/*
===========================================================================
Copyright (C) 1999-2005 Id Software, Inc.

This file is part of Quake III Arena source code.

Quake III Arena source code is free software; you can redistribute it
and/or modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the License,
or (at your option) any later version.

Quake III Arena source code is distributed in the hope that it will be
useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with Quake III Arena source code; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
===========================================================================
*/
#include "tr_local.h"

backEndData_t	*backEndData[SMP_FRAMES];
backEndState_t	backEnd;


void GL_Bind( const image_t* image )
{
	int texnum;

	if ( !image ) {
		ri.Printf( PRINT_WARNING, "GL_Bind: NULL image\n" );
		texnum = tr.defaultImage->texnum;
	} else {
		texnum = image->texnum;
	}

	if ( r_nobind->integer && tr.defaultImage && !backEnd.projection2D ) {
		texnum = tr.defaultImage->texnum;
	}

	if ( glState.texID[glState.currenttmu] != texnum ) {
		glState.texID[glState.currenttmu] = texnum;
		qglBindTexture( GL_TEXTURE_2D, texnum );
	}
}


void GL_SelectTexture( int unit )
{
	if ( glState.currenttmu == unit )
		return;

	if ( unit >= MAX_TMUS )
		ri.Error( ERR_DROP, "GL_SelectTexture: unit = %i", unit );

	qglActiveTextureARB( GL_TEXTURE0_ARB + unit );
	qglClientActiveTextureARB( GL_TEXTURE0_ARB + unit );

	glState.currenttmu = unit;
}


/*
** GL_Cull
*/
void GL_Cull( int cullType ) {
	if ( glState.faceCulling == cullType ) {
		return;
	}

	glState.faceCulling = cullType;

	if ( cullType == CT_TWO_SIDED )
	{
		qglDisable( GL_CULL_FACE );
	}
	else
	{
		qglEnable( GL_CULL_FACE );

		if ( cullType == CT_BACK_SIDED )
		{
			if ( backEnd.viewParms.isMirror )
			{
				qglCullFace( GL_FRONT );
			}
			else
			{
				qglCullFace( GL_BACK );
			}
		}
		else
		{
			if ( backEnd.viewParms.isMirror )
			{
				qglCullFace( GL_BACK );
			}
			else
			{
				qglCullFace( GL_FRONT );
			}
		}
	}
}

/*
** GL_TexEnv
*/
void GL_TexEnv( int env )
{
	if ( env == glState.texEnv[glState.currenttmu] )
	{
		return;
	}

	glState.texEnv[glState.currenttmu] = env;

	switch ( env )
	{
	case GL_MODULATE:
	case GL_REPLACE:
	case GL_DECAL:
	case GL_ADD:
		qglTexEnvf( GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, env );
		break;
	default:
		ri.Error( ERR_DROP, "GL_TexEnv: invalid env '%d' passed\n", env );
		break;
	}
}

/*
** GL_State
**
** This routine is responsible for setting the most commonly changed state
** in Q3.
*/
void GL_State( unsigned long stateBits )
{
	unsigned long diff = stateBits ^ glState.glStateBits;

	if ( !diff )
	{
		return;
	}

	//
	// check depthFunc bits
	//
	if ( diff & GLS_DEPTHFUNC_EQUAL )
	{
		if ( stateBits & GLS_DEPTHFUNC_EQUAL )
		{
			qglDepthFunc( GL_EQUAL );
		}
		else
		{
			qglDepthFunc( GL_LEQUAL );
		}
	}

	//
	// check blend bits
	//
	if ( diff & ( GLS_SRCBLEND_BITS | GLS_DSTBLEND_BITS ) )
	{
		GLenum srcFactor, dstFactor;

		if ( stateBits & ( GLS_SRCBLEND_BITS | GLS_DSTBLEND_BITS ) )
		{
			switch ( stateBits & GLS_SRCBLEND_BITS )
			{
			case GLS_SRCBLEND_ZERO:
				srcFactor = GL_ZERO;
				break;
			case GLS_SRCBLEND_ONE:
				srcFactor = GL_ONE;
				break;
			case GLS_SRCBLEND_DST_COLOR:
				srcFactor = GL_DST_COLOR;
				break;
			case GLS_SRCBLEND_ONE_MINUS_DST_COLOR:
				srcFactor = GL_ONE_MINUS_DST_COLOR;
				break;
			case GLS_SRCBLEND_SRC_ALPHA:
				srcFactor = GL_SRC_ALPHA;
				break;
			case GLS_SRCBLEND_ONE_MINUS_SRC_ALPHA:
				srcFactor = GL_ONE_MINUS_SRC_ALPHA;
				break;
			case GLS_SRCBLEND_DST_ALPHA:
				srcFactor = GL_DST_ALPHA;
				break;
			case GLS_SRCBLEND_ONE_MINUS_DST_ALPHA:
				srcFactor = GL_ONE_MINUS_DST_ALPHA;
				break;
			case GLS_SRCBLEND_ALPHA_SATURATE:
				srcFactor = GL_SRC_ALPHA_SATURATE;
				break;
			default:
				srcFactor = GL_ONE;		// to get warning to shut up
				ri.Error( ERR_DROP, "GL_State: invalid src blend state bits\n" );
				break;
			}

