int main(){
const uint64_t N = 1000;
clock_t start_t = 0;
clock_t end_t = 0;
double *A_matrix, *B_matrix, *A_matrix_dev, *B_matrix_dev, *result_cpu, *result_gpu;
TIMEIT_START;
// allocate memmory on HOST
A_matrix = (double*)malloc(N*N*sizeof(double));
B_matrix = (double*)malloc(N*N*sizeof(double));
result_cpu = (double*)malloc(N*N*sizeof(double));
result_gpu = (double*)malloc(N*N*sizeof(double));
TIMEIT_END(clock_t host_alloc_t);
print_timeit(host_alloc_t,"allocating memory on host:");
fill_matrix(A_matrix,N);
fill_matrix(B_matrix,N);
//print_matrix(A_matrix,N);
printf("\n");
TIMEIT_START;
mult_matrix(A_matrix,B_matrix,result_cpu,N);
TIMEIT_END(clock_t cpu_mult_t);
print_timeit(cpu_mult_t,"multiplying matrix by a matrix with into another one:");
//print_matrix(result_cpu,N);
TIMEIT_START;
cudaMalloc((void **) &A_matrix_dev,N*N*sizeof(double));
cudaMalloc((void **) &B_matrix_dev,N*N*sizeof(double));
cudaMalloc((void **) &result_gpu,N*N*sizeof(double));
TIMEIT_END(clock_t gpu_alloc_t);
print_timeit(gpu_alloc_t,"allocation memory on device:");
TIMEIT_START;
cudaMemcpy(A_matrix_dev, A_matrix, N*N*sizeof(double), cudaMemcpyHostToDevice);
cudaMemcpy(B_matrix_dev, B_matrix, N*N*sizeof(double), cudaMemcpyHostToDevice);
TIMEIT_END(clock_t gpu_copy_t);
print_timeit(gpu_copy_t,"allocation memory on device:");
free(A_matrix);
free(B_matrix);
free(result_cpu);
free(result_gpu);
return EXIT_SUCCESS;
}
static void um_arraystore_compact_with_info(UndoMesh *um, const UndoMesh *um_ref)
{
#ifdef DEBUG_PRINT
size_t size_expanded_prev, size_compacted_prev;
BLI_array_store_at_size_calc_memory_usage(&um_arraystore.bs_stride, &size_expanded_prev, &size_compacted_prev);
#endif
#ifdef DEBUG_TIME
TIMEIT_START(mesh_undo_compact);
#endif
um_arraystore_compact(um, um_ref);
#ifdef DEBUG_TIME
TIMEIT_END(mesh_undo_compact);
#endif
#ifdef DEBUG_PRINT
{
size_t size_expanded, size_compacted;
BLI_array_store_at_size_calc_memory_usage(&um_arraystore.bs_stride, &size_expanded, &size_compacted);
const double percent_total = size_expanded ?
(((double)size_compacted / (double)size_expanded) * 100.0) : -1.0;
size_t size_expanded_step = size_expanded - size_expanded_prev;
size_t size_compacted_step = size_compacted - size_compacted_prev;
const double percent_step = size_expanded_step ?
(((double)size_compacted_step / (double)size_expanded_step) * 100.0) : -1.0;
printf("overall memory use: %.8f%% of expanded size\n", percent_total);
printf("step memory use: %.8f%% of expanded size\n", percent_step);
}
#endif
}
/**
* A version of #BLI_polyfill_calc that uses a memory arena to avoid re-allocations.
*/
void BLI_polyfill_calc_arena(const float (*coords)[2],
const uint coords_tot,
const int coords_sign,
uint (*r_tris)[3],
struct MemArena *arena)
{
PolyFill pf;
PolyIndex *indices = BLI_memarena_alloc(arena, sizeof(*indices) * coords_tot);
#ifdef DEBUG_TIME
TIMEIT_START(polyfill2d);
#endif
polyfill_prepare(&pf,
coords,
coords_tot,
coords_sign,
r_tris,
/* cache */
indices);
#ifdef USE_KDTREE
if (pf.coords_tot_concave) {
pf.kdtree.nodes = BLI_memarena_alloc(arena, sizeof(*pf.kdtree.nodes) * pf.coords_tot_concave);
pf.kdtree.nodes_map = memset(
BLI_memarena_alloc(arena, sizeof(*pf.kdtree.nodes_map) * coords_tot),
0xff,
sizeof(*pf.kdtree.nodes_map) * coords_tot);
}
else {
pf.kdtree.totnode = 0;
}
#endif
polyfill_calc(&pf);
/* indices are no longer needed,
* caller can clear arena */
#ifdef DEBUG_TIME
TIMEIT_END(polyfill2d);
#endif
}
/**
* Triangulates the given (convex or concave) simple polygon to a list of triangle vertices.
*
* \param coords: 2D coordinates describing vertices of the polygon,
* in either clockwise or counterclockwise order.
* \param coords_tot: Total points in the array.
* \param coords_sign: Pass this when we know the sign in advance to avoid extra calculations.
*
* \param r_tris: This array is filled in with triangle indices in clockwise order.
* The length of the array must be ``coords_tot - 2``.
* Indices are guaranteed to be assigned to unique triangles, with valid indices,
* even in the case of degenerate input (self intersecting polygons, zero area ears... etc).
*/
void BLI_polyfill_calc(const float (*coords)[2],
const uint coords_tot,
const int coords_sign,
uint (*r_tris)[3])
{
PolyFill pf;
PolyIndex *indices = BLI_array_alloca(indices, coords_tot);
#ifdef DEBUG_TIME
TIMEIT_START(polyfill2d);
#endif
polyfill_prepare(&pf,
coords,
coords_tot,
coords_sign,
r_tris,
/* cache */
indices);
#ifdef USE_KDTREE
if (pf.coords_tot_concave) {
pf.kdtree.nodes = BLI_array_alloca(pf.kdtree.nodes, pf.coords_tot_concave);
pf.kdtree.nodes_map = memset(BLI_array_alloca(pf.kdtree.nodes_map, coords_tot),
0xff,
sizeof(*pf.kdtree.nodes_map) * coords_tot);
}
else {
pf.kdtree.totnode = 0;
}
#endif
polyfill_calc(&pf);
#ifdef DEBUG_TIME
TIMEIT_END(polyfill2d);
#endif
}
void draw_smoke_volume(SmokeDomainSettings *sds, Object *ob,
const float min[3], const float max[3],
const float viewnormal[3])
{
if (!sds->tex || !sds->tex_shadow) {
fprintf(stderr, "Could not allocate 3D texture for volume rendering!\n");
return;
}
const bool use_fire = (sds->active_fields & SM_ACTIVE_FIRE) && sds->tex_flame;
GPUShader *shader = GPU_shader_get_builtin_shader(
(use_fire) ? GPU_SHADER_SMOKE_FIRE : GPU_SHADER_SMOKE);
if (!shader) {
fprintf(stderr, "Unable to create GLSL smoke shader.\n");
return;
}
const float ob_sizei[3] = {
1.0f / fabsf(ob->size[0]),
1.0f / fabsf(ob->size[1]),
1.0f / fabsf(ob->size[2])
};
const float size[3] = { max[0] - min[0], max[1] - min[1], max[2] - min[2] };
const float invsize[3] = { 1.0f / size[0], 1.0f / size[1], 1.0f / size[2] };
#ifdef DEBUG_DRAW_TIME
TIMEIT_START(draw);
#endif
/* setup smoke shader */
int soot_location = GPU_shader_get_uniform(shader, "soot_texture");
