1
0
Fork 0
forked from fte/fteqw
fteqw/engine/gl/gl_backend.c
Spoike fa9252cefa splitscreen cleaned up more.
demo menu can now leave quake dir.
scissor+line rendering abstracted from csqc.
added a rain particle effect to the 'high' particle set.
added support for parsing ezquake's koi stuff. Still only generates utf-8.
implemented some string-buffer builtins from dp that have been stubs for quite some time.
http code now supports/uses gzipped downloads properly.
added support for non-blocking tcp connects.
#pragma optimize makes more sense with the gui version now.

git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@4397 fc73d0e0-1445-4013-8a0c-d673dee63da5
2013-06-23 02:17:02 +00:00

4696 lines
131 KiB
C

#include "quakedef.h"
//#define FORCESTATE
void DumpGLState(void);
#ifdef GLQUAKE
#define r_refract_fboival gl_config.ext_framebuffer_objects
#include "glquake.h"
#include "shader.h"
#ifdef _WIN32
#include <malloc.h>
#else
#include <alloca.h>
#endif
#ifdef FORCESTATE
#pragma warningmsg("FORCESTATE is active")
#endif
#ifdef ANDROID
/*android appears to have a bug, and requires f and not i*/
#define qglTexEnvi qglTexEnvf
#endif
extern cvar_t gl_overbright;
extern cvar_t gl_ati_truform;
extern cvar_t r_wireframe;
static const char LIGHTPASS_SHADER[] = "\
{\n\
program rtlight%s\n\
\
{\n\
map $diffuse\n\
blendfunc add\n\
}\n\
{\n\
map $normalmap\n\
}\n\
{\n\
map $specular\n\
}\n\
{\n\
map $lightcubemap\n\
}\n\
{\n\
map $shadowmap\n\
}\n\
{\n\
map $loweroverlay\n\
}\n\
{\n\
map $upperoverlay\n\
}\n\
}";
enum
{
LSHADER_STANDARD,
LSHADER_CUBE,
LSHADER_SMAP,
LSHADER_SPOT,
LSHADER_MODES
};
extern cvar_t r_glsl_offsetmapping, r_noportals;
static void BE_SendPassBlendDepthMask(unsigned int sbits);
void GLBE_SubmitBatch(batch_t *batch);
static qboolean GLBE_RegisterLightShader(int mode);
struct {
//internal state
struct {
int lastpasstmus;
// int vbo_colour;
// int vbo_texcoords[SHADER_PASS_MAX];
// int vbo_deforms; //holds verticies... in case you didn't realise.
const shader_t *shader_light[1u<<LSHADER_MODES];
qboolean inited_shader_light[1u<<LSHADER_MODES];
const shader_t *crepskyshader;
const shader_t *crepopaqueshader;
GLhandleARB allblackshader;
qboolean initeddepthnorm;
const shader_t *depthnormshader;
texid_t tex_normals;
texid_t tex_diffuse;
int fbo_diffuse;
int rb_depth;
int rb_stencil;
texid_t tex_sourcecol; /*this is used by $sourcecolour tgen*/
texid_t tex_sourcedepth;
int fbo_depthless;
int fbo_reflection;
texid_t tex_reflection; /*basically a portal rendered to texture*/
texid_t tex_refraction; /*the (culled) underwater view*/
texid_t tex_refractiondepth; /*the (culled) underwater view*/
texid_t tex_ripplemap; /*temp image for waves and things.*/
qboolean force2d;
int currenttmu;
int blendmode[SHADER_TMU_MAX];
int texenvmode[SHADER_TMU_MAX];
int currenttextures[SHADER_TMU_MAX];
GLenum curtexturetype[SHADER_TMU_MAX];
polyoffset_t curpolyoffset;
unsigned int curcull;
texid_t curshadowmap;
unsigned int shaderbits;
unsigned int sha_attr;
int currentprogram;
int lastuniform; /*program which was last set, so using the same prog for multiple surfaces on the same ent (ie: world) does not require lots of extra uniform chnges*/
batch_t dummybatch;
vbo_t dummyvbo;
int colourarraytype;
int currentvbo;
int currentebo;
int currentvao;
mesh_t **meshes;
unsigned int meshcount;
float modelmatrix[16];
float modelmatrixinv[16];
float modelviewmatrix[16];
vec4_t pendingcolourflat;
int pendingcolourvbo;
void *pendingcolourpointer;
int curcolourvbo;
void *curcolourpointer;
int pendingvertexvbo;
void *pendingvertexpointer;
int curvertexvbo;
void *curvertexpointer;
int streamvbo[64];
int streamebo[64];
int streamvao[64];
int streamid;
int pendingtexcoordparts[SHADER_TMU_MAX];
int pendingtexcoordvbo[SHADER_TMU_MAX];
void *pendingtexcoordpointer[SHADER_TMU_MAX];
float identitylighting; //set to how bright lightmaps should be (reduced for overbright or realtime_world_lightmaps)
texid_t temptexture; //$current
texid_t fogtexture;
texid_t normalisationcubemap;
float fogfar;
float depthrange;
batch_t **mbatches; //model batches (ie: not world)
};
//exterior state (paramters)
struct {
backendmode_t mode;
unsigned int flags;
int oldwidth, oldheight;
vbo_t *sourcevbo;
const shader_t *curshader;
const entity_t *curentity;
const batch_t *curbatch;
const texnums_t *curtexnums;
const mfog_t *fog;
float curtime;
float updatetime;
int lightmode;
vec3_t lightorg;
vec3_t lightcolours;
vec3_t lightcolourscale;
float lightradius;
texid_t lighttexture;
texid_t lightcubemap;
float lightprojmatrix[16]; /*world space*/
vec4_t lightshadowmapinfo;
};
int wbatch;
int maxwbatches;
batch_t *wbatches;
} shaderstate;
static void BE_PolyOffset(qboolean pushdepth)
{
polyoffset_t po;
po.factor = shaderstate.curshader->polyoffset.factor;
po.unit = shaderstate.curshader->polyoffset.unit;
if (pushdepth)
{
/*some quake doors etc are flush with the walls that they're meant to be hidden behind, or plats the same height as the floor, etc
we move them back very slightly using polygonoffset to avoid really ugly z-fighting*/
extern cvar_t r_polygonoffset_submodel_offset, r_polygonoffset_submodel_factor;
po.factor += r_polygonoffset_submodel_factor.value;
po.unit += r_polygonoffset_submodel_offset.value;
}
if (shaderstate.mode == BEM_DEPTHONLY)
{
po.factor += 5;
po.unit += 25;
}
#ifndef FORCESTATE
if (shaderstate.curpolyoffset.factor != po.factor || shaderstate.curpolyoffset.unit != po.unit)
#endif
{
shaderstate.curpolyoffset = po;
if (shaderstate.curpolyoffset.factor || shaderstate.curpolyoffset.unit)
{
qglEnable(GL_POLYGON_OFFSET_FILL);
qglPolygonOffset(shaderstate.curpolyoffset.factor, shaderstate.curpolyoffset.unit);
}
else
qglDisable(GL_POLYGON_OFFSET_FILL);
}
}
void GL_TexEnv(GLenum mode)
{
#ifndef FORCESTATE
if (mode != shaderstate.texenvmode[shaderstate.currenttmu])
#endif
{
qglTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, mode);
shaderstate.texenvmode[shaderstate.currenttmu] = mode;
}
}
static void BE_SetPassBlendMode(int tmu, int pbm)
{
#ifndef FORCESTATE
if (shaderstate.blendmode[tmu] != pbm)
#endif
{
shaderstate.blendmode[tmu] = pbm;
#ifndef FORCESTATE
if (shaderstate.currenttmu != tmu)
#endif
GL_SelectTexture(tmu);
switch (pbm)
{
case PBM_DOTPRODUCT:
GL_TexEnv(GL_COMBINE_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE0_RGB_ARB, GL_TEXTURE);
qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE1_RGB_ARB, GL_PREVIOUS_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_RGB_ARB, GL_DOT3_RGB_ARB);
qglTexEnvf(GL_TEXTURE_ENV, GL_RGB_SCALE_ARB, 1);
break;
case PBM_MODULATE_PREV_COLOUR:
GL_TexEnv(GL_COMBINE_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE0_RGB_ARB, GL_PRIMARY_COLOR_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE1_RGB_ARB, GL_PREVIOUS_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_RGB_ARB, GL_MODULATE);
qglTexEnvf(GL_TEXTURE_ENV, GL_RGB_SCALE_ARB, 1);
break;
case PBM_REPLACELIGHT:
if (shaderstate.identitylighting != 1)
goto forcemod;
GL_TexEnv(GL_REPLACE);
break;
case PBM_REPLACE:
GL_TexEnv(GL_REPLACE);
break;
case PBM_DECAL:
if (tmu == 0)
goto forcemod;
GL_TexEnv(GL_DECAL);
break;
case PBM_ADD:
if (tmu == 0)
goto forcemod;
GL_TexEnv(GL_ADD);
break;
case PBM_OVERBRIGHT:
GL_TexEnv(GL_COMBINE_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE0_RGB_ARB, GL_TEXTURE);
qglTexEnvi(GL_TEXTURE_ENV, GL_SOURCE1_RGB_ARB, GL_PREVIOUS_ARB);
qglTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_RGB_ARB, GL_MODULATE);
qglTexEnvf(GL_TEXTURE_ENV, GL_RGB_SCALE_ARB, 1<<gl_overbright.ival);
break;
default:
case PBM_MODULATE:
forcemod:
GL_TexEnv(GL_MODULATE);
break;
}
}
}
/*OpenGL requires glDepthMask(GL_TRUE) or glClear(GL_DEPTH_BUFFER_BIT) will fail*/
void GL_ForceDepthWritable(void)
{
#ifndef FORCESTATE
if (!(shaderstate.shaderbits & SBITS_MISC_DEPTHWRITE))
#endif
{
shaderstate.shaderbits |= SBITS_MISC_DEPTHWRITE;
qglDepthMask(GL_TRUE);
}
}
void GL_SetShaderState2D(qboolean is2d)
{
shaderstate.updatetime = realtime;
shaderstate.force2d = is2d;
if (is2d)
memcpy(shaderstate.modelviewmatrix, r_refdef.m_view, sizeof(shaderstate.modelviewmatrix));
BE_SelectMode(BEM_STANDARD);
if (cl.paused || cls.state < ca_active)
shaderstate.updatetime = r_refdef.time;
else
shaderstate.updatetime = cl.servertime;
BE_SelectEntity(&r_worldentity);
shaderstate.curtime = shaderstate.updatetime - shaderstate.curentity->shaderTime;
}
void GL_SelectTexture(int target)
{
shaderstate.currenttmu = target;
if (qglActiveTextureARB)
qglActiveTextureARB(target + mtexid0);
else if (qglSelectTextureSGIS)
qglSelectTextureSGIS(target + mtexid0);
}
void GL_SelectVBO(int vbo)
{
#ifndef FORCESTATE
if (shaderstate.currentvbo != vbo)
#endif
{
shaderstate.currentvbo = vbo;
qglBindBufferARB(GL_ARRAY_BUFFER_ARB, shaderstate.currentvbo);
}
}
void GL_DeselectVAO(void)
{
if (shaderstate.currentvao)
{
qglBindVertexArray(0);
shaderstate.currentvao = 0;
}
}
void GL_SelectEBO(int vbo)
{
//EBO is part of the current VAO, so keep things matching that
#ifndef FORCESTATE
if (shaderstate.currentebo != vbo)
#endif
{
shaderstate.currentebo = vbo;
qglBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, shaderstate.currentebo);
}
}
void GL_MTBind(int tmu, int target, texid_t texnum)
{
GL_SelectTexture(tmu);
#ifndef FORCESTATE
if (shaderstate.currenttextures[tmu] == texnum.num)
return;
#endif
shaderstate.currenttextures[tmu] = texnum.num;
if (target)
qglBindTexture (target, texnum.num);
if (
#ifndef FORCESTATE
shaderstate.curtexturetype[tmu] != target &&
#endif
!gl_config.nofixedfunc)
{
if (shaderstate.curtexturetype[tmu])
qglDisable(shaderstate.curtexturetype[tmu]);
shaderstate.curtexturetype[tmu] = target;
if (target)
qglEnable(target);
}
}
void GL_LazyBind(int tmu, int target, texid_t texnum)
{
#ifndef FORCESTATE
if (shaderstate.currenttextures[tmu] != texnum.num)
#endif
{
GL_SelectTexture(tmu);
shaderstate.currenttextures[shaderstate.currenttmu] = texnum.num;
#ifndef FORCESTATE
if (shaderstate.curtexturetype[tmu] != target)
#endif
{
if (shaderstate.curtexturetype[tmu])
qglBindTexture (shaderstate.curtexturetype[tmu], 0);
if (gl_config.nofixedfunc)
{
shaderstate.curtexturetype[tmu] = target;
}
else
{
if (shaderstate.curtexturetype[tmu])
qglDisable(shaderstate.curtexturetype[tmu]);
shaderstate.curtexturetype[tmu] = target;
if (target)
qglEnable(target);
}
}
if (target)
qglBindTexture (target, texnum.num);
}
}
static void BE_ApplyAttributes(unsigned int bitstochange, unsigned int bitstoendisable)
{
unsigned int i;
//legacy colour attribute (including flat shaded)
if ((bitstochange) & (1u<<VATTR_LEG_COLOUR))
{
if (!shaderstate.pendingcolourpointer && !shaderstate.pendingcolourvbo)
{
if (shaderstate.curcolourpointer || shaderstate.curcolourvbo)
{
qglShadeModel(GL_FLAT);
qglDisableClientState(GL_COLOR_ARRAY);
}
shaderstate.curcolourpointer = NULL;
shaderstate.curcolourvbo = 0;
qglColor4fv(shaderstate.pendingcolourflat);
}
else
{
#ifndef FORCESTATE
if (shaderstate.curcolourpointer != shaderstate.pendingcolourpointer || shaderstate.pendingcolourvbo != shaderstate.curcolourvbo)
#endif
{
if (!shaderstate.curcolourpointer && !shaderstate.curcolourvbo)
{
if (qglShadeModel)
qglShadeModel(GL_SMOOTH);
bitstoendisable |= (1u<<VATTR_LEG_COLOUR);
}
shaderstate.curcolourpointer = shaderstate.pendingcolourpointer;
shaderstate.curcolourvbo = shaderstate.pendingcolourvbo;
GL_SelectVBO(shaderstate.curcolourvbo);
qglColorPointer(4, shaderstate.colourarraytype, 0, shaderstate.curcolourpointer);
}
if ((bitstoendisable) & (1u<<VATTR_LEG_COLOUR))
{
qglEnableClientState(GL_COLOR_ARRAY);
}
}
}
else
{
if ((bitstoendisable) & (1u<<VATTR_LEG_COLOUR))
{
qglDisableClientState(GL_COLOR_ARRAY);
}
}
//legacy tmus
if ((bitstoendisable|bitstochange) >= (1u<<VATTR_LEG_TMU0))
{
for (i = VATTR_LEG_TMU0; (bitstoendisable|bitstochange) >= (1u<<i); i++)
{
if ((bitstochange) & (1u<<i))
{
qglClientActiveTextureARB(i-VATTR_LEG_TMU0 + mtexid0);
if (bitstoendisable & (1u<<i))
qglEnableClientState(GL_TEXTURE_COORD_ARRAY);
GL_SelectVBO(shaderstate.pendingtexcoordvbo[i-VATTR_LEG_TMU0]);
qglTexCoordPointer(shaderstate.pendingtexcoordparts[i-VATTR_LEG_TMU0], GL_FLOAT, 0, shaderstate.pendingtexcoordpointer[i-VATTR_LEG_TMU0]);
}
#ifndef FORCESTATE
else if (bitstoendisable & (1u<<i))
#endif
{
qglClientActiveTextureARB(i-VATTR_LEG_TMU0 + mtexid0);
if (bitstochange & (1u<<i))
qglEnableClientState(GL_TEXTURE_COORD_ARRAY);
else
qglDisableClientState(GL_TEXTURE_COORD_ARRAY);
}
}
}
//legacy vertex coords
if ((bitstochange) & (1u<<VATTR_LEG_VERTEX))
{
#ifndef FORCESTATE
if (shaderstate.currentvao || shaderstate.curvertexpointer != shaderstate.pendingvertexpointer || shaderstate.pendingvertexvbo != shaderstate.curvertexvbo)
#endif
{
shaderstate.curvertexpointer = shaderstate.pendingvertexpointer;
shaderstate.curvertexvbo = shaderstate.pendingvertexvbo;
GL_SelectVBO(shaderstate.curvertexvbo);
qglVertexPointer(3, GL_FLOAT, VECV_STRIDE, shaderstate.curvertexpointer);
}
if ((bitstoendisable) & (1u<<VATTR_LEG_VERTEX))
{
qglEnableClientState(GL_VERTEX_ARRAY);
}
}
else
{
if ((bitstoendisable) & (1u<<VATTR_LEG_VERTEX))
{
qglDisableClientState(GL_VERTEX_ARRAY);
}
}
if (!((bitstochange|bitstoendisable) & ~((1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR))))
return;
for (i = 0; i < VATTR_LEG_FIRST; i++)
{
if ((bitstochange) & (1u<<i))
{
switch (i)
{
case VATTR_VERTEX1:
/*we still do vertex transforms for billboards and shadows and such*/
GL_SelectVBO(shaderstate.pendingvertexvbo);
qglVertexAttribPointer(i, 3, GL_FLOAT, GL_FALSE, VECV_STRIDE, shaderstate.pendingvertexpointer);
break;
case VATTR_VERTEX2:
if (!shaderstate.sourcevbo->coord2.gl.vbo && !shaderstate.sourcevbo->coord2.gl.addr)
{
GL_SelectVBO(shaderstate.pendingvertexvbo);
qglVertexAttribPointer(i, 3, GL_FLOAT, GL_FALSE, VECV_STRIDE, shaderstate.pendingvertexpointer);
}
else
{
GL_SelectVBO(shaderstate.sourcevbo->coord2.gl.vbo);
qglVertexAttribPointer(VATTR_VERTEX2, 3, GL_FLOAT, GL_FALSE, VECV_STRIDE, shaderstate.sourcevbo->coord2.gl.addr);
}
break;
case VATTR_COLOUR:
if (shaderstate.sourcevbo->colours.gl.addr)
{
GL_SelectVBO(shaderstate.sourcevbo->colours.gl.vbo);
qglVertexAttribPointer(VATTR_COLOUR, 4, shaderstate.colourarraytype, ((shaderstate.colourarraytype==GL_FLOAT)?GL_FALSE:GL_TRUE), 0, shaderstate.sourcevbo->colours.gl.addr);
break;
}
break;
case VATTR_TEXCOORD:
GL_SelectVBO(shaderstate.sourcevbo->texcoord.gl.vbo);
qglVertexAttribPointer(VATTR_TEXCOORD, 2, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->texcoord.gl.addr);
break;
case VATTR_LMCOORD:
if (!shaderstate.sourcevbo->lmcoord[0].gl.vbo && !shaderstate.sourcevbo->lmcoord[0].gl.addr)
{
GL_SelectVBO(shaderstate.sourcevbo->texcoord.gl.vbo);
qglVertexAttribPointer(VATTR_LMCOORD, 2, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->texcoord.gl.addr);
}
else
{
GL_SelectVBO(shaderstate.sourcevbo->lmcoord[0].gl.vbo);
qglVertexAttribPointer(VATTR_LMCOORD, 2, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->lmcoord[0].gl.addr);
}
break;
case VATTR_LMCOORD2:
GL_SelectVBO(shaderstate.sourcevbo->lmcoord[1].gl.vbo);
qglVertexAttribPointer(VATTR_LMCOORD2, 2, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->lmcoord[1].gl.addr);
break;
case VATTR_LMCOORD3:
GL_SelectVBO(shaderstate.sourcevbo->lmcoord[2].gl.vbo);
qglVertexAttribPointer(VATTR_LMCOORD3, 2, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->lmcoord[2].gl.addr);
break;
case VATTR_LMCOORD4:
GL_SelectVBO(shaderstate.sourcevbo->lmcoord[3].gl.vbo);