			switch ( stateBits & GLS_DSTBLEND_BITS )
			{
			case GLS_DSTBLEND_ZERO:
				dstFactor = GL_ZERO;
				break;
			case GLS_DSTBLEND_ONE:
				dstFactor = GL_ONE;
				break;
			case GLS_DSTBLEND_SRC_COLOR:
				dstFactor = GL_SRC_COLOR;
				break;
			case GLS_DSTBLEND_ONE_MINUS_SRC_COLOR:
				dstFactor = GL_ONE_MINUS_SRC_COLOR;
				break;
			case GLS_DSTBLEND_SRC_ALPHA:
				dstFactor = GL_SRC_ALPHA;
				break;
			case GLS_DSTBLEND_ONE_MINUS_SRC_ALPHA:
				dstFactor = GL_ONE_MINUS_SRC_ALPHA;
				break;
			case GLS_DSTBLEND_DST_ALPHA:
				dstFactor = GL_DST_ALPHA;
				break;
			case GLS_DSTBLEND_ONE_MINUS_DST_ALPHA:
				dstFactor = GL_ONE_MINUS_DST_ALPHA;
				break;
			default:
				dstFactor = GL_ONE;		// to get warning to shut up
				ri.Error( ERR_DROP, "GL_State: invalid dst blend state bits\n" );
				break;
			}

			qglEnable( GL_BLEND );
			qglBlendFunc( srcFactor, dstFactor );
		}
		else
		{
			qglDisable( GL_BLEND );
		}
	}

	//
	// check depthmask
	//
	if ( diff & GLS_DEPTHMASK_TRUE )
	{
		if ( stateBits & GLS_DEPTHMASK_TRUE )
		{
			qglDepthMask( GL_TRUE );
		}
		else
		{
			qglDepthMask( GL_FALSE );
		}
	}

	//
	// fill/line mode
	//
	if ( diff & GLS_POLYMODE_LINE )
	{
		if ( stateBits & GLS_POLYMODE_LINE )
		{
			qglPolygonMode( GL_FRONT_AND_BACK, GL_LINE );
		}
		else
		{
			qglPolygonMode( GL_FRONT_AND_BACK, GL_FILL );
		}
	}

	//
	// depthtest
	//
	if ( diff & GLS_DEPTHTEST_DISABLE )
	{
		if ( stateBits & GLS_DEPTHTEST_DISABLE )
		{
			qglDisable( GL_DEPTH_TEST );
		}
		else
		{
			qglEnable( GL_DEPTH_TEST );
		}
	}

	//
	// alpha test
	//
	if ( diff & GLS_ATEST_BITS )
	{
		switch ( stateBits & GLS_ATEST_BITS )
		{
		case 0:
			qglDisable( GL_ALPHA_TEST );
			break;
		case GLS_ATEST_GT_0:
			qglEnable( GL_ALPHA_TEST );
			qglAlphaFunc( GL_GREATER, 0.0f );
			break;
		case GLS_ATEST_LT_80:
			qglEnable( GL_ALPHA_TEST );
			qglAlphaFunc( GL_LESS, 0.5f );
			break;
		case GLS_ATEST_GE_80:
			qglEnable( GL_ALPHA_TEST );
			qglAlphaFunc( GL_GEQUAL, 0.5f );
			break;
		default:
			assert( 0 );
			break;
		}
	}

	glState.glStateBits = stateBits;
}


// a player has predicted a teleport, but hasn't arrived yet

static void RB_Hyperspace()
{
	float c = 0.25 + 0.5 * sin( M_PI * (backEnd.refdef.time & 0x01FF) / 0x0200 );
	qglClearColor( c, c, c, 1 );
	qglClear( GL_COLOR_BUFFER_BIT );
}


static void SetViewportAndScissor()
{
	qglMatrixMode(GL_PROJECTION);
	qglLoadMatrixf( backEnd.viewParms.projectionMatrix );
	qglMatrixMode(GL_MODELVIEW);

