Merge pull request #317 from 0lvin/render_speedup_review

Cleanup code in soft render

src/client/menu/videomenu.c

@@ -42,7 +42,7 @@ extern cvar_t *vid_gamma;
 extern cvar_t *vid_fullscreen;
 extern cvar_t *vid_renderer;
 static cvar_t *r_vsync;
-static cvar_t *r_anisotropic;
+static cvar_t *gl_anisotropic;
 static cvar_t *gl_msaa_samples;
 
 static menuframework_s s_opengl_menu;
@@ -126,11 +126,11 @@ AnisotropicCallback(void *s)
 
 	if (list->curvalue == 0)
 	{
-		Cvar_SetValue("r_anisotropic", 0);
+		Cvar_SetValue("gl_anisotropic", 0);
 	}
 	else
 	{
-		Cvar_SetValue("r_anisotropic", pow(2, list->curvalue));
+		Cvar_SetValue("gl_anisotropic", pow(2, list->curvalue));
 	}
 }
 
@@ -362,9 +362,9 @@ VID_MenuInit(void)
 		r_vsync = Cvar_Get("r_vsync", "1", CVAR_ARCHIVE);
 	}
 
-	if (!r_anisotropic)
+	if (!gl_anisotropic)
 	{
-		r_anisotropic = Cvar_Get("r_anisotropic", "0", CVAR_ARCHIVE);
+		gl_anisotropic = Cvar_Get("gl_anisotropic", "0", CVAR_ARCHIVE);
 	}
 
 	if (!gl_msaa_samples)
@@ -462,13 +462,13 @@ VID_MenuInit(void)
 	s_af_list.generic.callback = AnisotropicCallback;
 	s_af_list.itemnames = pow2_names;
 	s_af_list.curvalue = 0;
-	if (r_anisotropic->value)
+	if (gl_anisotropic->value)
 	{
 		do
 		{
 			s_af_list.curvalue++;
 		} while (pow2_names[s_af_list.curvalue] &&
-				pow(2, s_af_list.curvalue) <= r_anisotropic->value);
+				pow(2, s_af_list.curvalue) <= gl_anisotropic->value);
 		s_af_list.curvalue--;
 	}
 

src/client/refresh/gl1/gl1_image.c

@@ -202,18 +202,18 @@ R_TextureMode(char *string)
 	/* clamp selected anisotropy */
 	if (gl_config.anisotropic)
 	{
-		if (r_anisotropic->value > gl_config.max_anisotropy)
+		if (gl_anisotropic->value > gl_config.max_anisotropy)
 		{
-			ri.Cvar_SetValue("r_anisotropic", gl_config.max_anisotropy);
+			ri.Cvar_SetValue("gl_anisotropic", gl_config.max_anisotropy);
 		}
-		else if (r_anisotropic->value < 1.0)
+		else if (gl_anisotropic->value < 1.0)
 		{
-			ri.Cvar_SetValue("r_anisotropic", 1.0);
+			ri.Cvar_SetValue("gl_anisotropic", 1.0);
 		}
 	}
 	else
 	{
-		ri.Cvar_SetValue("r_anisotropic", 0.0);
+		ri.Cvar_SetValue("gl_anisotropic", 0.0);
 	}
 
 	/* change all the existing mipmap texture objects */
@@ -228,10 +228,10 @@ R_TextureMode(char *string)
 					gl_filter_max);
 
 			/* Set anisotropic filter if supported and enabled */
-			if (gl_config.anisotropic && r_anisotropic->value)
+			if (gl_config.anisotropic && gl_anisotropic->value)
 			{
 				glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT,
-						r_anisotropic->value);
+						gl_anisotropic->value);
 			}
 		}
 	}
@@ -788,10 +788,10 @@ R_Upload32(unsigned *data, int width, int height, qboolean mipmap)
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, gl_filter_max);
 	}
 
-	if (mipmap && gl_config.anisotropic && r_anisotropic->value)
+	if (mipmap && gl_config.anisotropic && gl_anisotropic->value)
 	{
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT,
-				r_anisotropic->value);
+				gl_anisotropic->value);
 	}
 	return res;
 }

src/client/refresh/gl1/gl1_main.c

@@ -124,7 +124,7 @@ cvar_t *r_vsync;
 cvar_t *gl_texturemode;
 cvar_t *gl1_texturealphamode;
 cvar_t *gl1_texturesolidmode;
-cvar_t *r_anisotropic;
+cvar_t *gl_anisotropic;
 cvar_t *r_lockpvs;
 cvar_t *gl_msaa_samples;
 
@@ -1241,7 +1241,7 @@ R_Register(void)
 	gl_texturemode = ri.Cvar_Get("gl_texturemode", "GL_LINEAR_MIPMAP_NEAREST", CVAR_ARCHIVE);
 	gl1_texturealphamode = ri.Cvar_Get("gl1_texturealphamode", "default", CVAR_ARCHIVE);
 	gl1_texturesolidmode = ri.Cvar_Get("gl1_texturesolidmode", "default", CVAR_ARCHIVE);
-	r_anisotropic = ri.Cvar_Get("r_anisotropic", "0", CVAR_ARCHIVE);
+	gl_anisotropic = ri.Cvar_Get("gl_anisotropic", "0", CVAR_ARCHIVE);
 	r_lockpvs = ri.Cvar_Get("r_lockpvs", "0", 0);
 
 	gl1_palettedtexture = ri.Cvar_Get("gl1_palettedtexture", "0", CVAR_ARCHIVE);
@@ -1676,11 +1676,11 @@ RI_BeginFrame(float camera_separation)
 	}
 
 	/* texturemode stuff */
-	if (gl_texturemode->modified || (gl_config.anisotropic && r_anisotropic->modified))
+	if (gl_texturemode->modified || (gl_config.anisotropic && gl_anisotropic->modified))
 	{
 		R_TextureMode(gl_texturemode->string);
 		gl_texturemode->modified = false;
-		r_anisotropic->modified = false;
+		gl_anisotropic->modified = false;
 	}
 
 	if (gl1_texturealphamode->modified)

src/client/refresh/gl1/header/local.h

@@ -215,7 +215,7 @@ extern cvar_t *gl1_flashblend;
 extern cvar_t *r_modulate;
 extern cvar_t *gl_drawbuffer;
 extern cvar_t *r_vsync;
-extern cvar_t *r_anisotropic;
+extern cvar_t *gl_anisotropic;
 extern cvar_t *gl_texturemode;
 extern cvar_t *gl1_texturealphamode;
 extern cvar_t *gl1_texturesolidmode;

src/client/refresh/gl3/gl3_image.c

@@ -74,18 +74,18 @@ GL3_TextureMode(char *string)
 	/* clamp selected anisotropy */
 	if (gl3config.anisotropic)
 	{
-		if (r_anisotropic->value > gl3config.max_anisotropy)
+		if (gl_anisotropic->value > gl3config.max_anisotropy)
 		{
-			ri.Cvar_SetValue("r_anisotropic", gl3config.max_anisotropy);
+			ri.Cvar_SetValue("gl_anisotropic", gl3config.max_anisotropy);
 		}
-		else if (r_anisotropic->value < 1.0)
+		else if (gl_anisotropic->value < 1.0)
 		{
-			ri.Cvar_SetValue("r_anisotropic", 1.0);
+			ri.Cvar_SetValue("gl_anisotropic", 1.0);
 		}
 	}
 	else
 	{
-		ri.Cvar_SetValue("r_anisotropic", 0.0);
+		ri.Cvar_SetValue("gl_anisotropic", 0.0);
 	}
 
 	gl3image_t *glt;
@@ -101,9 +101,9 @@ GL3_TextureMode(char *string)
 			glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, gl_filter_max);
 
 			/* Set anisotropic filter if supported and enabled */
-			if (gl3config.anisotropic && r_anisotropic->value)
+			if (gl3config.anisotropic && gl_anisotropic->value)
 			{
-				glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, r_anisotropic->value);
+				glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, gl_anisotropic->value);
 			}
 		}
 	}
@@ -200,9 +200,9 @@ GL3_Upload32(unsigned *data, int width, int height, qboolean mipmap)
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, gl_filter_max);
 	}
 