int spec_location = GPU_shader_get_uniform(shader, "spectrum_texture");
int shadow_location = GPU_shader_get_uniform(shader, "shadow_texture");
int flame_location = GPU_shader_get_uniform(shader, "flame_texture");
int actcol_location = GPU_shader_get_uniform(shader, "active_color");
int stepsize_location = GPU_shader_get_uniform(shader, "step_size");
int densityscale_location = GPU_shader_get_uniform(shader, "density_scale");
int invsize_location = GPU_shader_get_uniform(shader, "invsize");
int ob_sizei_location = GPU_shader_get_uniform(shader, "ob_sizei");
int min_location = GPU_shader_get_uniform(shader, "min");
GPU_shader_bind(shader);
GPU_texture_bind(sds->tex, 0);
GPU_shader_uniform_texture(shader, soot_location, sds->tex);
GPU_texture_bind(sds->tex_shadow, 1);
GPU_shader_uniform_texture(shader, shadow_location, sds->tex_shadow);
GPUTexture *tex_spec = NULL;
if (use_fire) {
GPU_texture_bind(sds->tex_flame, 2);
GPU_shader_uniform_texture(shader, flame_location, sds->tex_flame);
tex_spec = create_flame_spectrum_texture();
GPU_texture_bind(tex_spec, 3);
GPU_shader_uniform_texture(shader, spec_location, tex_spec);
}
float active_color[3] = { 0.9, 0.9, 0.9 };
float density_scale = 10.0f;
if ((sds->active_fields & SM_ACTIVE_COLORS) == 0)
mul_v3_v3(active_color, sds->active_color);
GPU_shader_uniform_vector(shader, actcol_location, 3, 1, active_color);
GPU_shader_uniform_vector(shader, stepsize_location, 1, 1, &sds->dx);
GPU_shader_uniform_vector(shader, densityscale_location, 1, 1, &density_scale);
GPU_shader_uniform_vector(shader, min_location, 3, 1, min);
GPU_shader_uniform_vector(shader, ob_sizei_location, 3, 1, ob_sizei);
GPU_shader_uniform_vector(shader, invsize_location, 3, 1, invsize);
/* setup slicing information */
const int max_slices = 256;
const int max_points = max_slices * 12;
VolumeSlicer slicer;
copy_v3_v3(slicer.min, min);
copy_v3_v3(slicer.max, max);
copy_v3_v3(slicer.size, size);
slicer.verts = MEM_mallocN(sizeof(float) * 3 * max_points, "smoke_slice_vertices");
const int num_points = create_view_aligned_slices(&slicer, max_slices, viewnormal);
/* setup buffer and draw */
int gl_depth = 0, gl_blend = 0;
glGetBooleanv(GL_BLEND, (GLboolean *)&gl_blend);
glGetBooleanv(GL_DEPTH_TEST, (GLboolean *)&gl_depth);
glEnable(GL_DEPTH_TEST);
glEnable(GL_BLEND);
glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
GLuint vertex_buffer;
glGenBuffers(1, &vertex_buffer);
glBindBuffer(GL_ARRAY_BUFFER, vertex_buffer);
glBufferData(GL_ARRAY_BUFFER, sizeof(float) * 3 * num_points, &slicer.verts[0][0], GL_STATIC_DRAW);
glEnableClientState(GL_VERTEX_ARRAY);
glVertexPointer(3, GL_FLOAT, 0, NULL);
glDrawArrays(GL_TRIANGLES, 0, num_points);
glDisableClientState(GL_VERTEX_ARRAY);
#ifdef DEBUG_DRAW_TIME
printf("Draw Time: %f\n", (float)TIMEIT_VALUE(draw));
TIMEIT_END(draw);
#endif
/* cleanup */
glBindBuffer(GL_ARRAY_BUFFER, 0);
glDeleteBuffers(1, &vertex_buffer);
GPU_texture_unbind(sds->tex);
GPU_texture_unbind(sds->tex_shadow);
if (use_fire) {
GPU_texture_unbind(sds->tex_flame);
GPU_texture_unbind(tex_spec);
GPU_texture_free(tex_spec);
}
MEM_freeN(slicer.verts);
GPU_shader_unbind();
if (!gl_blend) {
glDisable(GL_BLEND);
}
if (gl_depth) {
glEnable(GL_DEPTH_TEST);
}
}
static void correctivesmooth_modifier_do(
ModifierData *md, Object *ob, DerivedMesh *dm,
float (*vertexCos)[3], unsigned int numVerts,
struct BMEditMesh *em)
{
CorrectiveSmoothModifierData *csmd = (CorrectiveSmoothModifierData *)md;
const bool force_delta_cache_update =
/* XXX, take care! if mesh data its self changes we need to forcefully recalculate deltas */
((csmd->rest_source == MOD_CORRECTIVESMOOTH_RESTSOURCE_ORCO) &&
(((ID *)ob->data)->tag & LIB_TAG_ID_RECALC));
bool use_only_smooth = (csmd->flag & MOD_CORRECTIVESMOOTH_ONLY_SMOOTH) != 0;
MDeformVert *dvert = NULL;
int defgrp_index;
modifier_get_vgroup(ob, dm, csmd->defgrp_name, &dvert, &defgrp_index);
/* if rest bind_coords not are defined, set them (only run during bind) */
if ((csmd->rest_source == MOD_CORRECTIVESMOOTH_RESTSOURCE_BIND) &&
/* signal to recalculate, whoever sets MUST also free bind coords */
(csmd->bind_coords_num == (unsigned int)-1))
{
BLI_assert(csmd->bind_coords == NULL);
csmd->bind_coords = MEM_dupallocN(vertexCos);
csmd->bind_coords_num = numVerts;
BLI_assert(csmd->bind_coords != NULL);
}
if (UNLIKELY(use_only_smooth)) {
smooth_verts(csmd, dm, dvert, defgrp_index, vertexCos, numVerts);
return;
}
if ((csmd->rest_source == MOD_CORRECTIVESMOOTH_RESTSOURCE_BIND) && (csmd->bind_coords == NULL)) {
modifier_setError(md, "Bind data required");
goto error;
}
/* If the number of verts has changed, the bind is invalid, so we do nothing */
if (csmd->rest_source == MOD_CORRECTIVESMOOTH_RESTSOURCE_BIND) {
if (csmd->bind_coords_num != numVerts) {
modifier_setError(md, "Bind vertex count mismatch: %u to %u", csmd->bind_coords_num, numVerts);
goto error;
}
}
else {
/* MOD_CORRECTIVESMOOTH_RESTSOURCE_ORCO */
if (ob->type != OB_MESH) {
modifier_setError(md, "Object is not a mesh");
goto error;
}
else {
unsigned int me_numVerts = (unsigned int)((em) ? em->bm->totvert : ((Mesh *)ob->data)->totvert);
if (me_numVerts != numVerts) {
modifier_setError(md, "Original vertex count mismatch: %u to %u", me_numVerts, numVerts);
goto error;
}
}
}
/* check to see if our deltas are still valid */
if (!csmd->delta_cache || (csmd->delta_cache_num != numVerts) || force_delta_cache_update) {
const float (*rest_coords)[3];
bool is_rest_coords_alloc = false;
if (csmd->rest_source == MOD_CORRECTIVESMOOTH_RESTSOURCE_BIND) {
/* caller needs to do sanity check here */
csmd->bind_coords_num = numVerts;
rest_coords = (const float (*)[3])csmd->bind_coords;
}
else {
int me_numVerts;
rest_coords = (const float (*)[3]) ((em) ?