	// set the window clipping
	qglViewport( backEnd.viewParms.viewportX, backEnd.viewParms.viewportY,
		backEnd.viewParms.viewportWidth, backEnd.viewParms.viewportHeight );
	qglScissor( backEnd.viewParms.viewportX, backEnd.viewParms.viewportY,
		backEnd.viewParms.viewportWidth, backEnd.viewParms.viewportHeight );
}


// any mirrored or portaled views have already been drawn
// so prepare to actually render the visible surfaces for this view

static void RB_BeginDrawingView()
{
	int clearBits = 0;

	// sync with gl if needed
	if ( r_finish->integer == 1 && !glState.finishCalled ) {
		qglFinish();
		glState.finishCalled = qtrue;
	}
	if ( r_finish->integer == 0 ) {
		glState.finishCalled = qtrue;
	}

	// we will need to change the projection matrix before drawing 2D images again
	backEnd.projection2D = qfalse;
	backEnd.pc = backEnd.pc3D;

	//
	// set the modelview matrix for the viewer
	//
	SetViewportAndScissor();

	// ensures that depth writes are enabled for the depth clear
	GL_State( GLS_DEFAULT );
	// clear relevant buffers
	clearBits = GL_DEPTH_BUFFER_BIT;

	if ( r_measureOverdraw->integer )
	{
		clearBits |= GL_STENCIL_BUFFER_BIT;
	}
	if ( r_fastsky->integer && !( backEnd.refdef.rdflags & RDF_NOWORLDMODEL ) )
	{
		clearBits |= GL_COLOR_BUFFER_BIT;
		// tr.sunLight could have colored fastsky properly for the last 9 years,
		// ... if the code had actually been right >:(  but, it's a bad idea to trust mappers anyway
		qglClearColor( 0.0f, 0.0f, 0.0f, 1.0f );
	}
	qglClear( clearBits );

	if ( backEnd.refdef.rdflags & RDF_HYPERSPACE ) {
		RB_Hyperspace();
		return;
	}

	glState.faceCulling = -1;		// force face culling to set next time

	// clip to the plane of the portal
	if ( backEnd.viewParms.isPortal ) {
		float	plane[4];
		double	plane2[4];

		plane[0] = backEnd.viewParms.portalPlane.normal[0];
		plane[1] = backEnd.viewParms.portalPlane.normal[1];
		plane[2] = backEnd.viewParms.portalPlane.normal[2];
		plane[3] = backEnd.viewParms.portalPlane.dist;

		plane2[0] = DotProduct (backEnd.viewParms.orient.axis[0], plane);
		plane2[1] = DotProduct (backEnd.viewParms.orient.axis[1], plane);
		plane2[2] = DotProduct (backEnd.viewParms.orient.axis[2], plane);
		plane2[3] = DotProduct (plane, backEnd.viewParms.orient.origin) - plane[3];

		qglLoadMatrixf( s_flipMatrix );
		qglClipPlane (GL_CLIP_PLANE0, plane2);
		qglEnable (GL_CLIP_PLANE0);
	} else {
		qglDisable (GL_CLIP_PLANE0);
	}
}


static void RB_RenderDrawSurfList( const drawSurf_t* drawSurfs, int numDrawSurfs )
{
	int i;
	const shader_t* shader = NULL;
	unsigned int sort = (unsigned int)-1;

	// save original time for entity shader offsets
	double originalTime = backEnd.refdef.floatTime;

	// clear the z buffer, set the modelview, etc
	RB_BeginDrawingView();

	// draw everything
	int oldEntityNum = -1;
	backEnd.currentEntity = &tr.worldEntity;
	qbool oldDepthRange = qfalse;
	qbool depthRange = qfalse;

	backEnd.pc[RB_SURFACES] += numDrawSurfs;

	const drawSurf_t* drawSurf;
	for ( i = 0, drawSurf = drawSurfs; i < numDrawSurfs; ++i, ++drawSurf ) {
		if ( drawSurf->sort == sort ) {
			// fast path, same as previous sort
			rb_surfaceTable[ *drawSurf->surface ]( drawSurf->surface );
			continue;
		}

		int fogNum;
		const shader_t* shaderPrev = shader;
		int entityNum;
		R_DecomposeSort( drawSurf->sort, &entityNum, &shader, &fogNum );

		// "entityMergable" shaders can have surfaces from multiple refentities
		// merged into a single batch, like (CONCEPTUALLY) smoke and blood puff sprites
		// only legacy code still uses them though, because refents are so heavyweight:
		// modern code just billboards in cgame and submits raw polys, all of which are
		// ENTITYNUM_WORLD and thus automatically take the "same sort" fast path

		if ( !shader->entityMergable || ((sort ^ drawSurf->sort) & ~QSORT_ENTITYNUM_MASK) ) {
			if (shaderPrev)
				RB_EndSurface();
			RB_BeginSurface( shader, fogNum );
		}

		sort = drawSurf->sort;

