cvar_t r_glsl_vertextextureblend_usebothalphas = {CVAR_SAVE, "r_glsl_vertextextureblend_usebothalphas", "0", "use both alpha layers on vertex blended surfaces, each alpha layer sets amount of 'blend leak' on another layer, requires mod_q3shader_force_terrain_alphaflag on."};
cvar_t r_framedatasize = {CVAR_SAVE, "r_framedatasize", "0.5", "size of renderer data cache used during one frame (for skeletal animation caching, light processing, etc)"};
-cvar_t r_bufferdatasize[R_BUFFERDATA_COUNT] =
+cvar_t r_buffermegs[R_BUFFERDATA_COUNT] =
{
- {CVAR_SAVE, "r_bufferdatasize_vertex", "4", "vertex buffer size for one frame"},
- {CVAR_SAVE, "r_bufferdatasize_index16", "1", "index buffer size for one frame (16bit indices)"},
- {CVAR_SAVE, "r_bufferdatasize_index32", "1", "index buffer size for one frame (32bit indices)"},
- {CVAR_SAVE, "r_bufferdatasize_uniform", "0.25", "uniform buffer size for one frame"},
+ {CVAR_SAVE, "r_buffermegs_vertex", "4", "vertex buffer size for one frame"},
+ {CVAR_SAVE, "r_buffermegs_index16", "1", "index buffer size for one frame (16bit indices)"},
+ {CVAR_SAVE, "r_buffermegs_index32", "1", "index buffer size for one frame (32bit indices)"},
+ {CVAR_SAVE, "r_buffermegs_uniform", "0.25", "uniform buffer size for one frame"},
};
extern cvar_t v_glslgamma;
if (p->loc_Texture_ReflectCube >= 0) {p->tex_Texture_ReflectCube = sampler;qglUniform1i(p->loc_Texture_ReflectCube , sampler);sampler++;}
if (p->loc_Texture_BounceGrid >= 0) {p->tex_Texture_BounceGrid = sampler;qglUniform1i(p->loc_Texture_BounceGrid , sampler);sampler++;}
// get the uniform block indices so we can bind them
- p->ubiloc_Skeletal_Transform12_UniformBlock = qglGetUniformBlockIndex(p->program, "Skeletal_Transform12_UniformBlock");
+ if (vid.support.arb_uniform_buffer_object)
+ p->ubiloc_Skeletal_Transform12_UniformBlock = qglGetUniformBlockIndex(p->program, "Skeletal_Transform12_UniformBlock");
+ else
+ p->ubiloc_Skeletal_Transform12_UniformBlock = -1;
// clear the uniform block bindings
p->ubibind_Skeletal_Transform12_UniformBlock = -1;
// bind the uniform blocks in use
Cvar_RegisterVariable(&r_glsl_vertextextureblend_usebothalphas);
Cvar_RegisterVariable(&r_framedatasize);
for (i = 0;i < R_BUFFERDATA_COUNT;i++)
- Cvar_RegisterVariable(&r_bufferdatasize[i]);
+ Cvar_RegisterVariable(&r_buffermegs[i]);
Cvar_RegisterVariable(&r_batch_dynamicbuffer);
if (gamemode == GAME_NEHAHRA || gamemode == GAME_TENEBRAE)
Cvar_SetValue("r_fullbrights", 0);
//==================================================================================
-// avoid reusing the same buffer objects on consecutive buffers
-#define R_BUFFERDATA_CYCLE 2
+// avoid reusing the same buffer objects on consecutive frames
+#define R_BUFFERDATA_CYCLE 3
typedef struct r_bufferdata_buffer_s
{
for (type = 0;type < R_BUFFERDATA_COUNT;type++)
{
// free all buffers
- p = &r_bufferdata_buffer[r_bufferdata_cycle][type];
+ p = &r_bufferdata_buffer[cycle][type];
while (*p)
{
mem = *p;
}
// resize buffer as needed (this actually makes a new one, the old one will be recycled next frame)
-static void R_BufferData_Resize(r_bufferdata_type_t type, qboolean mustgrow)
+static void R_BufferData_Resize(r_bufferdata_type_t type, qboolean mustgrow, size_t minsize)
{
r_bufferdata_buffer_t *mem = r_bufferdata_buffer[r_bufferdata_cycle][type];
size_t size;
- size = (size_t)(r_bufferdatasize[type].value * 1024*1024);
- size = bound(65536, size, 512*1024*1024);
+ float newvalue = r_buffermegs[type].value;
+
+ // increase the cvar if we have to (but only if we already have a mem)
+ if (mustgrow && mem)
+ newvalue *= 2.0f;
+ newvalue = bound(0.25f, newvalue, 256.0f);
+ while (newvalue * 1024*1024 < minsize)
+ newvalue *= 2.0f;
+
+ // clamp the cvar to valid range
