]> de.git.xonotic.org Git - xonotic/darkplaces.git/blobdiff - r_shadow.c
nudge dlights out of walls if they're too close
[xonotic/darkplaces.git] / r_shadow.c
index 50de89788c7513570ea2222a6efcc91cd5fd72e0..058d5589b96c125b7e19b444ae747487b6b80f49 100644 (file)
@@ -14,8 +14,6 @@ extern void R_Shadow_EditLights_Init(void);
 int r_shadowstage = SHADOWSTAGE_NONE;
 int r_shadow_reloadlights = false;
 
-int r_shadow_lightingmode = 0;
-
 mempool_t *r_shadow_mempool;
 
 int maxshadowelements;
@@ -25,10 +23,13 @@ qbyte *trianglefacinglight;
 int *trianglefacinglightlist;
 
 int maxshadowvertices;
-float *shadowvertices;
+float *shadowvertex3f;
+int maxvertexupdate;
+int *vertexupdate;
+int vertexupdatenum;
 
 rtexturepool_t *r_shadow_texturepool;
-rtexture_t *r_shadow_normalscubetexture;
+rtexture_t *r_shadow_normalcubetexture;
 rtexture_t *r_shadow_attenuation2dtexture;
 rtexture_t *r_shadow_attenuation3dtexture;
 rtexture_t *r_shadow_blankbumptexture;
@@ -38,13 +39,15 @@ rtexture_t *r_shadow_blankwhitetexture;
 cvar_t r_shadow_lightattenuationpower = {0, "r_shadow_lightattenuationpower", "0.5"};
 cvar_t r_shadow_lightattenuationscale = {0, "r_shadow_lightattenuationscale", "1"};
 cvar_t r_shadow_lightintensityscale = {0, "r_shadow_lightintensityscale", "1"};
-cvar_t r_shadow_realtime = {0, "r_shadow_realtime", "0"};
+cvar_t r_shadow_realtime_world = {0, "r_shadow_realtime_world", "0"};
+cvar_t r_shadow_realtime_dlight = {0, "r_shadow_realtime_dlight", "0"};
+cvar_t r_shadow_visiblevolumes = {0, "r_shadow_visiblevolumes", "0"};
 cvar_t r_shadow_gloss = {0, "r_shadow_gloss", "1"};
 cvar_t r_shadow_debuglight = {0, "r_shadow_debuglight", "-1"};
 cvar_t r_shadow_scissor = {0, "r_shadow_scissor", "1"};
 cvar_t r_shadow_bumpscale_bumpmap = {0, "r_shadow_bumpscale_bumpmap", "4"};
 cvar_t r_shadow_bumpscale_basetexture = {0, "r_shadow_bumpscale_basetexture", "0"};
-cvar_t r_shadow_shadownudge = {0, "r_shadow_shadownudge", "1"};
+cvar_t r_shadow_polygonoffset = {0, "r_shadow_polygonoffset", "-1"};
 cvar_t r_shadow_portallight = {0, "r_shadow_portallight", "1"};
 cvar_t r_shadow_projectdistance = {0, "r_shadow_projectdistance", "100000"};
 cvar_t r_shadow_texture3d = {0, "r_shadow_texture3d", "1"};
@@ -66,11 +69,14 @@ void r_shadow_start(void)
        maxshadowelements = 0;
        shadowelements = NULL;
        maxshadowvertices = 0;
-       shadowvertices = NULL;
+       shadowvertex3f = NULL;
+       maxvertexupdate = 0;
+       vertexupdate = NULL;
+       vertexupdatenum = 0;
        maxtrianglefacinglight = 0;
        trianglefacinglight = NULL;
        trianglefacinglightlist = NULL;
-       r_shadow_normalscubetexture = NULL;
+       r_shadow_normalcubetexture = NULL;
        r_shadow_attenuation2dtexture = NULL;
        r_shadow_attenuation3dtexture = NULL;
        r_shadow_blankbumptexture = NULL;
@@ -85,7 +91,7 @@ void r_shadow_shutdown(void)
 {
        R_Shadow_ClearWorldLights();
        r_shadow_reloadlights = true;
-       r_shadow_normalscubetexture = NULL;
+       r_shadow_normalcubetexture = NULL;
        r_shadow_attenuation2dtexture = NULL;
        r_shadow_attenuation3dtexture = NULL;
        r_shadow_blankbumptexture = NULL;
@@ -95,7 +101,10 @@ void r_shadow_shutdown(void)
        maxshadowelements = 0;
        shadowelements = NULL;
        maxshadowvertices = 0;
-       shadowvertices = NULL;
+       shadowvertex3f = NULL;
+       maxvertexupdate = 0;
+       vertexupdate = NULL;
+       vertexupdatenum = 0;
        maxtrianglefacinglight = 0;
        trianglefacinglight = NULL;
        trianglefacinglightlist = NULL;
@@ -113,13 +122,15 @@ void R_Shadow_Init(void)
        Cvar_RegisterVariable(&r_shadow_lightattenuationpower);
        Cvar_RegisterVariable(&r_shadow_lightattenuationscale);
        Cvar_RegisterVariable(&r_shadow_lightintensityscale);
-       Cvar_RegisterVariable(&r_shadow_realtime);
+       Cvar_RegisterVariable(&r_shadow_realtime_world);
+       Cvar_RegisterVariable(&r_shadow_realtime_dlight);
+       Cvar_RegisterVariable(&r_shadow_visiblevolumes);
        Cvar_RegisterVariable(&r_shadow_gloss);
        Cvar_RegisterVariable(&r_shadow_debuglight);
        Cvar_RegisterVariable(&r_shadow_scissor);
        Cvar_RegisterVariable(&r_shadow_bumpscale_bumpmap);
        Cvar_RegisterVariable(&r_shadow_bumpscale_basetexture);
-       Cvar_RegisterVariable(&r_shadow_shadownudge);
+       Cvar_RegisterVariable(&r_shadow_polygonoffset);
        Cvar_RegisterVariable(&r_shadow_portallight);
        Cvar_RegisterVariable(&r_shadow_projectdistance);
        Cvar_RegisterVariable(&r_shadow_texture3d);
@@ -127,90 +138,42 @@ void R_Shadow_Init(void)
        R_RegisterModule("R_Shadow", r_shadow_start, r_shadow_shutdown, r_shadow_newmap);
 }
 
