X-Git-Url: http://de.git.xonotic.org/?a=blobdiff_plain;f=gl_backend.c;h=1ab98969ce70260e7872419c838aef201af7c7d8;hb=9168bc2cbb2230dd8147be0b5aa31cf4a1404745;hp=48d49829a25568c44ae6f9df4920cfbc9996fcdb;hpb=d96a2757b0c114e7132e7e82444b676274105f08;p=xonotic%2Fdarkplaces.git

diff --git a/gl_backend.c b/gl_backend.c
index 48d49829..1ab98969 100644
--- a/gl_backend.c
+++ b/gl_backend.c
@@ -1,9 +1,27 @@
 
 #include "quakedef.h"
+#include "image.h"
+#include "jpeg.h"
 
-cvar_t gl_mesh_maxverts = {0, "gl_mesh_maxverts", "1024"};
+//#define MESH_VAR
+#define MESH_BATCH
+
+// 65536 is the max addressable on a Geforce 256 up until Geforce3
+// (excluding MX), seems a reasonable number...
+cvar_t gl_mesh_maxverts = {0, "gl_mesh_maxverts", "65536"};
 cvar_t gl_mesh_floatcolors = {0, "gl_mesh_floatcolors", "1"};
 cvar_t gl_mesh_drawrangeelements = {0, "gl_mesh_drawrangeelements", "1"};
+#ifdef MESH_VAR
+cvar_t gl_mesh_vertex_array_range = {0, "gl_mesh_vertex_array_range", "0"};
+cvar_t gl_mesh_vertex_array_range_readfrequency = {0, "gl_mesh_vertex_array_range_readfrequency", "0.2"};
+cvar_t gl_mesh_vertex_array_range_writefrequency = {0, "gl_mesh_vertex_array_range_writefrequency", "0.2"};
+cvar_t gl_mesh_vertex_array_range_priority = {0, "gl_mesh_vertex_array_range_priority", "0.7"};
+#endif
+#ifdef MESH_BATCH
+cvar_t gl_mesh_batching = {0, "gl_mesh_batching", "1"};
+#endif
+cvar_t gl_mesh_copyarrays = {0, "gl_mesh_copyarrays", "1"};
+cvar_t gl_delayfinish = {CVAR_SAVE, "gl_delayfinish", "0"};
 
 cvar_t r_render = {0, "r_render", "1"};
 cvar_t gl_dither = {CVAR_SAVE, "gl_dither", "1"}; // whether or not to use dithering
@@ -50,7 +68,7 @@ void GL_PrintError(int errornumber, char *filename, int linenumber)
 		break;
 #endif
 #ifdef GL_TABLE_TOO_LARGE
-    case GL_TABLE_TOO_LARGE:
+	case GL_TABLE_TOO_LARGE:
 		Con_Printf("GL_TABLE_TOO_LARGE at %s:%i\n", filename, linenumber);
 		break;
 #endif
@@ -70,65 +88,201 @@ void SCR_ScreenShot_f (void);
 // these are externally accessible
 int r_lightmapscalebit;
 float r_colorscale;
-float *varray_vertex;
-float *varray_color;
-float *varray_texcoord[MAX_TEXTUREUNITS];
+GLfloat *varray_vertex3f, *varray_buf_vertex3f;
+GLfloat *varray_color4f, *varray_buf_color4f;
+GLfloat *varray_texcoord3f[MAX_TEXTUREUNITS], *varray_buf_texcoord3f[MAX_TEXTUREUNITS];
+GLfloat *varray_texcoord2f[MAX_TEXTUREUNITS], *varray_buf_texcoord2f[MAX_TEXTUREUNITS];
+static qbyte *varray_buf_color4b;
 int mesh_maxverts;
+#ifdef MESH_VAR
+int mesh_var;
+float mesh_var_readfrequency;
+float mesh_var_writefrequency;
+float mesh_var_priority;
+#endif
+int varray_offset = 0, varray_offsetnext = 0;
+GLuint *varray_buf_elements3i;
+int mesh_maxelements = 32768;
+#ifdef MESH_BATCH
+int gl_batchvertexfirst = 0;
+int gl_batchvertexcount = 0;
+int gl_batchelementcount = 0;
+#endif
 
 static matrix4x4_t backend_viewmatrix;
 static matrix4x4_t backend_modelmatrix;
 static matrix4x4_t backend_modelviewmatrix;
 static matrix4x4_t backend_glmodelviewmatrix;
+static matrix4x4_t backend_projectmatrix;
 
 static int backendunits, backendactive;
-static qbyte *varray_bcolor;
 static mempool_t *gl_backend_mempool;
 
+/*
+note: here's strip order for a terrain row:
+0--1--2--3--4
+|\ |\ |\ |\ |
+| \| \| \| \|
+A--B--C--D--E
+
+A0B, 01B, B1C, 12C, C2D, 23D, D3E, 34E
+
+*elements++ = i + row;
+*elements++ = i;
+*elements++ = i + row + 1;
+*elements++ = i;
+*elements++ = i + 1;
+*elements++ = i + row + 1;
+*/
+
+void GL_Backend_AllocElementsArray(void)
+{
+	if (varray_buf_elements3i)
+		Mem_Free(varray_buf_elements3i);
+	varray_buf_elements3i = Mem_Alloc(gl_backend_mempool, mesh_maxelements * sizeof(GLuint));
+}
+
+void GL_Backend_FreeElementArray(void)
+{
+	if (varray_buf_elements3i)
+		Mem_Free(varray_buf_elements3i);
+	varray_buf_elements3i = NULL;
+}
+
+void GL_Backend_CheckCvars(void)
+{
+	if (gl_mesh_maxverts.integer < 1024)
+		Cvar_SetValueQuick(&gl_mesh_maxverts, 1024);
+	if (gl_mesh_maxverts.integer > 65536)
+		Cvar_SetValueQuick(&gl_mesh_maxverts, 65536);
+#ifdef MESH_VAR
+	if (gl_mesh_vertex_array_range.integer && !gl_support_var)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range, 0);
+	if (gl_mesh_vertex_array_range_readfrequency.value < 0)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range_readfrequency, 0);
+	if (gl_mesh_vertex_array_range_readfrequency.value > 1)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range_readfrequency, 1);
+	if (gl_mesh_vertex_array_range_writefrequency.value < 0)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range_writefrequency, 0);
+	if (gl_mesh_vertex_array_range_writefrequency.value > 1)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range_writefrequency, 1);
+	if (gl_mesh_vertex_array_range_priority.value < 0)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range_priority, 0);
+	if (gl_mesh_vertex_array_range_priority.value > 1)
+		Cvar_SetValueQuick(&gl_mesh_vertex_array_range_priority, 1);
+#endif
+}
+
 int polygonelements[768];
 
+static void R_Mesh_CacheArray_Startup(void);
+static void R_Mesh_CacheArray_Shutdown(void);
 void GL_Backend_AllocArrays(void)
 {
-	int i;
+	int i, size;
+	qbyte *data;
 
-	for (i = 0;i < POLYGONELEMENTS_MAXPOINTS - 2;i++)
+	if (!gl_backend_mempool)
 	{
-		polygonelements[i * 3 + 0] = 0;
-		polygonelements[i * 3 + 1] = i + 1;
-		polygonelements[i * 3 + 2] = i + 2;
+		gl_backend_mempool = Mem_AllocPool("GL_Backend");
+		varray_buf_vertex3f = NULL;
+		varray_buf_color4f = NULL;
+		varray_buf_color4b = NULL;
+		varray_buf_elements3i = NULL;
+		for (i = 0;i < MAX_TEXTUREUNITS;i++)
+			varray_buf_texcoord3f[i] = varray_buf_texcoord2f[i] = NULL;
 	}
 
-	if (!gl_backend_mempool)
-		gl_backend_mempool = Mem_AllocPool("GL_Backend");
+	if (varray_buf_vertex3f)
+#ifdef MESH_VAR
+		VID_FreeVertexArrays(varray_buf_vertex3f);
+#else
+		Mem_Free(varray_buf_vertex3f);
+#endif
+	varray_buf_vertex3f = NULL;
+	varray_buf_color4f = NULL;
+	varray_buf_color4b = NULL;
+	for (i = 0;i < MAX_TEXTUREUNITS;i++)
+		varray_buf_texcoord3f[i] = varray_buf_texcoord2f[i] = NULL;
+
+	mesh_maxverts = gl_mesh_maxverts.integer;
+	size = mesh_maxverts * (sizeof(float[3]) + sizeof(float[4]) + sizeof(qbyte[4]) + (sizeof(float[3]) + sizeof(float[2])) * backendunits);
+#ifdef MESH_VAR
+	mesh_var = gl_mesh_vertex_array_range.integer && gl_support_var;
+	mesh_var_readfrequency = gl_mesh_vertex_array_range_readfrequency.value;
+	mesh_var_writefrequency = gl_mesh_vertex_array_range_writefrequency.value;
+	mesh_var_priority = gl_mesh_vertex_array_range_priority.value;
+	data = VID_AllocVertexArrays(gl_backend_mempool, size, gl_mesh_vertex_array_range.integer, gl_mesh_vertex_array_range_readfrequency.value, gl_mesh_vertex_array_range_writefrequency.value, gl_mesh_vertex_array_range_priority.value);
+#else
+	data = Mem_Alloc(gl_backend_mempool, size);
+#endif
 