-	if (mipmap && gl3config.anisotropic && r_anisotropic->value)
+	if (mipmap && gl3config.anisotropic && gl_anisotropic->value)
 	{
-		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, r_anisotropic->value);
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, gl_anisotropic->value);
 	}
 
 	return res;

src/client/refresh/gl3/gl3_main.c

@@ -87,7 +87,7 @@ cvar_t *r_mode;
 cvar_t *r_customwidth;
 cvar_t *r_customheight;
 cvar_t *vid_gamma;
-cvar_t *r_anisotropic;
+cvar_t *gl_anisotropic;
 cvar_t *gl_texturemode;
 cvar_t *gl_drawbuffer;
 cvar_t *r_clear;
@@ -216,7 +216,7 @@ GL3_Register(void)
 	gl_nobind = ri.Cvar_Get("gl_nobind", "0", 0);
 
 	gl_texturemode = ri.Cvar_Get("gl_texturemode", "GL_LINEAR_MIPMAP_NEAREST", CVAR_ARCHIVE);
-	r_anisotropic = ri.Cvar_Get("r_anisotropic", "0", CVAR_ARCHIVE);
+	gl_anisotropic = ri.Cvar_Get("gl_anisotropic", "0", CVAR_ARCHIVE);
 
 	vid_fullscreen = ri.Cvar_Get("vid_fullscreen", "0", CVAR_ARCHIVE);
 	vid_gamma = ri.Cvar_Get("vid_gamma", "1.2", CVAR_ARCHIVE);
@@ -276,7 +276,7 @@ GL3_Register(void)
 	//gl_texturemode = ri.Cvar_Get("gl_texturemode", "GL_LINEAR_MIPMAP_NEAREST", CVAR_ARCHIVE);
 	gl1_texturealphamode = ri.Cvar_Get("gl1_texturealphamode", "default", CVAR_ARCHIVE);
 	gl1_texturesolidmode = ri.Cvar_Get("gl1_texturesolidmode", "default", CVAR_ARCHIVE);
-	//r_anisotropic = ri.Cvar_Get("r_anisotropic", "0", CVAR_ARCHIVE);
+	//gl_anisotropic = ri.Cvar_Get("gl_anisotropic", "0", CVAR_ARCHIVE);
 	//r_lockpvs = ri.Cvar_Get("r_lockpvs", "0", 0);
 
 	//gl1_palettedtexture = ri.Cvar_Get("gl1_palettedtexture", "0", CVAR_ARCHIVE); NOPE.
@@ -1657,11 +1657,11 @@ GL3_BeginFrame(float camera_separation)
 	}
 
 	/* texturemode stuff */
-	if (gl_texturemode->modified || (gl3config.anisotropic && r_anisotropic->modified))
+	if (gl_texturemode->modified || (gl3config.anisotropic && gl_anisotropic->modified))
 	{
 		GL3_TextureMode(gl_texturemode->string);
 		gl_texturemode->modified = false;
-		r_anisotropic->modified = false;
+		gl_anisotropic->modified = false;
 	}
 
 	if (r_vsync->modified)

src/client/refresh/gl3/header/local.h

@@ -498,7 +498,7 @@ extern cvar_t *r_drawworld;
 extern cvar_t *vid_gamma;
 extern cvar_t *gl3_intensity;
 extern cvar_t *gl3_intensity_2D;
-extern cvar_t *r_anisotropic;
+extern cvar_t *gl_anisotropic;
 
 extern cvar_t *r_lightlevel;
 extern cvar_t *gl3_overbrightbits;

src/client/refresh/soft/header/local.h

@@ -375,7 +375,6 @@ extern int	d_pix_min, d_pix_max, d_pix_mul;
 
 extern pixel_t	*d_viewbuffer;
 extern zvalue_t	*d_pzbuffer;
-extern unsigned int d_zwidth;
 
 extern int	d_minmip;
 extern float	d_scalemip[NUM_MIPS-1];
@@ -384,7 +383,6 @@ extern float	d_scalemip[NUM_MIPS-1];
 
 extern int	cachewidth;
 extern pixel_t	*cacheblock;
-extern int	r_screenwidth;
 
 extern int	r_drawnpolycount;
 
@@ -508,7 +506,7 @@ typedef struct {
 	int		sfrac, tfrac, light;
 	zvalue_t	zi;
 } spanpackage_t;
-extern spanpackage_t	*triangle_spans;
+extern spanpackage_t	*triangle_spans, *triangles_max;
 
 extern byte	**warp_rowptr;
 extern int	*warp_column;
@@ -523,6 +521,7 @@ extern float	aliasxscale, aliasyscale, aliasxcenter, aliasycenter;
 extern int	r_outofsurfaces;
 extern int	r_outofedges;
 extern int	r_outofverts;
+extern int	r_outoftriangles;
 
 extern mvertex_t	*r_pcurrentvertbase;
 

src/client/refresh/soft/sw_edge.c

@@ -786,7 +786,7 @@ D_FlatFillSurface (surf_t *surf, pixel_t color)
 	{
 		pixel_t   *pdest;
 
-		pdest = d_viewbuffer + r_screenwidth*span->v + span->u;
+		pdest = d_viewbuffer + vid.width*span->v + span->u;
 		memset(pdest,  color&0xFF, span->count * sizeof(pixel_t));
 	}
 }

src/client/refresh/soft/sw_main.c

@@ -59,10 +59,6 @@ typedef struct swstate_s
 
 	unsigned char	gammatable[256];
 	unsigned char	currentpalette[1024];
-
-	// SDL colors
-	Uint32 	palette_colors[256];
-
 } swstate_t;
 
 static swstate_t sw_state;
@@ -71,10 +67,12 @@ void	*colormap;
 float	r_time1;
 int	r_numallocatededges;
 int	r_numallocatedverts;
+int	r_numallocatedtriangles;
 float	r_aliasuvscale = 1.0;
 int	r_outofsurfaces;
 int	r_outofedges;
 int	r_outofverts;
+int	r_outoftriangles;
 
 qboolean	r_dowarp;
 
@@ -101,8 +99,6 @@ float		xscaleinv, yscaleinv;
 float		xscaleshrink, yscaleshrink;
 float		aliasxscale, aliasyscale, aliasxcenter, aliasycenter;
 
-int		r_screenwidth;
-
 mplane_t	screenedge[4];
 
 //
@@ -137,7 +133,7 @@ cvar_t	*sw_surfcacheoverride;
 cvar_t	*sw_waterwarp;
 static cvar_t	*sw_overbrightbits;
 cvar_t	*sw_custom_particles;
-cvar_t	*r_anisotropic;
+cvar_t	*sw_texture_filtering;
 
 cvar_t	*r_drawworld;
 static cvar_t	*r_drawentities;
@@ -197,12 +193,9 @@ pixel_t		*cacheblock;
 int		cachewidth;
 pixel_t		*d_viewbuffer;
 zvalue_t	*d_pzbuffer;
-unsigned int	d_zwidth;
 
 qboolean	insubmodel;
 
-static qboolean	sdl_palette_outdated;
-
 static struct texture_buffer {
 	image_t	image;
 	byte	buffer[1024];
@@ -290,7 +283,7 @@ R_Register (void)
 	sw_waterwarp = ri.Cvar_Get ("sw_waterwarp", "1", 0);
 	sw_overbrightbits = ri.Cvar_Get("sw_overbrightbits", "1.0", CVAR_ARCHIVE);
 	sw_custom_particles = ri.Cvar_Get("sw_custom_particles", "0", CVAR_ARCHIVE);
-	r_anisotropic = ri.Cvar_Get("r_anisotropic", "0", CVAR_ARCHIVE);
+	sw_texture_filtering = ri.Cvar_Get("sw_texture_filtering", "0", CVAR_ARCHIVE);
 	r_mode = ri.Cvar_Get( "r_mode", "0", CVAR_ARCHIVE );
 
 	r_lefthand = ri.Cvar_Get( "hand", "0", CVAR_USERINFO | CVAR_ARCHIVE );
@@ -332,7 +325,9 @@ R_UnRegister (void)
 	ri.Cmd_RemoveCommand( "imagelist" );
 }
 
-static void SWimp_DestroyRender(void);
+static void RE_ShutdownContext(void);
+static void SWimp_CreateRender(void);
+static int RE_InitContext(void *win);
 
 /*
 ===============
@@ -402,7 +397,7 @@ RE_Shutdown (void)
 	Mod_FreeAll ();
 	R_ShutdownImages ();
 
-	SWimp_DestroyRender();
+	RE_ShutdownContext();
 }
 
 /*
@@ -528,6 +523,35 @@ R_ReallocateMapBuffers (void)
 
