lilium-voyager/code/renderergl2/tr_surface.c

1917 lines
51 KiB
C
Raw Normal View History

/*
===========================================================================
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
===========================================================================
*/
// tr_surf.c
#include "tr_local.h"
/*
THIS ENTIRE FILE IS BACK END
backEnd.currentEntity will be valid.
Tess_Begin has already been called for the surface's shader.
The modelview matrix will be set.
It is safe to actually issue drawing commands here if you don't want to
use the shader system.
*/
//============================================================================
/*
==============
RB_CheckOverflow
==============
*/
void RB_CheckOverflow( int verts, int indexes ) {
if (tess.numVertexes + verts < SHADER_MAX_VERTEXES
&& tess.numIndexes + indexes < SHADER_MAX_INDEXES) {
return;
}
RB_EndSurface();
if ( verts >= SHADER_MAX_VERTEXES ) {
ri.Error(ERR_DROP, "RB_CheckOverflow: verts > MAX (%d > %d)", verts, SHADER_MAX_VERTEXES );
}
if ( indexes >= SHADER_MAX_INDEXES ) {
ri.Error(ERR_DROP, "RB_CheckOverflow: indices > MAX (%d > %d)", indexes, SHADER_MAX_INDEXES );
}
2013-09-16 07:54:26 +00:00
RB_BeginSurface(tess.shader, tess.fogNum, tess.cubemapIndex );
}
2014-10-14 08:50:44 +00:00
void RB_CheckVao(vao_t *vao)
{
if (vao != glState.currentVao)
{
RB_EndSurface();
2013-09-16 07:54:26 +00:00
RB_BeginSurface(tess.shader, tess.fogNum, tess.cubemapIndex);
2014-10-14 08:50:44 +00:00
R_BindVao(vao);
}
2014-10-14 08:50:44 +00:00
if (vao != tess.vao)
tess.useInternalVao = qfalse;
}
/*
==============
RB_AddQuadStampExt
==============
*/
void RB_AddQuadStampExt( vec3_t origin, vec3_t left, vec3_t up, float color[4], float s1, float t1, float s2, float t2 ) {
vec3_t normal;
int16_t iNormal[4];
uint16_t iColor[4];
int ndx;
RB_CheckVao(tess.vao);
RB_CHECKOVERFLOW( 4, 6 );
ndx = tess.numVertexes;
// triangle indexes for a simple quad
tess.indexes[ tess.numIndexes ] = ndx;
tess.indexes[ tess.numIndexes + 1 ] = ndx + 1;
tess.indexes[ tess.numIndexes + 2 ] = ndx + 3;
tess.indexes[ tess.numIndexes + 3 ] = ndx + 3;
tess.indexes[ tess.numIndexes + 4 ] = ndx + 1;
tess.indexes[ tess.numIndexes + 5 ] = ndx + 2;
tess.xyz[ndx][0] = origin[0] + left[0] + up[0];
tess.xyz[ndx][1] = origin[1] + left[1] + up[1];
tess.xyz[ndx][2] = origin[2] + left[2] + up[2];
tess.xyz[ndx+1][0] = origin[0] - left[0] + up[0];
tess.xyz[ndx+1][1] = origin[1] - left[1] + up[1];
tess.xyz[ndx+1][2] = origin[2] - left[2] + up[2];
tess.xyz[ndx+2][0] = origin[0] - left[0] - up[0];
tess.xyz[ndx+2][1] = origin[1] - left[1] - up[1];
tess.xyz[ndx+2][2] = origin[2] - left[2] - up[2];
tess.xyz[ndx+3][0] = origin[0] + left[0] - up[0];
tess.xyz[ndx+3][1] = origin[1] + left[1] - up[1];
tess.xyz[ndx+3][2] = origin[2] + left[2] - up[2];
// constant normal all the way around
VectorSubtract( vec3_origin, backEnd.viewParms.or.axis[0], normal );
R_VaoPackNormal(iNormal, normal);
VectorCopy4(iNormal, tess.normal[ndx]);
VectorCopy4(iNormal, tess.normal[ndx + 1]);
VectorCopy4(iNormal, tess.normal[ndx + 2]);
VectorCopy4(iNormal, tess.normal[ndx + 3]);
// standard square texture coordinates
VectorSet2(tess.texCoords[ndx], s1, t1);
VectorSet2(tess.lightCoords[ndx], s1, t1);
VectorSet2(tess.texCoords[ndx+1], s2, t1);
VectorSet2(tess.lightCoords[ndx+1], s2, t1);
VectorSet2(tess.texCoords[ndx+2], s2, t2);
VectorSet2(tess.lightCoords[ndx+2], s2, t2);
VectorSet2(tess.texCoords[ndx+3], s1, t2);
VectorSet2(tess.lightCoords[ndx+3], s1, t2);
// constant color all the way around
// should this be identity and let the shader specify from entity?
R_VaoPackColor(iColor, color);
VectorCopy4(iColor, tess.color[ndx]);
VectorCopy4(iColor, tess.color[ndx + 1]);
VectorCopy4(iColor, tess.color[ndx + 2]);
VectorCopy4(iColor, tess.color[ndx + 3]);
tess.numVertexes += 4;
tess.numIndexes += 6;
}
/*
==============
RB_AddQuadStamp
==============
*/
void RB_AddQuadStamp( vec3_t origin, vec3_t left, vec3_t up, float color[4] ) {
RB_AddQuadStampExt( origin, left, up, color, 0, 0, 1, 1 );
}
/*
==============
RB_InstantQuad
based on Tess_InstantQuad from xreal
==============
*/
void RB_InstantQuad2(vec4_t quadVerts[4], vec2_t texCoords[4])
{
GLimp_LogComment("--- RB_InstantQuad2 ---\n");
tess.numVertexes = 0;
tess.numIndexes = 0;
tess.firstIndex = 0;
VectorCopy4(quadVerts[0], tess.xyz[tess.numVertexes]);
VectorCopy2(texCoords[0], tess.texCoords[tess.numVertexes]);
tess.numVertexes++;
VectorCopy4(quadVerts[1], tess.xyz[tess.numVertexes]);
VectorCopy2(texCoords[1], tess.texCoords[tess.numVertexes]);
tess.numVertexes++;
VectorCopy4(quadVerts[2], tess.xyz[tess.numVertexes]);
VectorCopy2(texCoords[2], tess.texCoords[tess.numVertexes]);
tess.numVertexes++;
VectorCopy4(quadVerts[3], tess.xyz[tess.numVertexes]);
VectorCopy2(texCoords[3], tess.texCoords[tess.numVertexes]);
tess.numVertexes++;
tess.indexes[tess.numIndexes++] = 0;
tess.indexes[tess.numIndexes++] = 1;
tess.indexes[tess.numIndexes++] = 2;
tess.indexes[tess.numIndexes++] = 0;
tess.indexes[tess.numIndexes++] = 2;
tess.indexes[tess.numIndexes++] = 3;
2014-10-14 08:50:44 +00:00
RB_UpdateTessVao(ATTR_POSITION | ATTR_TEXCOORD);
R_DrawElements(tess.numIndexes, tess.firstIndex);
tess.numIndexes = 0;
tess.numVertexes = 0;
tess.firstIndex = 0;
}
void RB_InstantQuad(vec4_t quadVerts[4])
{
vec2_t texCoords[4];
2013-03-06 10:02:01 +00:00
VectorSet2(texCoords[0], 0.0f, 0.0f);
VectorSet2(texCoords[1], 1.0f, 0.0f);
VectorSet2(texCoords[2], 1.0f, 1.0f);
VectorSet2(texCoords[3], 0.0f, 1.0f);
GLSL_BindProgram(&tr.textureColorShader);
GLSL_SetUniformMat4(&tr.textureColorShader, UNIFORM_MODELVIEWPROJECTIONMATRIX, glState.modelviewProjection);
GLSL_SetUniformVec4(&tr.textureColorShader, UNIFORM_COLOR, colorWhite);
2013-03-06 10:02:01 +00:00
RB_InstantQuad2(quadVerts, texCoords);
}
/*
==============
RB_SurfaceSprite
==============
*/
static void RB_SurfaceSprite( void ) {
vec3_t left, up;
float radius;
float rotation;
float colors[4];
trRefEntity_t *ent = backEnd.currentEntity;
// calculate the xyz locations for the four corners
#ifdef ELITEFORCE
radius = backEnd.currentEntity->e.data.sprite.radius;
rotation = backEnd.currentEntity->e.data.sprite.rotation;
#else
radius = backEnd.currentEntity->e.radius;
rotation = backEnd.currentEntity->e.rotation;
#endif
if ( rotation == 0 ) {
VectorScale( backEnd.viewParms.or.axis[1], radius, left );
VectorScale( backEnd.viewParms.or.axis[2], radius, up );
} else {