+ newvalue = bound(0.25f, newvalue, 256.0f);
+ if (r_buffermegs[type].value != newvalue)
+ Cvar_SetValueQuick(&r_buffermegs[type], newvalue);
+
+ // calculate size in bytes
+ size = (size_t)(newvalue * 1024*1024);
+ size = bound(131072, size, 256*1024*1024);
+
+ // allocate a new buffer if the size is different (purge old one later)
+ // or if we were told we must grow the buffer
if (!mem || mem->size != size || mustgrow)
{
mem = (r_bufferdata_buffer_t *)Mem_Alloc(r_main_mempool, sizeof(*mem));
{
if (r_bufferdata_buffer[r_bufferdata_cycle][type])
{
- R_BufferData_Resize(type, false);
+ R_BufferData_Resize((r_bufferdata_type_t)type, false, 131072);
// free all but the head buffer, this is how we recycle obsolete
// buffers after they are no longer in use
p = &r_bufferdata_buffer[r_bufferdata_cycle][type]->purge;
}
}
-r_meshbuffer_t *R_BufferData_Store(size_t datasize, void *data, r_bufferdata_type_t type, int *returnbufferoffset, qboolean allowfail)
+r_meshbuffer_t *R_BufferData_Store(size_t datasize, void *data, r_bufferdata_type_t type, int *returnbufferoffset)
{
r_bufferdata_buffer_t *mem;
int offset = 0;
int padsize;
- float newvalue;
*returnbufferoffset = 0;
else
padsize = (datasize + 15) & ~15;
- while (!r_bufferdata_buffer[r_bufferdata_cycle][type] || r_bufferdata_buffer[r_bufferdata_cycle][type]->current + padsize > r_bufferdata_buffer[r_bufferdata_cycle][type]->size)
- {
- // emergency - we ran out of space, allocate more memory
- newvalue = bound(0.25f, r_bufferdatasize[type].value * 2.0f, 256.0f);
- // if we're already at the limit, just fail (if allowfail is false we might run out of video ram)
- if (newvalue == r_bufferdatasize[type].value && allowfail)
- return NULL;
- Cvar_SetValueQuick(&r_bufferdatasize[type], newvalue);
- R_BufferData_Resize(type, true);
- }
+ // if we ran out of space in this buffer we must allocate a new one
+ if (!r_bufferdata_buffer[r_bufferdata_cycle][type] || r_bufferdata_buffer[r_bufferdata_cycle][type]->current + padsize > r_bufferdata_buffer[r_bufferdata_cycle][type]->size)
+ R_BufferData_Resize(type, true, padsize);
+
+ // if the resize did not give us enough memory, fail
+ if (!r_bufferdata_buffer[r_bufferdata_cycle][type] || r_bufferdata_buffer[r_bufferdata_cycle][type]->current + padsize > r_bufferdata_buffer[r_bufferdata_cycle][type]->size)
+ Sys_Error("R_BufferData_Store: failed to create a new buffer of sufficient size\n");
mem = r_bufferdata_buffer[r_bufferdata_cycle][type];
offset = mem->current;
if (r_gpuskeletal && model->num_bones > 0 && model->surfmesh.data_skeletalindex4ub)
{
// cache the skeleton so the vertex shader can use it
- 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];
r_refdef.stats[r_stat_animcache_skeletal_count] += 1;
r_refdef.stats[r_stat_animcache_skeletal_bones] += model->num_bones;
r_refdef.stats[r_stat_animcache_skeletal_maxbones] = max(r_refdef.stats[r_stat_animcache_skeletal_maxbones], model->num_bones);
ent->animcache_skeletaltransform3x4 = (float *)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);
- }
- }
+ Mod_Skeletal_BuildTransforms(model, ent->frameblend, ent->skeleton, NULL, ent->animcache_skeletaltransform3x4);
// note: this can fail if the buffer is at the grow limit
ent->animcache_skeletaltransform3x4size = sizeof(float[3][4]) * model->num_bones;
- ent->animcache_skeletaltransform3x4buffer = R_BufferData_Store(ent->animcache_skeletaltransform3x4size, ent->animcache_skeletaltransform3x4, R_BUFFERDATA_UNIFORM, &ent->animcache_skeletaltransform3x4offset, true);
+ ent->animcache_skeletaltransform3x4buffer = R_BufferData_Store(ent->animcache_skeletaltransform3x4size, ent->animcache_skeletaltransform3x4, R_BUFFERDATA_UNIFORM, &ent->animcache_skeletaltransform3x4offset);
