>1];
-
-		dest += vid.width;
-		frac += fracstep;
-	} while (--count);
-}
-
-/**	\brief The R_DrawWallColumn_16 function
-	LAME cutnpaste: same as R_DrawColumn_16 but wraps around 256
-	instead of 128 for the tall sky textures (256x240)
-*/
-void R_DrawWallColumn_16(void)
-{
-	INT32 count;
-	INT16 *dest;
-	fixed_t frac, fracstep;
-
-	count = dc_yh - dc_yl + 1;
-
-	// Zero length, column does not exceed a pixel.
-	if (count <= 0)
-		return;
-
-#ifdef RANGECHECK
-	if (dc_x >= vid.width || dc_yl < 0 || dc_yh >= vid.height)
-		I_Error("R_DrawWallColumn_16: %d to %d at %d", dc_yl, dc_yh, dc_x);
-#endif
-
-	dest = (INT16 *)(void *)(ylookup[dc_yl] + columnofs[dc_x]);
-
-	fracstep = dc_iscale;
-	frac = dc_texturemid + (dc_yl - centery)*fracstep;
-
-	do
-	{
-		*dest = hicolormaps[((INT16 *)(void *)dc_source)[(frac>>FRACBITS)&255]>>1];
-
-		dest += vid.width;
-		frac += fracstep;
-	} while (--count);
-}
-
-/**	\brief The R_DrawTranslucentColumn_16 function
-		LAME cutnpaste: same as R_DrawColumn_16 but does
-		translucent
-*/
-void R_DrawTranslucentColumn_16(void)
-{
-	INT32 count;
-	INT16 *dest;
-	fixed_t frac, fracstep;
-
-	// check out coords for src*
-	if ((dc_yl < 0) || (dc_x >= vid.width))
-		return;
-
-	count = dc_yh - dc_yl;
-	if (count < 0)
-		return;
-
-#ifdef RANGECHECK
-	if (dc_x >= vid.width || dc_yl < 0 || dc_yh >= vid.height)
-		I_Error("R_DrawTranslucentColumn_16: %d to %d at %d", dc_yl, dc_yh, dc_x);
-#endif
-
-	// FIXME. As above.
-	dest = (INT16 *)(void *)(ylookup[dc_yl] + columnofs[dc_x]);
-
-	// Looks familiar.
-	fracstep = dc_iscale;
-	frac = dc_texturemid + (dc_yl - centery)*fracstep;
-
-	// Here we do an additional index re-mapping.
-	do
-	{
-		*dest = (INT16)((INT16)((color8to16[dc_source[frac>>FRACBITS]]>>1) & 0x39ce)
-			+ (INT16)(((*dest & HIMASK1)) & 0x7fff));
-
-		dest += vid.width;
-		frac += fracstep;
-	} while (count--);
-}
-
-/**	\brief The R_DrawTranslatedColumn_16 function
-	?
-*/
-void R_DrawTranslatedColumn_16(void)
-{
-	INT32 count;
-	INT16 *dest;
-	fixed_t frac, fracstep;
-
-	count = dc_yh - dc_yl;
-	if (count < 0)
-		return;
-
-#ifdef RANGECHECK
-	if (dc_x >= vid.width || dc_yl < 0 || dc_yh >= vid.height)
-		I_Error("R_DrawTranslatedColumn_16: %d to %d at %d", dc_yl, dc_yh, dc_x);
-#endif
-
-	dest = (INT16 *)(void *)(ylookup[dc_yl] + columnofs[dc_x]);
-
-	// Looks familiar.
-	fracstep = dc_iscale;
-	frac = dc_texturemid + (dc_yl - centery)*fracstep;
-
-	// Here we do an additional index re-mapping.
-	do
-	{
-		*dest = color8to16[dc_colormap[dc_translation[dc_source[frac>>FRACBITS]]]];
-		dest += vid.width;
-
-		frac += fracstep;
-	} while (count--);
-}
-
-// ==========================================================================
-// SPANS
-// ==========================================================================
-
-/**	\brief The R_*_16 function
-	Draws the actual span.
-*/
-void R_DrawSpan_16(void)
-{
-	fixed_t xfrac, yfrac;
-	INT16 *dest;
-	INT32 count, spot;
-
-#ifdef RANGECHECK
-	if (ds_x2 < ds_x1 || ds_x1 < 0 || ds_x2 >= vid.width || ds_y > vid.height)
-		I_Error("R_DrawSpan_16: %d to %d at %d", ds_x1, ds_x2, ds_y);
-#endif
-
-	xfrac = ds_xfrac;
-	yfrac = ds_yfrac;
-
-	dest = (INT16 *)(void *)(ylookup[ds_y] + columnofs[ds_x1]);
-
-	// We do not check for zero spans here?
-	count = ds_x2 - ds_x1;
-
-	if (count <= 0) // We do now!
-		return;
-
-	do
-	{
-		// Current texture index in u, v.
-		spot = ((yfrac>>(16-6))&(63*64)) + ((xfrac>>16)&63);
-
-		// Lookup pixel from flat texture tile, re-index using light/colormap.
-		*dest++ = hicolormaps[((INT16 *)(void *)ds_source)[spot]>>1];
-
-		// Next step in u, v.
-		xfrac += ds_xstep;
-		yfrac += ds_ystep;
-	} while (count--);
-}

src/r_local.h

@@ -34,8 +34,7 @@
 #include "r_textures.h"
 #include "r_things.h"
 #include "r_draw.h"
-
+#include "r_context.h"
-extern drawseg_t *firstseg;
 
 void SplitScreen_OnChange(void);
 

src/r_main.c

@@ -53,6 +53,22 @@ INT64 mytotal = 0;
 // Fineangles in the SCREENWIDTH wide window.
 #define FIELDOFVIEW 2048
 
+typedef struct renderdata_s
+{
+	rendercontext_t context;
+	I_thread_handle thread;
+	I_Atomicval_t running;
+	I_Atomicval_t shouldquit;
+	I_Atomicval_t framewaiting;
+	I_Atomicval_t framefinished;
+} renderdata_t;
+
+static renderdata_t *renderdatas;
+
+INT32 numrendercontexts = MAX_RENDER_THREADS;
+INT32 numusablerendercontexts = 0;
+boolean renderthreaded = false;
+
 // increment every time a check is made
 size_t validcount = 1;
 
@@ -102,20 +118,21 @@ extracolormap_t *extra_colormaps = NULL;
 // Render stats
 precise_t ps_prevframetime = 0;
 precise_t ps_rendercalltime = 0;
+precise_t ps_postprocesstime = 0;
 precise_t ps_uitime = 0;
 precise_t ps_swaptime = 0;
 
-precise_t ps_bsptime = 0;
+precise_t ps_bsptime[MAX_RENDER_THREADS] = {0};
 
-precise_t ps_sw_spritecliptime = 0;
+precise_t ps_sw_spritecliptime[MAX_RENDER_THREADS] = {0};
-precise_t ps_sw_portaltime = 0;
+precise_t ps_sw_portaltime[MAX_RENDER_THREADS] = {0};
-precise_t ps_sw_planetime = 0;
+precise_t ps_sw_planetime[MAX_RENDER_THREADS] = {0};
-precise_t ps_sw_maskedtime = 0;
+precise_t ps_sw_maskedtime[MAX_RENDER_THREADS] = {0};
 
-int ps_numbspcalls = 0;
+int ps_numbspcalls[MAX_RENDER_THREADS] = {0};
-int ps_numsprites = 0;
+int ps_numsprites[MAX_RENDER_THREADS] = {0};
-int ps_numdrawnodes = 0;
+int ps_numdrawnodes[MAX_RENDER_THREADS] = {0};
-int ps_numpolyobjects = 0;
+int ps_numpolyobjects[MAX_RENDER_THREADS] = {0};
 
 static CV_PossibleValue_t drawdist_cons_t[] = {
 	{256, "256"},	{512, "512"},	{768, "768"},
@@ -132,7 +149,7 @@ static CV_PossibleValue_t drawdist_precip_cons_t[] = {
 
 static CV_PossibleValue_t fov_cons_t[] = {{60*FRACUNIT, "MIN"}, {179*FRACUNIT, "MAX"}, {0, NULL}};
 static CV_PossibleValue_t translucenthud_cons_t[] = {{0, "MIN"}, {10, "MAX"}, {0, NULL}};
-static CV_PossibleValue_t maxportals_cons_t[] = {{0, "MIN"}, {12, "MAX"}, {0, NULL}}; // lmao rendering 32 portals, you're a card
+static CV_PossibleValue_t maxportals_cons_t[] = {{0, "MIN"}, {12, "MAX"}, {0, NULL}};
 static CV_PossibleValue_t homremoval_cons_t[] = {{0, "No"}, {1, "Yes"}, {2, "Flash"}, {0, NULL}};
 
 static void Fov_OnChange(void);
@@ -163,9 +180,11 @@ consvar_t cv_drawdist_precip = CVAR_INIT ("drawdist_precip", "1024", CV_SAVE, dr
 //consvar_t cv_precipdensity = CVAR_INIT ("precipdensity", "Moderate", CV_SAVE, precipdensity_cons_t, NULL);
 consvar_t cv_fov = CVAR_INIT ("fov", "90", CV_FLOAT|CV_CALL, fov_cons_t, Fov_OnChange);
 
-// Okay, whoever said homremoval causes a performance hit should be shot.
 consvar_t cv_homremoval = CVAR_INIT ("homremoval", "No", CV_SAVE, homremoval_cons_t, NULL);
 
+static CV_PossibleValue_t numthreads_cons_t[] = {{1, "MIN"}, {8, "MAX"}, {0, NULL}};
+consvar_t cv_numthreads = CVAR_INIT ("numthreads", "2", CV_SAVE, numthreads_cons_t, NULL);
+
 consvar_t cv_maxportals = CVAR_INIT ("maxportals", "2", CV_SAVE, maxportals_cons_t, NULL);
 
 consvar_t cv_renderstats = CVAR_INIT ("renderstats", "Off", 0, CV_OnOff, NULL);
@@ -411,35 +430,6 @@ angle_t R_PointToAngleEx(INT32 x2, INT32 y2, INT32 x1, INT32 y1)
 		0;
 }
 
-//
-// R_ScaleFromGlobalAngle
-// Returns the texture mapping scale for the current line (horizontal span)
-//  at the given angle.
-// rw_distance must be calculated first.
-//
-// killough 5/2/98: reformatted, cleaned up
-//
-// note: THIS IS USED ONLY FOR WALLS!
-fixed_t R_ScaleFromGlobalAngle(angle_t visangle)
-{
-	angle_t anglea = ANGLE_90 + (visangle-viewangle);
-	angle_t angleb = ANGLE_90 + (visangle-rw_normalangle);
-	fixed_t den = FixedMul(rw_distance, FINESINE(anglea>>ANGLETOFINESHIFT));
-	// proff 11/06/98: Changed for high-res
-	fixed_t num = FixedMul(projectiony, FINESINE(angleb>>ANGLETOFINESHIFT));
-
-	if (den > num>>16)
-	{
-		num = FixedDiv(num, den);
-		if (num > 64*FRACUNIT)
-			return 64*FRACUNIT;
-		if (num < 256)
-			return 256;
-		return num;
-	}
-	return 64*FRACUNIT;
-}
-
 //
 // R_DoCulling
 // Checks viewz and top/bottom heights of an item against culling planes
@@ -883,6 +873,285 @@ void R_ApplyViewMorph(void)
 			vid.width*vid.bpp, vid.height, vid.width*vid.bpp, vid.width);
 }
 
+static void R_PortalFrame(rendercontext_t *context, portal_t *portal)
+{
+	bspcontext_t *bspcontext = &context->bspcontext;
+
+	context->viewcontext.x = portal->viewx;
+	context->viewcontext.y = portal->viewy;
+	context->viewcontext.z = portal->viewz;
+
+	context->viewcontext.angle = portal->viewangle;
+	context->viewcontext.sin = FINESINE(portal->viewangle>>ANGLETOFINESHIFT);
+	context->viewcontext.cos = FINECOSINE(portal->viewangle>>ANGLETOFINESHIFT);
+
+	bspcontext->portalclipstart = portal->start;
+	bspcontext->portalclipend = portal->end;
+
+	if (portal->clipline != -1)
+	{
+		line_t *clipline = &lines[portal->clipline];
+		bspcontext->portalclipline = clipline;
+		bspcontext->portalcullsector = clipline->frontsector;
+		context->viewcontext.sector = clipline->frontsector;
+	}
+	else
+	{
+		bspcontext->portalclipline = NULL;
+		bspcontext->portalcullsector = NULL;
+		context->viewcontext.sector = R_PointInSubsector(context->viewcontext.x, context->viewcontext.y)->sector;
+	}
+}
+
+static void Mask_Pre (rendercontext_t *context, maskcount_t* m)
+{
+	m->drawsegs[0] = context->bspcontext.ds_p - context->bspcontext.drawsegs;
+	m->vissprites[0] = context->spritecontext.visspritecount;
+	m->viewx = context->viewcontext.x;
+	m->viewy = context->viewcontext.y;
+	m->viewz = context->viewcontext.z;
+	m->viewsector = context->viewcontext.sector;
+}
+
+static void Mask_Post (rendercontext_t *context, maskcount_t* m)
+{
+	m->drawsegs[1] = context->bspcontext.ds_p - context->bspcontext.drawsegs;
+	m->vissprites[1] = context->spritecontext.visspritecount;
+}
+
+static void R_RenderViewContext(rendercontext_t *rendercontext)
+{
+	UINT8			nummasks	= 1;
+	maskcount_t*	masks		= malloc(sizeof(maskcount_t));
+
+	bspcontext_t *bspcontext = &rendercontext->bspcontext;
+	planecontext_t *planecontext = &rendercontext->planecontext;
+
+	INT32 i = rendercontext->num;
+
+	// The head node is the last node output.
+	Mask_Pre(rendercontext, &masks[nummasks - 1]);
+	bspcontext->curdrawsegs = bspcontext->ds_p;
+
+	ps_numbspcalls[i] = ps_numpolyobjects[i] = ps_numdrawnodes[i] = 0;
+
+	ps_bsptime[i] = I_GetPreciseTime();
+	R_RenderBSPNode(rendercontext, (INT32)numnodes - 1);
+	ps_bsptime[i] = I_GetPreciseTime() - ps_bsptime[i];
+	ps_numsprites[i] = rendercontext->spritecontext.visspritecount;
+
+	Mask_Post(rendercontext, &masks[nummasks - 1]);
+
+	ps_sw_spritecliptime[i] = I_GetPreciseTime();
+	R_ClipSprites(rendercontext, bspcontext->drawsegs, NULL);
+	ps_sw_spritecliptime[i] = I_GetPreciseTime() - ps_sw_spritecliptime[i];
+
+	// Add skybox portals caused by sky visplanes.
+	if (cv_skybox.value && skyboxmo[0])
+		Portal_AddSkyboxPortals(rendercontext);
+
+	// Portal rendering. Hijacks the BSP traversal.
+	ps_sw_portaltime[i] = I_GetPreciseTime();
+	if (bspcontext->portal_base)
+	{
+		portal_t *portal = bspcontext->portal_base;
+
+		for (; portal; portal = bspcontext->portal_base)
+		{
+			bspcontext->portalrender = portal->pass; // Recursiveness depth.
+
+			R_ClearFFloorClips(planecontext);
+
+			// Apply the viewpoint stored for the portal.
+			R_PortalFrame(rendercontext, portal);
+
+			// Hack in the clipsegs to delimit the starting
+			// clipping for sprites and possibly other similar
+			// future items.
+			R_ClearClipSegs(bspcontext, portal->start, portal->end);
+
+			// Hack in the top/bottom clip values for the window
+			// that were previously stored.
+			Portal_ClipApply(planecontext, portal);
+
+			masks = realloc(masks, (++nummasks)*sizeof(maskcount_t));
+
+			Mask_Pre(rendercontext, &masks[nummasks - 1]);
+			bspcontext->curdrawsegs = bspcontext->ds_p;
+
+			// Render the BSP from the new viewpoint, and clip
+			// any sprites with the new clipsegs and window.
+			R_RenderBSPNode(rendercontext, (INT32)numnodes - 1);
+			Mask_Post(rendercontext, &masks[nummasks - 1]);
+
+			R_ClipSprites(rendercontext, bspcontext->ds_p - (masks[nummasks - 1].drawsegs[1] - masks[nummasks - 1].drawsegs[0]), portal);
+
+			Portal_Remove(bspcontext, portal);
+		}
+	}
+	ps_sw_portaltime[i] = I_GetPreciseTime() - ps_sw_portaltime[i];
+
+	ps_sw_planetime[i] = I_GetPreciseTime();
+	R_DrawPlanes(rendercontext);
+	ps_sw_planetime[i] = I_GetPreciseTime() - ps_sw_planetime[i];
+
+	// draw mid texture and sprite
+	// And now 3D floors/sides!
+	ps_sw_maskedtime[i] = I_GetPreciseTime();
+	R_DrawMasked(rendercontext, masks, nummasks);
+	ps_sw_maskedtime[i] = I_GetPreciseTime() - ps_sw_maskedtime[i];
+
+	free(masks);
+
+#ifdef HAVE_THREADS
+	if (renderthreaded)
+	{
+		INT32 x = rendercontext->begincolumn;
+		INT32 w = (rendercontext->endcolumn - rendercontext->begincolumn);
+		INT32 c = 35 | V_NOSCALESTART;
+
+		V_DrawFill(x, 0, w, 1, c);
+		V_DrawFill(x, viewheight - 1, w, 1, c);
+
+		V_DrawFill(x, 0, 1, viewheight, c);
+		V_DrawFill(x+w-1, 0, 1, viewheight, c);
+	}
+#endif
+}
+
+static void R_ResetContext(rendercontext_t *context, INT32 leftclip, INT32 rightclip)
+{
+	context->begincolumn = leftclip;
+	context->endcolumn = rightclip;
+
+	// Clear buffers.
+	R_ClearPlanes(&context->planecontext);
+
+	if (viewmorph.use)
+	{
+		INT32 left = max(leftclip, viewmorph.x1);
+		INT32 right = min(viewwidth-viewmorph.x1-1, rightclip);
+
+		context->bspcontext.portalclipstart = left;
+		context->bspcontext.portalclipend = right;
+
+		R_ClearClipSegs(&context->bspcontext, left, right);
+
+		memcpy(context->planecontext.ceilingclip, viewmorph.ceilingclip, sizeof(INT16)*vid.width);
+		memcpy(context->planecontext.floorclip, viewmorph.floorclip, sizeof(INT16)*vid.width);
+	}
+	else
+	{
+		context->bspcontext.portalclipstart = leftclip;
+		context->bspcontext.portalclipend = rightclip;
+
+		R_ClearClipSegs(&context->bspcontext, leftclip, rightclip);
+	}
+
+	R_ClearDrawSegs(&context->bspcontext);
+	R_ClearSprites(&context->spritecontext);
+
+	Portal_InitList(&context->bspcontext);
+}
+
+#ifdef HAVE_THREADS
+static void R_ContextThreadFunc(void *userdata)
+{
+	renderdata_t *data = (renderdata_t *)userdata;
+
+	I_atomic_exchange(&data->running, 1);
+
+	while (!I_atomic_load(&data->shouldquit))
+	{
+		if (I_atomic_exchange(&data->framewaiting, 0))
+		{
+			R_RenderViewContext(&data->context);
+			I_atomic_exchange(&data->framefinished, 1);
+		}
+	}
+
+	I_atomic_exchange(&data->running, 0);
+}
+#endif
+
+static void R_StartThreads(void)
+{
+#ifdef HAVE_THREADS
+	INT32 currcontext;
+
+	for (currcontext = 0; currcontext < numusablerendercontexts; ++currcontext)
+	{
+		renderdata_t *renderdata = &renderdatas[currcontext];
+		rendercontext_t *context = &renderdata->context;
+
+		context->colcontext.func = colfuncs[BASEDRAWFUNC];
+		context->spancontext.func = spanfuncs[BASEDRAWFUNC];
+
+		I_atomic_exchange(&renderdata->running, 0);
+		I_atomic_exchange(&renderdata->shouldquit, 0);
+		I_atomic_exchange(&renderdata->framewaiting, 0);
+		I_atomic_exchange(&renderdata->framefinished, 0);
+
+		renderdata->thread = I_spawn_thread("software-renderer", (I_thread_fn)R_ContextThreadFunc, context);
+	}
+#endif
+}
+
+void R_StopThreads(void)
+{
+#ifdef HAVE_THREADS
+	INT32 currcontext;
+
+	if (!numusablerendercontexts)
+		return;
+
+	for (currcontext = 0; currcontext < numusablerendercontexts; ++currcontext)
+		I_atomic_exchange(&renderdatas[currcontext].shouldquit, 1);
+
+	for (;;) {
+		INT32 stoppedthreads = 0;
+
+		for (currcontext = 0; currcontext < numusablerendercontexts; ++currcontext)
+			stoppedthreads += !I_atomic_load(&renderdatas[currcontext].running);
+
+		if (stoppedthreads == numusablerendercontexts)
+			break;
+	}
+
+	numusablerendercontexts = 0;
+#endif
+}
+
+static void R_InitContexts(void)
+{
+	INT32 currcontext;
+	INT32 currstart;
+	INT32 incrementby;
+
+	currstart = 0;
+	incrementby = vid.width / numrendercontexts;
+
+	renderdatas = Z_Calloc(sizeof(renderdata_t) * numrendercontexts, PU_STATIC, NULL);
+
+	for (currcontext = 0; currcontext < numrendercontexts; ++currcontext)
+	{
+		renderdata_t *renderdata = &renderdatas[currcontext];
+		rendercontext_t *context = &renderdata->context;
+
+		context->num = currcontext;
+		context->begincolumn = max(currstart, 0);
+		currstart += incrementby;
+		context->endcolumn = min(currstart, vid.width);
+
+		context->planecontext.freehead = &context->planecontext.freetail;
+
+		R_ResetContext(&renderdata->context, context->begincolumn, context->endcolumn);
+		R_InitDrawNodes(&renderdata->context.spritecontext.nodebankhead);
+
+		context->colcontext.func = colfuncs[BASEDRAWFUNC];
+		context->spancontext.func = spanfuncs[BASEDRAWFUNC];
+	}
+}
 