BKE_editmesh_vertexCos_get_orco(em, &me_numVerts) :
BKE_mesh_vertexCos_get(ob->data, &me_numVerts));
BLI_assert((unsigned int)me_numVerts == numVerts);
is_rest_coords_alloc = true;
}
#ifdef DEBUG_TIME
TIMEIT_START(corrective_smooth_deltas);
#endif
calc_deltas(csmd, dm, dvert, defgrp_index, rest_coords, numVerts);
#ifdef DEBUG_TIME
TIMEIT_END(corrective_smooth_deltas);
#endif
if (is_rest_coords_alloc) {
MEM_freeN((void *)rest_coords);
}
}
if (csmd->rest_source == MOD_CORRECTIVESMOOTH_RESTSOURCE_BIND) {
/* this could be a check, but at this point it _must_ be valid */
BLI_assert(csmd->bind_coords_num == numVerts && csmd->delta_cache);
}
#ifdef DEBUG_TIME
TIMEIT_START(corrective_smooth);
#endif
/* do the actual delta mush */
smooth_verts(csmd, dm, dvert, defgrp_index, vertexCos, numVerts);
{
unsigned int i;
float (*tangent_spaces)[3][3];
/* calloc, since values are accumulated */
tangent_spaces = MEM_callocN((size_t)numVerts * sizeof(float[3][3]), __func__);
calc_tangent_spaces(dm, vertexCos, tangent_spaces);
for (i = 0; i < numVerts; i++) {
float delta[3];
#ifdef USE_TANGENT_CALC_INLINE
calc_tangent_ortho(tangent_spaces[i]);
#endif
mul_v3_m3v3(delta, tangent_spaces[i], csmd->delta_cache[i]);
add_v3_v3(vertexCos[i], delta);
}
MEM_freeN(tangent_spaces);
}
#ifdef DEBUG_TIME
TIMEIT_END(corrective_smooth);
#endif
return;
/* when the modifier fails to execute */
error:
MEM_SAFE_FREE(csmd->delta_cache);
csmd->delta_cache_num = 0;
}
void draw_smoke_volume(SmokeDomainSettings *sds, Object *ob,
const float min[3], const float max[3],
const float viewnormal[3])
{
if (!sds->tex || !sds->tex_shadow) {
fprintf(stderr, "Could not allocate 3D texture for volume rendering!\n");
return;
}
const bool use_fire = (sds->active_fields & SM_ACTIVE_FIRE) && sds->tex_flame;
GPUShader *shader = GPU_shader_get_builtin_shader(GPU_SHADER_SMOKE);
if (!shader) {
fprintf(stderr, "Unable to create GLSL smoke shader.\n");
return;
}
GPUShader *fire_shader = NULL;
if (use_fire) {
fire_shader = GPU_shader_get_builtin_shader(GPU_SHADER_SMOKE_FIRE);
if (!fire_shader) {
fprintf(stderr, "Unable to create GLSL fire shader.\n");
return;
}
}
const float ob_sizei[3] = {
1.0f / fabsf(ob->size[0]),
1.0f / fabsf(ob->size[1]),
1.0f / fabsf(ob->size[2])
};
const float size[3] = { max[0] - min[0], max[1] - min[1], max[2] - min[2] };
const float invsize[3] = { 1.0f / size[0], 1.0f / size[1], 1.0f / size[2] };
#ifdef DEBUG_DRAW_TIME
TIMEIT_START(draw);
#endif
/* setup slicing information */
const int max_slices = 256;
const int max_points = max_slices * 12;
VolumeSlicer slicer;
copy_v3_v3(slicer.min, min);
copy_v3_v3(slicer.max, max);
copy_v3_v3(slicer.size, size);
slicer.verts = MEM_mallocN(sizeof(float) * 3 * max_points, "smoke_slice_vertices");
const int num_points = create_view_aligned_slices(&slicer, max_slices, viewnormal);
/* setup buffer and draw */
int gl_depth = 0, gl_blend = 0, gl_depth_write = 0;
glGetBooleanv(GL_BLEND, (GLboolean *)&gl_blend);
glGetBooleanv(GL_DEPTH_TEST, (GLboolean *)&gl_depth);
glGetBooleanv(GL_DEPTH_WRITEMASK, (GLboolean *)&gl_depth_write);
glEnable(GL_DEPTH_TEST);
glDepthMask(GL_FALSE);
glEnable(GL_BLEND);
glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
draw_buffer(sds, shader, &slicer, ob_sizei, invsize, num_points, false);
/* Draw fire separately (T47639). */
if (use_fire) {
glBlendFunc(GL_ONE, GL_ONE);
draw_buffer(sds, fire_shader, &slicer, ob_sizei, invsize, num_points, true);
}
#ifdef DEBUG_DRAW_TIME
printf("Draw Time: %f\n", (float)TIMEIT_VALUE(draw));
TIMEIT_END(draw);
#endif
MEM_freeN(slicer.verts);
glDepthMask(gl_depth_write);
if (!gl_blend) {
glDisable(GL_BLEND);
}
if (gl_depth) {
glEnable(GL_DEPTH_TEST);
}
}
static DerivedMesh *applyModifier(ModifierData *md, Object *ob, DerivedMesh *derivedData,
ModifierApplyFlag UNUSED(flag))
{
WeightVGProximityModifierData *wmd = (WeightVGProximityModifierData *) md;
DerivedMesh *dm = derivedData;
MDeformVert *dvert = NULL;
MDeformWeight **dw, **tdw;
int numVerts;
float (*v_cos)[3] = NULL; /* The vertices coordinates. */
Object *obr = NULL; /* Our target object. */
int defgrp_index;
float *tw = NULL;
float *org_w = NULL;
float *new_w = NULL;
int *tidx, *indices = NULL;
int numIdx = 0;
int i;
/* Flags. */
#if 0
const bool do_prev = (wmd->modifier.mode & eModifierMode_DoWeightPreview) != 0;
#endif
#ifdef USE_TIMEIT
TIMEIT_START(perf);
#endif
/* Get number of verts. */
numVerts = dm->getNumVerts(dm);
/* Check if we can just return the original mesh.
* Must have verts and therefore verts assigned to vgroups to do anything useful!
*/
if ((numVerts == 0) || BLI_listbase_is_empty(&ob->defbase))
return dm;
/* Get our target object. */
obr = wmd->proximity_ob_target;
if (obr == NULL)
return dm;
/* Get vgroup idx from its name. */
defgrp_index = defgroup_name_index(ob, wmd->defgrp_name);
if (defgrp_index == -1)
return dm;
dvert = CustomData_duplicate_referenced_layer(&dm->vertData, CD_MDEFORMVERT, numVerts);
/* If no vertices were ever added to an object's vgroup, dvert might be NULL.
* As this modifier never add vertices to vgroup, just return. */
if (!dvert)
return dm;
/* Find out which vertices to work on (all vertices in vgroup), and get their relevant weight.
*/
tidx = MEM_mallocN(sizeof(int) * numVerts, "WeightVGProximity Modifier, tidx");
tw = MEM_mallocN(sizeof(float) * numVerts, "WeightVGProximity Modifier, tw");
tdw = MEM_mallocN(sizeof(MDeformWeight *) * numVerts, "WeightVGProximity Modifier, tdw");
for (i = 0; i < numVerts; i++) {
MDeformWeight *_dw = defvert_find_index(&dvert[i], defgrp_index);
if (_dw) {
tidx[numIdx] = i;
tw[numIdx] = _dw->weight;
tdw[numIdx++] = _dw;
}
}
/* If no vertices found, return org data! */
if (numIdx == 0) {
MEM_freeN(tidx);
MEM_freeN(tw);
MEM_freeN(tdw);
return dm;
}
if (numIdx != numVerts) {
indices = MEM_mallocN(sizeof(int) * numIdx, "WeightVGProximity Modifier, indices");
memcpy(indices, tidx, sizeof(int) * numIdx);
org_w = MEM_mallocN(sizeof(float) * numIdx, "WeightVGProximity Modifier, org_w");
memcpy(org_w, tw, sizeof(float) * numIdx);
dw = MEM_mallocN(sizeof(MDeformWeight *) * numIdx, "WeightVGProximity Modifier, dw");
memcpy(dw, tdw, sizeof(MDeformWeight *) * numIdx);
MEM_freeN(tw);
MEM_freeN(tdw);
}
else {
org_w = tw;
dw = tdw;
}
new_w = MEM_mallocN(sizeof(float) * numIdx, "WeightVGProximity Modifier, new_w");
MEM_freeN(tidx);
/* Get our vertex coordinates. */
v_cos = MEM_mallocN(sizeof(float[3]) * numIdx, "WeightVGProximity Modifier, v_cos");
if (numIdx != numVerts) {
/* XXX In some situations, this code can be up to about 50 times more performant
* than simply using getVertCo for each affected vertex...