 		R_Printf(PRINT_ALL, "Allocated %d verts\n", r_numallocatedverts);
 	}
+
+	if (!r_numallocatedtriangles || r_outoftriangles)
+	{
+		if (triangle_spans)
+		{
+			free(triangle_spans);
+		}
+
+		if (r_outoftriangles)
+		{
+			//R_Printf(PRINT_ALL, "%s: not enough %d(+%d) triangles\n",
+			//		    __func__, r_numallocatedtriangles, r_outoftriangles);
+			r_numallocatedtriangles *= 2;
+		}
+
+		if (r_numallocatedtriangles < vid.height)
+			r_numallocatedtriangles = vid.height;
+
+		triangle_spans  = malloc(r_numallocatedtriangles * sizeof(spanpackage_t));
+		if (!triangle_spans)
+		{
+			R_Printf(PRINT_ALL, "%s: Couldn't malloc %d bytes\n",
+				 __func__, (int)(r_numallocatedtriangles * sizeof(spanpackage_t)));
+			return;
+		}
+		triangles_max = &triangle_spans[r_numallocatedtriangles];
+
+		R_Printf(PRINT_ALL, "Allocated %d triangles\n", r_numallocatedtriangles);
+	}
 }
 
 
@@ -1269,12 +1293,11 @@ R_GammaCorrectAndSetPalette( const unsigned char *palette )
 
 	for ( i = 0; i < 256; i++ )
 	{
-		sw_state.currentpalette[i*4+0] = sw_state.gammatable[palette[i*4+0]];
+		sw_state.currentpalette[i*4+0] = sw_state.gammatable[palette[i*4+2]]; // blue
-		sw_state.currentpalette[i*4+1] = sw_state.gammatable[palette[i*4+1]];
+		sw_state.currentpalette[i*4+1] = sw_state.gammatable[palette[i*4+1]]; // green
-		sw_state.currentpalette[i*4+2] = sw_state.gammatable[palette[i*4+2]];
+		sw_state.currentpalette[i*4+2] = sw_state.gammatable[palette[i*4+0]]; // red
+		sw_state.currentpalette[i*4+3] = 0xFF; // alpha
 	}
-
-	sdl_palette_outdated = true;
 }
 
 /*
@@ -1508,6 +1531,12 @@ RE_IsVsyncActive(void)
 	}
 }
 
+static int RE_PrepareForWindow(void)
+{
+	int flags = SDL_SWSURFACE;
+	return flags;
+}
+
 /*
 ===============
 GetRefAPI
@@ -1546,6 +1575,9 @@ GetRefAPI(refimport_t imp)
 	re.Init = RE_Init;
 	re.IsVSyncActive = RE_IsVsyncActive;
 	re.Shutdown = RE_Shutdown;
+	re.InitContext = RE_InitContext;
+	re.ShutdownContext = RE_ShutdownContext;
+	re.PrepareForWindow = RE_PrepareForWindow;
 
 	re.SetPalette = RE_SetPalette;
 	re.BeginFrame = RE_BeginFrame;
@@ -1571,106 +1603,26 @@ GetRefAPI(refimport_t imp)
  */
 
 static SDL_Window	*window = NULL;
-static SDL_Surface	*surface = NULL;
 static SDL_Texture	*texture = NULL;
 static SDL_Renderer	*renderer = NULL;
 
-/*
- * Sets the window icon
- */
-static void
-SetSDLIcon()
-{
-	/* The 64x64 32bit window icon */
-	#include "../../vid/icon/q2icon64.h"
-
-	/* these masks are needed to tell SDL_CreateRGBSurface(From)
-	   to assume the data it gets is byte-wise RGB(A) data */
-	Uint32 rmask, gmask, bmask, amask;
-#if SDL_BYTEORDER == SDL_BIG_ENDIAN
-	int shift = (q2icon64.bytes_per_pixel == 3) ? 8 : 0;
-	rmask = 0xff000000 >> shift;
-	gmask = 0x00ff0000 >> shift;
-	bmask = 0x0000ff00 >> shift;
-	amask = 0x000000ff >> shift;
-#else /* little endian, like x86 */
-	rmask = 0x000000ff;
-	gmask = 0x0000ff00;
-	bmask = 0x00ff0000;
-	amask = (q2icon64.bytes_per_pixel == 3) ? 0 : 0xff000000;
-#endif
-
-	SDL_Surface* icon = SDL_CreateRGBSurfaceFrom((void*)q2icon64.pixel_data, q2icon64.width,
-		q2icon64.height, q2icon64.bytes_per_pixel*8, q2icon64.bytes_per_pixel*q2icon64.width,
-		rmask, gmask, bmask, amask);
-
-	SDL_SetWindowIcon(window, icon);
-
-	SDL_FreeSurface(icon);
-}
-
 static int
-IsFullscreen()
+RE_InitContext(void *win)
-{
-	if (SDL_GetWindowFlags(window) & SDL_WINDOW_FULLSCREEN_DESKTOP) {
-		return 1;
-	} else if (SDL_GetWindowFlags(window) & SDL_WINDOW_FULLSCREEN) {
-		return 2;
-	} else {
-		return 0;
-	}
-}
-
-static qboolean
-GetWindowSize(int* w, int* h)
-{
-	if(window == NULL || w == NULL || h == NULL)
-		return false;
-
-	SDL_DisplayMode m;
-	if(SDL_GetWindowDisplayMode(window, &m) != 0)
-	{
-		Com_Printf("Can't get Displaymode: %s\n", SDL_GetError());
-		return false;
-	}
-	*w = m.w;
-	*h = m.h;
-
-	return true;
-}
-
-static int
-R_InitContext(SDL_Window *win)
 {
 	char title[40] = {0};
 
 	if(win == NULL)
 	{
-		ri.Sys_Error(ERR_FATAL, "R_InitContext() must not be called with NULL argument!");
+		ri.Sys_Error(ERR_FATAL, "RE_InitContext() must not be called with NULL argument!");
 		return false;
 	}
 
-	window = win;
+	window = (SDL_Window *)win;
 
 	/* Window title - set here so we can display renderer name in it */
 	snprintf(title, sizeof(title), "Yamagi Quake II %s - Soft Render", YQ2VERSION);
 	SDL_SetWindowTitle(window, title);
 
-	return true;
-}
-
-static qboolean
-CreateSDLWindow(int flags, int w, int h)
-{
-	Uint32 Rmask, Gmask, Bmask, Amask;
-	int bpp;
-	int windowPos = SDL_WINDOWPOS_UNDEFINED;
-	if (!SDL_PixelFormatEnumToMasks(SDL_PIXELFORMAT_ARGB8888, &bpp, &Rmask, &Gmask, &Bmask, &Amask))
-		return 0;
-
-	// TODO: support fullscreen on different displays with SDL_WINDOWPOS_UNDEFINED_DISPLAY(displaynum)
-	window = SDL_CreateWindow("Yamagi Quake II", windowPos, windowPos, w, h, flags);
-
 	if (r_vsync->value)
 	{
 		renderer = SDL_CreateRenderer(window, -1, SDL_RENDERER_ACCELERATED | SDL_RENDERER_PRESENTVSYNC);
@@ -1680,17 +1632,30 @@ CreateSDLWindow(int flags, int w, int h)
 		renderer = SDL_CreateRenderer(window, -1, SDL_RENDERER_ACCELERATED);
 	}
 
-	surface = SDL_CreateRGBSurface(0, w, h, bpp, Rmask, Gmask, Bmask, Amask);
+	/* Select the color for drawing. It is set to black here. */
+	SDL_SetRenderDrawColor(renderer, 0, 0, 0, 0);
+
+	/* Clear the entire screen to our selected color. */
+	SDL_RenderClear(renderer);
+
+	/* Up until now everything was drawn behind the scenes.
+	   This will show the new, black contents of the window. */
+	SDL_RenderPresent(renderer);
 
 	texture = SDL_CreateTexture(renderer,
+#if SDL_BYTEORDER == SDL_BIG_ENDIAN
+				    SDL_PIXELFORMAT_BGRA8888,
+#else
 				    SDL_PIXELFORMAT_ARGB8888,
+#endif
 				    SDL_TEXTUREACCESS_STREAMING,
-				    w, h);
+				    vid.width, vid.height);
-	return window != NULL;
+
+	return true;
 }
 
 static void
-SWimp_DestroyRender(void)
+RE_ShutdownContext(void)
 {
 	if (vid_buffer)
 	{
@@ -1788,24 +1753,11 @@ SWimp_DestroyRender(void)
 	}
 	texture = NULL;
 
-	if (surface)
-	{
-		SDL_FreeSurface(surface);
-	}
-	surface = NULL;
-
 	if (renderer)
 	{
 		SDL_DestroyRenderer(renderer);
 	}
 	renderer = NULL;
-
-	/* Is the surface used? */
-	if (window)
-	{
-		SDL_DestroyWindow(window);
-	}
-	window = NULL;
 }
 