 //
 // R_SetViewSize
@@ -960,7 +1229,7 @@ void R_ExecuteSetViewSize(void)
 			yslopetab[i] = FixedDiv(centerx*FRACUNIT, dy);
 		}
 
-		if (ds_su)
+		/*if (ds_su)
 			Z_Free(ds_su);
 		if (ds_sv)
 			Z_Free(ds_sv);
@@ -968,7 +1237,7 @@ void R_ExecuteSetViewSize(void)
 			Z_Free(ds_sz);
 
 		ds_su = ds_sv = ds_sz = NULL;
-		ds_sup = ds_svp = ds_szp = NULL;
+		ds_sup = ds_svp = ds_szp = NULL;*/
 	}
 
 	memset(scalelight, 0xFF, sizeof(scalelight));
@@ -1024,7 +1293,7 @@ void R_Init(void)
 	//I_OutputMsg("\nR_InitTranslucencyTables\n");
 	R_InitTranslucencyTables();
 
-	R_InitDrawNodes();
+	R_InitContexts();
 
 	framecount = 0;
 }
@@ -1394,63 +1663,57 @@ boolean R_IsViewpointThirdPerson(player_t *player, boolean skybox)
 	return false;
 }
 
-static void R_PortalFrame(portal_t *portal)
-{
-	viewx = portal->viewx;
-	viewy = portal->viewy;
-	viewz = portal->viewz;
-
-	viewangle = portal->viewangle;
-	viewsin = FINESINE(viewangle>>ANGLETOFINESHIFT);
-	viewcos = FINECOSINE(viewangle>>ANGLETOFINESHIFT);
-
-	portalclipstart = portal->start;
-	portalclipend = portal->end;
-
-	if (portal->clipline != -1)
-	{
-		portalclipline = &lines[portal->clipline];
-		portalcullsector = portalclipline->frontsector;
-		viewsector = portalclipline->frontsector;
-	}
-	else
-	{
-		portalclipline = NULL;
-		portalcullsector = NULL;
-		viewsector = R_PointInSubsector(viewx, viewy)->sector;
-	}
-}
-
-static void Mask_Pre (maskcount_t* m)
-{
-	m->drawsegs[0] = ds_p - drawsegs;
-	m->vissprites[0] = visspritecount;
-	m->viewx = viewx;
-	m->viewy = viewy;
-	m->viewz = viewz;
-	m->viewsector = viewsector;
-}
-
-static void Mask_Post (maskcount_t* m)
-{
-	m->drawsegs[1] = ds_p - drawsegs;
-	m->vissprites[1] = visspritecount;
-}
-
 // ================
 // R_RenderView
 // ================
 
-//                     FAB NOTE FOR WIN32 PORT !! I'm not finished already,
+static void R_PrepareViewContext(viewcontext_t *context)
-// but I suspect network may have problems with the video buffer being locked
+{
-// for all duration of rendering, and being released only once at the end..
+	context->x = viewx;
-// I mean, there is a win16lock() or something that lasts all the rendering,
+	context->y = viewy;
-// so maybe we should release screen lock before each netupdate below..?
+	context->z = viewz;
+
+	context->angle = viewangle;
+	context->sin = viewsin;
+	context->cos = viewcos;
+
+	context->sector = viewsector;
+	context->player = viewplayer;
+	context->mobj = r_viewmobj;
+}
+
+static void R_PrepareContexts(void)
+{
+	INT32 currcontext;
+	INT32 currstart = 0;
+	INT32 desiredwidth = viewwidth / numusablerendercontexts;
+	INT32 i;
+
+	for (currcontext = 0; currcontext < numusablerendercontexts; ++currcontext)
+	{
+		rendercontext_t *context = &renderdatas[currcontext].context;
+
+		context->buffer.screens[0] = topleft;
+		for (i = 1; i < NUMSCREENS; i++)
+			context->buffer.screens[i] = screens[i];
+
+		context->begincolumn = max(currstart, 0);
+		currstart += desiredwidth;
+		context->endcolumn = min(currstart, viewwidth);
+
+		R_ResetContext(context, context->begincolumn, context->endcolumn);
+		R_PrepareViewContext(&context->viewcontext);
+
+		if (context->spritecontext.sectorvisited == NULL)
+			context->spritecontext.sectorvisited = Z_Calloc(sizeof(boolean) * numsectors, PU_LEVEL, NULL);
+		memset(context->spritecontext.sectorvisited, 0, sizeof(boolean) * numsectors);
+	}
+}
 
 void R_RenderPlayerView(player_t *player)
 {
-	UINT8			nummasks	= 1;
+	INT32 currcontext;
-	maskcount_t*	masks		= malloc(sizeof(maskcount_t));
+	INT32 finishedcontexts = 0;
 
 	if (cv_homremoval.value && player == &players[displayplayer]) // if this is display player 1
 	{
@@ -1462,117 +1725,49 @@ void R_RenderPlayerView(player_t *player)
 
 	R_SetupFrame(player);
 	framecount++;
-	validcount++;
-
-	// Clear buffers.
-	R_ClearPlanes();
-	if (viewmorph.use)
-	{
-		portalclipstart = viewmorph.x1;
-		portalclipend = viewwidth-viewmorph.x1-1;
-		R_PortalClearClipSegs(portalclipstart, portalclipend);
-		memcpy(ceilingclip, viewmorph.ceilingclip, sizeof(INT16)*vid.width);
-		memcpy(floorclip, viewmorph.floorclip, sizeof(INT16)*vid.width);
-	}
-	else
-	{
-		portalclipstart = 0;
-		portalclipend = viewwidth;
-		R_ClearClipSegs();
-	}
-	R_ClearDrawSegs();
-	R_ClearSprites();
-	Portal_InitList();
 
 	// check for new console commands.
 	NetUpdate();
 
-	// The head node is the last node output.
+#ifdef HAVE_THREADS
-
+	if (cv_numthreads.value != numusablerendercontexts)
-	Mask_Pre(&masks[nummasks - 1]);
-	curdrawsegs = ds_p;
-//profile stuff ---------------------------------------------------------
-#ifdef TIMING
-	mytotal = 0;
-	ProfZeroTimer();
-#endif
-	ps_numbspcalls = ps_numpolyobjects = ps_numdrawnodes = 0;
-	ps_bsptime = I_GetPreciseTime();
-	R_RenderBSPNode((INT32)numnodes - 1);
-	ps_bsptime = I_GetPreciseTime() - ps_bsptime;
-	ps_numsprites = visspritecount;
-#ifdef TIMING
-	RDMSR(0x10, &mycount);
-	mytotal += mycount; // 64bit add
-
-	CONS_Debug(DBG_RENDER, "RenderBSPNode: 0x%d %d\n", *((INT32 *)&mytotal + 1), (INT32)mytotal);
-#endif
-//profile stuff ---------------------------------------------------------
-	Mask_Post(&masks[nummasks - 1]);
-
-	ps_sw_spritecliptime = I_GetPreciseTime();
-	R_ClipSprites(drawsegs, NULL);
-	ps_sw_spritecliptime = I_GetPreciseTime() - ps_sw_spritecliptime;
-
-
-	// Add skybox portals caused by sky visplanes.
-	if (cv_skybox.value && skyboxmo[0])
-		Portal_AddSkyboxPortals();
-
-	// Portal rendering. Hijacks the BSP traversal.
-	ps_sw_portaltime = I_GetPreciseTime();
-	if (portal_base)
 	{
-		portal_t *portal;
+		if (renderthreaded)
+			R_StopThreads();
 
-		for(portal = portal_base; portal; portal = portal_base)
+		numusablerendercontexts = cv_numthreads.value;
+		renderthreaded = (numusablerendercontexts > 1);
+
+		if (renderthreaded)
+			R_StartThreads();
+	}
+#else
+	numusablerendercontexts = 1;
+	renderthreaded = false;
+#endif
+
+	R_PrepareContexts();
+
+	for (currcontext = 0; currcontext < numusablerendercontexts; currcontext++)
 	{
-			portalrender = portal->pass; // Recursiveness depth.
+#ifdef HAVE_THREADS
-
+		if (renderthreaded)
-			R_ClearFFloorClips();
+			I_atomic_exchange(&renderdatas[currcontext].framewaiting, 1);
-
+		else
-			// Apply the viewpoint stored for the portal.
+#endif
-			R_PortalFrame(portal);
+		{
-
+			R_RenderViewContext(&renderdatas[currcontext].context);
-			// Hack in the clipsegs to delimit the starting
+			finishedcontexts++;
-			// clipping for sprites and possibly other similar
-			// future items.
-			R_PortalClearClipSegs(portal->start, portal->end);
-
-			// Hack in the top/bottom clip values for the window
-			// that were previously stored.
-			Portal_ClipApply(portal);
-
-			validcount++;
-
-			masks = realloc(masks, (++nummasks)*sizeof(maskcount_t));
-
-			Mask_Pre(&masks[nummasks - 1]);
-			curdrawsegs = ds_p;
-
-			// Render the BSP from the new viewpoint, and clip
-			// any sprites with the new clipsegs and window.
-			R_RenderBSPNode((INT32)numnodes - 1);
-			Mask_Post(&masks[nummasks - 1]);
-
-			R_ClipSprites(ds_p - (masks[nummasks - 1].drawsegs[1] - masks[nummasks - 1].drawsegs[0]), portal);
-
-			Portal_Remove(portal);
 		}
 	}
-	ps_sw_portaltime = I_GetPreciseTime() - ps_sw_portaltime;
 
-	ps_sw_planetime = I_GetPreciseTime();
+#ifdef HAVE_THREADS
-	R_DrawPlanes();
+	while (renderthreaded && finishedcontexts != numusablerendercontexts)
-	ps_sw_planetime = I_GetPreciseTime() - ps_sw_planetime;
+	{
-
+		for (currcontext = 0; currcontext < numusablerendercontexts; currcontext++)
-	// draw mid texture and sprite
+			finishedcontexts += I_atomic_exchange(&renderdatas[currcontext].framefinished, 0);
-	// And now 3D floors/sides!
+	}
-	ps_sw_maskedtime = I_GetPreciseTime();
+#endif
-	R_DrawMasked(masks, nummasks);
-	ps_sw_maskedtime = I_GetPreciseTime() - ps_sw_maskedtime;
-
-	free(masks);
 }
 
 // =========================================================================
@@ -1639,6 +1834,7 @@ void R_RegisterEngineStuff(void)
 	CV_RegisterVar(&cv_translucenthud);
 
 	CV_RegisterVar(&cv_maxportals);
+	CV_RegisterVar(&cv_numthreads);
 
 	CV_RegisterVar(&cv_movebob);
 }

src/r_main.h

@@ -29,6 +29,10 @@ extern fixed_t centerxfrac, centeryfrac;
 extern fixed_t projection, projectiony;
 extern fixed_t fovtan;
 
+extern INT32 numrendercontexts;
+extern INT32 numusablerendercontexts;
+extern boolean renderthreaded;
+
 // WARNING: a should be unsigned but to add with 2048, it isn't!
 #define AIMINGTODY(a) FixedDiv((FINETANGENT((2048+(((INT32)a)>>ANGLETOFINESHIFT)) & FINEMASK)*160), fovtan)
 
@@ -70,7 +74,6 @@ angle_t R_PointToAngleEx(INT32 x2, INT32 y2, INT32 x1, INT32 y1);
 fixed_t R_PointToDist(fixed_t x, fixed_t y);
 fixed_t R_PointToDist2(fixed_t px2, fixed_t py2, fixed_t px1, fixed_t py1);
 
-fixed_t R_ScaleFromGlobalAngle(angle_t visangle);
 subsector_t *R_PointInSubsector(fixed_t x, fixed_t y);
 subsector_t *R_PointInSubsectorOrNull(fixed_t x, fixed_t y);
 
@@ -80,20 +83,21 @@ boolean R_DoCulling(line_t *cullheight, line_t *viewcullheight, fixed_t vz, fixe
 
 extern precise_t ps_prevframetime; // time when previous frame was rendered
 extern precise_t ps_rendercalltime;
+extern precise_t ps_postprocesstime;
 extern precise_t ps_uitime;
 extern precise_t ps_swaptime;
 
-extern precise_t ps_bsptime;
+extern precise_t ps_bsptime[MAX_RENDER_THREADS];
 
-extern precise_t ps_sw_spritecliptime;
+extern precise_t ps_sw_spritecliptime[MAX_RENDER_THREADS];
-extern precise_t ps_sw_portaltime;
+extern precise_t ps_sw_portaltime[MAX_RENDER_THREADS];
-extern precise_t ps_sw_planetime;
+extern precise_t ps_sw_planetime[MAX_RENDER_THREADS];
-extern precise_t ps_sw_maskedtime;
+extern precise_t ps_sw_maskedtime[MAX_RENDER_THREADS];
 
-extern int ps_numbspcalls;
+extern int ps_numbspcalls[MAX_RENDER_THREADS];
-extern int ps_numsprites;
+extern int ps_numsprites[MAX_RENDER_THREADS];
-extern int ps_numdrawnodes;
+extern int ps_numdrawnodes[MAX_RENDER_THREADS];
-extern int ps_numpolyobjects;
+extern int ps_numpolyobjects[MAX_RENDER_THREADS];
 
 //
 // REFRESH - the actual rendering functions.
@@ -131,6 +135,8 @@ void R_SkyboxFrame(player_t *player);
 boolean R_ViewpointHasChasecam(player_t *player);
 boolean R_IsViewpointThirdPerson(player_t *player, boolean skybox);
 
+void R_StopThreads(void);
+
 // Called by D_Display.
 void R_RenderPlayerView(player_t *player);
 

src/r_plane.h

@@ -55,52 +55,9 @@ typedef struct visplane_s
 	pslope_t *slope;
 } visplane_t;
 
-extern visplane_t *visplanes[MAXVISPLANES];
-extern visplane_t *floorplane;
-extern visplane_t *ceilingplane;
-
-// Visplane related.
-extern INT16 *lastopening, *openings;
-extern size_t maxopenings;
-
-extern INT16 floorclip[MAXVIDWIDTH], ceilingclip[MAXVIDWIDTH];
-extern fixed_t frontscale[MAXVIDWIDTH], yslopetab[MAXVIDHEIGHT*16];
-extern fixed_t cachedheight[MAXVIDHEIGHT];
-extern fixed_t cacheddistance[MAXVIDHEIGHT];
-extern fixed_t cachedxstep[MAXVIDHEIGHT];
-extern fixed_t cachedystep[MAXVIDHEIGHT];
-
-extern fixed_t *yslope;
-extern lighttable_t **planezlight;
-
-void R_InitPlanes(void);
-void R_ClearPlanes(void);
-void R_ClearFFloorClips (void);
-
-void R_DrawPlanes(void);
-visplane_t *R_FindPlane(fixed_t height, INT32 picnum, INT32 lightlevel, fixed_t xoff, fixed_t yoff, angle_t plangle,
-	extracolormap_t *planecolormap, ffloor_t *ffloor, polyobj_t *polyobj, pslope_t *slope);
-visplane_t *R_CheckPlane(visplane_t *pl, INT32 start, INT32 stop);
-void R_ExpandPlane(visplane_t *pl, INT32 start, INT32 stop);
-void R_PlaneBounds(visplane_t *plane);
-
-void R_CheckFlatLength(size_t size);
-boolean R_CheckPowersOfTwo(void);
-
-// Draws a single visplane.
-void R_DrawSinglePlane(visplane_t *pl);
-
-// Calculates the slope vectors needed for tilted span drawing.
-void R_SetSlopePlane(pslope_t *slope, fixed_t xpos, fixed_t ypos, fixed_t zpos, fixed_t xoff, fixed_t yoff, angle_t angle, angle_t plangle);
-void R_SetScaledSlopePlane(pslope_t *slope, fixed_t xpos, fixed_t ypos, fixed_t zpos, fixed_t xs, fixed_t ys, fixed_t xoff, fixed_t yoff, angle_t angle, angle_t plangle);
-void R_CalculateSlopeVectors(void);
-
-// Sets the slope vector pointers for the current tilted span.
-void R_SetTiltedSpan(INT32 span);
-
 typedef struct planemgr_s
 {
-	visplane_t *plane;
+	struct visplane_s *plane;
 	fixed_t height;
 	fixed_t f_pos; // F for Front sector
 	fixed_t b_pos; // B for Back sector
@@ -117,10 +74,36 @@ typedef struct planemgr_s
 
 	struct ffloor_s *ffloor;
 	polyobj_t *polyobj;
-} visffloor_t;
+} planemgr_t;
 
-extern visffloor_t ffloor[MAXFFLOORS];
+// Visplane related.
-extern INT32 numffloors;
+extern fixed_t yslopetab[MAXVIDHEIGHT*16];
+extern fixed_t *yslope;
 
-void Portal_AddSkyboxPortals (void);
+struct planecontext_s;
+struct rendercontext_s;
+struct viewcontext_s;
+
+void R_InitPlanes(void);
+void R_ClearPlanes(struct planecontext_s *context);
+void R_ClearFFloorClips (struct planecontext_s *context);
+
+void R_DrawPlanes(struct rendercontext_s *planecontext);
+visplane_t *R_FindPlane(struct planecontext_s *planecontext, struct viewcontext_s *viewcontext,
+	fixed_t height, INT32 picnum, INT32 lightlevel, fixed_t xoff, fixed_t yoff, angle_t plangle,
+	extracolormap_t *planecolormap, ffloor_t *ffloor, polyobj_t *polyobj, pslope_t *slope);
+visplane_t *R_CheckPlane(struct planecontext_s *planecontext, visplane_t *pl, INT32 start, INT32 stop);
+void R_ExpandPlane(visplane_t *pl, INT32 start, INT32 stop);
+void R_PlaneBounds(visplane_t *plane);
+
+// Draws a single visplane.
+void R_DrawSinglePlane(struct rendercontext_s *rendercontext, visplane_t *pl);
+
+// Calculates the slope vectors needed for tilted span drawing.
+void R_SetSlopePlane(pslope_t *slope, spancontext_t *ds, fixed_t xpos, fixed_t ypos, fixed_t zpos, fixed_t xoff, fixed_t yoff, angle_t angle, angle_t plangle);
+void R_SetScaledSlopePlane(pslope_t *slope, spancontext_t *ds, fixed_t xpos, fixed_t ypos, fixed_t zpos, fixed_t xs, fixed_t ys, fixed_t xoff, fixed_t yoff, angle_t angle, angle_t plangle);
+void R_CalculateSlopeVectors(spancontext_t *ds);
+
+// Sets the slope vector pointers for the current tilted span.
+void R_SetTiltedSpan(spancontext_t *ds, INT32 span);
 #endif

src/r_portal.c

@@ -14,27 +14,17 @@
 #include "r_portal.h"
 #include "r_plane.h"
 #include "r_main.h"
+#include "r_context.h"
 #include "doomstat.h"
 #include "p_spec.h" // Skybox viewpoints
 #include "z_zone.h"
 #include "r_things.h"
 #include "r_sky.h"
 
-UINT8 portalrender;			/**< When rendering a portal, it establishes the depth of the current BSP traversal. */
+void Portal_InitList (bspcontext_t *bspcontext)
-
-// Linked list for portals.
-portal_t *portal_base, *portal_cap;
-
-line_t *portalclipline;
-sector_t *portalcullsector;
-INT32 portalclipstart, portalclipend;
-
-boolean portalline; // is curline a portal seg?
-
-void Portal_InitList (void)
 {
-	portalrender = 0;
+	bspcontext->portalrender = 0;
-	portal_base = portal_cap = NULL;
+	bspcontext->portal_base = bspcontext->portal_cap = NULL;
 }
 