		//
		// change the modelview matrix if needed
		//
		if ( entityNum != oldEntityNum ) {
			depthRange = qfalse;

			if ( entityNum != ENTITYNUM_WORLD ) {
				backEnd.currentEntity = &backEnd.refdef.entities[entityNum];
				if (backEnd.currentEntity->intShaderTime)
				    backEnd.refdef.floatTime = originalTime - (double)(backEnd.currentEntity->e.shaderTime.iShaderTime) / 1000.0;
				else
				    backEnd.refdef.floatTime = originalTime - backEnd.currentEntity->e.shaderTime.fShaderTime;
				// we have to reset the shaderTime as well otherwise image animations start
				// from the wrong frame
				tess.shaderTime = backEnd.refdef.floatTime - tess.shader->timeOffset;

				// set up the transformation matrix
				R_RotateForEntity( backEnd.currentEntity, &backEnd.viewParms, &backEnd.orient );

				if ( backEnd.currentEntity->e.renderfx & RF_DEPTHHACK ) {
					// hack the depth range to prevent view model from poking into walls
					depthRange = qtrue;
				}
			} else {
				backEnd.currentEntity = &tr.worldEntity;
				backEnd.refdef.floatTime = originalTime;
				backEnd.orient = backEnd.viewParms.world;
				// we have to reset the shaderTime as well otherwise image animations on
				// the world (like water) continue with the wrong frame
				tess.shaderTime = backEnd.refdef.floatTime - tess.shader->timeOffset;
			}

			qglLoadMatrixf( backEnd.orient.modelMatrix );

			//
			// change depthrange if needed
			//
			if ( oldDepthRange != depthRange ) {
				if ( depthRange ) {
					qglDepthRange( 0, 0.3 );
				} else {
					qglDepthRange( 0, 1 );
				}
				oldDepthRange = depthRange;
			}

			oldEntityNum = entityNum;
		}

		// add the triangles for this surface
		rb_surfaceTable[ *drawSurf->surface ]( drawSurf->surface );
	}

	backEnd.refdef.floatTime = originalTime;

	// draw the contents of the last shader batch
	if (shader) {
		RB_EndSurface();
	}

	// go back to the world modelview matrix
	qglLoadMatrixf( backEnd.viewParms.world.modelMatrix );
	if ( depthRange ) {
		qglDepthRange( 0, 1 );
	}

	// add light flares on lights that aren't obscured
	RB_RenderFlares();
}


static void RB_RenderLitSurfList( dlight_t* dl )
{
	const shader_t* shader = NULL;

	int				entityNum, oldEntityNum;
	qbool			depthRange, oldDepthRange;
	unsigned int sort = (unsigned int)-1;

	// save original time for entity shader offsets
	double originalTime = backEnd.refdef.floatTime;

	//int litsurfs = backEnd.pc[RB_LIT_SURFACES];

	// draw everything
	oldEntityNum = -1;
	backEnd.currentEntity = &tr.worldEntity;
	oldDepthRange = qfalse;
	depthRange = qfalse;

	for ( litSurf_t* litSurf = dl->head; litSurf; litSurf = litSurf->next ) {
		++backEnd.pc[RB_LIT_SURFACES];
		if ( litSurf->sort == sort ) {
			// fast path, same as previous sort
			rb_surfaceTable[ *litSurf->surface ]( litSurf->surface );
			continue;
		}

		int fogNum;
		const shader_t* shaderPrev = shader;
		R_DecomposeSort( litSurf->sort, &entityNum, &shader, &fogNum );

		// anything BEFORE opaque is sky/portal, anything AFTER it should never have been added
		//assert( shader->sort == SS_OPAQUE );
		// !!! but MIRRORS can trip that assert, so just do this for now
		if ( shader->sort < SS_OPAQUE )
			continue;

		if (shaderPrev)
			RB_EndSurface();
		RB_BeginSurface( shader, fogNum );

		sort = litSurf->sort;

		//
		// change the modelview matrix if needed
		//
		if ( entityNum != oldEntityNum ) {
			depthRange = qfalse;

			if ( entityNum != ENTITYNUM_WORLD ) {
				backEnd.currentEntity = &backEnd.refdef.entities[entityNum];
				if (backEnd.currentEntity->intShaderTime)
				    backEnd.refdef.floatTime = originalTime - (double)backEnd.currentEntity->e.shaderTime.iShaderTime;
				else
                    backEnd.refdef.floatTime = originalTime - backEnd.currentEntity->e.shaderTime.fShaderTime;
				// we have to reset the shaderTime as well otherwise image animations start
				// from the wrong frame
				tess.shaderTime = backEnd.refdef.floatTime - tess.shader->timeOffset;