 /*
@@ -1814,165 +1766,10 @@ point math used in R_ScanEdges() overflows at width 2048 !!
 */
 char shift_size;
 
-/*
-** SWimp_InitGraphics
-**
-** This initializes the software refresh's implementation specific
-** graphics subsystem.  In the case of Windows it creates DIB or
-** DDRAW surfaces.
-**
-** The necessary width and height parameters are grabbed from
-** vid.width and vid.height.
-*/
-static qboolean
-SWimp_InitGraphics(int fullscreen, int *pwidth, int *pheight)
-{
-	int flags;
-	int curWidth, curHeight;
-	int width = *pwidth;
-	int height = *pheight;
-	unsigned int fs_flag = 0;
-
-	if (fullscreen == 1) {
-		fs_flag = SDL_WINDOW_FULLSCREEN_DESKTOP;
-	} else if (fullscreen == 2) {
-		fs_flag = SDL_WINDOW_FULLSCREEN;
-	}
-
-	if (GetWindowSize(&curWidth, &curHeight) && (curWidth == width) && (curHeight == height))
-	{
-		/* If we want fullscreen, but aren't */
-		if (fullscreen != IsFullscreen())
-		{
-			SDL_SetWindowFullscreen(window, fs_flag);
-
-			ri.Cvar_SetValue("vid_fullscreen", fullscreen);
-		}
-
-		/* Are we now? */
-		if (fullscreen == IsFullscreen())
-		{
-			return true;
-		}
-	}
-
-	SWimp_DestroyRender();
-
-	// let the sound and input subsystems know about the new window
-	ri.Vid_NewWindow (vid.width, vid.height);
-
-	flags = SDL_SWSURFACE;
-	if (fs_flag)
-	{
-		flags |= fs_flag;
-	}
-
-	while (1)
-	{
-		if (!CreateSDLWindow(flags, width, height))
-		{
-			Sys_Error("(SOFTSDL) SDL SetVideoMode failed: %s\n", SDL_GetError());
-			return false;
-		}
-		else
-		{
-			break;
-		}
-	}
-
-	if(!R_InitContext(window))
-	{
-		// InitContext() should have logged an error
-		return false;
-	}
-
-	/* Note: window title is now set in re.InitContext() to include renderer name */
-	/* Set the window icon - For SDL2, this must be done after creating the window */
-	SetSDLIcon();
-
-	/* No cursor */
-	SDL_ShowCursor(0);
-
-	vid_buffer = malloc(vid.height * vid.width * sizeof(pixel_t));
-
-	sintable = malloc((vid.width+CYCLE) * sizeof(int));
-	intsintable = malloc((vid.width+CYCLE) * sizeof(int));
-	blanktable = malloc((vid.width+CYCLE) * sizeof(int));
-
-	newedges = malloc(vid.width * sizeof(edge_t *));
-	removeedges = malloc(vid.width * sizeof(edge_t *));
-
-	// 1 extra for spanpackage that marks end
-	triangle_spans = malloc((vid.width + 1) * sizeof(spanpackage_t));
-
-	warp_rowptr = malloc((vid.width+AMP2*2) * sizeof(byte*));
-	warp_column = malloc((vid.width+AMP2*2) * sizeof(int));
-
-	edge_basespans = malloc((vid.width*2) * sizeof(espan_t));
-
-	// count of "out of items"
-	r_outofsurfaces = r_outofedges = r_outofverts = 0;
-	// pointers to allocated buffers
-	finalverts = NULL;
-	r_edges = NULL;
-	lsurfs = NULL;
-	// curently allocated items
-	r_cnumsurfs = r_numallocatededges = r_numallocatedverts = 0;
-
-	R_ReallocateMapBuffers();
-
-	r_warpbuffer = malloc(vid.height * vid.width * sizeof(pixel_t));
-
-	if ((vid.width >= 2048) && (sizeof(shift20_t) == 4)) // 2k+ resolution and 32 == shift20_t
-	{
-		shift_size = 18;
-	}
-	else
-	{
-		shift_size = 20;
-	}
-
-	R_InitTurb ();
-
-	vid_polygon_spans = malloc(sizeof(espan_t) * (vid.height + 1));
-
-	memset(sw_state.currentpalette, 0, sizeof(sw_state.currentpalette));
-	memset(sw_state.palette_colors, 0, sizeof(sw_state.palette_colors));
-
-	sdl_palette_outdated = true;
-
-	return true;
-}
-
-
-static void
-RE_SDLPaletteConvert (void)
-{
-	int i;
-	const unsigned char *palette = sw_state.currentpalette;
-	Uint32 *sdl_palette = sw_state.palette_colors;
-
-	if (!sdl_palette_outdated)
-		return;
-
-	sdl_palette_outdated = false;
-	for ( i = 0; i < 256; i++ )
-	{
-		if (surface)
-		{
-			sdl_palette[i] = SDL_MapRGB(surface->format,
-					palette[i * 4 + 0], // red
-					palette[i * 4 + 1], // green
-					palette[i * 4 + 2]  //blue
-					);
-		}
-	}
-}
-
 static void
 RE_CopyFrame (Uint32 * pixels, int pitch)
 {
-	RE_SDLPaletteConvert();
+	Uint32 *sdl_palette = (Uint32 *)sw_state.currentpalette;
 
 	// no gaps between images rows
 	if (pitch == vid.width)
@@ -1986,7 +1783,7 @@ RE_CopyFrame (Uint32 * pixels, int pitch)
 
 		for (pixels_pos = pixels; pixels_pos < max_pixels; pixels_pos++)
 		{
-			*pixels_pos = sw_state.palette_colors[*buffer_pos];
+			*pixels_pos = sdl_palette[*buffer_pos];
 			buffer_pos++;
 		}
 	}
@@ -1999,7 +1796,7 @@ RE_CopyFrame (Uint32 * pixels, int pitch)
 		{
 			for (x=0; x < vid.width; x ++)
 			{
-				pixels[x] = sw_state.palette_colors[vid_buffer[buffer_pos + x]];
+				pixels[x] = sdl_palette[vid_buffer[buffer_pos + x]];
 			}
 			pixels += pitch;
 			buffer_pos += vid.width;
@@ -2019,14 +1816,16 @@ static void
 RE_EndFrame (void)
 {
 	int pitch;
+	Uint32 * pixels;
 
-	Uint32 * pixels = (Uint32 *)surface->pixels;
+	if (SDL_LockTexture(texture, NULL, (void**)&pixels, &pitch))
-	pitch = surface->pitch / sizeof(Uint32);
+	{
+		Com_Printf("Can't lock texture: %s\n", SDL_GetError());
+		return;
+	}
+	RE_CopyFrame (pixels, pitch / sizeof(Uint32));
+	SDL_UnlockTexture(texture);
 
-	RE_CopyFrame (pixels, pitch);
-
-	SDL_UpdateTexture(texture, NULL, surface->pixels, surface->pitch);
-	SDL_RenderClear(renderer);
 	SDL_RenderCopy(renderer, texture, NULL, NULL);
 	SDL_RenderPresent(renderer);
 }
@@ -2049,16 +1848,66 @@ SWimp_SetMode(int *pwidth, int *pheight, int mode, int fullscreen )
 
 	R_Printf( PRINT_ALL, " %d %d\n", *pwidth, *pheight);
 
-	if ( !SWimp_InitGraphics(fullscreen, pwidth, pheight) ) {
+	if (!ri.GLimp_InitGraphics(fullscreen, pwidth, pheight))
+	{
 		// failed to set a valid mode in windowed mode
 		return rserr_invalid_mode;
 	}
 