-	varray_vertex = Mem_Alloc(gl_backend_mempool, mesh_maxverts * sizeof(float[4]));
-	varray_color = Mem_Alloc(gl_backend_mempool, mesh_maxverts * sizeof(float[4]));
-	varray_bcolor = Mem_Alloc(gl_backend_mempool, mesh_maxverts * sizeof(qbyte[4]));
+	varray_buf_vertex3f = (void *)data;data += sizeof(float[3]) * mesh_maxverts;
+	varray_buf_color4f = (void *)data;data += sizeof(float[4]) * mesh_maxverts;
 	for (i = 0;i < backendunits;i++)
-		varray_texcoord[i] = Mem_Alloc(gl_backend_mempool, mesh_maxverts * sizeof(float[4]));
+	{
+		varray_buf_texcoord3f[i] = (void *)data;data += sizeof(float[3]) * mesh_maxverts;
+		varray_buf_texcoord2f[i] = (void *)data;data += sizeof(float[2]) * mesh_maxverts;
+	}
 	for (;i < MAX_TEXTUREUNITS;i++)
-		varray_texcoord[i] = NULL;
+		varray_buf_texcoord3f[i] = varray_buf_texcoord2f[i] = NULL;
+	varray_buf_color4b = (void *)data;data += sizeof(qbyte[4]) * mesh_maxverts;
+
+	GL_Backend_AllocElementsArray();
+
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		CHECKGLERROR
+		qglVertexArrayRangeNV(size, varray_buf_vertex3f);
+		CHECKGLERROR
+	}
+#endif
+
+	R_Mesh_CacheArray_Startup();
 }
 
-void GL_Backend_FreeArrays(int resizingbuffers)
+void GL_Backend_FreeArrays(void)
 {
 	int i;
-	if (resizingbuffers)
-		Mem_EmptyPool(gl_backend_mempool);
-	else
-		Mem_FreePool(&gl_backend_mempool);
-	varray_vertex = NULL;
-	varray_color = NULL;
-	varray_bcolor = NULL;
+
+	R_Mesh_CacheArray_Shutdown();
+
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		CHECKGLERROR
+		qglDisableClientState(GL_VERTEX_ARRAY_RANGE_NV);
+		CHECKGLERROR
+	}
+#endif
+
+	if (varray_buf_vertex3f)
+#ifdef MESH_VAR
+		VID_FreeVertexArrays(varray_buf_vertex3f);
+#else
+		Mem_Free(varray_buf_vertex3f);
+#endif
+	varray_buf_vertex3f = NULL;
+	varray_buf_color4f = NULL;
+	varray_buf_color4b = NULL;
 	for (i = 0;i < MAX_TEXTUREUNITS;i++)
-		varray_texcoord[i] = NULL;
+		varray_buf_texcoord3f[i] = varray_buf_texcoord2f[i] = NULL;
+	varray_buf_elements3i = NULL;
+
+	Mem_FreePool(&gl_backend_mempool);
 }
 
 static void gl_backend_start(void)
 {
+	GL_Backend_CheckCvars();
+
 	Con_Printf("OpenGL Backend started with gl_mesh_maxverts %i\n", gl_mesh_maxverts.integer);
 	if (qglDrawRangeElements != NULL)
 	{
+		CHECKGLERROR
 		qglGetIntegerv(GL_MAX_ELEMENTS_VERTICES, &gl_maxdrawrangeelementsvertices);
+		CHECKGLERROR
 		qglGetIntegerv(GL_MAX_ELEMENTS_INDICES, &gl_maxdrawrangeelementsindices);
 		CHECKGLERROR
 		Con_Printf("glDrawRangeElements detected (max vertices %i, max indices %i)\n", gl_maxdrawrangeelementsvertices, gl_maxdrawrangeelementsindices);
@@ -143,6 +297,21 @@ static void gl_backend_start(void)
 
 	GL_Backend_AllocArrays();
 
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		CHECKGLERROR
+		qglEnableClientState(GL_VERTEX_ARRAY_RANGE_NV);
+		CHECKGLERROR
+	}
+#endif
+	varray_offset = varray_offsetnext = 0;
+#ifdef MESH_BATCH
+	gl_batchvertexfirst = 0;
+	gl_batchvertexcount = 0;
+	gl_batchelementcount = 0;
+#endif
+
 	backendactive = true;
 }
 
@@ -153,24 +322,23 @@ static void gl_backend_shutdown(void)
 
 	Con_Printf("OpenGL Backend shutting down\n");
 
-	GL_Backend_FreeArrays(false);
-}
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		CHECKGLERROR
+		qglDisableClientState(GL_VERTEX_ARRAY_RANGE_NV);
+		CHECKGLERROR
+	}
+#endif
 
-void GL_Backend_CheckCvars(void)
-{
-	// 21760 is (65536 / 3) rounded off to a multiple of 128
-	if (gl_mesh_maxverts.integer < 1024)
-		Cvar_SetValueQuick(&gl_mesh_maxverts, 1024);
-	if (gl_mesh_maxverts.integer > 21760)
-		Cvar_SetValueQuick(&gl_mesh_maxverts, 21760);
+	GL_Backend_FreeArrays();
 }
 
-void GL_Backend_ResizeArrays(int numtriangles)
+void GL_Backend_ResizeArrays(int numvertices)
 {
-	Cvar_SetValueQuick(&gl_mesh_maxverts, numtriangles);
+	Cvar_SetValueQuick(&gl_mesh_maxverts, numvertices);
 	GL_Backend_CheckCvars();
 	mesh_maxverts = gl_mesh_maxverts.integer;
-	GL_Backend_FreeArrays(true);
 	GL_Backend_AllocArrays();
 }
 
@@ -180,9 +348,19 @@ static void gl_backend_newmap(void)
 
 void gl_backend_init(void)
 {
+	int i;
+
+	for (i = 0;i < POLYGONELEMENTS_MAXPOINTS - 2;i++)
+	{
+		polygonelements[i * 3 + 0] = 0;
+		polygonelements[i * 3 + 1] = i + 1;
+		polygonelements[i * 3 + 2] = i + 2;
+	}
+
 	Cvar_RegisterVariable(&r_render);
 	Cvar_RegisterVariable(&gl_dither);
 	Cvar_RegisterVariable(&gl_lockarrays);
+	Cvar_RegisterVariable(&gl_delayfinish);
 #ifdef NORENDER
 	Cvar_SetValue("r_render", 0);
 #endif
@@ -190,7 +368,16 @@ void gl_backend_init(void)
 	Cvar_RegisterVariable(&gl_mesh_maxverts);
 	Cvar_RegisterVariable(&gl_mesh_floatcolors);
 	Cvar_RegisterVariable(&gl_mesh_drawrangeelements);
-	GL_Backend_CheckCvars();
+#ifdef MESH_VAR
+	Cvar_RegisterVariable(&gl_mesh_vertex_array_range);
+	Cvar_RegisterVariable(&gl_mesh_vertex_array_range_readfrequency);
+	Cvar_RegisterVariable(&gl_mesh_vertex_array_range_writefrequency);
+	Cvar_RegisterVariable(&gl_mesh_vertex_array_range_priority);
+#endif
+#ifdef MESH_BATCH
+	Cvar_RegisterVariable(&gl_mesh_batching);
+#endif
+	Cvar_RegisterVariable(&gl_mesh_copyarrays);
 	R_RegisterModule("GL_Backend", gl_backend_start, gl_backend_shutdown, gl_backend_newmap);
 }
 
@@ -221,7 +408,7 @@ void GL_SetupView_Orientation_FromEntity (vec3_t origin, vec3_t angles)
 	memset(&backend_modelmatrix, 0, sizeof(backend_modelmatrix));
 }
 
-void GL_SetupView_Mode_Perspective (double aspect, double fovx, double fovy, double zNear, double zFar)
+void GL_SetupView_Mode_Perspective (double fovx, double fovy, double zNear, double zFar)
 {
 	double xmax, ymax;
 
@@ -230,7 +417,7 @@ void GL_SetupView_Mode_Perspective (double aspect, double fovx, double fovy, dou
 
 	// set up viewpoint
 	qglMatrixMode(GL_PROJECTION);CHECKGLERROR
-	qglLoadIdentity ();CHECKGLERROR
+	qglLoadIdentity();CHECKGLERROR
 	// pyramid slopes
 	xmax = zNear * tan(fovx * M_PI / 360.0);
 	ymax = zNear * tan(fovy * M_PI / 360.0);
@@ -240,6 +427,55 @@ void GL_SetupView_Mode_Perspective (double aspect, double fovx, double fovy, dou
 	GL_SetupView_Orientation_Identity();
 }
 