 /** Store the clipping window for a portal in its given range.
@@ -43,7 +33,7 @@ void Portal_InitList (void)
  * the function is called, so it is useful for converting one-sided
  * lines into portals.
  */
-void Portal_ClipRange (portal_t* portal)
+void Portal_ClipRange (planecontext_t *planecontext, portal_t* portal)
 {
 	INT32 start	= portal->start;
 	INT32 end	= portal->end;
@@ -54,18 +44,18 @@ void Portal_ClipRange (portal_t* portal)
 	INT32 i;
 	for (i = 0; i < end-start; i++)
 	{
-		*ceil = ceilingclip[start+i];
+		*ceil = planecontext->ceilingclip[start+i];
 		ceil++;
-		*floor = floorclip[start+i];
+		*floor = planecontext->floorclip[start+i];
 		floor++;
-		*scale = frontscale[start+i];
+		*scale = planecontext->frontscale[start+i];
 		scale++;
 	}
 }
 
 /** Apply the clipping window from a portal.
  */
-void Portal_ClipApply (const portal_t* portal)
+void Portal_ClipApply (planecontext_t *planecontext, const portal_t* portal)
 {
 	INT32 i;
 	INT32 start	= portal->start;
@@ -76,28 +66,28 @@ void Portal_ClipApply (const portal_t* portal)
 
 	for (i = 0; i < end-start; i++)
 	{
-		ceilingclip[start+i] = *ceil;
+		planecontext->ceilingclip[start+i] = *ceil;
 		ceil++;
-		floorclip[start+i] = *floor;
+		planecontext->floorclip[start+i] = *floor;
 		floor++;
-		frontscale[start+i] = *scale;
+		planecontext->frontscale[start+i] = *scale;
 		scale++;
 	}
 
 	// HACKS FOLLOW
 	for (i = 0; i < start; i++)
 	{
-		floorclip[i] = -1;
+		planecontext->floorclip[i] = -1;
-		ceilingclip[i] = (INT16)viewheight;
+		planecontext->ceilingclip[i] = (INT16)viewheight;
 	}
 	for (i = end; i < vid.width; i++)
 	{
-		floorclip[i] = -1;
+		planecontext->floorclip[i] = -1;
-		ceilingclip[i] = (INT16)viewheight;
+		planecontext->ceilingclip[i] = (INT16)viewheight;
 	}
 }
 
-static portal_t* Portal_Add (const INT16 x1, const INT16 x2)
+static portal_t* Portal_Add (bspcontext_t *context, const INT16 x1, const INT16 x2)
 {
 	portal_t *portal		= Z_Malloc(sizeof(portal_t), PU_LEVEL, NULL);
 	INT16 *ceilingclipsave	= Z_Malloc(sizeof(INT16)*(x2-x1 + 1), PU_LEVEL, NULL);
@@ -105,15 +95,15 @@ static portal_t* Portal_Add (const INT16 x1, const INT16 x2)
 	fixed_t *frontscalesave	= Z_Malloc(sizeof(fixed_t)*(x2-x1 + 1), PU_LEVEL, NULL);
 
 	// Linked list.
-	if (!portal_base)
+	if (!context->portal_base)
 	{
-		portal_base	= portal;
+		context->portal_base = portal;
-		portal_cap	= portal;
+		context->portal_cap = portal;
 	}
 	else
 	{
-		portal_cap->next = portal;
+		context->portal_cap->next = portal;
-		portal_cap = portal;
+		context->portal_cap = portal;
 	}
 	portal->next = NULL;
 
@@ -125,14 +115,14 @@ static portal_t* Portal_Add (const INT16 x1, const INT16 x2)
 	portal->end		= x2;
 
 	// Increase recursion level.
-	portal->pass = portalrender+1;
+	portal->pass = context->portalrender+1;
 
 	return portal;
 }
 
-void Portal_Remove (portal_t* portal)
+void Portal_Remove (bspcontext_t *context, portal_t* portal)
 {
-	portal_base = portal->next;
+	context->portal_base = portal->next;
 	Z_Free(portal->ceilingclip);
 	Z_Free(portal->floorclip);
 	Z_Free(portal->frontscale);
@@ -149,9 +139,10 @@ void Portal_Remove (portal_t* portal)
  * When the portal renders, it will create the illusion of
  * the two lines being seamed together.
  */
-void Portal_Add2Lines (const INT32 line1, const INT32 line2, const INT32 x1, const INT32 x2)
+void Portal_Add2Lines (rendercontext_t *context,
+    const INT32 line1, const INT32 line2, const INT32 x1, const INT32 x2)
 {
-	portal_t* portal = Portal_Add(x1, x2);
+	portal_t* portal = Portal_Add(&context->bspcontext, x1, x2);
 
 	// Offset the portal view by the linedef centers
 	line_t* start	= &lines[line1];
@@ -172,20 +163,20 @@ void Portal_Add2Lines (const INT32 line1, const INT32 line2, const INT32 x1, con
 	dest_c.x = (dest->v1->x + dest->v2->x) / 2;
 	dest_c.y = (dest->v1->y + dest->v2->y) / 2;
 
-	disttopoint = R_PointToDist2(start_c.x, start_c.y, viewx, viewy);
+	disttopoint = R_PointToDist2(start_c.x, start_c.y, context->viewcontext.x, context->viewcontext.y);
-	angtopoint = R_PointToAngle2(start_c.x, start_c.y, viewx, viewy);
+	angtopoint = R_PointToAngle2(start_c.x, start_c.y, context->viewcontext.x, context->viewcontext.y);
 	angtopoint += dangle;
 
 	portal->viewx = dest_c.x + FixedMul(FINECOSINE(angtopoint>>ANGLETOFINESHIFT), disttopoint);
 	portal->viewy = dest_c.y + FixedMul(FINESINE(angtopoint>>ANGLETOFINESHIFT), disttopoint);
-	portal->viewz = viewz + dest->frontsector->floorheight - start->frontsector->floorheight;
+	portal->viewz = context->viewcontext.z + dest->frontsector->floorheight - start->frontsector->floorheight;
-	portal->viewangle = viewangle + dangle;
+	portal->viewangle = context->viewcontext.angle + dangle;
 
 	portal->clipline = line2;
 
-	Portal_ClipRange(portal);
+	Portal_ClipRange(&context->planecontext, portal);
 
-	portalline = true; // this tells R_StoreWallRange that curline is a portal seg
+	context->bspcontext.portalline = true; // this tells R_StoreWallRange that curline is a portal seg
 }
 
 /** Store the clipping window for a portal using a visplane.
@@ -256,7 +247,7 @@ static boolean TrimVisplaneBounds (const visplane_t* plane, INT16* start, INT16*
  * Applies the necessary offsets and rotation to give
  * a depth illusion to the skybox.
  */
-void Portal_AddSkybox (const visplane_t* plane)
+void Portal_AddSkybox (bspcontext_t *bspcontext, viewcontext_t *viewcontext, const visplane_t* plane)
 {
 	INT16 start, end;
 	mapheader_t *mh;
@@ -265,14 +256,14 @@ void Portal_AddSkybox (const visplane_t* plane)
 	if (TrimVisplaneBounds(plane, &start, &end))
 		return;
 
-	portal = Portal_Add(start, end);
+	portal = Portal_Add(bspcontext, start, end);
 
 	Portal_ClipVisplane(plane, portal);
 
 	portal->viewx = skyboxmo[0]->x;
 	portal->viewy = skyboxmo[0]->y;
 	portal->viewz = skyboxmo[0]->z;
-	portal->viewangle = viewangle + skyboxmo[0]->angle;
+	portal->viewangle = viewcontext->angle + skyboxmo[0]->angle;
 
 	mh = mapheaderinfo[gamemap-1];
 
@@ -283,14 +274,14 @@ void Portal_AddSkybox (const visplane_t* plane)
 		angle_t ang = skyboxmo[0]->angle>>ANGLETOFINESHIFT;
 
 		if (mh->skybox_scalex > 0)
-			x = (viewx - skyboxmo[1]->x) / mh->skybox_scalex;
+			x = (viewcontext->x - skyboxmo[1]->x) / mh->skybox_scalex;
 		else if (mh->skybox_scalex < 0)
-			x = (viewx - skyboxmo[1]->x) * -mh->skybox_scalex;
+			x = (viewcontext->x - skyboxmo[1]->x) * -mh->skybox_scalex;
 
 		if (mh->skybox_scaley > 0)
-			y = (viewy - skyboxmo[1]->y) / mh->skybox_scaley;
+			y = (viewcontext->y - skyboxmo[1]->y) / mh->skybox_scaley;
 		else if (mh->skybox_scaley < 0)
-			y = (viewy - skyboxmo[1]->y) * -mh->skybox_scaley;
+			y = (viewcontext->y - skyboxmo[1]->y) * -mh->skybox_scaley;
 
 		// Apply transform to account for the skybox viewport angle.
 		portal->viewx += FixedMul(x,FINECOSINE(ang)) - FixedMul(y,  FINESINE(ang));
@@ -298,9 +289,9 @@ void Portal_AddSkybox (const visplane_t* plane)
 	}
 
 	if (mh->skybox_scalez > 0)
-		portal->viewz += viewz / mh->skybox_scalez;
+		portal->viewz += viewcontext->z / mh->skybox_scalez;
 	else if (mh->skybox_scalez < 0)
-		portal->viewz += viewz * -mh->skybox_scalez;
+		portal->viewz += viewcontext->z * -mh->skybox_scalez;
 
 	portal->clipline = -1;
 }
@@ -308,7 +299,7 @@ void Portal_AddSkybox (const visplane_t* plane)
 /** Creates portals for the currently existing sky visplanes.
  * The visplanes are also removed and cleared from the list.
  */
-void Portal_AddSkyboxPortals (void)
+void Portal_AddSkyboxPortals (rendercontext_t *context)
 {
 	visplane_t *pl;
 	INT32 i;
@@ -316,11 +307,11 @@ void Portal_AddSkyboxPortals (void)
 
 	for (i = 0; i < MAXVISPLANES; i++, pl++)
 	{
-		for (pl = visplanes[i]; pl; pl = pl->next)
+		for (pl = context->planecontext.visplanes[i]; pl; pl = pl->next)
 		{
 			if (pl->picnum == skyflatnum)
 			{
-				Portal_AddSkybox(pl);
+				Portal_AddSkybox(&context->bspcontext, &context->viewcontext, pl);
 
 				pl->minx = 0;
 				pl->maxx = -1;

src/r_portal.h

@@ -41,21 +41,19 @@ typedef struct portal_s
 	fixed_t *frontscale;/**< Temporary screen bottom clipping array. */
 } portal_t;
 
-extern portal_t* portal_base;
+struct bspcontext_s;
-extern portal_t* portal_cap;
+struct planecontext_s;
-extern UINT8 portalrender;
+struct rendercontext_s;
+struct viewcontext_s;
 
-extern line_t *portalclipline;
+void Portal_InitList	(struct bspcontext_s *bspcontext);
-extern sector_t *portalcullsector;
+void Portal_Remove		(struct bspcontext_s *context, portal_t* portal);
-extern INT32 portalclipstart, portalclipend;
+void Portal_Add2Lines	(struct rendercontext_s *context,
+                        const INT32 line1, const INT32 line2, const INT32 x1, const INT32 x2);
+void Portal_AddSkybox	(struct bspcontext_s *bspcontext, struct viewcontext_s *viewcontext, const visplane_t* plane);
 
-void Portal_InitList	(void);
+void Portal_ClipRange (struct planecontext_s *planecontext, portal_t* portal);
-void Portal_Remove		(portal_t* portal);
+void Portal_ClipApply (struct planecontext_s *planecontext, const portal_t* portal);
-void Portal_Add2Lines	(const INT32 line1, const INT32 line2, const INT32 x1, const INT32 x2);
-void Portal_AddSkybox	(const visplane_t* plane);
 
-void Portal_ClipRange (portal_t* portal);
+void Portal_AddSkyboxPortals (struct rendercontext_s *context);
-void Portal_ClipApply (const portal_t* portal);
-
-void Portal_AddSkyboxPortals (void);
 #endif

src/r_segs.h

@@ -18,9 +18,12 @@
 #pragma interface
 #endif
 
+struct rendercontext_s;
+struct wallcontext_s;
+
 transnum_t R_GetLinedefTransTable(fixed_t alpha);
-void R_RenderMaskedSegRange(drawseg_t *ds, INT32 x1, INT32 x2);
+void R_RenderMaskedSegRange(struct rendercontext_s *context, drawseg_t *ds, INT32 x1, INT32 x2);
-void R_RenderThickSideRange(drawseg_t *ds, INT32 x1, INT32 x2, ffloor_t *pffloor);
+void R_RenderThickSideRange(struct rendercontext_s *context, drawseg_t *ds, INT32 x1, INT32 x2, ffloor_t *pffloor);
-void R_StoreWallRange(INT32 start, INT32 stop);
+void R_StoreWallRange(struct rendercontext_s *context, struct wallcontext_s *wallcontext, INT32 start, INT32 stop);
 
 #endif

src/r_sky.c

@@ -54,9 +54,6 @@ INT32 globallevelskynum;
 
 	Called at loadlevel after skytexture is set, or when sky texture changes.
 
-	\warning wallcolfunc should be set at R_ExecuteSetViewSize()
-	I don't bother because we don't use low detail anymore
-
 	\return	void
 */
 void R_SetupSkyDraw(void)

src/r_splats.c

@@ -13,37 +13,22 @@
 #include "r_draw.h"
 #include "r_main.h"
 #include "r_splats.h"
+#include "r_context.h"
 #include "r_bsp.h"
 #include "p_local.h"
 #include "p_slopes.h"
 #include "w_wad.h"
 #include "z_zone.h"
 
-struct rastery_s *prastertab; // for ASM code
+static void prepare_rastertab(planecontext_t *planecontext);
-
-static struct rastery_s rastertab[MAXVIDHEIGHT];
-static void prepare_rastertab(void);
 
 // ==========================================================================
 //                                                               FLOOR SPLATS
 // ==========================================================================
 
-static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, vissprite_t *vis);
+static void R_RasterizeFloorSplat(rendercontext_t *context, floorsplat_t *pSplat, vector2_t *verts, vissprite_t *vis);
 
-#ifdef USEASM
+static void rasterize_segment_tex(planecontext_t *planecontext, INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32 tv1, INT32 tv2, INT32 tc, INT32 dir)
-void ASMCALL rasterize_segment_tex_asm(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32 tv1, INT32 tv2, INT32 tc, INT32 dir);
-#endif
-
-static void rasterize_segment_tex(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32 tv1, INT32 tv2, INT32 tc, INT32 dir)
-{
-#ifdef USEASM
-	if (R_ASM)
-	{
-		rasterize_segment_tex_asm(x1, y1, x2, y2, tv1, tv2, tc, dir);
-		return;
-	}
-	else
-#endif
 {
 	fixed_t xs, xe, count;
 	fixed_t dx0, dx1;
@@ -66,9 +51,9 @@ static void rasterize_segment_tex(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32
 		{
 			for (;;)
 			{
-					rastertab[y1].maxx = xs;
+				planecontext->rastertab[y1].maxx = xs;
-					rastertab[y1].tx2 = xe;
+				planecontext->rastertab[y1].tx2 = xe;
-					rastertab[y1].ty2 = tc;
+				planecontext->rastertab[y1].ty2 = tc;
 
 				xs += dx0;
 				xe += dx1;
@@ -81,9 +66,9 @@ static void rasterize_segment_tex(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32
 		{
 			for (;;)
 			{
-					rastertab[y1].maxx = xs;
+				planecontext->rastertab[y1].maxx = xs;
-					rastertab[y1].tx2 = tc;
+				planecontext->rastertab[y1].tx2 = tc;
-					rastertab[y1].ty2 = xe;
+				planecontext->rastertab[y1].ty2 = xe;
 
 				xs += dx0;
 				xe += dx1;
@@ -108,9 +93,9 @@ static void rasterize_segment_tex(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32
 		{
 			for (;;)
 			{
-					rastertab[y2].minx = xs;
+				planecontext->rastertab[y2].minx = xs;
-					rastertab[y2].tx1 = xe;
+				planecontext->rastertab[y2].tx1 = xe;
-					rastertab[y2].ty1 = tc;
+				planecontext->rastertab[y2].ty1 = tc;
 
 				xs += dx0;
 				xe += dx1;
@@ -123,9 +108,9 @@ static void rasterize_segment_tex(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32
 		{
 			for (;;)
 			{
-					rastertab[y2].minx = xs;
+				planecontext->rastertab[y2].minx = xs;
-					rastertab[y2].tx1 = tc;
+				planecontext->rastertab[y2].tx1 = tc;
-					rastertab[y2].ty1 = xe;
+				planecontext->rastertab[y2].ty1 = xe;
 
 				xs += dx0;
 				xe += dx1;
@@ -136,9 +121,8 @@ static void rasterize_segment_tex(INT32 x1, INT32 y1, INT32 x2, INT32 y2, INT32
 		}
 	}
 }
-}
 
-void R_DrawFloorSplat(vissprite_t *spr)
+void R_DrawFloorSplat(rendercontext_t *context, vissprite_t *spr)
 {
 	floorsplat_t splat;
 	mobj_t *mobj = spr->mobj;
@@ -308,7 +292,7 @@ void R_DrawFloorSplat(vissprite_t *spr)
 		v2d[i].y = (centeryfrac + FixedMul(rot_z, yscale))>>FRACBITS;
 	}
 
-	R_RasterizeFloorSplat(&splat, v2d, spr);
+	R_RasterizeFloorSplat(context, &splat, v2d, spr);
 }
 
 // --------------------------------------------------------------------------
@@ -316,7 +300,7 @@ void R_DrawFloorSplat(vissprite_t *spr)
 // fill the polygon with linear interpolation, call span drawer for each
 // scan line
 // --------------------------------------------------------------------------
-static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, vissprite_t *vis)
+static void R_RasterizeFloorSplat(rendercontext_t *context, floorsplat_t *pSplat, vector2_t *verts, vissprite_t *vis)
 {
 	// rasterizing
 	INT32 miny = viewheight + 1, maxy = 0;
@@ -325,9 +309,13 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 	fixed_t planeheight = 0;
 	fixed_t step;
 
+	planecontext_t *planecontext = &context->planecontext;
+
+	colcontext_t *dc = &context->colcontext;
+	spancontext_t *ds = &context->spancontext;
 	int spanfunctype = SPANDRAWFUNC_SPRITE;
 
-	prepare_rastertab();
+	prepare_rastertab(planecontext);
 
 #define RASTERPARAMS(vnum1, vnum2, tv1, tv2, tc, dir) \
     x1 = verts[vnum1].x; \
@@ -372,7 +360,7 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
         } \
         y2 = vid.height - 1; \
     } \
-    rasterize_segment_tex(x1, ry1, x2, y2, tv1, tv2, tc, dir); \
+    rasterize_segment_tex(planecontext, x1, ry1, x2, y2, tv1, tv2, tc, dir); \
     if (ry1 < miny) \
         miny = ry1; \
     if (ry1 > maxy) \
@@ -387,27 +375,27 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 	// do segment d -> left side of texture
 	RASTERPARAMS(0,3,pSplat->width-1,0,0,1);
 
-	ds_source = (UINT8 *)pSplat->pic;
+	ds->source = (UINT8 *)pSplat->pic;
-	ds_flatwidth = pSplat->width;
+	ds->flatwidth = pSplat->width;
-	ds_flatheight = pSplat->height;
+	ds->flatheight = pSplat->height;
 
-	if (R_CheckPowersOfTwo())
+	if (R_CheckPowersOfTwo(ds->flatwidth, ds->flatheight))
-		R_CheckFlatLength(ds_flatwidth * ds_flatheight);
+		R_CheckFlatLength(ds, ds->flatwidth * ds->flatheight);
 
 	if (pSplat->slope)
 	{
-		R_SetTiltedSpan(0);
+		R_SetTiltedSpan(ds, 0);
-		R_SetScaledSlopePlane(pSplat->slope, vis->viewpoint.x, vis->viewpoint.y, vis->viewpoint.z, pSplat->xscale, pSplat->yscale, -pSplat->verts[0].x, pSplat->verts[0].y, vis->viewpoint.angle, pSplat->angle);
+		R_SetScaledSlopePlane(pSplat->slope, ds, vis->viewpoint.x, vis->viewpoint.y, vis->viewpoint.z, pSplat->xscale, pSplat->yscale, -pSplat->verts[0].x, pSplat->verts[0].y, vis->viewpoint.angle, pSplat->angle);
-		R_CalculateSlopeVectors();
+		R_CalculateSlopeVectors(ds);
 		spanfunctype = SPANDRAWFUNC_TILTEDSPRITE;
 	}
 	else
 	{
-		planeheight = abs(pSplat->z - vis->viewpoint.z);
+		planecontext->planeheight = abs(pSplat->z - vis->viewpoint.z);
 
 		if (pSplat->angle)
 		{
-			memset(cachedheight, 0, sizeof(cachedheight));
+			memset(planecontext->cachedheight, 0, sizeof(planecontext->cachedheight));
 
 			// Add the view offset, rotated by the plane angle.
 			fixed_t a = -pSplat->verts[0].x + vis->viewpoint.x;
@@ -423,22 +411,22 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 		}
 	}
 