-	R_GammaCorrectAndSetPalette( ( const unsigned char * ) d_8to24table );
+	SWimp_CreateRender();
 
 	return retval;
 }
 
+static void
+SWimp_CreateRender(void)
+{
+	vid_buffer = malloc(vid.height * vid.width * sizeof(pixel_t));
+
+	sintable = malloc((vid.width+CYCLE) * sizeof(int));
+	intsintable = malloc((vid.width+CYCLE) * sizeof(int));
+	blanktable = malloc((vid.width+CYCLE) * sizeof(int));
+
+	newedges = malloc(vid.width * sizeof(edge_t *));
+	removeedges = malloc(vid.width * sizeof(edge_t *));
+
+	warp_rowptr = malloc((vid.width+AMP2*2) * sizeof(byte*));
+	warp_column = malloc((vid.width+AMP2*2) * sizeof(int));
+
+	edge_basespans = malloc((vid.width*2) * sizeof(espan_t));
+
+	// count of "out of items"
+	r_outofsurfaces = r_outofedges = r_outofverts = r_outoftriangles = 0;
+	// pointers to allocated buffers
+	finalverts = NULL;
+	r_edges = NULL;
+	lsurfs = NULL;
+	triangle_spans = NULL;
+	// curently allocated items
+	r_cnumsurfs = r_numallocatededges = r_numallocatedverts = r_numallocatedtriangles = 0;
+
+	R_ReallocateMapBuffers();
+
+	r_warpbuffer = malloc(vid.height * vid.width * sizeof(pixel_t));
+
+	if ((vid.width >= 2048) && (sizeof(shift20_t) == 4)) // 2k+ resolution and 32 == shift20_t
+	{
+		shift_size = 18;
+	}
+	else
+	{
+		shift_size = 20;
+	}
+
+	R_InitTurb ();
+
+	vid_polygon_spans = malloc(sizeof(espan_t) * (vid.height + 1));
+
+	memset(sw_state.currentpalette, 0, sizeof(sw_state.currentpalette));
+
+	R_GammaCorrectAndSetPalette( ( const unsigned char * ) d_8to24table );
+}
+
 // this is only here so the functions in q_shared.c and q_shwin.c can link
 void
 Sys_Error (char *error, ...)
@@ -2116,9 +1965,9 @@ R_ScreenShot_f(void)
 	{
 		for (y=0; y < vid.height; y ++) {
 			int buffer_pos = y * vid.width + x;
-			buffer[buffer_pos * 3 + 0] = palette[vid_buffer[buffer_pos] * 4 + 0]; // red
+			buffer[buffer_pos * 3 + 0] = palette[vid_buffer[buffer_pos] * 4 + 2]; // red
 			buffer[buffer_pos * 3 + 1] = palette[vid_buffer[buffer_pos] * 4 + 1]; // green
-			buffer[buffer_pos * 3 + 2] = palette[vid_buffer[buffer_pos] * 4 + 2]; // blue
+			buffer[buffer_pos * 3 + 2] = palette[vid_buffer[buffer_pos] * 4 + 0]; // blue
 		}
 	}
 

src/client/refresh/soft/sw_misc.c

@@ -47,8 +47,6 @@ D_ViewChanged (void)
 	if (yscale > xscale)
 		scale_for_mip = yscale;
 
-	d_zwidth = vid.width;
-
 	d_pix_min = r_refdef.vrect.height / 240;
 	if (d_pix_min < 1)
 		d_pix_min = 1;
@@ -360,7 +358,6 @@ R_SetupFrame (void)
 		vrect.height = r_newrefdef.height;
 
 		d_viewbuffer = r_warpbuffer;
-		r_screenwidth = vid.width;
 	}
 	else
 	{
@@ -370,7 +367,6 @@ R_SetupFrame (void)
 		vrect.height = r_newrefdef.height;
 
 		d_viewbuffer = vid_buffer;
-		r_screenwidth = vid.width;
 	}
 
 	R_ViewChanged (&vrect);
@@ -392,6 +388,7 @@ R_SetupFrame (void)
 	r_outofsurfaces = 0;
 	r_outofverts = 0;
 	r_outofedges = 0;
+	r_outoftriangles = 0;
 
 	// d_setup
 	d_minmip = sw_mipcap->value;

src/client/refresh/soft/sw_part.c

@@ -83,8 +83,8 @@ R_DrawParticle(particle_t *pparticle, int level)
 	** compute addresses of zbuffer, framebuffer, and
 	** compute the Z-buffer reference value.
 	*/
-	pz = d_pzbuffer + (d_zwidth * v) + u;
+	pz = d_pzbuffer + (vid.width * v) + u;
-	pdest = d_viewbuffer + r_screenwidth * v + u;
+	pdest = d_viewbuffer + vid.width * v + u;
 	izi = (int)(zi * 0x8000);
 
 	/*
@@ -106,7 +106,7 @@ R_DrawParticle(particle_t *pparticle, int level)
 	{
 		switch (level) {
 		case PARTICLE_33 :
-			for ( ; count ; count--, pz += d_zwidth, pdest += r_screenwidth)
+			for ( ; count ; count--, pz += vid.width, pdest += vid.width)
 			{
 				//FIXME--do it in blocks of 8?
 				for (i=0 ; i<pix ; i++)
@@ -123,7 +123,7 @@ R_DrawParticle(particle_t *pparticle, int level)
 		case PARTICLE_66 :
 		{
 			int color_part = (color<<8);
-			for ( ; count ; count--, pz += d_zwidth, pdest += r_screenwidth)
+			for ( ; count ; count--, pz += vid.width, pdest += vid.width)
 			{
 				for (i=0 ; i<pix ; i++)
 				{
@@ -138,7 +138,7 @@ R_DrawParticle(particle_t *pparticle, int level)
 		}
 
 		default:  //100
-			for ( ; count ; count--, pz += d_zwidth, pdest += r_screenwidth)
+			for ( ; count ; count--, pz += vid.width, pdest += vid.width)
 			{
 				for (i=0 ; i<pix ; i++)
 				{
@@ -160,7 +160,7 @@ R_DrawParticle(particle_t *pparticle, int level)
 
 		switch (level) {
 		case PARTICLE_33 :
-			for ( ; count ; count--, pz += d_zwidth, pdest += r_screenwidth)
+			for ( ; count ; count--, pz += vid.width, pdest += vid.width)
 			{
 				//FIXME--do it in blocks of 8?
 				for (i=0 ; i<pix ; i++)
@@ -178,7 +178,7 @@ R_DrawParticle(particle_t *pparticle, int level)
 		case PARTICLE_66 :
 		{
 			int color_part = (color<<8);
-			for ( ; count ; count--, pz += d_zwidth, pdest += r_screenwidth)
+			for ( ; count ; count--, pz += vid.width, pdest += vid.width)
 			{
 				for (i=0 ; i<pix ; i++)
 				{
@@ -194,7 +194,7 @@ R_DrawParticle(particle_t *pparticle, int level)
 		}
 
 		default:  //100
-			for ( ; count ; count--, pz += d_zwidth, pdest += r_screenwidth)
+			for ( ; count ; count--, pz += vid.width, pdest += vid.width)
 			{
 				for (i=0 ; i<pix ; i++)
 				{

src/client/refresh/soft/sw_poly.c

@@ -622,8 +622,8 @@ R_PolygonDrawSpans(espan_t *pspan, int iswater )
 	{
 		int	count;
 
-		s_spanletvars.pdest   = d_viewbuffer + (r_screenwidth * pspan->v) + pspan->u;
+		s_spanletvars.pdest   = d_viewbuffer + (vid.width * pspan->v) + pspan->u;
-		s_spanletvars.pz      = d_pzbuffer + (d_zwidth * pspan->v) + pspan->u;
+		s_spanletvars.pz      = d_pzbuffer + (vid.width * pspan->v) + pspan->u;
 		s_spanletvars.u       = pspan->u;
 		s_spanletvars.v       = pspan->v;
 		count = pspan->count;

src/client/refresh/soft/sw_polyse.c

@@ -65,10 +65,9 @@ static int	r_sstepx, r_tstepx, r_lstepy, r_sstepy, r_tstepy;
 static zvalue_t	r_zistepx, r_zistepy;
 static int	d_aspancount, d_countextrastep;
 
-static spanpackage_t	*a_spans;
 static spanpackage_t	*d_pedgespanpackage;
 
-spanpackage_t	*triangle_spans;
+spanpackage_t	*triangle_spans, *triangles_max;
 
 static int	ystart;
 static pixel_t	*d_pdest, *d_ptex;
@@ -208,8 +207,6 @@ R_DrawTriangle(const finalvert_t *a, const finalvert_t *b, const finalvert_t *c)
 
 	if ( d_xdenom < 0 )
 	{
-		a_spans = triangle_spans;
-
 		r_p0[0] = a->u;	// u
 		r_p0[1] = a->v;	// v
 		r_p0[2] = a->s;	// s
@@ -236,17 +233,16 @@ R_DrawTriangle(const finalvert_t *a, const finalvert_t *b, const finalvert_t *c)
 	}
 }
 