				// set up the transformation matrix
				R_RotateForEntity( backEnd.currentEntity, &backEnd.viewParms, &backEnd.orient );

				R_TransformDlights( 1, dl, &backEnd.orient );
				GL2_DynLights_SetupLight();

				if ( backEnd.currentEntity->e.renderfx & RF_DEPTHHACK ) {
					// hack the depth range to prevent view model from poking into walls
					depthRange = qtrue;
				}
			} else {
				backEnd.currentEntity = &tr.worldEntity;
				backEnd.refdef.floatTime = originalTime;
				backEnd.orient = backEnd.viewParms.world;
				// we have to reset the shaderTime as well otherwise image animations on
				// the world (like water) continue with the wrong frame
				tess.shaderTime = backEnd.refdef.floatTime - tess.shader->timeOffset;
				R_TransformDlights( 1, dl, &backEnd.orient );
				GL2_DynLights_SetupLight();
			}

			qglLoadMatrixf( backEnd.orient.modelMatrix );

			//
			// change depthrange if needed
			//
			if ( oldDepthRange != depthRange ) {
				if ( depthRange ) {
					qglDepthRange( 0, 0.3 );
				} else {
					qglDepthRange( 0, 1 );
				}
				oldDepthRange = depthRange;
			}

			oldEntityNum = entityNum;
		}

		// add the triangles for this surface
		rb_surfaceTable[ *litSurf->surface ]( litSurf->surface );
	}

	backEnd.refdef.floatTime = originalTime;

	// draw the contents of the last shader batch
	if (shader) {
		RB_EndSurface();
	}

	// go back to the world modelview matrix
	qglLoadMatrixf( backEnd.viewParms.world.modelMatrix );
	if ( depthRange ) {
		qglDepthRange( 0, 1 );
	}
/*
	if ( r_speeds->integer ) {
		ri.Printf( PRINT_ALL, "light %d: %d surfaces\n", dl - backEnd.refdef.dlights, backEnd.pc[RB_LIT_SURFACES] - litsurfs );
	}
*/
}


/*
============================================================================

RENDER BACK END THREAD FUNCTIONS

============================================================================
*/


static void RB_SetGL2D()
{
	backEnd.projection2D = qtrue;
	backEnd.pc = backEnd.pc2D;

	// set 2D virtual screen size
	qglViewport( 0, 0, glConfig.vidWidth, glConfig.vidHeight );
	qglScissor( 0, 0, glConfig.vidWidth, glConfig.vidHeight );
	qglMatrixMode(GL_PROJECTION);
	qglLoadIdentity();
	qglOrtho( 0, glConfig.vidWidth, glConfig.vidHeight, 0, 0, 1 );
	qglMatrixMode(GL_MODELVIEW);
	qglLoadIdentity();

	GL_State( GLS_DEPTHTEST_DISABLE |
			  GLS_SRCBLEND_SRC_ALPHA |
			  GLS_DSTBLEND_ONE_MINUS_SRC_ALPHA );

	qglDisable( GL_CULL_FACE );
	qglDisable( GL_CLIP_PLANE0 );

	// set time for 2D shaders
	backEnd.refdef.time = ri.Milliseconds();
	backEnd.refdef.floatTime = (double)backEnd.refdef.time / 1000.0;
}


/*
=============
RE_StretchRaw

FIXME: not exactly backend
Stretches a raw 32 bit power of 2 bitmap image over the given screen rectangle.
Used for cinematics.
=============
*/
void RE_StretchRaw (int x, int y, int w, int h, int cols, int rows, const byte *data, int client, qbool dirty) {
	int			i, j;
	int			start, end;

	if ( !tr.registered ) {
		return;
	}
	R_SyncRenderThread();

	// we definately want to sync every frame for the cinematics
	qglFinish();

	start = end = 0;
	if ( r_speeds->integer ) {
		start = ri.Milliseconds();
	}

	// make sure rows and cols are powers of 2
	for ( i = 0 ; ( 1 << i ) < cols ; i++ ) {
	}
	for ( j = 0 ; ( 1 << j ) < rows ; j++ ) {
	}
	if ( ( 1 << i ) != cols || ( 1 << j ) != rows) {
		ri.Error (ERR_DROP, "Draw_StretchRaw: size not a power of 2: %i by %i", cols, rows);
	}

	GL_Bind( tr.scratchImage[client] );

