cvar_t r_viewscale_fpsscaling_stepmax = {CVAR_SAVE, "r_viewscale_fpsscaling_stepmax", "1.00", "largest adjustment to hit the target framerate (this value prevents wild overshooting of the estimate)"};
cvar_t r_viewscale_fpsscaling_target = {CVAR_SAVE, "r_viewscale_fpsscaling_target", "70", "desired framerate"};
+cvar_t r_glsl_skeletal = {CVAR_SAVE, "r_glsl_skeletal", "1", "render skeletal models faster using a gpu-skinning technique"};
cvar_t r_glsl_deluxemapping = {CVAR_SAVE, "r_glsl_deluxemapping", "1", "use per pixel lighting on deluxemap-compiled q3bsp maps (or a value of 2 forces deluxemap shading even without deluxemaps)"};
cvar_t r_glsl_offsetmapping = {CVAR_SAVE, "r_glsl_offsetmapping", "0", "offset mapping effect (also known as parallax mapping or virtual displacement mapping)"};
cvar_t r_glsl_offsetmapping_steps = {CVAR_SAVE, "r_glsl_offsetmapping_steps", "2", "offset mapping steps (note: too high values may be not supported by your GPU)"};
cvar_t r_batch_multidraw = {CVAR_SAVE, "r_batch_multidraw", "1", "issue multiple glDrawElements calls when rendering a batch of surfaces with the same texture (otherwise the index data is copied to make it one draw)"};
cvar_t r_batch_multidraw_mintriangles = {CVAR_SAVE, "r_batch_multidraw_mintriangles", "0", "minimum number of triangles to activate multidraw path (copying small groups of triangles may be faster)"};
+cvar_t r_batch_debugdynamicvertexpath = {CVAR_SAVE, "r_batch_debugdynamicvertexpath", "0", "force the dynamic batching code path for debugging purposes"};
cvar_t r_glsl_saturation = {CVAR_SAVE, "r_glsl_saturation", "1", "saturation multiplier (only working in glsl!)"};
cvar_t r_glsl_saturation_redcompensate = {CVAR_SAVE, "r_glsl_saturation_redcompensate", "0", "a 'vampire sight' addition to desaturation effect, does compensation for red color, r_glsl_restart is required"};
{"#define USEBOUNCEGRIDDIRECTIONAL\n", " bouncegriddirectional"}, // TODO make this a static parm
{"#define USETRIPPY\n", " trippy"},
{"#define USEDEPTHRGB\n", " depthrgb"},
- {"#define USEALPHAGENVERTEX\n", "alphagenvertex"}
+ {"#define USEALPHAGENVERTEX\n", " alphagenvertex"},
+ {"#define USESKELETAL\n", " skeletal"}
};
// NOTE: MUST MATCH ORDER OF SHADERMODE_* ENUMS!
int loc_ShadowMap_Parameters;
int loc_ShadowMap_TextureScale;
int loc_SpecularPower;
+ int loc_Skeletal_Transform12;
int loc_UserVec1;
int loc_UserVec2;
int loc_UserVec3;
p->loc_ShadowMap_Parameters = qglGetUniformLocation(p->program, "ShadowMap_Parameters");
p->loc_ShadowMap_TextureScale = qglGetUniformLocation(p->program, "ShadowMap_TextureScale");
p->loc_SpecularPower = qglGetUniformLocation(p->program, "SpecularPower");
+ p->loc_Skeletal_Transform12 = qglGetUniformLocation(p->program, "Skeletal_Transform12");
p->loc_UserVec1 = qglGetUniformLocation(p->program, "UserVec1");
p->loc_UserVec2 = qglGetUniformLocation(p->program, "UserVec2");
p->loc_UserVec3 = qglGetUniformLocation(p->program, "UserVec3");
R_Mesh_TexCoordPointer(2, 3, GL_FLOAT, sizeof(float[3]), rsurface.batchtvector3f, rsurface.batchtvector3f_vertexbuffer, rsurface.batchtvector3f_bufferoffset);
R_Mesh_TexCoordPointer(3, 3, GL_FLOAT, sizeof(float[3]), rsurface.batchnormal3f, rsurface.batchnormal3f_vertexbuffer, rsurface.batchnormal3f_bufferoffset);
R_Mesh_TexCoordPointer(4, 2, GL_FLOAT, sizeof(float[2]), rsurface.batchtexcoordlightmap2f, rsurface.batchtexcoordlightmap2f_vertexbuffer, rsurface.batchtexcoordlightmap2f_bufferoffset);