-	ds_colormap = vis->colormap;
+	ds->colormap = vis->colormap;
-	ds_translation = R_GetSpriteTranslation(vis);
+	ds->translation = R_GetSpriteTranslation(vis);
-	if (ds_translation == NULL)
+	if (ds->translation == NULL)
-		ds_translation = colormaps;
+		ds->translation = colormaps;
 
 	if (vis->extra_colormap)
 	{
-		if (!ds_colormap)
+		if (!ds->colormap)
-			ds_colormap = vis->extra_colormap->colormap;
+			ds->colormap = vis->extra_colormap->colormap;
 		else
-			ds_colormap = &vis->extra_colormap->colormap[ds_colormap - colormaps];
+			ds->colormap = &vis->extra_colormap->colormap[ds->colormap - colormaps];
 	}
 
 	if (vis->transmap)
 	{
-		ds_transmap = vis->transmap;
+		ds->transmap = vis->transmap;
 
 		if (pSplat->slope)
 			spanfunctype = SPANDRAWFUNC_TILTEDTRANSSPRITE;
@@ -446,12 +434,12 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 			spanfunctype = SPANDRAWFUNC_TRANSSPRITE;
 	}
 	else
-		ds_transmap = NULL;
+		ds->transmap = NULL;
 
-	if (ds_powersoftwo)
+	if (ds->powersoftwo)
-		spanfunc = spanfuncs[spanfunctype];
+		ds->func = spanfuncs[spanfunctype];
 	else
-		spanfunc = spanfuncs_npo2[spanfunctype];
+		ds->func = spanfuncs_npo2[spanfunctype];
 
 	if (maxy >= vid.height)
 		maxy = vid.height-1;
@@ -460,8 +448,8 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 	{
 		boolean cliptab[MAXVIDWIDTH+1];
 
-		x1 = rastertab[y].minx>>FRACBITS;
+		x1 = planecontext->rastertab[y].minx>>FRACBITS;
-		x2 = rastertab[y].maxx>>FRACBITS;
+		x2 = planecontext->rastertab[y].maxx>>FRACBITS;
 
 		if (x1 > x2)
 		{
@@ -482,7 +470,7 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 			continue;
 
 		for (i = x1; i <= x2; i++)
-			cliptab[i] = (y >= mfloorclip[i]);
+			cliptab[i] = (y >= dc->mfloorclip[i]);
 
 		// clip left
 		while (cliptab[x1])
@@ -516,10 +504,10 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 			angle_t planecos = FINECOSINE(angle);
 			angle_t planesin = FINESINE(angle);
 
-			if (planeheight != cachedheight[y])
+			if (planecontext->planeheight != planecontext->cachedheight[y])
 			{
-				cachedheight[y] = planeheight;
+				planecontext->cachedheight[y] = planeheight;
-				distance = cacheddistance[y] = FixedMul(planeheight, yslope[y]);
+				distance = planecontext->cacheddistance[y] = FixedMul(planeheight, yslope[y]);
 				span = abs(centery - y);
 
 				if (span) // Don't divide by zero
@@ -530,43 +518,43 @@ static void R_RasterizeFloorSplat(floorsplat_t *pSplat, vector2_t *verts, visspr
 				else
 					xstep = ystep = FRACUNIT;
 
-				cachedxstep[y] = xstep;
+				planecontext->cachedxstep[y] = xstep;
-				cachedystep[y] = ystep;
+				planecontext->cachedystep[y] = ystep;
 			}
 			else
 			{
-				distance = cacheddistance[y];
+				distance = planecontext->cacheddistance[y];
-				xstep = cachedxstep[y];
+				xstep = planecontext->cachedxstep[y];
-				ystep = cachedystep[y];
+				ystep = planecontext->cachedystep[y];
 			}
 
-			ds_xstep = FixedDiv(xstep, pSplat->xscale);
+			ds->xstep = FixedDiv(xstep, pSplat->xscale);
-			ds_ystep = FixedDiv(ystep, pSplat->yscale);
+			ds->ystep = FixedDiv(ystep, pSplat->yscale);
 
-			ds_xfrac = FixedDiv(offsetx + FixedMul(planecos, distance) + (x1 - centerx) * xstep, pSplat->xscale);
+			ds->xfrac = FixedDiv(offsetx + FixedMul(planecos, distance) + (x1 - centerx) * xstep, pSplat->xscale);
-			ds_yfrac = FixedDiv(offsety - FixedMul(planesin, distance) + (x1 - centerx) * ystep, pSplat->yscale);
+			ds->yfrac = FixedDiv(offsety - FixedMul(planesin, distance) + (x1 - centerx) * ystep, pSplat->yscale);
 		}
 
-		ds_y = y;
+		ds->y = y;
-		ds_x1 = x1;
+		ds->x1 = x1;
-		ds_x2 = x2;
+		ds->x2 = x2;
-		spanfunc();
+		ds->func(ds);
 
-		rastertab[y].minx = INT32_MAX;
+		planecontext->rastertab[y].minx = INT32_MAX;
-		rastertab[y].maxx = INT32_MIN;
+		planecontext->rastertab[y].maxx = INT32_MIN;
 	}
 
 	if (pSplat->angle && !pSplat->slope)
-		memset(cachedheight, 0, sizeof(cachedheight));
+		memset(planecontext->cachedheight, 0, sizeof(planecontext->cachedheight));
 }
 
-static void prepare_rastertab(void)
+static void prepare_rastertab(planecontext_t *planecontext)
 {
 	INT32 i;
-	prastertab = rastertab;
+	planecontext->prastertab = planecontext->rastertab;
 	for (i = 0; i < vid.height; i++)
 	{
-		rastertab[i].minx = INT32_MAX;
+		planecontext->rastertab[i].minx = INT32_MAX;
-		rastertab[i].maxx = INT32_MIN;
+		planecontext->rastertab[i].maxx = INT32_MIN;
 	}
 }

src/r_splats.h

@@ -26,7 +26,6 @@ struct rastery_s
 	fixed_t tx1, ty1;   // start points in texture at this line
 	fixed_t tx2, ty2;   // end points in texture at this line
 };
-extern struct rastery_s *prastertab; // for ASM code
 
 typedef struct floorsplat_s
 {
@@ -41,6 +40,8 @@ typedef struct floorsplat_s
 	mobj_t *mobj; // Mobj it is tied to
 } floorsplat_t;
 
-void R_DrawFloorSplat(vissprite_t *spr);
+struct rendercontext_s;
+
+void R_DrawFloorSplat(struct rendercontext_s *context, vissprite_t *spr);
 
 #endif /*__R_SPLATS_H__*/

src/r_state.h

@@ -90,6 +90,8 @@ extern sector_t *viewsector;
 extern player_t *viewplayer;
 extern mobj_t *r_viewmobj;
 
+extern consvar_t cv_numthreads;
+
 extern consvar_t cv_allowmlook;
 extern consvar_t cv_maxportals;
 
@@ -99,10 +101,4 @@ extern angle_t doubleclipangle;
 extern INT32 viewangletox[FINEANGLES/2];
 extern angle_t xtoviewangle[MAXVIDWIDTH+1];
 
-extern fixed_t rw_distance;
-extern angle_t rw_normalangle;
-
-// angle to line origin
-extern angle_t rw_angle1;
-
 #endif

src/r_textures.c

@@ -34,6 +34,15 @@
 
 #include <errno.h>
 
+#ifdef HAVE_THREADS
+static I_mutex r_texture_mutex;
+#  define Lock_state()    I_lock_mutex(&r_texture_mutex)
+#  define Unlock_state() I_unlock_mutex(r_texture_mutex)
+#else
+#  define Lock_state()
+#  define Unlock_state()
+#endif
+
 //
 // TEXTURE_T CACHING
 // When a texture is first needed, it counts the number of composite columns
@@ -508,8 +517,10 @@ INT32 R_GetTextureNum(INT32 texnum)
 //
 void R_CheckTextureCache(INT32 tex)
 {
+	Lock_state();
 	if (!texturecache[tex])
 		R_GenerateTexture(tex);
+	Unlock_state();
 }
 
 //
@@ -525,9 +536,11 @@ UINT8 *R_GetColumn(fixed_t tex, INT32 col)
 	else
 		col &= (width - 1);
 
+	Lock_state();
 	data = texturecache[tex];
 	if (!data)
 		data = R_GenerateTexture(tex);
+	Unlock_state();
 
 	return data + LONG(texturecolumnofs[tex][col]);
 }
@@ -542,7 +555,7 @@ void *R_GetFlat(lumpnum_t flatlumpnum)
 //
 // If needed, convert a texture or patch to a flat.
 //
-void *R_GetLevelFlat(levelflat_t *levelflat)
+void *R_GetLevelFlat(levelflat_t *levelflat, UINT16 *flatwidth, UINT16 *flatheight)
 {
 	boolean isleveltexture = (levelflat->type == LEVELFLAT_TEXTURE);
 	texture_t *texture = (isleveltexture ? textures[levelflat->u.texture.num] : NULL);
@@ -555,8 +568,8 @@ void *R_GetLevelFlat(levelflat_t *levelflat)
 		if (texture->flat)
 		{
 			flatdata = texture->flat;
-			ds_flatwidth = texture->width;
+			*flatwidth = texture->width;
-			ds_flatheight = texture->height;
+			*flatheight = texture->height;
 			texturechanged = false;
 		}
 		else
@@ -566,12 +579,14 @@ void *R_GetLevelFlat(levelflat_t *levelflat)
 	// If the texture changed, or the flat wasn't generated, convert.
 	if (levelflat->picture == NULL || texturechanged)
 	{
+		Lock_state();
+
 		// Level texture
 		if (isleveltexture)
 		{
 			levelflat->picture = R_GenerateTextureAsFlat(levelflat->u.texture.num);
-			ds_flatwidth = levelflat->width = texture->width;
+			*flatwidth = levelflat->width = texture->width;
-			ds_flatheight = levelflat->height = texture->height;
+			*flatheight = levelflat->height = texture->height;
 		}
 		else
 		{
@@ -584,8 +599,8 @@ void *R_GetLevelFlat(levelflat_t *levelflat)
 				levelflat->width = (UINT16)pngwidth;
 				levelflat->height = (UINT16)pngheight;
 
-				ds_flatwidth = levelflat->width;
+				*flatwidth = levelflat->width;
-				ds_flatheight = levelflat->height;
+				*flatheight = levelflat->height;
 			}
 			else
 #endif
@@ -595,8 +610,8 @@ void *R_GetLevelFlat(levelflat_t *levelflat)
 				size_t size;
 				softwarepatch_t *patch = W_CacheLumpNum(levelflat->u.flat.lumpnum, PU_CACHE);
 
-				levelflat->width = ds_flatwidth = SHORT(patch->width);
+				levelflat->width = *flatwidth = SHORT(patch->width);
-				levelflat->height = ds_flatheight = SHORT(patch->height);
+				levelflat->height = *flatheight = SHORT(patch->height);
 
 				levelflat->picture = Z_Malloc(levelflat->width * levelflat->height, PU_LEVEL, NULL);
 				converted = Picture_FlatConvert(PICFMT_DOOMPATCH, patch, PICFMT_FLAT, 0, &size, levelflat->width, levelflat->height, SHORT(patch->topoffset), SHORT(patch->leftoffset), 0);
@@ -604,11 +619,13 @@ void *R_GetLevelFlat(levelflat_t *levelflat)
 				Z_Free(converted);
 			}
 		}
+
+		Unlock_state();
 	}
 	else
 	{
-		ds_flatwidth = levelflat->width;
+		*flatwidth = levelflat->width;
-		ds_flatheight = levelflat->height;
+		*flatheight = levelflat->height;
 	}
 
 	levelflat->u.texture.lastnum = levelflat->u.texture.num;
@@ -621,23 +638,14 @@ void *R_GetLevelFlat(levelflat_t *levelflat)
 //
 // R_CheckPowersOfTwo
 //
-// Sets ds_powersoftwo true if the flat's dimensions are powers of two, and returns that.
+// Checks if the flat's dimensions are powers of two.
 //
-boolean R_CheckPowersOfTwo(void)
+boolean R_CheckPowersOfTwo(UINT16 flatwidth, UINT16 flatheight)
 {
-	boolean wpow2 = (!(ds_flatwidth & (ds_flatwidth - 1)));
+	boolean wpow2 = (!(flatwidth & (flatwidth - 1)));
-	boolean hpow2 = (!(ds_flatheight & (ds_flatheight - 1)));
+	boolean hpow2 = (!(flatheight & (flatheight - 1)));
 
-	// Initially, the flat isn't powers-of-two-sized.
+	return ((flatwidth == flatheight) && (wpow2 && hpow2));
-	ds_powersoftwo = false;
-
-	// But if the width and height are powers of two,
-	// and are EQUAL, then it's okay :]
-	if ((ds_flatwidth == ds_flatheight) && (wpow2 && hpow2))
-		ds_powersoftwo = true;
-
-	// Just return ds_powersoftwo.
-	return ds_powersoftwo;
 }
 
 //
@@ -645,58 +653,58 @@ boolean R_CheckPowersOfTwo(void)
 //
 // Determine the flat's dimensions from its lump length.
 //
-void R_CheckFlatLength(size_t size)
+void R_CheckFlatLength(spancontext_t *ds, size_t size)
 {
 	switch (size)
 	{
 		case 4194304: // 2048x2048 lump
-			nflatmask = 0x3FF800;
+			ds->nflatmask = 0x3FF800;
-			nflatxshift = 21;
+			ds->nflatxshift = 21;
-			nflatyshift = 10;
+			ds->nflatyshift = 10;
-			nflatshiftup = 5;
+			ds->nflatshiftup = 5;
-			ds_flatwidth = ds_flatheight = 2048;
+			ds->flatwidth = ds->flatheight = 2048;
 			break;
 		case 1048576: // 1024x1024 lump
-			nflatmask = 0xFFC00;
+			ds->nflatmask = 0xFFC00;
-			nflatxshift = 22;
+			ds->nflatxshift = 22;
-			nflatyshift = 12;
+			ds->nflatyshift = 12;
-			nflatshiftup = 6;
+			ds->nflatshiftup = 6;
-			ds_flatwidth = ds_flatheight = 1024;
+			ds->flatwidth = ds->flatheight = 1024;
 			break;
 		case 262144:// 512x512 lump
-			nflatmask = 0x3FE00;
+			ds->nflatmask = 0x3FE00;
-			nflatxshift = 23;
+			ds->nflatxshift = 23;
-			nflatyshift = 14;
+			ds->nflatyshift = 14;
-			nflatshiftup = 7;
+			ds->nflatshiftup = 7;
-			ds_flatwidth = ds_flatheight = 512;
+			ds->flatwidth = ds->flatheight = 512;
 			break;
 		case 65536: // 256x256 lump
-			nflatmask = 0xFF00;
+			ds->nflatmask = 0xFF00;
-			nflatxshift = 24;
+			ds->nflatxshift = 24;
-			nflatyshift = 16;
+			ds->nflatyshift = 16;
-			nflatshiftup = 8;
+			ds->nflatshiftup = 8;
-			ds_flatwidth = ds_flatheight = 256;
+			ds->flatwidth = ds->flatheight = 256;
 			break;
 		case 16384: // 128x128 lump
-			nflatmask = 0x3F80;
+			ds->nflatmask = 0x3F80;
-			nflatxshift = 25;
+			ds->nflatxshift = 25;
-			nflatyshift = 18;
+			ds->nflatyshift = 18;
-			nflatshiftup = 9;
+			ds->nflatshiftup = 9;
-			ds_flatwidth = ds_flatheight = 128;
+			ds->flatwidth = ds->flatheight = 128;
 			break;
 		case 1024: // 32x32 lump
-			nflatmask = 0x3E0;
+			ds->nflatmask = 0x3E0;
-			nflatxshift = 27;
+			ds->nflatxshift = 27;
-			nflatyshift = 22;
+			ds->nflatyshift = 22;
-			nflatshiftup = 11;
+			ds->nflatshiftup = 11;
-			ds_flatwidth = ds_flatheight = 32;
+			ds->flatwidth = ds->flatheight = 32;
 			break;
 		default: // 64x64 lump
-			nflatmask = 0xFC0;
+			ds->nflatmask = 0xFC0;
-			nflatxshift = 26;
+			ds->nflatxshift = 26;
-			nflatyshift = 20;
+			ds->nflatyshift = 20;
-			nflatshiftup = 10;
+			ds->nflatshiftup = 10;
-			ds_flatwidth = ds_flatheight = 64;
+			ds->flatwidth = ds->flatheight = 64;
 			break;
 	}
 }

src/r_textures.h

@@ -18,6 +18,7 @@
 #include "r_state.h"
 #include "p_setup.h" // levelflats
 #include "r_data.h"
+#include "r_draw.h"
 
 #ifdef __GNUG__
 #pragma interface
@@ -86,12 +87,12 @@ void R_CheckTextureCache(INT32 tex);
 void R_ClearTextureNumCache(boolean btell);
 
 // Retrieve texture data.
-void *R_GetLevelFlat(levelflat_t *levelflat);
+void *R_GetLevelFlat(levelflat_t *levelflat, UINT16 *flatwidth, UINT16 *flatheight);
 UINT8 *R_GetColumn(fixed_t tex, INT32 col);
 void *R_GetFlat(lumpnum_t flatnum);
 
-boolean R_CheckPowersOfTwo(void);
+boolean R_CheckPowersOfTwo(UINT16 flatwidth, UINT16 flatheight);
-void R_CheckFlatLength(size_t size);
+void R_CheckFlatLength(spancontext_t *ds, size_t size);
 
 // Returns the texture number for the texture name.
 INT32 R_TextureNumForName(const char *name);

src/r_things.h

@@ -27,6 +27,7 @@
 
 #define FEETADJUST (4<<FRACBITS) // R_AddSingleSpriteDef
 
+void R_InitSprites(void);
 boolean R_AddSingleSpriteDef(const char *sprname, spritedef_t *spritedef, UINT16 wadnum, UINT16 startlump, UINT16 endlump);
 
 //faB: find sprites in wadfile, replace existing, add new ones
@@ -37,23 +38,16 @@ void R_AddSpriteDefs(UINT16 wadnum);
 // MASKED COLUMN DRAWING
 // ---------------------
 
-// vars for R_DrawMaskedColumn
+void R_DrawMaskedColumn(colcontext_t *dc, column_t *column);
-extern INT16 *mfloorclip;
+void R_DrawFlippedMaskedColumn(colcontext_t *dc, column_t *column);
-extern INT16 *mceilingclip;
-extern fixed_t spryscale;
-extern fixed_t sprtopscreen;
-extern fixed_t sprbotscreen;
-extern fixed_t windowtop;
-extern fixed_t windowbottom;
-extern INT32 lengthcol;
-
-void R_DrawMaskedColumn(column_t *column);
-void R_DrawFlippedMaskedColumn(column_t *column);
 
 // ----------------
 // SPRITE RENDERING
 // ----------------
 
+struct rendercontext_s;
+struct spritecontext_s;
+
 // Constant arrays used for psprite clipping
 //  and initializing clipping.
 extern INT16 negonearray[MAXVIDWIDTH];
@@ -62,17 +56,18 @@ extern INT16 screenheightarray[MAXVIDWIDTH];
 fixed_t R_GetShadowZ(mobj_t *thing, pslope_t **shadowslope);
 
 //SoM: 6/5/2000: Light sprites correctly!
-void R_AddSprites(sector_t *sec, INT32 lightlevel);
+void R_AddSprites(struct rendercontext_s *context, sector_t *sec, INT32 lightlevel);
-void R_InitSprites(void);
+void R_ClearSprites(struct spritecontext_s *spritecontext);
-void R_ClearSprites(void);
 
-boolean R_ThingVisible (mobj_t *thing);
+boolean R_ThingVisible (mobj_t *thing, mobj_t *viewmobj);
 
-boolean R_ThingVisibleWithinDist (mobj_t *thing,
+boolean R_ThingVisibleWithinDist (fixed_t viewpos_x, fixed_t viewpos_y,
+		mobj_t *thing, mobj_t *viewmobj,
 		fixed_t        draw_dist,
 		fixed_t nights_draw_dist);
 
-boolean R_PrecipThingVisible (precipmobj_t *precipthing,
+boolean R_PrecipThingVisible (fixed_t viewpos_x, fixed_t viewpos_y,
+		precipmobj_t *precipthing,
 		fixed_t precip_draw_dist);
 
 boolean R_ThingHorizontallyFlipped (mobj_t *thing);
@@ -99,7 +94,7 @@ typedef struct
 	sector_t* viewsector;
 } maskcount_t;
 
-void R_DrawMasked(maskcount_t* masks, UINT8 nummasks);
+void R_DrawMasked(struct rendercontext_s *context, maskcount_t *masks, UINT8 nummasks);
 
 // ----------
 // VISSPRITES
@@ -210,10 +205,7 @@ typedef struct vissprite_s
 	INT32 dispoffset; // copy of info->dispoffset, affects ordering but not drawing
 } vissprite_t;
 