+void GL_SetupView_Mode_PerspectiveInfiniteFarClip (double fovx, double fovy, double zNear)
+{
+	float nudge, m[16];
+
+	if (!r_render.integer)
+		return;
+
+	// set up viewpoint
+	qglMatrixMode(GL_PROJECTION);CHECKGLERROR
+	qglLoadIdentity();CHECKGLERROR
+	// set view pyramid
+	nudge = 1.0 - 1.0 / (1<<23);
+	m[ 0] = 1.0 / tan(fovx * M_PI / 360.0);
+	m[ 1] = 0;
+	m[ 2] = 0;
+	m[ 3] = 0;
+	m[ 4] = 0;
+	m[ 5] = 1.0 / tan(fovy * M_PI / 360.0);
+	m[ 6] = 0;
+	m[ 7] = 0;
+	m[ 8] = 0;
+	m[ 9] = 0;
+	m[10] = -1 * nudge;
+	m[11] = -1 * nudge;
+	m[12] = 0;
+	m[13] = 0;
+	m[14] = -2 * zNear * nudge;
+	m[15] = 0;
+	qglLoadMatrixf(m);
+	qglMatrixMode(GL_MODELVIEW);CHECKGLERROR
+	GL_SetupView_Orientation_Identity();
+	backend_projectmatrix.m[0][0] = m[0];
+	backend_projectmatrix.m[1][0] = m[1];
+	backend_projectmatrix.m[2][0] = m[2];
+	backend_projectmatrix.m[3][0] = m[3];
+	backend_projectmatrix.m[0][1] = m[4];
+	backend_projectmatrix.m[1][1] = m[5];
+	backend_projectmatrix.m[2][1] = m[6];
+	backend_projectmatrix.m[3][1] = m[7];
+	backend_projectmatrix.m[0][2] = m[8];
+	backend_projectmatrix.m[1][2] = m[9];
+	backend_projectmatrix.m[2][2] = m[10];
+	backend_projectmatrix.m[3][2] = m[11];
+	backend_projectmatrix.m[0][3] = m[12];
+	backend_projectmatrix.m[1][3] = m[13];
+	backend_projectmatrix.m[2][3] = m[14];
+	backend_projectmatrix.m[3][3] = m[15];
+}
+
 void GL_SetupView_Mode_Ortho (double x1, double y1, double x2, double y2, double zNear, double zFar)
 {
 	if (!r_render.integer)
@@ -247,7 +483,7 @@ void GL_SetupView_Mode_Ortho (double x1, double y1, double x2, double y2, double
 
 	// set up viewpoint
 	qglMatrixMode(GL_PROJECTION);CHECKGLERROR
-	qglLoadIdentity ();CHECKGLERROR
+	qglLoadIdentity();CHECKGLERROR
 	qglOrtho(x1, x2, y2, y1, zNear, zFar);
 	qglMatrixMode(GL_MODELVIEW);CHECKGLERROR
 	GL_SetupView_Orientation_Identity();
@@ -255,9 +491,11 @@ void GL_SetupView_Mode_Ortho (double x1, double y1, double x2, double y2, double
 
 typedef struct gltextureunit_s
 {
-	unsigned int t1d, t2d, t3d, tcubemap;
-	unsigned int arrayenabled;
-	float rgbscale;
+	int t1d, t2d, t3d, tcubemap;
+	int arrayenabled, arrayis3d;
+	void *pointer_texcoord;
+	float rgbscale, alphascale;
+	int combinergb, combinealpha;
 	// FIXME: add more combine stuff
 }
 gltextureunit_t;
@@ -273,6 +511,12 @@ static struct
 	int clientunit;
 	gltextureunit_t units[MAX_TEXTUREUNITS];
 	int colorarray;
+	float color4f[4];
+	int lockrange_first;
+	int lockrange_count;
+	int pointervertexcount;
+	void *pointer_vertex;
+	void *pointer_color;
 }
 gl_state;
 
@@ -291,17 +535,15 @@ void GL_SetupTextureState(void)
 		unit->t2d = 0;
 		unit->t3d = 0;
 		unit->tcubemap = 0;
-		unit->rgbscale = 1;
 		unit->arrayenabled = false;
+		unit->arrayis3d = false;
+		unit->pointer_texcoord = NULL;
+		unit->rgbscale = 1;
+		unit->alphascale = 1;
+		unit->combinergb = GL_MODULATE;
+		unit->combinealpha = GL_MODULATE;
 		qglDisableClientState(GL_TEXTURE_COORD_ARRAY);CHECKGLERROR
-		if (gl_texture3d)
-		{
-			qglTexCoordPointer(3, GL_FLOAT, sizeof(float[4]), varray_texcoord[i]);CHECKGLERROR
-		}
-		else
-		{
-			qglTexCoordPointer(2, GL_FLOAT, sizeof(float[4]), varray_texcoord[i]);CHECKGLERROR
-		}
+		qglTexCoordPointer(2, GL_FLOAT, sizeof(float[2]), varray_buf_texcoord2f[i]);CHECKGLERROR
 		qglDisable(GL_TEXTURE_1D);CHECKGLERROR
 		qglDisable(GL_TEXTURE_2D);CHECKGLERROR
 		if (gl_texture3d)
@@ -318,7 +560,7 @@ void GL_SetupTextureState(void)
 			qglTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_RGB_ARB, GL_MODULATE);CHECKGLERROR
 			qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE0_RGB_ARB, GL_TEXTURE);CHECKGLERROR
 			qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE1_RGB_ARB, GL_PREVIOUS_ARB);CHECKGLERROR
-			qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE2_RGB_ARB, GL_CONSTANT_ARB);CHECKGLERROR
+			qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE2_RGB_ARB, GL_TEXTURE);CHECKGLERROR // for GL_INTERPOLATE_ARB mode
 			qglTexEnvi(GL_TEXTURE_ENV, GL_OPERAND0_RGB_ARB, GL_SRC_COLOR);CHECKGLERROR
 			qglTexEnvi(GL_TEXTURE_ENV, GL_OPERAND1_RGB_ARB, GL_SRC_COLOR);CHECKGLERROR
 			qglTexEnvi(GL_TEXTURE_ENV, GL_OPERAND2_RGB_ARB, GL_SRC_ALPHA);CHECKGLERROR
@@ -348,7 +590,14 @@ void GL_Backend_ResetState(void)
 	gl_state.blend = false;
 	gl_state.depthmask = GL_TRUE;
 	gl_state.colorarray = false;
+	gl_state.color4f[0] = gl_state.color4f[1] = gl_state.color4f[2] = gl_state.color4f[3] = 1;
+	gl_state.lockrange_first = 0;
+	gl_state.lockrange_count = 0;
+	gl_state.pointervertexcount = 0;
+	gl_state.pointer_vertex = NULL;
+	gl_state.pointer_color = NULL;
 
+	CHECKGLERROR
 	qglDisableClientState(GL_COLOR_ARRAY);CHECKGLERROR
 	qglDisableClientState(GL_VERTEX_ARRAY);CHECKGLERROR
 
@@ -358,16 +607,17 @@ void GL_Backend_ResetState(void)
 	qglBlendFunc(gl_state.blendfunc1, gl_state.blendfunc2);CHECKGLERROR
 	qglDisable(GL_BLEND);CHECKGLERROR
 	qglDepthMask(gl_state.depthmask);CHECKGLERROR
-	qglVertexPointer(3, GL_FLOAT, sizeof(float[4]), varray_vertex);CHECKGLERROR
+	qglVertexPointer(3, GL_FLOAT, sizeof(float[3]), varray_buf_vertex3f);CHECKGLERROR
 	qglEnableClientState(GL_VERTEX_ARRAY);CHECKGLERROR
 	if (gl_mesh_floatcolors.integer)
 	{
-		qglColorPointer(4, GL_FLOAT, sizeof(float[4]), varray_color);CHECKGLERROR
+		qglColorPointer(4, GL_FLOAT, sizeof(float[4]), varray_buf_color4f);CHECKGLERROR
 	}
 	else
 	{
-		qglColorPointer(4, GL_UNSIGNED_BYTE, sizeof(qbyte[4]), varray_bcolor);CHECKGLERROR
+		qglColorPointer(4, GL_UNSIGNED_BYTE, sizeof(qbyte[4]), varray_buf_color4b);CHECKGLERROR
 	}
+	GL_Color(0, 0, 0, 0);
 	GL_Color(1, 1, 1, 1);
 