qglVertexAttribPointer(VATTR_LMCOORD4, 2, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->lmcoord[3].gl.addr);
break;
case VATTR_NORMALS:
if (!shaderstate.sourcevbo->normals.gl.addr)
{
shaderstate.sha_attr &= ~(1u<<i);
qglDisableVertexAttribArray(i);
continue;
}
GL_SelectVBO(shaderstate.sourcevbo->normals.gl.vbo);
qglVertexAttribPointer(VATTR_NORMALS, 3, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->normals.gl.addr);
break;
case VATTR_SNORMALS:
if (!shaderstate.sourcevbo->svector.gl.addr)
{
shaderstate.sha_attr &= ~(1u<<i);
qglDisableVertexAttribArray(i);
continue;
}
GL_SelectVBO(shaderstate.sourcevbo->svector.gl.vbo);
qglVertexAttribPointer(VATTR_SNORMALS, 3, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->svector.gl.addr);
break;
case VATTR_TNORMALS:
if (!shaderstate.sourcevbo->tvector.gl.addr)
{
shaderstate.sha_attr &= ~(1u<<i);
qglDisableVertexAttribArray(i);
continue;
}
GL_SelectVBO(shaderstate.sourcevbo->tvector.gl.vbo);
qglVertexAttribPointer(VATTR_TNORMALS, 3, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->tvector.gl.addr);
break;
case VATTR_BONENUMS:
GL_SelectVBO(shaderstate.sourcevbo->bonenums.gl.vbo);
qglVertexAttribPointer(VATTR_BONENUMS, 4, GL_UNSIGNED_BYTE, GL_FALSE, 0, shaderstate.sourcevbo->bonenums.gl.addr);
break;
case VATTR_BONEWEIGHTS:
GL_SelectVBO(shaderstate.sourcevbo->boneweights.gl.vbo);
qglVertexAttribPointer(VATTR_BONEWEIGHTS, 4, GL_FLOAT, GL_FALSE, 0, shaderstate.sourcevbo->boneweights.gl.addr);
break;
}
if ((bitstoendisable) & (1u<<i))
{
qglEnableVertexAttribArray(i);
}
}
else
{
if ((bitstoendisable) & (1u<<i))
{
qglDisableVertexAttribArray(i);
}
}
}
}
static void BE_EnableShaderAttributes(unsigned int progattrmask, int usevao)
{
unsigned int bitstochange, bitstoendisable;
if (shaderstate.currentvao != usevao)
{
shaderstate.currentvao = usevao;
qglBindVertexArray(usevao);
}
if (shaderstate.currentvao)
{
bitstochange = shaderstate.sourcevbo->vaodynamic&progattrmask;
#if 0
bitstoendisable = 0;
#else
bitstoendisable = shaderstate.sourcevbo->vaoenabled^progattrmask;
if (bitstoendisable)
bitstochange |= bitstoendisable;
shaderstate.sourcevbo->vaoenabled = progattrmask;
#endif
if (bitstochange & (1u<<VATTR_LEG_ELEMENTS))
{
qglBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, shaderstate.sourcevbo->indicies.gl.vbo);
}
}
else
{
bitstochange = progattrmask;
bitstoendisable = progattrmask^shaderstate.sha_attr;
shaderstate.sha_attr = progattrmask;
#ifndef FORCESTATE
if (shaderstate.currentebo != shaderstate.sourcevbo->indicies.gl.vbo)
#endif
{
shaderstate.currentebo = shaderstate.sourcevbo->indicies.gl.vbo;
qglBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, shaderstate.currentebo);
}
}
if (bitstochange || bitstoendisable)
BE_ApplyAttributes(bitstochange, bitstoendisable);
}
void GLBE_SetupVAO(vbo_t *vbo, unsigned int vaodynamic, unsigned int vaostatic)
{
if (qglGenVertexArrays)
{
qglGenVertexArrays(1, &vbo->vao);
if ((vaostatic & VATTR_VERTEX1) && !gl_config.nofixedfunc)
vaostatic = (vaostatic & ~VATTR_VERTEX1) | VATTR_LEG_VERTEX;
if ((vaodynamic & VATTR_VERTEX1) && !gl_config.nofixedfunc)
vaodynamic = (vaodynamic & ~VATTR_VERTEX1) | VATTR_LEG_VERTEX;
shaderstate.curvertexpointer = NULL;
shaderstate.curvertexvbo = 0;
shaderstate.sourcevbo = vbo;
shaderstate.pendingvertexvbo = shaderstate.sourcevbo->coord.gl.vbo;
shaderstate.pendingvertexpointer = shaderstate.sourcevbo->coord.gl.addr;
shaderstate.colourarraytype = GL_FLOAT;
shaderstate.currentvao = vbo->vao;
qglBindVertexArray(vbo->vao);
qglBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, shaderstate.sourcevbo->indicies.gl.vbo);
BE_ApplyAttributes(vaostatic, vaodynamic|vaostatic);
GL_SelectVBO(shaderstate.sourcevbo->coord.gl.vbo);
vbo->vaoenabled = vaodynamic|vaostatic;
vbo->vaodynamic = vaodynamic;
shaderstate.curvertexpointer = NULL;
shaderstate.curvertexvbo = 0;
}
else
{
GL_DeselectVAO();
/*always select the coord vbo and indicies ebo, for easy bufferdata*/
GL_SelectEBO(vbo->indicies.gl.vbo);
GL_SelectVBO(vbo->coord.gl.vbo);
}
}
void GL_SelectProgram(int program)
{
if (shaderstate.currentprogram != program)
{
qglUseProgramObjectARB(program);
shaderstate.currentprogram = program;
}
}
static void GL_DeSelectProgram(void)
{
if (shaderstate.currentprogram != 0)
{
qglUseProgramObjectARB(0);
shaderstate.currentprogram = 0;
}
}
void GLBE_RenderShadowBuffer(unsigned int numverts, int vbo, vecV_t *verts, unsigned numindicies, int ibo, index_t *indicies)
{
shaderstate.pendingvertexvbo = vbo;
shaderstate.pendingvertexpointer = verts;
shaderstate.sourcevbo = &shaderstate.dummyvbo;
shaderstate.dummyvbo.indicies.gl.vbo = ibo;
if (shaderstate.allblackshader)
{
GL_SelectProgram(shaderstate.allblackshader);
BE_EnableShaderAttributes(gl_config.nofixedfunc?(1u<<VATTR_VERTEX1):(1u<<VATTR_LEG_VERTEX), 0);
if (shaderstate.allblackshader != shaderstate.lastuniform)
{
float m16[16];
Matrix4_Multiply(r_refdef.m_projection, shaderstate.modelviewmatrix, m16);
qglUniformMatrix4fvARB(qglGetUniformLocationARB(shaderstate.allblackshader, "m_modelviewprojection"), 1, false, m16);
}
shaderstate.lastuniform = shaderstate.allblackshader;
qglDrawRangeElements(GL_TRIANGLES, 0, numverts, numindicies, GL_INDEX_TYPE, indicies);
}
else
{
GL_DeSelectProgram();
BE_EnableShaderAttributes((1u<<VATTR_LEG_VERTEX), 0);
//draw cached world shadow mesh
qglDrawRangeElements(GL_TRIANGLES, 0, numverts, numindicies, GL_INDEX_TYPE, indicies);
}
RQuantAdd(RQUANT_SHADOWFACES, numindicies);
shaderstate.dummyvbo.indicies.gl.vbo = 0;
shaderstate.sourcevbo = NULL;
}
void GL_CullFace(unsigned int sflags)
{
#ifndef FORCESTATE
if (shaderstate.curcull == sflags)
return;
#endif
shaderstate.curcull = sflags;
if (shaderstate.curcull & SHADER_CULL_FRONT)
{
qglEnable(GL_CULL_FACE);
qglCullFace(r_refdef.flipcull?GL_BACK:GL_FRONT);
}
else if (shaderstate.curcull & SHADER_CULL_BACK)
{
qglEnable(GL_CULL_FACE);
qglCullFace(r_refdef.flipcull?GL_FRONT:GL_BACK);
}
else
{
qglDisable(GL_CULL_FACE);
}
}
void R_FetchTopColour(int *retred, int *retgreen, int *retblue)
{
int i;
unsigned int cv = shaderstate.curentity->topcolour;
if (cv >= 16)
{
*retred = (((cv&0xff0000)>>16)**((unsigned char*)&d_8to24rgbtable[15]+0))>>8;
*retgreen = (((cv&0x00ff00)>>8)**((unsigned char*)&d_8to24rgbtable[15]+1))>>8;
*retblue = (((cv&0x0000ff)>>0)**((unsigned char*)&d_8to24rgbtable[15]+2))>>8;
return;
}
if (cv >= 0)
i = cv;
else
i = TOP_RANGE>>4;
if (i > 8)
{
i<<=4;
}
else
{
i<<=4;
i+=15;
}
i*=3;
*retred = host_basepal[i+0];
*retgreen = host_basepal[i+1];
*retblue = host_basepal[i+2];
/* if (!gammaworks)
{
*retred = gammatable[*retred];
*retgreen = gammatable[*retgreen];
*retblue = gammatable[*retblue];
}*/
}
void R_FetchBottomColour(int *retred, int *retgreen, int *retblue)
{
int i;
unsigned int cv = shaderstate.curentity->bottomcolour;
if (cv >= 16)
{
*retred = (((cv&0xff0000)>>16)**((unsigned char*)&d_8to24rgbtable[15]+0))>>8;
*retgreen = (((cv&0x00ff00)>>8)**((unsigned char*)&d_8to24rgbtable[15]+1))>>8;
*retblue = (((cv&0x0000ff)>>0)**((unsigned char*)&d_8to24rgbtable[15]+2))>>8;
return;
}
if (cv >= 0)
i = cv;
else
i = BOTTOM_RANGE>>4;
if (i > 8)
{
i<<=4;
}
else
{
i<<=4;
i+=15;
}
i*=3;
*retred = host_basepal[i+0];
*retgreen = host_basepal[i+1];
*retblue = host_basepal[i+2];
/* if (!gammaworks)
{
*retred = gammatable[*retred];
*retgreen = gammatable[*retgreen];
*retblue = gammatable[*retblue];
}*/
}
static void RevertToKnownState(void)
{
if (shaderstate.currentvao)
qglBindVertexArray(0);
shaderstate.currentvao = 0;
shaderstate.curvertexvbo = ~0;
GL_SelectVBO(0);
// GL_SelectEBO(0);
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
GL_SelectTexture(0);
if (!gl_config.nofixedfunc)
{
BE_SetPassBlendMode(0, PBM_REPLACE);
qglColor3f(1,1,1);
GL_DeSelectProgram();
}
shaderstate.shaderbits &= ~(SBITS_MISC_DEPTHEQUALONLY|SBITS_MISC_DEPTHCLOSERONLY|SBITS_MASK_BITS);
shaderstate.shaderbits |= SBITS_MISC_DEPTHWRITE;
shaderstate.shaderbits &= ~(SBITS_BLEND_BITS);
qglDisable(GL_BLEND);
qglDepthFunc(GL_LEQUAL);
qglDepthMask(GL_TRUE);
qglColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
}
void PPL_RevertToKnownState(void)
{
RevertToKnownState();
}
void R_IBrokeTheArrays(void)
{
RevertToKnownState();
}
#ifdef RTLIGHTS
//called from gl_shadow
void GLBE_SetupForShadowMap(texid_t shadowmaptex, int texwidth, int texheight, float shadowscale)
{
shaderstate.lightshadowmapinfo[0] = 1.0/texwidth;
shaderstate.lightshadowmapinfo[1] = 1.0/texheight;
shaderstate.lightshadowmapinfo[2] = 1.0;
shaderstate.lightshadowmapinfo[3] = shadowscale;
shaderstate.curshadowmap = shadowmaptex;
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
shaderstate.shaderbits &= ~SBITS_MISC_DEPTHWRITE;
if (qglShadeModel)
qglShadeModel(GL_FLAT);
BE_SetPassBlendMode(0, PBM_REPLACE);
GL_ForceDepthWritable();
// qglColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
BE_SelectMode(BEM_DEPTHONLY);
}
#endif
static void T_Gen_CurrentRender(int tmu)
{
int vwidth, vheight;
if (r_refdef.recurse)
return;
if (r_config.texture_non_power_of_two)
{
vwidth = vid.pixelwidth;
vheight = vid.pixelheight;
}
else
{
vwidth = 1;
vheight = 1;
while (vwidth < vid.pixelwidth)
{
vwidth *= 2;
}
while (vheight < vid.pixelheight)
{
vheight *= 2;
}
}
// copy the scene to texture
if (!TEXVALID(shaderstate.temptexture))
TEXASSIGN(shaderstate.temptexture, GL_AllocNewTexture("***$currentrender***", vwidth, vheight, 0));
GL_MTBind(tmu, GL_TEXTURE_2D, shaderstate.temptexture);
qglCopyTexImage2D(GL_TEXTURE_2D, 0, GL_RGB, 0, 0, vwidth, vheight, 0);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
}
static void Shader_BindTextureForPass(int tmu, const shaderpass_t *pass)
{
extern texid_t missing_texture;
extern texid_t missing_texture_gloss;
extern texid_t scenepp_postproc_cube;
extern texid_t r_whiteimage;
texid_t t;
switch(pass->texgen)
{
default:
case T_GEN_SINGLEMAP:
t = pass->anim_frames[0];
break;
case T_GEN_ANIMMAP:
t = pass->anim_frames[(int)(pass->anim_fps * shaderstate.curtime) % pass->anim_numframes];
break;
case T_GEN_LIGHTMAP:
if (shaderstate.curbatch->lightmap[0] < 0)
t = r_whiteimage;
else
t = lightmap[shaderstate.curbatch->lightmap[0]]->lightmap_texture;
break;
case T_GEN_DELUXMAP:
{
int lmi = shaderstate.curbatch->lightmap[0];
if (lmi < 0 || !lightmap[lmi]->hasdeluxe)
t = r_nulltex; //fixme
else
t = lightmap[lmi+1]->lightmap_texture;
}
break;
case T_GEN_DIFFUSE:
if (shaderstate.curtexnums && TEXVALID(shaderstate.curtexnums->base))
t = shaderstate.curtexnums->base;
else
t = missing_texture;
break;
case T_GEN_NORMALMAP:
t = shaderstate.curtexnums?shaderstate.curtexnums->bump:r_nulltex; /*FIXME: nulltex is not correct*/
break;
case T_GEN_SPECULAR:
if (TEXVALID(shaderstate.curtexnums->specular))
t = shaderstate.curtexnums->specular;
else
t = missing_texture_gloss;
break;
case T_GEN_UPPEROVERLAY:
t = shaderstate.curtexnums->upperoverlay;
break;
case T_GEN_LOWEROVERLAY:
t = shaderstate.curtexnums->loweroverlay;
break;
case T_GEN_FULLBRIGHT:
t = shaderstate.curtexnums->fullbright;
break;
case T_GEN_SHADOWMAP:
t = shaderstate.curshadowmap;
break;
case T_GEN_LIGHTCUBEMAP:
GL_LazyBind(tmu, GL_TEXTURE_CUBE_MAP_ARB, shaderstate.lightcubemap);
return;
case T_GEN_CUBEMAP:
t = pass->anim_frames[0];
GL_LazyBind(tmu, GL_TEXTURE_CUBE_MAP_ARB, t);
return;
case T_GEN_SOURCECUBE:
t = scenepp_postproc_cube;
GL_LazyBind(tmu, GL_TEXTURE_CUBE_MAP_ARB, t);
return;
case T_GEN_3DMAP:
t = pass->anim_frames[0];
GL_LazyBind(tmu, GL_TEXTURE_3D, t);
return;
case T_GEN_VIDEOMAP:
#ifdef NOMEDIA
t = shaderstate.curtexnums?shaderstate.curtexnums->base:r_nulltex;
#else
t = Media_UpdateForShader(pass->cin);
#endif
break;
case T_GEN_CURRENTRENDER:
T_Gen_CurrentRender(tmu);
return;
case T_GEN_SOURCECOLOUR:
t = shaderstate.tex_sourcecol;
break;
case T_GEN_SOURCEDEPTH:
t = shaderstate.tex_sourcedepth;
break;
case T_GEN_REFLECTION:
t = shaderstate.tex_reflection;
break;
case T_GEN_REFRACTION:
if (!r_refract_fboival)
{
T_Gen_CurrentRender(tmu);
return;
}
t = shaderstate.tex_refraction;
break;
case T_GEN_REFRACTIONDEPTH:
t = shaderstate.tex_refractiondepth;
break;
case T_GEN_RIPPLEMAP:
t = shaderstate.tex_ripplemap;
break;
}
GL_LazyBind(tmu, GL_TEXTURE_2D, t);
}
/*========================================== matrix functions =====================================*/
typedef vec3_t mat3_t[3];
static mat3_t axisDefault={{1, 0, 0},
{0, 1, 0},
{0, 0, 1}};
static void Matrix3_Transpose (mat3_t in, mat3_t out)
{
out[0][0] = in[0][0];
out[1][1] = in[1][1];
out[2][2] = in[2][2];
out[0][1] = in[1][0];
out[0][2] = in[2][0];
out[1][0] = in[0][1];
out[1][2] = in[2][1];
out[2][0] = in[0][2];
out[2][1] = in[1][2];
}
static void Matrix3_Multiply_Vec3 (mat3_t a, vec3_t b, vec3_t product)
{
product[0] = a[0][0]*b[0] + a[0][1]*b[1] + a[0][2]*b[2];
product[1] = a[1][0]*b[0] + a[1][1]*b[1] + a[1][2]*b[2];
product[2] = a[2][0]*b[0] + a[2][1]*b[1] + a[2][2]*b[2];
}
static int Matrix3_Compare(mat3_t in, mat3_t out)
{
return memcmp(in, out, sizeof(mat3_t));
}
//end matrix functions
/*========================================== tables for deforms =====================================*/
#define frand() (rand()*(1.0/RAND_MAX))
#define FTABLE_SIZE 1024
#define FTABLE_CLAMP(x) (((int)((x)*FTABLE_SIZE) & (FTABLE_SIZE-1)))
#define FTABLE_EVALUATE(table,x) (table ? table[FTABLE_CLAMP(x)] : frand()*((x)-floor(x)))
static float r_sintable[FTABLE_SIZE];
static float r_triangletable[FTABLE_SIZE];
static float r_squaretable[FTABLE_SIZE];
static float r_sawtoothtable[FTABLE_SIZE];
static float r_inversesawtoothtable[FTABLE_SIZE];
static float *FTableForFunc ( unsigned int func )
{
switch (func)
{
case SHADER_FUNC_SIN:
return r_sintable;
case SHADER_FUNC_TRIANGLE:
return r_triangletable;
case SHADER_FUNC_SQUARE:
return r_squaretable;
case SHADER_FUNC_SAWTOOTH:
return r_sawtoothtable;
case SHADER_FUNC_INVERSESAWTOOTH:
return r_inversesawtoothtable;
}
//bad values allow us to crash (so I can debug em)
return NULL;
}
void Shader_LightPass(char *shortname, shader_t *s, const void *args)
{
char shadertext[8192*2];
sprintf(shadertext, LIGHTPASS_SHADER, "");
Shader_DefaultScript(shortname, s, shadertext);
}
void GenerateFogTexture(texid_t *tex, float density, float zscale)
{
#define FOGS 256
#define FOGT 32
byte_vec4_t fogdata[FOGS*FOGT];
int s, t;
float f, z;
static float fogdensity, fogzscale;
if (TEXVALID(*tex) && density == fogdensity && zscale == fogzscale)
return;
fogdensity = density;
fogzscale = zscale;
for(s = 0; s < FOGS; s++)
for(t = 0; t < FOGT; t++)
{
z = (float)s / (FOGS-1);
z *= zscale;
if (0)//q3
f = pow(z, 0.5);
else if (1)//GL_EXP
f = 1-exp(-density * z);
else //GL_EXP2
f = 1-exp(-(density*density) * z);
if (f < 0)
f = 0;
if (f > 1)
f = 1;
fogdata[t*FOGS + s][0] = 255;
fogdata[t*FOGS + s][1] = 255;
fogdata[t*FOGS + s][2] = 255;
fogdata[t*FOGS + s][3] = 255*f;
}
if (!TEXVALID(*tex))
*tex = R_AllocNewTexture("***fog***", FOGS, FOGT, 0);
R_Upload(*tex, "fog", TF_RGBA32, fogdata, NULL, FOGS, FOGT, IF_CLAMP|IF_NOMIPMAP);
}
void GLBE_Shutdown(void)
{
BZ_Free(shaderstate.wbatches);
shaderstate.wbatches = NULL;
shaderstate.maxwbatches = 0;
}
void GLBE_Init(void)
{
int i;
double t;
GLBE_Shutdown();
memset(&shaderstate, 0, sizeof(shaderstate));
shaderstate.curentity = &r_worldentity;
be_maxpasses = gl_mtexarbable;
gl_stencilbits = 0;
if (gl_config.glversion >= 3.0 && gl_config.nofixedfunc)
{
//docs say this line should be okay in gl3+. nvidia do not seem to agree. GL_STENCIL_BITS is depricated however. so for now, just assume.