-
-/*
-===================
-R_PolysetScanLeftEdge_C
-====================
-*/
 static void
-R_PolysetScanLeftEdge_C(int height)
+R_PushEdgesSpan()
 {
-	do
+	if (d_pedgespanpackage >= triangles_max)
 	{
+		// no space any more
+		r_outoftriangles++;
+		return;
+	}
+
 	d_pedgespanpackage->pdest = d_pdest;
 	d_pedgespanpackage->pz = d_pz;
 	d_pedgespanpackage->count = d_aspancount;
@@ -260,6 +256,19 @@ R_PolysetScanLeftEdge_C(int height)
 	d_pedgespanpackage->zi = d_zi;
 
 	d_pedgespanpackage++;
+}
+
+/*
+===================
+R_PolysetScanLeftEdge_C
+====================
+*/
+static void
+R_PolysetScanLeftEdge_C(int height)
+{
+	do
+	{
+		R_PushEdgesSpan();
 
 		errorterm += erroradjustup;
 		if (errorterm >= 0)
@@ -359,8 +368,6 @@ R_PolysetSetUpForLineScan(fixed8_t startvertu, fixed8_t startvertv,
 	int		tm, tn;
 	adivtab_t	*ptemp;
 
-	// TODO: implement x86 version
-
 	errorterm = -1;
 
 	tm = endvertu - startvertu;
@@ -782,7 +789,7 @@ R_RasterizeAliasPolySmooth (void)
 	//
 	// scan out the top (and possibly only) part of the left edge
 	//
-	d_pedgespanpackage = a_spans;
+	d_pedgespanpackage = triangle_spans;
 
 	ystart = plefttop[1];
 	d_aspancount = plefttop[0] - prighttop[0];
@@ -796,35 +803,23 @@ R_RasterizeAliasPolySmooth (void)
 	d_light = plefttop[4];
 	d_zi = plefttop[5];
 
-	d_pdest = d_viewbuffer + ystart * r_screenwidth + plefttop[0];
+	d_pdest = d_viewbuffer + ystart * vid.width + plefttop[0];
-	d_pz = d_pzbuffer + ystart * d_zwidth + plefttop[0];
+	d_pz = d_pzbuffer + ystart * vid.width + plefttop[0];
 
 	if (initialleftheight == 1)
 	{
-		d_pedgespanpackage->pdest = d_pdest;
+		R_PushEdgesSpan();
-		d_pedgespanpackage->pz = d_pz;
-		d_pedgespanpackage->count = d_aspancount;
-		d_pedgespanpackage->ptex = d_ptex;
-
-		d_pedgespanpackage->sfrac = d_sfrac;
-		d_pedgespanpackage->tfrac = d_tfrac;
-
-		// FIXME: need to clamp l, s, t, at both ends?
-		d_pedgespanpackage->light = d_light;
-		d_pedgespanpackage->zi = d_zi;
-
-		d_pedgespanpackage++;
 	}
 	else
 	{
 		R_PolysetSetUpForLineScan(plefttop[0], plefttop[1],
 					  pleftbottom[0], pleftbottom[1]);
 		{
-			d_pzbasestep = d_zwidth + ubasestep;
+			d_pzbasestep = vid.width + ubasestep;
 			d_pzextrastep = d_pzbasestep + 1;
 		}
 
-		d_pdestbasestep = r_screenwidth + ubasestep;
+		d_pdestbasestep = vid.width + ubasestep;
 		d_pdestextrastep = d_pdestbasestep + 1;
 
 		// TODO: can reuse partial expressions here
@@ -886,35 +881,23 @@ R_RasterizeAliasPolySmooth (void)
 		d_light = plefttop[4];
 		d_zi = plefttop[5];
 
-		d_pdest = d_viewbuffer + ystart * r_screenwidth + plefttop[0];
+		d_pdest = d_viewbuffer + ystart * vid.width + plefttop[0];
-		d_pz = d_pzbuffer + ystart * d_zwidth + plefttop[0];
+		d_pz = d_pzbuffer + ystart * vid.width + plefttop[0];
 
 		if (height == 1)
 		{
-			d_pedgespanpackage->pdest = d_pdest;
+			R_PushEdgesSpan();
-			d_pedgespanpackage->pz = d_pz;
-			d_pedgespanpackage->count = d_aspancount;
-			d_pedgespanpackage->ptex = d_ptex;
-
-			d_pedgespanpackage->sfrac = d_sfrac;
-			d_pedgespanpackage->tfrac = d_tfrac;
-
-			// FIXME: need to clamp l, s, t, at both ends?
-			d_pedgespanpackage->light = d_light;
-			d_pedgespanpackage->zi = d_zi;
-
-			d_pedgespanpackage++;
 		}
 		else
 		{
 			R_PolysetSetUpForLineScan(plefttop[0], plefttop[1],
 								  pleftbottom[0], pleftbottom[1]);
 
-			d_pdestbasestep = r_screenwidth + ubasestep;
+			d_pdestbasestep = vid.width + ubasestep;
 			d_pdestextrastep = d_pdestbasestep + 1;
 
 			{
-				d_pzbasestep = d_zwidth + ubasestep;
+				d_pzbasestep = vid.width + ubasestep;
 				d_pzextrastep = d_pzbasestep + 1;
 			}
 
@@ -952,15 +935,21 @@ R_RasterizeAliasPolySmooth (void)
 
 	// scan out the top (and possibly only) part of the right edge, updating the
 	// count field
-	d_pedgespanpackage = a_spans;
+	d_pedgespanpackage = triangle_spans;
 
 	R_PolysetSetUpForLineScan(prighttop[0], prighttop[1],
 						  prightbottom[0], prightbottom[1]);
 	d_aspancount = 0;
 	d_countextrastep = ubasestep + 1;
-	originalcount = a_spans[initialrightheight].count;
+	if ((triangle_spans + initialrightheight) >= triangles_max)
-	a_spans[initialrightheight].count = -999999; // mark end of the spanpackages
+	{
-	(*d_pdrawspans) (a_spans);
+		// we dont have enough triangles for save full height
+		r_outoftriangles++;
+		return;
+	}
+	originalcount = triangle_spans[initialrightheight].count;
+	triangle_spans[initialrightheight].count = -999999; // mark end of the spanpackages
+	(*d_pdrawspans) (triangle_spans);
 
 	// scan out the bottom part of the right edge, if it exists
 	if (pedgetable->numrightedges == 2)
@@ -968,7 +957,7 @@ R_RasterizeAliasPolySmooth (void)
 		int				height;
 		spanpackage_t	*pstart;
 
-		pstart = a_spans + initialrightheight;
+		pstart = triangle_spans + initialrightheight;
 		pstart->count = originalcount;
 
 		d_aspancount = prightbottom[0] - prighttop[0];
@@ -982,7 +971,14 @@ R_RasterizeAliasPolySmooth (void)
 							  prightbottom[0], prightbottom[1]);
 
 		d_countextrastep = ubasestep + 1;
-		a_spans[initialrightheight + height].count = -999999; // mark end of the spanpackages
+
+		if ((triangle_spans + initialrightheight + height) >= triangles_max)
+		{
+			// we dont have enough triangles for save full height
+			r_outoftriangles++;
+			return;
+		}
+		triangle_spans[initialrightheight + height].count = -999999; // mark end of the spanpackages
 		(*d_pdrawspans) (pstart);
 	}
 }

src/client/refresh/soft/sw_scan.c

@@ -25,7 +25,6 @@ Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
 
 
 #define SPANSTEP_SHIFT	4
-#define SPANSTEP	(1 << SPANSTEP_SHIFT)
 
 byte	**warp_rowptr;
 int	*warp_column;
@@ -100,7 +99,7 @@ Return safe span step for u/z and v/z
 =============
 */
 static int
-D_DrawSpanGetStep(float d_zistepu, float d_zistepv, int cachewidth)
+D_DrawSpanGetStep(float d_zistepu, float d_zistepv)
 {
 	int	spanzshift = SPANSTEP_SHIFT;
 	int	spanzshift_value = (1 << spanzshift);
@@ -115,24 +114,46 @@ D_DrawSpanGetStep(float d_zistepu, float d_zistepv, int cachewidth)
 	    (int)(d_zistepv_shifted * spanzshift_value) == 0)
 	{
 		// search next safe value
-		while (spanzshift_value < vid.width &&
+		do
-		       (int)(d_zistepu_shifted * spanzshift_value) == 0 &&
-		       (int)(d_zistepv_shifted * spanzshift_value) == 0)
 		{
 			spanzshift ++;
 			spanzshift_value <<= 1;
-		}
 
-		// step back to last safe value
 			if ((int)(d_zistepu_shifted * spanzshift_value) != 0 ||
 			    (int)(d_zistepv_shifted * spanzshift_value) != 0)
 			{
-			spanzshift --;
+				// step back to last safe value
+				return spanzshift - 1;
 			}
+
+		}
+		while (spanzshift_value < vid.width);
 	}
 	return spanzshift;
 }
 