}
else if (ent->animcache_vertex3f)
{
}
else
{
- for (texturewidth = 1;texturewidth < waterwidth ;texturewidth *= 2);
- for (textureheight = 1;textureheight < waterheight;textureheight *= 2);
- for (camerawidth = 1;camerawidth <= waterwidth; camerawidth *= 2); camerawidth /= 2;
- for (cameraheight = 1;cameraheight <= waterheight;cameraheight *= 2); cameraheight /= 2;
+ for (texturewidth = 1;texturewidth < waterwidth ;texturewidth *= 2);
+ for (textureheight = 1;textureheight < waterheight;textureheight *= 2);
+ for (camerawidth = 1;camerawidth * 2 <= waterwidth ;camerawidth *= 2);
+ for (cameraheight = 1;cameraheight * 2 <= waterheight;cameraheight *= 2);
}
// allocate textures as needed
R_SortEntities();
R_AnimCache_ClearCache();
- R_FrameData_NewFrame();
- R_BufferData_NewFrame();
/* adjust for stereo display */
if(R_Stereo_Active())
rsurface.batchelement3s[i] = rsurface.batchelement3i[i];
}
// upload buffer data for the copytriangles batch
- if (vid.forcevbo || (r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object))
+ if ((r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object && gl_vbo.integer) || vid.forcevbo)
{
if (rsurface.batchelement3s)
- rsurface.batchelement3s_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(short[3]), rsurface.batchelement3s, R_BUFFERDATA_INDEX16, &rsurface.batchelement3s_bufferoffset, !vid.forcevbo);
+ rsurface.batchelement3s_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(short[3]), rsurface.batchelement3s, R_BUFFERDATA_INDEX16, &rsurface.batchelement3s_bufferoffset);
else if (rsurface.batchelement3i)
- rsurface.batchelement3i_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(int[3]), rsurface.batchelement3i, R_BUFFERDATA_INDEX32, &rsurface.batchelement3i_bufferoffset, !vid.forcevbo);
+ rsurface.batchelement3i_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(int[3]), rsurface.batchelement3i, R_BUFFERDATA_INDEX32, &rsurface.batchelement3i_bufferoffset);
}
}
else
}
// upload buffer data for the dynamic batch
- if (vid.forcevbo || (r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object))
+ if ((r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object && gl_vbo.integer) || vid.forcevbo)
{
if (rsurface.batchvertexmesh)
- rsurface.batchvertexmesh_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(r_vertexmesh_t), rsurface.batchvertexmesh, R_BUFFERDATA_VERTEX, &rsurface.batchvertexmesh_bufferoffset, !vid.forcevbo);
+ rsurface.batchvertexmesh_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(r_vertexmesh_t), rsurface.batchvertexmesh, R_BUFFERDATA_VERTEX, &rsurface.batchvertexmesh_bufferoffset);
else
{
if (rsurface.batchvertex3f)
- rsurface.batchvertex3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchvertex3f, R_BUFFERDATA_VERTEX, &rsurface.batchvertex3f_bufferoffset, !vid.forcevbo);
+ rsurface.batchvertex3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchvertex3f, R_BUFFERDATA_VERTEX, &rsurface.batchvertex3f_bufferoffset);
if (rsurface.batchsvector3f)
- rsurface.batchsvector3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchsvector3f, R_BUFFERDATA_VERTEX, &rsurface.batchsvector3f_bufferoffset, !vid.forcevbo);
+ rsurface.batchsvector3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchsvector3f, R_BUFFERDATA_VERTEX, &rsurface.batchsvector3f_bufferoffset);
if (rsurface.batchtvector3f)
- rsurface.batchtvector3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchtvector3f, R_BUFFERDATA_VERTEX, &rsurface.batchtvector3f_bufferoffset, !vid.forcevbo);
+ rsurface.batchtvector3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchtvector3f, R_BUFFERDATA_VERTEX, &rsurface.batchtvector3f_bufferoffset);
if (rsurface.batchnormal3f)