-void R_Shadow_ProjectVertices(float *verts, int numverts, const float *relativelightorigin, float projectdistance)
-{
-       int i;
-       float *in, *out, diff[4];
-       in = verts;
-       out = verts + numverts * 4;
-       for (i = 0;i < numverts;i++, in += 4, out += 4)
-       {
-               VectorSubtract(in, relativelightorigin, diff);
-               VectorNormalizeFast(diff);
-               VectorMA(in, projectdistance, diff, out);
-               VectorMA(in, r_shadow_shadownudge.value, diff, in);
-       }
-}
-
-int R_Shadow_MakeTriangleShadowFlags(const int *elements, const float *vertex, int numtris, qbyte *facing, int *list, const float *relativelightorigin)
+int R_Shadow_MakeTriangleShadowFlags_Vertex3f(const int *elements, const float *vertex, int numtris, qbyte *facing, int *list, const float *relativelightorigin)
 {
        int i, tris = 0;
        const float *v0, *v1, *v2;
        for (i = 0;i < numtris;i++, elements += 3)
        {
                // calculate triangle facing flag
-               v0 = vertex + elements[0] * 4;
-               v1 = vertex + elements[1] * 4;
-               v2 = vertex + elements[2] * 4;
-               // we do not need to normalize the surface normal because both sides
-               // of the comparison use it, therefore they are both multiplied the
-               // same amount...  furthermore the subtract can be done on the
-               // vectors, saving a little bit of math in the dotproducts
-#if 1
-               // fast version
-               // subtracts v1 from v0 and v2, combined into a crossproduct,
-               // combined with a dotproduct of the light location relative to the
-               // first point of the triangle (any point works, since any triangle
-               // is obviously flat), and finally a comparison to determine if the
-               // light is infront of the triangle (the goal of this statement)
-               if((relativelightorigin[0] - v0[0]) * ((v0[1] - v1[1]) * (v2[2] - v1[2]) - (v0[2] - v1[2]) * (v2[1] - v1[1]))
-                + (relativelightorigin[1] - v0[1]) * ((v0[2] - v1[2]) * (v2[0] - v1[0]) - (v0[0] - v1[0]) * (v2[2] - v1[2]))
-                + (relativelightorigin[2] - v0[2]) * ((v0[0] - v1[0]) * (v2[1] - v1[1]) - (v0[1] - v1[1]) * (v2[0] - v1[0])) > 0)
-               {
-                       facing[i] = true;
-                       list[tris++] = i;
-               }
-               else
-                       facing[i] = false;
-#else
-               // readable version
-               {
-               float dir0[3], dir1[3], temp[3];
-
-               // calculate two mostly perpendicular edge directions
-               VectorSubtract(v0, v1, dir0);
-               VectorSubtract(v2, v1, dir1);
-
-               // we have two edge directions, we can calculate a third vector from
-               // them, which is the direction of the surface normal (it's magnitude
-               // is not 1 however)
-               CrossProduct(dir0, dir1, temp);
-
-               // this is entirely unnecessary, but kept for clarity
-               //VectorNormalize(temp);
-
-               // compare distance of light along normal, with distance of any point
-               // of the triangle along the same normal (the triangle is planar,
-               // I.E. flat, so all points give the same answer)
-               // the normal is not normalized because it is used on both sides of
-               // the comparison, so it's magnitude does not matter
-               if (DotProduct(relativelightorigin, temp) >= DotProduct(v0, temp))
+               v0 = vertex + elements[0] * 3;
+               v1 = vertex + elements[1] * 3;
+               v2 = vertex + elements[2] * 3;
+               if(PointInfrontOfTriangle(relativelightorigin, v0, v1, v2))
                {
                        facing[i] = true;
                        list[tris++] = i;
                }
                else
                        facing[i] = false;
-               }
-#endif
        }
        return tris;
 }
 
-int R_Shadow_BuildShadowVolumeTriangles(const int *elements, const int *neighbors, int numverts, const qbyte *facing, const int *facinglist, int numfacing, int *out)
+int R_Shadow_BuildShadowVolume(const int *elements, const int *neighbors, int numverts, const qbyte *facing, const int *facinglist, int numfacing, int *out, float *vertices, const float *relativelightorigin, float projectdistance)
 {
-       int i, tris;
+       int i, j, tris, vertexpointeradjust = numverts * 3;
        const int *e, *n;
+       float *vin, *vout;
+
+       if (maxvertexupdate < numverts)
+       {
+               maxvertexupdate = numverts;
+               if (vertexupdate)
+                       Mem_Free(vertexupdate);
+               vertexupdate = Mem_Alloc(r_shadow_mempool, maxvertexupdate * sizeof(int));
+       }
+       vertexupdatenum++;
+
        // check each frontface for bordering backfaces,
        // and cast shadow polygons from those edges,
        // also create front and back caps for shadow volume
@@ -228,6 +191,19 @@ int R_Shadow_BuildShadowVolumeTriangles(const int *elements, const int *neighbor
        for (i = 0;i < numfacing;i++)
        {
                e = elements + facinglist[i] * 3;
+               // generate vertices if needed
+               for (j = 0;j < 3;j++)
+               {
+                       if (vertexupdate[e[j]] != vertexupdatenum)
+                       {
+                               vertexupdate[e[j]] = vertexupdatenum;
+                               vin = vertices + e[j] * 3;
+                               vout = vin + vertexpointeradjust;
+                               vout[0] = relativelightorigin[0] + projectdistance * (vin[0] - relativelightorigin[0]);
+                               vout[1] = relativelightorigin[1] + projectdistance * (vin[1] - relativelightorigin[1]);
+                               vout[2] = relativelightorigin[2] + projectdistance * (vin[2] - relativelightorigin[2]);
+                       }
+               }
                out[0] = e[2] + numverts;
                out[1] = e[1] + numverts;
                out[2] = e[0] + numverts;
@@ -313,11 +289,11 @@ float *R_Shadow_VertexBuffer(int numvertices)
        if (maxshadowvertices < numvertices)
        {
                maxshadowvertices = numvertices;
-               if (shadowvertices)
-                       Mem_Free(shadowvertices);
-               shadowvertices = Mem_Alloc(r_shadow_mempool, maxshadowvertices * sizeof(float[4]));
+               if (shadowvertex3f)
+                       Mem_Free(shadowvertex3f);
+               shadowvertex3f = Mem_Alloc(r_shadow_mempool, maxshadowvertices * sizeof(float[3]));
        }
-       return shadowvertices;
+       return shadowvertex3f;
 }
 
 void R_Shadow_Volume(int numverts, int numtris, int *elements, int *neighbors, vec3_t relativelightorigin, float lightradius, float projectdistance)
@@ -347,8 +323,8 @@ void R_Shadow_Volume(int numverts, int numtris, int *elements, int *neighbors, v
 // description:
 // draws the shadow volumes of the model.
 // requirements:
-// vertex locations must already be in vertices before use.
-// vertices must have capacity for numverts * 2.
+// vertex locations must already be in varray_vertex3f before use.
+// varray_vertex3f must have capacity for numverts * 2.
 
        // make sure trianglefacinglight is big enough for this volume
        if (maxtrianglefacinglight < numtris)
@@ -359,18 +335,17 @@ void R_Shadow_Volume(int numverts, int numtris, int *elements, int *neighbors, v
                R_Shadow_ResizeShadowElements(numtris);
 
        // check which triangles are facing the light
-       tris = R_Shadow_MakeTriangleShadowFlags(elements, varray_vertex, numtris, trianglefacinglight, trianglefacinglightlist, relativelightorigin);
-       if (!tris)
-               return;
-
-       // output triangle elements
-       tris = R_Shadow_BuildShadowVolumeTriangles(elements, neighbors, numverts, trianglefacinglight, trianglefacinglightlist, tris, shadowelements);
+       tris = R_Shadow_MakeTriangleShadowFlags_Vertex3f(elements, varray_vertex3f, numtris, trianglefacinglight, trianglefacinglightlist, relativelightorigin);
        if (!tris)
                return;
 