+ R_Mesh_TexCoordPointer(5, 2, GL_FLOAT, sizeof(float[2]), NULL, NULL, 0);
+ R_Mesh_TexCoordPointer(6, 4, GL_UNSIGNED_BYTE | 0x80000000, sizeof(unsigned char[4]), rsurface.batchskeletalindex4ub, rsurface.batchskeletalindex4ub_vertexbuffer, rsurface.batchskeletalindex4ub_bufferoffset);
+ R_Mesh_TexCoordPointer(7, 4, GL_UNSIGNED_BYTE, sizeof(unsigned char[4]), rsurface.batchskeletalweight4ub, rsurface.batchskeletalweight4ub_vertexbuffer, rsurface.batchskeletalweight4ub_bufferoffset);
}
else
{
- RSurf_PrepareVerticesForBatch(BATCHNEED_VERTEXMESH_VERTEX | BATCHNEED_VERTEXMESH_NORMAL | BATCHNEED_VERTEXMESH_VECTOR | (rsurface.modellightmapcolor4f ? BATCHNEED_VERTEXMESH_VERTEXCOLOR : 0) | BATCHNEED_VERTEXMESH_TEXCOORD | (rsurface.uselightmaptexture ? BATCHNEED_VERTEXMESH_LIGHTMAP : 0) | BATCHNEED_ALLOWMULTIDRAW, texturenumsurfaces, texturesurfacelist);
+ RSurf_PrepareVerticesForBatch(BATCHNEED_VERTEXMESH_VERTEX | BATCHNEED_VERTEXMESH_NORMAL | BATCHNEED_VERTEXMESH_VECTOR | (rsurface.modellightmapcolor4f ? BATCHNEED_VERTEXMESH_VERTEXCOLOR : 0) | BATCHNEED_VERTEXMESH_TEXCOORD | (rsurface.uselightmaptexture ? BATCHNEED_VERTEXMESH_LIGHTMAP : 0) | (rsurface.entityskeletaltransform3x4 ? BATCHNEED_VERTEXMESH_SKELETAL : 0) | BATCHNEED_ALLOWMULTIDRAW, texturenumsurfaces, texturesurfacelist);
R_Mesh_PrepareVertices_Mesh(rsurface.batchnumvertices, rsurface.batchvertexmesh, rsurface.batchvertexmeshbuffer);
}
+ // this has to be after RSurf_PrepareVerticesForBatch
+ if (rsurface.batchskeletaltransform3x4)
+ permutation |= SHADERPERMUTATION_SKELETAL;
R_SetupShader_SetPermutationGLSL(mode, permutation);
if (r_glsl_permutation->loc_ModelToReflectCube >= 0) {Matrix4x4_ToArrayFloatGL(&rsurface.matrix, m16f);qglUniformMatrix4fv(r_glsl_permutation->loc_ModelToReflectCube, 1, false, m16f);}
if (mode == SHADERMODE_LIGHTSOURCE)
}
}
if (r_glsl_permutation->tex_Texture_BounceGrid >= 0) R_Mesh_TexBind(r_glsl_permutation->tex_Texture_BounceGrid, r_shadow_bouncegridtexture);
+ if (r_glsl_permutation->loc_Skeletal_Transform12 >= 0 && rsurface.batchskeletalnumtransforms > 0)
+ qglUniform4fv(r_glsl_permutation->loc_Skeletal_Transform12, rsurface.batchskeletalnumtransforms*3, rsurface.batchskeletaltransform3x4);
CHECKGLERROR
break;
case RENDERPATH_GL11:
Cvar_RegisterVariable(&r_test);
Cvar_RegisterVariable(&r_batch_multidraw);
Cvar_RegisterVariable(&r_batch_multidraw_mintriangles);
+ Cvar_RegisterVariable(&r_batch_debugdynamicvertexpath);
+ Cvar_RegisterVariable(&r_glsl_skeletal);
Cvar_RegisterVariable(&r_glsl_saturation);
Cvar_RegisterVariable(&r_glsl_saturation_redcompensate);
Cvar_RegisterVariable(&r_glsl_vertextextureblend_usebothalphas);
ent->animcache_vertexmesh = NULL;
ent->animcache_vertex3fbuffer = NULL;
ent->animcache_vertexmeshbuffer = NULL;
+ ent->animcache_skeletaltransform3x4 = NULL;
}
}
{
dp_model_t *model = ent->model;
int numvertices;
+
+ // cache skeletal animation data first (primarily for gpu-skinning)
+ if (!ent->animcache_skeletaltransform3x4 && model->num_bones > 0 && model->surfmesh.data_skeletalindex4ub)
+ {
+ int i;
+ int blends;
+ const skeleton_t *skeleton = ent->skeleton;
+ const frameblend_t *frameblend = ent->frameblend;
+ float *boneposerelative;
+ float m[12];
+ static float bonepose[256][12];
+ ent->animcache_skeletaltransform3x4 = R_FrameData_Alloc(sizeof(float[3][4]) * model->num_bones);
+ boneposerelative = ent->animcache_skeletaltransform3x4;
+ if (skeleton && !skeleton->relativetransforms)