+/*
+=============
+D_DrawZSpanGetStepValue
+
+Return safe span step for izistep
+=============
+*/
+static int
+D_DrawZSpanGetStepValue(zvalue_t izistep)
+{
+	// check that we can draw parallel surfaces to screen surface
+	// (both ends have same z value)
+	int	count = 1;
+
+	while ((izistep * count) >> SHIFT16XYZ == 0 && count < vid.width)
+	{
+		count <<= 1;
+	}
+
+	return count;
+}
+
 /*
 =============
 D_DrawTurbulentSpan
@@ -169,14 +190,18 @@ TurbulentPow2 (espan_t *pspan)
 	float	sdivzpow2stepu, tdivzpow2stepu, zipow2stepu;
 	pixel_t	*r_turb_pbase;
 	int	*r_turb_turb;
+	int	spanstep_shift, spanstep_value;
+
+	spanstep_shift = D_DrawSpanGetStep(d_zistepu, d_zistepv);
+	spanstep_value = (1 << spanstep_shift);
 
 	r_turb_turb = sintable + ((int)(r_newrefdef.time*SPEED)&(CYCLE-1));
 
 	r_turb_pbase = (unsigned char *)cacheblock;
 
-	sdivzpow2stepu = d_sdivzstepu * SPANSTEP;
+	sdivzpow2stepu = d_sdivzstepu * spanstep_value;
-	tdivzpow2stepu = d_tdivzstepu * SPANSTEP;
+	tdivzpow2stepu = d_tdivzstepu * spanstep_value;
-	zipow2stepu = d_zistepu * SPANSTEP;
+	zipow2stepu = d_zistepu * spanstep_value;
 
 	do
 	{
@@ -184,7 +209,7 @@ TurbulentPow2 (espan_t *pspan)
 		float sdivz, tdivz, zi, z, du, dv;
 		pixel_t	*r_turb_pdest;
 
-		r_turb_pdest = d_viewbuffer + (r_screenwidth * pspan->v) + pspan->u;
+		r_turb_pdest = d_viewbuffer + (vid.width * pspan->v) + pspan->u;
 
 		count = pspan->count;
 
@@ -219,8 +244,8 @@ TurbulentPow2 (espan_t *pspan)
 			r_turb_tstep = 0;	// ditto
 
 			// calculate s and t at the far end of the span
-			if (count >= SPANSTEP)
+			if (count >= spanstep_value)
-				r_turb_spancount = SPANSTEP;
+				r_turb_spancount = spanstep_value;
 			else
 				r_turb_spancount = count;
 
@@ -238,21 +263,21 @@ TurbulentPow2 (espan_t *pspan)
 				snext = (int)(sdivz * z) + sadjust;
 				if (snext > bbextents)
 					snext = bbextents;
-				else if (snext < SPANSTEP)
+				else if (snext < spanstep_value)
 					// prevent round-off error on <0 steps from
 					//  from causing overstepping & running off the
 					//  edge of the texture
-					snext = SPANSTEP;
+					snext = spanstep_value;
 
 				tnext = (int)(tdivz * z) + tadjust;
 				if (tnext > bbextentt)
 					tnext = bbextentt;
-				else if (tnext < SPANSTEP)
+				else if (tnext < spanstep_value)
 					// guard against round-off error on <0 steps
-					tnext = SPANSTEP;
+					tnext = spanstep_value;
 
-				r_turb_sstep = (snext - r_turb_s) >> SPANSTEP_SHIFT;
+				r_turb_sstep = (snext - r_turb_s) >> spanstep_shift;
-				r_turb_tstep = (tnext - r_turb_t) >> SPANSTEP_SHIFT;
+				r_turb_tstep = (tnext - r_turb_t) >> spanstep_shift;
 			}
 			else
 			{
@@ -268,18 +293,18 @@ TurbulentPow2 (espan_t *pspan)
 				snext = (int)(sdivz * z) + sadjust;
 				if (snext > bbextents)
 					snext = bbextents;
-				else if (snext < SPANSTEP)
+				else if (snext < spanstep_value)
 					// prevent round-off error on <0 steps from
 					//  from causing overstepping & running off the
 					//  edge of the texture
-					snext = SPANSTEP;
+					snext = spanstep_value;
 
 				tnext = (int)(tdivz * z) + tadjust;
 				if (tnext > bbextentt)
 					tnext = bbextentt;
-				else if (tnext < SPANSTEP)
+				else if (tnext < spanstep_value)
 					// guard against round-off error on <0 steps
-					tnext = SPANSTEP;
+					tnext = spanstep_value;
 
 				if (r_turb_spancount > 1)
 				{
@@ -320,14 +345,18 @@ NonTurbulentPow2 (espan_t *pspan)
 	float sdivzpow2stepu, tdivzpow2stepu, zipow2stepu;
 	pixel_t	*r_turb_pbase;
 	int	*r_turb_turb;
+	int	spanstep_shift, spanstep_value;
+
+	spanstep_shift = D_DrawSpanGetStep(d_zistepu, d_zistepv);
+	spanstep_value = (1 << spanstep_shift);
 
 	r_turb_turb = blanktable;
 
 	r_turb_pbase = (unsigned char *)cacheblock;
 
-	sdivzpow2stepu = d_sdivzstepu * SPANSTEP;
+	sdivzpow2stepu = d_sdivzstepu * spanstep_value;
-	tdivzpow2stepu = d_tdivzstepu * SPANSTEP;
+	tdivzpow2stepu = d_tdivzstepu * spanstep_value;
-	zipow2stepu = d_zistepu * SPANSTEP;
+	zipow2stepu = d_zistepu * spanstep_value;
 
 	do
 	{
@@ -335,7 +364,7 @@ NonTurbulentPow2 (espan_t *pspan)
 		float sdivz, tdivz, zi, z, dv, du;
 		pixel_t	*r_turb_pdest;
 
-		r_turb_pdest = d_viewbuffer + (r_screenwidth * pspan->v) + pspan->u;
+		r_turb_pdest = d_viewbuffer + (vid.width * pspan->v) + pspan->u;
 
 		count = pspan->count;
 
@@ -370,8 +399,8 @@ NonTurbulentPow2 (espan_t *pspan)
 			r_turb_tstep = 0;	// ditto
 
 			// calculate s and t at the far end of the span
-			if (count >= SPANSTEP)
+			if (count >= spanstep_value)
-				r_turb_spancount = SPANSTEP;
+				r_turb_spancount = spanstep_value;
 			else
 				r_turb_spancount = count;
 
@@ -389,21 +418,21 @@ NonTurbulentPow2 (espan_t *pspan)
 				snext = (int)(sdivz * z) + sadjust;
 				if (snext > bbextents)
 					snext = bbextents;
-				else if (snext < SPANSTEP)
+				else if (snext < spanstep_value)
 					// prevent round-off error on <0 steps from
 					//  from causing overstepping & running off the
 					//  edge of the texture
-					snext = SPANSTEP;
+					snext = spanstep_value;
 
 				tnext = (int)(tdivz * z) + tadjust;
 				if (tnext > bbextentt)
 					tnext = bbextentt;
-				else if (tnext < SPANSTEP)
+				else if (tnext < spanstep_value)
 					// guard against round-off error on <0 steps
-					tnext = SPANSTEP;
+					tnext = spanstep_value;
 
-				r_turb_sstep = (snext - r_turb_s) >> SPANSTEP_SHIFT;
+				r_turb_sstep = (snext - r_turb_s) >> spanstep_shift;
-				r_turb_tstep = (tnext - r_turb_t) >> SPANSTEP_SHIFT;
+				r_turb_tstep = (tnext - r_turb_t) >> spanstep_shift;
 			}
 			else
 			{
@@ -419,18 +448,18 @@ NonTurbulentPow2 (espan_t *pspan)
 				snext = (int)(sdivz * z) + sadjust;
 				if (snext > bbextents)
 					snext = bbextents;
-				else if (snext < SPANSTEP)
+				else if (snext < spanstep_value)
 					// prevent round-off error on <0 steps from
 					//  from causing overstepping & running off the
 					//  edge of the texture
-					snext = SPANSTEP;
+					snext = spanstep_value;
 
 				tnext = (int)(tdivz * z) + tadjust;
 				if (tnext > bbextentt)
 					tnext = bbextentt;
-				else if (tnext < SPANSTEP)
+				else if (tnext < spanstep_value)
 					// guard against round-off error on <0 steps
-					tnext = SPANSTEP;
+					tnext = spanstep_value;
 
 				if (r_turb_spancount > 1)
 				{
@@ -459,7 +488,7 @@ NonTurbulentPow2 (espan_t *pspan)
 //====================
 