//qglGetFramebufferAttachmentParameteriv(GL_FRAMEBUFFER_EXT, GL_STENCIL, GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE, &gl_stencilbits);
gl_stencilbits = 8;
}
else
qglGetIntegerv(GL_STENCIL_BITS, &gl_stencilbits);
for (i = 0; i < FTABLE_SIZE; i++)
{
t = (double)i / (double)FTABLE_SIZE;
r_sintable[i] = sin(t * 2*M_PI);
if (t < 0.25)
r_triangletable[i] = t * 4.0;
else if (t < 0.75)
r_triangletable[i] = 2 - 4.0 * t;
else
r_triangletable[i] = (t - 0.75) * 4.0 - 1.0;
if (t < 0.5)
r_squaretable[i] = 1.0f;
else
r_squaretable[i] = -1.0f;
r_sawtoothtable[i] = t;
r_inversesawtoothtable[i] = 1.0 - t;
}
shaderstate.identitylighting = 1;
for (i = 0; i < MAXLIGHTMAPS; i++)
shaderstate.dummybatch.lightmap[i] = -1;
#ifdef RTLIGHTS
if (r_shadow_realtime_dlight.ival || r_shadow_realtime_world.ival)
{
if (r_shadow_shadowmapping.ival)
GLBE_RegisterLightShader(LSHADER_SMAP);
else
GLBE_RegisterLightShader(0);
}
#endif
gl_overbright.modified = true; /*in case the d3d renderer does the same*/
/*lock the cvar down if the backend can't actually do it*/
if (!gl_config.tex_env_combine && !gl_config.nofixedfunc && gl_overbright.ival)
Cvar_ApplyLatchFlag(&gl_overbright, "0", CVAR_RENDERERLATCH);
shaderstate.shaderbits = ~SBITS_ATEST_BITS;
BE_SendPassBlendDepthMask(0);
currententity = &r_worldentity;
shaderstate.fogtexture = r_nulltex;
//make sure the world draws correctly
r_worldentity.shaderRGBAf[0] = 1;
r_worldentity.shaderRGBAf[1] = 1;
r_worldentity.shaderRGBAf[2] = 1;
r_worldentity.shaderRGBAf[3] = 1;
r_worldentity.axis[0][0] = 1;
r_worldentity.axis[1][1] = 1;
r_worldentity.axis[2][2] = 1;
r_worldentity.light_avg[0] = 1;
r_worldentity.light_avg[1] = 1;
r_worldentity.light_avg[2] = 1;
R_InitFlashblends();
//only do this where we have to.
if (qglBufferDataARB && gl_config.nofixedfunc)
{
memset(&shaderstate.streamvbo, 0, sizeof(shaderstate.streamvbo));
memset(&shaderstate.streamebo, 0, sizeof(shaderstate.streamebo));
memset(&shaderstate.streamvao, 0, sizeof(shaderstate.streamvao));
qglGenBuffersARB(sizeof(shaderstate.streamvbo)/sizeof(shaderstate.streamvbo[0]), shaderstate.streamvbo);
qglGenBuffersARB(sizeof(shaderstate.streamebo)/sizeof(shaderstate.streamebo[0]), shaderstate.streamebo);
if (qglGenVertexArrays)
qglGenVertexArrays(sizeof(shaderstate.streamvao)/sizeof(shaderstate.streamvao[0]), shaderstate.streamvao);
}
}
//end tables
#define MAX_ARRAY_VERTS 65535
static avec4_t coloursarray[MAX_ARRAY_VERTS];
static float texcoordarray[SHADER_PASS_MAX][MAX_ARRAY_VERTS*2];
static vecV_t vertexarray[MAX_ARRAY_VERTS];
/*========================================== texture coord generation =====================================*/
static void tcgen_environment(float *st, unsigned int numverts, float *xyz, float *normal)
{
int i;
vec3_t viewer, reflected;
float d;
vec3_t rorg;
RotateLightVector(shaderstate.curentity->axis, shaderstate.curentity->origin, r_origin, rorg);
for (i = 0 ; i < numverts ; i++, xyz += sizeof(vecV_t)/sizeof(vec_t), normal += 3, st += 2 )
{
VectorSubtract (rorg, xyz, viewer);
VectorNormalizeFast (viewer);
d = DotProduct (normal, viewer);
reflected[0] = normal[0]*2*d - viewer[0];
reflected[1] = normal[1]*2*d - viewer[1];
reflected[2] = normal[2]*2*d - viewer[2];
st[0] = 0.5 + reflected[1] * 0.5;
st[1] = 0.5 - reflected[2] * 0.5;
}
}
static void tcgen_fog(float *st, unsigned int numverts, float *xyz, mfog_t *fog)
{
int i;
float z;
vec4_t zmat;
//generate a simple matrix to calc only the projected z coord
zmat[0] = -shaderstate.modelviewmatrix[2];
zmat[1] = -shaderstate.modelviewmatrix[6];
zmat[2] = -shaderstate.modelviewmatrix[10];
zmat[3] = -shaderstate.modelviewmatrix[14];
Vector4Scale(zmat, shaderstate.fogfar, zmat);
for (i = 0 ; i < numverts ; i++, xyz += sizeof(vecV_t)/sizeof(vec_t), st += 2 )
{
z = DotProduct(xyz, zmat) + zmat[3];
st[0] = z;
st[1] = realtime - (int)realtime;
}
}
static float *tcgen(unsigned int tcgen, int cnt, float *dst, const mesh_t *mesh)
{
int i;
vecV_t *src;
switch (tcgen)
{
default:
case TC_GEN_BASE:
return (float*)mesh->st_array;
case TC_GEN_LIGHTMAP:
if (!mesh->lmst_array)
return (float*)mesh->st_array;
else
return (float*)mesh->lmst_array;
case TC_GEN_NORMAL:
return (float*)mesh->normals_array;
case TC_GEN_SVECTOR:
return (float*)mesh->snormals_array;
case TC_GEN_TVECTOR:
return (float*)mesh->tnormals_array;
case TC_GEN_ENVIRONMENT:
if (!mesh->normals_array)
return (float*)mesh->st_array;
tcgen_environment(dst, cnt, (float*)mesh->xyz_array, (float*)mesh->normals_array);
return dst;
// case TC_GEN_DOTPRODUCT:
// return mesh->st_array[0];
case TC_GEN_VECTOR:
src = mesh->xyz_array;
for (i = 0; i < cnt; i++, dst += 2)
{
static vec3_t tc_gen_s = { 1.0f, 0.0f, 0.0f };
static vec3_t tc_gen_t = { 0.0f, 1.0f, 0.0f };
dst[0] = DotProduct(tc_gen_s, src[i]);
dst[1] = DotProduct(tc_gen_t, src[i]);
}
return dst;
}
}
/*src and dst can be the same address when tcmods are chained*/
static void tcmod(const tcmod_t *tcmod, int cnt, const float *src, float *dst, const mesh_t *mesh)
{
float *table;
float t1, t2;
float cost, sint;
int j;
#define R_FastSin(x) sin((x)*(2*M_PI))
switch (tcmod->type)
{
case SHADER_TCMOD_ROTATE:
cost = tcmod->args[0] * shaderstate.curtime;
sint = R_FastSin(cost);
cost = R_FastSin(cost + 0.25);
for (j = 0; j < cnt; j++, dst+=2,src+=2)
{
t1 = cost * (src[0] - 0.5f) - sint * (src[1] - 0.5f) + 0.5f;
t2 = cost * (src[1] - 0.5f) + sint * (src[0] - 0.5f) + 0.5f;
dst[0] = t1;
dst[1] = t2;
}
break;
case SHADER_TCMOD_SCALE:
t1 = tcmod->args[0];
t2 = tcmod->args[1];
for (j = 0; j < cnt; j++, dst+=2,src+=2)
{
dst[0] = src[0] * t1;
dst[1] = src[1] * t2;
}
break;
case SHADER_TCMOD_TURB:
t1 = tcmod->args[2] + shaderstate.curtime * tcmod->args[3];
t2 = tcmod->args[1];
for (j = 0; j < cnt; j++, dst+=2,src+=2)
{
dst[0] = src[0] + R_FastSin (src[0]*t2+t1) * t2;
dst[1] = src[1] + R_FastSin (src[1]*t2+t1) * t2;
}
break;
case SHADER_TCMOD_STRETCH:
table = FTableForFunc(tcmod->args[0]);
t2 = tcmod->args[3] + shaderstate.curtime * tcmod->args[4];
t1 = FTABLE_EVALUATE(table, t2) * tcmod->args[2] + tcmod->args[1];
t1 = t1 ? 1.0f / t1 : 1.0f;
t2 = 0.5f - 0.5f * t1;
for (j = 0; j < cnt; j++, dst+=2,src+=2)
{
dst[0] = src[0] * t1 + t2;
dst[1] = src[1] * t1 + t2;
}
break;
case SHADER_TCMOD_SCROLL:
t1 = tcmod->args[0] * shaderstate.curtime;
t2 = tcmod->args[1] * shaderstate.curtime;
for (j = 0; j < cnt; j++, dst += 2, src+=2)
{
dst[0] = src[0] + t1;
dst[1] = src[1] + t2;
}
break;
case SHADER_TCMOD_TRANSFORM:
for (j = 0; j < cnt; j++, dst+=2, src+=2)
{
t1 = src[0];
t2 = src[1];
dst[0] = t1 * tcmod->args[0] + t2 * tcmod->args[2] + tcmod->args[4];
dst[1] = t2 * tcmod->args[1] + t1 * tcmod->args[3] + tcmod->args[5];
}
break;
default:
break;
}
}
static void GenerateTCFog(int passnum, mfog_t *fog)
{
int m;
mesh_t *mesh;
for (m = 0; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
tcgen_fog(texcoordarray[passnum]+mesh->vbofirstvert*2, mesh->numvertexes, (float*)mesh->xyz_array, fog);
}
shaderstate.pendingtexcoordparts[passnum] = 2;
shaderstate.pendingtexcoordvbo[passnum] = 0;
shaderstate.pendingtexcoordpointer[passnum] = texcoordarray[passnum];
}
static void GenerateTCMods(const shaderpass_t *pass, int passnum)
{
#if 1
int i, m;
float *src;
mesh_t *mesh;
for (m = 0; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
src = tcgen(pass->tcgen, mesh->numvertexes, texcoordarray[passnum]+mesh->vbofirstvert*2, mesh);
//tcgen might return unmodified info
if (pass->numtcmods)
{
tcmod(&pass->tcmods[0], mesh->numvertexes, src, texcoordarray[passnum]+mesh->vbofirstvert*2, mesh);
for (i = 1; i < pass->numtcmods; i++)
{
tcmod(&pass->tcmods[i], mesh->numvertexes, texcoordarray[passnum]+mesh->vbofirstvert*2, texcoordarray[passnum]+mesh->vbofirstvert*2, mesh);
}
src = texcoordarray[passnum]+mesh->vbofirstvert*2;
}
else if (src != texcoordarray[passnum]+mesh->vbofirstvert*2)
{
//this shouldn't actually ever be true
memcpy(texcoordarray[passnum]+mesh->vbofirstvert*2, src, 8*mesh->numvertexes);
}
}
shaderstate.pendingtexcoordparts[passnum] = 2;
shaderstate.pendingtexcoordvbo[passnum] = 0;
shaderstate.pendingtexcoordpointer[passnum] = texcoordarray[passnum];
#else
GL_DeselectVAO();
if (!shaderstate.vbo_texcoords[passnum])
{
shaderstate.vbo_texcoords[passnum] = 0;
qglGenBuffersARB(1, &shaderstate.vbo_texcoords[passnum]);
}
GL_SelectVBO(shaderstate.vbo_texcoords[passnum]);
{
qglBufferDataARB(GL_ARRAY_BUFFER_ARB, MAX_ARRAY_VERTS*sizeof(float)*2, NULL, GL_STREAM_DRAW_ARB);
for (; meshlist; meshlist = meshlist->next)
{
int i;
float *src;
src = tcgen(pass, meshlist->numvertexes, texcoordarray[passnum], meshlist);
//tcgen might return unmodified info
if (pass->numtcmods)
{
tcmod(&pass->tcmods[0], meshlist->numvertexes, src, texcoordarray[passnum], meshlist);
for (i = 1; i < pass->numtcmods; i++)
{
tcmod(&pass->tcmods[i], meshlist->numvertexes, texcoordarray[passnum], texcoordarray[passnum], meshlist);
}
src = texcoordarray[passnum];
}
qglBufferSubDataARB(GL_ARRAY_BUFFER_ARB, meshlist->vbofirstvert*8, meshlist->numvertexes*8, src);
}
}
shaderstate.pendingtexcoordparts[passnum] = 2;
shaderstate.pendingtexcoordvbo[passnum] = shaderstate.vbo_texcoords[passnum];
shaderstate.pendingtexcoordpointer[passnum] = NULL;
#endif
}
//end texture coords
/*========================================== colour generation =====================================*/
//source is always packed
//dest is packed too
static void colourgen(const shaderpass_t *pass, int cnt, vec4_t *src, vec4_t *dst, const mesh_t *mesh)
{
switch (pass->rgbgen)
{
case RGB_GEN_ENTITY:
while((cnt)--)
{
dst[cnt][0] = shaderstate.curentity->shaderRGBAf[0];
dst[cnt][1] = shaderstate.curentity->shaderRGBAf[1];
dst[cnt][2] = shaderstate.curentity->shaderRGBAf[2];
}
break;
case RGB_GEN_ONE_MINUS_ENTITY:
while((cnt)--)
{
dst[cnt][0] = 1-shaderstate.curentity->shaderRGBAf[0];
dst[cnt][1] = 1-shaderstate.curentity->shaderRGBAf[1];
dst[cnt][2] = 1-shaderstate.curentity->shaderRGBAf[2];
}
break;
case RGB_GEN_VERTEX_LIGHTING:
if (shaderstate.identitylighting != 1)
{
if (!src)
{
while((cnt)--)
{
dst[cnt][0] = shaderstate.identitylighting;
dst[cnt][1] = shaderstate.identitylighting;
dst[cnt][2] = shaderstate.identitylighting;
}
break;
}
while((cnt)--)
{
dst[cnt][0] = src[cnt][0]*shaderstate.identitylighting;
dst[cnt][1] = src[cnt][1]*shaderstate.identitylighting;
dst[cnt][2] = src[cnt][2]*shaderstate.identitylighting;
}
break;
}
case RGB_GEN_VERTEX_EXACT:
if (!src)
{
while((cnt)--)
{
dst[cnt][0] = 1;
dst[cnt][1] = 1;
dst[cnt][2] = 1;
}
break;
}
while((cnt)--)
{
dst[cnt][0] = src[cnt][0];
dst[cnt][1] = src[cnt][1];
dst[cnt][2] = src[cnt][2];
}
break;
case RGB_GEN_ONE_MINUS_VERTEX:
while((cnt)--)
{
dst[cnt][0] = 1-src[cnt][0];
dst[cnt][1] = 1-src[cnt][1];
dst[cnt][2] = 1-src[cnt][2];
}
break;
case RGB_GEN_IDENTITY_LIGHTING:
if (shaderstate.curbatch->lightstyle[0] != 255)
{
while((cnt)--)
{
dst[cnt][0] = shaderstate.identitylighting * d_lightstylevalue[shaderstate.curbatch->lightstyle[0]]/256.0f;
dst[cnt][1] = shaderstate.identitylighting * d_lightstylevalue[shaderstate.curbatch->lightstyle[0]]/256.0f;
dst[cnt][2] = shaderstate.identitylighting * d_lightstylevalue[shaderstate.curbatch->lightstyle[0]]/256.0f;
}
}
else
{
//compensate for overbrights
while((cnt)--)
{
dst[cnt][0] = shaderstate.identitylighting;
dst[cnt][1] = shaderstate.identitylighting;
dst[cnt][2] = shaderstate.identitylighting;
}
}
break;
default:
case RGB_GEN_IDENTITY:
while((cnt)--)
{
dst[cnt][0] = 1;
dst[cnt][1] = 1;
dst[cnt][2] = 1;
}
break;
case RGB_GEN_CONST:
while((cnt)--)
{
dst[cnt][0] = pass->rgbgen_func.args[0];
dst[cnt][1] = pass->rgbgen_func.args[1];
dst[cnt][2] = pass->rgbgen_func.args[2];
}
break;
case RGB_GEN_LIGHTING_DIFFUSE:
//collect lighting details for mobile entities
if (!mesh->normals_array)
{
while((cnt)--)
{
dst[cnt][0] = 1;
dst[cnt][1] = 1;
dst[cnt][2] = 1;
}
}
else
{
R_LightArrays(shaderstate.curentity, mesh->xyz_array, dst, cnt, mesh->normals_array, shaderstate.identitylighting);
}
break;
case RGB_GEN_WAVE:
{
float *table;
float c;
table = FTableForFunc(pass->rgbgen_func.type);
c = pass->rgbgen_func.args[2] + shaderstate.curtime * pass->rgbgen_func.args[3];
c = FTABLE_EVALUATE(table, c) * pass->rgbgen_func.args[1] + pass->rgbgen_func.args[0];
c = bound(0.0f, c, 1.0f);
while((cnt)--)
{
dst[cnt][0] = c;
dst[cnt][1] = c;
dst[cnt][2] = c;
}
}
break;
case RGB_GEN_TOPCOLOR:
if (cnt)
{
int r, g, b;
R_FetchTopColour(&r, &g, &b);
dst[0][0] = r/255.0f;
dst[0][1] = g/255.0f;
dst[0][2] = b/255.0f;
while((cnt)--)
{
dst[cnt][0] = dst[0][0];
dst[cnt][1] = dst[0][1];
dst[cnt][2] = dst[0][2];
}
}
break;
case RGB_GEN_BOTTOMCOLOR:
if (cnt)
{
int r, g, b;
R_FetchBottomColour(&r, &g, &b);
dst[0][0] = r/255.0f;
dst[0][1] = g/255.0f;
dst[0][2] = b/255.0f;
while((cnt)--)
{
dst[cnt][0] = dst[0][0];
dst[cnt][1] = dst[0][1];
dst[cnt][2] = dst[0][2];
}
}
break;
}
}
static void deformgen(const deformv_t *deformv, int cnt, vecV_t *src, vecV_t *dst, const mesh_t *mesh)
{
float *table;
int j, k;
float args[4];
float deflect;
switch (deformv->type)
{
default:
case DEFORMV_NONE:
if (src != dst)
memcpy(dst, src, sizeof(*src)*cnt);
break;
case DEFORMV_WAVE:
if (!mesh->normals_array)
{
if (src != dst)
memcpy(dst, src, sizeof(*src)*cnt);
return;
}
args[0] = deformv->func.args[0];
args[1] = deformv->func.args[1];
args[3] = deformv->func.args[2] + deformv->func.args[3] * shaderstate.curtime;
table = FTableForFunc(deformv->func.type);
for ( j = 0; j < cnt; j++ )
{
deflect = deformv->args[0] * (src[j][0]+src[j][1]+src[j][2]) + args[3];
deflect = FTABLE_EVALUATE(table, deflect) * args[1] + args[0];
// Deflect vertex along its normal by wave amount
VectorMA(src[j], deflect, mesh->normals_array[j], dst[j]);
}
break;
case DEFORMV_NORMAL:
//normal does not actually move the verts, but it does change the normals array
//we don't currently support that.