 	GL_SetupTextureState();
@@ -377,6 +627,10 @@ void GL_UseColorArray(void)
 {
 	if (!gl_state.colorarray)
 	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
 		gl_state.colorarray = true;
 		qglEnableClientState(GL_COLOR_ARRAY);CHECKGLERROR
 	}
@@ -386,10 +640,67 @@ void GL_Color(float cr, float cg, float cb, float ca)
 {
 	if (gl_state.colorarray)
 	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
 		gl_state.colorarray = false;
 		qglDisableClientState(GL_COLOR_ARRAY);CHECKGLERROR
+		gl_state.color4f[0] = cr;
+		gl_state.color4f[1] = cg;
+		gl_state.color4f[2] = cb;
+		gl_state.color4f[3] = ca;
+		qglColor4f(cr, cg, cb, ca);
 	}
-	qglColor4f(cr, cg, cb, ca);
+	else
+	{
+		if (gl_state.color4f[0] != cr || gl_state.color4f[1] != cg || gl_state.color4f[2] != cb || gl_state.color4f[3] != ca)
+		{
+#ifdef MESH_BATCH
+			if (gl_batchelementcount)
+				R_Mesh_EndBatch();
+#endif
+			gl_state.color4f[0] = cr;
+			gl_state.color4f[1] = cg;
+			gl_state.color4f[2] = cb;
+			gl_state.color4f[3] = ca;
+			qglColor4f(cr, cg, cb, ca);
+		}
+	}
+}
+
+void GL_LockArrays(int first, int count)
+{
+	if (gl_state.lockrange_count != count || gl_state.lockrange_first != first)
+	{
+		if (gl_state.lockrange_count)
+		{
+			gl_state.lockrange_count = 0;
+			CHECKGLERROR
+			qglUnlockArraysEXT();
+			CHECKGLERROR
+		}
+		if (count && gl_supportslockarrays && gl_lockarrays.integer)
+		{
+			gl_state.lockrange_first = first;
+			gl_state.lockrange_count = count;
+			CHECKGLERROR
+			qglLockArraysEXT(first, count);
+			CHECKGLERROR
+		}
+	}
+}
+
+void GL_TransformToScreen(const vec4_t in, vec4_t out)
+{
+	vec4_t temp;
+	float iw;
+	Matrix4x4_Transform4 (&backend_viewmatrix, in, temp);
+	Matrix4x4_Transform4 (&backend_projectmatrix, temp, out);
+	iw = 1.0f / out[3];
+	out[0] = r_refdef.x + (out[0] * iw + 1.0f) * r_refdef.width * 0.5f;
+	out[1] = r_refdef.y + (out[1] * iw + 1.0f) * r_refdef.height * 0.5f;
+	out[2] = out[2] * iw;
 }
 
 // called at beginning of frame
@@ -400,74 +711,59 @@ void R_Mesh_Start(void)
 	CHECKGLERROR
 
 	GL_Backend_CheckCvars();
-	if (mesh_maxverts != gl_mesh_maxverts.integer)
+	if (mesh_maxverts != gl_mesh_maxverts.integer
+#ifdef MESH_VAR
+	 || mesh_var != (gl_mesh_vertex_array_range.integer && gl_support_var)
+	 || mesh_var_readfrequency != gl_mesh_vertex_array_range_readfrequency.value
+	 || mesh_var_writefrequency != gl_mesh_vertex_array_range_writefrequency.value
+	 || mesh_var_priority != gl_mesh_vertex_array_range_priority.value
+#endif
+		)
 		GL_Backend_ResizeArrays(gl_mesh_maxverts.integer);
 
 	GL_Backend_ResetState();
+#ifdef MESH_VAR
+	if (!mesh_var)
+	{
+		gl_batchvertexfirst = gl_batchvertexcount = gl_batchelementcount = 0;
+		varray_offset = varray_offsetnext = 0;
+	}
+#else
+	varray_offset = varray_offsetnext = 0;
+#endif
 }
 
 int gl_backend_rebindtextures;
 
-void GL_ConvertColorsFloatToByte(int numverts)
+void GL_ConvertColorsFloatToByte(int first, int count)
 {
-	int i, k, total;
-	// LordHavoc: to avoid problems with aliasing (treating memory as two
-	// different types - exactly what this is doing), these must be volatile
-	// (or a union)
-	volatile int *icolor;
-	volatile float *fcolor;
-	qbyte *bcolor;
-
-	total = numverts * 4;
+	int i, k;
+	union {float f[4];int i[4];} *color4fi;
+	struct {GLubyte c[4];} *color4b;
 
 	// shift float to have 8bit fraction at base of number
-	fcolor = varray_color;
-	for (i = 0;i < total;)
+	color4fi = (void *)(varray_buf_color4f + first * 4);
+	for (i = 0;i < count;i++, color4fi++)
 	{
-		fcolor[i    ] += 32768.0f;
-		fcolor[i + 1] += 32768.0f;
-		fcolor[i + 2] += 32768.0f;
-		fcolor[i + 3] += 32768.0f;
-		i += 4;
+		color4fi->f[0] += 32768.0f;
+		color4fi->f[1] += 32768.0f;
+		color4fi->f[2] += 32768.0f;
+		color4fi->f[3] += 32768.0f;
 	}
 
 	// then read as integer and kill float bits...
-	icolor = (int *)varray_color;
-	bcolor = varray_bcolor;
-	for (i = 0;i < total;)
-	{
-		k = icolor[i    ] & 0x7FFFFF;if (k > 255) k = 255;bcolor[i    ] = (qbyte) k;
-		k = icolor[i + 1] & 0x7FFFFF;if (k > 255) k = 255;bcolor[i + 1] = (qbyte) k;
-		k = icolor[i + 2] & 0x7FFFFF;if (k > 255) k = 255;bcolor[i + 2] = (qbyte) k;
-		k = icolor[i + 3] & 0x7FFFFF;if (k > 255) k = 255;bcolor[i + 3] = (qbyte) k;
-		i += 4;
-	}
-}
-
-void GL_DrawRangeElements(int firstvert, int endvert, int indexcount, int *index)
-{
-	int arraylocked = false;
-	c_meshs++;
-	c_meshelements += indexcount;
-	if (gl_supportslockarrays && gl_lockarrays.integer)
+	color4fi = (void *)(varray_buf_color4f + first * 4);
+	color4b = (void *)(varray_buf_color4b + first * 4);
+	for (i = 0;i < count;i++, color4fi++, color4b++)
 	{
-		qglLockArraysEXT(firstvert, endvert - firstvert);
-		CHECKGLERROR
-		arraylocked = true;
-	}
-	if (gl_mesh_drawrangeelements.integer && qglDrawRangeElements != NULL)
-		qglDrawRangeElements(GL_TRIANGLES, firstvert, endvert, indexcount, GL_UNSIGNED_INT, (GLuint *) index);
-	else
-		qglDrawElements(GL_TRIANGLES, indexcount, GL_UNSIGNED_INT, (GLuint *) index);
-	CHECKGLERROR
-	if (arraylocked)
-	{
-		qglUnlockArraysEXT();
-		CHECKGLERROR
-		arraylocked = false;
+		k = color4fi->i[0] & 0x7FFFFF;color4b->c[0] = (GLubyte) min(k, 255);
+		k = color4fi->i[1] & 0x7FFFFF;color4b->c[1] = (GLubyte) min(k, 255);
+		k = color4fi->i[2] & 0x7FFFFF;color4b->c[2] = (GLubyte) min(k, 255);
+		k = color4fi->i[3] & 0x7FFFFF;color4b->c[3] = (GLubyte) min(k, 255);
 	}
 }
 
+/*
 // enlarges geometry buffers if they are too small
 void _R_Mesh_ResizeCheck(int numverts)
 {
@@ -478,19 +774,220 @@ void _R_Mesh_ResizeCheck(int numverts)
 		GL_Backend_ResetState();
 	}
 }
+*/
 
-// renders the mesh
-void R_Mesh_Draw(int numverts, int numtriangles, int *elements)
+void R_Mesh_EndBatch(void)
 {
-	BACKENDACTIVECHECK
+#ifdef MESH_BATCH
+	if (gl_batchelementcount)
+	{
+		if (gl_state.pointervertexcount)
+			Host_Error("R_Mesh_EndBatch: called with pointers enabled\n");
+
+		if (gl_state.colorarray && !gl_mesh_floatcolors.integer && gl_state.pointer_color == NULL)
+			GL_ConvertColorsFloatToByte(gl_batchvertexfirst, gl_batchvertexcount);
+		if (r_render.integer)
+		{
+			//int i;for (i = 0;i < gl_batchelementcount;i++) if (varray_buf_elements3i[i] < gl_batchvertexfirst || varray_buf_elements3i[i] >= (gl_batchvertexfirst + gl_batchvertexcount)) Host_Error("R_Mesh_EndBatch: invalid element #%i (value %i) outside range %i-%i\n", i, varray_buf_elements3i[i], gl_batchvertexfirst, gl_batchvertexfirst + gl_batchvertexcount);
+			CHECKGLERROR
+			GL_LockArrays(gl_batchvertexfirst, gl_batchvertexcount);
+			if (gl_mesh_drawrangeelements.integer && qglDrawRangeElements != NULL)
+			{
+				qglDrawRangeElements(GL_TRIANGLES, gl_batchvertexfirst, gl_batchvertexfirst + gl_batchvertexcount, gl_batchelementcount, GL_UNSIGNED_INT, (const GLuint *) varray_buf_elements3i);CHECKGLERROR
+			}
+			else
+			{
+				qglDrawElements(GL_TRIANGLES, gl_batchelementcount, GL_UNSIGNED_INT, (const GLuint *) varray_buf_elements3i);CHECKGLERROR
+			}
+			GL_LockArrays(0, 0);
+		}
+		gl_batchelementcount = 0;
+		gl_batchvertexcount = 0;
+	}
+#endif
+}
 