        // by clever use of elements we can construct the whole shadow from
        // the unprojected vertices and the projected vertices
-       R_Shadow_ProjectVertices(varray_vertex, numverts, relativelightorigin, projectdistance);
+
+       // output triangle elements and vertices
+       tris = R_Shadow_BuildShadowVolume(elements, neighbors, numverts, trianglefacinglight, trianglefacinglightlist, tris, shadowelements, varray_vertex3f, relativelightorigin, projectdistance);
+       if (!tris)
+               return;
 
        if (r_shadowstage == SHADOWSTAGE_STENCIL)
        {
@@ -400,8 +375,8 @@ void R_Shadow_RenderShadowMeshVolume(shadowmesh_t *firstmesh)
                for (mesh = firstmesh;mesh;mesh = mesh->next)
                {
                        R_Mesh_GetSpace(mesh->numverts);
-                       memcpy(varray_vertex, mesh->verts, mesh->numverts * sizeof(float[4]));
-                       R_Mesh_Draw(mesh->numverts, mesh->numtriangles, mesh->elements);
+                       R_Mesh_CopyVertex3f(mesh->vertex3f, mesh->numverts);
+                       R_Mesh_Draw(mesh->numverts, mesh->numtriangles, mesh->element3i);
                        c_rtcached_shadowmeshes++;
                        c_rtcached_shadowtris += mesh->numtriangles;
                }
@@ -412,8 +387,8 @@ void R_Shadow_RenderShadowMeshVolume(shadowmesh_t *firstmesh)
        for (mesh = firstmesh;mesh;mesh = mesh->next)
        {
                R_Mesh_GetSpace(mesh->numverts);
-               memcpy(varray_vertex, mesh->verts, mesh->numverts * sizeof(float[4]));
-               R_Mesh_Draw(mesh->numverts, mesh->numtriangles, mesh->elements);
+               R_Mesh_CopyVertex3f(mesh->vertex3f, mesh->numverts);
+               R_Mesh_Draw(mesh->numverts, mesh->numtriangles, mesh->element3i);
                c_rtcached_shadowmeshes++;
                c_rtcached_shadowtris += mesh->numtriangles;
        }
@@ -448,56 +423,61 @@ static void R_Shadow_MakeTextures(void)
        data[2] = 255;
        data[3] = 255;
        r_shadow_blankwhitetexture = R_LoadTexture2D(r_shadow_texturepool, "blankwhite", 1, 1, data, TEXTYPE_RGBA, TEXF_PRECACHE, NULL);
-       for (side = 0;side < 6;side++)
+       if (gl_texturecubemap)
        {
-               for (y = 0;y < NORMSIZE;y++)
+               for (side = 0;side < 6;side++)
                {
-                       for (x = 0;x < NORMSIZE;x++)
+                       for (y = 0;y < NORMSIZE;y++)
                        {
-                               s = (x + 0.5f) * (2.0f / NORMSIZE) - 1.0f;
-                               t = (y + 0.5f) * (2.0f / NORMSIZE) - 1.0f;
-                               switch(side)
+                               for (x = 0;x < NORMSIZE;x++)
                                {
-                               case 0:
-                                       v[0] = 1;
-                                       v[1] = -t;
-                                       v[2] = -s;
-                                       break;
-                               case 1:
-                                       v[0] = -1;
-                                       v[1] = -t;
-                                       v[2] = s;
-                                       break;
-                               case 2:
-                                       v[0] = s;
-                                       v[1] = 1;
-                                       v[2] = t;
-                                       break;
-                               case 3:
-                                       v[0] = s;
-                                       v[1] = -1;
-                                       v[2] = -t;
-                                       break;
-                               case 4:
-                                       v[0] = s;
-                                       v[1] = -t;
-                                       v[2] = 1;
-                                       break;
-                               case 5:
-                                       v[0] = -s;
-                                       v[1] = -t;
-                                       v[2] = -1;
-                                       break;
+                                       s = (x + 0.5f) * (2.0f / NORMSIZE) - 1.0f;
+                                       t = (y + 0.5f) * (2.0f / NORMSIZE) - 1.0f;
+                                       switch(side)
+                                       {
+                                       case 0:
+                                               v[0] = 1;
+                                               v[1] = -t;
+                                               v[2] = -s;
+                                               break;
+                                       case 1:
+                                               v[0] = -1;
+                                               v[1] = -t;
+                                               v[2] = s;
+                                               break;
+                                       case 2:
+                                               v[0] = s;
+                                               v[1] = 1;
+                                               v[2] = t;
+                                               break;
+                                       case 3:
+                                               v[0] = s;
+                                               v[1] = -1;
+                                               v[2] = -t;
+                                               break;
+                                       case 4:
+                                               v[0] = s;
+                                               v[1] = -t;
+                                               v[2] = 1;
+                                               break;
+                                       case 5:
+                                               v[0] = -s;
+                                               v[1] = -t;
+                                               v[2] = -1;
+                                               break;
+                                       }
+                                       intensity = 127.0f / sqrt(DotProduct(v, v));
+                                       data[((side*NORMSIZE+y)*NORMSIZE+x)*4+0] = 128.0f + intensity * v[0];
+                                       data[((side*NORMSIZE+y)*NORMSIZE+x)*4+1] = 128.0f + intensity * v[1];
+                                       data[((side*NORMSIZE+y)*NORMSIZE+x)*4+2] = 128.0f + intensity * v[2];
+                                       data[((side*NORMSIZE+y)*NORMSIZE+x)*4+3] = 255;
                                }
-                               intensity = 127.0f / sqrt(DotProduct(v, v));
-                               data[((side*NORMSIZE+y)*NORMSIZE+x)*4+0] = 128.0f + intensity * v[0];
-                               data[((side*NORMSIZE+y)*NORMSIZE+x)*4+1] = 128.0f + intensity * v[1];
-                               data[((side*NORMSIZE+y)*NORMSIZE+x)*4+2] = 128.0f + intensity * v[2];
-                               data[((side*NORMSIZE+y)*NORMSIZE+x)*4+3] = 255;
                        }
                }
+               r_shadow_normalcubetexture = R_LoadTextureCubeMap(r_shadow_texturepool, "normalcube", NORMSIZE, data, TEXTYPE_RGBA, TEXF_PRECACHE | TEXF_CLAMP, NULL);
        }
-       r_shadow_normalscubetexture = R_LoadTextureCubeMap(r_shadow_texturepool, "normalscube", NORMSIZE, data, TEXTYPE_RGBA, TEXF_PRECACHE | TEXF_CLAMP, NULL);
+       else
+               r_shadow_normalcubetexture = NULL;
        for (y = 0;y < ATTEN2DSIZE;y++)
        {
                for (x = 0;x < ATTEN2DSIZE;x++)
@@ -556,6 +536,22 @@ void R_Shadow_Stage_Begin(void)
         || r_shadow_lightattenuationpower.value != r_shadow_attenpower
         || r_shadow_lightattenuationscale.value != r_shadow_attenscale)
                R_Shadow_MakeTextures();
+
+       memset(&m, 0, sizeof(m));
+       m.blendfunc1 = GL_ONE;
+       m.blendfunc2 = GL_ZERO;
+       R_Mesh_State(&m);
+       GL_Color(0, 0, 0, 1);
+       qglDisable(GL_SCISSOR_TEST);
+       r_shadowstage = SHADOWSTAGE_NONE;
+
+       c_rt_lights = c_rt_clears = c_rt_scissored = 0;
+       c_rt_shadowmeshes = c_rt_shadowtris = c_rt_lightmeshes = c_rt_lighttris = 0;
+       c_rtcached_shadowmeshes = c_rtcached_shadowtris = 0;
+}
+
+void R_Shadow_LoadWorldLightsIfNeeded(void)
+{
        if (r_shadow_reloadlights && cl.worldmodel)
        {
                R_Shadow_ClearWorldLights();
@@ -568,17 +564,6 @@ void R_Shadow_Stage_Begin(void)
                                R_Shadow_LoadWorldLightsFromMap_LightArghliteTyrlite();
                }
        }
-
-       memset(&m, 0, sizeof(m));
-       m.blendfunc1 = GL_ONE;
-       m.blendfunc2 = GL_ZERO;
-       R_Mesh_State(&m);
-       GL_Color(0, 0, 0, 1);
-       r_shadowstage = SHADOWSTAGE_NONE;
-
-       c_rt_lights = c_rt_clears = c_rt_scissored = 0;
-       c_rt_shadowmeshes = c_rt_shadowtris = c_rt_lightmeshes = c_rt_lighttris = 0;
-       c_rtcached_shadowmeshes = c_rtcached_shadowtris = 0;
 }
 
 void R_Shadow_Stage_ShadowVolumes(void)
@@ -832,7 +817,9 @@ int R_Shadow_ScissorForBBox(const float *mins, const float *maxs)
        if (!r_shadow_scissor.integer)
                return false;
        // if view is inside the box, just say yes it's visible
-       if (BoxesOverlap(r_origin, r_origin, mins, maxs))
+       // LordHavoc: for some odd reason scissor seems broken without stencil
+       // (?!?  seems like a driver bug) so abort if gl_stencil is false
+       if (!gl_stencil || BoxesOverlap(r_origin, r_origin, mins, maxs))
        {
                qglDisable(GL_SCISSOR_TEST);
                return false;
@@ -854,8 +841,7 @@ int R_Shadow_ScissorForBBox(const float *mins, const float *maxs)
        if (DotProduct(vpn, v2) <= f)
        {
                // entirely behind nearclip plane
-               qglDisable(GL_SCISSOR_TEST);
-               return false;
+               return true;
        }
        if (DotProduct(vpn, v) >= f)
        {
@@ -994,91 +980,133 @@ int R_Shadow_ScissorForBBox(const float *mins, const float *maxs)
        return false;
 }
 