*/
float (*tv_cos)[3] = MEM_mallocN(sizeof(float[3]) * numVerts, "WeightVGProximity Modifier, tv_cos");
dm->getVertCos(dm, tv_cos);
for (i = 0; i < numIdx; i++)
copy_v3_v3(v_cos[i], tv_cos[indices[i]]);
MEM_freeN(tv_cos);
}
else
dm->getVertCos(dm, v_cos);
/* Compute wanted distances. */
if (wmd->proximity_mode == MOD_WVG_PROXIMITY_OBJECT) {
const float dist = get_ob2ob_distance(ob, obr);
for (i = 0; i < numIdx; i++)
new_w[i] = dist;
}
else if (wmd->proximity_mode == MOD_WVG_PROXIMITY_GEOMETRY) {
const short use_trgt_verts = (wmd->proximity_flags & MOD_WVG_PROXIMITY_GEOM_VERTS);
const short use_trgt_edges = (wmd->proximity_flags & MOD_WVG_PROXIMITY_GEOM_EDGES);
const short use_trgt_faces = (wmd->proximity_flags & MOD_WVG_PROXIMITY_GEOM_FACES);
if (use_trgt_verts || use_trgt_edges || use_trgt_faces) {
DerivedMesh *target_dm = obr->derivedFinal;
bool free_target_dm = false;
if (!target_dm) {
if (ELEM(obr->type, OB_CURVE, OB_SURF, OB_FONT))
target_dm = CDDM_from_curve(obr);
else if (obr->type == OB_MESH) {
Mesh *me = (Mesh *)obr->data;
if (me->edit_btmesh)
target_dm = CDDM_from_editbmesh(me->edit_btmesh, false, false);
else
target_dm = CDDM_from_mesh(me);
}
free_target_dm = true;
}
/* We must check that we do have a valid target_dm! */
if (target_dm) {
SpaceTransform loc2trgt;
float *dists_v = use_trgt_verts ? MEM_mallocN(sizeof(float) * numIdx, "dists_v") : NULL;
float *dists_e = use_trgt_edges ? MEM_mallocN(sizeof(float) * numIdx, "dists_e") : NULL;
float *dists_f = use_trgt_faces ? MEM_mallocN(sizeof(float) * numIdx, "dists_f") : NULL;
BLI_SPACE_TRANSFORM_SETUP(&loc2trgt, ob, obr);
get_vert2geom_distance(numIdx, v_cos, dists_v, dists_e, dists_f,
target_dm, &loc2trgt);
for (i = 0; i < numIdx; i++) {
new_w[i] = dists_v ? dists_v[i] : FLT_MAX;
if (dists_e)
new_w[i] = min_ff(dists_e[i], new_w[i]);
if (dists_f)
new_w[i] = min_ff(dists_f[i], new_w[i]);
}
if (free_target_dm) target_dm->release(target_dm);
if (dists_v) MEM_freeN(dists_v);
if (dists_e) MEM_freeN(dists_e);
if (dists_f) MEM_freeN(dists_f);
}
/* Else, fall back to default obj2vert behavior. */
else {
get_vert2ob_distance(numIdx, v_cos, new_w, ob, obr);
}
}
else {
get_vert2ob_distance(numIdx, v_cos, new_w, ob, obr);
}
}
/* Map distances to weights. */
do_map(ob, new_w, numIdx, wmd->min_dist, wmd->max_dist, wmd->falloff_type);
/* Do masking. */
weightvg_do_mask(numIdx, indices, org_w, new_w, ob, dm, wmd->mask_constant,
wmd->mask_defgrp_name, wmd->modifier.scene, wmd->mask_texture,
wmd->mask_tex_use_channel, wmd->mask_tex_mapping,
wmd->mask_tex_map_obj, wmd->mask_tex_uvlayer_name);
/* Update vgroup. Note we never add nor remove vertices from vgroup here. */
weightvg_update_vg(dvert, defgrp_index, dw, numIdx, indices, org_w, false, 0.0f, false, 0.0f);
/* If weight preview enabled... */
#if 0 /* XXX Currently done in mod stack :/ */
if (do_prev)
DM_update_weight_mcol(ob, dm, 0, org_w, numIdx, indices);
#endif
/* Freeing stuff. */
MEM_freeN(org_w);
MEM_freeN(new_w);
MEM_freeN(dw);
if (indices)
MEM_freeN(indices);
MEM_freeN(v_cos);
#ifdef USE_TIMEIT
TIMEIT_END(perf);
#endif
/* Return the vgroup-modified mesh. */
return dm;
}
static DerivedMesh *applyModifier(ModifierData *md, Object *ob,
DerivedMesh *derivedData,
ModifierApplyFlag UNUSED(flag))
{
DecimateModifierData *dmd = (DecimateModifierData *) md;
DerivedMesh *dm = derivedData, *result = NULL;
BMesh *bm;
bool calc_face_normal;
float *vweights = NULL;
#ifdef USE_TIMEIT
TIMEIT_START(decim);
#endif
/* set up front so we dont show invalid info in the UI */
dmd->face_count = dm->getNumPolys(dm);
switch (dmd->mode) {
case MOD_DECIM_MODE_COLLAPSE:
if (dmd->percent == 1.0f) {
return dm;
}
calc_face_normal = true;
break;
case MOD_DECIM_MODE_UNSUBDIV:
if (dmd->iter == 0) {
return dm;
}
calc_face_normal = false;
break;
case MOD_DECIM_MODE_DISSOLVE:
if (dmd->angle == 0.0f) {
return dm;
}
calc_face_normal = true;
break;
default:
return dm;
}
if (dmd->face_count <= 3) {
modifier_setError(md, "Modifier requires more than 3 input faces");
return dm;
}
if (dmd->mode == MOD_DECIM_MODE_COLLAPSE) {
if (dmd->defgrp_name[0]) {
MDeformVert *dvert;
int defgrp_index;
modifier_get_vgroup(ob, dm, dmd->defgrp_name, &dvert, &defgrp_index);
if (dvert) {
const unsigned int vert_tot = dm->getNumVerts(dm);
unsigned int i;
vweights = MEM_mallocN(vert_tot * sizeof(float), __func__);
if (dmd->flag & MOD_DECIM_FLAG_INVERT_VGROUP) {
for (i = 0; i < vert_tot; i++) {
const float f = 1.0f - defvert_find_weight(&dvert[i], defgrp_index);
vweights[i] = f > BM_MESH_DECIM_WEIGHT_EPS ? (1.0f / f) : BM_MESH_DECIM_WEIGHT_MAX;
}
}
else {
for (i = 0; i < vert_tot; i++) {
const float f = defvert_find_weight(&dvert[i], defgrp_index);
vweights[i] = f > BM_MESH_DECIM_WEIGHT_EPS ? (1.0f / f) : BM_MESH_DECIM_WEIGHT_MAX;
}
}
}
}
}
bm = DM_to_bmesh(dm, calc_face_normal);
switch (dmd->mode) {
case MOD_DECIM_MODE_COLLAPSE:
{
const int do_triangulate = (dmd->flag & MOD_DECIM_FLAG_TRIANGULATE) != 0;
BM_mesh_decimate_collapse(bm, dmd->percent, vweights, do_triangulate);
break;
}
case MOD_DECIM_MODE_UNSUBDIV:
{
BM_mesh_decimate_unsubdivide(bm, dmd->iter);
break;
}
case MOD_DECIM_MODE_DISSOLVE:
{
const int do_dissolve_boundaries = (dmd->flag & MOD_DECIM_FLAG_ALL_BOUNDARY_VERTS) != 0;
BM_mesh_decimate_dissolve(bm, dmd->angle, do_dissolve_boundaries, (BMO_Delimit)dmd->delimit);
break;
}
}
if (vweights) {
MEM_freeN(vweights);
}
/* update for display only */
dmd->face_count = bm->totface;
result = CDDM_from_bmesh(bm, false);
BLI_assert(bm->vtoolflagpool == NULL &&
bm->etoolflagpool == NULL &&
bm->ftoolflagpool == NULL); /* make sure we never alloc'd these */
BLI_assert(bm->vtable == NULL &&
bm->etable == NULL &&
bm->ftable == NULL);
BM_mesh_free(bm);
#ifdef USE_TIMEIT
TIMEIT_END(decim);
#endif
result->dirty = DM_DIRTY_NORMALS;
return result;
}
void draw_smoke_volume(SmokeDomainSettings *sds, Object *ob,
GPUTexture *tex, float min[3], float max[3],