if (src != dst)
memcpy(dst, src, sizeof(*src)*cnt);
/*
args[0] = deformv->args[1] * shaderstate.curtime;
for ( j = 0; j < cnt; j++ )
{
args[1] = normalsArray[j][2] * args[0];
deflect = deformv->args[0] * R_FastSin(args[1]);
normalsArray[j][0] *= deflect;
deflect = deformv->args[0] * R_FastSin(args[1] + 0.25);
normalsArray[j][1] *= deflect;
VectorNormalizeFast(normalsArray[j]);
}
*/ break;
case DEFORMV_MOVE:
table = FTableForFunc(deformv->func.type);
deflect = deformv->func.args[2] + shaderstate.curtime * deformv->func.args[3];
deflect = FTABLE_EVALUATE(table, deflect) * deformv->func.args[1] + deformv->func.args[0];
for ( j = 0; j < cnt; j++ )
VectorMA(src[j], deflect, deformv->args, dst[j]);
break;
case DEFORMV_BULGE:
args[0] = deformv->args[0]/(2*M_PI);
args[1] = deformv->args[1];
args[2] = shaderstate.curtime * deformv->args[2]/(2*M_PI);
for (j = 0; j < cnt; j++)
{
deflect = R_FastSin(mesh->st_array[j][0]*args[0] + args[2])*args[1];
dst[j][0] = src[j][0]+deflect*mesh->normals_array[j][0];
dst[j][1] = src[j][1]+deflect*mesh->normals_array[j][1];
dst[j][2] = src[j][2]+deflect*mesh->normals_array[j][2];
}
break;
case DEFORMV_AUTOSPRITE:
if (mesh->numindexes < 6)
break;
for (j = 0; j < cnt-3; j+=4, src+=4, dst+=4)
{
vec3_t mid, d;
float radius;
mid[0] = 0.25*(src[0][0] + src[1][0] + src[2][0] + src[3][0]);
mid[1] = 0.25*(src[0][1] + src[1][1] + src[2][1] + src[3][1]);
mid[2] = 0.25*(src[0][2] + src[1][2] + src[2][2] + src[3][2]);
VectorSubtract(src[0], mid, d);
radius = 2*VectorLength(d);
for (k = 0; k < 4; k++)
{
dst[k][0] = mid[0] + radius*((mesh->st_array[k][0]-0.5)*r_refdef.m_view[0+0]-(mesh->st_array[k][1]-0.5)*r_refdef.m_view[0+1]);
dst[k][1] = mid[1] + radius*((mesh->st_array[k][0]-0.5)*r_refdef.m_view[4+0]-(mesh->st_array[k][1]-0.5)*r_refdef.m_view[4+1]);
dst[k][2] = mid[2] + radius*((mesh->st_array[k][0]-0.5)*r_refdef.m_view[8+0]-(mesh->st_array[k][1]-0.5)*r_refdef.m_view[8+1]);
}
}
break;
case DEFORMV_AUTOSPRITE2:
if (mesh->numindexes < 6)
break;
for (k = 0; k < mesh->numindexes; k += 6)
{
int long_axis, short_axis;
vec3_t axis;
float len[3];
mat3_t m0, m1, m2, result;
float *quad[4];
vec3_t rot_centre, tv;
quad[0] = (float *)(dst + mesh->indexes[k+0]);
quad[1] = (float *)(dst + mesh->indexes[k+1]);
quad[2] = (float *)(dst + mesh->indexes[k+2]);
for (j = 2; j >= 0; j--)
{
quad[3] = (float *)(dst + mesh->indexes[k+3+j]);
if (!VectorEquals (quad[3], quad[0]) &&
!VectorEquals (quad[3], quad[1]) &&
!VectorEquals (quad[3], quad[2]))
{
break;
}
}
// build a matrix were the longest axis of the billboard is the Y-Axis
VectorSubtract(quad[1], quad[0], m0[0]);
VectorSubtract(quad[2], quad[0], m0[1]);
VectorSubtract(quad[2], quad[1], m0[2]);
len[0] = DotProduct(m0[0], m0[0]);
len[1] = DotProduct(m0[1], m0[1]);
len[2] = DotProduct(m0[2], m0[2]);
if ((len[2] > len[1]) && (len[2] > len[0]))
{
if (len[1] > len[0])
{
long_axis = 1;
short_axis = 0;
}
else
{
long_axis = 0;
short_axis = 1;
}
}
else if ((len[1] > len[2]) && (len[1] > len[0]))
{
if (len[2] > len[0])
{
long_axis = 2;
short_axis = 0;
}
else
{
long_axis = 0;
short_axis = 2;
}
}
else //if ( (len[0] > len[1]) && (len[0] > len[2]) )
{
if (len[2] > len[1])
{
long_axis = 2;
short_axis = 1;
}
else
{
long_axis = 1;
short_axis = 2;
}
}
if (DotProduct(m0[long_axis], m0[short_axis]))
{
VectorNormalize2(m0[long_axis], axis);
VectorCopy(axis, m0[1]);
if (axis[0] || axis[1])
{
VectorVectors(m0[1], m0[2], m0[0]);
}
else
{
VectorVectors(m0[1], m0[0], m0[2]);
}
}
else
{
VectorNormalize2(m0[long_axis], axis);
VectorNormalize2(m0[short_axis], m0[0]);
VectorCopy(axis, m0[1]);
CrossProduct(m0[0], m0[1], m0[2]);
}
for (j = 0; j < 3; j++)
rot_centre[j] = (quad[0][j] + quad[1][j] + quad[2][j] + quad[3][j]) * 0.25;
if (shaderstate.curentity)
{
VectorAdd(shaderstate.curentity->origin, rot_centre, tv);
}
else
{
VectorCopy(rot_centre, tv);
}
VectorSubtract(r_origin, tv, tv);
// filter any longest-axis-parts off the camera-direction
deflect = -DotProduct(tv, axis);
VectorMA(tv, deflect, axis, m1[2]);
VectorNormalizeFast(m1[2]);
VectorCopy(axis, m1[1]);
CrossProduct(m1[1], m1[2], m1[0]);
Matrix3_Transpose(m1, m2);
Matrix3_Multiply(m2, m0, result);
for (j = 0; j < 4; j++)
{
VectorSubtract(quad[j], rot_centre, tv);
Matrix3_Multiply_Vec3(result, tv, quad[j]);
VectorAdd(rot_centre, quad[j], quad[j]);
}
}
break;
// case DEFORMV_PROJECTION_SHADOW:
// break;
}
}
static void GenerateVertexBlends(const shader_t *shader)
{
int i, m;
mesh_t *meshlist;
vecV_t *ov, *iv1, *iv2;
float w1, w2;
for (m = 0; m < shaderstate.meshcount; m++)
{
meshlist = shaderstate.meshes[m];
ov = vertexarray+meshlist->vbofirstvert;
iv1 = meshlist->xyz_array;
iv2 = meshlist->xyz2_array;
w1 = meshlist->xyz_blendw[0];
w2 = meshlist->xyz_blendw[1];
for (i = 0; i < meshlist->numvertexes; i++)
{
ov[i][0] = iv1[i][0]*w1 + iv2[i][0]*w2;
ov[i][1] = iv1[i][1]*w1 + iv2[i][1]*w2;
ov[i][2] = iv1[i][2]*w1 + iv2[i][2]*w2;
}
for (i = 0; i < shader->numdeforms; i++)
{
deformgen(&shader->deforms[i], meshlist->numvertexes, vertexarray+meshlist->vbofirstvert, vertexarray+meshlist->vbofirstvert, meshlist);
}
}
shaderstate.pendingvertexpointer = vertexarray;
shaderstate.pendingvertexvbo = 0;
}
static void GenerateVertexDeforms(const shader_t *shader)
{
int i, m;
mesh_t *meshlist;
for (m = 0; m < shaderstate.meshcount; m++)
{
meshlist = shaderstate.meshes[m];
deformgen(&shader->deforms[0], meshlist->numvertexes, meshlist->xyz_array, vertexarray+meshlist->vbofirstvert, meshlist);
for (i = 1; i < shader->numdeforms; i++)
{
deformgen(&shader->deforms[i], meshlist->numvertexes, vertexarray+meshlist->vbofirstvert, vertexarray+meshlist->vbofirstvert, meshlist);
}
}
shaderstate.pendingvertexpointer = vertexarray;
shaderstate.pendingvertexvbo = 0;
}
/*======================================alpha ===============================*/
static void alphagen(const shaderpass_t *pass, int cnt, avec4_t *const src, avec4_t *dst, const mesh_t *mesh)
{
float *table;
float t;
float f;
vec3_t v1, v2;
int i;
switch (pass->alphagen)
{
default:
case ALPHA_GEN_IDENTITY:
if (shaderstate.flags & BEF_FORCETRANSPARENT)
{
while(cnt--)
dst[cnt][3] = shaderstate.curentity->shaderRGBAf[3];
}
else
{
while(cnt--)
dst[cnt][3] = 1;
}
break;
case ALPHA_GEN_CONST:
t = pass->alphagen_func.args[0];
while(cnt--)
dst[cnt][3] = t;
break;
case ALPHA_GEN_WAVE:
table = FTableForFunc(pass->alphagen_func.type);
f = pass->alphagen_func.args[2] + shaderstate.curtime * pass->alphagen_func.args[3];
f = FTABLE_EVALUATE(table, f) * pass->alphagen_func.args[1] + pass->alphagen_func.args[0];
t = bound(0.0f, f, 1.0f);
while(cnt--)
dst[cnt][3] = t;
break;
case ALPHA_GEN_PORTAL:
//FIXME: should this be per-vert?
if (r_refdef.recurse)
f = 1;
else
{
VectorAdd(mesh->xyz_array[0], shaderstate.curentity->origin, v1);
VectorSubtract(r_origin, v1, v2);
f = VectorLength(v2) * (1.0 / shaderstate.curshader->portaldist);
f = bound(0.0f, f, 1.0f);
}
while(cnt--)
dst[cnt][3] = f;
break;
case ALPHA_GEN_VERTEX:
if (!src)
{
while(cnt--)
{
dst[cnt][3] = 1;
}
break;
}
while(cnt--)
{
dst[cnt][3] = src[cnt][3];
}
break;
case ALPHA_GEN_ENTITY:
f = bound(0, shaderstate.curentity->shaderRGBAf[3], 1);
while(cnt--)
{
dst[cnt][3] = f;
}
break;
case ALPHA_GEN_SPECULAR:
{
mat3_t axis;
AngleVectors(shaderstate.curentity->angles, axis[0], axis[1], axis[2]);
VectorSubtract(r_origin, shaderstate.curentity->origin, v1);
if (!Matrix3_Compare(axis, axisDefault))
{
Matrix3_Multiply_Vec3(axis, v1, v2);
}
else
{
VectorCopy(v1, v2);
}
for (i = 0; i < cnt; i++)
{
VectorSubtract(v2, mesh->xyz_array[i], v1);
f = DotProduct(v1, mesh->normals_array[i] ) * Q_rsqrt(DotProduct(v1,v1));
f = f * f * f * f * f;
dst[i][3] = bound (0.0f, f, 1.0f);
}
}
break;
}
}
static void GenerateColourMods(const shaderpass_t *pass)
{
unsigned int m;
mesh_t *meshlist;
meshlist = shaderstate.meshes[0];
if (pass->flags & SHADER_PASS_NOCOLORARRAY && qglColor4fv)
{
colourgen(pass, 1, meshlist->colors4f_array, &shaderstate.pendingcolourflat, meshlist);
alphagen(pass, 1, meshlist->colors4f_array, &shaderstate.pendingcolourflat, meshlist);
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
}
else
{
extern cvar_t r_nolightdir;
if (pass->rgbgen == RGB_GEN_LIGHTING_DIFFUSE)
{
if (shaderstate.mode == BEM_DEPTHDARK || shaderstate.mode == BEM_DEPTHONLY)
{
shaderstate.pendingcolourflat[0] = shaderstate.pendingcolourflat[1] = shaderstate.pendingcolourflat[2] = 0;
alphagen(pass, 1, meshlist->colors4f_array, &shaderstate.pendingcolourflat, meshlist);
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
return;
}
if (shaderstate.mode == BEM_LIGHT)
{
shaderstate.pendingcolourflat[0] = shaderstate.pendingcolourflat[1] = shaderstate.pendingcolourflat[2] = 1;
alphagen(pass, 1, meshlist->colors4f_array, &shaderstate.pendingcolourflat, meshlist);
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
return;
}
if (r_nolightdir.ival)
{
VectorCopy(shaderstate.curentity->light_avg, shaderstate.pendingcolourflat);
shaderstate.pendingcolourflat[3] = shaderstate.curentity->shaderRGBAf[3];
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
return;
}
}
//if its vetex lighting, just use the vbo
if (((pass->rgbgen == RGB_GEN_VERTEX_LIGHTING && shaderstate.identitylighting == 1) || pass->rgbgen == RGB_GEN_VERTEX_EXACT) && pass->alphagen == ALPHA_GEN_VERTEX)
{
shaderstate.pendingcolourvbo = shaderstate.sourcevbo->colours.gl.vbo;
shaderstate.pendingcolourpointer = shaderstate.sourcevbo->colours.gl.addr;
return;
}
for (m = 0; m < shaderstate.meshcount; m++)
{
meshlist = shaderstate.meshes[m];
colourgen(pass, meshlist->numvertexes, meshlist->colors4f_array, coloursarray + meshlist->vbofirstvert, meshlist);
alphagen(pass, meshlist->numvertexes, meshlist->colors4f_array, coloursarray + meshlist->vbofirstvert, meshlist);
}
shaderstate.colourarraytype = GL_FLOAT;
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = coloursarray;
}
}
static void BE_GeneratePassTC(const shaderpass_t *pass, int tmu)
{
if (!pass->numtcmods)
{
//if there are no tcmods, pass through here as fast as possible
if (pass->tcgen == TC_GEN_BASE)
{
shaderstate.pendingtexcoordparts[tmu] = 2;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->texcoord.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->texcoord.gl.addr;
}
else if (pass->tcgen == TC_GEN_LIGHTMAP)
{
if (!shaderstate.sourcevbo->lmcoord[0].gl.addr)
{
shaderstate.pendingtexcoordparts[tmu] = 2;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->texcoord.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->texcoord.gl.addr;
}
else
{
shaderstate.pendingtexcoordparts[tmu] = 2;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->lmcoord[0].gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->lmcoord[0].gl.addr;
}
}
else if (pass->tcgen == TC_GEN_NORMAL)
{
shaderstate.pendingtexcoordparts[tmu] = 3;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->normals.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->normals.gl.addr;
}
else if (pass->tcgen == TC_GEN_SVECTOR)
{
shaderstate.pendingtexcoordparts[tmu] = 3;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->svector.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->svector.gl.addr;
}
else if (pass->tcgen == TC_GEN_TVECTOR)
{
shaderstate.pendingtexcoordparts[tmu] = 3;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->tvector.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->tvector.gl.addr;
}
else
{
//specular highlights and reflections have no fixed data, and must be generated.
GenerateTCMods(pass, tmu);
}
}
else
{
GenerateTCMods(pass, tmu);
}
}
static void BE_SendPassBlendDepthMask(unsigned int sbits)
{
unsigned int delta;
/*2d mode doesn't depth test or depth write*/
if (shaderstate.force2d)
{
#ifdef warningmsg
#pragma warningmsg("fixme: q3 doesn't seem to have this, why do we need it?")
#endif
sbits &= ~(SBITS_MISC_DEPTHWRITE|SBITS_MISC_DEPTHEQUALONLY);
sbits |= SBITS_MISC_NODEPTHTEST;
}
if (shaderstate.flags & (BEF_FORCEADDITIVE|BEF_FORCETRANSPARENT|BEF_FORCENODEPTH|BEF_FORCEDEPTHTEST|BEF_FORCEDEPTHWRITE))
{
if (shaderstate.flags & BEF_FORCEADDITIVE)
sbits = (sbits & ~(SBITS_MISC_DEPTHWRITE|SBITS_BLEND_BITS|SBITS_ATEST_BITS))
| (SBITS_SRCBLEND_SRC_ALPHA | SBITS_DSTBLEND_ONE);
else if (shaderstate.flags & BEF_FORCETRANSPARENT)
{
if ((sbits & SBITS_BLEND_BITS) == (SBITS_SRCBLEND_ONE| SBITS_DSTBLEND_ZERO) || !(sbits & SBITS_BLEND_BITS)) /*if transparency is forced, clear alpha test bits*/
sbits = (sbits & ~(SBITS_BLEND_BITS|SBITS_ATEST_BITS))
| (SBITS_SRCBLEND_SRC_ALPHA | SBITS_DSTBLEND_ONE_MINUS_SRC_ALPHA | SBITS_ATEST_GT0);
}
if (shaderstate.flags & BEF_FORCENODEPTH) /*EF_NODEPTHTEST dp extension*/
sbits |= SBITS_MISC_NODEPTHTEST;
else
{
if (shaderstate.flags & BEF_FORCEDEPTHTEST)
sbits &= ~SBITS_MISC_NODEPTHTEST;
if (shaderstate.flags & BEF_FORCEDEPTHWRITE)
sbits |= SBITS_MISC_DEPTHWRITE;
}
}
delta = sbits^shaderstate.shaderbits;
#ifdef FORCESTATE
delta |= ~0;
#endif
if (!delta)
return;
shaderstate.shaderbits = sbits;
if (delta & SBITS_BLEND_BITS)
{
if (sbits & SBITS_BLEND_BITS)
{
int src, dst;
/*unpack the src and dst factors*/
switch(sbits & SBITS_SRCBLEND_BITS)
{
case SBITS_SRCBLEND_ZERO: src = GL_ZERO; break;
default:
case SBITS_SRCBLEND_ONE: src = GL_ONE; break;
case SBITS_SRCBLEND_DST_COLOR: src = GL_DST_COLOR; break;
case SBITS_SRCBLEND_ONE_MINUS_DST_COLOR: src = GL_ONE_MINUS_DST_COLOR; break;
case SBITS_SRCBLEND_SRC_ALPHA: src = GL_SRC_ALPHA; break;
case SBITS_SRCBLEND_ONE_MINUS_SRC_ALPHA: src = GL_ONE_MINUS_SRC_ALPHA; break;
case SBITS_SRCBLEND_DST_ALPHA: src = GL_DST_ALPHA; break;
case SBITS_SRCBLEND_ONE_MINUS_DST_ALPHA: src = GL_ONE_MINUS_DST_ALPHA; break;
case SBITS_SRCBLEND_ALPHA_SATURATE: src = GL_SRC_ALPHA_SATURATE; break;
}
switch((sbits & SBITS_DSTBLEND_BITS)>>4)
{
case SBITS_DSTBLEND_ZERO>>4: dst = GL_ZERO; break;
default:
case SBITS_DSTBLEND_ONE>>4: dst = GL_ONE; break;
case SBITS_DSTBLEND_SRC_COLOR>>4: dst = GL_SRC_COLOR; break;
case SBITS_DSTBLEND_ONE_MINUS_SRC_COLOR>>4: dst = GL_ONE_MINUS_SRC_COLOR; break;
case SBITS_DSTBLEND_SRC_ALPHA>>4: dst = GL_SRC_ALPHA; break;
case SBITS_DSTBLEND_ONE_MINUS_SRC_ALPHA>>4: dst = GL_ONE_MINUS_SRC_ALPHA; break;
case SBITS_DSTBLEND_DST_ALPHA>>4: dst = GL_DST_ALPHA; break;
case SBITS_DSTBLEND_ONE_MINUS_DST_ALPHA>>4: dst = GL_ONE_MINUS_DST_ALPHA; break;
}
qglEnable(GL_BLEND);
qglBlendFunc(src, dst);
}
else
qglDisable(GL_BLEND);
}
#ifdef GL_ALPHA_TEST //alpha test doesn't exist in gles2
if ((delta & SBITS_ATEST_BITS) && !gl_config.nofixedfunc)
{
switch (sbits & SBITS_ATEST_BITS)
{
default:
qglDisable(GL_ALPHA_TEST);
break;
case SBITS_ATEST_GT0:
qglEnable(GL_ALPHA_TEST);
qglAlphaFunc(GL_GREATER, 0);
break;
case SBITS_ATEST_LT128:
qglEnable(GL_ALPHA_TEST);
qglAlphaFunc(GL_LESS, 0.5f);
break;
case SBITS_ATEST_GE128:
qglEnable(GL_ALPHA_TEST);
qglAlphaFunc(GL_GEQUAL, 0.5f);
break;
}
}
#endif
if (delta & SBITS_MISC_NODEPTHTEST)
{
if (sbits & SBITS_MISC_NODEPTHTEST)
qglDisable(GL_DEPTH_TEST);
else
qglEnable(GL_DEPTH_TEST);
}
if (delta & SBITS_MISC_DEPTHWRITE)
{
if (sbits & SBITS_MISC_DEPTHWRITE)
qglDepthMask(GL_TRUE);
else
qglDepthMask(GL_FALSE);
}
if (delta & (SBITS_MISC_DEPTHEQUALONLY|SBITS_MISC_DEPTHCLOSERONLY))
{
extern int gldepthfunc;
switch (sbits & (SBITS_MISC_DEPTHEQUALONLY|SBITS_MISC_DEPTHCLOSERONLY))
{
case SBITS_MISC_DEPTHEQUALONLY:
qglDepthFunc(GL_EQUAL);
break;
case SBITS_MISC_DEPTHCLOSERONLY:
if (gldepthfunc == GL_LEQUAL)
qglDepthFunc(GL_LESS);
else
qglDepthFunc(GL_GREATER);
break;
default:
qglDepthFunc(gldepthfunc);
break;
}
}
if (delta & (SBITS_MASK_BITS))
{
qglColorMask(
(sbits&SBITS_MASK_RED)?GL_FALSE:GL_TRUE,
(sbits&SBITS_MASK_GREEN)?GL_FALSE:GL_TRUE,
(sbits&SBITS_MASK_BLUE)?GL_FALSE:GL_TRUE,
(sbits&SBITS_MASK_ALPHA)?GL_FALSE:GL_TRUE
);
}
if ((delta & SBITS_TRUFORM) && qglPNTrianglesiATI)
{
if ((sbits & SBITS_TRUFORM) && gl_ati_truform.ival)
qglEnable(GL_PN_TRIANGLES_ATI);
else
qglDisable(GL_PN_TRIANGLES_ATI);
}
}
static void BE_SubmitMeshChain(void)