+ skeleton = NULL;
+ // resolve hierarchy and make relative transforms (deforms) which the shader wants
+ if (skeleton)
+ {
+ for (i = 0;i < model->num_bones;i++)
+ {
+ Matrix4x4_ToArray12FloatD3D(&skeleton->relativetransforms[i], m);
+ if (model->data_bones[i].parent >= 0)
+ R_ConcatTransforms(bonepose[model->data_bones[i].parent], m, bonepose[i]);
+ else
+ memcpy(bonepose[i], m, sizeof(m));
+
+ // create a relative deformation matrix to describe displacement
+ // from the base mesh, which is used by the actual weighting
+ R_ConcatTransforms(bonepose[i], model->data_baseboneposeinverse + i * 12, boneposerelative + i * 12);
+ }
+ }
+ else
+ {
+ for (i = 0;i < model->num_bones;i++)
+ {
+ const short * RESTRICT pose7s = model->data_poses7s + 7 * (frameblend[0].subframe * model->num_bones + i);
+ float lerp = frameblend[0].lerp,
+ tx = pose7s[0], ty = pose7s[1], tz = pose7s[2],
+ rx = pose7s[3] * lerp,
+ ry = pose7s[4] * lerp,
+ rz = pose7s[5] * lerp,
+ rw = pose7s[6] * lerp,
+ dx = tx*rw + ty*rz - tz*ry,
+ dy = -tx*rz + ty*rw + tz*rx,
+ dz = tx*ry - ty*rx + tz*rw,
+ dw = -tx*rx - ty*ry - tz*rz,
+ scale, sx, sy, sz, sw;
+ for (blends = 1;blends < MAX_FRAMEBLENDS && frameblend[blends].lerp > 0;blends++)
+ {
+ const short * RESTRICT pose7s = model->data_poses7s + 7 * (frameblend[blends].subframe * model->num_bones + i);
+ float lerp = frameblend[blends].lerp,
+ tx = pose7s[0], ty = pose7s[1], tz = pose7s[2],
+ qx = pose7s[3], qy = pose7s[4], qz = pose7s[5], qw = pose7s[6];
+ if(rx*qx + ry*qy + rz*qz + rw*qw < 0) lerp = -lerp;
+ qx *= lerp;
+ qy *= lerp;
+ qz *= lerp;
+ qw *= lerp;
+ rx += qx;
+ ry += qy;
+ rz += qz;
+ rw += qw;
+ dx += tx*qw + ty*qz - tz*qy;
+ dy += -tx*qz + ty*qw + tz*qx;
+ dz += tx*qy - ty*qx + tz*qw;
+ dw += -tx*qx - ty*qy - tz*qz;
+ }
+ scale = 1.0f / (rx*rx + ry*ry + rz*rz + rw*rw);
+ sx = rx * scale;
+ sy = ry * scale;
+ sz = rz * scale;
+ sw = rw * scale;
+ m[0] = sw*rw + sx*rx - sy*ry - sz*rz;
+ m[1] = 2*(sx*ry - sw*rz);
+ m[2] = 2*(sx*rz + sw*ry);
+ m[3] = model->num_posescale*(dx*sw - dy*sz + dz*sy - dw*sx);
+ m[4] = 2*(sx*ry + sw*rz);
+ m[5] = sw*rw + sy*ry - sx*rx - sz*rz;
+ m[6] = 2*(sy*rz - sw*rx);
+ m[7] = model->num_posescale*(dx*sz + dy*sw - dz*sx - dw*sy);
+ m[8] = 2*(sx*rz - sw*ry);
+ m[9] = 2*(sy*rz + sw*rx);
+ m[10] = sw*rw + sz*rz - sx*rx - sy*ry;
+ m[11] = model->num_posescale*(dy*sx + dz*sw - dx*sy - dw*sz);
+ if (i == r_skeletal_debugbone.integer)
+ m[r_skeletal_debugbonecomponent.integer % 12] += r_skeletal_debugbonevalue.value;
+ m[3] *= r_skeletal_debugtranslatex.value;
+ m[7] *= r_skeletal_debugtranslatey.value;
+ m[11] *= r_skeletal_debugtranslatez.value;
+ if (model->data_bones[i].parent >= 0)
+ R_ConcatTransforms(bonepose[model->data_bones[i].parent], m, bonepose[i]);
+ else
+ memcpy(bonepose[i], m, sizeof(m));
+ // create a relative deformation matrix to describe displacement
+ // from the base mesh, which is used by the actual weighting
+ R_ConcatTransforms(bonepose[i], model->data_baseboneposeinverse + i * 12, boneposerelative + i * 12);
+ }
+ }
+ }
+
// see if it's already cached this frame
if (ent->animcache_vertex3f)
{
// see if this ent is worth caching
if (!model || !model->Draw || !model->surfmesh.isanimated || !model->AnimateVertices)
return false;
+ // skip entity if the shader backend has a cheaper way
+ if (model->surfmesh.data_skeletalindex4ub && r_glsl_skeletal.integer)
+ {
+ switch (vid.renderpath)
+ {
