ioq3/code/renderergl1/tr_world.c

668 lines
15 KiB
C

/*
===========================================================================
Copyright (C) 1999-2005 Id Software, Inc.
This file is part of Quake III Arena source code.
Quake III Arena source code is free software; you can redistribute it
and/or modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the License,
or (at your option) any later version.
Quake III Arena source code is distributed in the hope that it will be
useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with Quake III Arena source code; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
===========================================================================
*/
#include "tr_local.h"
/*
=================
R_CullTriSurf
Returns true if the grid is completely culled away.
Also sets the clipped hint bit in tess
=================
*/
static qboolean R_CullTriSurf( srfTriangles_t *cv ) {
int boxCull;
boxCull = R_CullLocalBox( cv->bounds );
if ( boxCull == CULL_OUT ) {
return qtrue;
}
return qfalse;
}
/*
=================
R_CullGrid
Returns true if the grid is completely culled away.
Also sets the clipped hint bit in tess
=================
*/
static qboolean R_CullGrid( srfGridMesh_t *cv ) {
int boxCull;
int sphereCull;
if ( r_nocurves->integer ) {
return qtrue;
}
if ( tr.currentEntityNum != REFENTITYNUM_WORLD ) {
sphereCull = R_CullLocalPointAndRadius( cv->localOrigin, cv->meshRadius );
} else {
sphereCull = R_CullPointAndRadius( cv->localOrigin, cv->meshRadius );
}
// check for trivial reject
if ( sphereCull == CULL_OUT )
{
tr.pc.c_sphere_cull_patch_out++;
return qtrue;
}
// check bounding box if necessary
else if ( sphereCull == CULL_CLIP )
{
tr.pc.c_sphere_cull_patch_clip++;
boxCull = R_CullLocalBox( cv->meshBounds );
if ( boxCull == CULL_OUT )
{
tr.pc.c_box_cull_patch_out++;
return qtrue;
}
else if ( boxCull == CULL_IN )
{
tr.pc.c_box_cull_patch_in++;
}
else
{
tr.pc.c_box_cull_patch_clip++;
}
}
else
{
tr.pc.c_sphere_cull_patch_in++;
}
return qfalse;
}
/*
================
R_CullSurface
Tries to back face cull surfaces before they are lighted or
added to the sorting list.
This will also allow mirrors on both sides of a model without recursion.
================
*/
static qboolean R_CullSurface( surfaceType_t *surface, shader_t *shader ) {
srfSurfaceFace_t *sface;
float d;
if ( r_nocull->integer ) {
return qfalse;
}
if ( *surface == SF_GRID ) {
return R_CullGrid( (srfGridMesh_t *)surface );
}
if ( *surface == SF_TRIANGLES ) {
return R_CullTriSurf( (srfTriangles_t *)surface );
}
if ( *surface != SF_FACE ) {
return qfalse;
}
if ( shader->cullType == CT_TWO_SIDED ) {
return qfalse;
}
// face culling
if ( !r_facePlaneCull->integer ) {
return qfalse;
}
sface = ( srfSurfaceFace_t * ) surface;
d = DotProduct (tr.or.viewOrigin, sface->plane.normal);
// don't cull exactly on the plane, because there are levels of rounding
// through the BSP, ICD, and hardware that may cause pixel gaps if an
// epsilon isn't allowed here
if ( shader->cullType == CT_FRONT_SIDED ) {
if ( d < sface->plane.dist - 8 ) {
return qtrue;
}
} else {
if ( d > sface->plane.dist + 8 ) {
return qtrue;
}
}
return qfalse;
}
static int R_DlightFace( srfSurfaceFace_t *face, int dlightBits ) {
float d;
int i;
dlight_t *dl;
for ( i = 0 ; i < tr.refdef.num_dlights ; i++ ) {
if ( ! ( dlightBits & ( 1 << i ) ) ) {
continue;
}
dl = &tr.refdef.dlights[i];
d = DotProduct( dl->origin, face->plane.normal ) - face->plane.dist;
if ( d < -dl->radius || d > dl->radius ) {
// dlight doesn't reach the plane
dlightBits &= ~( 1 << i );
}
}
if ( !dlightBits ) {
tr.pc.c_dlightSurfacesCulled++;
}
face->dlightBits = dlightBits;
return dlightBits;
}
static int R_DlightGrid( srfGridMesh_t *grid, int dlightBits ) {
int i;
dlight_t *dl;
for ( i = 0 ; i < tr.refdef.num_dlights ; i++ ) {
if ( ! ( dlightBits & ( 1 << i ) ) ) {
continue;
}
dl = &tr.refdef.dlights[i];
if ( dl->origin[0] - dl->radius > grid->meshBounds[1][0]
|| dl->origin[0] + dl->radius < grid->meshBounds[0][0]
|| dl->origin[1] - dl->radius > grid->meshBounds[1][1]
|| dl->origin[1] + dl->radius < grid->meshBounds[0][1]
|| dl->origin[2] - dl->radius > grid->meshBounds[1][2]
|| dl->origin[2] + dl->radius < grid->meshBounds[0][2] ) {
// dlight doesn't reach the bounds
dlightBits &= ~( 1 << i );
}
}
if ( !dlightBits ) {
tr.pc.c_dlightSurfacesCulled++;
}
grid->dlightBits = dlightBits;
return dlightBits;
}
static int R_DlightTrisurf( srfTriangles_t *surf, int dlightBits ) {
// FIXME: more dlight culling to trisurfs...