int res[3], float dx, float UNUSED(base_scale), float viewnormal[3],
GPUTexture *tex_shadow, GPUTexture *tex_flame)
{
int i, j, k, n, good_index;
float d /*, d0 */ /* UNUSED */, dd, ds;
float *points = NULL;
int numpoints = 0;
float cor[3] = {1.0f, 1.0f, 1.0f};
int gl_depth = 0, gl_blend = 0;
/* draw slices of smoke is adapted from c++ code authored
* by: Johannes Schmid and Ingemar Rask, 2006, [email protected] */
float cv[][3] = {
{1.0f, 1.0f, 1.0f}, {-1.0f, 1.0f, 1.0f}, {-1.0f, -1.0f, 1.0f}, {1.0f, -1.0f, 1.0f},
{1.0f, 1.0f, -1.0f}, {-1.0f, 1.0f, -1.0f}, {-1.0f, -1.0f, -1.0f}, {1.0f, -1.0f, -1.0f}
};
/* edges have the form edges[n][0][xyz] + t*edges[n][1][xyz] */
float edges[12][2][3] = {
{{1.0f, 1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{-1.0f, 1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{-1.0f, -1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{1.0f, -1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{1.0f, -1.0f, 1.0f}, {0.0f, 2.0f, 0.0f}},
{{-1.0f, -1.0f, 1.0f}, {0.0f, 2.0f, 0.0f}},
{{-1.0f, -1.0f, -1.0f}, {0.0f, 2.0f, 0.0f}},
{{1.0f, -1.0f, -1.0f}, {0.0f, 2.0f, 0.0f}},
{{-1.0f, 1.0f, 1.0f}, {2.0f, 0.0f, 0.0f}},
{{-1.0f, -1.0f, 1.0f}, {2.0f, 0.0f, 0.0f}},
{{-1.0f, -1.0f, -1.0f}, {2.0f, 0.0f, 0.0f}},
{{-1.0f, 1.0f, -1.0f}, {2.0f, 0.0f, 0.0f}}
};
unsigned char *spec_data;
float *spec_pixels;
GPUTexture *tex_spec;
/* Fragment program to calculate the view3d of smoke */
/* using 4 textures, density, shadow, flame and flame spectrum */
const char *shader_basic =
"!!ARBfp1.0\n"
"PARAM dx = program.local[0];\n"
"PARAM darkness = program.local[1];\n"
"PARAM render = program.local[2];\n"
"PARAM f = {1.442695041, 1.442695041, 1.442695041, 0.01};\n"
"TEMP temp, shadow, flame, spec, value;\n"
"TEX temp, fragment.texcoord[0], texture[0], 3D;\n"
"TEX shadow, fragment.texcoord[0], texture[1], 3D;\n"
"TEX flame, fragment.texcoord[0], texture[2], 3D;\n"
"TEX spec, flame.r, texture[3], 1D;\n"
/* calculate shading factor from density */
"MUL value.r, temp.a, darkness.a;\n"
"MUL value.r, value.r, dx.r;\n"
"MUL value.r, value.r, f.r;\n"
"EX2 temp, -value.r;\n"
/* alpha */
"SUB temp.a, 1.0, temp.r;\n"
/* shade colors */
"MUL temp.r, temp.r, shadow.r;\n"
"MUL temp.g, temp.g, shadow.r;\n"
"MUL temp.b, temp.b, shadow.r;\n"
"MUL temp.r, temp.r, darkness.r;\n"
"MUL temp.g, temp.g, darkness.g;\n"
"MUL temp.b, temp.b, darkness.b;\n"
/* for now this just replace smoke shading if rendering fire */
"CMP result.color, render.r, temp, spec;\n"
"END\n";
/* color shader */
const char *shader_color =
"!!ARBfp1.0\n"
"PARAM dx = program.local[0];\n"
"PARAM darkness = program.local[1];\n"
"PARAM render = program.local[2];\n"
"PARAM f = {1.442695041, 1.442695041, 1.442695041, 1.442695041};\n"
"TEMP temp, shadow, flame, spec, value;\n"
"TEX temp, fragment.texcoord[0], texture[0], 3D;\n"
"TEX shadow, fragment.texcoord[0], texture[1], 3D;\n"
"TEX flame, fragment.texcoord[0], texture[2], 3D;\n"
"TEX spec, flame.r, texture[3], 1D;\n"
/* unpremultiply volume texture */
"RCP value.r, temp.a;\n"
"MUL temp.r, temp.r, value.r;\n"
"MUL temp.g, temp.g, value.r;\n"
"MUL temp.b, temp.b, value.r;\n"
/* calculate shading factor from density */
"MUL value.r, temp.a, darkness.a;\n"
"MUL value.r, value.r, dx.r;\n"
"MUL value.r, value.r, f.r;\n"
"EX2 value.r, -value.r;\n"
/* alpha */
"SUB temp.a, 1.0, value.r;\n"
/* shade colors */
"MUL temp.r, temp.r, shadow.r;\n"
"MUL temp.g, temp.g, shadow.r;\n"
"MUL temp.b, temp.b, shadow.r;\n"
"MUL temp.r, temp.r, value.r;\n"
"MUL temp.g, temp.g, value.r;\n"
"MUL temp.b, temp.b, value.r;\n"
/* for now this just replace smoke shading if rendering fire */
"CMP result.color, render.r, temp, spec;\n"
"END\n";
GLuint prog;
float size[3];
if (!tex) {
printf("Could not allocate 3D texture for 3D View smoke drawing.\n");
return;
}
#ifdef DEBUG_DRAW_TIME
TIMEIT_START(draw);
#endif
/* generate flame spectrum texture */
#define SPEC_WIDTH 256
#define FIRE_THRESH 7
#define MAX_FIRE_ALPHA 0.06f
#define FULL_ON_FIRE 100
spec_data = malloc(SPEC_WIDTH * 4 * sizeof(unsigned char));
flame_get_spectrum(spec_data, SPEC_WIDTH, 1500, 3000);
spec_pixels = malloc(SPEC_WIDTH * 4 * 16 * 16 * sizeof(float));
for (i = 0; i < 16; i++) {
for (j = 0; j < 16; j++) {
for (k = 0; k < SPEC_WIDTH; k++) {
int index = (j * SPEC_WIDTH * 16 + i * SPEC_WIDTH + k) * 4;
if (k >= FIRE_THRESH) {
spec_pixels[index] = ((float)spec_data[k * 4]) / 255.0f;
spec_pixels[index + 1] = ((float)spec_data[k * 4 + 1]) / 255.0f;
spec_pixels[index + 2] = ((float)spec_data[k * 4 + 2]) / 255.0f;
spec_pixels[index + 3] = MAX_FIRE_ALPHA * (
(k > FULL_ON_FIRE) ? 1.0f : (k - FIRE_THRESH) / ((float)FULL_ON_FIRE - FIRE_THRESH));
}
else {
spec_pixels[index] = spec_pixels[index + 1] = spec_pixels[index + 2] = spec_pixels[index + 3] = 0.0f;
}
}
}
}
tex_spec = GPU_texture_create_1D(SPEC_WIDTH, spec_pixels, NULL);
sub_v3_v3v3(size, max, min);
/* maxx, maxy, maxz */
cv[0][0] = max[0];
cv[0][1] = max[1];
cv[0][2] = max[2];
/* minx, maxy, maxz */
cv[1][0] = min[0];
cv[1][1] = max[1];
cv[1][2] = max[2];
/* minx, miny, maxz */
cv[2][0] = min[0];
cv[2][1] = min[1];
cv[2][2] = max[2];
/* maxx, miny, maxz */
cv[3][0] = max[0];
cv[3][1] = min[1];
cv[3][2] = max[2];
/* maxx, maxy, minz */
cv[4][0] = max[0];
cv[4][1] = max[1];
cv[4][2] = min[2];
/* minx, maxy, minz */
cv[5][0] = min[0];
cv[5][1] = max[1];
cv[5][2] = min[2];
/* minx, miny, minz */
cv[6][0] = min[0];
cv[6][1] = min[1];
cv[6][2] = min[2];
/* maxx, miny, minz */
cv[7][0] = max[0];
cv[7][1] = min[1];
cv[7][2] = min[2];
copy_v3_v3(edges[0][0], cv[4]); /* maxx, maxy, minz */
copy_v3_v3(edges[1][0], cv[5]); /* minx, maxy, minz */
copy_v3_v3(edges[2][0], cv[6]); /* minx, miny, minz */
copy_v3_v3(edges[3][0], cv[7]); /* maxx, miny, minz */
copy_v3_v3(edges[4][0], cv[3]); /* maxx, miny, maxz */
copy_v3_v3(edges[5][0], cv[2]); /* minx, miny, maxz */
copy_v3_v3(edges[6][0], cv[6]); /* minx, miny, minz */
copy_v3_v3(edges[7][0], cv[7]); /* maxx, miny, minz */