	// if the scratchImage isn't in the format we want, specify it as a new texture
	if ( cols != tr.scratchImage[client]->width || rows != tr.scratchImage[client]->height ) {
		tr.scratchImage[client]->width = cols;
		tr.scratchImage[client]->height = rows;
		qglTexImage2D( GL_TEXTURE_2D, 0, GL_RGB8, cols, rows, 0, GL_RGBA, GL_UNSIGNED_BYTE, data );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP );
	} else {
		if (dirty) {
			// otherwise, just subimage upload it so that drivers can tell we are going to be changing
			// it and don't try and do a texture compression
			qglTexSubImage2D( GL_TEXTURE_2D, 0, 0, 0, cols, rows, GL_RGBA, GL_UNSIGNED_BYTE, data );
		}
	}

	if ( r_speeds->integer ) {
		end = ri.Milliseconds();
		ri.Printf( PRINT_ALL, "qglTexSubImage2D %i, %i: %i msec\n", cols, rows, end - start );
	}

	RB_SetGL2D();

	qglColor3f( tr.identityLight, tr.identityLight, tr.identityLight );

	qglBegin (GL_QUADS);
	qglTexCoord2f ( 0.5f / cols,  0.5f / rows );
	qglVertex2f (x, y);
	qglTexCoord2f ( ( cols - 0.5f ) / cols ,  0.5f / rows );
	qglVertex2f (x+w, y);
	qglTexCoord2f ( ( cols - 0.5f ) / cols, ( rows - 0.5f ) / rows );
	qglVertex2f (x+w, y+h);
	qglTexCoord2f ( 0.5f / cols, ( rows - 0.5f ) / rows );
	qglVertex2f (x, y+h);
	qglEnd ();
}

void RE_UploadCinematic (int w, int h, int cols, int rows, const byte *data, int client, qbool dirty) {

	GL_Bind( tr.scratchImage[client] );

	// if the scratchImage isn't in the format we want, specify it as a new texture
	if ( cols != tr.scratchImage[client]->width || rows != tr.scratchImage[client]->height ) {
		tr.scratchImage[client]->width = cols;
		tr.scratchImage[client]->height = rows;
		qglTexImage2D( GL_TEXTURE_2D, 0, GL_RGB8, cols, rows, 0, GL_RGBA, GL_UNSIGNED_BYTE, data );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP );
		qglTexParameterf( GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP );
	} else {
		if (dirty) {
			// otherwise, just subimage upload it so that drivers can tell we are going to be changing
			// it and don't try and do a texture compression
			qglTexSubImage2D( GL_TEXTURE_2D, 0, 0, 0, cols, rows, GL_RGBA, GL_UNSIGNED_BYTE, data );
		}
	}
}


static const void* RB_SetColor( const void* data )
{
	const setColorCommand_t* cmd = (const setColorCommand_t*)data;

	backEnd.color2D[0] = (byte)(cmd->color[0] * 255);
	backEnd.color2D[1] = (byte)(cmd->color[1] * 255);
	backEnd.color2D[2] = (byte)(cmd->color[2] * 255);
	backEnd.color2D[3] = (byte)(cmd->color[3] * 255);

	return (const void*)(cmd + 1);
}


static const void* RB_StretchPic( const void* data )
{
	const stretchPicCommand_t* cmd = (const stretchPicCommand_t*)data;

	if ( !backEnd.projection2D )
		RB_SetGL2D();

	const shader_t* shader = cmd->shader;
	if ( shader != tess.shader ) {
		if ( tess.numIndexes ) {
			RB_EndSurface();
		}
		backEnd.currentEntity = &backEnd.entity2D;
		RB_BeginSurface( shader, 0 );
	}

	RB_CHECKOVERFLOW( 4, 6 );
	int numVerts = tess.numVertexes;
	int numIndexes = tess.numIndexes;

	tess.numVertexes += 4;
	tess.numIndexes += 6;

	tess.indexes[ numIndexes ] = numVerts + 3;
	tess.indexes[ numIndexes + 1 ] = numVerts + 0;
	tess.indexes[ numIndexes + 2 ] = numVerts + 2;
	tess.indexes[ numIndexes + 3 ] = numVerts + 2;
	tess.indexes[ numIndexes + 4 ] = numVerts + 0;
	tess.indexes[ numIndexes + 5 ] = numVerts + 1;