surf->dlightBits = dlightBits;
return dlightBits;
#if 0
int i;
dlight_t *dl;
for ( i = 0 ; i < tr.refdef.num_dlights ; i++ ) {
if ( ! ( dlightBits & ( 1 << i ) ) ) {
continue;
}
dl = &tr.refdef.dlights[i];
if ( dl->origin[0] - dl->radius > grid->meshBounds[1][0]
|| dl->origin[0] + dl->radius < grid->meshBounds[0][0]
|| dl->origin[1] - dl->radius > grid->meshBounds[1][1]
|| dl->origin[1] + dl->radius < grid->meshBounds[0][1]
|| dl->origin[2] - dl->radius > grid->meshBounds[1][2]
|| dl->origin[2] + dl->radius < grid->meshBounds[0][2] ) {
// dlight doesn't reach the bounds
dlightBits &= ~( 1 << i );
}
}
if ( !dlightBits ) {
tr.pc.c_dlightSurfacesCulled++;
}
grid->dlightBits = dlightBits;
return dlightBits;
#endif
}
/*
====================
R_DlightSurface
The given surface is going to be drawn, and it touches a leaf
that is touched by one or more dlights, so try to throw out
more dlights if possible.
====================
*/
static int R_DlightSurface( msurface_t *surf, int dlightBits ) {
if ( *surf->data == SF_FACE ) {
dlightBits = R_DlightFace( (srfSurfaceFace_t *)surf->data, dlightBits );
} else if ( *surf->data == SF_GRID ) {
dlightBits = R_DlightGrid( (srfGridMesh_t *)surf->data, dlightBits );
} else if ( *surf->data == SF_TRIANGLES ) {
dlightBits = R_DlightTrisurf( (srfTriangles_t *)surf->data, dlightBits );
} else {
dlightBits = 0;
}
if ( dlightBits ) {
tr.pc.c_dlightSurfaces++;
}
return dlightBits;
}
/*
======================
R_AddWorldSurface
======================
*/
static void R_AddWorldSurface( msurface_t *surf, int dlightBits ) {
if ( surf->viewCount == tr.viewCount ) {
return; // already in this view
}
surf->viewCount = tr.viewCount;
// FIXME: bmodel fog?
// try to cull before dlighting or adding
if ( R_CullSurface( surf->data, surf->shader ) ) {
return;
}
// check for dlighting
if ( dlightBits ) {
dlightBits = R_DlightSurface( surf, dlightBits );
dlightBits = ( dlightBits != 0 );
}
R_AddDrawSurf( surf->data, surf->shader, surf->fogIndex, dlightBits );
}
/*
=============================================================
BRUSH MODELS
=============================================================
*/
/*
=================
R_AddBrushModelSurfaces
=================
*/
void R_AddBrushModelSurfaces ( trRefEntity_t *ent ) {
bmodel_t *bmodel;
int clip;
model_t *pModel;
int i;
pModel = R_GetModelByHandle( ent->e.hModel );
bmodel = pModel->bmodel;
clip = R_CullLocalBox( bmodel->bounds );
if ( clip == CULL_OUT ) {
return;
}
R_SetupEntityLighting( &tr.refdef, ent );
R_DlightBmodel( bmodel );
for ( i = 0 ; i < bmodel->numSurfaces ; i++ ) {
R_AddWorldSurface( bmodel->firstSurface + i, tr.currentEntity->needDlights );
}
}
/*
=============================================================
WORLD MODEL
=============================================================
*/
/*
================
R_RecursiveWorldNode
================
*/
static void R_RecursiveWorldNode( mnode_t *node, int planeBits, int dlightBits ) {
do {
int newDlights[2];
// if the node wasn't marked as potentially visible, exit
if (node->visframe != tr.visCount) {
return;
}
// if the bounding volume is outside the frustum, nothing
// inside can be visible OPTIMIZE: don't do this all the way to leafs?