+void GL_Backend_RenumberElements(int *out, int count, const int *in, int offset)
+{
+	int i;
+	//if (offset)
+		for (i = 0;i < count;i++)
+			*out++ = *in++ + offset;
+	//else
+	//	memcpy(out, in, sizeof(*out) * count);
+}
+
+// gets vertex buffer space for use with a following R_Mesh_Draw
+// (can be multiple Draw calls per GetSpace)
+void R_Mesh_GetSpace(int numverts)
+{
+	int i;
+
+	if (gl_state.pointervertexcount)
+		Host_Error("R_Mesh_GetSpace: called with pointers enabled\n");
+	if (gl_state.lockrange_count)
+		Host_Error("R_Mesh_GetSpace: called with arrays locked\n");
+
+	varray_offset = varray_offsetnext;
+	if (varray_offset + numverts > mesh_maxverts)
+	{
+		//Con_Printf("R_Mesh_GetSpace: vertex buffer wrap\n");
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
+		varray_offset = 0;
+#ifdef MESH_VAR
+		if (mesh_var)
+		{
+			CHECKGLERROR
+			qglFlushVertexArrayRangeNV();
+			CHECKGLERROR
+		}
+#endif
+		if (numverts > mesh_maxverts)
+		{
+			GL_Backend_ResizeArrays(numverts + 100);
+			GL_Backend_ResetState();
+		}
+	}
+
+	varray_vertex3f = varray_buf_vertex3f + varray_offset * 3;
+	varray_color4f = varray_buf_color4f + varray_offset * 4;
+	for (i = 0;i < backendunits;i++)
+	{
+		varray_texcoord3f[i] = varray_buf_texcoord3f[i] + varray_offset * 3;
+		varray_texcoord2f[i] = varray_buf_texcoord2f[i] + varray_offset * 2;
+	}
+
+	varray_offsetnext = varray_offset + numverts;
+}
+
+// renders triangles using vertices from the most recent GetSpace call
+// (can be multiple Draw calls per GetSpace)
+void R_Mesh_Draw(int numverts, int numtriangles, const int *elements)
+{
+	int numelements = numtriangles * 3;
+	if (numtriangles == 0 || numverts == 0)
+	{
+		Con_Printf("R_Mesh_Draw(%d, %d, %08p);\n", numverts, numtriangles, elements);
+		return;
+	}
+	c_meshs++;
+	c_meshelements += numelements;
 	CHECKGLERROR
+	if (gl_state.pointervertexcount)
+	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
+		if (r_render.integer)
+		{
+			GL_LockArrays(0, gl_state.pointervertexcount);
+			if (gl_mesh_drawrangeelements.integer && qglDrawRangeElements != NULL)
+			{
+				qglDrawRangeElements(GL_TRIANGLES, 0, gl_state.pointervertexcount, numelements, GL_UNSIGNED_INT, elements);CHECKGLERROR
+			}
+			else
+			{
+				qglDrawElements(GL_TRIANGLES, numelements, GL_UNSIGNED_INT, elements);CHECKGLERROR
+			}
+			GL_LockArrays(0, 0);
+		}
+	}
+#ifdef MESH_BATCH
+	else if (gl_mesh_batching.integer)
+	{
+		if (mesh_maxelements < gl_batchelementcount + numelements)
+		{
+			//Con_Printf("R_Mesh_Draw: enlarging elements array\n");
+			if (gl_batchelementcount)
+				R_Mesh_EndBatch();
+			// round up to a multiple of 1024 and add another 1024 just for good measure
+			mesh_maxelements = (gl_batchelementcount + numelements + 1024 + 1023) & ~1023;
+			GL_Backend_AllocElementsArray();
+		}
+		if (varray_offset < gl_batchvertexfirst && gl_batchelementcount)
+			R_Mesh_EndBatch();
+		if (gl_batchelementcount == 0)
+		{
+			gl_batchvertexfirst = varray_offset;
+			gl_batchvertexcount = 0;
+		}
+		if (gl_batchvertexcount < varray_offsetnext - gl_batchvertexfirst)
+			gl_batchvertexcount = varray_offsetnext - gl_batchvertexfirst;
+		GL_Backend_RenumberElements(varray_buf_elements3i + gl_batchelementcount, numelements, elements, varray_offset);
+		//Con_Printf("off %i:%i, vertex %i:%i, element %i:%i\n", varray_offset, varray_offsetnext, gl_batchvertexfirst, gl_batchvertexfirst + gl_batchvertexcount, gl_batchelementcount, gl_batchelementcount + numelements);
+		gl_batchelementcount += numelements;
+		//{int i;for (i = 0;i < gl_batchelementcount;i++) if (varray_buf_elements3i[i] < gl_batchvertexfirst || varray_buf_elements3i[i] >= (gl_batchvertexfirst + gl_batchvertexcount)) Host_Error("R_Mesh_EndBatch: invalid element #%i (value %i) outside range %i-%i, there were previously %i elements and there are now %i elements, varray_offset is %i\n", i, varray_buf_elements3i[i], gl_batchvertexfirst, gl_batchvertexfirst + gl_batchvertexcount, gl_batchelementcount - numelements, gl_batchelementcount, varray_offset);}
+	}
+#endif
+	else
+	{
+		GL_Backend_RenumberElements(varray_buf_elements3i, numelements, elements, varray_offset);
+		if (r_render.integer)
+		{
+			GL_LockArrays(varray_offset, numverts);
+			if (gl_mesh_drawrangeelements.integer && qglDrawRangeElements != NULL)
+			{
+				qglDrawRangeElements(GL_TRIANGLES, varray_offset, varray_offset + numverts, numelements, GL_UNSIGNED_INT, varray_buf_elements3i);CHECKGLERROR
+			}
+			else
+			{
+				qglDrawElements(GL_TRIANGLES, numelements, GL_UNSIGNED_INT, varray_buf_elements3i);CHECKGLERROR
+			}
+			GL_LockArrays(0, 0);
+		}
+	}
+}
 
-	if (gl_state.colorarray && !gl_mesh_floatcolors.integer)
-		GL_ConvertColorsFloatToByte(numverts);
-	if (!r_render.integer)
+// renders triangles using vertices from the most recent GetSpace call
+// (can be multiple Draw calls per GetSpace)
+void R_Mesh_Draw_NoBatching(int numverts, int numtriangles, const int *elements)
+{
+	int numelements = numtriangles * 3;
+	if (numtriangles == 0 || numverts == 0)
+	{
+		Con_Printf("R_Mesh_Draw_NoBatching(%d, %d, %08p);\n", numverts, numtriangles, elements);
 		return;
-	GL_DrawRangeElements(0, numverts, numtriangles * 3, elements);
+	}
+	c_meshs++;
+	c_meshelements += numelements;
+	CHECKGLERROR
+	if (gl_state.pointervertexcount)
+	{
+		if (r_render.integer)
+		{
+			GL_LockArrays(0, gl_state.pointervertexcount);
+			if (gl_mesh_drawrangeelements.integer && qglDrawRangeElements != NULL)
+			{
+				qglDrawRangeElements(GL_TRIANGLES, 0, gl_state.pointervertexcount, numelements, GL_UNSIGNED_INT, elements);CHECKGLERROR
+			}
+			else
+			{
+				qglDrawElements(GL_TRIANGLES, numelements, GL_UNSIGNED_INT, elements);CHECKGLERROR
+			}
+			GL_LockArrays(0, 0);
+		}
+	}
+	else
+	{
+		GL_Backend_RenumberElements(varray_buf_elements3i, numelements, elements, varray_offset);
+		if (r_render.integer)
+		{
+			GL_LockArrays(varray_offset, numverts);
+			if (gl_mesh_drawrangeelements.integer && qglDrawRangeElements != NULL)
+			{
+				qglDrawRangeElements(GL_TRIANGLES, varray_offset, varray_offset + numverts, numelements, GL_UNSIGNED_INT, varray_buf_elements3i);CHECKGLERROR
+			}
+			else
+			{
+				qglDrawElements(GL_TRIANGLES, numelements, GL_UNSIGNED_INT, varray_buf_elements3i);CHECKGLERROR
+			}
+			GL_LockArrays(0, 0);
+		}
+	}
 }
 
