#ifndef DELAUNAY_3D_H #define DELAUNAY_3D_H /*************************************************************************/ /* delaunay_3d.h */ /*************************************************************************/ /* This file is part of: */ /* GODOT ENGINE */ /* https://godotengine.org */ /*************************************************************************/ /* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ /* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ /* */ /* Permission is hereby granted, free of charge, to any person obtaining */ /* a copy of this software and associated documentation files (the */ /* "Software"), to deal in the Software without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of the Software, and to */ /* permit persons to whom the Software is furnished to do so, subject to */ /* the following conditions: */ /* */ /* The above copyright notice and this permission notice shall be */ /* included in all copies or substantial portions of the Software. */ /* */ /* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ /* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ /* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ /* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ /* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ /* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /*************************************************************************/ #include "core/math/aabb.h" #include "core/math/projection.h" #include "core/math/vector3.h" #include "r128.h" class Delaunay3D { struct Simplex; struct Vector3i { int x; int y; int z; Vector3i() { x = 0; y = 0; z = 0; } Vector3i(const uint32_t p_x, const uint32_t p_y, const uint32_t p_z) { x = p_x; y = p_y; z = p_z; } Vector3i(const Vector3i &other) { x = other.x; y = other.y; z = other.z; } Vector3i(const Vector3 &v) { x = static_cast(v.x); y = static_cast(v.y); z = static_cast(v.z); } Vector3i &operator=(const Vector3i &other) { x = other.x; y = other.y; z = other.z; return *this; } }; enum { ACCEL_GRID_SIZE = 16 }; struct GridPos { Vector3i pos; List::Element *E = nullptr; }; struct Simplex { uint32_t points[4]; R128 circum_center_x; R128 circum_center_y; R128 circum_center_z; R128 circum_r2; LocalVector grid_positions; List::Element *SE = nullptr; _FORCE_INLINE_ Simplex() {} _FORCE_INLINE_ Simplex(uint32_t p_a, uint32_t p_b, uint32_t p_c, uint32_t p_d) { points[0] = p_a; points[1] = p_b; points[2] = p_c; points[3] = p_d; } }; struct Triangle { uint32_t triangle[3]; bool bad = false; _FORCE_INLINE_ bool operator==(const Triangle &p_triangle) const { return triangle[0] == p_triangle.triangle[0] && triangle[1] == p_triangle.triangle[1] && triangle[2] == p_triangle.triangle[2]; } _FORCE_INLINE_ Triangle() {} _FORCE_INLINE_ Triangle(uint32_t p_a, uint32_t p_b, uint32_t p_c) { if (p_a > p_b) { SWAP(p_a, p_b); } if (p_b > p_c) { SWAP(p_b, p_c); } if (p_a > p_b) { SWAP(p_a, p_b); } triangle[0] = p_a; triangle[1] = p_b; triangle[2] = p_c; } }; struct TriangleHasher { _FORCE_INLINE_ static uint32_t hash(const Triangle &p_triangle) { uint32_t h = hash_djb2_one_32(p_triangle.triangle[0]); h = hash_djb2_one_32(p_triangle.triangle[1], h); return