-void R_Shadow_VertexLighting(float *color, int numverts, const float *vertex, const float *normal, const float *lightcolor, const float *relativelightorigin, float lightradius)
+void R_Shadow_VertexLighting(int numverts, const float *vertex3f, const float *normal3f, const float *lightcolor, const matrix4x4_t *m)
 {
-       float dist, dot, intensity, iradius = 1.0f / lightradius, radius2 = lightradius * lightradius, v[3];
-       for (;numverts > 0;numverts--, vertex += 4, color += 4, normal += 4)
+       float *color4f = varray_color4f;
+       float dist, dot, intensity, v[3], n[3];
+       for (;numverts > 0;numverts--, vertex3f += 3, normal3f += 3, color4f += 4)
        {
-               VectorSubtract(vertex, relativelightorigin, v);
-               if ((dot = DotProduct(normal, v)) > 0 && (dist = DotProduct(v, v)) < radius2)
+               Matrix4x4_Transform(m, vertex3f, v);
+               if ((dist = DotProduct(v, v)) < 1)
                {
-                       dist = sqrt(dist);
-                       intensity = pow(1 - (dist * iradius), r_shadow_attenpower) * r_shadow_attenscale * dot / dist;
-                       VectorScale(lightcolor, intensity, color);
+                       Matrix4x4_Transform3x3(m, normal3f, n);
+                       if ((dot = DotProduct(n, v)) > 0)
+                       {
+                               dist = sqrt(dist);
+                               intensity = pow(1 - dist, r_shadow_attenpower) * r_shadow_attenscale * dot / sqrt(DotProduct(n,n));
+                               VectorScale(lightcolor, intensity, color4f);
+                               color4f[3] = 1;
+                       }
+                       else
+                       {
+                               VectorClear(color4f);
+                               color4f[3] = 1;
+                       }
                }
                else
-                       VectorClear(color);
+               {
+                       VectorClear(color4f);
+                       color4f[3] = 1;
+               }
        }
 }
 
-void R_Shadow_VertexLightingWithXYAttenuationTexture(float *color, int numverts, const float *vertex, const float *normal, const float *lightcolor, const float *relativelightorigin, float lightradius, const float *zdir)
+void R_Shadow_VertexLightingWithXYAttenuationTexture(int numverts, const float *vertex3f, const float *normal3f, const float *lightcolor, const matrix4x4_t *m)
 {
-       float dist, dot, intensity, iradius = 1.0f / lightradius, v[3];
-       for (;numverts > 0;numverts--, vertex += 4, color += 4, normal += 4)
+       float *color4f = varray_color4f;
+       float dist, dot, intensity, v[3], n[3];
+       for (;numverts > 0;numverts--, vertex3f += 3, normal3f += 3, color4f += 4)
        {
-               VectorSubtract(vertex, relativelightorigin, v);
-               if ((dot = DotProduct(normal, v)) > 0 && (dist = fabs(DotProduct(zdir, v))) < lightradius)
+               Matrix4x4_Transform(m, vertex3f, v);
+               if ((dist = fabs(v[2])) < 1)
                {
-                       intensity = pow(1 - (dist * iradius), r_shadow_attenpower) * r_shadow_attenscale * dot / sqrt(DotProduct(v,v));
-                       VectorScale(lightcolor, intensity, color);
+                       Matrix4x4_Transform3x3(m, normal3f, n);
+                       if ((dot = DotProduct(n, v)) > 0)
+                       {
+                               intensity = pow(1 - dist, r_shadow_attenpower) * r_shadow_attenscale * dot / sqrt(DotProduct(n,n));
+                               VectorScale(lightcolor, intensity, color4f);
+                               color4f[3] = 1;
+                       }
+                       else
+                       {
+                               VectorClear(color4f);
+                               color4f[3] = 1;
+                       }
                }
                else
-                       VectorClear(color);
+               {
+                       VectorClear(color4f);
+                       color4f[3] = 1;
+               }
        }
 }
 
 // FIXME: this should be done in a vertex program when possible
 // FIXME: if vertex program not available, this would really benefit from 3DNow! or SSE
-void R_Shadow_TransformVertices(float *out, int numverts, const float *vertex, const matrix4x4_t *matrix)
+void R_Shadow_Transform_Vertex3f_TexCoord3f(float *tc3f, int numverts, const float *vertex3f, const matrix4x4_t *matrix)
+{
+       do
+       {
+               tc3f[0] = vertex3f[0] * matrix->m[0][0] + vertex3f[1] * matrix->m[0][1] + vertex3f[2] * matrix->m[0][2] + matrix->m[0][3];
+               tc3f[1] = vertex3f[0] * matrix->m[1][0] + vertex3f[1] * matrix->m[1][1] + vertex3f[2] * matrix->m[1][2] + matrix->m[1][3];
+               tc3f[2] = vertex3f[0] * matrix->m[2][0] + vertex3f[1] * matrix->m[2][1] + vertex3f[2] * matrix->m[2][2] + matrix->m[2][3];
+               vertex3f += 3;
+               tc3f += 3;
+       }
+       while (--numverts);
+}
+
+void R_Shadow_Transform_Vertex3f_TexCoord2f(float *tc2f, int numverts, const float *vertex3f, const matrix4x4_t *matrix)
 {
        do
        {
-               Matrix4x4_Transform(matrix, vertex, out);
-               vertex += 4;
-               out += 4;
+               tc2f[0] = vertex3f[0] * matrix->m[0][0] + vertex3f[1] * matrix->m[0][1] + vertex3f[2] * matrix->m[0][2] + matrix->m[0][3];
+               tc2f[1] = vertex3f[0] * matrix->m[1][0] + vertex3f[1] * matrix->m[1][1] + vertex3f[2] * matrix->m[1][2] + matrix->m[1][3];
+               vertex3f += 3;
+               tc2f += 2;
        }
        while (--numverts);
 }
 
-void R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(float *out, int numverts, const float *vertex, const float *svectors, const float *tvectors, const float *normals, const vec3_t relativelightorigin)
+void R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(float *out3f, int numverts, const float *vertex3f, const float *svector3f, const float *tvector3f, const float *normal3f, const vec3_t relativelightorigin)
 {
        int i;
        float lightdir[3];
-       for (i = 0;i < numverts;i++, vertex += 4, svectors += 4, tvectors += 4, normals += 4, out += 4)
+       for (i = 0;i < numverts;i++, vertex3f += 3, svector3f += 3, tvector3f += 3, normal3f += 3, out3f += 3)
        {
-               VectorSubtract(vertex, relativelightorigin, lightdir);
+               VectorSubtract(vertex3f, relativelightorigin, lightdir);
                // the cubemap normalizes this for us
-               out[0] = DotProduct(svectors, lightdir);
-               out[1] = DotProduct(tvectors, lightdir);
-               out[2] = DotProduct(normals, lightdir);
+               out3f[0] = DotProduct(svector3f, lightdir);
+               out3f[1] = DotProduct(tvector3f, lightdir);
+               out3f[2] = DotProduct(normal3f, lightdir);
        }
 }
 