float s, c;
float ang;
ang = M_PI * rotation / 180;
s = sin( ang );
c = cos( ang );
VectorScale( backEnd.viewParms.or.axis[1], c * radius, left );
VectorMA( left, -s * radius, backEnd.viewParms.or.axis[2], left );
VectorScale( backEnd.viewParms.or.axis[2], c * radius, up );
VectorMA( up, s * radius, backEnd.viewParms.or.axis[1], up );
}
if ( backEnd.viewParms.isMirror ) {
VectorSubtract( vec3_origin, left, left );
}
VectorScale4(ent->e.shaderRGBA, 1.0f / 255.0f, colors);
RB_AddQuadStamp( ent->e.origin, left, up, colors );
}
/*
=============
RB_SurfacePolychain
=============
*/
static void RB_SurfacePolychain( srfPoly_t *p ) {
int i;
int numv;
RB_CheckVao(tess.vao);
RB_CHECKOVERFLOW( p->numVerts, 3*(p->numVerts - 2) );
// fan triangles into the tess array
numv = tess.numVertexes;
for ( i = 0; i < p->numVerts; i++ ) {
VectorCopy( p->verts[i].xyz, tess.xyz[numv] );
tess.texCoords[numv][0] = p->verts[i].st[0];
tess.texCoords[numv][1] = p->verts[i].st[1];
tess.color[numv][0] = (int)p->verts[i].modulate[0] * 257;
tess.color[numv][1] = (int)p->verts[i].modulate[1] * 257;
tess.color[numv][2] = (int)p->verts[i].modulate[2] * 257;
tess.color[numv][3] = (int)p->verts[i].modulate[3] * 257;
numv++;
}
// generate fan indexes into the tess array
for ( i = 0; i < p->numVerts-2; i++ ) {
tess.indexes[tess.numIndexes + 0] = tess.numVertexes;
tess.indexes[tess.numIndexes + 1] = tess.numVertexes + i + 1;
tess.indexes[tess.numIndexes + 2] = tess.numVertexes + i + 2;
tess.numIndexes += 3;
}
tess.numVertexes = numv;
}
static void RB_SurfaceVertsAndIndexes( int numVerts, srfVert_t *verts, int numIndexes, glIndex_t *indexes, int dlightBits, int pshadowBits)
{
2013-03-06 10:02:01 +00:00
int i;
glIndex_t *inIndex;
srfVert_t *dv;
float *xyz, *texCoords, *lightCoords;
int16_t *lightdir;
int16_t *normal;
int16_t *tangent;
glIndex_t *outIndex;
uint16_t *color;
2014-10-14 08:50:44 +00:00
RB_CheckVao(tess.vao);
RB_CHECKOVERFLOW( numVerts, numIndexes );
inIndex = indexes;
outIndex = &tess.indexes[ tess.numIndexes ];
for ( i = 0 ; i < numIndexes ; i++ ) {
*outIndex++ = tess.numVertexes + *inIndex++;
}
tess.numIndexes += numIndexes;
if ( tess.shader->vertexAttribs & ATTR_POSITION )
{
dv = verts;
xyz = tess.xyz[ tess.numVertexes ];
for ( i = 0 ; i < numVerts ; i++, dv++, xyz+=4 )
VectorCopy(dv->xyz, xyz);
}
if ( tess.shader->vertexAttribs & ATTR_NORMAL )
{
dv = verts;
normal = tess.normal[ tess.numVertexes ];
for ( i = 0 ; i < numVerts ; i++, dv++, normal+=4 )
VectorCopy4(dv->normal, normal);
}
if ( tess.shader->vertexAttribs & ATTR_TANGENT )
{
dv = verts;
tangent = tess.tangent[ tess.numVertexes ];
for ( i = 0 ; i < numVerts ; i++, dv++, tangent+=4 )
VectorCopy4(dv->tangent, tangent);
}
if ( tess.shader->vertexAttribs & ATTR_TEXCOORD )
{
dv = verts;
texCoords = tess.texCoords[tess.numVertexes];
for ( i = 0 ; i < numVerts ; i++, dv++, texCoords+=2 )
VectorCopy2(dv->st, texCoords);
}
if ( tess.shader->vertexAttribs & ATTR_LIGHTCOORD )
{
dv = verts;
lightCoords = tess.lightCoords[ tess.numVertexes ];
for ( i = 0 ; i < numVerts ; i++, dv++, lightCoords+=2 )
VectorCopy2(dv->lightmap, lightCoords);
}
if ( tess.shader->vertexAttribs & ATTR_COLOR )
{
dv = verts;
color = tess.color[ tess.numVertexes ];
for ( i = 0 ; i < numVerts ; i++, dv++, color+=4 )
VectorCopy4(dv->color, color);
}
if ( tess.shader->vertexAttribs & ATTR_LIGHTDIRECTION )
{
dv = verts;
lightdir = tess.lightdir[ tess.numVertexes ];
for ( i = 0 ; i < numVerts ; i++, dv++, lightdir+=4 )
VectorCopy4(dv->lightdir, lightdir);
}
#if 0 // nothing even uses vertex dlightbits
for ( i = 0 ; i < numVerts ; i++ ) {
tess.vertexDlightBits[ tess.numVertexes + i ] = dlightBits;
}
#endif
tess.dlightBits |= dlightBits;
tess.pshadowBits |= pshadowBits;
tess.numVertexes += numVerts;
}
static qboolean RB_SurfaceVaoCached(int numVerts, srfVert_t *verts, int numIndexes, glIndex_t *indexes, int dlightBits, int pshadowBits)
{
qboolean recycleVertexBuffer = qfalse;
qboolean recycleIndexBuffer = qfalse;
qboolean endSurface = qfalse;
if (!(!ShaderRequiresCPUDeforms(tess.shader) && !tess.shader->isSky && !tess.shader->isPortal))
return qfalse;
if (!numIndexes || !numVerts)
return qfalse;
VaoCache_BindVao();
tess.dlightBits |= dlightBits;
tess.pshadowBits |= pshadowBits;
VaoCache_CheckAdd(&endSurface, &recycleVertexBuffer, &recycleIndexBuffer, numVerts, numIndexes);
if (endSurface)
{
RB_EndSurface();
RB_BeginSurface(tess.shader, tess.fogNum, tess.cubemapIndex);
}
if (recycleVertexBuffer)
VaoCache_RecycleVertexBuffer();
if (recycleIndexBuffer)
VaoCache_RecycleIndexBuffer();
if (!tess.numVertexes)
VaoCache_InitQueue();
VaoCache_AddSurface(verts, numVerts, indexes, numIndexes);
tess.numIndexes += numIndexes;
tess.numVertexes += numVerts;
tess.useInternalVao = qfalse;
tess.useCacheVao = qtrue;
return qtrue;
}
/*
=============
RB_SurfaceTriangles
=============
*/
static void RB_SurfaceTriangles( srfBspSurface_t *srf ) {
if (RB_SurfaceVaoCached(srf->numVerts, srf->verts, srf->numIndexes,
srf->indexes, srf->dlightBits, srf->pshadowBits))
{
return;
}
RB_SurfaceVertsAndIndexes(srf->numVerts, srf->verts, srf->numIndexes,
srf->indexes, srf->dlightBits, srf->pshadowBits);
}
#ifdef ELITEFORCE
/*
==============
RB_SurfaceOrientedSprite
==============
*/
static void RB_SurfaceOrientedSprite( void )
{
vec3_t left, up;
float radius;
float rotation;
float colors[4];
trRefEntity_t *ent = backEnd.currentEntity;
// calculate the xyz locations for the four corners
radius = backEnd.currentEntity->e.data.sprite.radius;
rotation = backEnd.currentEntity->e.data.sprite.rotation;
if (rotation == 0)
{
VectorScale( backEnd.currentEntity->e.axis[1], radius, left );
VectorScale( backEnd.currentEntity->e.axis[2], radius, up );
}
else
{
float s, c;
float ang;
ang = M_PI * rotation / 180;
s = sin( ang );
c = cos( ang );
VectorScale( backEnd.currentEntity->e.axis[1], c * radius, left );
VectorMA( left, -s * radius, backEnd.currentEntity->e.axis[2], left );
VectorScale( backEnd.currentEntity->e.axis[2], c * radius, up );
VectorMA( up, s * radius, backEnd.currentEntity->e.axis[1], up );
}
if ( backEnd.viewParms.isMirror )
VectorSubtract( vec3_origin, left, left );
VectorScale4(ent->e.shaderRGBA, 1.0f / 255.0f, colors);
RB_AddQuadStamp( ent->e.origin, left, up, colors );
}
void RB_Line(vec3_t start, vec3_t end, vec3_t linedirection, vec3_t left,
vec3_t *corners, float starttex, float endtex, refEntity_t *e)
{
int ndx, numind;
RB_CHECKOVERFLOW( 4, 6 );
// Set up the triangles ..