 // restores backend state, used when done with 3D rendering
@@ -498,6 +995,11 @@ void R_Mesh_Finish(void)
 {
 	int i;
 	BACKENDACTIVECHECK
+#ifdef MESH_BATCH
+	if (gl_batchelementcount)
+		R_Mesh_EndBatch();
+#endif
+	GL_LockArrays(0, 0);
 
 	for (i = backendunits - 1;i >= 0;i--)
 	{
@@ -520,6 +1022,7 @@ void R_Mesh_Finish(void)
 		if (gl_combine.integer)
 		{
 			qglTexEnvi(GL_TEXTURE_ENV, GL_RGB_SCALE_ARB, 1);CHECKGLERROR
+			qglTexEnvi(GL_TEXTURE_ENV, GL_ALPHA_SCALE, 1);CHECKGLERROR
 		}
 	}
 	qglDisableClientState(GL_COLOR_ARRAY);CHECKGLERROR
@@ -535,6 +1038,10 @@ void R_Mesh_Matrix(const matrix4x4_t *matrix)
 {
 	if (memcmp(matrix, &backend_modelmatrix, sizeof(matrix4x4_t)))
 	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
 		backend_modelmatrix = *matrix;
 		Matrix4x4_Concat(&backend_modelviewmatrix, &backend_viewmatrix, matrix);
 		Matrix4x4_Transpose(&backend_glmodelviewmatrix, &backend_modelviewmatrix);
@@ -545,10 +1052,15 @@ void R_Mesh_Matrix(const matrix4x4_t *matrix)
 // sets up the requested state
 void R_Mesh_MainState(const rmeshstate_t *m)
 {
+	void *p;
 	BACKENDACTIVECHECK
 
 	if (gl_state.blendfunc1 != m->blendfunc1 || gl_state.blendfunc2 != m->blendfunc2)
 	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
 		qglBlendFunc(gl_state.blendfunc1 = m->blendfunc1, gl_state.blendfunc2 = m->blendfunc2);CHECKGLERROR
 		if (gl_state.blendfunc2 == GL_ZERO)
 		{
@@ -580,6 +1092,10 @@ void R_Mesh_MainState(const rmeshstate_t *m)
 	}
 	if (gl_state.depthdisable != m->depthdisable)
 	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
 		gl_state.depthdisable = m->depthdisable;
 		if (gl_state.depthdisable)
 			qglDisable(GL_DEPTH_TEST);
@@ -588,15 +1104,55 @@ void R_Mesh_MainState(const rmeshstate_t *m)
 	}
 	if (gl_state.depthmask != (m->blendfunc2 == GL_ZERO || m->depthwrite))
 	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
 		qglDepthMask(gl_state.depthmask = (m->blendfunc2 == GL_ZERO || m->depthwrite));CHECKGLERROR
 	}
+
+	if (gl_state.pointervertexcount != m->pointervertexcount)
+	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
+		gl_state.pointervertexcount = m->pointervertexcount;
+	}
+
+	p = gl_state.pointervertexcount ? m->pointer_vertex : NULL;
+	if (gl_state.pointer_vertex != p)
+	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
+		gl_state.pointer_vertex = p;
+		qglVertexPointer(3, GL_FLOAT, sizeof(float[3]), p ? p : varray_buf_vertex3f);CHECKGLERROR
+	}
+
+	p = gl_state.pointervertexcount ? m->pointer_color : NULL;
+	if (gl_state.pointer_color != p)
+	{
+#ifdef MESH_BATCH
+		if (gl_batchelementcount)
+			R_Mesh_EndBatch();
+#endif
+		gl_state.pointer_color = p;
+		if (p || gl_mesh_floatcolors.integer)
+			qglColorPointer(4, GL_FLOAT, sizeof(float[4]), p ? p : varray_buf_color4f);
+		else
+			qglColorPointer(4, GL_UNSIGNED_BYTE, sizeof(GLubyte[4]), p ? p : varray_buf_color4b);
+		CHECKGLERROR
+	}
 }
 
 void R_Mesh_TextureState(const rmeshstate_t *m)
 {
-	int i;
-	float rgbscale;
+	int i, combinergb, combinealpha;
+	float scale;
 	gltextureunit_t *unit;
+	void *p;
 
 	BACKENDACTIVECHECK
 
@@ -611,14 +1167,56 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 		unit = gl_state.units + i;
 		if (unit->t1d != m->tex1d[i] || unit->t2d != m->tex[i] || unit->t3d != m->tex3d[i] || unit->tcubemap != m->texcubemap[i])
 		{
-			if (gl_state.unit != i)
+			if (m->tex3d[i] || m->texcubemap[i])
 			{
-				qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+				if (!unit->arrayis3d)
+				{
+#ifdef MESH_BATCH
+					if (gl_batchelementcount)
+						R_Mesh_EndBatch();
+#endif
+					unit->arrayis3d = true;
+					if (gl_state.clientunit != i)
+					{
+						qglClientActiveTexture(GL_TEXTURE0_ARB + (gl_state.clientunit = i));CHECKGLERROR
+					}
+					qglTexCoordPointer(3, GL_FLOAT, sizeof(float[3]), varray_buf_texcoord3f[i]);
+				}
+				if (!unit->arrayenabled)
+				{
+#ifdef MESH_BATCH
+					if (gl_batchelementcount)
+						R_Mesh_EndBatch();
+#endif
+					unit->arrayenabled = true;
+					if (gl_state.clientunit != i)
+					{
+						qglClientActiveTexture(GL_TEXTURE0_ARB + (gl_state.clientunit = i));CHECKGLERROR
+					}
+					qglEnableClientState(GL_TEXTURE_COORD_ARRAY);CHECKGLERROR
+				}
 			}
-			if (m->tex1d[i] || m->tex[i] || m->tex3d[i] || m->texcubemap[i])
+			else if (m->tex1d[i] || m->tex[i])
 			{
+				if (unit->arrayis3d)
+				{
+#ifdef MESH_BATCH
+					if (gl_batchelementcount)
+						R_Mesh_EndBatch();
+#endif
+					unit->arrayis3d = false;
+					if (gl_state.clientunit != i)
+					{
+						qglClientActiveTexture(GL_TEXTURE0_ARB + (gl_state.clientunit = i));CHECKGLERROR
+					}
+					qglTexCoordPointer(2, GL_FLOAT, sizeof(float[2]), varray_buf_texcoord2f[i]);
+				}
 				if (!unit->arrayenabled)
 				{
+#ifdef MESH_BATCH
+					if (gl_batchelementcount)
+						R_Mesh_EndBatch();
+#endif
 					unit->arrayenabled = true;
 					if (gl_state.clientunit != i)
 					{
@@ -631,6 +1229,10 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 			{
 				if (unit->arrayenabled)
 				{
+#ifdef MESH_BATCH
+					if (gl_batchelementcount)
+						R_Mesh_EndBatch();
+#endif
 					unit->arrayenabled = false;
 					if (gl_state.clientunit != i)
 					{
@@ -641,6 +1243,14 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 			}
 			if (unit->t1d != m->tex1d[i])
 			{
+#ifdef MESH_BATCH
+				if (gl_batchelementcount)
+					R_Mesh_EndBatch();
+#endif
+				if (gl_state.unit != i)
+				{
+					qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+				}
 				if (m->tex1d[i])
 				{
 					if (unit->t1d == 0)
@@ -655,6 +1265,14 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 			}
 			if (unit->t2d != m->tex[i])
 			{
+#ifdef MESH_BATCH
+				if (gl_batchelementcount)
+					R_Mesh_EndBatch();
+#endif
+				if (gl_state.unit != i)
+				{
+					qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+				}
 				if (m->tex[i])
 				{
 					if (unit->t2d == 0)
@@ -669,6 +1287,14 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 			}
 			if (unit->t3d != m->tex3d[i])
 			{
+#ifdef MESH_BATCH
+				if (gl_batchelementcount)
+					R_Mesh_EndBatch();
+#endif
+				if (gl_state.unit != i)
+				{
+					qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+				}
 				if (m->tex3d[i])
 				{
 					if (unit->t3d == 0)
@@ -683,6 +1309,14 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 			}
 			if (unit->tcubemap != m->texcubemap[i])
 			{
+#ifdef MESH_BATCH
+				if (gl_batchelementcount)
+					R_Mesh_EndBatch();
+#endif
+				if (gl_state.unit != i)
+				{
+					qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+				}
 				if (m->texcubemap[i])
 				{
 					if (unit->tcubemap == 0)
@@ -696,14 +1330,94 @@ void R_Mesh_TextureState(const rmeshstate_t *m)
 				qglBindTexture(GL_TEXTURE_CUBE_MAP_ARB, (unit->tcubemap = m->texcubemap[i]));CHECKGLERROR
 			}
 		}
-		rgbscale = max(m->texrgbscale[i], 1);
-		if (gl_state.units[i].rgbscale != rgbscale)
+		combinergb = m->texcombinergb[i];
+		if (!combinergb)
+			combinergb = GL_MODULATE;
+		if (unit->combinergb != combinergb)
 		{
+#ifdef MESH_BATCH
+			if (gl_batchelementcount)
+				R_Mesh_EndBatch();
+#endif
 			if (gl_state.unit != i)
 			{
 				qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
 			}
-			qglTexEnvi(GL_TEXTURE_ENV, GL_RGB_SCALE_ARB, (gl_state.units[i].rgbscale = rgbscale));CHECKGLERROR
+			unit->combinergb = combinergb;
+			if (gl_combine.integer)
+			{
+				qglTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_RGB_ARB, unit->combinergb);CHECKGLERROR
+			}
+			else
+			{
+				qglTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, unit->combinergb);CHECKGLERROR
+			}
+		}
+		combinealpha = m->texcombinealpha[i];
+		if (!combinealpha)
+			combinealpha = GL_MODULATE;
+		if (unit->combinealpha != combinealpha)
+		{
+#ifdef MESH_BATCH
+			if (gl_batchelementcount)
+				R_Mesh_EndBatch();
+#endif
+			if (gl_state.unit != i)
+			{
+				qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+			}
+			unit->combinealpha = combinealpha;
+			if (gl_combine.integer)
+			{
+				qglTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_ALPHA_ARB, unit->combinealpha);CHECKGLERROR
+			}
+		}
+		scale = max(m->texrgbscale[i], 1);
+		if (gl_state.units[i].rgbscale != scale)
+		{
+#ifdef MESH_BATCH
+			if (gl_batchelementcount)
+				R_Mesh_EndBatch();
+#endif
+			if (gl_state.unit != i)
+			{
+				qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+			}
+			qglTexEnvi(GL_TEXTURE_ENV, GL_RGB_SCALE_ARB, (gl_state.units[i].rgbscale = scale));CHECKGLERROR
+		}
+		scale = max(m->texalphascale[i], 1);
+		if (gl_state.units[i].alphascale != scale)
+		{
+#ifdef MESH_BATCH
+			if (gl_batchelementcount)
+				R_Mesh_EndBatch();
+#endif
+			if (gl_state.unit != i)
+			{
+				qglActiveTexture(GL_TEXTURE0_ARB + (gl_state.unit = i));CHECKGLERROR
+			}
+			qglTexEnvi(GL_TEXTURE_ENV, GL_ALPHA_SCALE, (gl_state.units[i].alphascale = scale));CHECKGLERROR
+		}
+		if (unit->arrayenabled)
+		{
+			p = gl_state.pointervertexcount ? m->pointer_texcoord[i] : NULL;
+			if (unit->pointer_texcoord != p)
+			{
+#ifdef MESH_BATCH
+				if (gl_batchelementcount)
+					R_Mesh_EndBatch();
+#endif
+				unit->pointer_texcoord = p;
+				if (gl_state.clientunit != i)
+				{
+					qglClientActiveTexture(GL_TEXTURE0_ARB + (gl_state.clientunit = i));CHECKGLERROR
+				}
+				if (unit->arrayis3d)
+					qglTexCoordPointer(3, GL_FLOAT, sizeof(float[3]), p ? p : varray_buf_texcoord3f[i]);
+				else
+					qglTexCoordPointer(2, GL_FLOAT, sizeof(float[2]), p ? p : varray_buf_texcoord2f[i]);
+				CHECKGLERROR
+			}
 		}
 	}
 }
@@ -722,10 +1436,10 @@ void R_Mesh_State(const rmeshstate_t *m)
 ==============================================================================
 */
 