copy_v3_v3(edges[8][0], cv[1]); /* minx, maxy, maxz */
copy_v3_v3(edges[9][0], cv[2]); /* minx, miny, maxz */
copy_v3_v3(edges[10][0], cv[6]); /* minx, miny, minz */
copy_v3_v3(edges[11][0], cv[5]); /* minx, maxy, minz */
// printf("size x: %f, y: %f, z: %f\n", size[0], size[1], size[2]);
// printf("min[2]: %f, max[2]: %f\n", min[2], max[2]);
edges[0][1][2] = size[2];
edges[1][1][2] = size[2];
edges[2][1][2] = size[2];
edges[3][1][2] = size[2];
edges[4][1][1] = size[1];
edges[5][1][1] = size[1];
edges[6][1][1] = size[1];
edges[7][1][1] = size[1];
edges[8][1][0] = size[0];
edges[9][1][0] = size[0];
edges[10][1][0] = size[0];
edges[11][1][0] = size[0];
glGetBooleanv(GL_BLEND, (GLboolean *)&gl_blend);
glGetBooleanv(GL_DEPTH_TEST, (GLboolean *)&gl_depth);
glDepthMask(GL_FALSE);
glDisable(GL_DEPTH_TEST);
glEnable(GL_BLEND);
/* find cube vertex that is closest to the viewer */
for (i = 0; i < 8; i++) {
float x, y, z;
x = cv[i][0] - viewnormal[0] * size[0] * 0.5f;
y = cv[i][1] - viewnormal[1] * size[1] * 0.5f;
z = cv[i][2] - viewnormal[2] * size[2] * 0.5f;
if ((x >= min[0]) && (x <= max[0]) &&
(y >= min[1]) && (y <= max[1]) &&
(z >= min[2]) && (z <= max[2]))
{
break;
}
}
if (i >= 8) {
/* fallback, avoid using buffer over-run */
i = 0;
}
// printf("i: %d\n", i);
// printf("point %f, %f, %f\n", cv[i][0], cv[i][1], cv[i][2]);
if (GL_TRUE == glewIsSupported("GL_ARB_fragment_program")) {
glEnable(GL_FRAGMENT_PROGRAM_ARB);
glGenProgramsARB(1, &prog);
glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, prog);
/* set shader */
if (sds->active_fields & SM_ACTIVE_COLORS)
glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, (GLsizei)strlen(shader_color), shader_color);
else
glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, (GLsizei)strlen(shader_basic), shader_basic);
/* cell spacing */
glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 0, dx, dx, dx, 1.0);
/* custom parameter for smoke style (higher = thicker) */
if (sds->active_fields & SM_ACTIVE_COLORS)
glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 1, 1.0, 1.0, 1.0, 10.0);
else
glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 1, sds->active_color[0], sds->active_color[1], sds->active_color[2], 10.0);
}
else
printf("Your gfx card does not support 3D View smoke drawing.\n");
GPU_texture_bind(tex, 0);
if (tex_shadow)
GPU_texture_bind(tex_shadow, 1);
else
printf("No volume shadow\n");
if (tex_flame) {
GPU_texture_bind(tex_flame, 2);
GPU_texture_bind(tex_spec, 3);
}
if (!GPU_non_power_of_two_support()) {
cor[0] = (float)res[0] / (float)power_of_2_max_i(res[0]);
cor[1] = (float)res[1] / (float)power_of_2_max_i(res[1]);
cor[2] = (float)res[2] / (float)power_of_2_max_i(res[2]);
}
/* our slices are defined by the plane equation a*x + b*y +c*z + d = 0
* (a,b,c), the plane normal, are given by viewdir
* d is the parameter along the view direction. the first d is given by
* inserting previously found vertex into the plane equation */
/* d0 = (viewnormal[0]*cv[i][0] + viewnormal[1]*cv[i][1] + viewnormal[2]*cv[i][2]); */ /* UNUSED */
ds = (fabsf(viewnormal[0]) * size[0] + fabsf(viewnormal[1]) * size[1] + fabsf(viewnormal[2]) * size[2]);
dd = max_fff(sds->global_size[0], sds->global_size[1], sds->global_size[2]) / 128.f;
n = 0;
good_index = i;
// printf("d0: %f, dd: %f, ds: %f\n\n", d0, dd, ds);
points = MEM_callocN(sizeof(float) * 12 * 3, "smoke_points_preview");
while (1) {
float p0[3];
float tmp_point[3], tmp_point2[3];
if (dd * (float)n > ds)
break;
copy_v3_v3(tmp_point, viewnormal);
mul_v3_fl(tmp_point, -dd * ((ds / dd) - (float)n));
add_v3_v3v3(tmp_point2, cv[good_index], tmp_point);
d = dot_v3v3(tmp_point2, viewnormal);
// printf("my d: %f\n", d);
/* intersect_edges returns the intersection points of all cube edges with
* the given plane that lie within the cube */
numpoints = intersect_edges(points, viewnormal[0], viewnormal[1], viewnormal[2], -d, edges);
// printf("points: %d\n", numpoints);
if (numpoints > 2) {
copy_v3_v3(p0, points);
/* sort points to get a convex polygon */
for (i = 1; i < numpoints - 1; i++) {
for (j = i + 1; j < numpoints; j++) {
if (!convex(p0, viewnormal, &points[j * 3], &points[i * 3])) {
float tmp2[3];
copy_v3_v3(tmp2, &points[j * 3]);
copy_v3_v3(&points[j * 3], &points[i * 3]);
copy_v3_v3(&points[i * 3], tmp2);
}
}
}
/* render fire slice */
glBlendFunc(GL_SRC_ALPHA, GL_ONE);
glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 2, 1.0, 0.0, 0.0, 0.0);
glBegin(GL_POLYGON);
glColor3f(1.0, 1.0, 1.0);
for (i = 0; i < numpoints; i++) {
glTexCoord3d((points[i * 3 + 0] - min[0]) * cor[0] / size[0],
(points[i * 3 + 1] - min[1]) * cor[1] / size[1],
(points[i * 3 + 2] - min[2]) * cor[2] / size[2]);
glVertex3f(points[i * 3 + 0] / fabsf(ob->size[0]),
points[i * 3 + 1] / fabsf(ob->size[1]),
points[i * 3 + 2] / fabsf(ob->size[2]));
}
glEnd();
/* render smoke slice */
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 2, -1.0, 0.0, 0.0, 0.0);
glBegin(GL_POLYGON);
glColor3f(1.0, 1.0, 1.0);
for (i = 0; i < numpoints; i++) {
glTexCoord3d((points[i * 3 + 0] - min[0]) * cor[0] / size[0],
(points[i * 3 + 1] - min[1]) * cor[1] / size[1],
(points[i * 3 + 2] - min[2]) * cor[2] / size[2]);
glVertex3f(points[i * 3 + 0] / fabsf(ob->size[0]),
points[i * 3 + 1] / fabsf(ob->size[1]),
points[i * 3 + 2] / fabsf(ob->size[2]));
}
glEnd();
}
n++;
}
#ifdef DEBUG_DRAW_TIME
printf("Draw Time: %f\n", (float)TIMEIT_VALUE(draw));
TIMEIT_END(draw);
#endif
if (tex_shadow)
GPU_texture_unbind(tex_shadow);
GPU_texture_unbind(tex);
if (tex_flame) {
GPU_texture_unbind(tex_flame);
GPU_texture_unbind(tex_spec);
}
GPU_texture_free(tex_spec);
free(spec_data);
free(spec_pixels);
if (GLEW_ARB_fragment_program) {
glDisable(GL_FRAGMENT_PROGRAM_ARB);
glDeleteProgramsARB(1, &prog);
}
MEM_freeN(points);
if (!gl_blend) {
glDisable(GL_BLEND);
}
if (gl_depth) {
glEnable(GL_DEPTH_TEST);
glDepthMask(GL_TRUE);
}
}
/**
* \note This function sets the edge indices to invalid values.