if ( !r_nocull->integer ) {
int r;
if ( planeBits & 1 ) {
r = BoxOnPlaneSide(node->mins, node->maxs, &tr.viewParms.frustum[0]);
if (r == 2) {
return; // culled
}
if ( r == 1 ) {
planeBits &= ~1; // all descendants will also be in front
}
}
if ( planeBits & 2 ) {
r = BoxOnPlaneSide(node->mins, node->maxs, &tr.viewParms.frustum[1]);
if (r == 2) {
return; // culled
}
if ( r == 1 ) {
planeBits &= ~2; // all descendants will also be in front
}
}
if ( planeBits & 4 ) {
r = BoxOnPlaneSide(node->mins, node->maxs, &tr.viewParms.frustum[2]);
if (r == 2) {
return; // culled
}
if ( r == 1 ) {
planeBits &= ~4; // all descendants will also be in front
}
}
if ( planeBits & 8 ) {
r = BoxOnPlaneSide(node->mins, node->maxs, &tr.viewParms.frustum[3]);
if (r == 2) {
return; // culled
}
if ( r == 1 ) {
planeBits &= ~8; // all descendants will also be in front
}
}
}
if ( node->contents != -1 ) {
break;
}
// node is just a decision point, so go down both sides
// since we don't care about sort orders, just go positive to negative
// determine which dlights are needed
newDlights[0] = 0;
newDlights[1] = 0;
if ( dlightBits ) {
int i;
for ( i = 0 ; i < tr.refdef.num_dlights ; i++ ) {
dlight_t *dl;
float dist;
if ( dlightBits & ( 1 << i ) ) {
dl = &tr.refdef.dlights[i];
dist = DotProduct( dl->origin, node->plane->normal ) - node->plane->dist;
if ( dist > -dl->radius ) {
newDlights[0] |= ( 1 << i );
}
if ( dist < dl->radius ) {
newDlights[1] |= ( 1 << i );
}
}
}
}
// recurse down the children, front side first
R_RecursiveWorldNode (node->children[0], planeBits, newDlights[0] );
// tail recurse
node = node->children[1];
dlightBits = newDlights[1];
} while ( 1 );
{
// leaf node, so add mark surfaces
int c;
msurface_t *surf, **mark;
tr.pc.c_leafs++;
// add to z buffer bounds
if ( node->mins[0] < tr.viewParms.visBounds[0][0] ) {
tr.viewParms.visBounds[0][0] = node->mins[0];
}
if ( node->mins[1] < tr.viewParms.visBounds[0][1] ) {
tr.viewParms.visBounds[0][1] = node->mins[1];
}
if ( node->mins[2] < tr.viewParms.visBounds[0][2] ) {
tr.viewParms.visBounds[0][2] = node->mins[2];
}
if ( node->maxs[0] > tr.viewParms.visBounds[1][0] ) {
tr.viewParms.visBounds[1][0] = node->maxs[0];
}
if ( node->maxs[1] > tr.viewParms.visBounds[1][1] ) {
tr.viewParms.visBounds[1][1] = node->maxs[1];
}
if ( node->maxs[2] > tr.viewParms.visBounds[1][2] ) {
tr.viewParms.visBounds[1][2] = node->maxs[2];
}
// add the individual surfaces
mark = node->firstmarksurface;
c = node->nummarksurfaces;
while (c--) {
// the surface may have already been added if it
// spans multiple leafs
surf = *mark;
R_AddWorldSurface( surf, dlightBits );
mark++;
}
}
}
/*
===============
R_PointInLeaf
===============
*/
static mnode_t *R_PointInLeaf( const vec3_t p ) {
mnode_t *node;
float d;
cplane_t *plane;
if ( !tr.world ) {
ri.Error (ERR_DROP, "R_PointInLeaf: bad model");
}
node = tr.world->nodes;
while( 1 ) {
if (node->contents != -1) {
break;
}
plane = node->plane;
d = DotProduct (p,plane->normal) - plane->dist;
if (d > 0) {
node = node->children[0];
} else {
node = node->children[1];
}
}
return node;
}
/*
==============
R_ClusterPVS
==============
*/
static const byte *R_ClusterPVS (int cluster) {
if (!tr.world->vis || cluster < 0 || cluster >= tr.world->numClusters ) {
return tr.world->novis;
}
return tr.world->vis + cluster * tr.world->clusterBytes;
}
/*
=================
R_inPVS
=================
*/
qboolean R_inPVS( const vec3_t p1, const vec3_t p2 ) {
mnode_t *leaf;
byte *vis;
leaf = R_PointInLeaf( p1 );
vis = ri.CM_ClusterPVS( leaf->cluster ); // why not R_ClusterPVS ??