-extern UINT32 visspritecount;
+void R_ClipSprites(struct rendercontext_s *context, drawseg_t* dsstart, portal_t* portal);
-
-void R_ClipSprites(drawseg_t* dsstart, portal_t* portal);
-void R_ClipVisSprite(vissprite_t *spr, INT32 x1, INT32 x2, drawseg_t* dsstart, portal_t* portal);
 
 boolean R_SpriteIsFlashing(vissprite_t *vis);
 UINT8 *R_GetSpriteTranslation(vissprite_t *vis);
@@ -236,7 +228,7 @@ typedef struct drawnode_s
 	struct drawnode_s *prev;
 } drawnode_t;
 
-void R_InitDrawNodes(void);
+void R_InitDrawNodes(drawnode_t *nodebankhead);
 
 // -----------------------
 // SPRITE FRAME CHARACTERS

src/screen.c

@@ -39,20 +39,13 @@
 #include "hardware/hw_model.h"
 #endif
 
-
-#if defined (USEASM) && !defined (NORUSEASM)//&& (!defined (_MSC_VER) || (_MSC_VER <= 1200))
-#define RUSEASM //MSC.NET can't patch itself
-#endif
-
 // --------------------------------------------
 // assembly or c drawer routines for 8bpp/16bpp
 // --------------------------------------------
-void (*colfunc)(void);
+colfunc_t colfuncs[COLDRAWFUNC_MAX];
-void (*colfuncs[COLDRAWFUNC_MAX])(void);
 
-void (*spanfunc)(void);
+spanfunc_t spanfuncs[SPANDRAWFUNC_MAX];
-void (*spanfuncs[SPANDRAWFUNC_MAX])(void);
+spanfunc_t spanfuncs_npo2[SPANDRAWFUNC_MAX];
-void (*spanfuncs_npo2[SPANDRAWFUNC_MAX])(void);
 
 // ------------------
 // global video state
@@ -106,17 +99,10 @@ boolean R_SSE2 = false;
 
 void SCR_SetDrawFuncs(void)
 {
-	//
+	// Always run in 8bpp. todo: remove all 16bpp code?
-	//  setup the right draw routines for either 8bpp or 16bpp
-	//
-	if (true)//vid.bpp == 1) //Always run in 8bpp. todo: remove all 16bpp code?
-	{
 	colfuncs[BASEDRAWFUNC] = R_DrawColumn_8;
 	spanfuncs[BASEDRAWFUNC] = R_DrawSpan_8;
 
-		colfunc = colfuncs[BASEDRAWFUNC];
-		spanfunc = spanfuncs[BASEDRAWFUNC];
-
 	colfuncs[COLDRAWFUNC_FUZZY] = R_DrawTranslucentColumn_8;
 	colfuncs[COLDRAWFUNC_TRANS] = R_DrawTranslatedColumn_8;
 	colfuncs[COLDRAWFUNC_SHADE] = R_DrawShadeColumn_8;
@@ -154,47 +140,6 @@ void SCR_SetDrawFuncs(void)
 	spanfuncs_npo2[SPANDRAWFUNC_TILTEDTRANSSPRITE] = R_DrawTiltedTranslucentFloorSprite_NPO2_8;
 	spanfuncs_npo2[SPANDRAWFUNC_WATER] = R_DrawTranslucentWaterSpan_NPO2_8;
 	spanfuncs_npo2[SPANDRAWFUNC_TILTEDWATER] = R_DrawTiltedTranslucentWaterSpan_NPO2_8;
-		spanfuncs_npo2[SPANDRAWFUNC_FOG] = NULL; // Not needed
-
-#ifdef RUSEASM
-		if (R_ASM)
-		{
-			if (R_MMX)
-			{
-				colfuncs[BASEDRAWFUNC] = R_DrawColumn_8_MMX;
-				//colfuncs[COLDRAWFUNC_SHADE] = R_DrawShadeColumn_8_ASM;
-				//colfuncs[COLDRAWFUNC_FUZZY] = R_DrawTranslucentColumn_8_ASM;
-				colfuncs[COLDRAWFUNC_TWOSMULTIPATCH] = R_Draw2sMultiPatchColumn_8_MMX;
-				spanfuncs[BASEDRAWFUNC] = R_DrawSpan_8_MMX;
-			}
-			else
-			{
-				colfuncs[BASEDRAWFUNC] = R_DrawColumn_8_ASM;
-				//colfuncs[COLDRAWFUNC_SHADE] = R_DrawShadeColumn_8_ASM;
-				//colfuncs[COLDRAWFUNC_FUZZY] = R_DrawTranslucentColumn_8_ASM;
-				colfuncs[COLDRAWFUNC_TWOSMULTIPATCH] = R_Draw2sMultiPatchColumn_8_ASM;
-			}
-		}
-#endif
-	}
-/*	else if (vid.bpp > 1)
-	{
-		I_OutputMsg("using highcolor mode\n");
-		spanfunc = basespanfunc = R_DrawSpan_16;
-		transcolfunc = R_DrawTranslatedColumn_16;
-		transtransfunc = R_DrawTranslucentColumn_16; // No 16bit operation for this function
-
-		colfunc = basecolfunc = R_DrawColumn_16;
-		shadecolfunc = NULL; // detect error if used somewhere..
-		fuzzcolfunc = R_DrawTranslucentColumn_16;
-		walldrawerfunc = R_DrawWallColumn_16;
-	}*/
-	else
-		I_Error("unknown bytes per pixel mode %d\n", vid.bpp);
-/*
-	if (SCR_IsAspectCorrect(vid.width, vid.height))
-		CONS_Alert(CONS_WARNING, M_GetText("Resolution is not aspect-correct!\nUse a multiple of %dx%d\n"), BASEVIDWIDTH, BASEVIDHEIGHT);
-*/
 }
 
 void SCR_SetMode(void)

src/screen.h

@@ -112,58 +112,6 @@ typedef struct vmode_s
 #define NUMSPECIALMODES  4
 extern vmode_t specialmodes[NUMSPECIALMODES];
 
-// ---------------------------------------------
-// color mode dependent drawer function pointers
-// ---------------------------------------------
-
-#define BASEDRAWFUNC 0
-
-enum
-{
-	COLDRAWFUNC_BASE = BASEDRAWFUNC,
-	COLDRAWFUNC_FUZZY,
-	COLDRAWFUNC_TRANS,
-	COLDRAWFUNC_SHADE,
-	COLDRAWFUNC_SHADOWED,
-	COLDRAWFUNC_TRANSTRANS,
-	COLDRAWFUNC_TWOSMULTIPATCH,
-	COLDRAWFUNC_TWOSMULTIPATCHTRANS,
-	COLDRAWFUNC_FOG,
-
-	COLDRAWFUNC_MAX
-};
-
-extern void (*colfunc)(void);
-extern void (*colfuncs[COLDRAWFUNC_MAX])(void);
-
-enum
-{
-	SPANDRAWFUNC_BASE = BASEDRAWFUNC,
-	SPANDRAWFUNC_TRANS,
-	SPANDRAWFUNC_TILTED,
-	SPANDRAWFUNC_TILTEDTRANS,
-
-	SPANDRAWFUNC_SPLAT,
-	SPANDRAWFUNC_TRANSSPLAT,
-	SPANDRAWFUNC_TILTEDSPLAT,
-
-	SPANDRAWFUNC_SPRITE,
-	SPANDRAWFUNC_TRANSSPRITE,
-	SPANDRAWFUNC_TILTEDSPRITE,
-	SPANDRAWFUNC_TILTEDTRANSSPRITE,
-
-	SPANDRAWFUNC_WATER,
-	SPANDRAWFUNC_TILTEDWATER,
-
-	SPANDRAWFUNC_FOG,
-
-	SPANDRAWFUNC_MAX
-};
-
-extern void (*spanfunc)(void);
-extern void (*spanfuncs[SPANDRAWFUNC_MAX])(void);
-extern void (*spanfuncs_npo2[SPANDRAWFUNC_MAX])(void);
-
 // -----
 // CPUID
 // -----

src/sdl/CMakeLists.txt

@@ -128,11 +128,9 @@ if(${SDL2_FOUND})
 		if(${SRB2_CONFIG_YASM})
 			set(ASM_ASSEMBLER_TEMP ${CMAKE_ASM_YASM_COMPILER})
 			set(ASM_ASSEMBLER_OBJFORMAT ${CMAKE_ASM_YASM_OBJECT_FORMAT})
-			set_source_files_properties(${SRB2_NASM_SOURCES} LANGUAGE ASM_YASM)
 		else()
 			set(ASM_ASSEMBLER_TEMP ${CMAKE_ASM_NASM_COMPILER})
 			set(ASM_ASSEMBLER_OBJFORMAT ${CMAKE_ASM_NASM_OBJECT_FORMAT})
-			set_source_files_properties(${SRB2_NASM_SOURCES} LANGUAGE ASM_NASM)
 		endif()
 	endif()
 

src/sdl/i_system.c

@@ -177,6 +177,7 @@ static char returnWadPath[256];
 #include "../screen.h" //vid.WndParent
 #include "../d_net.h"
 #include "../g_game.h"
+#include "../r_main.h"
 #include "../filesrch.h"
 #include "endtxt.h"
 #include "sdlmain.h"
@@ -2272,6 +2273,7 @@ INT32 I_StartupSystem(void)
 #ifdef HAVE_THREADS
 	I_start_threads();
 	I_AddExitFunc(I_stop_threads);
+	I_AddExitFunc(R_StopThreads);
 #endif
 	I_StartupConsole();
 #ifdef NEWSIGNALHANDLER

src/sdl/i_threads.c

@@ -155,7 +155,7 @@ Worker (
 	return 0;
 }
 
-void
+I_thread_handle
 I_spawn_thread (
 		const char  * name,
 		I_thread_fn   entry,
@@ -189,6 +189,7 @@ I_spawn_thread (
 		}
 	}
 	I_unlock_mutex(i_thread_pool_mutex);
+	return (I_thread_handle)th;
 }
 
 int
@@ -354,3 +355,18 @@ I_wake_all_cond (
 	if (SDL_CondBroadcast(cond) == -1)
 		abort();
 }
+
+INT32
+I_atomic_load (
+		I_Atomicptr_t atomic
+){
+	return SDL_AtomicGet(atomic);
+}
+
+INT32
+I_atomic_exchange (
+	I_Atomicptr_t atomic,
+	INT32         val
+){
+	return SDL_AtomicSet(atomic, val);
+}