+ case RENDERPATH_GL20:
+ return false;
+ case RENDERPATH_GL11:
+ case RENDERPATH_GL13:
+ case RENDERPATH_GLES1:
+ case RENDERPATH_GLES2:
+ case RENDERPATH_D3D9:
+ case RENDERPATH_D3D10:
+ case RENDERPATH_D3D11:
+ case RENDERPATH_SOFT:
+ break;
+ }
+ }
// get some memory for this entity and generate mesh data
numvertices = model->surfmesh.num_vertices;
ent->animcache_vertex3f = (float *)R_FrameData_Alloc(sizeof(float[3])*numvertices);
rsurface.ent_alttextures = false;
rsurface.basepolygonfactor = r_refdef.polygonfactor;
rsurface.basepolygonoffset = r_refdef.polygonoffset;
+ rsurface.entityskeletaltransform3x4 = NULL;
+ rsurface.entityskeletalnumtransforms = 0;
rsurface.modelvertex3f = model->surfmesh.data_vertex3f;
rsurface.modelvertex3f_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
rsurface.modelvertex3f_bufferoffset = model->surfmesh.vbooffset_vertex3f;
rsurface.modeltexcoordlightmap2f = model->surfmesh.data_texcoordlightmap2f;
rsurface.modeltexcoordlightmap2f_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
rsurface.modeltexcoordlightmap2f_bufferoffset = model->surfmesh.vbooffset_texcoordlightmap2f;
+ rsurface.modelskeletalindex4ub = model->surfmesh.data_skeletalindex4ub;
+ rsurface.modelskeletalindex4ub_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
+ rsurface.modelskeletalindex4ub_bufferoffset = model->surfmesh.vbooffset_skeletalindex4ub;
+ rsurface.modelskeletalweight4ub = model->surfmesh.data_skeletalweight4ub;
+ rsurface.modelskeletalweight4ub_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
+ rsurface.modelskeletalweight4ub_bufferoffset = model->surfmesh.vbooffset_skeletalweight4ub;
rsurface.modelelement3i = model->surfmesh.data_element3i;
rsurface.modelelement3i_indexbuffer = model->surfmesh.data_element3i_indexbuffer;
rsurface.modelelement3i_bufferoffset = model->surfmesh.data_element3i_bufferoffset;
rsurface.batchtexcoordlightmap2f = NULL;
rsurface.batchtexcoordlightmap2f_vertexbuffer = NULL;
rsurface.batchtexcoordlightmap2f_bufferoffset = 0;
+ rsurface.batchskeletalindex4ub = NULL;
+ rsurface.batchskeletalindex4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalindex4ub_bufferoffset = 0;
+ rsurface.batchskeletalweight4ub = NULL;
+ rsurface.batchskeletalweight4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalweight4ub_bufferoffset = 0;
rsurface.batchvertexmesh = NULL;
rsurface.batchvertexmeshbuffer = NULL;
rsurface.batchvertex3fbuffer = NULL;
rsurface.basepolygonfactor += r_polygonoffset_submodel_factor.value;
rsurface.basepolygonoffset += r_polygonoffset_submodel_offset.value;
}
- if (model->surfmesh.isanimated && model->AnimateVertices)
+ // if the animcache code decided it should use the shader path, skip the deform step
+ rsurface.entityskeletaltransform3x4 = ent->animcache_vertex3f ? NULL : ent->animcache_skeletaltransform3x4;
+ rsurface.entityskeletalnumtransforms = rsurface.entityskeletaltransform3x4 ? model->num_bones : 0;
+ if (model->surfmesh.isanimated && model->AnimateVertices && !rsurface.entityskeletaltransform3x4)
{
if (ent->animcache_vertex3f)
{
rsurface.modeltexcoordlightmap2f = model->surfmesh.data_texcoordlightmap2f;
rsurface.modeltexcoordlightmap2f_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
rsurface.modeltexcoordlightmap2f_bufferoffset = model->surfmesh.vbooffset_texcoordlightmap2f;
+ rsurface.modelskeletalindex4ub = model->surfmesh.data_skeletalindex4ub;
+ rsurface.modelskeletalindex4ub_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