ndx = tess.numVertexes;
numind = tess.numIndexes;
tess.indexes[ numind ] = ndx;
tess.indexes[ numind + 1 ] = ndx + 1;
tess.indexes[ numind + 2 ] = ndx + 3;
tess.indexes[ numind + 3 ] = ndx + 3;
tess.indexes[ numind + 4 ] = ndx + 1;
tess.indexes[ numind + 5 ] = ndx + 2;
// now create the corner vertices
if(corners)
{
// we have the corner points for the start already given.
VectorCopy(corners[0], tess.xyz[ndx]);
VectorCopy(corners[1], tess.xyz[ndx+1]);
}
else
{
// start left corner
VectorAdd(start, left, tess.xyz[ndx]);
// start right corner
VectorSubtract(start, left, tess.xyz[ndx+1]);
}
// end right corner
VectorSubtract(end, left, tess.xyz[ndx+2]);
// end left corner
VectorAdd(end, left, tess.xyz[ndx+3]);
if(corners)
{
// save the end corner points here.
VectorCopy(tess.xyz[ndx+3], corners[0]);
VectorCopy(tess.xyz[ndx+2], corners[1]);
}
// Texture stuff....
tess.texCoords[ndx][0] = 0;
tess.texCoords[ndx][1] = starttex;
tess.texCoords[ndx+1][0] = 1;
tess.texCoords[ndx+1][1] = starttex;
tess.texCoords[ndx+2][0] = 1;
tess.texCoords[ndx+2][1] = endtex;
tess.texCoords[ndx+3][0] = 0;
tess.texCoords[ndx+3][1] = endtex;
tess.color[ndx][0] = tess.color[ndx+1][0] = tess.color[ndx+2][0] = tess.color[ndx+3][0] = e->shaderRGBA[0] * 257;
tess.color[ndx][1] = tess.color[ndx+1][1] = tess.color[ndx+2][1] = tess.color[ndx+3][1] = e->shaderRGBA[1] * 257;
tess.color[ndx][2] = tess.color[ndx+1][2] = tess.color[ndx+2][2] = tess.color[ndx+3][2] = e->shaderRGBA[2] * 257;
tess.color[ndx][3] = tess.color[ndx+1][3] = tess.color[ndx+2][3] = tess.color[ndx+3][3] = e->shaderRGBA[3] * 257;
tess.numVertexes += 4;
tess.numIndexes += 6;
}
// Create a normal vector and scale it to the right length
void RB_LineNormal(vec3_t vec1, vec3_t vec2, float scale, vec3_t result)
{
// Create the offset vector for the width of the line
CrossProduct(vec1, vec2, result);
// Normalize the offset vector first.
VectorNormalize(result);
// Scale the offset vector to the intended width.
VectorScale(result, scale / 2, result);
}
// Let's draw a thick line from A to B and dance happily around a christmas tree!
void RB_SurfaceLine( void )
{
refEntity_t *e;
vec3_t left; // I vote the green party...
vec3_t start, end, linedirection, start2origin;
shader_t *surfshader;
e = &backEnd.currentEntity->e;
// Getting up before 1:00 pm to attend HM I lectures finally paid off..
// Get the start and end point of the line
VectorCopy(e->origin, start);
VectorCopy(e->oldorigin, end);
// Direction vector for the line:
VectorSubtract(end, start, linedirection);
// Direction vector for the start to current point of view
VectorSubtract(backEnd.viewParms.or.origin, start, start2origin);
RB_LineNormal(start2origin, linedirection, e->data.line.width, left);
RB_Line(start, end, linedirection, left, NULL, 0, e->data.line.stscale, e);
// Hack to make the dreadnought lightning bolt work: set the correct cull type...
if((surfshader = R_GetShaderByHandle(e->customShader)))
surfshader->cullType = CT_TWO_SIDED;
}
void RB_SurfaceOrientedLine(void)
{
refEntity_t *e;
vec3_t left;
vec3_t linedirection;
shader_t *surfshader;
e = &backEnd.currentEntity->e;
VectorSubtract(e->oldorigin, e->origin, linedirection);
VectorCopy(e->axis[1], left);
VectorNormalize(left);
VectorScale(left, e->data.line.width / 2, left);
RB_Line(e->origin, e->oldorigin, linedirection, left, NULL, 0, 1, e);
surfshader = R_GetShaderByHandle(e->customShader);
surfshader->cullType = CT_TWO_SIDED;
}
// This time it's not a rectangle but a trapezoid I guess ...
void RB_SurfaceLine2( void )
{
refEntity_t *e;
vec3_t startleft, endleft; // I still vote the green party...
vec3_t start, end, linedirection, start2origin;
int ndx, numind;
RB_CHECKOVERFLOW( 6, 12 );
e = &backEnd.currentEntity->e;
// Set up the triangle ..
ndx = tess.numVertexes;
numind = tess.numIndexes;
tess.indexes[ numind ] = ndx;
tess.indexes[ numind + 1 ] = ndx + 1;
tess.indexes[ numind + 2 ] = ndx + 5;
tess.indexes[ numind + 3 ] = ndx + 5;
tess.indexes[ numind + 4 ] = ndx + 4;
tess.indexes[ numind + 5 ] = ndx + 1;
tess.indexes[ numind + 6 ] = ndx + 1;
tess.indexes[ numind + 7 ] = ndx + 4;
tess.indexes[ numind + 8 ] = ndx + 3;
tess.indexes[ numind + 9 ] = ndx + 3;
tess.indexes[ numind + 10 ] = ndx + 2;
tess.indexes[ numind + 11 ] = ndx + 1;
// Get the start and end point of the line
VectorCopy(e->origin, start);
VectorCopy(e->oldorigin, end);
// Direction vector for the line:
VectorSubtract(end, start, linedirection);
// Direction vector for the start to current point of view
VectorSubtract(backEnd.viewParms.or.origin, start, start2origin);
// Create the offset vector for the width of the line
CrossProduct(start2origin, linedirection, startleft);
// Normalize the offset vector first.
VectorNormalize(startleft);
// The normal of the triangle we just thought up:
// Scale the offset vector to the intended width.