src/tmap.nas

@@ -1,957 +0,0 @@
-;; SONIC ROBO BLAST 2
-;;-----------------------------------------------------------------------------
-;; Copyright (C) 1998-2000 by DooM Legacy Team.
-;; Copyright (C) 1999-2021 by Sonic Team Junior.
-;;
-;; This program is free software distributed under the
-;; terms of the GNU General Public License, version 2.
-;; See the 'LICENSE' file for more details.
-;;-----------------------------------------------------------------------------
-;; FILE:
-;;      tmap.nas
-;; DESCRIPTION:
-;;      Assembler optimised rendering code for software mode.
-;;      Draw wall columns.
-
-
-[BITS 32]
-
-%define FRACBITS 16
-%define TRANSPARENTPIXEL 255
-
-%ifdef LINUX
-%macro cextern 1
-[extern %1]
-%endmacro
-
-%macro cglobal 1
-[global %1]
-%endmacro
-
-%else
-%macro cextern 1
-%define %1 _%1
-[extern %1]
-%endmacro
-
-%macro cglobal 1
-%define %1 _%1
-[global %1]
-%endmacro
-
-%endif
-
-
-; The viddef_s structure. We only need the width field.
-struc viddef_s
-        resb 12
-.width: resb 4
-        resb 44
-endstruc
-
-;; externs
-;; columns
-cextern dc_x
-cextern dc_yl
-cextern dc_yh
-cextern ylookup
-cextern columnofs
-cextern dc_source
-cextern dc_texturemid
-cextern dc_texheight
-cextern dc_iscale
-cextern dc_hires
-cextern centery
-cextern centeryfrac
-cextern dc_colormap
-cextern dc_transmap
-cextern colormaps
-cextern vid
-cextern topleft
-
-; DELME
-cextern R_DrawColumn_8
-
-; polygon edge rasterizer
-cextern prastertab
-
-[SECTION .data]
-
-;;.align        4
-loopcount       dd      0
-pixelcount      dd      0
-tystep          dd      0
-
-[SECTION .text]
-
-;;----------------------------------------------------------------------
-;;
-;; R_DrawColumn : 8bpp column drawer
-;;
-;; New  optimised version 10-01-1998 by D.Fabrice and P.Boris
-;; Revised by G. Dick July 2010 to support the intervening twelve years'
-;; worth of changes to the renderer. Since I only vaguely know what I'm
-;; doing, this is probably rather suboptimal. Help appreciated!
-;;
-;;----------------------------------------------------------------------
-;; fracstep, vid.width in memory
-;; eax = accumulator
-;; ebx = colormap
-;; ecx = count
-;; edx = heightmask
-;; esi = source
-;; edi = dest
-;; ebp = frac
-;;----------------------------------------------------------------------
-
-cglobal R_DrawColumn_8_ASM
-;       align   16
-R_DrawColumn_8_ASM:
-        push    ebp                     ;; preserve caller's stack frame pointer
-        push    esi                     ;; preserve register variables
-        push    edi
-        push    ebx
-;;
-;; dest = ylookup[dc_yl] + columnofs[dc_x];
-;;
-        mov     ebp,[dc_yl]
-        mov     edi,[ylookup+ebp*4]
-        mov     ebx,[dc_x]
-        add     edi,[columnofs+ebx*4]  ;; edi = dest
-;;
-;; pixelcount = yh - yl + 1
-;;
-        mov     ecx,[dc_yh]
-        add     ecx,1
-        sub     ecx,ebp                 ;; pixel count
-        jle     near .done              ;; nothing to scale
-;;
-;; fracstep = dc_iscale;	// But we just use [dc_iscale]
-;; frac = (dc_texturemid + FixedMul((dc_yl << FRACBITS) - centeryfrac, fracstep));
-;;
-        mov     eax,ebp                 ;; dc_yl
-        shl     eax,FRACBITS
-        sub     eax,[centeryfrac]
-        imul    dword [dc_iscale]
-        shrd    eax,edx,FRACBITS
-        add     eax,[dc_texturemid]
-        mov     ebp,eax                 ;; ebp = frac
-
-        mov     ebx,[dc_colormap]
-
-        mov     esi,[dc_source]
-;;
-;; if (dc_hires) frac = 0;
-;;
-        test    byte [dc_hires],0x01
-        jz      .texheightcheck
-        xor     ebp,ebp
-
-;;
-;; Check for power of two
-;;
-.texheightcheck:
-        mov     edx,[dc_texheight]
-        sub     edx,1                   ;; edx = heightmask
-        test    edx,[dc_texheight]
-        jnz     .notpowertwo
-
-        test    ecx,0x01                ;; Test for odd no. pixels
-        jnz     .odd
-
-;;
-;; Texture height is a power of two, so we get modular arithmetic by
-;; masking
-;;
-.powertwo:
-        mov     eax,ebp                 ;; eax = frac
-        sar     eax,FRACBITS            ;; Integer part
-        and     eax,edx                 ;; eax &= heightmask
-        movzx   eax,byte [esi + eax]    ;; eax = texel
-        add     ebp,[dc_iscale]         ;; frac += fracstep
-        movzx   eax,byte [ebx+eax]      ;; Map through colormap
-        mov     [edi],al                ;; Write pixel
-                                        ;; dest += vid.width
-        add     edi,[vid + viddef_s.width]
-
-.odd:
-        mov     eax,ebp                 ;; eax = frac
-        sar     eax,FRACBITS            ;; Integer part
-        and     eax,edx                 ;; eax &= heightmask
-        movzx   eax,byte [esi + eax]    ;; eax = texel
-        add     ebp,[dc_iscale]         ;; frac += fracstep
-        movzx   eax,byte [ebx+eax]      ;; Map through colormap
-        mov     [edi],al                ;; Write pixel
-                                        ;; dest += vid.width
-        add     edi,[vid + viddef_s.width]
-
-
-        sub     ecx,2                   ;; count -= 2
-        jg      .powertwo
-
-        jmp     .done
-
-.notpowertwo:
-        add     edx,1
-        shl     edx,FRACBITS
-        test    ebp,ebp
-        jns     .notpowtwoloop
-
-.makefracpos:
-        add     ebp,edx                 ;; frac is negative; make it positive
-        js      .makefracpos
-
-.notpowtwoloop:
-        cmp     ebp,edx                 ;; Reduce mod height
-        jl      .writenonpowtwo
-        sub     ebp,edx
-        jmp     .notpowtwoloop
-
-.writenonpowtwo:
-        mov     eax,ebp                 ;; eax = frac
-        sar     eax,FRACBITS            ;; Integer part.
-        mov     bl,[esi + eax]          ;; ebx = colormap + texel
-        add     ebp,[dc_iscale]         ;; frac += fracstep
-        movzx   eax,byte [ebx]          ;; Map through colormap
-        mov     [edi],al                ;; Write pixel
-                                        ;; dest += vid.width
-        add     edi,[vid + viddef_s.width]
-
-        sub     ecx,1
-        jnz     .notpowtwoloop
-
-;;
-
-.done:
-        pop     ebx                     ;; restore register variables
-        pop     edi
-        pop     esi
-        pop     ebp                     ;; restore caller's stack frame pointer
-        ret
-
-
-;;----------------------------------------------------------------------
-;;
-;; R_Draw2sMultiPatchColumn : Like R_DrawColumn, but omits transparent
-;;                            pixels.
-;;
-;; New  optimised version 10-01-1998 by D.Fabrice and P.Boris
-;; Revised by G. Dick July 2010 to support the intervening twelve years'
-;; worth of changes to the renderer. Since I only vaguely know what I'm
-;; doing, this is probably rather suboptimal. Help appreciated!
-;;
-;;----------------------------------------------------------------------
-;; fracstep, vid.width in memory
-;; eax = accumulator
-;; ebx = colormap
-;; ecx = count
-;; edx = heightmask
-;; esi = source
-;; edi = dest
-;; ebp = frac
-;;----------------------------------------------------------------------
-
-cglobal R_Draw2sMultiPatchColumn_8_ASM
-;       align   16
-R_Draw2sMultiPatchColumn_8_ASM:
-        push    ebp                     ;; preserve caller's stack frame pointer
-        push    esi                     ;; preserve register variables
-        push    edi
-        push    ebx
-;;
-;; dest = ylookup[dc_yl] + columnofs[dc_x];
-;;
-        mov     ebp,[dc_yl]
-        mov     edi,[ylookup+ebp*4]
-        mov     ebx,[dc_x]
-        add     edi,[columnofs+ebx*4]  ;; edi = dest
-;;
-;; pixelcount = yh - yl + 1
-;;
-        mov     ecx,[dc_yh]
-        add     ecx,1
-        sub     ecx,ebp                 ;; pixel count
-        jle     near .done              ;; nothing to scale
-;;
-;; fracstep = dc_iscale;	// But we just use [dc_iscale]
-;; frac = (dc_texturemid + FixedMul((dc_yl << FRACBITS) - centeryfrac, fracstep));
-;;
-        mov     eax,ebp                 ;; dc_yl
-        shl     eax,FRACBITS
-        sub     eax,[centeryfrac]
-        imul    dword [dc_iscale]
-        shrd    eax,edx,FRACBITS
-        add     eax,[dc_texturemid]
-        mov     ebp,eax                 ;; ebp = frac
-
-        mov     ebx,[dc_colormap]
-
-        mov     esi,[dc_source]
-;;
-;; if (dc_hires) frac = 0;
-;;
-        test    byte [dc_hires],0x01
-        jz      .texheightcheck
-        xor     ebp,ebp
-
-;;
-;; Check for power of two
-;;
-.texheightcheck:
-        mov     edx,[dc_texheight]
-        sub     edx,1                   ;; edx = heightmask
-        test    edx,[dc_texheight]
-        jnz     .notpowertwo
-
-        test    ecx,0x01                ;; Test for odd no. pixels
-        jnz     .odd
-
-;;
-;; Texture height is a power of two, so we get modular arithmetic by
-;; masking
-;;
-.powertwo:
-        mov     eax,ebp                 ;; eax = frac
-        sar     eax,FRACBITS            ;; Integer part
-        and     eax,edx                 ;; eax &= heightmask
-        movzx   eax,byte [esi + eax]    ;; eax = texel
-        add     ebp,[dc_iscale]         ;; frac += fracstep
-        cmp     al,TRANSPARENTPIXEL     ;; Is pixel transparent?
-        je      .nextpowtwoeven         ;; If so, advance.
-        movzx   eax,byte [ebx+eax]      ;; Map through colormap
-        mov	    [edi],al                ;; Write pixel
-.nextpowtwoeven:
-                                        ;; dest += vid.width
-        add     edi,[vid + viddef_s.width]
-
-.odd:
-        mov     eax,ebp                 ;; eax = frac
-        sar     eax,FRACBITS            ;; Integer part
-        and     eax,edx                 ;; eax &= heightmask
-        movzx   eax,byte [esi + eax]    ;; eax = texel
-        add     ebp,[dc_iscale]         ;; frac += fracstep
-        cmp     al,TRANSPARENTPIXEL     ;; Is pixel transparent?
-        je      .nextpowtwoodd          ;; If so, advance.
-        movzx   eax,byte [ebx+eax]      ;; Map through colormap
-        mov     [edi],al                ;; Write pixel
-.nextpowtwoodd:
-                                        ;; dest += vid.width
-        add     edi,[vid + viddef_s.width]
-
-
-        sub     ecx,2                   ;; count -= 2
-        jg      .powertwo
-
-        jmp     .done
-
-.notpowertwo:
-        add     edx,1
-        shl     edx,FRACBITS
-        test    ebp,ebp
-        jns     .notpowtwoloop
-
-.makefracpos:
-        add     ebp,edx                 ;; frac is negative; make it positive
-        js      .makefracpos
-
-.notpowtwoloop:
-        cmp     ebp,edx                 ;; Reduce mod height
-        jl      .writenonpowtwo
-        sub     ebp,edx
-        jmp     .notpowtwoloop
-
-.writenonpowtwo:
-        mov     eax,ebp                 ;; eax = frac
-        sar     eax,FRACBITS            ;; Integer part.
-        mov     bl,[esi + eax]          ;; ebx = colormap + texel
-        add     ebp,[dc_iscale]         ;; frac += fracstep
-        cmp     bl,TRANSPARENTPIXEL     ;; Is pixel transparent?
-        je      .nextnonpowtwo          ;; If so, advance.
-        movzx   eax,byte [ebx]          ;; Map through colormap
-        mov     [edi],al                ;; Write pixel
-.nextnonpowtwo:
-                                        ;; dest += vid.width
-        add     edi,[vid + viddef_s.width]
-
-        sub     ecx,1
-        jnz     .notpowtwoloop
-
-;;
-
-.done:
-        pop     ebx                     ;; restore register variables
-        pop     edi
-        pop     esi
-        pop     ebp                     ;; restore caller's stack frame pointer
-        ret
-
-;;----------------------------------------------------------------------
-;; R_DrawTranslucentColumnA_8
-;;
-;; Vertical column texture drawer, with transparency. Replaces Doom2's
-;; 'fuzz' effect, which was not so beautiful.
-;; Transparency is always impressive in some way, don't know why...
-;;----------------------------------------------------------------------
-
-cglobal R_DrawTranslucentColumn_8_ASM
-R_DrawTranslucentColumn_8_ASM:
-        push    ebp                     ;; preserve caller's stack frame pointer
-        push    esi                     ;; preserve register variables
-        push    edi
-        push    ebx
-;;
-;; dest = ylookup[dc_yl] + columnofs[dc_x];
-;;
-        mov     ebp,[dc_yl]
-        mov     ebx,ebp
-        mov     edi,[ylookup+ebx*4]
-        mov     ebx,[dc_x]
-        add     edi,[columnofs+ebx*4]   ;; edi = dest
-;;
-;; pixelcount = yh - yl + 1
-;;
-        mov     eax,[dc_yh]
-        inc     eax
-        sub     eax,ebp                 ;; pixel count
-        mov     [pixelcount],eax        ;; save for final pixel
-        jle     near    vtdone         ;; nothing to scale
-;;
-;; frac = dc_texturemid - (centery-dc_yl)*fracstep;
-;;
-        mov     ecx,[dc_iscale]        ;; fracstep
-        mov     eax,[centery]
-        sub     eax,ebp
-        imul    eax,ecx
-        mov     edx,[dc_texturemid]
-        sub     edx,eax
-        mov     ebx,edx
-
-        shr     ebx,16                  ;; frac int.
-        and     ebx,0x7f
-        shl     edx,16                  ;; y frac up
-
-        mov     ebp,ecx
-        shl     ebp,16                  ;; fracstep f. up
-        shr     ecx,16                  ;; fracstep i. ->cl
-        and     cl,0x7f
-        push    cx
-        mov     ecx,edx
-        pop     cx
-        mov     edx,[dc_colormap]
-        mov     esi,[dc_source]
-;;
-;; lets rock :) !
-;;
-        mov     eax,[pixelcount]
-        shr     eax,0x2
-        test    byte [pixelcount],0x3
-        mov     ch,al                   ;; quad count
-        mov     eax,[dc_transmap]
-        je      vt4quadloop
-;;
-;;  do un-even pixel
-;;
-        test    byte [pixelcount],0x1
-        je      trf2
-
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        add     ecx,ebp
-        adc     bl,cl
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     dl,[edx]
-        mov     [edi],dl
-pf:     add     edi,0x12345678
-;;
-;;  do two non-quad-aligned pixels
-;;
-trf2:    test    byte [pixelcount],0x2
-        je      trf3
-
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        add     ecx,ebp
-        adc     bl,cl
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     dl,[edx]
-        mov     [edi],dl
-pg:     add     edi,0x12345678
-
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        add     ecx,ebp
-        adc     bl,cl
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     dl,[edx]
-        mov     [edi],dl
-ph:     add     edi,0x12345678
-;;
-;;  test if there was at least 4 pixels
-;;
-trf3:   test    ch,0xff                 ;; test quad count
-        je near vtdone
-
-;;
-;; ebp : ystep frac. upper 24 bits
-;; edx : y     frac. upper 24 bits
-;; ebx : y     i.    lower 7 bits,  masked for index
-;; ecx : ch = counter, cl = y step i.
-;; eax : colormap aligned 256
-;; esi : source texture column
-;; edi : dest screen
-;;
-vt4quadloop:
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     [tystep],ebp
-pi:     add     edi,0x12345678
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-pj:     sub     edi,0x12345678
-        mov     ebp,edi
-pk:     sub     edi,0x12345678
-        jmp short inloop
-align 4
-vtquadloop:
-        add     ecx,[tystep]
-        adc     bl,cl
-q1:     add     ebp,0x23456789
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     dl,[edx]
-        mov     [edi],dl
-        mov     al,[ebp]                ;; fetch dest   : index into colormap
-inloop:
-        add     ecx,[tystep]
-        adc     bl,cl
-q2:     add     edi,0x23456789
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     dl,[edx]
-        mov     [ebp+0x0],dl
-        mov     al,[edi]                ;; fetch dest   : index into colormap
-
-        add     ecx,[tystep]
-        adc     bl,cl
-q3:     add     ebp,0x23456789
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     dl,[edx]
-        mov     [edi],dl
-        mov     al,[ebp]                ;; fetch dest   : index into colormap
-
-        add     ecx,[tystep]
-        adc     bl,cl
-q4:     add     edi,0x23456789
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     dl,[edx]
-        mov     [ebp],dl
-        mov     al,[edi]                ;; fetch dest   : index into colormap
-
-        dec     ch
-        jne     vtquadloop
-vtdone:
-        pop     ebx
-        pop     edi
-        pop     esi
-        pop     ebp
-        ret
-
-;;----------------------------------------------------------------------
-;; R_DrawShadeColumn
-;;
-;;   for smoke..etc.. test.
-;;----------------------------------------------------------------------
-cglobal R_DrawShadeColumn_8_ASM
-R_DrawShadeColumn_8_ASM:
-        push    ebp                     ;; preserve caller's stack frame pointer
-        push    esi                     ;; preserve register variables
-        push    edi
-        push    ebx
-
-;;
-;; dest = ylookup[dc_yl] + columnofs[dc_x];
-;;
-        mov     ebp,[dc_yl]
-        mov     ebx,ebp
-        mov     edi,[ylookup+ebx*4]
-        mov     ebx,[dc_x]
-        add     edi,[columnofs+ebx*4]  ;; edi = dest
-;;
-;; pixelcount = yh - yl + 1
-;;
-        mov     eax,[dc_yh]
-        inc     eax
-        sub     eax,ebp                 ;; pixel count
-        mov     [pixelcount],eax       ;; save for final pixel
-        jle near shdone                ;; nothing to scale
-;;
-;; frac = dc_texturemid - (centery-dc_yl)*fracstep;
-;;
-        mov     ecx,[dc_iscale]        ;; fracstep
-        mov     eax,[centery]
-        sub     eax,ebp
-        imul    eax,ecx
-        mov     edx,[dc_texturemid]
-        sub     edx,eax
-        mov     ebx,edx
-        shr     ebx,16                  ;; frac int.
-        and     ebx,byte +0x7f
-        shl     edx,16                  ;; y frac up
-
-        mov     ebp,ecx
-        shl     ebp,16                  ;; fracstep f. up
-        shr     ecx,16                  ;; fracstep i. ->cl
-        and     cl,0x7f
-
-        mov     esi,[dc_source]
-;;
-;; lets rock :) !
-;;
-        mov     eax,[pixelcount]
-        mov     dh,al
-        shr     eax,2
-        mov     ch,al                   ;; quad count
-        mov     eax,[colormaps]
-        test    dh,3
-        je      sh4quadloop
-;;
-;;  do un-even pixel
-;;
-        test    dh,0x1
-        je      shf2
-
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        add     edx,ebp
-        adc     bl,cl
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     [edi],dl
-pl:     add     edi,0x12345678
-;;
-;;  do two non-quad-aligned pixels
-;;
-shf2:
-        test    dh,0x2
-        je      shf3
-
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        add     edx,ebp
-        adc     bl,cl
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     [edi],dl
-pm:     add     edi,0x12345678
-
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        add     edx,ebp
-        adc     bl,cl
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-        and     bl,0x7f
-        mov     dl,[eax]
-        mov     [edi],dl
-pn:     add     edi,0x12345678
-;;
-;;  test if there was at least 4 pixels
-;;
-shf3:
-        test    ch,0xff                 ;; test quad count
-        je near shdone
-
-;;
-;; ebp : ystep frac. upper 24 bits
-;; edx : y     frac. upper 24 bits
-;; ebx : y     i.    lower 7 bits,  masked for index
-;; ecx : ch = counter, cl = y step i.
-;; eax : colormap aligned 256
-;; esi : source texture column
-;; edi : dest screen
-;;
-sh4quadloop:
-        mov     dh,0x7f                 ;; prep mask
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     [tystep],ebp
-po:     add     edi,0x12345678
-        mov     al,[edi]                ;; fetch dest  : index into colormap
-pp:     sub     edi,0x12345678
-        mov     ebp,edi
-pq:     sub     edi,0x12345678
-        jmp short shinloop
-
-align  4
-shquadloop:
-        add     edx,[tystep]
-        adc     bl,cl
-        and     bl,dh
-q5:     add     ebp,0x12345678
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     [edi],dl
-        mov     al,[ebp]                ;; fetch dest : index into colormap
-shinloop:
-        add     edx,[tystep]
-        adc     bl,cl
-        and     bl,dh
-q6:     add     edi,0x12345678
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     [ebp],dl
-        mov     al,[edi]                ;; fetch dest : index into colormap
-
-        add     edx,[tystep]
-        adc     bl,cl
-        and     bl,dh
-q7:     add     ebp,0x12345678
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     [edi],dl
-        mov     al,[ebp]                ;; fetch dest : index into colormap
-
-        add     edx,[tystep]
-        adc     bl,cl
-        and     bl,dh
-q8:     add     edi,0x12345678
-        mov     dl,[eax]
-        mov     ah,[esi+ebx]            ;; fetch texel : colormap number
-        mov     [ebp],dl
-        mov     al,[edi]                ;; fetch dest : index into colormap
-
-        dec     ch
-        jne     shquadloop
-
-shdone:
-        pop     ebx                     ;; restore register variables
-        pop     edi
-        pop     esi
-        pop     ebp                     ;; restore caller's stack frame pointer
-        ret
-
-
-;; ========================================================================
-;;  Rasterization of the segments of a LINEAR polygne textur of manire.
-;;  It is thus a question of interpolating coordinate them at the edges of texture in
-;;  the time that the X-coordinates minx/maxx for each line.
-;;  the argument ' dir' indicates which edges of texture are Interpol?:
-;;    0:  segments associs at edge TOP? and BOTTOM? (constant TY)
-;;    1:  segments associs at the LEFT and RIGHT edge (constant TX)
-;; ========================================================================
-;;
-;;  void   rasterize_segment_tex( LONG x1, LONG y1, LONG x2, LONG y2, LONG tv1, LONG tv2, LONG tc, LONG dir );
-;;                                   ARG1     ARG2     ARG3     ARG4      ARG5      ARG6     ARG7       ARG8
-;;
-;;  Pour dir = 0, (tv1,tv2) = (tX1,tX2), tc = tY, en effet TY est constant.
-;;
-;;  Pour dir = 1, (tv1,tv2) = (tY1,tY2), tc = tX, en effet TX est constant.
-;;
-;;
-;;  Uses:  extern struct rastery *_rastertab;
-;;
-
-MINX            EQU    0
-MAXX            EQU    4
-TX1             EQU    8
-TY1             EQU    12
-TX2             EQU    16
-TY2             EQU    20
-RASTERY_SIZEOF  EQU    24
-
-cglobal rasterize_segment_tex_asm
-rasterize_segment_tex_asm:
-        push    ebp
-        mov     ebp,esp
-
-        sub     esp,byte +0x8           ;; allocate the local variables
-
-        push    ebx
-        push    esi
-        push    edi
-        o16 mov ax,es
-        push    eax
-
-;;        #define DX       [ebp-4]
-;;        #define TD       [ebp-8]
-
-        mov     eax,[ebp+0xc]           ;; y1
-        mov     ebx,[ebp+0x14]          ;; y2
-        cmp     ebx,eax
-        je near .L_finished             ;; special (y1==y2) segment horizontal, exit!
-
-        jg near .L_rasterize_right
-
-;;rasterize_left:       ;; one rasterize a segment LEFT of the polygne
-
-        mov     ecx,eax
-        sub     ecx,ebx
-        inc     ecx                     ;; y1-y2+1
-
-        mov     eax,RASTERY_SIZEOF
-        mul     ebx                     ;; * y2
-        mov     esi,[prastertab]
-        add     esi,eax                 ;; point into rastertab[y2]
-
-        mov     eax,[ebp+0x8]           ;; ARG1
-        sub     eax,[ebp+0x10]          ;; ARG3
-        shl     eax,0x10                ;;     ((x1-x2)<<PRE) ...
-        cdq
-        idiv    ecx                     ;; dx =     ...        / (y1-y2+1)
-        mov     [ebp-0x4],eax           ;; DX
-
-        mov     eax,[ebp+0x18]          ;; ARG5
-        sub     eax,[ebp+0x1c]          ;; ARG6
-        shl     eax,0x10
-        cdq
-        idiv    ecx                     ;;      tdx =((tx1-tx2)<<PRE) / (y1-y2+1)
-        mov     [ebp-0x8],eax           ;; idem tdy =((ty1-ty2)<<PRE) / (y1-y2+1)
-
-        mov     eax,[ebp+0x10]          ;; ARG3
-        shl     eax,0x10                ;; x = x2<<PRE
-
-        mov     ebx,[ebp+0x1c]          ;; ARG6
-        shl     ebx,0x10                ;; tx = tx2<<PRE    d0
-                                        ;; ty = ty2<<PRE    d1
-        mov     edx,[ebp+0x20]          ;; ARG7
-        shl     edx,0x10                ;; ty = ty<<PRE     d0
-                                        ;; tx = tx<<PRE     d1
-        push    ebp
-        mov     edi,[ebp-0x4]           ;; DX
-        cmp     dword [ebp+0x24],byte +0x0      ;; ARG8   direction ?
-
-        mov     ebp,[ebp-0x8]           ;; TD
-        je      .L_rleft_h_loop
-;;
-;; TY varies, TX is constant
-;;
-.L_rleft_v_loop:
-        mov     [esi+MINX],eax           ;; rastertab[y].minx = x
-          add     ebx,ebp
-        mov     [esi+TX1],edx           ;;             .tx1  = tx
-          add     eax,edi
-        mov     [esi+TY1],ebx           ;;             .ty1  = ty
-
-        ;;addl    DX, %eax        // x     += dx
-        ;;addl    TD, %ebx        // ty    += tdy
-
-        add     esi,RASTERY_SIZEOF      ;; next raster line into rastertab[]
-        dec     ecx
-        jne     .L_rleft_v_loop
-        pop     ebp
-        jmp     .L_finished
-;;
-;; TX varies, TY is constant
-;;
-.L_rleft_h_loop:
-        mov     [esi+MINX],eax           ;; rastertab[y].minx = x
-          add     eax,edi
-        mov     [esi+TX1],ebx           ;;             .tx1  = tx
-          add     ebx,ebp
-        mov     [esi+TY1],edx           ;;             .ty1  = ty
-
-        ;;addl    DX, %eax        // x     += dx
-        ;;addl    TD, %ebx        // tx    += tdx
-
-        add     esi,RASTERY_SIZEOF      ;; next raster line into rastertab[]
-        dec     ecx
-        jne     .L_rleft_h_loop
-        pop     ebp
-        jmp     .L_finished
-;;
-;; one rasterize a segment LINE of the polygne
-;;
-.L_rasterize_right:
-        mov     ecx,ebx
-        sub     ecx,eax
-        inc     ecx                     ;; y2-y1+1
-
-        mov     ebx,RASTERY_SIZEOF
-        mul     ebx                     ;;   * y1
-        mov     esi,[prastertab]
-        add     esi,eax                 ;;  point into rastertab[y1]
-
-        mov     eax,[ebp+0x10]          ;; ARG3
-        sub     eax,[ebp+0x8]           ;; ARG1
-        shl     eax,0x10                ;; ((x2-x1)<<PRE) ...
-        cdq
-        idiv    ecx                     ;;  dx =     ...        / (y2-y1+1)
-        mov     [ebp-0x4],eax           ;; DX
-
-        mov     eax,[ebp+0x1c]          ;; ARG6
-        sub     eax,[ebp+0x18]          ;; ARG5
-        shl     eax,0x10
-        cdq
-        idiv    ecx                     ;;       tdx =((tx2-tx1)<<PRE) / (y2-y1+1)
-        mov     [ebp-0x8],eax           ;;  idem tdy =((ty2-ty1)<<PRE) / (y2-y1+1)
-
-        mov     eax,[ebp+0x8]           ;; ARG1
-        shl     eax,0x10                ;; x  = x1<<PRE
-
-        mov     ebx,[ebp+0x18]          ;; ARG5
-        shl     ebx,0x10                ;; tx = tx1<<PRE    d0
-                                        ;; ty = ty1<<PRE    d1
-        mov     edx,[ebp+0x20]          ;; ARG7
-        shl     edx,0x10                ;; ty = ty<<PRE     d0
-                                        ;; tx = tx<<PRE     d1
-        push    ebp
-        mov     edi,[ebp-0x4]           ;; DX
-
-        cmp     dword [ebp+0x24], 0     ;; direction ?
-
-         mov     ebp,[ebp-0x8]          ;; TD
-        je      .L_rright_h_loop
-;;
-;; TY varies, TX is constant
-;;
-.L_rright_v_loop:
-
-        mov     [esi+MAXX],eax           ;; rastertab[y].maxx = x
-          add     ebx,ebp
-        mov     [esi+TX2],edx          ;;             .tx2  = tx
-          add     eax,edi
-        mov     [esi+TY2],ebx          ;;             .ty2  = ty
-
-        ;;addl    DX, %eax        // x     += dx
-        ;;addl    TD, %ebx        // ty    += tdy
-
-        add     esi,RASTERY_SIZEOF
-        dec     ecx
-        jne     .L_rright_v_loop
-
-        pop     ebp
-
-        jmp     short .L_finished
-;;
-;; TX varies, TY is constant
-;;
-.L_rright_h_loop:
-        mov     [esi+MAXX],eax           ;; rastertab[y].maxx = x
-          add     eax,edi
-        mov     [esi+TX2],ebx          ;;             .tx2  = tx
-          add     ebx,ebp
-        mov     [esi+TY2],edx          ;;             .ty2  = ty
-
-        ;;addl    DX, %eax        // x     += dx
-        ;;addl    TD, %ebx        // tx    += tdx
-
-        add     esi,RASTERY_SIZEOF
-        dec     ecx
-        jne     .L_rright_h_loop
-
-        pop     ebp
-
-.L_finished:
-        pop     eax
-        o16 mov es,ax
-        pop     edi
-        pop     esi
-        pop     ebx
-
-        mov     esp,ebp
-        pop     ebp
-        ret