*/
void BM_mesh_beautify_fill(
BMesh *bm, BMEdge **edge_array, const int edge_array_len,
const short flag, const short method,
const short oflag_edge, const short oflag_face)
{
Heap *eheap; /* edge heap */
HeapNode **eheap_table; /* edge index aligned table pointing to the eheap */
GSet **edge_state_arr = MEM_callocN((size_t)edge_array_len * sizeof(GSet *), __func__);
BLI_mempool *edge_state_pool = BLI_mempool_create(sizeof(EdRotState), 0, 512, BLI_MEMPOOL_NOP);
int i;
#ifdef DEBUG_TIME
TIMEIT_START(beautify_fill);
#endif
eheap = BLI_heap_new_ex((uint)edge_array_len);
eheap_table = MEM_mallocN(sizeof(HeapNode *) * (size_t)edge_array_len, __func__);
/* build heap */
for (i = 0; i < edge_array_len; i++) {
BMEdge *e = edge_array[i];
const float cost = bm_edge_calc_rotate_beauty(e, flag, method);
if (cost < 0.0f) {
eheap_table[i] = BLI_heap_insert(eheap, cost, e);
}
else {
eheap_table[i] = NULL;
}
BM_elem_index_set(e, i); /* set_dirty */
}
bm->elem_index_dirty |= BM_EDGE;
while (BLI_heap_is_empty(eheap) == false) {
BMEdge *e = BLI_heap_popmin(eheap);
i = BM_elem_index_get(e);
eheap_table[i] = NULL;
BLI_assert(BM_edge_face_count_is_equal(e, 2));
e = BM_edge_rotate(bm, e, false, BM_EDGEROT_CHECK_EXISTS);
BLI_assert(e == NULL || BM_edge_face_count_is_equal(e, 2));
if (LIKELY(e)) {
GSet *e_state_set = edge_state_arr[i];
/* add the new state into the set so we don't move into this state again
* note: we could add the previous state too but this isn't essential)
* for avoiding eternal loops */
EdRotState *e_state = BLI_mempool_alloc(edge_state_pool);
erot_state_current(e, e_state);
if (UNLIKELY(e_state_set == NULL)) {
edge_state_arr[i] = e_state_set = erot_gset_new(); /* store previous state */
}
BLI_assert(BLI_gset_haskey(e_state_set, (void *)e_state) == false);
BLI_gset_insert(e_state_set, e_state);
// printf(" %d -> %d, %d\n", i, BM_elem_index_get(e->v1), BM_elem_index_get(e->v2));
/* maintain the index array */
edge_array[i] = e;
BM_elem_index_set(e, i);
/* recalculate faces connected on the heap */
bm_edge_update_beauty_cost(e, eheap, eheap_table, edge_state_arr,
(const BMEdge **)edge_array, edge_array_len,
flag, method);
/* update flags */
if (oflag_edge) {
BMO_edge_flag_enable(bm, e, oflag_edge);
}
if (oflag_face) {
BMO_face_flag_enable(bm, e->l->f, oflag_face);
BMO_face_flag_enable(bm, e->l->radial_next->f, oflag_face);
}
}
}
BLI_heap_free(eheap, NULL);
MEM_freeN(eheap_table);
for (i = 0; i < edge_array_len; i++) {
if (edge_state_arr[i]) {
BLI_gset_free(edge_state_arr[i], NULL);
}
}
MEM_freeN(edge_state_arr);
BLI_mempool_destroy(edge_state_pool);
#ifdef DEBUG_TIME
TIMEIT_END(beautify_fill);
#endif
}
void draw_smoke_volume(SmokeDomainSettings *sds, Object *ob,
GPUTexture *tex, const float min[3], const float max[3],
const int res[3], float dx, float UNUSED(base_scale), const float viewnormal[3],
GPUTexture *tex_shadow, GPUTexture *tex_flame)
{
const float ob_sizei[3] = {
1.0f / fabsf(ob->size[0]),
1.0f / fabsf(ob->size[1]),
1.0f / fabsf(ob->size[2])};
int i, j, k, n, good_index;
float d /*, d0 */ /* UNUSED */, dd, ds;
float (*points)[3] = NULL;
int numpoints = 0;
float cor[3] = {1.0f, 1.0f, 1.0f};
int gl_depth = 0, gl_blend = 0;
int use_fire = (sds->active_fields & SM_ACTIVE_FIRE);
/* draw slices of smoke is adapted from c++ code authored
* by: Johannes Schmid and Ingemar Rask, 2006, [email protected] */
float cv[][3] = {
{1.0f, 1.0f, 1.0f}, {-1.0f, 1.0f, 1.0f}, {-1.0f, -1.0f, 1.0f}, {1.0f, -1.0f, 1.0f},
{1.0f, 1.0f, -1.0f}, {-1.0f, 1.0f, -1.0f}, {-1.0f, -1.0f, -1.0f}, {1.0f, -1.0f, -1.0f}
};
/* edges have the form edges[n][0][xyz] + t*edges[n][1][xyz] */
float edges[12][2][3] = {
{{1.0f, 1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{-1.0f, 1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{-1.0f, -1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{1.0f, -1.0f, -1.0f}, {0.0f, 0.0f, 2.0f}},
{{1.0f, -1.0f, 1.0f}, {0.0f, 2.0f, 0.0f}},
{{-1.0f, -1.0f, 1.0f}, {0.0f, 2.0f, 0.0f}},
{{-1.0f, -1.0f, -1.0f}, {0.0f, 2.0f, 0.0f}},
{{1.0f, -1.0f, -1.0f}, {0.0f, 2.0f, 0.0f}},
{{-1.0f, 1.0f, 1.0f}, {2.0f, 0.0f, 0.0f}},
{{-1.0f, -1.0f, 1.0f}, {2.0f, 0.0f, 0.0f}},
{{-1.0f, -1.0f, -1.0f}, {2.0f, 0.0f, 0.0f}},
{{-1.0f, 1.0f, -1.0f}, {2.0f, 0.0f, 0.0f}}
};
unsigned char *spec_data;
float *spec_pixels;
GPUTexture *tex_spec;
GPUProgram *smoke_program;
int progtype = (sds->active_fields & SM_ACTIVE_COLORS) ? GPU_PROGRAM_SMOKE_COLORED : GPU_PROGRAM_SMOKE;
float size[3];
if (!tex) {
printf("Could not allocate 3D texture for 3D View smoke drawing.\n");
return;
}
#ifdef DEBUG_DRAW_TIME
TIMEIT_START(draw);
#endif
/* generate flame spectrum texture */
#define SPEC_WIDTH 256
#define FIRE_THRESH 7
#define MAX_FIRE_ALPHA 0.06f
#define FULL_ON_FIRE 100
spec_data = malloc(SPEC_WIDTH * 4 * sizeof(unsigned char));
flame_get_spectrum(spec_data, SPEC_WIDTH, 1500, 3000);
spec_pixels = malloc(SPEC_WIDTH * 4 * 16 * 16 * sizeof(float));
for (i = 0; i < 16; i++) {
for (j = 0; j < 16; j++) {
for (k = 0; k < SPEC_WIDTH; k++) {
int index = (j * SPEC_WIDTH * 16 + i * SPEC_WIDTH + k) * 4;
if (k >= FIRE_THRESH) {
spec_pixels[index] = ((float)spec_data[k * 4]) / 255.0f;
spec_pixels[index + 1] = ((float)spec_data[k * 4 + 1]) / 255.0f;
spec_pixels[index + 2] = ((float)spec_data[k * 4 + 2]) / 255.0f;
spec_pixels[index + 3] = MAX_FIRE_ALPHA * (
(k > FULL_ON_FIRE) ? 1.0f : (k - FIRE_THRESH) / ((float)FULL_ON_FIRE - FIRE_THRESH));
}
else {
spec_pixels[index] = spec_pixels[index + 1] = spec_pixels[index + 2] = spec_pixels[index + 3] = 0.0f;
}
}
}
}
tex_spec = GPU_texture_create_1D(SPEC_WIDTH, spec_pixels, NULL);
#undef SPEC_WIDTH
#undef FIRE_THRESH
#undef MAX_FIRE_ALPHA
#undef FULL_ON_FIRE
sub_v3_v3v3(size, max, min);
/* maxx, maxy, maxz */
cv[0][0] = max[0];
cv[0][1] = max[1];
cv[0][2] = max[2];
/* minx, maxy, maxz */
cv[1][0] = min[0];
cv[1][1] = max[1];
cv[1][2] = max[2];
/* minx, miny, maxz */
cv[2][0] = min[0];
cv[2][1] = min[1];
cv[2][2] = max[2];
/* maxx, miny, maxz */
cv[3][0] = max[0];
cv[3][1] = min[1];
cv[3][2] = max[2];
/* maxx, maxy, minz */
cv[4][0] = max[0];
cv[4][1] = max[1];
cv[4][2] = min[2];