-qboolean SCR_ScreenShot(char *filename, int x, int y, int width, int height)
+qboolean SCR_ScreenShot(char *filename, int x, int y, int width, int height, qboolean jpeg)
 {
 	qboolean ret;
-	int i;
+	int i, j;
 	qbyte *buffer;
 
 	if (!r_render.integer)
@@ -737,10 +1451,18 @@ qboolean SCR_ScreenShot(char *filename, int x, int y, int width, int height)
 
 	// LordHavoc: compensate for v_overbrightbits when using hardware gamma
 	if (v_hwgamma.integer)
+	{
 		for (i = 0;i < width * height * 3;i++)
-			buffer[i] <<= v_overbrightbits.integer;
+		{
+			j = buffer[i] << v_overbrightbits.integer;
+			buffer[i] = (qbyte) (bound(0, j, 255));
+		}
+	}
 
-	ret = Image_WriteTGARGB_preflipped(filename, width, height, buffer);
+	if (jpeg)
+		ret = JPEG_SaveImage_preflipped (filename, width, height, buffer);
+	else
+		ret = Image_WriteTGARGB_preflipped (filename, width, height, buffer);
 
 	Mem_Free(buffer);
 	return ret;
@@ -754,8 +1476,16 @@ void R_ClearScreen(void)
 	{
 		// clear to black
 		qglClearColor(0,0,0,0);CHECKGLERROR
+		qglClearDepth(1);CHECKGLERROR
+		if (gl_stencil)
+		{
+			// LordHavoc: we use a stencil centered around 128 instead of 0,
+			// to avoid clamping interfering with strange shadow volume
+			// drawing orders
+			qglClearStencil(128);CHECKGLERROR
+		}
 		// clear the screen
-		qglClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);CHECKGLERROR
+		qglClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT | (gl_stencil ? GL_STENCIL_BUFFER_BIT : 0));CHECKGLERROR
 		// set dithering mode
 		if (gl_dither.integer)
 		{
@@ -778,9 +1508,12 @@ text to the screen.
 */
 void SCR_UpdateScreen (void)
 {
-	VID_Finish ();
+	if (gl_delayfinish.integer)
+	{
+		VID_Finish ();
 
-	R_TimeReport("finish");
+		R_TimeReport("finish");
+	}
 
 	if (r_textureunits.integer > gl_textureunits)
 		Cvar_SetValueQuick(&r_textureunits, gl_textureunits);
@@ -810,8 +1543,330 @@ void SCR_UpdateScreen (void)
 	// draw 2D stuff
 	R_DrawQueue();
 