VectorScale(startleft, e->data.line.width2 / 2, endleft);
VectorScale(startleft, e->data.line.width / 2, startleft);
// now create the corner vertices
// start left corner
VectorAdd(start, startleft, tess.xyz[ndx]);
// start point
VectorCopy(start, tess.xyz[ndx+1]);
// start right corner
VectorSubtract(start, startleft, tess.xyz[ndx+2]);
// end right corner
VectorSubtract(end, endleft, tess.xyz[ndx+3]);
// end point
VectorCopy(end, tess.xyz[ndx+4]);
// end left corner
VectorAdd(end, endleft, tess.xyz[ndx+5]);
// Texture stuff....
tess.texCoords[ndx][0] = 0;
tess.texCoords[ndx][1] = 0;
tess.texCoords[ndx+1][0] = 0.5;
tess.texCoords[ndx+1][1] = 0;
tess.texCoords[ndx+2][0] = 1;
tess.texCoords[ndx+2][1] = 0;
tess.texCoords[ndx+3][0] = 1;
tess.texCoords[ndx+3][1] = 1;
tess.texCoords[ndx+4][0] = 0.5;
tess.texCoords[ndx+4][1] = 1;
tess.texCoords[ndx+5][0] = 0;
tess.texCoords[ndx+5][1] = 1;
tess.color[ndx][0] = tess.color[ndx+1][0] = tess.color[ndx+2][0] =
tess.color[ndx+3][0] = tess.color[ndx+4][0] = tess.color[ndx+5][0] = e->shaderRGBA[0] * 257;
tess.color[ndx][1] = tess.color[ndx+1][1] = tess.color[ndx+2][1] =
tess.color[ndx+3][1] = tess.color[ndx+4][1] = tess.color[ndx+5][1] = e->shaderRGBA[1] * 257;
tess.color[ndx][2] = tess.color[ndx+1][2] = tess.color[ndx+2][2] =
tess.color[ndx+3][2] = tess.color[ndx+4][2] = tess.color[ndx+5][2] = e->shaderRGBA[2] * 257;
tess.color[ndx][3] = tess.color[ndx+1][3] = tess.color[ndx+2][3] =
tess.color[ndx+3][3] = tess.color[ndx+4][3] = tess.color[ndx+5][3] = e->shaderRGBA[3] * 257;
tess.numVertexes += 6;
tess.numIndexes += 12;
}
// Draw a cubic bezier curve for the imod weapon.
#define BEZIER_SEGMENTS 32
#define BEZIER_STEPSIZE 1.0 / BEZIER_SEGMENTS
void RB_SurfaceBezier(void)
{
refEntity_t *e = &backEnd.currentEntity->e;
double tvar = 0, one_tvar; // use double to not lose as much precision on cubing and squaring.
float starttc, endtc;
vec3_t segstartpos, segendpos, prevend[2];
vec3_t linedirection, start2origin, left;
int index;
VectorCopy(e->origin, segstartpos);
// start of the bezier curve is pointy
VectorCopy(segstartpos, prevend[0]);
VectorCopy(segstartpos, prevend[1]);
// iterate through all segments we have to create.
while(tvar <= 1.0)
{
// get the texture position for the rectangular approximating the bezier curve.
starttc = tvar / 2.0;
tvar += BEZIER_STEPSIZE;
endtc = tvar / 2.0;
one_tvar = 1.0 - tvar;
// get the endpoint for this segment.
for(index = 0; index < 3; index++)
{
// C(t) = P0 (t-1)^3 + 3 P1 t (t-1)^2 + 3 P2 t^2 (t-1) + P3 t^3
segendpos[index] =
e->origin[index] * one_tvar * one_tvar * one_tvar +
3 * e->data.bezier.control1[index] * tvar * one_tvar * one_tvar +
3 * e->data.bezier.control2[index] * tvar * tvar * one_tvar +
e->oldorigin[index] * tvar * tvar * tvar;
}
// Direction vector for the line:
VectorSubtract(segendpos, segstartpos, linedirection);
// Direction vector for the start to current point of view
VectorSubtract(backEnd.viewParms.or.origin, segstartpos, start2origin);
RB_LineNormal(start2origin, linedirection, e->data.bezier.width, left);
RB_Line(segstartpos, segendpos, linedirection, left, prevend, starttc, endtc, e);
// Our next segment starts where the old one ends.
VectorCopy(segendpos, segstartpos);
}
}
// this actually isn't a cylinder but a frustum, but for the sake of naming
// conventions we'll keep calling it a cylinder.
// number of rectangles we'll use to approximate a frustum of a cone.
#define CYLINDER_MAXPLANES 64
#define LOD_CONSTANT 4.0f
void RB_SurfaceCylinder(void)
{
int planes, index = 0; // calculate the number of planes based on the level of detail.
vec3_t bottom, top, bottom2top, zerodeg_bottom, zerodeg_top, faxis;
vec3_t bottomcorner, topcorner, vieworigin;
float anglestep, tcstep = 0, width, width2, height;
int ndx, numind;
refEntity_t *e = &backEnd.currentEntity->e;
// for LOD calculation:
vec3_t bottom2origin, top2origin, lodcalc, projection;
float maxradius, startdistance, enddistance, distance;
height = e->data.cylinder.height;
width = e->data.cylinder.width;
width2 = e->data.cylinder.width2;
maxradius = (width > width2) ? width : width2;
// get the start and end point of the frustum:
VectorCopy(backEnd.viewParms.or.origin, vieworigin);
VectorCopy(e->axis[0], faxis);
VectorCopy(e->origin, bottom);
VectorScale(faxis, height, bottom2top);
VectorAdd(bottom, bottom2top, top);
// Get distance from frustum:
// first distance from endpoints
VectorSubtract(vieworigin, bottom, bottom2origin);
VectorSubtract(vieworigin, top, top2origin);
// get projection of view origin on the middle line:
distance = DotProduct(bottom2origin, faxis);
startdistance = VectorLength(bottom2origin);
enddistance = VectorLength(top2origin);
if(distance < 0 || distance > height)
{
// projected point is not between bottom and top.
distance = (startdistance < enddistance) ? startdistance : enddistance;
}
else
{
VectorMA(bottom, distance, faxis, projection);
VectorSubtract(vieworigin, projection, lodcalc);
distance = VectorLength(lodcalc);
if(startdistance < distance)
distance = startdistance;
if(enddistance < distance)
distance = enddistance;
}
planes = LOD_CONSTANT * maxradius/distance * CYLINDER_MAXPLANES;
if(planes < 8)
planes = 8;
else if(planes > CYLINDER_MAXPLANES)
planes = CYLINDER_MAXPLANES;
RB_CHECKOVERFLOW(4 * planes, 6 * planes);
// create a normalized perpendicular vector to the frustum height.
PerpendicularVector(zerodeg_bottom, faxis);
// This vector gets rotated to create the top ring
VectorScale(zerodeg_bottom, width2 / 2.0f, zerodeg_top);
// Likewise the vector for the lower ring:
VectorScale(zerodeg_bottom, width / 2.0f, zerodeg_bottom);
anglestep = 360.0f / planes;
if(e->data.cylinder.wrap)
tcstep = e->data.cylinder.stscale / planes;
ndx = tess.numVertexes;
numind = tess.numIndexes;
// this loop is creating all surface planes.
while(index < planes)
{
index++;
// Set up the indexes for the new plane:
tess.indexes[numind++] = ndx;
tess.indexes[numind++] = ndx + 2;
tess.indexes[numind++] = ndx + 3;
tess.indexes[numind++] = ndx;
tess.indexes[numind++] = ndx + 3;
tess.indexes[numind++] = ndx + 1;
if(index <= 1)
{
VectorAdd(bottom, zerodeg_bottom, tess.xyz[ndx]);
VectorAdd(top, zerodeg_top, tess.xyz[ndx + 1]);
}
else
{
VectorCopy(bottomcorner, tess.xyz[ndx]);
VectorCopy(topcorner, tess.xyz[ndx + 1]);
}
// Create on the right side two new vertices, first the bottom, then the top one.