-void R_Shadow_GenTexCoords_Specular_NormalCubeMap(float *out, int numverts, const float *vertex, const float *svectors, const float *tvectors, const float *normals, const vec3_t relativelightorigin, const vec3_t relativeeyeorigin)
+void R_Shadow_GenTexCoords_Specular_NormalCubeMap(float *out3f, int numverts, const float *vertex3f, const float *svector3f, const float *tvector3f, const float *normal3f, const vec3_t relativelightorigin, const vec3_t relativeeyeorigin)
 {
        int i;
        float lightdir[3], eyedir[3], halfdir[3];
-       for (i = 0;i < numverts;i++, vertex += 4, svectors += 4, tvectors += 4, normals += 4, out += 4)
+       for (i = 0;i < numverts;i++, vertex3f += 3, svector3f += 3, tvector3f += 3, normal3f += 3, out3f += 3)
        {
-               VectorSubtract(vertex, relativelightorigin, lightdir);
+               VectorSubtract(vertex3f, relativelightorigin, lightdir);
                VectorNormalizeFast(lightdir);
-               VectorSubtract(vertex, relativeeyeorigin, eyedir);
+               VectorSubtract(vertex3f, relativeeyeorigin, eyedir);
                VectorNormalizeFast(eyedir);
                VectorAdd(lightdir, eyedir, halfdir);
                // the cubemap normalizes this for us
-               out[0] = DotProduct(svectors, halfdir);
-               out[1] = DotProduct(tvectors, halfdir);
-               out[2] = DotProduct(normals, halfdir);
+               out3f[0] = DotProduct(svector3f, halfdir);
+               out3f[1] = DotProduct(tvector3f, halfdir);
+               out3f[2] = DotProduct(normal3f, halfdir);
        }
 }
 
-void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *elements, const float *vertices, const float *svectors, const float *tvectors, const float *normals, const float *texcoords, const float *relativelightorigin, float lightradius, const float *lightcolor, const matrix4x4_t *matrix_modeltofilter, const matrix4x4_t *matrix_modeltoattenuationxyz, const matrix4x4_t *matrix_modeltoattenuationz, rtexture_t *basetexture, rtexture_t *bumptexture, rtexture_t *lightcubemap)
+void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *elements, const float *vertex3f, const float *svector3f, const float *tvector3f, const float *normal3f, const float *texcoord2f, const float *relativelightorigin, float lightradius, const float *lightcolor, const matrix4x4_t *matrix_modeltofilter, const matrix4x4_t *matrix_modeltoattenuationxyz, const matrix4x4_t *matrix_modeltoattenuationz, rtexture_t *basetexture, rtexture_t *bumptexture, rtexture_t *lightcubemap)
 {
        int renders;
-       float color[3];
+       float color[3], color2[3];
        rmeshstate_t m;
        memset(&m, 0, sizeof(m));
-       if (gl_dot3arb)
+       if (gl_dot3arb && gl_texturecubemap && gl_combine.integer && gl_stencil)
        {
                if (!bumptexture)
                        bumptexture = r_shadow_blankbumptexture;
@@ -1088,9 +1116,9 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                // limit mult to 64 for sanity sake
                if (r_shadow_texture3d.integer && r_textureunits.integer >= 4)
                {
-                       // 3/2 3D combine path
+                       // 3/2 3D combine path (Geforce3, Radeon 8500)
                        m.tex[0] = R_GetTexture(bumptexture);
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.tex3d[2] = R_GetTexture(r_shadow_attenuation3dtexture);
                        m.texcombinergb[0] = GL_REPLACE;
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
@@ -1099,10 +1127,10 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglDisable(GL_BLEND);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin);
-                       R_Shadow_TransformVertices(varray_texcoord[2], numverts, vertices, matrix_modeltoattenuationxyz);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin);
+                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[2], numverts, vertex3f, matrix_modeltoattenuationxyz);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1118,15 +1146,18 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
                        qglEnable(GL_BLEND);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
                                if (lightcubemap)
-                                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltofilter);
+                                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltofilter);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1134,31 +1165,31 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                }
                else if (r_shadow_texture3d.integer && r_textureunits.integer >= 2 && lightcubemap)
                {
-                       // 1/2/2 3D combine path
+                       // 1/2/2 3D combine path (original Radeon)
                        m.tex3d[0] = R_GetTexture(r_shadow_attenuation3dtexture);
                        R_Mesh_TextureState(&m);
                        qglColorMask(0,0,0,1);
                        qglDisable(GL_BLEND);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       R_Shadow_TransformVertices(varray_texcoord[0], numverts, vertices, matrix_modeltoattenuationxyz);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[0], numverts, vertex3f, matrix_modeltoattenuationxyz);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
 
                        m.tex[0] = R_GetTexture(bumptexture);
                        m.tex3d[0] = 0;
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[0] = GL_REPLACE;
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        R_Mesh_TextureState(&m);
                        qglBlendFunc(GL_DST_ALPHA, GL_ZERO);
                        qglEnable(GL_BLEND);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1171,15 +1202,18 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglColorMask(1,1,1,0);
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
                                if (lightcubemap)
-                                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltofilter);
+                                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltofilter);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1187,10 +1221,9 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                }
                else if (r_shadow_texture3d.integer && r_textureunits.integer >= 2 && !lightcubemap)
                {
-                       // 2/2 3D combine path
+                       // 2/2 3D combine path (original Radeon)
                        m.tex[0] = R_GetTexture(bumptexture);
-                       m.tex3d[0] = 0;
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[0] = GL_REPLACE;
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        R_Mesh_TextureState(&m);
@@ -1198,15 +1231,16 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglColorMask(0,0,0,1);
                        qglDisable(GL_BLEND);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
 
                        m.tex[0] = R_GetTexture(basetexture);
                        m.tex3d[1] = R_GetTexture(r_shadow_attenuation3dtexture);
+                       m.texcubemap[1] = 0;
                        m.texcombinergb[0] = GL_MODULATE;
                        m.texcombinergb[1] = GL_MODULATE;
                        R_Mesh_TextureState(&m);
@@ -1214,14 +1248,17 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
                        qglEnable(GL_BLEND);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                               R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltoattenuationxyz);
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                               R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltoattenuationxyz);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1229,9 +1266,9 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                }
                else if (r_textureunits.integer >= 4)
                {
-                       // 4/2 2D combine path
+                       // 4/2 2D combine path (Geforce3, Radeon 8500)
                        m.tex[0] = R_GetTexture(bumptexture);
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[0] = GL_REPLACE;
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        m.tex[2] = R_GetTexture(r_shadow_attenuation2dtexture);
@@ -1241,11 +1278,11 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglDisable(GL_BLEND);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin);
-                       R_Shadow_TransformVertices(varray_texcoord[2], numverts, vertices, matrix_modeltoattenuationxyz);
-                       R_Shadow_TransformVertices(varray_texcoord[3], numverts, vertices, matrix_modeltoattenuationz);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin);
+                       R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[2], numverts, vertex3f, matrix_modeltoattenuationxyz);
+                       R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[3], numverts, vertex3f, matrix_modeltoattenuationz);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1261,15 +1298,18 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
                        qglEnable(GL_BLEND);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
                                if (lightcubemap)
-                                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltofilter);
+                                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltofilter);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1277,7 +1317,7 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                }
                else
                {
-                       // 2/2/2 2D combine path
+                       // 2/2/2 2D combine path (any dot3 card)
                        m.tex[0] = R_GetTexture(r_shadow_attenuation2dtexture);
                        m.tex[1] = R_GetTexture(r_shadow_attenuation2dtexture);
                        R_Mesh_TextureState(&m);
@@ -1285,25 +1325,25 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglDisable(GL_BLEND);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       R_Shadow_TransformVertices(varray_texcoord[0], numverts, vertices, matrix_modeltoattenuationxyz);
-                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltoattenuationz);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[0], numverts, vertex3f, matrix_modeltoattenuationxyz);
+                       R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[1], numverts, vertex3f, matrix_modeltoattenuationz);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
 