{
int startv, starti, endv, endi;
int m;
mesh_t *mesh;
int batchtype = GL_TRIANGLES;
if (shaderstate.flags & BEF_LINES)
batchtype = GL_LINES;
#if 0
if (!shaderstate.currentebo)
{
if (shaderstate.meshcount == 1)
{
mesh = shaderstate.meshes[0];
qglDrawRangeElements(batchtype, mesh->vbofirstvert, mesh->vbofirstvert+mesh->numvertexes, mesh->numindexes, GL_INDEX_TYPE, shaderstate.sourcevbo->indicies + mesh->vbofirstelement);
RQuantAdd(RQUANT_DRAWS, 1);
return;
}
else
{
index_t *ilst;
mesh = shaderstate.meshes[0];
startv = mesh->vbofirstvert;
endv = startv + mesh->numvertexes;
endi = mesh->numindexes;
for (m = 1; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
endi += mesh->numindexes;
if (startv > mesh->vbofirstvert)
startv = mesh->vbofirstvert;
if (endv < mesh->vbofirstvert+mesh->numvertexes)
endv = mesh->vbofirstvert+mesh->numvertexes;
}
ilst = alloca(endi*sizeof(index_t));
endi = 0;
for (m = 0; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
for (starti = 0; starti < mesh->numindexes; )
ilst[endi++] = mesh->vbofirstvert + mesh->indexes[starti++];
}
qglDrawRangeElements(batchtype, startv, endv, endi, GL_INDEX_TYPE, ilst);
RQuantAdd(RQUANT_DRAWS, 1);
}
return;
}
#endif
/*
if (qglLockArraysEXT)
{
endv = 0;
startv = 0x7fffffff;
for (m = 0; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
starti = mesh->vbofirstvert;
if (starti < startv)
startv = starti;
endi = mesh->vbofirstvert+mesh->numvertexes;
if (endi > endv)
endv = endi;
}
qglLockArraysEXT(startv, endv);
}
*/
for (m = 0, mesh = shaderstate.meshes[0]; m < shaderstate.meshcount; )
{
startv = mesh->vbofirstvert;
starti = mesh->vbofirstelement;
endv = startv+mesh->numvertexes;
endi = starti+mesh->numindexes;
//find consecutive surfaces
for (++m; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
if (endi == mesh->vbofirstelement)
{
endv = mesh->vbofirstvert+mesh->numvertexes;
endi = mesh->vbofirstelement+mesh->numindexes;
}
else
{
break;
}
}
qglDrawRangeElements(batchtype, startv, endv, endi-starti, GL_INDEX_TYPE, (index_t*)shaderstate.sourcevbo->indicies.gl.addr + starti);
RQuantAdd(RQUANT_DRAWS, 1);
}
/*
if (qglUnlockArraysEXT)
qglUnlockArraysEXT();
*/
}
static void DrawPass(const shaderpass_t *pass)
{
int i, j, k;
int tmu;
int lastpass = pass->numMergedPasses;
unsigned int attr = (1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR);
for (i = 0; i < lastpass; i++)
{
if (pass[i].texgen == T_GEN_UPPEROVERLAY && !TEXVALID(shaderstate.curtexnums->upperoverlay))
continue;
if (pass[i].texgen == T_GEN_LOWEROVERLAY && !TEXVALID(shaderstate.curtexnums->loweroverlay))
continue;
if (pass[i].texgen == T_GEN_FULLBRIGHT && !TEXVALID(shaderstate.curtexnums->fullbright))
continue;
break;
}
if (i == lastpass)
return;
BE_SendPassBlendDepthMask(pass[i].shaderbits);
GenerateColourMods(pass+i);
tmu = 0;
for (; i < lastpass; i++)
{
if (pass[i].texgen == T_GEN_UPPEROVERLAY && !TEXVALID(shaderstate.curtexnums->upperoverlay))
continue;
if (pass[i].texgen == T_GEN_LOWEROVERLAY && !TEXVALID(shaderstate.curtexnums->loweroverlay))
continue;
if (pass[i].texgen == T_GEN_FULLBRIGHT && !TEXVALID(shaderstate.curtexnums->fullbright))
continue;
Shader_BindTextureForPass(tmu, pass+i);
attr |= (1u<<(VATTR_LEG_TMU0+tmu));
BE_GeneratePassTC(pass+i, tmu);
BE_SetPassBlendMode(tmu, pass[i].blendmode);
tmu++;
//add in
if (pass[i].texgen == T_GEN_LIGHTMAP)
{
//first pass should have been REPLACE
//second pass should be an ADD
//this depends upon rgbgens for light levels, so each pass *must* be pushed to hardware individually
for (j = 1; j < MAXLIGHTMAPS && shaderstate.curbatch->lightmap[j] >= 0; j++)
{
if (j == 1)
BE_SetPassBlendMode(tmu, PBM_REPLACE);
/*make sure no textures linger*/
for (k = tmu; k < shaderstate.lastpasstmus; k++)
{
GL_LazyBind(k, 0, r_nulltex);
}
shaderstate.lastpasstmus = tmu;
/*push it*/
BE_EnableShaderAttributes(attr, 0);
BE_SubmitMeshChain();
tmu = 0;
/*bind the light texture*/
GL_LazyBind(tmu, GL_TEXTURE_2D, lightmap[shaderstate.curbatch->lightmap[j]]->lightmap_texture);
/*set up the colourmod for this style's lighting*/
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
shaderstate.pendingcolourflat[0] = shaderstate.identitylighting * d_lightstylevalue[shaderstate.curbatch->lightstyle[j]]/256.0f;
shaderstate.pendingcolourflat[1] = shaderstate.identitylighting * d_lightstylevalue[shaderstate.curbatch->lightstyle[j]]/256.0f;
shaderstate.pendingcolourflat[2] = shaderstate.identitylighting * d_lightstylevalue[shaderstate.curbatch->lightstyle[j]]/256.0f;
shaderstate.pendingcolourflat[3] = 1;
/*pick the correct st coords for this lightmap pass*/
shaderstate.pendingtexcoordparts[tmu] = 2;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->lmcoord[j].gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->lmcoord[j].gl.addr;
BE_SetPassBlendMode(tmu, PBM_ADD);
BE_SendPassBlendDepthMask((pass[0].shaderbits & ~SBITS_BLEND_BITS) | SBITS_SRCBLEND_ONE | SBITS_DSTBLEND_ONE);
attr = (1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR);
attr |= (1u<<(VATTR_LEG_TMU0+tmu));
tmu++;
}
//might need to break the pass here
if (j > 1 && i != lastpass)
{
for (k = tmu; k < shaderstate.lastpasstmus; k++)
{
GL_LazyBind(k, 0, r_nulltex);
}
shaderstate.lastpasstmus = tmu;
BE_EnableShaderAttributes(attr, 0);
BE_SubmitMeshChain();
tmu = 0;
BE_SendPassBlendDepthMask(pass[i+1].shaderbits);
GenerateColourMods(&pass[i+1]);
}
}
}
if (!tmu)
return;
for (i = tmu; i < shaderstate.lastpasstmus; i++)
{
GL_LazyBind(i, 0, r_nulltex);
}
shaderstate.lastpasstmus = tmu;
BE_EnableShaderAttributes(attr, 0);
BE_SubmitMeshChain();
}
static void BE_Program_Set_Attributes(const program_t *prog, unsigned int perm, qboolean entunchanged)
{
vec4_t param4;
int r, g, b;
int i;
unsigned int ph;
const shaderprogparm_t *p;
/*don't bother setting it if the ent properties are unchanged (but do if the mesh changed)*/
if (entunchanged)
return;
for (i = 0; i < prog->numparams; i++)
{
p = &prog->parm[i];
ph = prog->permu[perm].parm[i];
if (ph == -1)
continue; /*not in this permutation*/
switch(p->type)
{
case SP_M_VIEW:
qglUniformMatrix4fvARB(ph, 1, false, r_refdef.m_view);
break;
case SP_M_PROJECTION:
qglUniformMatrix4fvARB(ph, 1, false, r_refdef.m_projection);
break;
case SP_M_MODELVIEW:
qglUniformMatrix4fvARB(ph, 1, false, shaderstate.modelviewmatrix);
break;
case SP_M_MODELVIEWPROJECTION:
{
float m16[16];
Matrix4_Multiply(r_refdef.m_projection, shaderstate.modelviewmatrix, m16);
qglUniformMatrix4fvARB(ph, 1, false, m16);
}
break;
case SP_M_INVMODELVIEWPROJECTION:
{
float m16[16], inv[16];
Matrix4_Multiply(r_refdef.m_projection, shaderstate.modelviewmatrix, m16);
Matrix4_Invert(m16, inv);
qglUniformMatrix4fvARB(ph, 1, false, inv);
}
break;
case SP_M_MODEL:
qglUniformMatrix4fvARB(ph, 1, false, shaderstate.modelmatrix);
break;
case SP_M_ENTBONES:
{
qglUniformMatrix3x4fv(ph, shaderstate.sourcevbo->numbones, false, shaderstate.sourcevbo->bones);
}
break;
case SP_M_INVVIEWPROJECTION:
{
float m16[16], inv[16];
Matrix4_Multiply(r_refdef.m_projection, r_refdef.m_view, m16);
Matrix4_Invert(m16, inv);
qglUniformMatrix4fvARB(ph, 1, false, inv);
}
break;
case SP_E_VBLEND:
qglUniform2fvARB(ph, 1, shaderstate.meshes[0]->xyz_blendw);
break;
case SP_E_LMSCALE:
if (perm & PERMUTATION_LIGHTSTYLES)
{
vec4_t colscale[MAXLIGHTMAPS];
int j, s;
for (j = 0; j < MAXLIGHTMAPS ; j++)
{
s = shaderstate.curbatch->lightstyle[j];
if (s == 255)
{
for (; j < MAXLIGHTMAPS ; j++)
{
colscale[j][0] = 0;
colscale[j][1] = 0;
colscale[j][2] = 0;
colscale[j][3] = 1;
}
break;
}
if (shaderstate.curentity->model && shaderstate.curentity->model->engineflags & MDLF_NEEDOVERBRIGHT)
{
float sc = (1<<bound(0, gl_overbright.ival, 2)) * shaderstate.identitylighting;
VectorSet(colscale[j], sc, sc, sc);
}
else
{
VectorSet(colscale[j], shaderstate.identitylighting, shaderstate.identitylighting, shaderstate.identitylighting);
}
colscale[j][3] = 1;
VectorScale(colscale[j], d_lightstylevalue[s]/256.0f, colscale[j]);
}
qglUniform4fvARB(ph, j, (GLfloat*)colscale);
shaderstate.lastuniform = 0;
}
else
{
vec4_t colscale[4];
if (shaderstate.curentity->model && shaderstate.curentity->model->engineflags & MDLF_NEEDOVERBRIGHT)
{
float sc = (1<<bound(0, gl_overbright.ival, 2)) * shaderstate.identitylighting;
VectorSet(colscale[0], sc, sc, sc);
}
else
{
VectorSet(colscale[0], shaderstate.identitylighting, shaderstate.identitylighting, shaderstate.identitylighting);
}
colscale[0][3] = 1;
qglUniform4fvARB(ph, 1, (GLfloat*)colscale);
}
break;
case SP_E_GLOWMOD:
qglUniform3fvARB(ph, 1, (GLfloat*)shaderstate.curentity->glowmod);
break;
case SP_E_ORIGIN:
qglUniform3fvARB(ph, 1, (GLfloat*)shaderstate.curentity->origin);
break;
case SP_E_COLOURS:
qglUniform4fvARB(ph, 1, (GLfloat*)shaderstate.curentity->shaderRGBAf);
break;
case SP_E_COLOURSIDENT:
if (shaderstate.flags & BEF_FORCECOLOURMOD)
qglUniform4fvARB(ph, 1, (GLfloat*)shaderstate.curentity->shaderRGBAf);
else
qglUniform4fARB(ph, 1, 1, 1, shaderstate.curentity->shaderRGBAf[3]);
break;
case SP_E_TOPCOLOURS:
R_FetchTopColour(&r, &g, &b);
param4[0] = r/255.0f;
param4[1] = g/255.0f;
param4[2] = b/255.0f;
qglUniform3fvARB(ph, 1, param4);
break;
case SP_E_BOTTOMCOLOURS:
R_FetchBottomColour(&r, &g, &b);
param4[0] = r/255.0f;
param4[1] = g/255.0f;
param4[2] = b/255.0f;
qglUniform3fvARB(ph, 1, param4);
break;
case SP_RENDERTEXTURESCALE:
if (r_config.texture_non_power_of_two)
{
param4[0] = 1;
param4[1] = 1;
}
else
{
r = 1;
g = 1;
while (r < vid.pixelwidth)
r *= 2;
while (g < vid.pixelheight)
g *= 2;
param4[0] = vid.pixelwidth/(float)r;
param4[1] = vid.pixelheight/(float)g;
}
param4[2] = 0;
param4[3] = 0;
qglUniform4fvARB(ph, 1, param4);
break;
case SP_LIGHTSCREEN:
{
float v[4], tempv[4];
v[0] = shaderstate.lightorg[0];
v[1] = shaderstate.lightorg[1];
v[2] = shaderstate.lightorg[2];
v[3] = 1;
Matrix4x4_CM_Transform4(shaderstate.modelviewmatrix, v, tempv);
Matrix4x4_CM_Transform4(r_refdef.m_projection, tempv, v);
v[3] *= 2;
v[0] = (v[0]/v[3]) + 0.5;
v[1] = (v[1]/v[3]) + 0.5;
v[2] = (v[2]/v[3]) + 0.5;
qglUniform3fvARB(ph, 1, v);
}
break;
case SP_LIGHTRADIUS:
qglUniform1fARB(ph, shaderstate.lightradius);
break;
case SP_LIGHTCOLOUR:
qglUniform3fvARB(ph, 1, shaderstate.lightcolours);
break;
case SP_W_FOG:
qglUniform4fvARB(ph, 1, r_refdef.gfog_rgbd);
break;
case SP_V_EYEPOS:
qglUniform3fvARB(ph, 1, r_origin);
break;
case SP_E_EYEPOS:
{
/*eye position in model space*/
vec3_t t2;
Matrix4x4_CM_Transform3(shaderstate.modelmatrixinv, r_origin, t2);
qglUniform3fvARB(ph, 1, t2);
}
break;
case SP_LIGHTPOSITION:
{
/*light position in model space*/
vec3_t t2;
Matrix4x4_CM_Transform3(shaderstate.modelmatrixinv, shaderstate.lightorg, t2);
qglUniform3fvARB(ph, 1, t2);
}
break;
case SP_LIGHTCOLOURSCALE:
qglUniform3fvARB(ph, 1, shaderstate.lightcolourscale);
break;
case SP_LIGHTPROJMATRIX:
{
float t[16];
Matrix4x4_CM_Projection_Far(t, 90, 90, 4, 3000);
qglUniformMatrix4fvARB(ph, 1, false, t);
}
break;
case SP_LIGHTCUBEMATRIX:
/*light's texture projection matrix*/
{
float t[16];
Matrix4_Multiply(shaderstate.modelmatrix, shaderstate.lightprojmatrix, t);
qglUniformMatrix4fvARB(ph, 1, false, t);
}
break;
case SP_LIGHTSHADOWMAPINFO:
qglUniform4fvARB(ph, 1, shaderstate.lightshadowmapinfo);
break;
/*static lighting info*/
case SP_E_L_DIR:
qglUniform3fvARB(ph, 1, (float*)shaderstate.curentity->light_dir);
break;
case SP_E_L_MUL:
qglUniform3fvARB(ph, 1, (float*)shaderstate.curentity->light_range);
break;
case SP_E_L_AMBIENT:
qglUniform3fvARB(ph, 1, (float*)shaderstate.curentity->light_avg);
break;
case SP_E_TIME:
qglUniform1fARB(ph, shaderstate.curtime);
break;
case SP_CONSTI:
case SP_TEXTURE:
qglUniform1iARB(ph, p->ival);
break;
case SP_CONSTF:
qglUniform1fARB(ph, p->fval);
break;
case SP_CVARI:
qglUniform1iARB(ph, ((cvar_t*)p->pval)->ival);
break;
case SP_CVARF:
qglUniform1fARB(ph, ((cvar_t*)p->pval)->value);
break;
case SP_CVAR3F:
{
cvar_t *var = (cvar_t*)p->pval;
char *vs = var->string;
vs = COM_Parse(vs);
param4[0] = atof(com_token);
vs = COM_Parse(vs);
param4[1] = atof(com_token);
vs = COM_Parse(vs);
param4[2] = atof(com_token);
qglUniform3fvARB(ph, 1, param4);
}
break;
default:
Host_EndGame("Bad shader program parameter type (%i)", p->type);
break;
}
}
}
static void BE_RenderMeshProgram(const shader_t *shader, const shaderpass_t *pass)
{
program_t *p = shader->prog;
int i;
int perm;
perm = 0;
if (shaderstate.sourcevbo->numbones)
{
if (p->permu[perm|PERMUTATION_SKELETAL].handle.glsl)
perm |= PERMUTATION_SKELETAL;
else
return;
}
if (p->permu[perm|PERMUTATION_FRAMEBLEND].handle.glsl && shaderstate.sourcevbo->coord2.gl.addr)
perm |= PERMUTATION_FRAMEBLEND;
if (TEXVALID(shaderstate.curtexnums->bump) && p->permu[perm|PERMUTATION_BUMPMAP].handle.glsl)
perm |= PERMUTATION_BUMPMAP;
if (TEXVALID(shaderstate.curtexnums->fullbright) && p->permu[perm|PERMUTATION_FULLBRIGHT].handle.glsl)
perm |= PERMUTATION_FULLBRIGHT;
if ((TEXVALID(shaderstate.curtexnums->loweroverlay) || TEXVALID(shaderstate.curtexnums->upperoverlay)) && p->permu[perm|PERMUTATION_UPPERLOWER].handle.glsl)
perm |= PERMUTATION_UPPERLOWER;
if (r_refdef.gfog_rgbd[3] && p->permu[perm|PERMUTATION_FOG].handle.glsl)
perm |= PERMUTATION_FOG;
if (p->permu[perm|PERMUTATION_DELUXE].handle.glsl && TEXVALID(shaderstate.curtexnums->bump) && shaderstate.curbatch->lightmap[0] >= 0 && lightmap[shaderstate.curbatch->lightmap[0]]->hasdeluxe)
perm |= PERMUTATION_DELUXE;
if (shaderstate.curbatch->lightmap[1] >= 0 && p->permu[perm|PERMUTATION_LIGHTSTYLES].handle.glsl)
perm |= PERMUTATION_LIGHTSTYLES;
GL_SelectProgram(p->permu[perm].handle.glsl);
#ifndef FORCESTATE
if (shaderstate.lastuniform == p->permu[perm].handle.glsl)
i = true;
else
#endif
{
i = false;
shaderstate.lastuniform = p->permu[perm].handle.glsl;
}
BE_Program_Set_Attributes(p, perm, i);
BE_SendPassBlendDepthMask(pass->shaderbits);
BE_EnableShaderAttributes(p->permu[perm].attrmask, shaderstate.sourcevbo->vao);
if (p->nofixedcompat)
{
for (i = 0; i < pass->numMergedPasses; i++)
{
Shader_BindTextureForPass(i, pass+i);
}
if (perm & PERMUTATION_LIGHTSTYLES)
{
GL_LazyBind(i++, GL_TEXTURE_2D, shaderstate.curbatch->lightmap[1] >= 0?lightmap[shaderstate.curbatch->lightmap[1]]->lightmap_texture:r_nulltex);
GL_LazyBind(i++, GL_TEXTURE_2D, shaderstate.curbatch->lightmap[2] >= 0?lightmap[shaderstate.curbatch->lightmap[2]]->lightmap_texture:r_nulltex);
GL_LazyBind(i++, GL_TEXTURE_2D, shaderstate.curbatch->lightmap[3] >= 0?lightmap[shaderstate.curbatch->lightmap[3]]->lightmap_texture:r_nulltex);
//we need this loop to fix up fixed-function stuff
for (; i < shaderstate.lastpasstmus; i++)
{
GL_LazyBind(i, 0, r_nulltex);
}
shaderstate.lastpasstmus = pass->numMergedPasses+3;
}
else
{
//we need this loop to fix up fixed-function stuff
for (; i < shaderstate.lastpasstmus; i++)
{
GL_LazyBind(i, 0, r_nulltex);
}
shaderstate.lastpasstmus = pass->numMergedPasses;
}
}
else
{
GenerateColourMods(pass);
for (i = 0; i < pass->numMergedPasses; i++)
{
Shader_BindTextureForPass(i, pass+i);
BE_GeneratePassTC(pass+i, i);
}
for (; i < shaderstate.lastpasstmus; i++)
{
GL_LazyBind(i, 0, r_nulltex);
}
shaderstate.lastpasstmus = pass->numMergedPasses;
}
BE_SubmitMeshChain();
}
qboolean GLBE_LightCullModel(vec3_t org, model_t *model)
{
#ifdef RTLIGHTS
if ((shaderstate.mode == BEM_LIGHT || shaderstate.mode == BEM_STENCIL || shaderstate.mode == BEM_DEPTHONLY))
{
float dist;
vec3_t disp;
if (model->type == mod_alias)
{
VectorSubtract(org, shaderstate.lightorg, disp);
dist = DotProduct(disp, disp);
if (dist > model->radius*model->radius + shaderstate.lightradius*shaderstate.lightradius)
return true;
}
else
{
int i;
for (i = 0; i < 3; i++)
{
if (shaderstate.lightorg[i]-shaderstate.lightradius > org[i] + model->maxs[i])
return true;
if (shaderstate.lightorg[i]+shaderstate.lightradius < org[i] + model->mins[i])
return true;
}
}
}
#endif
return false;
}
//Note: Be cautious about using BEM_LIGHT here, as it won't select the light.