if(index >= planes)
{
VectorAdd(bottom, zerodeg_bottom, bottomcorner);
VectorAdd(top, zerodeg_top, topcorner);
}
else
{
RotatePointAroundVector(bottomcorner, faxis, zerodeg_bottom, anglestep * index);
VectorAdd(bottom, bottomcorner, bottomcorner);
RotatePointAroundVector(topcorner, faxis, zerodeg_top, index * anglestep);
VectorAdd(top, topcorner, topcorner);
}
VectorCopy(bottomcorner, tess.xyz[ndx + 2]);
VectorCopy(topcorner, tess.xyz[ndx + 3]);
// Take care about the texture stuff
if(e->data.cylinder.wrap)
{
tess.texCoords[ndx][0] = tcstep * (index - 1);
tess.texCoords[ndx][1] = 0;
tess.texCoords[ndx+1][0] = tcstep * (index - 1);
tess.texCoords[ndx+1][1] = 1;
tess.texCoords[ndx+2][0] = tcstep * index;
tess.texCoords[ndx+2][1] = 0;
tess.texCoords[ndx+3][0] = tcstep * index;
tess.texCoords[ndx+3][1] = 1;
}
else
{
tess.texCoords[ndx][0] = 0;
tess.texCoords[ndx][1] = 0;
tess.texCoords[ndx+1][0] = 0;
tess.texCoords[ndx+1][1] = 1;
tess.texCoords[ndx+2][0] = e->data.cylinder.stscale;
tess.texCoords[ndx+2][1] = 0;
tess.texCoords[ndx+3][0] = e->data.cylinder.stscale;
tess.texCoords[ndx+3][1] = 1;
}
tess.color[ndx][0] = tess.color[ndx+1][0] = tess.color[ndx+2][0] = tess.color[ndx+3][0] = e->shaderRGBA[0] * 257;
tess.color[ndx][1] = tess.color[ndx+1][1] = tess.color[ndx+2][1] = tess.color[ndx+3][1] = e->shaderRGBA[1] * 257;
tess.color[ndx][2] = tess.color[ndx+1][2] = tess.color[ndx+2][2] = tess.color[ndx+3][2] = e->shaderRGBA[2] * 257;
tess.color[ndx][3] = tess.color[ndx+1][3] = tess.color[ndx+2][3] = tess.color[ndx+3][3] = 255 * 257;
ndx += 4;
}
tess.numVertexes = ndx;
tess.numIndexes = numind;
}
// Draws lightning using a multisegment line with jumpy connecting points.
#define ELECTRICITY_SEGMENTS 16
#define ELECTRICITY_STEPSIZE 1.0 / ELECTRICITY_SEGMENTS
void RB_SurfaceElectricity(void)
{
vec3_t start, end, linedirection, segstartpos, segendpos;
vec3_t prevend[2], left, start2origin, start2end;
refEntity_t *e = &backEnd.currentEntity->e;
float width, deviation, veclen;
int index;
VectorCopy(e->origin, start);
VectorCopy(e->oldorigin, end);
VectorSubtract(end, start, start2end);
veclen = VectorLength(start2end);
width = e->data.electricity.width;
deviation = e->data.electricity.deviation;
VectorCopy(start, segendpos);
for(index = 0; index < ELECTRICITY_SEGMENTS; index++)
{
VectorCopy(segendpos, segstartpos);
if(index > ELECTRICITY_SEGMENTS - 2)
{
// This is the last segment.
VectorCopy(end, segendpos);
}
else
{
if(index > ELECTRICITY_SEGMENTS - 3)
{
// The line should have somewhat deviated by now.
// Make the last two segments go to the endpoint
// get the middle point between startpos and endpos
VectorAdd(segstartpos, end, segendpos);
VectorScale(segendpos, 0.5f, segendpos);
}
// Slightly misplace the next point on the line.
segendpos[PITCH] += (start2end[PITCH] + crandom() * deviation * veclen) * ELECTRICITY_STEPSIZE;
segendpos[YAW] += (start2end[YAW] + crandom() * deviation * veclen) * ELECTRICITY_STEPSIZE;
segendpos[ROLL] += (start2end[ROLL] + crandom() * deviation * veclen / 2.0) * ELECTRICITY_STEPSIZE;
}
// Direction vector for the line:
VectorSubtract(segendpos, segstartpos, linedirection);
// Direction vector for the start to current point of view
VectorSubtract(backEnd.viewParms.or.origin, segstartpos, start2origin);
RB_LineNormal(start2origin, linedirection, width, left);
if(!index)
{
// this is our first segment, create the starting points
VectorAdd(segstartpos, left, prevend[0]);
VectorSubtract(segstartpos, left, prevend[1]);
}
RB_Line(segstartpos, segendpos, linedirection, left, prevend, 0, e->data.electricity.stscale, e);
}
}
#endif
/*
==============
RB_SurfaceBeam
==============
*/
static void RB_SurfaceBeam( void )
{
#define NUM_BEAM_SEGS 6
refEntity_t *e;
2013-03-06 10:02:01 +00:00
shaderProgram_t *sp = &tr.textureColorShader;
int i;
vec3_t perpvec;
vec3_t direction, normalized_direction;
vec3_t start_points[NUM_BEAM_SEGS], end_points[NUM_BEAM_SEGS];
vec3_t oldorigin, origin;
e = &backEnd.currentEntity->e;
oldorigin[0] = e->oldorigin[0];
oldorigin[1] = e->oldorigin[1];
oldorigin[2] = e->oldorigin[2];
origin[0] = e->origin[0];
origin[1] = e->origin[1];
origin[2] = e->origin[2];
normalized_direction[0] = direction[0] = oldorigin[0] - origin[0];
normalized_direction[1] = direction[1] = oldorigin[1] - origin[1];
normalized_direction[2] = direction[2] = oldorigin[2] - origin[2];
if ( VectorNormalize( normalized_direction ) == 0 )
return;
PerpendicularVector( perpvec, normalized_direction );
VectorScale( perpvec, 4, perpvec );
for ( i = 0; i < NUM_BEAM_SEGS ; i++ )
{
RotatePointAroundVector( start_points[i], normalized_direction, perpvec, (360.0/NUM_BEAM_SEGS)*i );
// VectorAdd( start_points[i], origin, start_points[i] );
VectorAdd( start_points[i], direction, end_points[i] );
}
GL_BindToTMU( tr.whiteImage, TB_COLORMAP );
GL_State( GLS_SRCBLEND_ONE | GLS_DSTBLEND_ONE );
// FIXME: Quake3 doesn't use this, so I never tested it
tess.numVertexes = 0;
tess.numIndexes = 0;
tess.firstIndex = 0;
for ( i = 0; i <= NUM_BEAM_SEGS; i++ ) {
VectorCopy(start_points[ i % NUM_BEAM_SEGS ], tess.xyz[tess.numVertexes++]);
VectorCopy(end_points [ i % NUM_BEAM_SEGS ], tess.xyz[tess.numVertexes++]);
}
for ( i = 0; i < NUM_BEAM_SEGS; i++ ) {
tess.indexes[tess.numIndexes++] = i * 2;
tess.indexes[tess.numIndexes++] = (i + 1) * 2;
tess.indexes[tess.numIndexes++] = 1 + i * 2;
tess.indexes[tess.numIndexes++] = 1 + i * 2;
tess.indexes[tess.numIndexes++] = (i + 1) * 2;
tess.indexes[tess.numIndexes++] = 1 + (i + 1) * 2;
}
// FIXME: A lot of this can probably be removed for speed, and refactored into a more convenient function
2014-10-14 08:50:44 +00:00
RB_UpdateTessVao(ATTR_POSITION);
2013-03-06 10:02:01 +00:00
GLSL_BindProgram(sp);
GLSL_SetUniformMat4(sp, UNIFORM_MODELVIEWPROJECTIONMATRIX, glState.modelviewProjection);
GLSL_SetUniformVec4(sp, UNIFORM_COLOR, colorRed);
GLSL_SetUniformInt(sp, UNIFORM_ALPHATEST, 0);
R_DrawElements(tess.numIndexes, tess.firstIndex);
tess.numIndexes = 0;
tess.numVertexes = 0;
tess.firstIndex = 0;
}
//================================================================================
static void DoRailCore( const vec3_t start, const vec3_t end, const vec3_t up, float len, float spanWidth )
{
float spanWidth2;
int vbase;
float t = len / 256.0f;
RB_CheckVao(tess.vao);
RB_CHECKOVERFLOW( 4, 6 );
vbase = tess.numVertexes;
spanWidth2 = -spanWidth;
// FIXME: use quad stamp?