+ rsurface.modelskeletalindex4ub_bufferoffset = model->surfmesh.vbooffset_skeletalindex4ub;
+ rsurface.modelskeletalweight4ub = model->surfmesh.data_skeletalweight4ub;
+ rsurface.modelskeletalweight4ub_vertexbuffer = model->surfmesh.vbo_vertexbuffer;
+ rsurface.modelskeletalweight4ub_bufferoffset = model->surfmesh.vbooffset_skeletalweight4ub;
rsurface.modelelement3i = model->surfmesh.data_element3i;
rsurface.modelelement3i_indexbuffer = model->surfmesh.data_element3i_indexbuffer;
rsurface.modelelement3i_bufferoffset = model->surfmesh.data_element3i_bufferoffset;
rsurface.batchtexcoordlightmap2f = NULL;
rsurface.batchtexcoordlightmap2f_vertexbuffer = NULL;
rsurface.batchtexcoordlightmap2f_bufferoffset = 0;
+ rsurface.batchskeletalindex4ub = NULL;
+ rsurface.batchskeletalindex4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalindex4ub_bufferoffset = 0;
+ rsurface.batchskeletalweight4ub = NULL;
+ rsurface.batchskeletalweight4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalweight4ub_bufferoffset = 0;
rsurface.batchvertexmesh = NULL;
rsurface.batchvertexmeshbuffer = NULL;
rsurface.batchvertex3fbuffer = NULL;
rsurface.ent_alttextures = false;
rsurface.basepolygonfactor = r_refdef.polygonfactor;
rsurface.basepolygonoffset = r_refdef.polygonoffset;
+ rsurface.entityskeletaltransform3x4 = NULL;
+ rsurface.entityskeletalnumtransforms = 0;
if (wanttangents)
{
rsurface.modelvertex3f = (float *)vertex3f;
rsurface.modeltexcoordlightmap2f = NULL;
rsurface.modeltexcoordlightmap2f_vertexbuffer = 0;
rsurface.modeltexcoordlightmap2f_bufferoffset = 0;
+ rsurface.modelskeletalindex4ub = NULL;
+ rsurface.modelskeletalindex4ub_vertexbuffer = NULL;
+ rsurface.modelskeletalindex4ub_bufferoffset = 0;
+ rsurface.modelskeletalweight4ub = NULL;
+ rsurface.modelskeletalweight4ub_vertexbuffer = NULL;
+ rsurface.modelskeletalweight4ub_bufferoffset = 0;
rsurface.modelelement3i = (int *)element3i;
rsurface.modelelement3i_indexbuffer = NULL;
rsurface.modelelement3i_bufferoffset = 0;
rsurface.batchtexcoordlightmap2f = NULL;
rsurface.batchtexcoordlightmap2f_vertexbuffer = NULL;
rsurface.batchtexcoordlightmap2f_bufferoffset = 0;
+ rsurface.batchskeletalindex4ub = NULL;
+ rsurface.batchskeletalindex4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalindex4ub_bufferoffset = 0;
+ rsurface.batchskeletalweight4ub = NULL;
+ rsurface.batchskeletalweight4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalweight4ub_bufferoffset = 0;
rsurface.batchvertexmesh = NULL;
rsurface.batchvertexmeshbuffer = NULL;
rsurface.batchvertex3fbuffer = NULL;
float scale;
float center[3], forward[3], right[3], up[3], v[3], newforward[3], newright[3], newup[3];
float waveparms[4];
+ unsigned char *ub;
q3shaderinfo_deform_t *deform;
const msurface_t *surface, *firstsurface;
r_vertexmesh_t *vertexmesh;
// check if any dynamic vertex processing must occur
dynamicvertex = false;
+ // a cvar to force the dynamic vertex path to be taken, for debugging
+ if (r_batch_debugdynamicvertexpath.integer)
+ dynamicvertex = true;
+
// if there is a chance of animated vertex colors, it's a dynamic batch
if ((batchneed & (BATCHNEED_VERTEXMESH_VERTEXCOLOR | BATCHNEED_ARRAY_VERTEXCOLOR)) && texturesurfacelist[0]->lightmapinfo)
{
if (batchneed & BATCHNEED_VERTEXMESH_VERTEXCOLOR) batchneed |= BATCHNEED_ARRAY_VERTEXCOLOR;
if (batchneed & BATCHNEED_VERTEXMESH_TEXCOORD) batchneed |= BATCHNEED_ARRAY_TEXCOORD;