- rsurface.batchnormal3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchnormal3f, R_BUFFERDATA_VERTEX, &rsurface.batchnormal3f_bufferoffset, !vid.forcevbo);
- if (rsurface.batchlightmapcolor4f && r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object)
- rsurface.batchlightmapcolor4f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[4]), rsurface.batchlightmapcolor4f, R_BUFFERDATA_VERTEX, &rsurface.batchlightmapcolor4f_bufferoffset, !vid.forcevbo);
- if (rsurface.batchtexcoordtexture2f && r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object)
- rsurface.batchtexcoordtexture2f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[2]), rsurface.batchtexcoordtexture2f, R_BUFFERDATA_VERTEX, &rsurface.batchtexcoordtexture2f_bufferoffset, !vid.forcevbo);
- if (rsurface.batchtexcoordlightmap2f && r_batch_dynamicbuffer.integer && vid.support.arb_vertex_buffer_object)
- rsurface.batchtexcoordlightmap2f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[2]), rsurface.batchtexcoordlightmap2f, R_BUFFERDATA_VERTEX, &rsurface.batchtexcoordlightmap2f_bufferoffset, !vid.forcevbo);
+ rsurface.batchnormal3f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[3]), rsurface.batchnormal3f, R_BUFFERDATA_VERTEX, &rsurface.batchnormal3f_bufferoffset);
+ if (rsurface.batchlightmapcolor4f)
+ rsurface.batchlightmapcolor4f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[4]), rsurface.batchlightmapcolor4f, R_BUFFERDATA_VERTEX, &rsurface.batchlightmapcolor4f_bufferoffset);
+ if (rsurface.batchtexcoordtexture2f)
+ rsurface.batchtexcoordtexture2f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[2]), rsurface.batchtexcoordtexture2f, R_BUFFERDATA_VERTEX, &rsurface.batchtexcoordtexture2f_bufferoffset);
+ if (rsurface.batchtexcoordlightmap2f)
+ rsurface.batchtexcoordlightmap2f_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(float[2]), rsurface.batchtexcoordlightmap2f, R_BUFFERDATA_VERTEX, &rsurface.batchtexcoordlightmap2f_bufferoffset);
if (rsurface.batchskeletalindex4ub)
- rsurface.batchskeletalindex4ub_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(unsigned char[4]), rsurface.batchskeletalindex4ub, R_BUFFERDATA_VERTEX, &rsurface.batchskeletalindex4ub_bufferoffset, !vid.forcevbo);
+ rsurface.batchskeletalindex4ub_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(unsigned char[4]), rsurface.batchskeletalindex4ub, R_BUFFERDATA_VERTEX, &rsurface.batchskeletalindex4ub_bufferoffset);
if (rsurface.batchskeletalweight4ub)
- rsurface.batchskeletalweight4ub_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(unsigned char[4]), rsurface.batchskeletalweight4ub, R_BUFFERDATA_VERTEX, &rsurface.batchskeletalweight4ub_bufferoffset, !vid.forcevbo);
+ rsurface.batchskeletalweight4ub_vertexbuffer = R_BufferData_Store(rsurface.batchnumvertices * sizeof(unsigned char[4]), rsurface.batchskeletalweight4ub, R_BUFFERDATA_VERTEX, &rsurface.batchskeletalweight4ub_bufferoffset);
}
if (rsurface.batchelement3s)
- rsurface.batchelement3s_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(short[3]), rsurface.batchelement3s, R_BUFFERDATA_INDEX16, &rsurface.batchelement3s_bufferoffset, !vid.forcevbo);
+ rsurface.batchelement3s_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(short[3]), rsurface.batchelement3s, R_BUFFERDATA_INDEX16, &rsurface.batchelement3s_bufferoffset);
else if (rsurface.batchelement3i)
- rsurface.batchelement3i_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(int[3]), rsurface.batchelement3i, R_BUFFERDATA_INDEX32, &rsurface.batchelement3i_bufferoffset, !vid.forcevbo);
+ rsurface.batchelement3i_indexbuffer = R_BufferData_Store(rsurface.batchnumtriangles * sizeof(int[3]), rsurface.batchelement3i, R_BUFFERDATA_INDEX32, &rsurface.batchelement3i_bufferoffset);
}
}
projects / xonotic / darkplaces.git / blobdiff