/* minx, maxy, minz */
cv[5][0] = min[0];
cv[5][1] = max[1];
cv[5][2] = min[2];
/* minx, miny, minz */
cv[6][0] = min[0];
cv[6][1] = min[1];
cv[6][2] = min[2];
/* maxx, miny, minz */
cv[7][0] = max[0];
cv[7][1] = min[1];
cv[7][2] = min[2];
copy_v3_v3(edges[0][0], cv[4]); /* maxx, maxy, minz */
copy_v3_v3(edges[1][0], cv[5]); /* minx, maxy, minz */
copy_v3_v3(edges[2][0], cv[6]); /* minx, miny, minz */
copy_v3_v3(edges[3][0], cv[7]); /* maxx, miny, minz */
copy_v3_v3(edges[4][0], cv[3]); /* maxx, miny, maxz */
copy_v3_v3(edges[5][0], cv[2]); /* minx, miny, maxz */
copy_v3_v3(edges[6][0], cv[6]); /* minx, miny, minz */
copy_v3_v3(edges[7][0], cv[7]); /* maxx, miny, minz */
copy_v3_v3(edges[8][0], cv[1]); /* minx, maxy, maxz */
copy_v3_v3(edges[9][0], cv[2]); /* minx, miny, maxz */
copy_v3_v3(edges[10][0], cv[6]); /* minx, miny, minz */
copy_v3_v3(edges[11][0], cv[5]); /* minx, maxy, minz */
// printf("size x: %f, y: %f, z: %f\n", size[0], size[1], size[2]);
// printf("min[2]: %f, max[2]: %f\n", min[2], max[2]);
edges[0][1][2] = size[2];
edges[1][1][2] = size[2];
edges[2][1][2] = size[2];
edges[3][1][2] = size[2];
edges[4][1][1] = size[1];
edges[5][1][1] = size[1];
edges[6][1][1] = size[1];
edges[7][1][1] = size[1];
edges[8][1][0] = size[0];
edges[9][1][0] = size[0];
edges[10][1][0] = size[0];
edges[11][1][0] = size[0];
glGetBooleanv(GL_BLEND, (GLboolean *)&gl_blend);
glGetBooleanv(GL_DEPTH_TEST, (GLboolean *)&gl_depth);
glEnable(GL_DEPTH_TEST);
glEnable(GL_BLEND);
/* find cube vertex that is closest to the viewer */
for (i = 0; i < 8; i++) {
float x, y, z;
x = cv[i][0] - viewnormal[0] * size[0] * 0.5f;
y = cv[i][1] - viewnormal[1] * size[1] * 0.5f;
z = cv[i][2] - viewnormal[2] * size[2] * 0.5f;
if ((x >= min[0]) && (x <= max[0]) &&
(y >= min[1]) && (y <= max[1]) &&
(z >= min[2]) && (z <= max[2]))
{
break;
}
}
if (i >= 8) {
/* fallback, avoid using buffer over-run */
i = 0;
}
// printf("i: %d\n", i);
// printf("point %f, %f, %f\n", cv[i][0], cv[i][1], cv[i][2]);
smoke_program = GPU_shader_get_builtin_program(progtype);
if (smoke_program) {
GPU_program_bind(smoke_program);
/* cell spacing */
GPU_program_parameter_4f(smoke_program, 0, dx, dx, dx, 1.0);
/* custom parameter for smoke style (higher = thicker) */
if (sds->active_fields & SM_ACTIVE_COLORS)
GPU_program_parameter_4f(smoke_program, 1, 1.0, 1.0, 1.0, 10.0);
else
GPU_program_parameter_4f(smoke_program, 1, sds->active_color[0], sds->active_color[1], sds->active_color[2], 10.0);
}
else
printf("Your gfx card does not support 3D View smoke drawing.\n");
GPU_texture_bind(tex, 0);
if (tex_shadow)
GPU_texture_bind(tex_shadow, 1);
else
printf("No volume shadow\n");
if (tex_flame) {
GPU_texture_bind(tex_flame, 2);
GPU_texture_bind(tex_spec, 3);
}
if (!GPU_non_power_of_two_support()) {
cor[0] = (float)res[0] / (float)power_of_2_max_u(res[0]);
cor[1] = (float)res[1] / (float)power_of_2_max_u(res[1]);
cor[2] = (float)res[2] / (float)power_of_2_max_u(res[2]);
}
cor[0] /= size[0];
cor[1] /= size[1];
cor[2] /= size[2];
/* our slices are defined by the plane equation a*x + b*y +c*z + d = 0
* (a,b,c), the plane normal, are given by viewdir
* d is the parameter along the view direction. the first d is given by
* inserting previously found vertex into the plane equation */
/* d0 = (viewnormal[0]*cv[i][0] + viewnormal[1]*cv[i][1] + viewnormal[2]*cv[i][2]); */ /* UNUSED */
ds = (fabsf(viewnormal[0]) * size[0] + fabsf(viewnormal[1]) * size[1] + fabsf(viewnormal[2]) * size[2]);
dd = max_fff(sds->global_size[0], sds->global_size[1], sds->global_size[2]) / 128.f;
n = 0;
good_index = i;
// printf("d0: %f, dd: %f, ds: %f\n\n", d0, dd, ds);
points = MEM_callocN(sizeof(*points) * 12, "smoke_points_preview");
while (1) {
float p0[3];
float tmp_point[3], tmp_point2[3];
if (dd * (float)n > ds)
break;
copy_v3_v3(tmp_point, viewnormal);
mul_v3_fl(tmp_point, -dd * ((ds / dd) - (float)n));
add_v3_v3v3(tmp_point2, cv[good_index], tmp_point);
d = dot_v3v3(tmp_point2, viewnormal);
// printf("my d: %f\n", d);
/* intersect_edges returns the intersection points of all cube edges with
* the given plane that lie within the cube */
numpoints = intersect_edges(points, viewnormal[0], viewnormal[1], viewnormal[2], -d, edges);
// printf("points: %d\n", numpoints);
if (numpoints > 2) {
copy_v3_v3(p0, points[0]);
/* sort points to get a convex polygon */
for (i = 1; i < numpoints - 1; i++) {
for (j = i + 1; j < numpoints; j++) {
if (!convex(p0, viewnormal, points[j], points[i])) {
swap_v3_v3(points[i], points[j]);
}
}
}
/* render fire slice */
if (use_fire) {
if (GLEW_VERSION_1_4)
glBlendFuncSeparate(GL_SRC_ALPHA, GL_ONE, GL_ONE, GL_ONE);
else
glBlendFunc(GL_SRC_ALPHA, GL_ONE);
GPU_program_parameter_4f(smoke_program, 2, 1.0, 0.0, 0.0, 0.0);
glBegin(GL_POLYGON);
glColor3f(1.0, 1.0, 1.0);
for (i = 0; i < numpoints; i++) {
glTexCoord3d((points[i][0] - min[0]) * cor[0],
(points[i][1] - min[1]) * cor[1],
(points[i][2] - min[2]) * cor[2]);
glVertex3f(points[i][0] * ob_sizei[0],
points[i][1] * ob_sizei[1],
points[i][2] * ob_sizei[2]);
}
glEnd();
}
/* render smoke slice */
if (GLEW_VERSION_1_4)
glBlendFuncSeparate(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
else
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
GPU_program_parameter_4f(smoke_program, 2, -1.0, 0.0, 0.0, 0.0);
glBegin(GL_POLYGON);
glColor3f(1.0, 1.0, 1.0);
for (i = 0; i < numpoints; i++) {
glTexCoord3d((points[i][0] - min[0]) * cor[0],
(points[i][1] - min[1]) * cor[1],
(points[i][2] - min[2]) * cor[2]);
glVertex3f(points[i][0] * ob_sizei[0],
points[i][1] * ob_sizei[1],
points[i][2] * ob_sizei[2]);
}
glEnd();
}
n++;
}
#ifdef DEBUG_DRAW_TIME
printf("Draw Time: %f\n", (float)TIMEIT_VALUE(draw));
TIMEIT_END(draw);
#endif
if (tex_shadow)
GPU_texture_unbind(tex_shadow);
GPU_texture_unbind(tex);
if (tex_flame) {
GPU_texture_unbind(tex_flame);
GPU_texture_unbind(tex_spec);
}
GPU_texture_free(tex_spec);
free(spec_data);
free(spec_pixels);
if (smoke_program)
GPU_program_unbind(smoke_program);
MEM_freeN(points);
if (!gl_blend) {
glDisable(GL_BLEND);
}
if (gl_depth) {
glEnable(GL_DEPTH_TEST);
}
}