                        m.tex[0] = R_GetTexture(bumptexture);
                        m.tex[1] = 0;
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[0] = GL_REPLACE;
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        R_Mesh_TextureState(&m);
                        qglBlendFunc(GL_DST_ALPHA, GL_ZERO);
                        qglEnable(GL_BLEND);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Diffuse_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1316,15 +1356,18 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        qglColorMask(1,1,1,0);
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
                                if (lightcubemap)
-                                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltofilter);
+                                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltofilter);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1349,14 +1392,20 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        m.blendfunc2 = GL_ONE;
                        R_Mesh_State(&m);
 #endif
-                       GL_UseColorArray();
-                       R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltoattenuationxyz);
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       R_Shadow_VertexLightingWithXYAttenuationTexture(varray_color, numverts, vertices, normals, color, relativelightorigin, lightradius, matrix_modeltofilter->m[2]);
-                       R_Mesh_Draw(numverts, numtriangles, elements);
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
+                       {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
+                               GL_UseColorArray();
+                               R_Mesh_GetSpace(numverts);
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                               R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[1], numverts, vertex3f, matrix_modeltoattenuationxyz);
+                               R_Shadow_VertexLightingWithXYAttenuationTexture(numverts, vertex3f, normal3f, color, matrix_modeltofilter);
+                               R_Mesh_Draw(numverts, numtriangles, elements);
+                       }
                }
                else
                {
@@ -1372,23 +1421,29 @@ void R_Shadow_DiffuseLighting(int numverts, int numtriangles, const int *element
                        m.blendfunc2 = GL_ONE;
                        R_Mesh_State(&m);
 #endif
-                       GL_UseColorArray();
-                       R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color);
-                       R_Shadow_VertexLighting(varray_color, numverts, vertices, normals, color, relativelightorigin, lightradius);
-                       R_Mesh_Draw(numverts, numtriangles, elements);
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
+                       {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
+                               GL_UseColorArray();
+                               R_Mesh_GetSpace(numverts);
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                               R_Shadow_VertexLighting(numverts, vertex3f, normal3f, color, matrix_modeltofilter);
+                               R_Mesh_Draw(numverts, numtriangles, elements);
+                       }
                }
        }
 }
 
-void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elements, const float *vertices, const float *svectors, const float *tvectors, const float *normals, const float *texcoords, const float *relativelightorigin, const float *relativeeyeorigin, float lightradius, const float *lightcolor, const matrix4x4_t *matrix_modeltofilter, const matrix4x4_t *matrix_modeltoattenuationxyz, const matrix4x4_t *matrix_modeltoattenuationz, rtexture_t *glosstexture, rtexture_t *bumptexture, rtexture_t *lightcubemap)
+void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elements, const float *vertex3f, const float *svector3f, const float *tvector3f, const float *normal3f, const float *texcoord2f, const float *relativelightorigin, const float *relativeeyeorigin, float lightradius, const float *lightcolor, const matrix4x4_t *matrix_modeltofilter, const matrix4x4_t *matrix_modeltoattenuationxyz, const matrix4x4_t *matrix_modeltoattenuationz, rtexture_t *glosstexture, rtexture_t *bumptexture, rtexture_t *lightcubemap)
 {
        int renders;
-       float color[3];
+       float color[3], color2[3];
        rmeshstate_t m;
-       if (!gl_dot3arb)
+       if (!gl_dot3arb || !gl_texturecubemap || !gl_combine.integer || !gl_stencil)
                return;
        memset(&m, 0, sizeof(m));
        if (!bumptexture)
@@ -1399,18 +1454,20 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
        {
                if (r_shadow_texture3d.integer && r_textureunits.integer >= 2 && lightcubemap /*&& gl_support_blendsquare*/) // FIXME: detect blendsquare!
                {
-                       // 2/0/0/0/1/2 3D combine blendsquare path
+                       // 2/0/0/1/2 3D combine blendsquare path
                        m.tex[0] = R_GetTexture(bumptexture);
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        R_Mesh_TextureState(&m);
                        qglColorMask(0,0,0,1);
-                       qglDisable(GL_BLEND);
+                       // this squares the result
+                       qglEnable(GL_BLEND);
+                       qglBlendFunc(GL_SRC_ALPHA, GL_ZERO);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Specular_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin, relativeeyeorigin);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Specular_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin, relativeeyeorigin);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1421,15 +1478,14 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        R_Mesh_TextureState(&m);
                        // square alpha in framebuffer a few times to make it shiny
                        qglBlendFunc(GL_ZERO, GL_DST_ALPHA);
-                       qglEnable(GL_BLEND);
                        // these comments are a test run through this math for intensity 0.5
-                       // 0.5 * 0.5 = 0.25
-                       // 0.25 * 0.25 = 0.0625
-                       // 0.0625 * 0.0625 = 0.00390625
-                       for (renders = 0;renders < 3;renders++)
+                       // 0.5 * 0.5 = 0.25 (done by the BlendFunc earlier)
+                       // 0.25 * 0.25 = 0.0625 (this is another pass)
+                       // 0.0625 * 0.0625 = 0.00390625 (this is another pass)
+                       for (renders = 0;renders < 2;renders++)
                        {
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                        }
                        c_rt_lightmeshes += 3;
@@ -1439,8 +1495,8 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        R_Mesh_TextureState(&m);
                        qglBlendFunc(GL_DST_ALPHA, GL_ZERO);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       R_Shadow_TransformVertices(varray_texcoord[0], numverts, vertices, matrix_modeltoattenuationxyz);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[0], numverts, vertex3f, matrix_modeltoattenuationxyz);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1452,15 +1508,18 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        qglColorMask(1,1,1,0);
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value * 0.25f, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value * 0.25f, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
                                if (lightcubemap)
-                                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltofilter);
+                                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltofilter);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1468,18 +1527,20 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                }
                else if (r_shadow_texture3d.integer && r_textureunits.integer >= 2 && !lightcubemap /*&& gl_support_blendsquare*/) // FIXME: detect blendsquare!
                {
-                       // 2/0/0/0/2 3D combine blendsquare path
+                       // 2/0/0/2 3D combine blendsquare path
                        m.tex[0] = R_GetTexture(bumptexture);
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        R_Mesh_TextureState(&m);
                        qglColorMask(0,0,0,1);
-                       qglDisable(GL_BLEND);
+                       // this squares the result
+                       qglEnable(GL_BLEND);
+                       qglBlendFunc(GL_SRC_ALPHA, GL_ZERO);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Specular_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin, relativeeyeorigin);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Specular_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin, relativeeyeorigin);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1490,36 +1551,40 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        R_Mesh_TextureState(&m);
                        // square alpha in framebuffer a few times to make it shiny
                        qglBlendFunc(GL_ZERO, GL_DST_ALPHA);
-                       qglEnable(GL_BLEND);
                        // these comments are a test run through this math for intensity 0.5
-                       // 0.5 * 0.5 = 0.25
-                       // 0.25 * 0.25 = 0.0625
-                       // 0.0625 * 0.0625 = 0.00390625
-                       for (renders = 0;renders < 3;renders++)
+                       // 0.5 * 0.5 = 0.25 (done by the BlendFunc earlier)
+                       // 0.25 * 0.25 = 0.0625 (this is another pass)
+                       // 0.0625 * 0.0625 = 0.00390625 (this is another pass)
+                       for (renders = 0;renders < 2;renders++)
                        {
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                        }
                        c_rt_lightmeshes += 3;
                        c_rt_lighttris += numtriangles * 3;
 