if (batchneed & BATCHNEED_VERTEXMESH_LIGHTMAP) batchneed |= BATCHNEED_ARRAY_LIGHTMAP;
+ if (batchneed & BATCHNEED_VERTEXMESH_SKELETAL) batchneed |= BATCHNEED_ARRAY_SKELETAL;
}
// if needsupdate, we have to do a dynamic vertex batch for sure
if (!rsurface.modelvertexmesh && (batchneed & (BATCHNEED_VERTEXMESH_VERTEX | BATCHNEED_VERTEXMESH_NORMAL | BATCHNEED_VERTEXMESH_VECTOR | BATCHNEED_VERTEXMESH_VERTEXCOLOR | BATCHNEED_VERTEXMESH_TEXCOORD | BATCHNEED_VERTEXMESH_LIGHTMAP)))
dynamicvertex = true;
+ // if we're going to have to apply the skeletal transform manually, we need to batch the skeletal data
+ if (dynamicvertex && rsurface.entityskeletaltransform3x4)
+ batchneed |= BATCHNEED_ARRAY_SKELETAL;
+
rsurface.batchvertex3f = rsurface.modelvertex3f;
rsurface.batchvertex3f_vertexbuffer = rsurface.modelvertex3f_vertexbuffer;
rsurface.batchvertex3f_bufferoffset = rsurface.modelvertex3f_bufferoffset;
rsurface.batchtexcoordlightmap2f = rsurface.modeltexcoordlightmap2f;
rsurface.batchtexcoordlightmap2f_vertexbuffer = rsurface.modeltexcoordlightmap2f_vertexbuffer;
rsurface.batchtexcoordlightmap2f_bufferoffset = rsurface.modeltexcoordlightmap2f_bufferoffset;
+ rsurface.batchskeletalindex4ub = rsurface.modelskeletalindex4ub;
+ rsurface.batchskeletalindex4ub_vertexbuffer = rsurface.modelskeletalindex4ub_vertexbuffer;
+ rsurface.batchskeletalindex4ub_bufferoffset = rsurface.modelskeletalindex4ub_bufferoffset;
+ rsurface.batchskeletalweight4ub = rsurface.modelskeletalweight4ub;
+ rsurface.batchskeletalweight4ub_vertexbuffer = rsurface.modelskeletalweight4ub_vertexbuffer;
+ rsurface.batchskeletalweight4ub_bufferoffset = rsurface.modelskeletalweight4ub_bufferoffset;
rsurface.batchvertex3fbuffer = rsurface.modelvertex3fbuffer;
rsurface.batchvertexmesh = rsurface.modelvertexmesh;
rsurface.batchvertexmeshbuffer = rsurface.modelvertexmeshbuffer;
rsurface.batchelement3s = rsurface.modelelement3s;
rsurface.batchelement3s_indexbuffer = rsurface.modelelement3s_indexbuffer;
rsurface.batchelement3s_bufferoffset = rsurface.modelelement3s_bufferoffset;
+ rsurface.batchskeletaltransform3x4 = rsurface.entityskeletaltransform3x4;
+ rsurface.batchskeletalnumtransforms = rsurface.entityskeletalnumtransforms;
// if any dynamic vertex processing has to occur in software, we copy the
// entire surface list together before processing to rebase the vertices
// now copy the vertex data into a combined array and make an index array
// (this is what Quake3 does all the time)
+ // we also apply any skeletal animation here that would have been done in
+ // the vertex shader, because most of the dynamic vertex animation cases
+ // need actual vertex positions and normals
//if (dynamicvertex)
{
rsurface.batchvertex3fbuffer = NULL;
rsurface.batchtexcoordlightmap2f = NULL;
rsurface.batchtexcoordlightmap2f_vertexbuffer = NULL;
rsurface.batchtexcoordlightmap2f_bufferoffset = 0;
+ rsurface.batchskeletalindex4ub = NULL;
+ rsurface.batchskeletalindex4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalindex4ub_bufferoffset = 0;
+ rsurface.batchskeletalweight4ub = NULL;
+ rsurface.batchskeletalweight4ub_vertexbuffer = NULL;
+ rsurface.batchskeletalweight4ub_bufferoffset = 0;
rsurface.batchelement3i = (int *)R_FrameData_Alloc(batchnumtriangles * sizeof(int[3]));
rsurface.batchelement3i_indexbuffer = NULL;
rsurface.batchelement3i_bufferoffset = 0;
rsurface.batchtexcoordtexture2f = (float *)R_FrameData_Alloc(batchnumvertices * sizeof(float[2]));