 // Enable custom filtering
-extern cvar_t	*r_anisotropic;
+extern cvar_t	*sw_texture_filtering;
 static const int filtering_kernel[2][2][2] = {
 	{
 		{0x1 << (SHIFT16XYZ-2), 0x0},
@@ -572,12 +601,12 @@ D_DrawSpansPow2 (espan_t *pspan)
 	int	texture_filtering;
 	int	spanstep_shift, spanstep_value;
 
-	spanstep_shift = D_DrawSpanGetStep(d_zistepu, d_zistepv, cachewidth);
+	spanstep_shift = D_DrawSpanGetStep(d_zistepu, d_zistepv);
 	spanstep_value = (1 << spanstep_shift);
 
 	pbase = (unsigned char *)cacheblock;
 
-	texture_filtering = (int)r_anisotropic->value;
+	texture_filtering = (int)sw_texture_filtering->value;
 	sdivzpow2stepu = d_sdivzstepu * spanstep_value;
 	tdivzpow2stepu = d_tdivzstepu * spanstep_value;
 	zipow2stepu = d_zistepu * spanstep_value;
@@ -588,7 +617,7 @@ D_DrawSpansPow2 (espan_t *pspan)
 		int	count, s, t;
 		float	sdivz, tdivz, zi, z, du, dv;
 
-		pdest = d_viewbuffer + (r_screenwidth * pspan->v) + pspan->u;
+		pdest = d_viewbuffer + (vid.width * pspan->v) + pspan->u;
 
 		count = pspan->count;
 
@@ -717,10 +746,12 @@ void
 D_DrawZSpans (espan_t *pspan)
 {
 	zvalue_t	izistep;
+	int		safe_step;
 
 	// FIXME: check for clamping/range problems
 	// we count on FP exceptions being turned off to avoid range problems
 	izistep = (int)(d_zistepu * 0x8000 * (float)SHIFT16XYZ_MULT);
+	safe_step = D_DrawZSpanGetStepValue(izistep);
 
 	do
 	{
@@ -730,7 +761,7 @@ D_DrawZSpans (espan_t *pspan)
 		float		zi;
 		float		du, dv;
 
-		pdest = d_pzbuffer + (d_zwidth * pspan->v) + pspan->u;
+		pdest = d_pzbuffer + (vid.width * pspan->v) + pspan->u;
 
 		count = pspan->count;
 
@@ -742,11 +773,29 @@ D_DrawZSpans (espan_t *pspan)
 		// we count on FP exceptions being turned off to avoid range problems
 		izi = (int)(zi * 0x8000 * (float)SHIFT16XYZ_MULT);
 
+		if (safe_step > 1)
+		{
+			const zvalue_t *tdest_max = pdest + count;
+			do
+			{
+				int	step;
+				zvalue_t izi_shifted = izi >> SHIFT16XYZ;
+
+				for(step = 0; (step < safe_step) && (pdest < tdest_max); step++)
+				{
+					*pdest++ = izi_shifted;
+				}
+				izi += (izistep * safe_step);
+			} while (pdest < tdest_max);
+		}
+		else
+		{
 			while (count > 0)
 			{
 				*pdest++ = izi >> SHIFT16XYZ;
 				izi += izistep;
 				count--;
 			}
+		}
 	} while ((pspan = pspan->pnext) != NULL);
 }

src/client/refresh/soft/sw_surf.c

@@ -80,7 +80,6 @@ R_DrawSurface (drawsurf_t *drawsurf)
 	int		smax, tmax, twidth;
 	int		u;
 	int		soffset, basetoffset, texwidth;
-	int		horzblockstep;
 	unsigned char	*pcolumndest;
 	image_t		*mt;
 
@@ -104,7 +103,7 @@ R_DrawSurface (drawsurf_t *drawsurf)
 	//==============================
 
 	// TODO: only needs to be set when there is a display settings change
-	horzblockstep = blocksize;
+	blocksize = blocksize;
 
 	smax = mt->width >> drawsurf->surfmip;
 	twidth = texwidth;
@@ -138,7 +137,7 @@ R_DrawSurface (drawsurf_t *drawsurf)
 		if (soffset >= smax)
 			soffset = 0;
 
-		pcolumndest += horzblockstep;
+		pcolumndest += blocksize;
 	}
 }
 

src/client/vid/glimp_sdl.c

@@ -36,6 +36,7 @@
 cvar_t *vid_displayrefreshrate;
 int glimp_refreshRate = -1;
 
+static int last_flags = 0;
 static SDL_Window* window = NULL;
 static qboolean initSuccessful = false;
 
@@ -252,8 +253,11 @@ GLimp_InitGraphics(int fullscreen, int *pwidth, int *pheight)
 	viddef.width = width;
 	viddef.height = height;
 
+	if(last_flags != -1 && (last_flags & SDL_WINDOW_OPENGL))
+	{
 		/* Reset SDL. */
 		SDL_GL_ResetAttributes();
+	}
 
 	/* Let renderer prepare things (set OpenGL attributes).
 	   FIXME: This is no longer necessary, the renderer
@@ -317,6 +321,8 @@ GLimp_InitGraphics(int fullscreen, int *pwidth, int *pheight)
 		}
 	}
 
+	last_flags = flags;
+
 	if (!re.InitContext(window))
 	{
 		/* InitContext() should have logged an error. */

src/client/vid/header/ref.h

@@ -226,7 +226,6 @@ typedef struct
 
 	qboolean	(IMPORT *Vid_GetModeInfo)(int *width, int *height, int mode);
 	void		(IMPORT *Vid_MenuInit)( void );
-	void		(IMPORT *Vid_NewWindow)( int width, int height );
 	// called with image data of width*height pixel which comp bytes per pixel (must be 3 or 4 for RGB or RGBA)
 	// expects the pixels data to be row-wise, starting at top left
 	void		(IMPORT *Vid_WriteScreenshot)( int width, int height, int comp, const void* data );

src/client/vid/vid.c

@@ -309,18 +309,6 @@ VID_Restart_f(void)
 	vid_fullscreen->modified = true;
 }
 
-/*
- * FIXME: This is only used by the softrenderer. The software
- * renderer should be ported to the API provided by refresh.c
- * and this call removed.
- */
-void
-VID_NewWindow(int width, int height)
-{
-	viddef.width = width;
-	viddef.height = height;
-}
-
 /*
  * Shuts the renderer down and unloads it.
  */
@@ -402,7 +390,6 @@ VID_LoadRenderer(void)
 	ri.Sys_Error = Com_Error;
 	ri.Vid_GetModeInfo = VID_GetModeInfo;
 	ri.Vid_MenuInit = VID_MenuInit;
-	ri.Vid_NewWindow = VID_NewWindow;
 	ri.Vid_WriteScreenshot = VID_WriteScreenshot;
 
 	// Exchange our export struct with the renderers import struct.
@@ -480,7 +467,7 @@ VID_CheckChanges(void)
 				Com_Printf("Retrying with soft...\n");
 				Cvar_Set("vid_renderer", "soft");
 			}
-			else if (strcmp("vid_renderer", "soft") == 0)
+			else if (strcmp(vid_renderer->string, "soft") == 0)
 			{
 				// Sorry, no usable renderer found.
 				Com_Error(ERR_FATAL, "No usable renderer found!\n");

src/common/cvar.c

@@ -39,7 +39,6 @@ typedef struct
 replacement_t replacements[] = {
 	{"cd_shuffle", "ogg_shuffle"},
 	{"cl_drawfps", "cl_showfps"},
-	{"gl_anisotropic", "r_anisotropic"},
 	{"gl_drawentities", "r_drawentities"},
 	{"gl_drawworld", "r_drawworld"},
 	{"gl_fullbright", "r_fullbright"},

stuff/cvarlist.md

@@ -163,15 +163,15 @@ Graphics (all renderers):
 * **r_vsync**: Enables the vsync: frames are synchronized with
   display refresh rate, should (but doesn't always) prevent tearing.
 
-* **r_anisotropic**: Anisotropic filtering. Possible values are
-  dependent on the GPU driver, most of them support `1`, `2`, `4`, `8`
-  and `16`. Anisotropic filtering gives a huge improvement to texture
-  quality by a negligible performance impact.
-
 
 Graphics (GL renderers only):
 -----------------------------
 
+* **gl_anisotropic**: Anisotropic filtering. Possible values are
+  dependent on the GPU driver, most of them support `1`, `2`, `4`, `8`
+  and `16`. Anisotropic filtering gives a huge improvement to texture
+  quality by a negligible performance impact.
+
 * **gl_msaa_samples**: Full scene anti aliasing samples. The number of
   samples depends on the GPU driver, most drivers support at least
   `2`, `4` and `8` samples. If an invalid value is set, the value is