void GLBE_SelectMode(backendmode_t mode)
{
extern int gldepthfunc;
if (mode != shaderstate.mode)
{
shaderstate.mode = mode;
shaderstate.flags = 0;
switch (mode)
{
default:
break;
case BEM_DEPTHONLY:
GL_DeSelectProgram();
/*BEM_DEPTHONLY does support mesh writing, but its not the only way its used... FIXME!*/
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
//we don't write or blend anything (maybe alpha test... but mneh)
BE_SendPassBlendDepthMask(SBITS_MISC_DEPTHWRITE | SBITS_MASK_BITS);
BE_SetPassBlendMode(0, PBM_REPLACE);
GL_CullFace(SHADER_CULL_FRONT);
break;
#ifdef RTLIGHTS
case BEM_STENCIL:
/*BEM_STENCIL doesn't support mesh writing*/
GLBE_PushOffsetShadow(false);
if (gl_config.nofixedfunc && !shaderstate.allblackshader)
{
char *defs[] = {NULL};
shaderstate.allblackshader = GLSlang_CreateProgram("allblackprogram", gl_config.gles?100:110, defs, "#include \"sys/skeletal.h\"\nvoid main(){gl_Position = skeletaltransform();}", "void main(){gl_FragColor=vec4(0.0,0.0,0.0,1.0);}", false);
}
//disable all tmus
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
if (!gl_config.nofixedfunc)
{
GL_DeSelectProgram();
//replace mode please
BE_SetPassBlendMode(0, PBM_REPLACE);
}
//we don't write or blend anything (maybe alpha test... but mneh)
BE_SendPassBlendDepthMask(SBITS_MISC_DEPTHCLOSERONLY | SBITS_MASK_BITS);
GL_CullFace(0);
//don't change cull stuff, and
//don't actually change stencil stuff - caller needs to be
//aware of how many times stuff is drawn, so they can do that themselves.
break;
case BEM_CREPUSCULAR:
if (!shaderstate.crepopaqueshader)
{
shaderstate.crepopaqueshader = R_RegisterShader("crepuscular_opaque",
"{\n"
"program crepuscular_opaque\n"
"}\n"
);
}
if (!shaderstate.crepskyshader)
{
shaderstate.crepskyshader = R_RegisterShader("crepuscular_sky",
"{\n"
"program crepuscular_sky\n"
"{\n"
"map $diffuse\n"
"}\n"
"{\n"
"map $fullbright\n"
"}\n"
"}\n"
);
}
break;
#endif
case BEM_FOG:
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
GL_LazyBind(0, GL_TEXTURE_2D, shaderstate.fogtexture);
shaderstate.lastpasstmus = 1;
Vector4Set(shaderstate.pendingcolourflat, 1, 1, 1, 1);
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
BE_SetPassBlendMode(0, PBM_MODULATE);
BE_SendPassBlendDepthMask(SBITS_SRCBLEND_SRC_ALPHA | SBITS_DSTBLEND_ONE_MINUS_SRC_ALPHA | SBITS_MISC_DEPTHEQUALONLY);
break;
}
}
}
void GLBE_SelectEntity(entity_t *ent)
{
float nd;
shaderstate.curentity = ent;
currententity = ent;
R_RotateForEntity(shaderstate.modelmatrix, shaderstate.modelviewmatrix, shaderstate.curentity, shaderstate.curentity->model);
Matrix4_Invert(shaderstate.modelmatrix, shaderstate.modelmatrixinv);
if (qglLoadMatrixf)
qglLoadMatrixf(shaderstate.modelviewmatrix);
if (shaderstate.curentity->flags & Q2RF_DEPTHHACK)
nd = 0.3;
else
nd = 1;
if (shaderstate.depthrange != nd)
{
shaderstate.depthrange = nd;
if (qglDepthRange)
qglDepthRange (gldepthmin, gldepthmin + shaderstate.depthrange*(gldepthmax-gldepthmin));
else if (qglDepthRangef)
qglDepthRangef (gldepthmin, gldepthmin + shaderstate.depthrange*(gldepthmax-gldepthmin));
}
shaderstate.lastuniform = 0;
shaderstate.curtime = shaderstate.updatetime - shaderstate.curentity->shaderTime;
}
#if 0
static void BE_SelectFog(vec3_t colour, float alpha, float density)
{
float zscale;
density /= 64;
zscale = 2048; /*this value is meant to be the distance at which fog the value becomes as good as fully fogged, just hack it to 2048...*/
GenerateFogTexture(&shaderstate.fogtexture, density, zscale);
shaderstate.fogfar = 1/zscale; /*scaler for z coords*/
qglColor4f(colour[0], colour[1], colour[2], alpha);
}
#endif
#ifdef RTLIGHTS
static qboolean GLBE_RegisterLightShader(int mode)
{
if (!shaderstate.inited_shader_light[mode])
{
char *name = va("rtlight%s%s%s",
(mode & (1u<<LSHADER_SMAP))?"#PCF":"",
(mode & (1u<<LSHADER_SPOT))?"#SPOT":"",
(mode & (1u<<LSHADER_CUBE))?"#CUBE":"");
shaderstate.inited_shader_light[mode] = true;
shaderstate.shader_light[mode] = R_RegisterCustom(name, Shader_LightPass, NULL);
//make sure it has a program and forget it if it doesn't, to save a compare.
if (!shaderstate.shader_light[mode] || !shaderstate.shader_light[mode]->prog)
{
shaderstate.shader_light[mode] = NULL;
}
}
if (shaderstate.shader_light[mode])
return true;
return false;
}
#endif
void GLBE_SelectDLight(dlight_t *dl, vec3_t colour)
{
float view[16];
int lmode;
extern cvar_t gl_specular;
extern cvar_t r_shadow_shadowmapping;
/*generate light projection information*/
float nearplane = 4;
if (dl->fov)
{
float proj[16];
Matrix4x4_CM_Projection_Far(proj, dl->fov, dl->fov, nearplane, dl->radius);
Matrix4x4_CM_ModelViewMatrixFromAxis(view, dl->axis[0], dl->axis[1], dl->axis[2], dl->origin);
Matrix4_Multiply(proj, view, shaderstate.lightprojmatrix);
}
else
{
// Matrix4x4_CM_Projection_Far(proj, 90, 90, nearplane, dl->radius);
Matrix4x4_CM_ModelViewMatrixFromAxis(shaderstate.lightprojmatrix, dl->axis[0], dl->axis[1], dl->axis[2], dl->origin);
}
/*simple info*/
shaderstate.lightradius = dl->radius;
VectorCopy(dl->origin, shaderstate.lightorg);
VectorCopy(dl->lightcolourscales, shaderstate.lightcolourscale);
shaderstate.lightcolourscale[2] *= gl_specular.value;
VectorCopy(colour, shaderstate.lightcolours);
#ifdef RTLIGHTS
shaderstate.lightcubemap = dl->cubetexture;
#endif
shaderstate.lastuniform = 0;
lmode = 0;
#ifdef RTLIGHTS
if (((dl->flags & LFLAG_SHADOWMAP) || r_shadow_shadowmapping.ival) && GLBE_RegisterLightShader(lmode | (1u<<LSHADER_SMAP)))
lmode |= 1u<<LSHADER_SMAP;
else
GLBE_RegisterLightShader(lmode); //make sure either shadowmapping or non-shadowmapping is loaded...
if (TEXVALID(shaderstate.lightcubemap) && GLBE_RegisterLightShader(lmode | (1u<<LSHADER_CUBE)))
lmode |= 1u<<LSHADER_CUBE;
else if (dl->fov && GLBE_RegisterLightShader(lmode | (1u<<LSHADER_SPOT)))
lmode |= 1u<<LSHADER_SPOT;
#endif
shaderstate.lightmode = lmode;
}
void GLBE_Scissor(srect_t *rect)
{
if (rect)
{
qglScissor(
floor(r_refdef.pxrect.x + rect->x*r_refdef.pxrect.width),
floor((r_refdef.pxrect.y + rect->y*r_refdef.pxrect.height) - r_refdef.pxrect.height),
ceil(rect->width * r_refdef.pxrect.width),
ceil(rect->height * r_refdef.pxrect.height));
qglEnable(GL_SCISSOR_TEST);
if (qglDepthBoundsEXT)
{
qglDepthBoundsEXT(rect->dmin, rect->dmax);
qglEnable(GL_DEPTH_BOUNDS_TEST_EXT);
}
}
else
{
qglDisable(GL_SCISSOR_TEST);
if (qglDepthBoundsEXT)
qglDisable(GL_DEPTH_BOUNDS_TEST_EXT);
}
}
void GLBE_PushOffsetShadow(qboolean pushdepth)
{
extern cvar_t r_polygonoffset_stencil_offset, r_polygonoffset_stencil_factor;
polyoffset_t po;
if (pushdepth)
{
/*some quake doors etc are flush with the walls that they're meant to be hidden behind, or plats the same height as the floor, etc
we move them back very slightly using polygonoffset to avoid really ugly z-fighting*/
extern cvar_t r_polygonoffset_submodel_offset, r_polygonoffset_submodel_factor;
po.factor = r_polygonoffset_submodel_factor.value + r_polygonoffset_stencil_factor.value;
po.unit = r_polygonoffset_submodel_offset.value + r_polygonoffset_stencil_offset.value;
}
else
{
po.factor = r_polygonoffset_stencil_factor.value;
po.unit = r_polygonoffset_stencil_offset.value;
}
#ifndef FORCESTATE
if (shaderstate.curpolyoffset.factor != po.factor || shaderstate.curpolyoffset.unit != po.unit)
#endif
{
shaderstate.curpolyoffset = po;
if (shaderstate.curpolyoffset.factor || shaderstate.curpolyoffset.unit)
{
qglEnable(GL_POLYGON_OFFSET_FILL);
qglPolygonOffset(shaderstate.curpolyoffset.factor, shaderstate.curpolyoffset.unit);
}
else
qglDisable(GL_POLYGON_OFFSET_FILL);
}
}
#ifdef RTLIGHTS
texid_t GenerateNormalisationCubeMap(void);
static void BE_LegacyLighting(void)
{
//bigfoot wants rtlight support without glsl, so here goes madness...
//register combiners for bumpmapping using 4 tmus...
float *col;
float *ldir;
vec3_t lightdir, rellight;
float scale;
int i, m;
mesh_t *mesh;
unsigned int attr = (1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR);
int tmu = 0;
BE_SendPassBlendDepthMask(SBITS_SRCBLEND_ONE | SBITS_DSTBLEND_ONE);
//rotate this into modelspace
Matrix4x4_CM_Transform3(shaderstate.modelmatrixinv, shaderstate.lightorg, rellight);
for (m = 0; m < shaderstate.meshcount; m++)
{
mesh = shaderstate.meshes[m];
col = coloursarray[0] + mesh->vbofirstvert*4;
ldir = texcoordarray[0] + mesh->vbofirstvert*3;
for (i = 0; i < mesh->numvertexes; i++, col+=4, ldir+=3)
{
VectorSubtract(rellight, mesh->xyz_array[i], lightdir);
scale = VectorNormalize(lightdir);
scale = 1 - (scale/shaderstate.lightradius);
VectorScale(shaderstate.lightcolours, scale, col);
col[3] = 1;
ldir[0] = -DotProduct(lightdir, mesh->snormals_array[i]);
ldir[1] = DotProduct(lightdir, mesh->tnormals_array[i]);
ldir[2] = DotProduct(lightdir, mesh->normals_array[i]);
}
}
if (shaderstate.curtexnums->bump.num && gl_config.arb_texture_cube_map && gl_config.arb_texture_env_dot3 && gl_config.arb_texture_env_combine && be_maxpasses >= 4)
{ //we could get this down to 2 tmus by arranging for the dot3 result to be written the alpha buffer. But then we'd need to have an alpha buffer too.
if (!shaderstate.normalisationcubemap.num)
shaderstate.normalisationcubemap = GenerateNormalisationCubeMap();
//tmu0: normalmap+replace+regular tex coords
GL_LazyBind(tmu, GL_TEXTURE_2D, shaderstate.curtexnums->bump);
BE_SetPassBlendMode(tmu, PBM_REPLACE);
shaderstate.pendingtexcoordparts[tmu] = 2;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->texcoord.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->texcoord.gl.addr;
attr |= (1u<<(VATTR_LEG_TMU0+tmu));
tmu++;
//tmu1: normalizationcubemap+dot3+lightdir
GL_LazyBind(tmu, GL_TEXTURE_CUBE_MAP_ARB, shaderstate.normalisationcubemap);
BE_SetPassBlendMode(tmu, PBM_DOTPRODUCT);
shaderstate.pendingtexcoordparts[tmu] = 3;
shaderstate.pendingtexcoordvbo[tmu] = 0;
shaderstate.pendingtexcoordpointer[tmu] = texcoordarray[0];
attr |= (1u<<(VATTR_LEG_TMU0+tmu));
tmu++;
//tmu2: $diffuse+multiply+regular tex coords
GL_LazyBind(tmu, GL_TEXTURE_2D, shaderstate.curtexnums->base); //texture not used, its just to make sure the code leaves it enabled.
BE_SetPassBlendMode(tmu, PBM_MODULATE);
shaderstate.pendingtexcoordparts[tmu] = 2;
shaderstate.pendingtexcoordvbo[tmu] = shaderstate.sourcevbo->texcoord.gl.vbo;
shaderstate.pendingtexcoordpointer[tmu] = shaderstate.sourcevbo->texcoord.gl.addr;
attr |= (1u<<(VATTR_LEG_TMU0+tmu));
tmu++;
//tmu3: $any+multiply-by-colour+notc
GL_LazyBind(tmu, GL_TEXTURE_2D, shaderstate.curtexnums->bump); //texture not used, its just to make sure the code leaves it enabled.
BE_SetPassBlendMode(tmu, PBM_MODULATE_PREV_COLOUR);
shaderstate.pendingtexcoordparts[tmu] = 0;
shaderstate.pendingtexcoordvbo[tmu] = 0;
shaderstate.pendingtexcoordpointer[tmu] = NULL;
tmu++;
//note we need 4 combiners in the first because we can't use the colour argument in the first without breaking the normals.
for (i = tmu; i < shaderstate.lastpasstmus; i++)
{
GL_LazyBind(i, 0, r_nulltex);
}
shaderstate.lastpasstmus = tmu;
}
else
{
attr |= (1u<<(VATTR_LEG_TMU0));
//tmu0: $diffuse+multiply+regular tex coords
//multiplies by vertex colours
GL_LazyBind(0, GL_TEXTURE_2D, shaderstate.curtexnums->base); //texture not used, its just to make sure the code leaves it enabled.
BE_SetPassBlendMode(0, PBM_MODULATE);
shaderstate.pendingtexcoordvbo[0] = shaderstate.sourcevbo->texcoord.gl.vbo;
shaderstate.pendingtexcoordpointer[0] = shaderstate.sourcevbo->texcoord.gl.addr;
for (i = 1; i < shaderstate.lastpasstmus; i++)
{
GL_LazyBind(i, 0, r_nulltex);
}
shaderstate.lastpasstmus = 1;
}
shaderstate.colourarraytype = GL_FLOAT;
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = coloursarray;
GL_DeSelectProgram();
BE_EnableShaderAttributes(attr, 0);
BE_SubmitMeshChain();
GL_LazyBind(1, 0, r_nulltex);
GL_LazyBind(2, 0, r_nulltex);
GL_LazyBind(3, 0, r_nulltex);
}
#endif
static void DrawMeshes(void)
{
const shaderpass_t *p;
int passno;
passno = 0;
if (shaderstate.force2d)
{
RQuantAdd(RQUANT_2DBATCHES, 1);
}
else if (shaderstate.curentity == &r_worldentity)
{
RQuantAdd(RQUANT_WORLDBATCHES, 1);
}
else
{
RQuantAdd(RQUANT_ENTBATCHES, 1);
}
#ifndef FORCESTATE
if (shaderstate.curcull != (shaderstate.curshader->flags & (SHADER_CULL_FRONT|SHADER_CULL_BACK)))
#endif
{
shaderstate.curcull = (shaderstate.curshader->flags & (SHADER_CULL_FRONT|SHADER_CULL_BACK));
if (shaderstate.curcull & SHADER_CULL_FRONT)
{
qglEnable(GL_CULL_FACE);
qglCullFace(r_refdef.flipcull?GL_BACK:GL_FRONT);
}
else if (shaderstate.curcull & SHADER_CULL_BACK)
{
qglEnable(GL_CULL_FACE);
qglCullFace(r_refdef.flipcull?GL_FRONT:GL_BACK);
}
else
{
qglDisable(GL_CULL_FACE);
}
}
if (shaderstate.sourcevbo->coord2.gl.addr && (shaderstate.curshader->numdeforms || !shaderstate.curshader->prog))
GenerateVertexBlends(shaderstate.curshader);
else if (shaderstate.curshader->numdeforms)
GenerateVertexDeforms(shaderstate.curshader);
else
{
shaderstate.pendingvertexpointer = shaderstate.sourcevbo->coord.gl.addr;
shaderstate.pendingvertexvbo = shaderstate.sourcevbo->coord.gl.vbo;
}
#ifdef FTE_TARGET_WEB
if (!shaderstate.pendingvertexvbo)
return;
if (!shaderstate.sourcevbo->indicies.gl.vbo)
return;
#endif
BE_PolyOffset(shaderstate.flags & BEF_PUSHDEPTH);
switch(shaderstate.mode)
{
case BEM_STENCIL:
Host_Error("Shader system is not meant to accept stencil meshes\n");
break;
#ifdef RTLIGHTS
case BEM_SMAPLIGHTSPOT:
// if (shaderstate.shader_spot->prog)
// BE_RenderMeshProgram(shaderstate.shader_spot, shaderstate.shader_spot->passes);
// break;
case BEM_SMAPLIGHT:
// if (shaderstate.shader_smap->prog)
// BE_RenderMeshProgram(shaderstate.shader_smap, shaderstate.shader_smap->passes);
// break;
case BEM_LIGHT:
if (!shaderstate.shader_light[shaderstate.lightmode])
BE_LegacyLighting();
else
BE_RenderMeshProgram(shaderstate.shader_light[shaderstate.lightmode], shaderstate.shader_light[shaderstate.lightmode]->passes);
break;
case BEM_DEPTHNORM:
BE_RenderMeshProgram(shaderstate.depthnormshader, shaderstate.depthnormshader->passes);
break;
#endif
case BEM_CREPUSCULAR:
if (shaderstate.curshader->flags & SHADER_SKY)
BE_RenderMeshProgram(shaderstate.crepskyshader, shaderstate.crepskyshader->passes);
else
BE_RenderMeshProgram(shaderstate.crepopaqueshader, shaderstate.crepopaqueshader->passes);
break;
case BEM_DEPTHONLY:
GL_DeSelectProgram();
#ifdef warningmsg
#pragma warningmsg("fixme: support alpha test")
#endif
BE_EnableShaderAttributes((1u<<VATTR_LEG_VERTEX), 0);
BE_SubmitMeshChain();
break;
case BEM_FOG:
GL_DeSelectProgram();
GenerateTCFog(0, NULL);
BE_EnableShaderAttributes((1u<<VATTR_LEG_VERTEX), 0);
BE_SubmitMeshChain();
break;
case BEM_WIREFRAME:
//FIXME: do this with a shader instead? its not urgent as we can draw the shader normally anyway, just faster.