	*(int *)tess.vertexColors[ numVerts ] =
		*(int *)tess.vertexColors[ numVerts + 1 ] =
		*(int *)tess.vertexColors[ numVerts + 2 ] =
		*(int *)tess.vertexColors[ numVerts + 3 ] = *(int *)backEnd.color2D;

	tess.xyz[ numVerts ][0] = cmd->x;
	tess.xyz[ numVerts ][1] = cmd->y;
	tess.xyz[ numVerts ][2] = 0;

	tess.texCoords[ numVerts ][0][0] = cmd->s1;
	tess.texCoords[ numVerts ][0][1] = cmd->t1;

	tess.xyz[ numVerts + 1 ][0] = cmd->x + cmd->w;
	tess.xyz[ numVerts + 1 ][1] = cmd->y;
	tess.xyz[ numVerts + 1 ][2] = 0;

	tess.texCoords[ numVerts + 1 ][0][0] = cmd->s2;
	tess.texCoords[ numVerts + 1 ][0][1] = cmd->t1;

	tess.xyz[ numVerts + 2 ][0] = cmd->x + cmd->w;
	tess.xyz[ numVerts + 2 ][1] = cmd->y + cmd->h;
	tess.xyz[ numVerts + 2 ][2] = 0;

	tess.texCoords[ numVerts + 2 ][0][0] = cmd->s2;
	tess.texCoords[ numVerts + 2 ][0][1] = cmd->t2;

	tess.xyz[ numVerts + 3 ][0] = cmd->x;
	tess.xyz[ numVerts + 3 ][1] = cmd->y + cmd->h;
	tess.xyz[ numVerts + 3 ][2] = 0;

	tess.texCoords[ numVerts + 3 ][0][0] = cmd->s1;
	tess.texCoords[ numVerts + 3 ][0][1] = cmd->t2;

	return (const void*)(cmd + 1);
}


static const void* RB_DrawSurfs( const void* data )
{
	const drawSurfsCommand_t* cmd = (const drawSurfsCommand_t*)data;

	// finish any 2D drawing if needed
	if ( tess.numIndexes )
		RB_EndSurface();

	backEnd.refdef = cmd->refdef;
	backEnd.viewParms = cmd->viewParms;

	tess.pass = shaderCommands_t::TP_BASE;

	RB_RenderDrawSurfList( cmd->drawSurfs, cmd->numDrawSurfs );

	tess.pass = shaderCommands_t::TP_LIGHT;

	if ( r_measureOverdraw->integer == 2 ) {
		// see exactly how many fragments are touched by lighting
		qglClear( GL_STENCIL_BUFFER_BIT );
		qglStencilOp( GL_KEEP, GL_KEEP, GL_INCR );
	}

	for ( int i = 0; i < backEnd.refdef.num_dlights; ++i ) {
		dlight_t* dl = &backEnd.refdef.dlights[i];
		if (!dl->head)
			continue;
		tess.light = dl;
		RB_RenderLitSurfList( dl );
	}

	tess.pass = shaderCommands_t::TP_BASE;

	return (const void*)(cmd + 1);
}


static const void* RB_BeginFrame( const void* data )
{
	const beginFrameCommand_t* cmd = (const beginFrameCommand_t*)data;

	GL2_BeginFrame();

	// clear screen for debugging
	if ( r_clear->integer ) {
		qglClearColor( 1, 0, 0.5, 1 );
		qglClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );
	}

	return (const void*)(cmd + 1);
}


/*
draw all the images to the screen, on top of whatever was there
this was used to test for texture thrashing in 1999, but is useless now

also called by RE_EndRegistration to touch them for residency
*/
void RB_ShowImages()
{
	if ( !backEnd.projection2D )
		RB_SetGL2D();

	qglClear( GL_COLOR_BUFFER_BIT );

	qglFinish();

	int start = ri.Milliseconds();

	for (int i = 0; i < tr.numImages; ++i) {
		const image_t* image = tr.images[i];

		float w = glConfig.vidWidth / 20;
		float h = glConfig.vidHeight / 15;
		float x = i % 20 * w;
		float y = i / 20 * h;

		// show in proportional size in mode 2
		if ( r_showImages->integer == 2 ) {
			w *= image->width / 512.0f;
			h *= image->height / 512.0f;
		}

		GL_Bind( image );
		qglBegin( GL_QUADS );
		qglTexCoord2f( 0, 0 );
		qglVertex2f( x, y );
		qglTexCoord2f( 1, 0 );
		qglVertex2f( x + w, y );
		qglTexCoord2f( 1, 1 );
		qglVertex2f( x + w, y + h );
		qglTexCoord2f( 0, 1 );
		qglVertex2f( x, y + h );
		qglEnd();
	}

	qglFinish();

	ri.Printf( PRINT_ALL, "%i msec to draw all images\n", ri.Milliseconds() - start );
}


static const void* RB_SwapBuffers( const void* data )
{
	// finish any 2D drawing if needed
	if ( tess.numIndexes ) {
		RB_EndSurface();
	}