-                       m.tex3d[0] = R_GetTexture(r_shadow_attenuation3dtexture);
-                       m.tex[1] = R_GetTexture(glosstexture);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       m.tex[0] = R_GetTexture(glosstexture);
+                       m.tex3d[1] = R_GetTexture(r_shadow_attenuation3dtexture);
                        R_Mesh_TextureState(&m);
                        qglColorMask(1,1,1,0);
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value * 0.25f, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value * 0.25f, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               R_Shadow_TransformVertices(varray_texcoord[0], numverts, vertices, matrix_modeltoattenuationxyz);
-                               memcpy(varray_texcoord[1], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                               R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltoattenuationxyz);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1527,18 +1592,20 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                }
                else if (r_textureunits.integer >= 2 /*&& gl_support_blendsquare*/) // FIXME: detect blendsquare!
                {
-                       // 2/0/0/0/2/2 2D combine blendsquare path
+                       // 2/0/0/2/2 2D combine blendsquare path
                        m.tex[0] = R_GetTexture(bumptexture);
-                       m.texcubemap[1] = R_GetTexture(r_shadow_normalscubetexture);
+                       m.texcubemap[1] = R_GetTexture(r_shadow_normalcubetexture);
                        m.texcombinergb[1] = GL_DOT3_RGBA_ARB;
                        R_Mesh_TextureState(&m);
                        qglColorMask(0,0,0,1);
-                       qglDisable(GL_BLEND);
+                       // this squares the result
+                       qglEnable(GL_BLEND);
+                       qglBlendFunc(GL_SRC_ALPHA, GL_ZERO);
                        GL_Color(1,1,1,1);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
-                       R_Shadow_GenTexCoords_Specular_NormalCubeMap(varray_texcoord[1], numverts, vertices, svectors, tvectors, normals, relativelightorigin, relativeeyeorigin);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
+                       R_Shadow_GenTexCoords_Specular_NormalCubeMap(varray_texcoord3f[1], numverts, vertex3f, svector3f, tvector3f, normal3f, relativelightorigin, relativeeyeorigin);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1549,15 +1616,14 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        R_Mesh_TextureState(&m);
                        // square alpha in framebuffer a few times to make it shiny
                        qglBlendFunc(GL_ZERO, GL_DST_ALPHA);
-                       qglEnable(GL_BLEND);
                        // these comments are a test run through this math for intensity 0.5
-                       // 0.5 * 0.5 = 0.25
-                       // 0.25 * 0.25 = 0.0625
-                       // 0.0625 * 0.0625 = 0.00390625
-                       for (renders = 0;renders < 3;renders++)
+                       // 0.5 * 0.5 = 0.25 (done by the BlendFunc earlier)
+                       // 0.25 * 0.25 = 0.0625 (this is another pass)
+                       // 0.0625 * 0.0625 = 0.00390625 (this is another pass)
+                       for (renders = 0;renders < 2;renders++)
                        {
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                        }
                        c_rt_lightmeshes += 3;
@@ -1568,9 +1634,9 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        R_Mesh_TextureState(&m);
                        qglBlendFunc(GL_DST_ALPHA, GL_ZERO);
                        R_Mesh_GetSpace(numverts);
-                       memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                       R_Shadow_TransformVertices(varray_texcoord[0], numverts, vertices, matrix_modeltoattenuationxyz);
-                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltoattenuationz);
+                       R_Mesh_CopyVertex3f(vertex3f, numverts);
+                       R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[0], numverts, vertex3f, matrix_modeltoattenuationxyz);
+                       R_Shadow_Transform_Vertex3f_TexCoord2f(varray_texcoord2f[1], numverts, vertex3f, matrix_modeltoattenuationz);
                        R_Mesh_Draw(numverts, numtriangles, elements);
                        c_rt_lightmeshes++;
                        c_rt_lighttris += numtriangles;
@@ -1581,15 +1647,18 @@ void R_Shadow_SpecularLighting(int numverts, int numtriangles, const int *elemen
                        qglColorMask(1,1,1,0);
                        qglBlendFunc(GL_DST_ALPHA, GL_ONE);
 
-                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value * 0.25f, color);
-                       for (renders = 0;renders < 64 && (color[0] > 0 || color[1] > 0 || color[2] > 0);renders++, color[0] = max(0, color[0] - 1.0f), color[1] = max(0, color[1] - 1.0f), color[2] = max(0, color[2] - 1.0f))
+                       VectorScale(lightcolor, r_colorscale * r_shadow_lightintensityscale.value * 0.25f, color2);
+                       for (renders = 0;renders < 64 && (color2[0] > 0 || color2[1] > 0 || color2[2] > 0);renders++, color2[0]--, color2[1]--, color2[2]--)
                        {
+                               color[0] = bound(0, color2[0], 1);
+                               color[1] = bound(0, color2[1], 1);
+                               color[2] = bound(0, color2[2], 1);
                                GL_Color(color[0], color[1], color[2], 1);
                                R_Mesh_GetSpace(numverts);
-                               memcpy(varray_vertex, vertices, numverts * sizeof(float[4]));
-                               memcpy(varray_texcoord[0], texcoords, numverts * sizeof(float[4]));
+                               R_Mesh_CopyVertex3f(vertex3f, numverts);
+                               R_Mesh_CopyTexCoord2f(0, texcoord2f, numverts);
                                if (lightcubemap)
-                                       R_Shadow_TransformVertices(varray_texcoord[1], numverts, vertices, matrix_modeltofilter);
+                                       R_Shadow_Transform_Vertex3f_TexCoord3f(varray_texcoord3f[1], numverts, vertex3f, matrix_modeltofilter);
                                R_Mesh_Draw(numverts, numtriangles, elements);
                                c_rt_lightmeshes++;
                                c_rt_lighttris += numtriangles;
@@ -1619,8 +1688,8 @@ vec3_t r_editlights_cursorlocation;
 static int castshadowcount = 1;
 void R_Shadow_NewWorldLight(vec3_t origin, float radius, vec3_t color, int style, const char *cubemapname, int castshadow)
 {
-       int i, j, k, l, maxverts, *mark, tris;
-       float *verts;
+       int i, j, k, l, maxverts = 256, *mark, tris;
+       float *vertex3f = NULL;
        worldlight_t *e;
        shadowmesh_t *mesh, *castmesh;
        mleaf_t *leaf;
@@ -1749,8 +1818,6 @@ void R_Shadow_NewWorldLight(vec3_t origin, float radius, vec3_t color, int style
 
                if (e->castshadows)
                {
-                       maxverts = 256;
-                       verts = NULL;
                        castshadowcount++;
                        for (j = 0;j < e->numsurfaces;j++)
                        {
@@ -1768,7 +1835,7 @@ void R_Shadow_NewWorldLight(vec3_t origin, float radius, vec3_t color, int style
                        for (j = 0;j < e->numsurfaces;j++)
                                if (e->surfaces[j]->castshadow == castshadowcount)
                                        for (surfmesh = e->surfaces[j]->mesh;surfmesh;surfmesh = surfmesh->chain)
-                                               Mod_ShadowMesh_AddMesh(r_shadow_mempool, castmesh, surfmesh->numverts, surfmesh->verts, surfmesh->numtriangles, surfmesh->index);
+                                               Mod_ShadowMesh_AddMesh(r_shadow_mempool, castmesh, surfmesh->numverts, surfmesh->vertex3f, surfmesh->numtriangles, surfmesh->element3i);
                        castmesh = Mod_ShadowMesh_Finish(r_shadow_mempool, castmesh);
 