GL_DeSelectProgram();
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
Vector4Set(shaderstate.pendingcolourflat, 1, 1, 1, 1);
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
BE_SetPassBlendMode(0, PBM_REPLACE);
BE_SendPassBlendDepthMask(shaderstate.curshader->passes[0].shaderbits | SBITS_MISC_NODEPTHTEST);
BE_EnableShaderAttributes((1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR), 0);
BE_SubmitMeshChain();
break;
case BEM_DEPTHDARK:
if ((shaderstate.curshader->flags & SHADER_HASLIGHTMAP) && !TEXVALID(shaderstate.curtexnums->fullbright) && !gl_config.nofixedfunc)
{
if (gl_config.arb_shader_objects)
{
char *defs[] = {NULL};
if (!shaderstate.allblackshader)
shaderstate.allblackshader = GLSlang_CreateProgram("allblackprogram", gl_config.gles?100:110, defs, "#include \"sys/skeletal.h\"\nvoid main(){gl_Position = skeletaltransform();}", "void main(){gl_FragColor=vec4(0.0,0.0,0.0,1.0);}", false);
GL_SelectProgram(shaderstate.allblackshader);
BE_SendPassBlendDepthMask(shaderstate.curshader->passes[0].shaderbits);
BE_EnableShaderAttributes(1u<<VATTR_LEG_VERTEX, 0);
if (shaderstate.allblackshader != shaderstate.lastuniform)
{
float m16[16];
Matrix4_Multiply(r_refdef.m_projection, shaderstate.modelviewmatrix, m16);
qglUniformMatrix4fvARB(qglGetUniformLocationARB(shaderstate.allblackshader, "m_modelviewprojection"), 1, false, m16);
}
BE_SubmitMeshChain();
shaderstate.lastuniform = shaderstate.allblackshader;
}
else
{
GL_DeSelectProgram();
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
Vector4Set(shaderstate.pendingcolourflat, 0, 0, 0, 1);
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
BE_SetPassBlendMode(0, PBM_REPLACE);
BE_SendPassBlendDepthMask(shaderstate.curshader->passes[0].shaderbits);
BE_EnableShaderAttributes((1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR), 0);
BE_SubmitMeshChain();
}
break;
}
//fallthrough
case BEM_STANDARD:
default:
if (shaderstate.curshader->prog)
{
BE_RenderMeshProgram(shaderstate.curshader, shaderstate.curshader->passes);
}
else if (gl_config.nofixedfunc)
break;
else
{
GL_DeSelectProgram();
while (passno < shaderstate.curshader->numpasses)
{
p = &shaderstate.curshader->passes[passno];
passno += p->numMergedPasses;
// if (p->flags & SHADER_PASS_DETAIL)
// continue;
DrawPass(p);
}
}
if (shaderstate.curbatch->fog)
{
GL_DeSelectProgram();
GenerateFogTexture(&shaderstate.fogtexture, shaderstate.curbatch->fog->shader->fog_dist, 2048);
shaderstate.fogfar = 1.0f/2048; /*scaler for z coords*/
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
GL_LazyBind(0, GL_TEXTURE_2D, shaderstate.fogtexture);
shaderstate.lastpasstmus = 1;
Vector4Scale(shaderstate.curbatch->fog->shader->fog_color, (1/255.0), shaderstate.pendingcolourflat);
shaderstate.pendingcolourvbo = 0;
shaderstate.pendingcolourpointer = NULL;
BE_SetPassBlendMode(0, PBM_MODULATE);
BE_SendPassBlendDepthMask(SBITS_SRCBLEND_SRC_ALPHA | SBITS_DSTBLEND_ONE_MINUS_SRC_ALPHA | SBITS_MISC_DEPTHEQUALONLY);
GenerateTCFog(0, shaderstate.curbatch->fog);
BE_EnableShaderAttributes((1u<<VATTR_LEG_VERTEX) | (1u<<VATTR_LEG_COLOUR) | (1u<<VATTR_LEG_TMU0), 0);
BE_SubmitMeshChain();
}
break;
}
}
static qboolean BE_GenTempMeshVBO(vbo_t **vbo, mesh_t *m)
{
*vbo = &shaderstate.dummyvbo;
//this code is shit shit shit.
if (shaderstate.streamvbo[0])
{
//use a local. can't use a static, that crashes the compiler due to memory use, so lets eat the malloc or stack or whatever because we really don't have a choice.
static char *buffer;
size_t len = 0;
if (!buffer)
buffer = malloc(65536 * 33 * sizeof(float));
//we're not doing vao... but just in case someone added that as an extension...
GL_DeselectVAO();
//cycle the vbo. this reduces the likelyhood that we'll have to wait for the vbo to no longer be in use.
//remember, we can't do nice things like orphan in webgl. we have to create a new buffer every single time.
//we can't stream with BufferSubData, because browsers are pure shite, and probably mmap and read back huge blocks of code as they translate the calls to direct3d nonsense.
//we can't use client memory and let the driver do the right thing, because lets face it, our driver is technically a browser. its safer to run everything in a scripted language than to fight the nonsense.
//how many buffers do we need? no idea. all this memory allocation is going to be shit for performance.
//but we really do not have a choice. This is the only way to 'stream' without dropping down to <1fps.
//although arguably we should build our entire hud+2d stuff into a single vbo each frame... meh. At least this keeps the memory use in the driver's 64bit memory space instead of the browser's 32bit one...
shaderstate.streamid = (shaderstate.streamid + 1) & (sizeof(shaderstate.streamvbo)/sizeof(shaderstate.streamvbo[0]) - 1);
shaderstate.dummyvbo.vao = shaderstate.streamvao[shaderstate.streamid];
shaderstate.dummyvbo.vaodynamic = ~0;
shaderstate.dummyvbo.vaoenabled = 0;
if (shaderstate.dummyvbo.vao)
{
qglBindVertexArray(shaderstate.dummyvbo.vao);
shaderstate.currentvao = shaderstate.dummyvbo.vao;
}
GL_SelectVBO(shaderstate.streamvbo[shaderstate.streamid]);
GL_SelectEBO(shaderstate.streamebo[shaderstate.streamid]);
memcpy(buffer+len, m->xyz_array, sizeof(*m->xyz_array) * m->numvertexes);
shaderstate.dummyvbo.coord.gl.addr = (void*)len;
shaderstate.dummyvbo.coord.gl.vbo = shaderstate.streamvbo[shaderstate.streamid];
len += sizeof(*m->xyz_array) * m->numvertexes;
if (m->xyz2_array)
{
memcpy(buffer+len, m->xyz2_array, sizeof(*m->xyz2_array) * m->numvertexes);
shaderstate.dummyvbo.coord2.gl.addr = (void*)len;
shaderstate.dummyvbo.coord2.gl.vbo = shaderstate.streamvbo[shaderstate.streamid];
len += sizeof(*m->xyz2_array) * m->numvertexes;
}
else
{
shaderstate.dummyvbo.coord2.gl.addr = NULL;
shaderstate.dummyvbo.coord2.gl.vbo = 0;
}
memcpy(buffer+len, m->st_array, sizeof(*m->st_array) * m->numvertexes);
shaderstate.dummyvbo.texcoord.gl.addr = (void*)len;
shaderstate.dummyvbo.texcoord.gl.vbo = shaderstate.streamvbo[shaderstate.streamid];
len += sizeof(*m->st_array) * m->numvertexes;
//FIXME: lightmaps
if (m->colors4f_array)
{
memcpy(buffer+len, m->colors4f_array, sizeof(*m->colors4f_array) * m->numvertexes);
shaderstate.dummyvbo.colours.gl.addr = (void*)len;
shaderstate.dummyvbo.colours.gl.vbo = shaderstate.streamvbo[shaderstate.streamid];
len += sizeof(*m->colors4f_array) * m->numvertexes;
shaderstate.colourarraytype = GL_FLOAT;
}
else if (m->colors4b_array)
{
memcpy(buffer+len, m->colors4b_array, sizeof(*m->colors4b_array) * m->numvertexes);
shaderstate.dummyvbo.colours.gl.addr = (void*)len;
shaderstate.dummyvbo.colours.gl.vbo = shaderstate.streamvbo[shaderstate.streamid];
len += sizeof(*m->colors4b_array) * m->numvertexes;
shaderstate.colourarraytype = GL_UNSIGNED_BYTE;
}
else
{
shaderstate.dummyvbo.colours.gl.addr = NULL;
shaderstate.dummyvbo.colours.gl.vbo = 0;
shaderstate.colourarraytype = GL_FLOAT;
}
//FIXME: normals, svector, tvector, bone nums, bone weights
//now we've got a single buffer in a single place, update the buffer
qglBufferDataARB(GL_ARRAY_BUFFER_ARB, len, buffer, GL_STREAM_DRAW_ARB);
//and finally the elements array, which is a much simpler affair
qglBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, sizeof(*m->indexes) * m->numindexes, m->indexes, GL_STREAM_DRAW_ARB);
shaderstate.dummyvbo.indicies.gl.addr = (void*)NULL;
shaderstate.dummyvbo.indicies.gl.vbo = shaderstate.streamebo[shaderstate.streamid];
}
else
{
//client memory. may be slower. will probably be faster.
shaderstate.dummyvbo.coord.gl.addr = m->xyz_array;
shaderstate.dummyvbo.coord2.gl.addr = m->xyz2_array;
shaderstate.dummyvbo.texcoord.gl.addr = m->st_array;
shaderstate.dummyvbo.indicies.gl.addr = m->indexes;
shaderstate.dummyvbo.normals.gl.addr = m->normals_array;
shaderstate.dummyvbo.svector.gl.addr = m->snormals_array;
shaderstate.dummyvbo.tvector.gl.addr = m->tnormals_array;
if (m->colors4f_array)
{
shaderstate.colourarraytype = GL_FLOAT;
shaderstate.dummyvbo.colours.gl.addr = m->colors4f_array;
}
else
{
shaderstate.colourarraytype = GL_UNSIGNED_BYTE;
shaderstate.dummyvbo.colours.gl.addr = m->colors4b_array;
}
shaderstate.dummyvbo.bonenums.gl.addr = m->bonenums;
shaderstate.dummyvbo.boneweights.gl.addr = m->boneweights;
}
shaderstate.dummyvbo.bones = m->bones;
shaderstate.dummyvbo.numbones = m->numbones;
shaderstate.meshcount = 1;
shaderstate.meshes = &m;
return true;
}
void GLBE_DrawMesh_List(shader_t *shader, int nummeshes, mesh_t **meshlist, vbo_t *vbo, texnums_t *texnums, unsigned int beflags)
{
shaderstate.curbatch = &shaderstate.dummybatch;
if (!vbo)
{
mesh_t *m;
shaderstate.curshader = shader;
shaderstate.flags = beflags;
TRACE(("GLBE_DrawMesh_List: shader %s\n", shader->name));
if (shaderstate.curentity != &r_worldentity)
{
GLBE_SelectEntity(&r_worldentity);
}
shaderstate.curtexnums = texnums;
while (nummeshes--)
{
m = *meshlist++;
if (!BE_GenTempMeshVBO(&shaderstate.sourcevbo, m))
continue;
shaderstate.meshcount = 1;
shaderstate.meshes = &m;
DrawMeshes();
}
}
else
{
shaderstate.sourcevbo = vbo;
shaderstate.colourarraytype = GL_FLOAT;
shaderstate.curshader = shader;
shaderstate.flags = beflags;
if (shaderstate.curentity != &r_worldentity)
{
GLBE_SelectEntity(&r_worldentity);
}
shaderstate.curtexnums = texnums;
shaderstate.meshcount = nummeshes;
shaderstate.meshes = meshlist;
DrawMeshes();
}
}
void GLBE_DrawMesh_Single(shader_t *shader, mesh_t *mesh, vbo_t *vbo, texnums_t *texnums, unsigned int beflags)
{
shader->next = NULL;
BE_DrawMesh_List(shader, 1, &mesh, NULL, texnums, beflags);
}
void GLBE_SubmitBatch(batch_t *batch)
{
shaderstate.curbatch = batch;
if (batch->vbo)
{
shaderstate.sourcevbo = batch->vbo;
shaderstate.colourarraytype = GL_FLOAT;
}
else
{
//we're only allowed one mesh per batch if there's no vbo info.
if (!BE_GenTempMeshVBO(&shaderstate.sourcevbo, batch->mesh[0]))
return;
}
shaderstate.curshader = batch->shader;
shaderstate.flags = batch->flags;
if (shaderstate.curentity != batch->ent)
{
GLBE_SelectEntity(batch->ent);
}
if (batch->skin)
shaderstate.curtexnums = batch->skin;
else
shaderstate.curtexnums = &shaderstate.curshader->defaulttextures;
if (0)
{
int i;
for (i = batch->firstmesh; i < batch->meshes; i++)
{
shaderstate.meshcount = 1;
shaderstate.meshes = &batch->mesh[i];
DrawMeshes();
}
}
else
{
shaderstate.meshcount = batch->meshes - batch->firstmesh;
shaderstate.meshes = batch->mesh+batch->firstmesh;
DrawMeshes();
}
}
static void GLBE_SubmitMeshesPortals(batch_t **worldlist, batch_t *dynamiclist)
{
batch_t *batch, *old;
int i;
/*attempt to draw portal shaders*/
if (shaderstate.mode == BEM_STANDARD)
{
for (i = 0; i < 2; i++)
{
for (batch = i?dynamiclist:worldlist[SHADER_SORT_PORTAL]; batch; batch = batch->next)
{
if (batch->meshes == batch->firstmesh)
continue;
if (batch->buildmeshes)
batch->buildmeshes(batch);
else
batch->shader = R_TextureAnimation(batch->ent->framestate.g[FS_REG].frame[0], batch->texture)->shader;
/*draw already-drawn portals as depth-only, to ensure that their contents are not harmed*/
GLBE_SelectMode(BEM_DEPTHONLY);
for (old = worldlist[SHADER_SORT_PORTAL]; old && old != batch; old = old->next)
{
if (old->meshes == old->firstmesh)
continue;
GLBE_SubmitBatch(old);
}
if (!old)
{
for (old = dynamiclist; old != batch; old = old->next)
{
if (old->meshes == old->firstmesh)
continue;
GLBE_SubmitBatch(old);
}
}
GLBE_SelectMode(BEM_STANDARD);
GLR_DrawPortal(batch, worldlist, 0);
/*clear depth again*/
GL_ForceDepthWritable();
qglClear(GL_DEPTH_BUFFER_BIT);
currententity = &r_worldentity;
shaderstate.curtime = shaderstate.updatetime - shaderstate.curentity->shaderTime;
}
}
}
}
static void GLBE_SubmitMeshesSortList(batch_t *sortlist)
{
batch_t *batch;
for (batch = sortlist; batch; batch = batch->next)
{
if (batch->meshes == batch->firstmesh)
continue;
if (batch->flags & BEF_NODLIGHT)
if (shaderstate.mode == BEM_LIGHT || shaderstate.mode == BEM_SMAPLIGHT)
continue;
if (batch->flags & BEF_NOSHADOWS)
if (shaderstate.mode == BEM_STENCIL || shaderstate.mode == BEM_DEPTHONLY) //fixme: depthonly is not just shadows.