leaf = R_PointInLeaf( p2 );
if ( !(vis[leaf->cluster>>3] & (1<<(leaf->cluster&7))) ) {
return qfalse;
}
return qtrue;
}
/*
===============
R_MarkLeaves
Mark the leaves and nodes that are in the PVS for the current
cluster
===============
*/
static void R_MarkLeaves (void) {
const byte *vis;
mnode_t *leaf, *parent;
int i;
int cluster;
// lockpvs lets designers walk around to determine the
// extent of the current pvs
if ( r_lockpvs->integer ) {
return;
}
// current viewcluster
leaf = R_PointInLeaf( tr.viewParms.pvsOrigin );
cluster = leaf->cluster;
// if the cluster is the same and the area visibility matrix
// hasn't changed, we don't need to mark everything again
// if r_showcluster was just turned on, remark everything
if ( tr.viewCluster == cluster && !tr.refdef.areamaskModified
&& !r_showcluster->modified ) {
return;
}
if ( r_showcluster->modified || r_showcluster->integer ) {
r_showcluster->modified = qfalse;
if ( r_showcluster->integer ) {
ri.Printf( PRINT_ALL, "cluster:%i area:%i\n", cluster, leaf->area );
}
}
tr.visCount++;
tr.viewCluster = cluster;
if ( r_novis->integer || tr.viewCluster == -1 ) {
for (i=0 ; i<tr.world->numnodes ; i++) {
if (tr.world->nodes[i].contents != CONTENTS_SOLID) {
tr.world->nodes[i].visframe = tr.visCount;
}
}
return;
}
vis = R_ClusterPVS (tr.viewCluster);
for (i=0,leaf=tr.world->nodes ; i<tr.world->numnodes ; i++, leaf++) {
cluster = leaf->cluster;
if ( cluster < 0 || cluster >= tr.world->numClusters ) {
continue;
}
// check general pvs
if ( !(vis[cluster>>3] & (1<<(cluster&7))) ) {
continue;
}
// check for door connection
if ( (tr.refdef.areamask[leaf->area>>3] & (1<<(leaf->area&7)) ) ) {
continue; // not visible
}
parent = leaf;
do {
if (parent->visframe == tr.visCount)
break;
parent->visframe = tr.visCount;
parent = parent->parent;
} while (parent);
}
}
/*
=============
R_AddWorldSurfaces
=============
*/
void R_AddWorldSurfaces (void) {
if ( !r_drawworld->integer ) {
return;
}
if ( tr.refdef.rdflags & RDF_NOWORLDMODEL ) {
return;
}
tr.currentEntityNum = REFENTITYNUM_WORLD;
tr.shiftedEntityNum = tr.currentEntityNum << QSORT_REFENTITYNUM_SHIFT;
// determine which leaves are in the PVS / areamask
R_MarkLeaves ();
// clear out the visible min/max
ClearBounds( tr.viewParms.visBounds[0], tr.viewParms.visBounds[1] );
// perform frustum culling and add all the potentially visible surfaces
if ( tr.refdef.num_dlights > 32 ) {
tr.refdef.num_dlights = 32 ;
}
R_RecursiveWorldNode( tr.world->nodes, 15, ( 1 << tr.refdef.num_dlights ) - 1 );
}