                        // cast shadow volume from castmesh
@@ -1780,24 +1847,23 @@ void R_Shadow_NewWorldLight(vec3_t origin, float radius, vec3_t color, int style
                                if (maxverts < castmesh->numverts * 2)
                                {
                                        maxverts = castmesh->numverts * 2;
-                                       if (verts)
-                                               Mem_Free(verts);
-                                       verts = NULL;
+                                       if (vertex3f)
+                                               Mem_Free(vertex3f);
+                                       vertex3f = NULL;
                                }
-                               if (verts == NULL && maxverts > 0)
-                                       verts = Mem_Alloc(r_shadow_mempool, maxverts * sizeof(float[4]));
+                               if (vertex3f == NULL && maxverts > 0)
+                                       vertex3f = Mem_Alloc(r_shadow_mempool, maxverts * sizeof(float[3]));
 
                                // now that we have the buffers big enough, construct shadow volume mesh
-                               memcpy(verts, castmesh->verts, castmesh->numverts * sizeof(float[4]));
-                               R_Shadow_ProjectVertices(verts, castmesh->numverts, e->origin, r_shadow_projectdistance.value);//, e->lightradius);
-                               tris = R_Shadow_MakeTriangleShadowFlags(castmesh->elements, verts, castmesh->numtriangles, trianglefacinglight, trianglefacinglightlist, e->origin);
-                               tris = R_Shadow_BuildShadowVolumeTriangles(castmesh->elements, castmesh->neighbors, castmesh->numverts, trianglefacinglight, trianglefacinglightlist, tris, shadowelements);
+                               memcpy(vertex3f, castmesh->vertex3f, castmesh->numverts * sizeof(float[3]));
+                               tris = R_Shadow_MakeTriangleShadowFlags_Vertex3f(castmesh->element3i, vertex3f, castmesh->numtriangles, trianglefacinglight, trianglefacinglightlist, e->origin);
+                               tris = R_Shadow_BuildShadowVolume(castmesh->element3i, castmesh->neighbor3i, castmesh->numverts, trianglefacinglight, trianglefacinglightlist, tris, shadowelements, vertex3f, e->origin, r_shadow_projectdistance.value);
                                // add the constructed shadow volume mesh
-                               Mod_ShadowMesh_AddMesh(r_shadow_mempool, e->shadowvolume, castmesh->numverts, verts, tris, shadowelements);
+                               Mod_ShadowMesh_AddMesh(r_shadow_mempool, e->shadowvolume, castmesh->numverts, vertex3f, tris, shadowelements);
                        }
-                       if (verts)
-                               Mem_Free(verts);
-                       verts = NULL;
+                       if (vertex3f)
+                               Mem_Free(vertex3f);
+                       vertex3f = NULL;
                        // we're done with castmesh now
                        Mod_ShadowMesh_Free(castmesh);
                        e->shadowvolume = Mod_ShadowMesh_Finish(r_shadow_mempool, e->shadowvolume);
@@ -1863,24 +1929,7 @@ void R_DrawLightSprite(int texnum, const vec3_t origin, vec_t scale, float cr, f
        R_Mesh_State(&m);
 
        GL_Color(cr * r_colorscale, cg * r_colorscale, cb * r_colorscale, ca);
-       R_Mesh_GetSpace(4);
-       varray_texcoord[0][ 0] = 0;varray_texcoord[0][ 1] = 0;
-       varray_texcoord[0][ 4] = 0;varray_texcoord[0][ 5] = 1;
-       varray_texcoord[0][ 8] = 1;varray_texcoord[0][ 9] = 1;
-       varray_texcoord[0][12] = 1;varray_texcoord[0][13] = 0;
-       varray_vertex[0] = origin[0] - vright[0] * scale - vup[0] * scale;
-       varray_vertex[1] = origin[1] - vright[1] * scale - vup[1] * scale;
-       varray_vertex[2] = origin[2] - vright[2] * scale - vup[2] * scale;
-       varray_vertex[4] = origin[0] - vright[0] * scale + vup[0] * scale;
-       varray_vertex[5] = origin[1] - vright[1] * scale + vup[1] * scale;
-       varray_vertex[6] = origin[2] - vright[2] * scale + vup[2] * scale;
-       varray_vertex[8] = origin[0] + vright[0] * scale + vup[0] * scale;
-       varray_vertex[9] = origin[1] + vright[1] * scale + vup[1] * scale;
-       varray_vertex[10] = origin[2] + vright[2] * scale + vup[2] * scale;
-       varray_vertex[12] = origin[0] + vright[0] * scale - vup[0] * scale;
-       varray_vertex[13] = origin[1] + vright[1] * scale - vup[1] * scale;
-       varray_vertex[14] = origin[2] + vright[2] * scale - vup[2] * scale;
-       R_Mesh_Draw(4, 2, polygonelements);
+       R_DrawSpriteMesh(origin, vright, vup, scale, -scale, -scale, scale);
 }
 
 void R_Shadow_DrawCursorCallback(const void *calldata1, int calldata2)
@@ -1958,9 +2007,9 @@ void R_Shadow_LoadWorldLights(void)
                Con_Printf("No map loaded.\n");
                return;
        }
-       COM_StripExtension(cl.worldmodel->name, name);
+       FS_StripExtension(cl.worldmodel->name, name);
        strcat(name, ".rtlights");
-       lightsstring = COM_LoadFile(name, false);
+       lightsstring = FS_LoadFile(name, false);
        if (lightsstring)
        {
                s = lightsstring;
@@ -2015,7 +2064,7 @@ void R_Shadow_SaveWorldLights(void)
                Con_Printf("No map loaded.\n");
                return;
        }
-       COM_StripExtension(cl.worldmodel->name, name);
+       FS_StripExtension(cl.worldmodel->name, name);
        strcat(name, ".rtlights");
        bufchars = bufmaxchars = 0;
        buf = NULL;
@@ -2041,7 +2090,7 @@ void R_Shadow_SaveWorldLights(void)
                }
        }
        if (bufchars)
-               COM_WriteFile(name, buf, bufchars);
+               FS_WriteFile(name, buf, bufchars);
        if (buf)
                Mem_Free(buf);
 }
@@ -2056,9 +2105,9 @@ void R_Shadow_LoadLightsFile(void)
                Con_Printf("No map loaded.\n");
                return;
        }
-       COM_StripExtension(cl.worldmodel->name, name);
+       FS_StripExtension(cl.worldmodel->name, name);
        strcat(name, ".lights");
-       lightsstring = COM_LoadFile(name, false);
+       lightsstring = FS_LoadFile(name, false);
        if (lightsstring)
        {
                s = lightsstring;
@@ -2249,18 +2298,6 @@ void R_Shadow_SetCursorLocationForView(void)
        r_editlights_cursorlocation[2] = floor(endpos[2] / r_editlights_cursorgrid.value + 0.5f) * r_editlights_cursorgrid.value;
 }
 
-void R_Shadow_UpdateLightingMode(void)
-{
-       r_shadow_lightingmode = 0;
-       if (r_shadow_realtime.integer)
-       {
-               if (r_shadow_worldlightchain)
-                       r_shadow_lightingmode = 2;
-               else
-                       r_shadow_lightingmode = 1;
-       }
-}
-
 void R_Shadow_UpdateWorldLightSelection(void)
 {
        R_Shadow_SetCursorLocationForView();