VectorMA( start, spanWidth, up, tess.xyz[tess.numVertexes] );
tess.texCoords[tess.numVertexes][0] = 0;
tess.texCoords[tess.numVertexes][1] = 0;
tess.color[tess.numVertexes][0] = backEnd.currentEntity->e.shaderRGBA[0] * 0.25f * 257.0f;
tess.color[tess.numVertexes][1] = backEnd.currentEntity->e.shaderRGBA[1] * 0.25f * 257.0f;
tess.color[tess.numVertexes][2] = backEnd.currentEntity->e.shaderRGBA[2] * 0.25f * 257.0f;
tess.numVertexes++;
VectorMA( start, spanWidth2, up, tess.xyz[tess.numVertexes] );
tess.texCoords[tess.numVertexes][0] = 0;
tess.texCoords[tess.numVertexes][1] = 1;
tess.color[tess.numVertexes][0] = backEnd.currentEntity->e.shaderRGBA[0] * 257;
tess.color[tess.numVertexes][1] = backEnd.currentEntity->e.shaderRGBA[1] * 257;
tess.color[tess.numVertexes][2] = backEnd.currentEntity->e.shaderRGBA[2] * 257;
tess.numVertexes++;
VectorMA( end, spanWidth, up, tess.xyz[tess.numVertexes] );
tess.texCoords[tess.numVertexes][0] = t;
tess.texCoords[tess.numVertexes][1] = 0;
tess.color[tess.numVertexes][0] = backEnd.currentEntity->e.shaderRGBA[0] * 257;
tess.color[tess.numVertexes][1] = backEnd.currentEntity->e.shaderRGBA[1] * 257;
tess.color[tess.numVertexes][2] = backEnd.currentEntity->e.shaderRGBA[2] * 257;
tess.numVertexes++;
VectorMA( end, spanWidth2, up, tess.xyz[tess.numVertexes] );
tess.texCoords[tess.numVertexes][0] = t;
tess.texCoords[tess.numVertexes][1] = 1;
tess.color[tess.numVertexes][0] = backEnd.currentEntity->e.shaderRGBA[0] * 257;
tess.color[tess.numVertexes][1] = backEnd.currentEntity->e.shaderRGBA[1] * 257;
tess.color[tess.numVertexes][2] = backEnd.currentEntity->e.shaderRGBA[2] * 257;
tess.numVertexes++;
tess.indexes[tess.numIndexes++] = vbase;
tess.indexes[tess.numIndexes++] = vbase + 1;
tess.indexes[tess.numIndexes++] = vbase + 2;
tess.indexes[tess.numIndexes++] = vbase + 2;
tess.indexes[tess.numIndexes++] = vbase + 1;
tess.indexes[tess.numIndexes++] = vbase + 3;
}
static void DoRailDiscs( int numSegs, const vec3_t start, const vec3_t dir, const vec3_t right, const vec3_t up )
{
int i;
vec3_t pos[4];
vec3_t v;
int spanWidth = r_railWidth->integer;
float c, s;
float scale;
if ( numSegs > 1 )
numSegs--;
if ( !numSegs )
return;
scale = 0.25;
for ( i = 0; i < 4; i++ )
{
c = cos( DEG2RAD( 45 + i * 90 ) );
s = sin( DEG2RAD( 45 + i * 90 ) );
v[0] = ( right[0] * c + up[0] * s ) * scale * spanWidth;
v[1] = ( right[1] * c + up[1] * s ) * scale * spanWidth;
v[2] = ( right[2] * c + up[2] * s ) * scale * spanWidth;
VectorAdd( start, v, pos[i] );
if ( numSegs > 1 )
{
// offset by 1 segment if we're doing a long distance shot
VectorAdd( pos[i], dir, pos[i] );
}
}
RB_CheckVao(tess.vao);
for ( i = 0; i < numSegs; i++ )
{
int j;
RB_CHECKOVERFLOW( 4, 6 );
for ( j = 0; j < 4; j++ )
{
VectorCopy( pos[j], tess.xyz[tess.numVertexes] );
tess.texCoords[tess.numVertexes][0] = (j < 2);
tess.texCoords[tess.numVertexes][1] = (j && j != 3);
tess.color[tess.numVertexes][0] = backEnd.currentEntity->e.shaderRGBA[0] * 257;
tess.color[tess.numVertexes][1] = backEnd.currentEntity->e.shaderRGBA[1] * 257;
tess.color[tess.numVertexes][2] = backEnd.currentEntity->e.shaderRGBA[2] * 257;
tess.numVertexes++;
VectorAdd( pos[j], dir, pos[j] );
}
tess.indexes[tess.numIndexes++] = tess.numVertexes - 4 + 0;
tess.indexes[tess.numIndexes++] = tess.numVertexes - 4 + 1;
tess.indexes[tess.numIndexes++] = tess.numVertexes - 4 + 3;
tess.indexes[tess.numIndexes++] = tess.numVertexes - 4 + 3;
tess.indexes[tess.numIndexes++] = tess.numVertexes - 4 + 1;
tess.indexes[tess.numIndexes++] = tess.numVertexes - 4 + 2;
}
}
/*
** RB_SurfaceRailRinges
*/
static void RB_SurfaceRailRings( void ) {
refEntity_t *e;
int numSegs;
int len;
vec3_t vec;
vec3_t right, up;
vec3_t start, end;
e = &backEnd.currentEntity->e;
VectorCopy( e->oldorigin, start );
VectorCopy( e->origin, end );
// compute variables
VectorSubtract( end, start, vec );
len = VectorNormalize( vec );
MakeNormalVectors( vec, right, up );
numSegs = ( len ) / r_railSegmentLength->value;
if ( numSegs <= 0 ) {
numSegs = 1;
}
VectorScale( vec, r_railSegmentLength->value, vec );
DoRailDiscs( numSegs, start, vec, right, up );
}
/*
** RB_SurfaceRailCore
*/
static void RB_SurfaceRailCore( void ) {
refEntity_t *e;
int len;
vec3_t right;
vec3_t vec;
vec3_t start, end;
vec3_t v1, v2;
e = &backEnd.currentEntity->e;
VectorCopy( e->oldorigin, start );
VectorCopy( e->origin, end );
VectorSubtract( end, start, vec );
len = VectorNormalize( vec );
// compute side vector
VectorSubtract( start, backEnd.viewParms.or.origin, v1 );
VectorNormalize( v1 );
VectorSubtract( end, backEnd.viewParms.or.origin, v2 );
VectorNormalize( v2 );
CrossProduct( v1, v2, right );
VectorNormalize( right );
DoRailCore( start, end, right, len, r_railCoreWidth->integer );
}
/*
** RB_SurfaceLightningBolt
*/
static void RB_SurfaceLightningBolt( void ) {
refEntity_t *e;
int len;
vec3_t right;
vec3_t vec;
vec3_t start, end;
vec3_t v1, v2;
int i;
e = &backEnd.currentEntity->e;
VectorCopy( e->oldorigin, end );
VectorCopy( e->origin, start );
// compute variables
VectorSubtract( end, start, vec );
len = VectorNormalize( vec );
// compute side vector
VectorSubtract( start, backEnd.viewParms.or.origin, v1 );
VectorNormalize( v1 );
VectorSubtract( end, backEnd.viewParms.or.origin, v2 );
VectorNormalize( v2 );
CrossProduct( v1, v2, right );
VectorNormalize( right );
for ( i = 0 ; i < 4 ; i++ ) {
vec3_t temp;
DoRailCore( start, end, right, len, 8 );
RotatePointAroundVector( temp, vec, right, 45 );
VectorCopy( temp, right );
}
}
static void LerpMeshVertexes(mdvSurface_t *surf, float backlerp)
{
float *outXyz;
int16_t *outNormal, *outTangent;
mdvVertex_t *newVerts;
int vertNum;
newVerts = surf->verts + backEnd.currentEntity->e.frame * surf->numVerts;
outXyz = tess.xyz[tess.numVertexes];
outNormal = tess.normal[tess.numVertexes];
outTangent = tess.tangent[tess.numVertexes];
if (backlerp == 0)
{
//
// just copy the vertexes
//
for (vertNum=0 ; vertNum < surf->numVerts ; vertNum++)
{
VectorCopy(newVerts->xyz, outXyz);
VectorCopy4(newVerts->normal, outNormal);
VectorCopy4(newVerts->tangent, outTangent);
newVerts++;
outXyz += 4;
outNormal += 4;
outTangent += 4;
}
}
else
{
//
// interpolate and copy the vertex and normal
//
mdvVertex_t *oldVerts;
oldVerts = surf->verts + backEnd.currentEntity->e.oldframe * surf->numVerts;
for (vertNum=0 ; vertNum < surf->numVerts ; vertNum++)
{
VectorLerp(newVerts->xyz, oldVerts->xyz, backlerp, outXyz);
outNormal[0] = (int16_t)(newVerts->normal[0] * (1.0f - backlerp) + oldVerts->normal[0] * backlerp);
outNormal[1] = (int16_t)(newVerts->normal[1] * (1.0f - backlerp) + oldVerts->normal[1] * backlerp);
outNormal[2] = (int16_t)(newVerts->normal[2] * (1.0f - backlerp) + oldVerts->normal[2] * backlerp);
outNormal[3] = 0;
outTangent[0] = (int16_t)(newVerts->tangent[0] * (1.0f - backlerp) + oldVerts->tangent[0] * backlerp);
outTangent[1] = (int16_t)(newVerts->tangent[1] * (1.0f - backlerp) + oldVerts->tangent[1] * backlerp);
outTangent[2] = (int16_t)(newVerts->tangent[2] * (1.0f - backlerp) + oldVerts->tangent[2] * backlerp);
outTangent[3] = newVerts->tangent[3];
newVerts++;
oldVerts++;
outXyz += 4;
outNormal += 4;
outTangent += 4;
}
}
}
/*
=============
RB_SurfaceMesh
=============
*/
static void RB_SurfaceMesh(mdvSurface_t *surface) {
int j;
float backlerp;
mdvSt_t *texCoords;
int Bob, Doug;
int numVerts;
if ( backEnd.currentEntity->e.oldframe == backEnd.currentEntity->e.frame ) {
backlerp = 0;
} else {
backlerp = backEnd.currentEntity->e.backlerp;
}
RB_CheckVao(tess.vao);
RB_CHECKOVERFLOW( surface->numVerts, surface->numIndexes );
LerpMeshVertexes (surface, backlerp);
Bob = tess.numIndexes;
Doug = tess.numVertexes;
for (j = 0 ; j < surface->numIndexes ; j++) {
tess.indexes[Bob + j] = Doug + surface->indexes[j];
}
tess.numIndexes += surface->numIndexes;
texCoords = surface->st;
numVerts = surface->numVerts;
for ( j = 0; j < numVerts; j++ ) {
tess.texCoords[Doug + j][0] = texCoords[j].st[0];
tess.texCoords[Doug + j][1] = texCoords[j].st[1];
// FIXME: fill in lightmapST for completeness?