continue;
if (batch->buildmeshes)
{
TRACE(("GLBE_SubmitMeshesSortList: build\n"));
batch->buildmeshes(batch);
}
else if (batch->texture)
batch->shader = R_TextureAnimation(batch->ent->framestate.g[FS_REG].frame[0], batch->texture)->shader;
TRACE(("GLBE_SubmitMeshesSortList: shader %s\n", batch->shader->name));
if (batch->shader->flags & SHADER_NODRAW)
continue;
if (batch->shader->flags & SHADER_NODLIGHT)
if (shaderstate.mode == BEM_LIGHT || shaderstate.mode == BEM_SMAPLIGHT)
continue;
if (batch->shader->flags & SHADER_SKY)
{
if (shaderstate.mode == BEM_STANDARD || shaderstate.mode == BEM_DEPTHDARK)
{
if (!batch->shader->prog)
{
R_DrawSkyChain (batch);
continue;
}
}
else if (shaderstate.mode != BEM_FOG && shaderstate.mode != BEM_CREPUSCULAR && shaderstate.mode != BEM_WIREFRAME)
continue;
}
if ((batch->shader->flags & (SHADER_HASREFLECT | SHADER_HASREFRACT | SHADER_HASRIPPLEMAP)) && shaderstate.mode != BEM_WIREFRAME)
{
int oldbem;
//these flags require rendering some view as an fbo
if (r_refdef.recurse)
continue;
if (shaderstate.mode != BEM_STANDARD && shaderstate.mode != BEM_DEPTHDARK)
continue;
oldbem = shaderstate.mode;
if ((batch->shader->flags & SHADER_HASREFLECT) && gl_config.ext_framebuffer_objects)
{
vrect_t orect = r_refdef.vrect, oprect = r_refdef.pxrect;
if (!shaderstate.tex_reflection.num)
{
shaderstate.tex_reflection = GL_AllocNewTexture("***tex_reflection***", vid.pixelwidth/2, vid.pixelheight/2, 0);
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_reflection);
qglTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, vid.pixelwidth/2, vid.pixelheight/2, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
}
GL_ForceDepthWritable();
GLBE_RenderToTexture(r_nulltex, r_nulltex, shaderstate.tex_reflection, r_nulltex, true);
qglViewport (0, 0, vid.pixelwidth/2, vid.pixelheight/2);
r_refdef.vrect.x = 0;
r_refdef.vrect.y = 0;
r_refdef.vrect.width = vid.width/2;
r_refdef.vrect.height = vid.height/2;
r_refdef.pxrect.x = 0;
r_refdef.pxrect.width = vid.pixelwidth/2;
r_refdef.pxrect.height = vid.pixelheight/2;
r_refdef.pxrect.y = r_refdef.pxrect.height;
GL_ForceDepthWritable();
qglClearColor(0, 0, 0, 0);
qglClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
GLR_DrawPortal(batch, cl.worldmodel->batches, 1);
GLBE_RenderToTexture(r_nulltex, r_nulltex, r_nulltex, r_nulltex, false);
qglViewport (0, 0, vid.pixelwidth, vid.pixelheight);
r_refdef.vrect = orect;
r_refdef.pxrect = oprect;
}
if (batch->shader->flags & (SHADER_HASREFRACT|SHADER_HASREFRACTDEPTH))
{
if (r_refract_fboival)
{
vrect_t ovrect = r_refdef.vrect, oprect = r_refdef.pxrect;
GL_ForceDepthWritable();
if (!shaderstate.tex_refraction.num)
{
shaderstate.tex_refraction = GL_AllocNewTexture("***tex_refraction***", vid.pixelwidth/2, vid.pixelheight/2, 0);
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_refraction);
qglTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, vid.pixelwidth/2, vid.pixelheight/2, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
}
if (batch->shader->flags & SHADER_HASREFRACTDEPTH)
{
if (!shaderstate.tex_refractiondepth.num)
{
shaderstate.tex_refractiondepth = GL_AllocNewTexture("***tex_refractiondepth***", vid.pixelwidth/2, vid.pixelheight/2, 0);
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_refractiondepth);
qglTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT24_ARB, vid.pixelwidth/2, vid.pixelheight/2, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_BYTE, NULL);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
}
GLBE_RenderToTexture(r_nulltex, r_nulltex, shaderstate.tex_refraction, shaderstate.tex_refractiondepth, true);
}
else
GLBE_RenderToTexture(r_nulltex, r_nulltex, shaderstate.tex_refraction, r_nulltex, true);
qglViewport (0, 0, vid.pixelwidth/2, vid.pixelheight/2);
r_refdef.vrect.x = 0;
r_refdef.vrect.y = 0;
r_refdef.vrect.width = vid.width/2;
r_refdef.vrect.height = vid.height/2;
r_refdef.pxrect.x = 0;
r_refdef.pxrect.width = vid.pixelwidth/2;
r_refdef.pxrect.height = vid.pixelheight/2;
r_refdef.pxrect.y = r_refdef.pxrect.height;
GL_ForceDepthWritable();
qglClearColor(0, 0, 0, 0);
qglClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
GLR_DrawPortal(batch, cl.worldmodel->batches, ((batch->shader->flags & SHADER_HASREFRACTDEPTH)?3:2)); //fixme
GLBE_RenderToTexture(r_nulltex, r_nulltex, r_nulltex, r_nulltex, false);
qglViewport (0, 0, vid.pixelwidth, vid.pixelheight);
r_refdef.vrect = ovrect;
r_refdef.pxrect = oprect;
}
else
GLR_DrawPortal(batch, cl.worldmodel->batches, 3);
}
if ((batch->shader->flags & SHADER_HASRIPPLEMAP) && gl_config.ext_framebuffer_objects)
{
vrect_t orect = r_refdef.vrect, oprect = r_refdef.pxrect;
if (!shaderstate.tex_ripplemap.num)
{
shaderstate.tex_ripplemap = GL_AllocNewTexture("***tex_ripplemap***", vid.pixelwidth/2, vid.pixelheight/2, 0);
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_ripplemap);
qglTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA16F_ARB, vid.pixelwidth/2, vid.pixelheight/2, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
}
GLBE_RenderToTexture(r_nulltex, r_nulltex, shaderstate.tex_ripplemap, r_nulltex, false);
qglViewport (0, 0, vid.pixelwidth/2, vid.pixelheight/2);
r_refdef.vrect.x = 0;
r_refdef.vrect.y = 0;
r_refdef.vrect.width = vid.width/2;
r_refdef.vrect.height = vid.height/2;
r_refdef.pxrect.x = 0;
r_refdef.pxrect.width = vid.pixelwidth/2;
r_refdef.pxrect.height = vid.pixelheight/2;
r_refdef.pxrect.y = r_refdef.pxrect.height;
qglClearColor(0, 0, 0, 0);
qglClear(GL_COLOR_BUFFER_BIT);
// r_refdef.waterheight = DotProduct(batch->mesh[0]->xyz_array[0], batch->mesh[0]->normals_array[0]);
r_refdef.recurse = true; //paranoid, should stop potential infinite loops
GLBE_SubmitMeshes(true, SHADER_SORT_RIPPLE, SHADER_SORT_RIPPLE);
r_refdef.recurse = false;
GLBE_RenderToTexture(r_nulltex, r_nulltex, r_nulltex, r_nulltex, false);
qglViewport (0, 0, vid.pixelwidth, vid.pixelheight);
r_refdef.vrect = orect;
r_refdef.pxrect = oprect;
}
BE_SelectMode(oldbem);
}
GLBE_SubmitBatch(batch);
}
}
void GLBE_SubmitMeshes (qboolean drawworld, int start, int stop)
{
model_t *model = cl.worldmodel;
int i;
for (i = start; i <= stop; i++)
{
if (drawworld)
{
if (i == SHADER_SORT_PORTAL && !r_noportals.ival && !r_refdef.recurse)
GLBE_SubmitMeshesPortals(model->batches, shaderstate.mbatches[i]);
GLBE_SubmitMeshesSortList(model->batches[i]);
}
GLBE_SubmitMeshesSortList(shaderstate.mbatches[i]);
}
}
static void BE_UpdateLightmaps(void)
{
lightmapinfo_t *lm;
int lmidx;
int glformat, gltype;
switch (lightmap_bytes)
{
case 4:
glformat = lightmap_bgra?GL_BGRA_EXT:GL_RGBA;
gltype = GL_UNSIGNED_INT_8_8_8_8_REV;
break;
case 3:
glformat = lightmap_bgra?GL_BGR_EXT:GL_RGB;
gltype = GL_UNSIGNED_BYTE;
break;
default:
glformat = GL_LUMINANCE;
gltype = GL_UNSIGNED_BYTE;
break;
}
for (lmidx = 0; lmidx < numlightmaps; lmidx++)
{
lm = lightmap[lmidx];
if (!lm)
continue;
if (lm->modified)
{
lm->modified = false;
if (!TEXVALID(lm->lightmap_texture))
{
TEXASSIGN(lm->lightmap_texture, R_AllocNewTexture("***lightmap***", lm->width, lm->height, IF_LINEAR|IF_NOMIPMAP));
GL_MTBind(0, GL_TEXTURE_2D, lm->lightmap_texture);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
qglTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA,
lm->width, lm->height, 0, glformat, gltype,
lm->lightmaps);
}
else
{
GL_MTBind(0, GL_TEXTURE_2D, lm->lightmap_texture);
qglTexSubImage2D(GL_TEXTURE_2D, 0, 0, lm->rectchange.t,
lm->width, lm->rectchange.h, glformat, gltype,
lm->lightmaps+(lm->rectchange.t) *lm->width*lightmap_bytes);
}
lm->rectchange.l = LMBLOCK_WIDTH;
lm->rectchange.t = LMBLOCK_HEIGHT;
lm->rectchange.h = 0;
lm->rectchange.w = 0;
}
}
}
batch_t *GLBE_GetTempBatch(void)
{
batch_t *b;
if (shaderstate.wbatch >= shaderstate.maxwbatches)
{
shaderstate.wbatch++;
return NULL;
}
b = &shaderstate.wbatches[shaderstate.wbatch++];
b->fog = NULL;
return b;
}
/*called from shadowmapping code*/
#ifdef RTLIGHTS
void GLBE_BaseEntTextures(void)
{
batch_t *batches[SHADER_SORT_COUNT];
batch_t **ob = shaderstate.mbatches;
shaderstate.mbatches = batches;
BE_GenModelBatches(batches);
GLBE_SubmitMeshes(false, SHADER_SORT_PORTAL, SHADER_SORT_DECAL);
GLBE_SelectEntity(&r_worldentity);
shaderstate.mbatches = ob;
}
#endif
void GLBE_RenderToTexture(texid_t sourcecol, texid_t sourcedepth, texid_t destcol, texid_t destdepth, qboolean usedepth)
{
shaderstate.tex_sourcecol = sourcecol;
shaderstate.tex_sourcedepth = sourcedepth;
if (!destcol.num)
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
else
{
if (usedepth)
{
if (!shaderstate.fbo_diffuse)
{
qglGenFramebuffersEXT(1, &shaderstate.fbo_diffuse);
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, shaderstate.fbo_diffuse);
//create an unnamed depth buffer
qglGenRenderbuffersEXT(1, &shaderstate.rb_depth);
qglBindRenderbufferEXT(GL_RENDERBUFFER_EXT, shaderstate.rb_depth);
if (gl_config.ext_packed_depth_stencil)
{
qglRenderbufferStorageEXT(GL_RENDERBUFFER_EXT, GL_DEPTH24_STENCIL8_EXT, vid.pixelwidth/2, vid.pixelheight/2);
qglFramebufferRenderbufferEXT(GL_FRAMEBUFFER_EXT, GL_STENCIL_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, shaderstate.rb_depth);
}
else
{
qglRenderbufferStorageEXT(GL_RENDERBUFFER_EXT, GL_DEPTH_COMPONENT24_ARB, vid.pixelwidth/2, vid.pixelheight/2);
qglGenRenderbuffersEXT(1, &shaderstate.rb_stencil);
qglBindRenderbufferEXT(GL_RENDERBUFFER_EXT, shaderstate.rb_stencil);
qglRenderbufferStorageEXT(GL_RENDERBUFFER_EXT, GL_STENCIL_INDEX8_EXT, vid.pixelwidth/2, vid.pixelheight/2);
qglFramebufferRenderbufferEXT(GL_FRAMEBUFFER_EXT, GL_STENCIL_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, shaderstate.rb_stencil);
}
qglFramebufferRenderbufferEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, shaderstate.rb_depth);
qglDrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
qglReadBuffer(GL_NONE);
}
else
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, shaderstate.fbo_diffuse);
if (destdepth.num)
qglFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_TEXTURE_2D, destdepth.num, 0);
else
qglFramebufferRenderbufferEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, shaderstate.rb_depth);
}
else
{
if (!shaderstate.fbo_depthless)
{
qglGenFramebuffersEXT(1, &shaderstate.fbo_depthless);
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, shaderstate.fbo_depthless);
qglDrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
qglReadBuffer(GL_NONE);
}
else
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, shaderstate.fbo_depthless);
}
qglFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_2D, destcol.num, 0);
}
}
void GLBE_DrawLightPrePass(qbyte *vis)
{
if (!shaderstate.initeddepthnorm)
{
shaderstate.initeddepthnorm = true;
shaderstate.depthnormshader = R_RegisterShader("lpp_depthnorm",
"{\n"
"program lpp_depthnorm\n"
"{\n"
"map $normalmap\n"
"tcgen base\n"
"}\n"
"}\n"
);
}
if (!shaderstate.depthnormshader)
{
Con_Printf("%s requires content support\n", r_lightprepass.name);
r_lightprepass.ival = 0;
return;
}
/*do portals*/
BE_SelectMode(BEM_STANDARD);
GLBE_SubmitMeshes(true, SHADER_SORT_PORTAL, SHADER_SORT_PORTAL);
BE_SelectMode(BEM_DEPTHNORM);
if (!shaderstate.depthnormshader)
{
BE_SelectMode(BEM_STANDARD);
return;
}
if (!TEXVALID(shaderstate.tex_normals))
{
shaderstate.tex_normals = GL_AllocNewTexture("***prepass normals***", vid.pixelwidth, vid.pixelheight, 0);
r_lightprepass.modified = true;
}
if (r_lightprepass.modified)
{
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_normals);
qglTexImage2D(GL_TEXTURE_2D, 0, (r_lightprepass.ival==2)?GL_RGBA32F_ARB:GL_RGBA16F_ARB, vid.pixelwidth, vid.pixelheight, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
r_lightprepass.modified = false;
}
if (!TEXVALID(shaderstate.tex_diffuse))
{
int drb;
shaderstate.tex_diffuse = GL_AllocNewTexture("***prepass diffuse***", vid.pixelwidth, vid.pixelheight, 0);
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_diffuse);
qglTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, vid.pixelwidth, vid.pixelheight, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
GL_MTBind(0, GL_TEXTURE_2D, shaderstate.tex_normals);
qglTexImage2D(GL_TEXTURE_2D, 0, (r_lightprepass.ival==2)?GL_RGBA32F_ARB:GL_RGBA16F_ARB, vid.pixelwidth, vid.pixelheight, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);
r_lightprepass.modified = false;
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
qglGenFramebuffersEXT(1, &shaderstate.fbo_diffuse);
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, shaderstate.fbo_diffuse);
qglGenRenderbuffersEXT(1, &drb);
qglBindRenderbufferEXT(GL_RENDERBUFFER_EXT, drb);
qglRenderbufferStorageEXT(GL_RENDERBUFFER_EXT, GL_DEPTH_COMPONENT24_ARB, vid.pixelwidth, vid.pixelheight);
qglFramebufferRenderbufferEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, drb);
qglDrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
// qglReadBuffer(GL_NONE);
}
else
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, shaderstate.fbo_diffuse);
/*set the FB up to draw surface info*/
qglFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_2D, shaderstate.tex_normals.num, 0);
GL_ForceDepthWritable();
qglClear(GL_DEPTH_BUFFER_BIT);
if (GL_FRAMEBUFFER_COMPLETE_EXT != qglCheckFramebufferStatusEXT(GL_FRAMEBUFFER_EXT))
{
Con_Printf("Bad framebuffer\n");
return;
}
/*draw surfaces that can be drawn this way*/
GLBE_SubmitMeshes(true, SHADER_SORT_OPAQUE, SHADER_SORT_OPAQUE);
/*reconfigure - now drawing diffuse light info using the previous fb image as a source image*/
shaderstate.tex_sourcecol = shaderstate.tex_normals;
qglFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_2D, shaderstate.tex_diffuse.num, 0);
BE_SelectMode(BEM_STANDARD);
qglClearColor (0,0,0,0);
qglClear(GL_COLOR_BUFFER_BIT);
GLBE_SelectEntity(&r_worldentity);
/*now draw the prelights*/
GLBE_SubmitMeshes(true, SHADER_SORT_PRELIGHT, SHADER_SORT_PRELIGHT);
/*final reconfigure - now drawing final surface data onto true framebuffer*/
qglBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
shaderstate.tex_sourcecol = shaderstate.tex_diffuse;
qglDrawBuffer(GL_BACK);
/*now draw the postlight passes (this includes blended stuff which will NOT be lit)*/
GLBE_SelectEntity(&r_worldentity);
GLBE_SubmitMeshes(true, SHADER_SORT_SKY, SHADER_SORT_NEAREST);
#ifdef RTLIGHTS
/*regular lighting now*/
GLBE_SelectEntity(&r_worldentity);
Sh_DrawLights(vis);
#endif
shaderstate.tex_sourcecol = r_nulltex;
shaderstate.tex_sourcedepth = r_nulltex;
qglClearColor (1,0,0,1);
}
void GLBE_DrawWorld (qboolean drawworld, qbyte *vis)
{
extern cvar_t r_shadow_realtime_world, r_shadow_realtime_world_lightmaps;
batch_t *batches[SHADER_SORT_COUNT];
batch_t **ob = shaderstate.mbatches;
RSpeedLocals();
shaderstate.mbatches = batches;
shaderstate.depthrange = 0;
TRACE(("GLBE_DrawWorld: %i %p\n", drawworld, vis));
if (!r_refdef.recurse)
{
GL_DoSwap();
if (shaderstate.wbatch + 50 > shaderstate.maxwbatches)
{
int newm = shaderstate.wbatch + 100;
shaderstate.wbatches = BZ_Realloc(shaderstate.wbatches, newm * sizeof(*shaderstate.wbatches));
memset(shaderstate.wbatches + shaderstate.maxwbatches, 0, (newm - shaderstate.maxwbatches) * sizeof(*shaderstate.wbatches));
shaderstate.maxwbatches = newm;
}
shaderstate.wbatch = 0;
}
if (shaderstate.oldwidth != vid.pixelwidth || shaderstate.oldheight != vid.pixelheight)
{
if (shaderstate.tex_reflection.num)
{
R_DestroyTexture(shaderstate.tex_reflection);
shaderstate.tex_reflection = r_nulltex;
}
if (shaderstate.tex_refraction.num)
{
R_DestroyTexture(shaderstate.tex_refraction);
shaderstate.tex_refraction = r_nulltex;
}
if (shaderstate.tex_refractiondepth.num)
{
R_DestroyTexture(shaderstate.tex_refractiondepth);
shaderstate.tex_refractiondepth = r_nulltex;
}
if (shaderstate.temptexture.num)
{
R_DestroyTexture(shaderstate.temptexture);
shaderstate.temptexture = r_nulltex;
}
if (shaderstate.fbo_diffuse)
{
qglDeleteFramebuffersEXT(1, &shaderstate.fbo_diffuse);
shaderstate.fbo_diffuse = 0;
}
if (shaderstate.rb_depth)
qglDeleteRenderbuffersEXT(1, &shaderstate.rb_depth);
shaderstate.rb_depth = 0;
if (shaderstate.rb_stencil)
qglDeleteRenderbuffersEXT(1, &shaderstate.rb_stencil);
shaderstate.rb_stencil = 0;
shaderstate.oldwidth = vid.pixelwidth;
shaderstate.oldheight = vid.pixelheight;
while(shaderstate.lastpasstmus>0)
{
GL_LazyBind(--shaderstate.lastpasstmus, 0, r_nulltex);
}
#ifdef RTLIGHTS
Sh_Reset();
#endif
}
BE_GenModelBatches(batches);
R_GenDlightBatches(batches);
shaderstate.curentity = &r_worldentity;
if (cl.paused || cls.state < ca_active)
shaderstate.updatetime = r_refdef.time;
else
shaderstate.updatetime = cl.servertime;
GLBE_SelectEntity(&r_worldentity);
BE_UpdateLightmaps();
if (drawworld)
{
if (gl_overbright.modified)
{
int i;
gl_overbright.modified = false;
if (gl_overbright.ival > 2)
gl_overbright.ival = 2;
for (i = 0; i < SHADER_TMU_MAX; i++)
shaderstate.blendmode[i] = -1;
}
#ifdef RTLIGHTS
if (vis && r_shadow_realtime_world.ival)
shaderstate.identitylighting = r_shadow_realtime_world_lightmaps.value;
else
#endif
shaderstate.identitylighting = 1;
// shaderstate.identitylighting /= 1<<gl_overbright.ival;
#ifdef RTLIGHTS
if (r_lightprepass.ival)
{
GLBE_DrawLightPrePass(vis);
}
else
#endif
{
if (shaderstate.identitylighting == 0)
BE_SelectMode(BEM_DEPTHDARK);
else
BE_SelectMode(BEM_STANDARD);
RSpeedRemark();
GLBE_SubmitMeshes(true, SHADER_SORT_PORTAL, SHADER_SORT_DECAL);
RSpeedEnd(RSPEED_WORLD);
}
#ifdef RTLIGHTS
if (drawworld)
{
RSpeedRemark();
TRACE(("GLBE_DrawWorld: drawing lights\n"));
GLBE_SelectEntity(&r_worldentity);
Sh_DrawLights(vis);
RSpeedEnd(RSPEED_STENCILSHADOWS);
TRACE(("GLBE_DrawWorld: lights drawn\n"));
}
#endif
shaderstate.identitylighting = 1;
GLBE_SubmitMeshes(true, SHADER_SORT_DECAL, SHADER_SORT_NEAREST);
/* if (r_refdef.gfog_alpha)
{
BE_SelectMode(BEM_FOG);
BE_SelectFog(r_refdef.gfog_rgb, r_refdef.gfog_alpha, r_refdef.gfog_density);
GLBE_SubmitMeshes(true, batches, SHADER_SORT_PORTAL, SHADER_SORT_NEAREST);
}
*/
#ifdef GL_LINE //no gles
if (r_wireframe.ival && qglPolygonMode)
{
BE_SelectMode(BEM_WIREFRAME);
qglPolygonMode(GL_FRONT_AND_BACK, GL_LINE);
GLBE_SubmitMeshes(true, SHADER_SORT_PORTAL, SHADER_SORT_NEAREST);
BE_SelectMode(BEM_STANDARD);
qglPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
}
#endif
}
else
{
GLBE_SubmitMeshes(false, SHADER_SORT_PORTAL, SHADER_SORT_NEAREST);
#ifdef GL_LINE //no gles
if (r_wireframe.ival && qglPolygonMode)
{
BE_SelectMode(BEM_WIREFRAME);
qglPolygonMode(GL_FRONT_AND_BACK, GL_LINE);
GLBE_SubmitMeshes(false, SHADER_SORT_PORTAL, SHADER_SORT_NEAREST);
BE_SelectMode(BEM_STANDARD);
qglPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
}
#endif
}
GLBE_SelectEntity(&r_worldentity);
shaderstate.curtime = shaderstate.updatetime = realtime;
shaderstate.identitylighting = 1;
shaderstate.mbatches = ob;
TRACE(("GLBE_DrawWorld: drawn everything\n"));
}
void GLBE_VBO_Begin(vbobctx_t *ctx, unsigned int maxsize)
{
ctx->maxsize = maxsize;
ctx->pos = 0;
ctx->fallback = NULL;
if (qglBufferDataARB)
{
ctx->vboid[0] = 0;
ctx->vboid[1] = 0;
qglGenBuffersARB(2, ctx->vboid);
GL_SelectVBO(ctx->vboid[0]);
//WARNING: in emscripten/webgl, we should probably not pass null.
qglBufferDataARB(GL_ARRAY_BUFFER_ARB, ctx->maxsize, NULL, GL_STATIC_DRAW_ARB);
}
else
ctx->fallback = BZ_Malloc(maxsize);
}
void GLBE_VBO_Data(vbobctx_t *ctx, void *data, unsigned int size, vboarray_t *varray)
{
if (ctx->fallback)
{
memcpy((char*)ctx->fallback + ctx->pos, data, size);
varray->gl.vbo = 0;
varray->gl.addr = (char*)ctx->fallback + ctx->pos;
}
else
{
qglBufferSubDataARB(GL_ARRAY_BUFFER_ARB, ctx->pos, size, data);
varray->gl.vbo = ctx->vboid[0];
varray->gl.addr = (void*)ctx->pos;
}
ctx->pos += size;
}
void GLBE_VBO_Finish(vbobctx_t *ctx, void *edata, unsigned int esize, vboarray_t *earray)
{
if (ctx->pos > ctx->maxsize)
Sys_Error("BE_VBO_Finish: too much data given\n");
if (ctx->fallback)
{
void *d = BZ_Malloc(esize);
memcpy(d, edata, esize);
earray->gl.vbo = 0;
earray->gl.addr = d;
}
else
{
GL_SelectEBO(ctx->vboid[1]);
qglBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, esize, edata, GL_STATIC_DRAW_ARB);
earray->gl.vbo = ctx->vboid[1];
earray->gl.addr = NULL;
}
}
void GLBE_VBO_Destroy(vboarray_t *vearray)
{
if (vearray->gl.vbo)
qglDeleteBuffersARB(1, &vearray->gl.vbo);
else
BZ_Free(vearray->gl.addr);
}
#endif