-	// tell driver to commit it's partially full geometry queue to the rendering queue
-	// (this doesn't wait for the commands themselves to complete)
-	qglFlush();
+	if (gl_delayfinish.integer)
+	{
+		// tell driver to commit it's partially full geometry queue to the rendering queue
+		// (this doesn't wait for the commands themselves to complete)
+		qglFlush();
+	}
+	else
+	{
+		VID_Finish ();
+
+		R_TimeReport("finish");
+	}
+}
+
+// utility functions
+
+void R_Mesh_CopyVertex3f(const float *vertex3f, int numverts)
+{
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		float *out = varray_vertex3f;
+		while (--numverts)
+		{
+			*out++ = *vertex3f++;
+			*out++ = *vertex3f++;
+			*out++ = *vertex3f++;
+		}
+	}
+	else
+#endif
+		memcpy(varray_vertex3f, vertex3f, numverts * sizeof(float[3]));
+}
+
+void R_Mesh_CopyTexCoord2f(int tmu, const float *texcoord2f, int numverts)
+{
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		float *out = varray_texcoord2f[tmu];
+		while (numverts--)
+		{
+			*out++ = *texcoord2f++;
+			*out++ = *texcoord2f++;
+		}
+	}
+	else
+#endif
+		memcpy(varray_texcoord2f[tmu], texcoord2f, numverts * sizeof(float[2]));
+}
+
+void R_Mesh_CopyColor4f(const float *color4f, int numverts)
+{
+#ifdef MESH_VAR
+	if (mesh_var)
+	{
+		float *out = varray_color4f;
+		while (numverts--)
+		{
+			*out++ = *color4f++;
+			*out++ = *color4f++;
+			*out++ = *color4f++;
+			*out++ = *color4f++;
+		}
+	}
+	else
+#endif
+		memcpy(varray_color4f, color4f, numverts * sizeof(float[4]));
+}
+
+void R_ScrollTexCoord2f (float *out2f, const float *in2f, int numverts, float s, float t)
+{
+	while (numverts--)
+	{
+		*out2f++ = *in2f++ + s;
+		*out2f++ = *in2f++ + t;
+	}
+}
+
+//===========================================================================
+// vertex array caching subsystem
+//===========================================================================
+
+typedef struct rcachearraylink_s
+{
+	struct rcachearraylink_s *next, *prev;
+	struct rcachearrayitem_s *data;
+}
+rcachearraylink_t;
+
+typedef struct rcachearrayitem_s
+{
+	// the original request structure
+	rcachearrayrequest_t request;
+	// active
+	int active;
+	// offset into r_mesh_rcachedata
+	int offset;
+	// for linking this into the sequential list
+	rcachearraylink_t sequentiallink;
+	// for linking this into the lookup list
+	rcachearraylink_t hashlink;
+}
+rcachearrayitem_t;
+
+#define RCACHEARRAY_HASHSIZE 65536
+#define RCACHEARRAY_ITEMS 4096
+#define RCACHEARRAY_DEFAULTSIZE (4 << 20)
+
+// all active items are linked into this chain in sorted order
+static rcachearraylink_t r_mesh_rcachesequentialchain;
+// all inactive items are linked into this chain in unknown order
+static rcachearraylink_t r_mesh_rcachefreechain;
+// all active items are also linked into these chains (using their hashlink)
+static rcachearraylink_t r_mesh_rcachechain[RCACHEARRAY_HASHSIZE];
+
+// all items are stored here, whether active or inactive
+static rcachearrayitem_t r_mesh_rcacheitems[RCACHEARRAY_ITEMS];
+
+// size of data buffer
+static int r_mesh_rcachedata_size = RCACHEARRAY_DEFAULTSIZE;
+// data buffer
+static qbyte r_mesh_rcachedata[RCACHEARRAY_DEFAULTSIZE];
+
+// current state
+static int r_mesh_rcachedata_offset;
+static rcachearraylink_t *r_mesh_rcachesequentialchain_current;
+
+static void R_Mesh_CacheArray_Startup(void)
+{
+	int i;
+	rcachearraylink_t *l;
+	// prepare all the linked lists
+	l = &r_mesh_rcachesequentialchain;l->next = l->prev = l;l->data = NULL;
+	l = &r_mesh_rcachefreechain;l->next = l->prev = l;l->data = NULL;
+	memset(&r_mesh_rcachechain, 0, sizeof(r_mesh_rcachechain));
+	for (i = 0;i < RCACHEARRAY_HASHSIZE;i++)
+	{
+		l = &r_mesh_rcachechain[i];
+		l->next = l->prev = l;
+		l->data = NULL;
+	}
+	memset(&r_mesh_rcacheitems, 0, sizeof(r_mesh_rcacheitems));
+	for (i = 0;i < RCACHEARRAY_ITEMS;i++)
+	{
+		r_mesh_rcacheitems[i].hashlink.data = r_mesh_rcacheitems[i].sequentiallink.data = &r_mesh_rcacheitems[i];
+		l = &r_mesh_rcacheitems[i].sequentiallink;
+		l->next = &r_mesh_rcachefreechain;
+		l->prev = l->next->prev;
+		l->next->prev = l->prev->next = l;
+	}
+	// clear other state
+	r_mesh_rcachedata_offset = 0;
+	r_mesh_rcachesequentialchain_current = &r_mesh_rcachesequentialchain;
+}
+
+static void R_Mesh_CacheArray_Shutdown(void)
+{
+}
+
+/*
+static void R_Mesh_CacheArray_ValidateState(int num)
+{
+	rcachearraylink_t *l, *lhead;
+	lhead = &r_mesh_rcachesequentialchain;
+	if (r_mesh_rcachesequentialchain_current == lhead)
+		return;
+	for (l = lhead->next;l != lhead;l = l->next)
+		if (r_mesh_rcachesequentialchain_current == l)
+			return;
+	Sys_Error("%i", num);
+}
+*/
+
+int R_Mesh_CacheArray(rcachearrayrequest_t *r)
+{
+	rcachearraylink_t *l, *lhead, *lnext;
+	rcachearrayitem_t *d;
+	int hashindex, offset, offsetend;
+
+	//R_Mesh_CacheArray_ValidateState(3);
+	// calculate a hashindex to choose a cache chain
+	r->data = NULL;
+	hashindex = CRC_Block((void *)r, sizeof(*r)) % RCACHEARRAY_HASHSIZE;
+
+	// is it already cached?
+	for (lhead = &r_mesh_rcachechain[hashindex], l = lhead->next;l != lhead;l = l->next)
+	{
+		if (!memcmp(&l->data->request, r, sizeof(l->data->request)))
+		{
+			// we have it cached already
+			r->data = r_mesh_rcachedata + l->data->offset;
+			return false;
+		}
+	}
+
+	// we need to add a new cache item, this means finding a place for the new
+	// data and making sure we have a free item available, lots of work...
+
+	// check if buffer needs to wrap
+	if (r_mesh_rcachedata_offset + r->data_size > r_mesh_rcachedata_size)
+	{
+		/*
+		if (r->data_size * 10 > r_mesh_rcachedata_size)
+		{
+			// realloc whole cache
+		}
+		*/
+		// reset back to start
+		r_mesh_rcachedata_offset = 0;
+		r_mesh_rcachesequentialchain_current = &r_mesh_rcachesequentialchain;
+	}
+	offset = r_mesh_rcachedata_offset;
+	r_mesh_rcachedata_offset += r->data_size;
+	offsetend = r_mesh_rcachedata_offset;
+	//R_Mesh_CacheArray_ValidateState(4);
+
+	/*
+	{
+		int n;
+		for (lhead = &r_mesh_rcachesequentialchain, l = lhead->next, n = 0;l != lhead;l = l->next, n++);
+		Con_Printf("R_Mesh_CacheArray: new data range %i:%i, %i items are already linked\n", offset, offsetend, n);
+	}
+	*/
+
+	// make room for the new data (remove old items)
+	lhead = &r_mesh_rcachesequentialchain;
+	l = r_mesh_rcachesequentialchain_current;
+	if (l == lhead)
+		l = l->next;
+	while (l != lhead && l->data->offset < offsetend && l->data->offset + l->data->request.data_size > offset)
+	{
+	//r_mesh_rcachesequentialchain_current = l;
+	//R_Mesh_CacheArray_ValidateState(8);
+		lnext = l->next;
+		// if at the end of the chain, wrap around
+		if (lnext == lhead)
+			lnext = lnext->next;
+	//r_mesh_rcachesequentialchain_current = lnext;
+	//R_Mesh_CacheArray_ValidateState(10);
+
+		// unlink from sequential chain
+		l->next->prev = l->prev;
+		l->prev->next = l->next;
+	//R_Mesh_CacheArray_ValidateState(11);
+		// link into free chain
+		l->next = &r_mesh_rcachefreechain;
+		l->prev = l->next->prev;
+		l->next->prev = l->prev->next = l;
+	//R_Mesh_CacheArray_ValidateState(12);
+
+		l = &l->data->hashlink;
+		// unlink from hash chain
+		l->next->prev = l->prev;
+		l->prev->next = l->next;
+
+		l = lnext;
+	//r_mesh_rcachesequentialchain_current = l;
+	//R_Mesh_CacheArray_ValidateState(9);
+	}
+	//r_mesh_rcachesequentialchain_current = l;
+	//R_Mesh_CacheArray_ValidateState(5);
+	// gobble an extra item if we have no free items available
+	if (r_mesh_rcachefreechain.next == &r_mesh_rcachefreechain)
+	{
+		lnext = l->next;
+
+		// unlink from sequential chain
+		l->next->prev = l->prev;
+		l->prev->next = l->next;
+		// link into free chain
+		l->next = &r_mesh_rcachefreechain;
+		l->prev = l->next->prev;
+		l->next->prev = l->prev->next = l;
+
+		l = &l->data->hashlink;
+		// unlink from hash chain
+		l->next->prev = l->prev;
+		l->prev->next = l->next;
+
+		l = lnext;
+	}
+	r_mesh_rcachesequentialchain_current = l;
+	//R_Mesh_CacheArray_ValidateState(6);
+
+	// now take an item from the free chain
+	l = r_mesh_rcachefreechain.next;
+	// set it up
+	d = l->data;
+	d->request = *r;
+	d->offset = offset;
+	// unlink
+	l->next->prev = l->prev;
+	l->prev->next = l->next;
+	// relink to sequential
+	l->next = r_mesh_rcachesequentialchain_current->prev;
+	l->prev = l->next->prev;
+	while (l->next->data && l->data && l->next->data->offset <= d->offset)
+	{
+		//Con_Printf(">\n");
+		l->next = l->next->next;
+		l->prev = l->prev->next;
+	}
+	while (l->prev->data && l->data && l->prev->data->offset >= d->offset)
+	{
+		//Con_Printf("<\n");
+		l->prev = l->prev->prev;
+		l->next = l->next->prev;
+	}
+	l->next->prev = l->prev->next = l;
+	// also link into hash chain
+	l = &l->data->hashlink;
+	l->next = &r_mesh_rcachechain[hashindex];
+	l->prev = l->next->prev;
+	l->prev->next = l;
+	l->next->prev = l->prev->next = l;
+
+
+	//r_mesh_rcachesequentialchain_current = d->sequentiallink.next;
+
+	//R_Mesh_CacheArray_ValidateState(7);
+	// and finally set the data pointer
+	r->data = r_mesh_rcachedata + d->offset;
+	// and tell the caller to fill the array
+	return true;
 }