src/tmap_asm.s

@@ -1,322 +0,0 @@
-// SONIC ROBO BLAST 2
-//-----------------------------------------------------------------------------
-// Copyright (C) 1998-2000 by DooM Legacy Team.
-// Copyright (C) 1999-2021 by Sonic Team Junior.
-//
-// This program is free software distributed under the
-// terms of the GNU General Public License, version 2.
-// See the 'LICENSE' file for more details.
-//-----------------------------------------------------------------------------
-/// \file  tmap_asm.s
-/// \brief ???
-
-//.comm _dc_colormap,4
-//.comm _dc_x,4
-//.comm _dc_yl,4
-//.comm _dc_yh,4
-//.comm _dc_iscale,4
-//.comm _dc_texturemid,4
-//.comm _dc_source,4
-//.comm _ylookup,4
-//.comm _columnofs,4
-//.comm _loopcount,4
-//.comm _pixelcount,4
-.data
-_pixelcount:
-.long 0x00000000
-_loopcount:
-.long 0x00000000
-.align 8
-_mmxcomm:
-.long 0x00000000
-.text
-
-        .align 4
-.globl _R_DrawColumn8_NOMMX
-_R_DrawColumn8_NOMMX:
-   pushl %ebp
-   pushl %esi
-   pushl %edi
-   pushl %ebx
-	movl _dc_yl,%edx
-	movl _dc_yh,%eax
-	subl %edx,%eax
-	leal 1(%eax),%ebx
-	testl %ebx,%ebx
-	jle rdc8ndone
-	movl _dc_x,%eax
-        movl _ylookup, %edi
-	movl (%edi,%edx,4),%esi
-	movl _columnofs, %edi
-	addl (%edi,%eax,4),%esi
-	movl _dc_iscale,%edi
-	movl %edx,%eax
-	imull %edi,%eax
-	movl _dc_texturemid,%ecx
-	addl %eax,%ecx
-
-	movl _dc_source,%ebp
-   xorl %edx, %edx
-   subl $0x12345678, %esi
-.globl rdc8nwidth1
-rdc8nwidth1:
-	.align 4,0x90
-rdc8nloop:
-	movl %ecx,%eax
-	shrl $16,%eax
-	addl %edi,%ecx
-	andl $127,%eax
-	addl $0x12345678,%esi
-.globl rdc8nwidth2
-rdc8nwidth2:
-	movb (%eax,%ebp),%dl
-	movl _dc_colormap,%eax
-	movb (%eax,%edx),%al
-	movb %al,(%esi)
-	decl %ebx
-	jne rdc8nloop
-rdc8ndone:
-   popl %ebx
-   popl %edi
-   popl %esi
-   popl %ebp
-   ret
-
-//
-// Optimised specifically for P54C/P55C (aka Pentium with/without MMX)
-// By ES 1998/08/01
-//
-
-.globl _R_DrawColumn_8_Pentium
-_R_DrawColumn_8_Pentium:
-	pushl %ebp
-        pushl %ebx
-	pushl %esi
-        pushl %edi
-	movl _dc_yl,%eax        // Top pixel
-	movl _dc_yh,%ebx        // Bottom pixel
-        movl _ylookup, %edi
-	movl (%edi,%ebx,4),%ecx
-	subl %eax,%ebx          // ebx=number of pixels-1
-	jl rdc8pdone            // no pixel to draw, done
-	jnz rdc8pmany
-	movl _dc_x,%edx         // Special case: only one pixel
-        movl _columnofs, %edi
-	addl (%edi,%edx,4),%ecx // dest pixel at (%ecx)
-	movl _dc_iscale,%esi
-	imull %esi,%eax
-	movl _dc_texturemid,%edi
-	addl %eax,%edi          // texture index in edi
-	movl _dc_colormap,%edx
-   	shrl $16, %edi
-   	movl _dc_source,%ebp
-	andl $127,%edi
-	movb (%edi,%ebp),%dl    // read texture pixel
-	movb (%edx),%al	        // lookup for light
-	movb %al,0(%ecx) 	// write it
-	jmp rdc8pdone		// done!
-.align 4, 0x90
-rdc8pmany:			// draw >1 pixel
-	movl _dc_x,%edx
-        movl _columnofs, %edi
-	movl (%edi,%edx,4),%edx
-	leal 0x12345678(%edx, %ecx), %edi  // edi = two pixels above bottom
-.globl rdc8pwidth5
-rdc8pwidth5:  // DeadBeef = -2*SCREENWIDTH
-        movl _dc_iscale,%edx	// edx = fracstep
-	imull %edx,%eax
-   	shll $9, %edx           // fixme: Should get 7.25 fix as input
-	movl _dc_texturemid,%ecx
-	addl %eax,%ecx          // ecx = frac
-	movl _dc_colormap,%eax  // eax = lighting/special effects LUT
-   	shll $9, %ecx
-   	movl _dc_source,%esi    // esi = source ptr
-
-	imull $0x12345678, %ebx // ebx = negative offset to pixel
-.globl rdc8pwidth6
-rdc8pwidth6:  // DeadBeef = -SCREENWIDTH
-
-// Begin the calculation of the two first pixels
-        leal (%ecx, %edx), %ebp
-	shrl $25, %ecx
-	movb (%esi, %ecx), %al
-	leal (%edx, %ebp), %ecx
-	shrl $25, %ebp
-        movb (%eax), %dl
-
-// The main loop
-rdc8ploop:
-	movb (%esi,%ebp), %al		// load 1
-        leal (%ecx, %edx), %ebp         // calc frac 3
-
-	shrl $25, %ecx                  // shift frac 2
-        movb %dl, 0x12345678(%edi, %ebx)// store 0
-.globl rdc8pwidth1
-rdc8pwidth1:  // DeadBeef = 2*SCREENWIDTH
-
-        movb (%eax), %al                // lookup 1
-
-        movb %al, 0x12345678(%edi, %ebx)// store 1
-.globl rdc8pwidth2
-rdc8pwidth2:  // DeadBeef = 3*SCREENWIDTH
-        movb (%esi, %ecx), %al          // load 2
-
-        leal (%ebp, %edx), %ecx         // calc frac 4
-
-        shrl $25, %ebp                  // shift frac 3
-        movb (%eax), %dl                // lookup 2
-
-        addl $0x12345678, %ebx          // counter
-.globl rdc8pwidth3
-rdc8pwidth3:  // DeadBeef = 2*SCREENWIDTH
-        jl rdc8ploop                    // loop
-
-// End of loop. Write extra pixel or just exit.
-        jnz rdc8pdone
-        movb %dl, 0x12345678(%edi, %ebx)// Write odd pixel
-.globl rdc8pwidth4
-rdc8pwidth4:  // DeadBeef = 2*SCREENWIDTH
-
-rdc8pdone:
-
-        popl %edi
-	popl %esi
-        popl %ebx
-	popl %ebp
-        ret
-
-//
-// MMX asm version, optimised for K6
-// By ES 1998/07/05
-//
-
-.globl _R_DrawColumn_8_K6_MMX
-_R_DrawColumn_8_K6_MMX:
-	pushl %ebp
-        pushl %ebx
-	pushl %esi
-        pushl %edi
-
-        movl %esp, %eax // Push 8 or 12, so that (%esp) gets aligned by 8
-        andl $7,%eax
-        addl $8,%eax
-        movl %eax, _mmxcomm // Temp storage in mmxcomm: (%esp) is used instead
-        subl %eax,%esp
-
-	movl _dc_yl,%edx        // Top pixel
-	movl _dc_yh,%ebx        // Bottom pixel
-        movl _ylookup, %edi
-	movl (%edi,%ebx,4),%ecx
-	subl %edx,%ebx         // ebx=number of pixels-1
-	jl 0x12345678            // no pixel to draw, done
-.globl rdc8moffs1
-rdc8moffs1:
-	jnz rdc8mmany
-	movl _dc_x,%eax         // Special case: only one pixel
-        movl _columnofs, %edi
-	addl (%edi,%eax,4),%ecx  // dest pixel at (%ecx)
-	movl _dc_iscale,%esi
-	imull %esi,%edx
-	movl _dc_texturemid,%edi
-	addl %edx,%edi         // texture index in edi
-	movl _dc_colormap,%edx
-   	shrl $16, %edi
-   	movl _dc_source,%ebp
-	andl $127,%edi
-	movb (%edi,%ebp),%dl  // read texture pixel
-	movb (%edx),%al	 // lookup for light
-	movb %al,0(%ecx) 	 // write it
-	jmp rdc8mdone		 // done!
-.globl rdc8moffs2
-rdc8moffs2:
-.align 4, 0x90
-rdc8mmany:			 // draw >1 pixel
-	movl _dc_x,%eax
-        movl _columnofs, %edi
-	movl (%edi,%eax,4),%eax
-	leal 0x12345678(%eax, %ecx), %esi  // esi = two pixels above bottom
-.globl rdc8mwidth3
-rdc8mwidth3:  // DeadBeef = -2*SCREENWIDTH
-        movl _dc_iscale,%ecx	 // ecx = fracstep
-	imull %ecx,%edx
-   	shll $9, %ecx           // fixme: Should get 7.25 fix as input
-	movl _dc_texturemid,%eax
-	addl %edx,%eax         // eax = frac
-	movl _dc_colormap,%edx  // edx = lighting/special effects LUT
-   	shll $9, %eax
-	leal (%ecx, %ecx), %edi
-   	movl _dc_source,%ebp    // ebp = source ptr
-	movl %edi, 0(%esp)     // Start moving frac and fracstep to MMX regs
-
-	imull $0x12345678, %ebx  // ebx = negative offset to pixel
-.globl rdc8mwidth5
-rdc8mwidth5:  // DeadBeef = -SCREENWIDTH
-
-	movl %edi, 4(%esp)
-	leal (%eax, %ecx), %edi
-	movq 0(%esp), %mm1     // fracstep:fracstep in mm1
-	movl %eax, 0(%esp)
-	shrl $25, %eax
-	movl %edi, 4(%esp)
-	movzbl (%ebp, %eax), %eax
-	movq 0(%esp), %mm0     // frac:frac in mm0
-
-	paddd %mm1, %mm0
-	shrl $25, %edi
-	movq %mm0, %mm2
-	psrld $25, %mm2         // texture index in mm2
-	paddd %mm1, %mm0
-	movq %mm2, 0(%esp)
-
-.globl rdc8mloop
-rdc8mloop:                      		// The main loop
-	movq %mm0, %mm2                    // move 4-5 to temp reg
-	movzbl (%ebp, %edi), %edi 		// read 1
-
-	psrld $25, %mm2 			// shift 4-5
-	movb (%edx,%eax), %cl 		// lookup 0
-
-	movl 0(%esp), %eax 			// load 2
-	addl $0x12345678, %ebx 		// counter
-.globl rdc8mwidth2
-rdc8mwidth2:  // DeadBeef = 2*SCREENWIDTH
-
-	movb %cl, (%esi, %ebx)		// write 0
-	movb (%edx,%edi), %ch 		// lookup 1
-
-	movb %ch, 0x12345678(%esi, %ebx) 	// write 1
-.globl rdc8mwidth1
-rdc8mwidth1:  // DeadBeef = SCREENWIDTH
-	movl 4(%esp), %edi			// load 3
-
-	paddd %mm1, %mm0 			// frac 6-7
-	movzbl (%ebp, %eax), %eax 		// lookup 2
-
-	movq %mm2, 0(%esp) 		     // store texture index 4-5
-	jl rdc8mloop
-
-	jnz rdc8mno_odd
-	movb (%edx,%eax), %cl  // write the last odd pixel
-	movb %cl, 0x12345678(%esi)
-.globl rdc8mwidth4
-rdc8mwidth4:  // DeadBeef = 2*SCREENWIDTH
-rdc8mno_odd:
-
-.globl rdc8mdone
-rdc8mdone:
-        emms
-
-        addl _mmxcomm, %esp
-        popl %edi
-	popl %esi
-        popl %ebx
-	popl %ebp
-        ret
-
-// Need some extra space to align run-time
-.globl R_DrawColumn_8_K6_MMX_end
-R_DrawColumn_8_K6_MMX_end:
-nop;nop;nop;nop;nop;nop;nop;nop;
-nop;nop;nop;nop;nop;nop;nop;nop;
-nop;nop;nop;nop;nop;nop;nop;nop;
-nop;nop;nop;nop;nop;nop;nop;