if (batchneed & BATCHNEED_ARRAY_LIGHTMAP)
rsurface.batchtexcoordlightmap2f = (float *)R_FrameData_Alloc(batchnumvertices * sizeof(float[2]));
+ if (batchneed & BATCHNEED_ARRAY_SKELETAL)
+ {
+ rsurface.batchskeletalindex4ub = (unsigned char *)R_FrameData_Alloc(batchnumvertices * sizeof(unsigned char[4]));
+ rsurface.batchskeletalweight4ub = (unsigned char *)R_FrameData_Alloc(batchnumvertices * sizeof(unsigned char[4]));
+ }
numvertices = 0;
numtriangles = 0;
for (i = 0;i < texturenumsurfaces;i++)
else
memset(rsurface.batchtexcoordlightmap2f + 2*numvertices, 0, surfacenumvertices * sizeof(float[2]));
}
+ if (batchneed & BATCHNEED_ARRAY_SKELETAL)
+ {
+ if (rsurface.modelskeletalindex4ub)
+ {
+ memcpy(rsurface.batchskeletalindex4ub + 4*numvertices, rsurface.modelskeletalindex4ub + 4*surfacefirstvertex, surfacenumvertices * sizeof(unsigned char[4]));
+ memcpy(rsurface.batchskeletalweight4ub + 4*numvertices, rsurface.modelskeletalweight4ub + 4*surfacefirstvertex, surfacenumvertices * sizeof(unsigned char[4]));
+ }
+ else
+ {
+ memset(rsurface.batchskeletalindex4ub + 4*numvertices, 0, surfacenumvertices * sizeof(unsigned char[4]));
+ memset(rsurface.batchskeletalweight4ub + 4*numvertices, 0, surfacenumvertices * sizeof(unsigned char[4]));
+ ub = rsurface.batchskeletalweight4ub + 4*numvertices;
+ for (j = 0;j < surfacenumvertices;j++)
+ ub[j*4] = 255;
+ }
+ }
}
RSurf_RenumberElements(rsurface.modelelement3i + 3*surfacefirsttriangle, rsurface.batchelement3i + 3*numtriangles, 3*surfacenumtriangles, numvertices - surfacefirstvertex);
numvertices += surfacenumvertices;
rsurface.batchnumtriangles = batchnumtriangles;
}
+ // apply skeletal animation that would have been done in the vertex shader
+ if (rsurface.batchskeletaltransform3x4)
+ {
+ const unsigned char *si;
+ const unsigned char *sw;
+ const float *t[4];
+ const float *b = rsurface.batchskeletaltransform3x4;
+ float *vp, *vs, *vt, *vn;
+ float w[4];
+ float m[3][4], n[3][4];
+ float tp[3], ts[3], tt[3], tn[3];
+ si = rsurface.batchskeletalindex4ub;
+ sw = rsurface.batchskeletalweight4ub;
+ vp = rsurface.batchvertex3f;
+ vs = rsurface.batchsvector3f;
+ vt = rsurface.batchtvector3f;
+ vn = rsurface.batchnormal3f;
+ memset(m[0], 0, sizeof(m));
+ memset(n[0], 0, sizeof(n));
+ for (i = 0;i < batchnumvertices;i++)
+ {
+ t[0] = b + si[0]*12;
+ if (sw[0] == 255)
+ {
+ // common case - only one matrix
+ m[0][0] = t[0][ 0];
+ m[0][1] = t[0][ 1];
+ m[0][2] = t[0][ 2];
+ m[0][3] = t[0][ 3];
+ m[1][0] = t[0][ 4];
+ m[1][1] = t[0][ 5];
+ m[1][2] = t[0][ 6];
+ m[1][3] = t[0][ 7];
+ m[2][0] = t[0][ 8];
+ m[2][1] = t[0][ 9];
+ m[2][2] = t[0][10];
+ m[2][3] = t[0][11];
+ }
+ else if (sw[2] + sw[3])
+ {
+ // blend 4 matrices
+ t[1] = b + si[1]*12;
+ t[2] = b + si[2]*12;
+ t[3] = b + si[3]*12;
+ w[0] = sw[0] * (1.0f / 255.0f);
+ w[1] = sw[1] * (1.0f / 255.0f);
+ w[2] = sw[2] * (1.0f / 255.0f);
+ w[3] = sw[3] * (1.0f / 255.0f);
+ // blend the matrices
+ m[0][0] = t[0][ 0] * w[0] + t[1][ 0] * w[1] + t[2][ 0] * w[2] + t[3][ 0] * w[3];
+ m[0][1] = t[0][ 1] * w[0] + t[1][ 1] * w[1] + t[2][ 1] * w[2] + t[3][ 1] * w[3];
+ m[0][2] = t[0][ 2] * w[0] + t[1][ 2] * w[1] + t[2][ 2] * w[2] + t[3][ 2] * w[3];
+ m[0][3] = t[0][ 3] * w[0] + t[1][ 3] * w[1] + t[2][ 3] * w[2] + t[3][ 3] * w[3];
+ m[1][0] = t[0][ 4] * w[0] + t[1][ 4] * w[1] + t[2][ 4] * w[2] + t[3][ 4] * w[3];
+ m[1][1] = t[0][ 5] * w[0] + t[1][ 5] * w[1] + t[2][ 5] * w[2] + t[3][ 5] * w[3];