}
tess.numVertexes += surface->numVerts;
}
/*
==============
RB_SurfaceFace
==============
*/
static void RB_SurfaceFace( srfBspSurface_t *srf ) {
if (RB_SurfaceVaoCached(srf->numVerts, srf->verts, srf->numIndexes,
srf->indexes, srf->dlightBits, srf->pshadowBits))
{
return;
}
RB_SurfaceVertsAndIndexes(srf->numVerts, srf->verts, srf->numIndexes,
srf->indexes, srf->dlightBits, srf->pshadowBits);
}
static float LodErrorForVolume( vec3_t local, float radius ) {
vec3_t world;
float d;
// never let it go negative
if ( r_lodCurveError->value < 0 ) {
return 0;
}
world[0] = local[0] * backEnd.or.axis[0][0] + local[1] * backEnd.or.axis[1][0] +
local[2] * backEnd.or.axis[2][0] + backEnd.or.origin[0];
world[1] = local[0] * backEnd.or.axis[0][1] + local[1] * backEnd.or.axis[1][1] +
local[2] * backEnd.or.axis[2][1] + backEnd.or.origin[1];
world[2] = local[0] * backEnd.or.axis[0][2] + local[1] * backEnd.or.axis[1][2] +
local[2] * backEnd.or.axis[2][2] + backEnd.or.origin[2];
VectorSubtract( world, backEnd.viewParms.or.origin, world );
d = DotProduct( world, backEnd.viewParms.or.axis[0] );
if ( d < 0 ) {
d = -d;
}
d -= radius;
if ( d < 1 ) {
d = 1;
}
return r_lodCurveError->value / d;
}
/*
=============
RB_SurfaceGrid
Just copy the grid of points and triangulate
=============
*/
static void RB_SurfaceGrid( srfBspSurface_t *srf ) {
int i, j;
float *xyz;
float *texCoords, *lightCoords;
int16_t *normal;
int16_t *tangent;
uint16_t *color;
int16_t *lightdir;
srfVert_t *dv;
int rows, irows, vrows;
int used;
int widthTable[MAX_GRID_SIZE];
int heightTable[MAX_GRID_SIZE];
float lodError;
int lodWidth, lodHeight;
int numVertexes;
int dlightBits;
int pshadowBits;
//int *vDlightBits;
if (RB_SurfaceVaoCached(srf->numVerts, srf->verts, srf->numIndexes,
srf->indexes, srf->dlightBits, srf->pshadowBits))
{
return;
}
RB_CheckVao(tess.vao);
2013-01-24 22:53:08 +00:00
dlightBits = srf->dlightBits;
tess.dlightBits |= dlightBits;
2013-01-24 22:53:08 +00:00
pshadowBits = srf->pshadowBits;
tess.pshadowBits |= pshadowBits;
// determine the allowable discrepance
lodError = LodErrorForVolume( srf->lodOrigin, srf->lodRadius );
// determine which rows and columns of the subdivision
// we are actually going to use
widthTable[0] = 0;
lodWidth = 1;
for ( i = 1 ; i < srf->width-1 ; i++ ) {
if ( srf->widthLodError[i] <= lodError ) {
widthTable[lodWidth] = i;
lodWidth++;
}
}
widthTable[lodWidth] = srf->width-1;
lodWidth++;
heightTable[0] = 0;
lodHeight = 1;
for ( i = 1 ; i < srf->height-1 ; i++ ) {
if ( srf->heightLodError[i] <= lodError ) {
heightTable[lodHeight] = i;
lodHeight++;
}
}
heightTable[lodHeight] = srf->height-1;
lodHeight++;
// very large grids may have more points or indexes than can be fit
// in the tess structure, so we may have to issue it in multiple passes
used = 0;
while ( used < lodHeight - 1 ) {
// see how many rows of both verts and indexes we can add without overflowing
do {
vrows = ( SHADER_MAX_VERTEXES - tess.numVertexes ) / lodWidth;
irows = ( SHADER_MAX_INDEXES - tess.numIndexes ) / ( lodWidth * 6 );
// if we don't have enough space for at least one strip, flush the buffer
if ( vrows < 2 || irows < 1 ) {
RB_EndSurface();
2013-09-16 07:54:26 +00:00
RB_BeginSurface(tess.shader, tess.fogNum, tess.cubemapIndex );
} else {
break;
}
} while ( 1 );
rows = irows;
if ( vrows < irows + 1 ) {
rows = vrows - 1;
}
if ( used + rows > lodHeight ) {
rows = lodHeight - used;
}
numVertexes = tess.numVertexes;
xyz = tess.xyz[numVertexes];
normal = tess.normal[numVertexes];
tangent = tess.tangent[numVertexes];
texCoords = tess.texCoords[numVertexes];
lightCoords = tess.lightCoords[numVertexes];
color = tess.color[numVertexes];
lightdir = tess.lightdir[numVertexes];
//vDlightBits = &tess.vertexDlightBits[numVertexes];
for ( i = 0 ; i < rows ; i++ ) {
for ( j = 0 ; j < lodWidth ; j++ ) {
dv = srf->verts + heightTable[ used + i ] * srf->width
+ widthTable[ j ];
if ( tess.shader->vertexAttribs & ATTR_POSITION )
{
VectorCopy(dv->xyz, xyz);
xyz += 4;
}
if ( tess.shader->vertexAttribs & ATTR_NORMAL )
{
VectorCopy4(dv->normal, normal);
normal += 4;
}
if ( tess.shader->vertexAttribs & ATTR_TANGENT )
{
VectorCopy4(dv->tangent, tangent);
tangent += 4;
}
if ( tess.shader->vertexAttribs & ATTR_TEXCOORD )
{
VectorCopy2(dv->st, texCoords);
texCoords += 2;
}
if ( tess.shader->vertexAttribs & ATTR_LIGHTCOORD )
{
VectorCopy2(dv->lightmap, lightCoords);
lightCoords += 2;
}
if ( tess.shader->vertexAttribs & ATTR_COLOR )
{
VectorCopy4(dv->color, color);
color += 4;
}
if ( tess.shader->vertexAttribs & ATTR_LIGHTDIRECTION )
{
VectorCopy4(dv->lightdir, lightdir);
lightdir += 4;
}
//*vDlightBits++ = dlightBits;
}
}
// add the indexes
{
int numIndexes;
int w, h;
h = rows - 1;
w = lodWidth - 1;
numIndexes = tess.numIndexes;
for (i = 0 ; i < h ; i++) {
for (j = 0 ; j < w ; j++) {
int v1, v2, v3, v4;
// vertex order to be reckognized as tristrips
v1 = numVertexes + i*lodWidth + j + 1;
v2 = v1 - 1;
v3 = v2 + lodWidth;
v4 = v3 + 1;
tess.indexes[numIndexes] = v2;
tess.indexes[numIndexes+1] = v3;
tess.indexes[numIndexes+2] = v1;
tess.indexes[numIndexes+3] = v1;
tess.indexes[numIndexes+4] = v3;
tess.indexes[numIndexes+5] = v4;
numIndexes += 6;
}
}
tess.numIndexes = numIndexes;
}
tess.numVertexes += rows * lodWidth;
used += rows - 1;
}
}
/*