src/tmap_mmx.nas

@@ -1,674 +0,0 @@
-;; SONIC ROBO BLAST 2
-;;-----------------------------------------------------------------------------
-;; Copyright (C) 1998-2000 by DOSDOOM.
-;; Copyright (C) 2010-2021 by Sonic Team Junior.
-;;
-;; This program is free software distributed under the
-;; terms of the GNU General Public License, version 2.
-;; See the 'LICENSE' file for more details.
-;;-----------------------------------------------------------------------------
-;; FILE:
-;;      tmap_mmx.nas
-;; DESCRIPTION:
-;;      Assembler optimised rendering code for software mode, using SIMD
-;;      instructions.
-;;      Draw wall columns.
-
-
-[BITS 32]
-
-%define FRACBITS 16
-%define TRANSPARENTPIXEL 255
-
-%ifdef LINUX
-%macro cextern 1
-[extern %1]
-%endmacro
-
-%macro cglobal 1
-[global %1]
-%endmacro
-
-%else
-%macro cextern 1
-%define %1 _%1
-[extern %1]
-%endmacro
-
-%macro cglobal 1
-%define %1 _%1
-[global %1]
-%endmacro
-
-%endif
-
-
-; The viddef_s structure. We only need the width field.
-struc viddef_s
-		resb 12
-.width: resb 4
-		resb 44
-endstruc
-
-
-;; externs
-;; columns
-cextern dc_colormap
-cextern dc_x
-cextern dc_yl
-cextern dc_yh
-cextern dc_iscale
-cextern dc_texturemid
-cextern dc_texheight
-cextern dc_source
-cextern dc_hires
-cextern centery
-cextern centeryfrac
-cextern dc_transmap
-
-cextern R_DrawColumn_8_ASM
-cextern R_Draw2sMultiPatchColumn_8_ASM
-
-;; spans
-cextern nflatshiftup
-cextern nflatxshift
-cextern nflatyshift
-cextern nflatmask
-cextern ds_xfrac
-cextern ds_yfrac
-cextern ds_xstep
-cextern ds_ystep
-cextern ds_x1
-cextern ds_x2
-cextern ds_y
-cextern ds_source
-cextern ds_colormap
-
-cextern ylookup
-cextern columnofs
-cextern vid
-
-[SECTION .data]
-
-nflatmask64		dq		0
-
-
-[SECTION .text]
-
-;;----------------------------------------------------------------------
-;;
-;; R_DrawColumn : 8bpp column drawer
-;;
-;; MMX column drawer.
-;;
-;;----------------------------------------------------------------------
-;; eax = accumulator
-;; ebx = colormap
-;; ecx = count
-;; edx = accumulator
-;; esi = source
-;; edi = dest
-;; ebp = vid.width
-;; mm0 = accumulator
-;; mm1 = heightmask, twice
-;; mm2 = 2 * fracstep, twice
-;; mm3 = pair of consecutive fracs
-;;----------------------------------------------------------------------
-
-
-cglobal R_DrawColumn_8_MMX
-R_DrawColumn_8_MMX:
-		push		ebp						;; preserve caller's stack frame pointer
-		push		esi						;; preserve register variables
-		push		edi
-		push		ebx
-
-;;
-;; Our algorithm requires that the texture height be a power of two.
-;; If not, fall back to the non-MMX drawer.
-;;
-.texheightcheck:
-		mov			edx, [dc_texheight]
-		sub			edx, 1					;; edx = heightmask
-		test		edx, [dc_texheight]
-		jnz			near .usenonMMX
-
-		mov			ebp, edx				;; Keep a copy of heightmask in a
-											;; GPR for the time being.
-
-;;
-;; Fill mm1 with heightmask
-;;
-		movd		mm1, edx				;; low dword = heightmask
-		punpckldq	mm1, mm1				;; copy low dword to high dword
-
-;;
-;; dest = ylookup[dc_yl] + columnofs[dc_x];
-;;
-		mov			eax, [dc_yl]
-		mov			edi, [ylookup+eax*4]
-		mov			ebx, [dc_x]
-		add			edi, [columnofs+ebx*4]	;; edi = dest
-
-
-;;
-;; pixelcount = yh - yl + 1
-;;
-		mov			ecx, [dc_yh]
-		add			ecx, 1
-		sub			ecx, eax				;; pixel count
-		jle			near .done				;; nothing to scale
-
-;;
-;; fracstep = dc_iscale;
-;;
-		movd		mm2, [dc_iscale]		;; fracstep in low dword
-		punpckldq	mm2, mm2				;; copy to high dword
-
-		mov			ebx, [dc_colormap]
-		mov			esi, [dc_source]
-
-;;
-;; frac = (dc_texturemid + FixedMul((dc_yl << FRACBITS) - centeryfrac, fracstep));
-;;
-											;; eax == dc_yl already
-		shl			eax, FRACBITS
-		sub			eax, [centeryfrac]
-		imul		dword [dc_iscale]
-		shrd		eax, edx, FRACBITS
-		add			eax, [dc_texturemid]
-
-;;
-;; if (dc_hires) frac = 0;
-;;
-		test		byte [dc_hires], 0x01
-		jz			.mod2
-		xor			eax, eax
-
-
-;;
-;; Do mod-2 pixel.
-;;
-.mod2:
-		test		ecx, 1
-		jz			.pairprepare
-		mov			edx, eax				;; edx = frac
-		add			eax, [dc_iscale]		;; eax += fracstep
-		sar			edx, FRACBITS
-		and			edx, ebp				;; edx &= heightmask
-		movzx		edx, byte [esi + edx]
-		movzx		edx, byte [ebx + edx]
-		mov			[edi], dl
-
-		add			edi, [vid + viddef_s.width]
-		sub			ecx, 1
-		jz			.done
-
-.pairprepare:
-;;
-;; Prepare for the main loop.
-;;
-		movd		mm3, eax				;; Low dword = frac
-		movq		mm4, mm3				;; Copy to intermediate register
-		paddd		mm4, mm2				;; dwords of mm4 += fracstep
-		punpckldq	mm3, mm4				;; Low dword = first frac, high = second
-		pslld		mm2, 1					;; fracstep *= 2
-
-;;
-;; ebp = vid.width
-;;
-		mov			ebp, [vid + viddef_s.width]
-
-		align		16
-.pairloop:
-		movq		mm0, mm3				;; 3B 1u.
-		psrad		mm0, FRACBITS			;; 4B 1u.
-		pand		mm0, mm1				;; 3B 1u. frac &= heightmask
-		paddd		mm3, mm2				;; 3B 1u. frac += fracstep
-
-		movd		eax, mm0				;; 3B 1u. Get first frac
-;; IFETCH boundary
-		movzx		eax, byte [esi + eax]	;; 4B 1u. Texture map
-		movzx		eax, byte [ebx + eax]	;; 4B 1u. Colormap
-
-		punpckhdq	mm0, mm0				;; 3B 1(2)u. low dword = high dword
-		movd		edx, mm0				;; 3B 1u. Get second frac
-		mov			[edi], al				;; 2B 1(2)u. First pixel
-;; IFETCH boundary
-
-		movzx		edx, byte [esi + edx]	;; 4B 1u. Texture map
-		movzx		edx, byte [ebx + edx]	;; 4B 1u. Colormap
-		mov			[edi + 1*ebp], dl		;; 3B 1(2)u. Second pixel
-
-		lea			edi, [edi + 2*ebp]		;; 3B 1u. edi += 2 * vid.width
-;; IFETCH boundary
-		sub			ecx, 2					;; 3B 1u. count -= 2
-		jnz			.pairloop				;; 2B 1u. if(count != 0) goto .pairloop
-
-
-.done:
-;;
-;; Clear MMX state, or else FPU operations will go badly awry.
-;;
-		emms
-
-		pop			ebx
-		pop			edi
-		pop			esi
-		pop			ebp
-		ret
-
-.usenonMMX:
-		call		R_DrawColumn_8_ASM
-		jmp			.done
-
-
-;;----------------------------------------------------------------------
-;;
-;; R_Draw2sMultiPatchColumn : Like R_DrawColumn, but omits transparent
-;;                            pixels.
-;;
-;; MMX column drawer.
-;;
-;;----------------------------------------------------------------------
-;; eax = accumulator
-;; ebx = colormap
-;; ecx = count
-;; edx = accumulator
-;; esi = source
-;; edi = dest
-;; ebp = vid.width
-;; mm0 = accumulator
-;; mm1 = heightmask, twice
-;; mm2 = 2 * fracstep, twice
-;; mm3 = pair of consecutive fracs
-;;----------------------------------------------------------------------
-
-
-cglobal R_Draw2sMultiPatchColumn_8_MMX
-R_Draw2sMultiPatchColumn_8_MMX:
-		push		ebp						;; preserve caller's stack frame pointer
-		push		esi						;; preserve register variables
-		push		edi
-		push		ebx
-
-;;
-;; Our algorithm requires that the texture height be a power of two.
-;; If not, fall back to the non-MMX drawer.
-;;
-.texheightcheck:
-		mov			edx, [dc_texheight]
-		sub			edx, 1					;; edx = heightmask
-		test		edx, [dc_texheight]
-		jnz			near .usenonMMX
-
-		mov			ebp, edx				;; Keep a copy of heightmask in a
-											;; GPR for the time being.
-
-;;
-;; Fill mm1 with heightmask
-;;
-		movd		mm1, edx				;; low dword = heightmask
-		punpckldq	mm1, mm1				;; copy low dword to high dword
-
-;;
-;; dest = ylookup[dc_yl] + columnofs[dc_x];
-;;
-		mov			eax, [dc_yl]
-		mov			edi, [ylookup+eax*4]
-		mov			ebx, [dc_x]
-		add			edi, [columnofs+ebx*4]	;; edi = dest
-
-
-;;
-;; pixelcount = yh - yl + 1
-;;
-		mov			ecx, [dc_yh]
-		add			ecx, 1
-		sub			ecx, eax				;; pixel count
-		jle			near .done				;; nothing to scale
-;;
-;; fracstep = dc_iscale;
-;;
-		movd		mm2, [dc_iscale]		;; fracstep in low dword
-		punpckldq	mm2, mm2				;; copy to high dword
-
-		mov			ebx, [dc_colormap]
-		mov			esi, [dc_source]
-
-;;
-;; frac = (dc_texturemid + FixedMul((dc_yl << FRACBITS) - centeryfrac, fracstep));
-;;
-											;; eax == dc_yl already
-		shl			eax, FRACBITS
-		sub			eax, [centeryfrac]
-		imul		dword [dc_iscale]
-		shrd		eax, edx, FRACBITS
-		add			eax, [dc_texturemid]
-
-;;
-;; if (dc_hires) frac = 0;
-;;
-		test		byte [dc_hires], 0x01
-		jz			.mod2
-		xor			eax, eax
-
-
-;;
-;; Do mod-2 pixel.
-;;
-.mod2:
-		test		ecx, 1
-		jz			.pairprepare
-		mov			edx, eax				;; edx = frac
-		add			eax, [dc_iscale]		;; eax += fracstep
-		sar			edx, FRACBITS
-		and			edx, ebp				;; edx &= heightmask
-		movzx		edx, byte [esi + edx]
-		cmp			dl, TRANSPARENTPIXEL
-		je			.nextmod2
-		movzx		edx, byte [ebx + edx]
-		mov			[edi], dl
-
-.nextmod2:
-		add			edi, [vid + viddef_s.width]
-		sub			ecx, 1
-		jz			.done
-
-.pairprepare:
-;;
-;; Prepare for the main loop.
-;;
-		movd		mm3, eax				;; Low dword = frac
-		movq		mm4, mm3				;; Copy to intermediate register
-		paddd		mm4, mm2				;; dwords of mm4 += fracstep
-		punpckldq	mm3, mm4				;; Low dword = first frac, high = second
-		pslld		mm2, 1					;; fracstep *= 2
-
-;;
-;; ebp = vid.width
-;;
-		mov			ebp, [vid + viddef_s.width]
-
-		align		16
-.pairloop:
-		movq		mm0, mm3				;; 3B 1u.
-		psrad		mm0, FRACBITS			;; 4B 1u.
-		pand		mm0, mm1				;; 3B 1u. frac &= heightmask
-		paddd		mm3, mm2				;; 3B 1u. frac += fracstep
-
-		movd		eax, mm0				;; 3B 1u. Get first frac
-;; IFETCH boundary
-		movzx		eax, byte [esi + eax]	;; 4B 1u. Texture map
-		punpckhdq	mm0, mm0				;; 3B 1(2)u. low dword = high dword
-		movd		edx, mm0				;; 3B 1u. Get second frac
-		cmp			al, TRANSPARENTPIXEL	;; 2B 1u.
-		je			.secondinpair			;; 2B 1u.
-;; IFETCH boundary
-		movzx		eax, byte [ebx + eax]	;; 4B 1u. Colormap
-		mov			[edi], al				;; 2B 1(2)u. First pixel
-
-.secondinpair:
-		movzx		edx, byte [esi + edx]	;; 4B 1u. Texture map
-		cmp			dl, TRANSPARENTPIXEL	;; 2B 1u.
-		je			.nextpair				;; 2B 1u.
-;; IFETCH boundary
-		movzx		edx, byte [ebx + edx]	;; 4B 1u. Colormap
-		mov			[edi + 1*ebp], dl		;; 3B 1(2)u. Second pixel
-
-.nextpair:
-		lea			edi, [edi + 2*ebp]		;; 3B 1u. edi += 2 * vid.width
-		sub			ecx, 2					;; 3B 1u. count -= 2
-		jnz			.pairloop				;; 2B 1u. if(count != 0) goto .pairloop
-
-
-.done:
-;;
-;; Clear MMX state, or else FPU operations will go badly awry.
-;;
-		emms
-
-		pop			ebx
-		pop			edi
-		pop			esi
-		pop			ebp
-		ret
-
-.usenonMMX:
-		call		R_Draw2sMultiPatchColumn_8_ASM
-		jmp			.done
-
-
-;;----------------------------------------------------------------------
-;;
-;; R_DrawSpan : 8bpp span drawer
-;;
-;; MMX span drawer.
-;;
-;;----------------------------------------------------------------------
-;; eax = accumulator
-;; ebx = colormap
-;; ecx = count
-;; edx = accumulator
-;; esi = source
-;; edi = dest
-;; ebp = two pixels
-;; mm0 = accumulator
-;; mm1 = xposition
-;; mm2 = yposition
-;; mm3 = 2 * xstep
-;; mm4 = 2 * ystep
-;; mm5 = nflatxshift
-;; mm6 = nflatyshift
-;; mm7 = accumulator
-;;----------------------------------------------------------------------
-
-cglobal R_DrawSpan_8_MMX
-R_DrawSpan_8_MMX:
-		push		ebp						;; preserve caller's stack frame pointer
-		push		esi						;; preserve register variables
-		push		edi
-		push		ebx
-
-;;
-;; esi = ds_source
-;; ebx = ds_colormap
-;;
-		mov			esi, [ds_source]
-		mov			ebx, [ds_colormap]
-
-;;
-;; edi = ylookup[ds_y] + columnofs[ds_x1]
-;;
-		mov			eax, [ds_y]
-		mov			edi, [ylookup + eax*4]
-		mov			edx, [ds_x1]
-		add			edi, [columnofs + edx*4]
-
-;;
-;; ecx = ds_x2 - ds_x1 + 1
-;;
-		mov			ecx, [ds_x2]
-		sub			ecx, edx
-		add			ecx, 1
-
-;;
-;; Needed for fracs and steps
-;;
-		movd		mm7, [nflatshiftup]
-
-;;
-;; mm3 = xstep
-;;
-		movd		mm3, [ds_xstep]
-		pslld		mm3, mm7
-		punpckldq	mm3, mm3
-
-;;
-;; mm4 = ystep
-;;
-		movd		mm4, [ds_ystep]
-		pslld		mm4, mm7
-		punpckldq	mm4, mm4
-
-;;
-;; mm1 = pair of consecutive xpositions
-;;
-		movd		mm1, [ds_xfrac]
-		pslld		mm1, mm7
-		movq		mm6, mm1
-		paddd		mm6, mm3
-		punpckldq	mm1, mm6
-
-;;
-;; mm2 = pair of consecutive ypositions
-;;
-		movd		mm2, [ds_yfrac]
-		pslld		mm2, mm7
-		movq		mm6, mm2
-		paddd		mm6, mm4
-		punpckldq	mm2, mm6
-
-;;
-;; mm5 = nflatxshift
-;; mm6 = nflatyshift
-;;
-		movd		mm5, [nflatxshift]
-		movd		mm6, [nflatyshift]
-
-;;
-;; Mask is in memory due to lack of registers.
-;;
-		mov			eax, [nflatmask]
-		mov			[nflatmask64], eax
-		mov			[nflatmask64 + 4], eax
-
-
-;;
-;; Go until we reach a dword boundary.
-;;
-.unaligned:
-		test		edi, 3
-		jz			.alignedprep
-.stragglers:
-		cmp			ecx, 0
-		je			.done					;; If ecx == 0, we're finished.
-
-;;
-;; eax = ((yposition >> nflatyshift) & nflatmask) | (xposition >> nflatxshift)
-;;
-		movq		mm0, mm1				;; mm0 = xposition
-		movq		mm7, mm2				;; mm7 = yposition
-		paddd		mm1, mm3				;; xposition += xstep (once!)
-		paddd		mm2, mm4				;; yposition += ystep (once!)
-		psrld		mm0, mm5				;; shift
-		psrld		mm7, mm6				;; shift
-		pand		mm7, [nflatmask64]		;; mask
-		por			mm0, mm7				;; or x and y together
-
-		movd		eax, mm0				;; eax = index of first pixel
-		movzx		eax, byte [esi + eax]	;; al = source[eax]
-		movzx		eax, byte [ebx + eax]	;; al = colormap[al]
-
-		mov			[edi], al
-		add			edi, 1
-
-		sub			ecx, 1
-		jmp			.unaligned
-
-
-.alignedprep:
-;;
-;; We can double the steps now.
-;;
-		pslld		mm3, 1
-		pslld		mm4, 1
-
-
-;;
-;; Generate chunks of four pixels.
-;;
-.alignedloop:
-
-;;
-;; Make sure we have at least four pixels.
-;;
-		cmp			ecx, 4
-		jl			.prestragglers
-
-;;
-;; First two pixels.
-;;
-		movq		mm0, mm1				;; mm0 = xposition
-		movq		mm7, mm2				;; mm7 = yposition
-		paddd		mm1, mm3				;; xposition += xstep
-		paddd		mm2, mm4				;; yposition += ystep
-		psrld		mm0, mm5				;; shift
-		psrld		mm7, mm6				;; shift
-		pand		mm7, [nflatmask64]		;; mask
-		por			mm0, mm7				;; or x and y together
-
-		movd		eax, mm0				;; eax = index of first pixel
-		movzx		eax, byte [esi + eax]	;; al = source[eax]
-		movzx		ebp, byte [ebx + eax]	;; ebp = colormap[al]
-
-		punpckhdq	mm0, mm0				;; both dwords = high dword
-		movd		eax, mm0				;; eax = index of second pixel
-		movzx		eax, byte [esi + eax]	;; al = source[eax]
-		movzx		eax, byte [ebx + eax]	;; al = colormap[al]
-		shl			eax, 8					;; get pixel in right byte
-		or			ebp, eax				;; put pixel in ebp
-
-;;
-;; Next two pixels.
-;;
-		movq		mm0, mm1				;; mm0 = xposition
-		movq		mm7, mm2				;; mm7 = yposition
-		paddd		mm1, mm3				;; xposition += xstep
-		paddd		mm2, mm4				;; yposition += ystep
-		psrld		mm0, mm5				;; shift
-		psrld		mm7, mm6				;; shift
-		pand		mm7, [nflatmask64]		;; mask
-		por			mm0, mm7				;; or x and y together
-
-		movd		eax, mm0				;; eax = index of third pixel
-		movzx		eax, byte [esi + eax]	;; al = source[eax]
-		movzx		eax, byte [ebx + eax]	;; al = colormap[al]
-		shl			eax, 16					;; get pixel in right byte
-		or			ebp, eax				;; put pixel in ebp
-
-		punpckhdq	mm0, mm0				;; both dwords = high dword
-		movd		eax, mm0				;; eax = index of second pixel
-		movzx		eax, byte [esi + eax]	;; al = source[eax]
-		movzx		eax, byte [ebx + eax]	;; al = colormap[al]
-		shl			eax, 24					;; get pixel in right byte
-		or			ebp, eax				;; put pixel in ebp
-
-;;
-;; Write pixels.
-;;
-		mov			[edi], ebp
-		add			edi, 4
-
-		sub			ecx, 4
-		jmp			.alignedloop
-
-.prestragglers:
-;;
-;; Back to one step at a time.
-;;
-		psrad		mm3, 1
-		psrad		mm4, 1
-		jmp			.stragglers
-
-.done:
-;;
-;; Clear MMX state, or else FPU operations will go badly awry.
-;;
-		emms
-
-		pop			ebx
-		pop			edi
-		pop			esi
-		pop			ebp
-		ret

src/tmap_vc.nas

@@ -1,48 +0,0 @@
-;; SONIC ROBO BLAST 2
-;;-----------------------------------------------------------------------------
-;; Copyright (C) 1998-2000 by DooM Legacy Team.
-;; Copyright (C) 1999-2021 by Sonic Team Junior.
-;;
-;; This program is free software distributed under the
-;; terms of the GNU General Public License, version 2.
-;; See the 'LICENSE' file for more details.
-;;-----------------------------------------------------------------------------
-;; FILE:
-;;      tmap_vc.nas
-;; DESCRIPTION:
-;;      Assembler optimised math code for Visual C++.
-
-
-[BITS 32]
-
-%macro cglobal 1
-%define %1 _%1
-[global %1]
-%endmacro
-
-[SECTION .text write]
-
-;----------------------------------------------------------------------------
-;fixed_t FixedMul (fixed_t a, fixed_t b)
-;----------------------------------------------------------------------------
-cglobal FixedMul
-;       align   16
-FixedMul:
-        mov     eax,[esp+4]
-        imul    dword [esp+8]
-        shrd    eax,edx,16
-        ret
-
-;----------------------------------------------------------------------------
-;fixed_t FixedDiv2 (fixed_t a, fixed_t b);
-;----------------------------------------------------------------------------
-cglobal FixedDiv2
-;       align   16
-FixedDiv2:
-        mov     eax,[esp+4]
-        mov     edx,eax                 ;; these two instructions allow the next
-        sar     edx,31                  ;; two to pair, on the Pentium processor.
-        shld    edx,eax,16
-        sal     eax,16
-        idiv    dword [esp+8]
-        ret

src/w_wad.c

@@ -82,6 +82,14 @@
 #define O_BINARY 0
 #endif
 
+#ifdef HAVE_THREADS
+static I_mutex wad_mutex;
+#  define Lock_state()    I_lock_mutex(&wad_mutex)
+#  define Unlock_state() I_unlock_mutex(wad_mutex)
+#else
+#  define Lock_state()
+#  define Unlock_state()
+#endif
 
 typedef struct
 {
@@ -1922,6 +1930,8 @@ void *W_CacheLumpNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 	if (!TestValidLump(wad,lump))
 		return NULL;
 
+	Lock_state();
+
 	lumpcache = wadfiles[wad]->lumpcache;
 	if (!lumpcache[lump])
 	{
@@ -1931,6 +1941,8 @@ void *W_CacheLumpNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 	else
 		Z_ChangeTag(lumpcache[lump], tag);
 
+	Unlock_state();
+
 	return lumpcache[lump];
 }
 
@@ -1955,9 +1967,13 @@ void *W_CacheLumpNumForce(lumpnum_t lumpnum, INT32 tag)
 	if (!TestValidLump(wad,lump))
 		return NULL;
 
+	Lock_state();
+
 	ptr = Z_Malloc(W_LumpLengthPwad(wad, lump), tag, NULL);
 	W_ReadLumpHeaderPwad(wad, lump, ptr, 0, 0);  // read the lump in full
 
+	Unlock_state();
+
 	return ptr;
 }
 
@@ -1975,15 +1991,23 @@ static inline boolean W_IsLumpCachedPWAD(UINT16 wad, UINT16 lump, void *ptr)
 	if (!TestValidLump(wad, lump))
 		return false;
 
+	Lock_state();
+
 	lcache = wadfiles[wad]->lumpcache[lump];
 
 	if (ptr)
 	{
-		if (ptr == lcache)
+		if (ptr == lcache) {
+			Unlock_state();
 			return true;
 		}
-	else if (lcache)
+	}
+	else if (lcache) {
+		Unlock_state();
 		return true;
+	}
+
+	Unlock_state();
 
 	return false;
 }
@@ -2007,15 +2031,23 @@ static inline boolean W_IsPatchCachedPWAD(UINT16 wad, UINT16 lump, void *ptr)
 	if (!TestValidLump(wad, lump))
 		return false;
 
+	Lock_state();
+
 	lcache = wadfiles[wad]->patchcache[lump];
 
 	if (ptr)
 	{
-		if (ptr == lcache)
+		if (ptr == lcache) {
+			Unlock_state();
 			return true;
 		}
-	else if (lcache)
+	}
+	else if (lcache) {
+		Unlock_state();
 		return true;
+	}
+
+	Unlock_state();
 
 	return false;
 }
@@ -2048,7 +2080,7 @@ void *W_CacheLumpName(const char *name, INT32 tag)
 // Cache a patch into heap memory, convert the patch format as necessary
 //
 
-void *W_CacheSoftwarePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
+static void *W_CacheSoftwarePatch(UINT16 wad, UINT16 lump, INT32 tag)
 {
 	lumpcache_t *lumpcache = NULL;
 
@@ -2082,11 +2114,6 @@ void *W_CacheSoftwarePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 	return lumpcache[lump];
 }
 
-void *W_CacheSoftwarePatchNum(lumpnum_t lumpnum, INT32 tag)
-{
-	return W_CacheSoftwarePatchNumPwad(WADFILENUM(lumpnum),LUMPNUM(lumpnum),tag);
-}
-
 void *W_CachePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 {
 	patch_t *patch;
@@ -2094,16 +2121,23 @@ void *W_CachePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 	if (!TestValidLump(wad, lump))
 		return NULL;
 
-	patch = W_CacheSoftwarePatchNumPwad(wad, lump, tag);
+	Lock_state();
+
+	patch = W_CacheSoftwarePatch(wad, lump, tag);
 
 #ifdef HWRENDER
 	// Software-only compile cache the data without conversion
 	if (rendermode == render_soft || rendermode == render_none)
 #endif
+	{
+		Unlock_state();
 		return (void *)patch;
+	}
 
 #ifdef HWRENDER
 	Patch_CreateGL(patch);
+	Unlock_state();
+
 	return (void *)patch;
 #endif
 }
@@ -2118,6 +2152,8 @@ void W_UnlockCachedPatch(void *patch)
 	if (!patch)
 		return;
 
+	Lock_state();
+
 	// The hardware code does its own memory management, as its patches
 	// have different lifetimes from software's.
 #ifdef HWRENDER
@@ -2126,6 +2162,8 @@ void W_UnlockCachedPatch(void *patch)
 	else
 #endif
 		Z_Unlock(patch);
+
+	Unlock_state();
 }
 
 void *W_CachePatchName(const char *name, INT32 tag)

src/w_wad.h

@@ -207,11 +207,6 @@ void *W_CachePatchLongName(const char *name, INT32 tag);
 void *W_CachePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag);
 void *W_CachePatchNum(lumpnum_t lumpnum, INT32 tag);
 
-// Returns a Software patch.
-// Performs any necessary conversions from PNG images.
-void *W_CacheSoftwarePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag);
-void *W_CacheSoftwarePatchNum(lumpnum_t lumpnum, INT32 tag);
-
 void W_UnlockCachedPatch(void *patch);
 
 void W_VerifyFileMD5(UINT16 wadfilenum, const char *matchmd5);

src/z_zone.c

@@ -39,6 +39,15 @@
 #include "hardware/hw_main.h" // For hardware memory info
 #endif
 
+#ifdef HAVE_THREADS
+static I_mutex Z_mutex;
+#  define Lock_state()    I_lock_mutex(&Z_mutex)
+#  define Unlock_state() I_unlock_mutex(Z_mutex)
+#else
+#  define Lock_state()
+#  define Unlock_state()
+#endif
+
 #ifdef HAVE_VALGRIND
 #include "valgrind.h"
 static boolean Z_calloc = false;
@@ -203,6 +212,8 @@ void Z_Free(void *ptr)
 	if (ptr == NULL)
 		return;
 
+	Lock_state();
+
 #ifdef ZDEBUG2
 	CONS_Debug(DBG_MEMORY, "Z_Free %s:%d\n", file, line);
 #endif
@@ -237,6 +248,8 @@ void Z_Free(void *ptr)
 	block->prev->next = block->next;
 	block->next->prev = block->prev;
 	free(block);
+
+	Unlock_state();
 }
 
 /** malloc() that doesn't accept failure.
@@ -295,6 +308,8 @@ void *Z_MallocAlign(size_t size, INT32 tag, void *user, INT32 alignbits)
 	void *given;
 	size_t blocksize = extrabytes + sizeof *hdr + size;
 
+	Lock_state();
+
 #ifdef ZDEBUG2
 	CONS_Debug(DBG_MEMORY, "Z_Malloc %s:%d\n", file, line);
 #endif
@@ -359,6 +374,8 @@ void *Z_MallocAlign(size_t size, INT32 tag, void *user, INT32 alignbits)
 		I_Error("Z_Malloc: attempted to allocate purgable block "
 			"(size %s) with no user", sizeu1(size));
 
+	Unlock_state();
+
 	return given;
 }
 
@@ -381,14 +398,19 @@ void *Z_Calloc2(size_t size, INT32 tag, void *user, INT32 alignbits, const char
 void *Z_CallocAlign(size_t size, INT32 tag, void *user, INT32 alignbits)
 #endif
 {
+	void *mem;
+	Lock_state();
 #ifdef VALGRIND_MEMPOOL_ALLOC
 	Z_calloc = true;
 #endif
 #ifdef ZDEBUG
-	return memset(Z_Malloc2    (size, tag, user, alignbits, file, line), 0, size);
+	mem = Z_Malloc2    (size, tag, user, alignbits, file, line);
 #else
-	return memset(Z_MallocAlign(size, tag, user, alignbits            ), 0, size);
+	mem = Z_MallocAlign(size, tag, user, alignbits            );
 #endif
+	memset(mem, 0, size);
+	Unlock_state();
+	return mem;
 }
 
 /** The Z_ReallocAlign function.