+ m[1][2] = t[0][ 6] * w[0] + t[1][ 6] * w[1] + t[2][ 6] * w[2] + t[3][ 6] * w[3];
+ m[1][3] = t[0][ 7] * w[0] + t[1][ 7] * w[1] + t[2][ 7] * w[2] + t[3][ 7] * w[3];
+ m[2][0] = t[0][ 8] * w[0] + t[1][ 8] * w[1] + t[2][ 8] * w[2] + t[3][ 8] * w[3];
+ m[2][1] = t[0][ 9] * w[0] + t[1][ 9] * w[1] + t[2][ 9] * w[2] + t[3][ 9] * w[3];
+ m[2][2] = t[0][10] * w[0] + t[1][10] * w[1] + t[2][10] * w[2] + t[3][10] * w[3];
+ m[2][3] = t[0][11] * w[0] + t[1][11] * w[1] + t[2][11] * w[2] + t[3][11] * w[3];
+ }
+ else
+ {
+ // blend 2 matrices
+ t[1] = b + si[1]*12;
+ w[0] = sw[0] * (1.0f / 255.0f);
+ w[1] = sw[1] * (1.0f / 255.0f);
+ // blend the matrices
+ m[0][0] = t[0][ 0] * w[0] + t[1][ 0] * w[1];
+ m[0][1] = t[0][ 1] * w[0] + t[1][ 1] * w[1];
+ m[0][2] = t[0][ 2] * w[0] + t[1][ 2] * w[1];
+ m[0][3] = t[0][ 3] * w[0] + t[1][ 3] * w[1];
+ m[1][0] = t[0][ 4] * w[0] + t[1][ 4] * w[1];
+ m[1][1] = t[0][ 5] * w[0] + t[1][ 5] * w[1];
+ m[1][2] = t[0][ 6] * w[0] + t[1][ 6] * w[1];
+ m[1][3] = t[0][ 7] * w[0] + t[1][ 7] * w[1];
+ m[2][0] = t[0][ 8] * w[0] + t[1][ 8] * w[1];
+ m[2][1] = t[0][ 9] * w[0] + t[1][ 9] * w[1];
+ m[2][2] = t[0][10] * w[0] + t[1][10] * w[1];
+ m[2][3] = t[0][11] * w[0] + t[1][11] * w[1];
+ }
+ si += 4;
+ sw += 4;
+ // modify the vertex
+ VectorCopy(vp, tp);
+ vp[0] = tp[0] * m[0][0] + tp[1] * m[0][1] + tp[2] * m[0][2] + m[0][3];
+ vp[1] = tp[0] * m[1][0] + tp[1] * m[1][1] + tp[2] * m[1][2] + m[1][3];
+ vp[2] = tp[0] * m[2][0] + tp[1] * m[2][1] + tp[2] * m[2][2] + m[2][3];
+ vp += 3;
+ if (vn)
+ {
+ // the normal transformation matrix is a set of cross products...
+ CrossProduct(m[1], m[0], n[0]);
+ CrossProduct(m[2], m[0], n[1]);
+ CrossProduct(m[0], m[1], n[2]);
+ VectorCopy(vn, tn);
+ vn[0] = tn[0] * n[0][0] + tn[1] * n[0][1] + tn[2] * n[0][2];
+ vn[1] = tn[0] * n[1][0] + tn[1] * n[1][1] + tn[2] * n[1][2];
+ vn[2] = tn[0] * n[2][0] + tn[1] * n[2][1] + tn[2] * n[2][2];
+ VectorNormalize(vn);
+ vn += 3;
+ if (vs)
+ {
+ VectorCopy(vs, ts);
+ vs[0] = ts[0] * n[0][0] + ts[1] * n[0][1] + ts[2] * n[0][2];
+ vs[1] = ts[0] * n[1][0] + ts[1] * n[1][1] + ts[2] * n[1][2];
+ vs[2] = ts[0] * n[2][0] + ts[1] * n[2][1] + ts[2] * n[2][2];
+ VectorNormalize(vs);
+ vs += 3;
+ VectorCopy(vt, tt);
+ vt[0] = tt[0] * n[0][0] + tt[1] * n[0][1] + tt[2] * n[0][2];
+ vt[1] = tt[0] * n[1][0] + tt[1] * n[1][1] + tt[2] * n[1][2];
+ vt[2] = tt[0] * n[2][0] + tt[1] * n[2][1] + tt[2] * n[2][2];
+ VectorNormalize(vt);
+ vt += 3;
+ }
+ }
+ }
+ rsurface.batchskeletaltransform3x4 = NULL;
+ rsurface.batchskeletalnumtransforms = 0;
+ }
+
// q1bsp surfaces rendered in vertex color mode have to have colors
// calculated based on lightstyles
if ((batchneed & (BATCHNEED_VERTEXMESH_VERTEXCOLOR | BATCHNEED_ARRAY_VERTEXCOLOR)) && texturesurfacelist[0]->lightmapinfo)
if ((batchneed & BATCHNEED_VERTEXMESH_LIGHTMAP) && rsurface.batchtexcoordlightmap2f)
for (j = 0, vertexmesh = rsurface.batchvertexmesh;j < batchnumvertices;j++, vertexmesh++)
Vector2Copy(rsurface.batchtexcoordlightmap2f + 2*j, vertexmesh->texcoordlightmap2f);
+ if ((batchneed & BATCHNEED_VERTEXMESH_SKELETAL) && rsurface.batchskeletalindex4ub)
+ {
+ for (j = 0, vertexmesh = rsurface.batchvertexmesh;j < batchnumvertices;j++, vertexmesh++)
+ {
+ Vector4Copy(rsurface.batchskeletalindex4ub + 4*j, vertexmesh->skeletalindex4ub);
+ Vector4Copy(rsurface.batchskeletalweight4ub + 4*j, vertexmesh->skeletalweight4ub);
+ }
+ }
}
}
projects / xonotic / darkplaces.git / blobdiff