===========================================================================
NULL MODEL
===========================================================================
*/
/*
===================
RB_SurfaceAxis
Draws x/y/z lines from the origin for orientation debugging
===================
*/
static void RB_SurfaceAxis( void ) {
// FIXME: implement this
#if 0
GL_BindToTMU( tr.whiteImage, TB_COLORMAP );
GL_State( GLS_DEFAULT );
qglLineWidth( 3 );
qglBegin( GL_LINES );
qglColor3f( 1,0,0 );
qglVertex3f( 0,0,0 );
qglVertex3f( 16,0,0 );
qglColor3f( 0,1,0 );
qglVertex3f( 0,0,0 );
qglVertex3f( 0,16,0 );
qglColor3f( 0,0,1 );
qglVertex3f( 0,0,0 );
qglVertex3f( 0,0,16 );
qglEnd();
qglLineWidth( 1 );
#endif
}
//===========================================================================
/*
====================
RB_SurfaceEntity
Entities that have a single procedurally generated surface
====================
*/
static void RB_SurfaceEntity( surfaceType_t *surfType ) {
switch( backEnd.currentEntity->e.reType ) {
#ifdef ELITEFORCE
case RT_ALPHAVERTPOLY:
#endif
case RT_SPRITE:
RB_SurfaceSprite();
break;
case RT_BEAM:
RB_SurfaceBeam();
break;
case RT_RAIL_CORE:
RB_SurfaceRailCore();
break;
case RT_RAIL_RINGS:
RB_SurfaceRailRings();
break;
case RT_LIGHTNING:
RB_SurfaceLightningBolt();
break;
#ifdef ELITEFORCE
case RT_ORIENTEDSPRITE:
RB_SurfaceOrientedSprite();
break;
case RT_LINE:
RB_SurfaceLine();
break;
case RT_ORIENTEDLINE:
RB_SurfaceOrientedLine();
break;
case RT_LINE2:
RB_SurfaceLine2();
break;
case RT_BEZIER:
RB_SurfaceBezier();
break;
case RT_CYLINDER:
RB_SurfaceCylinder();
break;
case RT_ELECTRICITY:
RB_SurfaceElectricity();
break;
#endif
default:
RB_SurfaceAxis();
break;
}
}
static void RB_SurfaceBad( surfaceType_t *surfType ) {
ri.Printf( PRINT_ALL, "Bad surface tesselated.\n" );
}
static void RB_SurfaceFlare(srfFlare_t *surf)
{
if (r_flares->integer)
RB_AddFlare(surf, tess.fogNum, surf->origin, surf->color, surf->normal);
}
2014-10-14 08:50:44 +00:00
void RB_SurfaceVaoMdvMesh(srfVaoMdvMesh_t * surface)
{
//mdvModel_t *mdvModel;
//mdvSurface_t *mdvSurface;
refEntity_t *refEnt;
2014-10-14 08:50:44 +00:00
GLimp_LogComment("--- RB_SurfaceVaoMdvMesh ---\n");
if (ShaderRequiresCPUDeforms(tess.shader))
{
RB_SurfaceMesh(surface->mdvSurface);
return;
}
2014-10-14 08:50:44 +00:00
if(!surface->vao)
return;
2014-10-14 08:50:44 +00:00
//RB_CheckVao(surface->vao);
RB_EndSurface();
2013-09-16 07:54:26 +00:00
RB_BeginSurface(tess.shader, tess.fogNum, tess.cubemapIndex);
2014-10-14 08:50:44 +00:00
R_BindVao(surface->vao);
2014-10-14 08:50:44 +00:00
tess.useInternalVao = qfalse;
2014-10-23 23:51:10 +00:00
tess.numIndexes = surface->numIndexes;
tess.numVertexes = surface->numVerts;
//mdvModel = surface->mdvModel;
//mdvSurface = surface->mdvSurface;
refEnt = &backEnd.currentEntity->e;
2014-10-23 23:51:10 +00:00
glState.vertexAttribsInterpolation = (refEnt->oldframe == refEnt->frame) ? 0.0f : refEnt->backlerp;
2014-10-14 13:15:06 +00:00
if (surface->mdvModel->numFrames > 1)
2014-10-23 23:51:10 +00:00
{
int frameOffset, attribIndex;
vaoAttrib_t *vAtb;
2014-10-14 13:15:06 +00:00
glState.vertexAnimation = qtrue;
2014-10-23 23:51:10 +00:00
if (glRefConfig.vertexArrayObject)
{
2016-07-25 04:29:13 +00:00
qglBindBuffer(GL_ARRAY_BUFFER, surface->vao->vertexesVBO);
2014-10-23 23:51:10 +00:00
}
frameOffset = refEnt->frame * surface->vao->frameSize;
attribIndex = ATTR_INDEX_POSITION;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset + frameOffset));
2014-10-23 23:51:10 +00:00
attribIndex = ATTR_INDEX_NORMAL;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset + frameOffset));
2014-10-23 23:51:10 +00:00
attribIndex = ATTR_INDEX_TANGENT;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset + frameOffset));
2014-10-23 23:51:10 +00:00
frameOffset = refEnt->oldframe * surface->vao->frameSize;
attribIndex = ATTR_INDEX_POSITION2;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset + frameOffset));
2014-10-23 23:51:10 +00:00
attribIndex = ATTR_INDEX_NORMAL2;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset + frameOffset));
2014-10-23 23:51:10 +00:00
attribIndex = ATTR_INDEX_TANGENT2;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset + frameOffset));
2014-10-23 23:51:10 +00:00
if (!glRefConfig.vertexArrayObject)
{
attribIndex = ATTR_INDEX_TEXCOORD;
vAtb = &surface->vao->attribs[attribIndex];
2016-07-25 04:29:13 +00:00
qglVertexAttribPointer(attribIndex, vAtb->count, vAtb->type, vAtb->normalized, vAtb->stride, BUFFER_OFFSET(vAtb->offset));
2014-10-23 23:51:10 +00:00
}
}
RB_EndSurface();
// So we don't lerp surfaces that shouldn't be lerped
glState.vertexAnimation = qfalse;
}
static void RB_SurfaceSkip( void *surf ) {
}
void (*rb_surfaceTable[SF_NUM_SURFACE_TYPES])( void *) = {
(void(*)(void*))RB_SurfaceBad, // SF_BAD,
(void(*)(void*))RB_SurfaceSkip, // SF_SKIP,
(void(*)(void*))RB_SurfaceFace, // SF_FACE,
(void(*)(void*))RB_SurfaceGrid, // SF_GRID,
(void(*)(void*))RB_SurfaceTriangles, // SF_TRIANGLES,
(void(*)(void*))RB_SurfacePolychain, // SF_POLY,
(void(*)(void*))RB_SurfaceMesh, // SF_MDV,
(void(*)(void*))RB_MDRSurfaceAnim, // SF_MDR,
(void(*)(void*))RB_IQMSurfaceAnim, // SF_IQM,
(void(*)(void*))RB_SurfaceFlare, // SF_FLARE,
(void(*)(void*))RB_SurfaceEntity, // SF_ENTITY
2014-10-14 08:50:44 +00:00
(void(*)(void*))RB_SurfaceVaoMdvMesh, // SF_VAO_MDVMESH
};