	// texture swapping test
	if ( r_showImages->integer ) {
		RB_ShowImages();
	}

	const swapBuffersCommand_t* cmd = (const swapBuffersCommand_t*)data;

	// we measure overdraw by reading back the stencil buffer and
	// counting up the number of increments that have happened
	if ( r_measureOverdraw->integer ) {
		RI_AutoPtr stencilReadback( glConfig.vidWidth * glConfig.vidHeight );
		qglReadPixels( 0, 0, glConfig.vidWidth, glConfig.vidHeight, GL_STENCIL_INDEX, GL_UNSIGNED_BYTE, stencilReadback );
		/*
		long sum = 0;
		for (int i = 0; i < glConfig.vidWidth * glConfig.vidHeight; ++i)
			sum += stencilReadback[i];
		backEnd.pc.c_overDraw += sum;
		*/
		// convert the stencil data into human-comprehensible colors
		RI_AutoPtr stencilRGBA( glConfig.vidWidth * glConfig.vidHeight * 4 );
		for (int i = 0; i < glConfig.vidWidth * glConfig.vidHeight; ++i) {
			unsigned* p = stencilRGBA.Get<unsigned>() + i;
			switch ( stencilReadback[i] ) {
				default: *p = 0xFFFFFFFF; break;
				case 0: *p = LittleLong( 0xFF000000 ); break;
				case 1: *p = LittleLong( 0xFF0000FF ); break;
				case 2: *p = LittleLong( 0xFF00FF00 ); break;
				case 3: *p = LittleLong( 0xFF00FFFF ); break;
				case 4: *p = LittleLong( 0xFFFF0000 ); break;
				case 5: *p = LittleLong( 0xFFFF00FF ); break;
				case 6: *p = LittleLong( 0xFFFFFF00 ); break;
			}
		}
		RB_SetGL2D();
		qglDisable( GL_TEXTURE_2D );
		qglRasterPos3f( 0, glConfig.vidHeight, 0 );
		qglDrawPixels( glConfig.vidWidth, glConfig.vidHeight, GL_RGBA, GL_UNSIGNED_BYTE, stencilRGBA );
		qglEnable( GL_TEXTURE_2D );
	}

	if ( !backEnd.projection2D )
		RB_SetGL2D();
	GL2_EndFrame();

	if ( !glState.finishCalled ) {
		qglFinish();
	}

	GLimp_EndFrame();

	backEnd.projection2D = qfalse;
	backEnd.pc = backEnd.pc3D;

	return (const void*)(cmd + 1);
}


/*
====================
RB_ExecuteRenderCommands

This function will be called synchronously if running without
smp extensions, or asynchronously by another thread.
====================
*/
void RB_ExecuteRenderCommands( const void *data )
{
	int ms = ri.Milliseconds();

#ifdef USE_R_SMP
	if ( !r_smp->integer || data == backEndData[0]->commands.cmds ) {
		backEnd.smpFrame = 0;
	} else {
		backEnd.smpFrame = 1;
	}
#else
    backEnd.smpFrame = 0;
#endif
    
	while ( 1 ) {
		data = PADP(data, sizeof(void *));

		switch ( *(const int *)data ) {
		case RC_SET_COLOR:
			data = RB_SetColor( data );
			break;
		case RC_STRETCH_PIC:
			data = RB_StretchPic( data );
			break;
		case RC_DRAW_SURFS:
			data = RB_DrawSurfs( data );
			break;
		case RC_BEGIN_FRAME:
			data = RB_BeginFrame( data );
			break;
		case RC_SWAP_BUFFERS:
			data = RB_SwapBuffers( data );
			break;
		case RC_SCREENSHOT:
			data = RB_TakeScreenshotCmd( (const screenshotCommand_t*)data );
			break;
		case RC_VIDEOFRAME:
			data = RB_TakeVideoFrameCmd( data );
			break;

		case RC_END_OF_LIST:
		default:
			// stop rendering on this thread
			// we can't really "charge" 2D/3D properly, so it all counts as 3D
			backEnd.pc3D[RB_MSEC] = ri.Milliseconds() - ms;
			return;
		}
	}

}


/*
================
RB_RenderThread
================
*/
void RB_RenderThread( void ) {
	const void	*data;

	// wait for either a rendering command or a quit command
	while ( 1 ) {
		// sleep until we have work to do
		data = GLimp_RendererSleep();

		if ( !data ) {
			return;	// all done, renderer is shutting down
		}

		renderThreadActive = qtrue;

		RB_ExecuteRenderCommands( data );

		renderThreadActive = qfalse;
	}
}