hash_fmix32(hash_djb2_one_32(p_triangle.triangle[2], h)); } }; _FORCE_INLINE_ static void circum_sphere_compute(const Vector3 *p_points, Simplex *p_simplex) { // the only part in the algorithm where there may be precision errors is this one, so ensure that // we do it as maximum precision as possible R128 v0_x = p_points[p_simplex->points[0]].x; R128 v0_y = p_points[p_simplex->points[0]].y; R128 v0_z = p_points[p_simplex->points[0]].z; R128 v1_x = p_points[p_simplex->points[1]].x; R128 v1_y = p_points[p_simplex->points[1]].y; R128 v1_z = p_points[p_simplex->points[1]].z; R128 v2_x = p_points[p_simplex->points[2]].x; R128 v2_y = p_points[p_simplex->points[2]].y; R128 v2_z = p_points[p_simplex->points[2]].z; R128 v3_x = p_points[p_simplex->points[3]].x; R128 v3_y = p_points[p_simplex->points[3]].y; R128 v3_z = p_points[p_simplex->points[3]].z; //Create the rows of our "unrolled" 3x3 matrix R128 row1_x = v1_x - v0_x; R128 row1_y = v1_y - v0_y; R128 row1_z = v1_z - v0_z; R128 row2_x = v2_x - v0_x; R128 row2_y = v2_y - v0_y; R128 row2_z = v2_z - v0_z; R128 row3_x = v3_x - v0_x; R128 row3_y = v3_y - v0_y; R128 row3_z = v3_z - v0_z; R128 sq_lenght1 = row1_x * row1_x + row1_y * row1_y + row1_z * row1_z; R128 sq_lenght2 = row2_x * row2_x + row2_y * row2_y + row2_z * row2_z; R128 sq_lenght3 = row3_x * row3_x + row3_y * row3_y + row3_z * row3_z; //Compute the determinant of said matrix R128 determinant = row1_x * (row2_y * row3_z - row3_y * row2_z) - row2_x * (row1_y * row3_z - row3_y * row1_z) + row3_x * (row1_y * row2_z - row2_y * row1_z); // Compute the volume of the tetrahedron, and precompute a scalar quantity for re-use in the formula R128 volume = determinant / R128(6.f); R128 i12volume = R128(1.f) / (volume * R128(12.f)); R128 center_x = v0_x + i12volume * ((row2_y * row3_z - row3_y * row2_z) * sq_lenght1 - (row1_y * row3_z - row3_y * row1_z) * sq_lenght2 + (row1_y * row2_z - row2_y * row1_z) * sq_lenght3); R128 center_y = v0_y + i12volume * (-(row2_x * row3_z - row3_x * row2_z) * sq_lenght1 + (row1_x * row3_z - row3_x * row1_z) * sq_lenght2 - (row1_x * row2_z - row2_x * row1_z) * sq_lenght3); R128 center_z = v0_z + i12volume * ((row2_x * row3_y - row3_x * row2_y) * sq_lenght1 - (row1_x * row3_y - row3_x * row1_y) * sq_lenght2 + (row1_x * row2_y - row2_x * row1_y) * sq_lenght3); //Once we know the center, the radius is clearly the distance to any vertex R128 rel1_x = center_x - v0_x; R128 rel1_y = center_y - v0_y; R128 rel1_z = center_z - v0_z; R128 radius1 = rel1_x * rel1_x + rel1_y * rel1_y + rel1_z * rel1_z; p_simplex->circum_center_x = center_x; p_simplex->circum_center_y = center_y; p_simplex->circum_center_z = center_z; p_simplex->circum_r2 = radius1; } _FORCE_INLINE_ static bool simplex_contains(const Vector3 *p_points, const Simplex &p_simplex, uint32_t p_vertex) { R128 v_x = p_points[p_vertex].x; R128 v_y = p_points[p_vertex].y; R128 v_z = p_points[p_vertex].z; R128 rel2_x = p_simplex.circum_center_x - v_x; R128 rel2_y = p_simplex.circum_center_y - v_y; R128 rel2_z = p_simplex.circum_center_z - v_z; R128 radius2 = rel2_x * rel2_x + rel2_y * rel2_y + rel2_z * rel2_z; return radius2 < (p_simplex.circum_r2 - R128(0.00001)); } static bool simplex_is_coplanar(const Vector3 *p_points, const Simplex &p_simplex) { Plane p(p_points[p_simplex.points[0]], p_points[p_simplex.points[1]], p_points[p_simplex.points[2]]); if (ABS(p.distance_to(p_points[p_simplex.points[3]])) < CMP_EPSILON) { return true; } Projection cm; cm.matrix[0][0] = p_points[p_simplex.points[0]].x; cm.matrix[0][1] = p_points[p_simplex.points[1]].x; cm.matrix[0][2] = p_points[p_simplex.points[2]].x; cm.matrix[0][3] = p_points[p_simplex.points[3]].x; cm.matrix[1][0] = p_points[p_simplex.points[0]].y; cm.matrix[1][1] = p_points[p_simplex.points[1]].y; cm.matrix[1][2] = p_points[p_simplex.points[2]].y; cm.matrix[1][3] = p_points[p_simplex.points[3]].y; cm.matrix[2][0] = p_points[p_simplex.points[0]].z; cm.matrix[2][1] = p_points[p_simplex.points[1]].z; cm.matrix[2][2] = p_points[p_simplex.points[2]].z; cm.matrix[2][3] = p_points[p_simplex.points[3]].z; cm.matrix[3][0] = 1.0; cm.matrix[3][1] = 1.0; cm.matrix[3][2] = 1.0; cm.matrix[3][3] = 1.0; return ABS(camera_matrix_determinant(cm)) <= CMP_EPSILON; } static float camera_matrix_determinant(const Projection &m) { return m.matrix[0][3] * m.matrix[1][2] * m.matrix[2][1] * m.matrix[3][0] - m.matrix[0][2] * m.matrix[1][3] * m.matrix[2][1] * m.matrix[3][0] - m.matrix[0][3] * m.matrix[1][1] * m.matrix[2][2] * m.matrix[3][0] + m.matrix[0][1] * m.matrix[1][3] * m.matrix[2][2] * m.matrix[3][0] + m.matrix[0][2] * m.matrix[1][1] * m.matrix[2][3] * m.matrix[3][0] - m.matrix[0][1] * m.matrix[1][2] * m.matrix[2][3] * m.matrix[3][0] - m.matrix[0][3] * m.matrix[1][2] * m.matrix[2][0] * m.matrix[3][1] + m.matrix[0][2] * m.matrix[1][3] * m.matrix[2][0] * m.matrix[3][1] + m.matrix[0][3] * m.matrix[1][0] * m.matrix[2][2] * m.matrix[3][1] - m.matrix[0][0] * m.matrix[1][3] * m.matrix[2][2] * m.matrix[3][1] - m.matrix[0][2] * m.matrix[1][0] * m.matrix[2][3] * m.matrix[3][1] + m.matrix[0][0] * m.matrix[1][2] * m.matrix[2][3] * m.matrix[3][1] + m.matrix[0][3] * m.matrix[1][1] * m.matrix[2][0] * m.matrix[3][2] - m.matrix[0][1] * m.matrix[1][3] * m.matrix[2][0] * m.matrix[3][2] - m.matrix[0][3] * m.matrix[1][0] * m.matrix[2][1] * m.matrix[3][2] + m.matrix[0][0] * m.matrix[1][3] * m.matrix[2][1] * m.matrix[3][2] + m.matrix[0][1] * m.matrix[1][0] * m.matrix[2][3] * m.matrix[3][2] - m.matrix[0][0] * m.matrix[1][1] * m.matrix[2][3] * m.matrix[3][2] - m.matrix[0][2] * m.matrix[1][1] * m.matrix[2][0] * m.matrix[3][3] + m.matrix[0][1] * m.matrix[1][2] * m.matrix[2][0] * m.matrix[3][3] + m.matrix[0][2] * m.matrix[1][0] * m.matrix[2][1] * m.matrix[3][3] - m.matrix[0][0] * m.matrix[1][2] * m.matrix[2][1] * m.matrix[3][3] - m.matrix[0][1] * m.matrix[1][0] * m.matrix[2][2] * m.matrix[3][3] + m.matrix[0][0] * m.matrix[1][1] * m.matrix[2][2] * m.matrix[3][3]; } public: struct OutputSimplex { uint32_t points[4]; }; static Vector tetrahedralize(const Vector &p_points) { uint32_t point_count = p_points.size(); Vector3 *points = (Vector3 *)memalloc(sizeof(Vector3) * (point_count + 4)); { const Vector3 *src_points = p_points.ptr(); AABB rect; for (uint32_t i = 0; i < point_count; i++) { Vector3 point = src_points[i]; if (i == 0) { rect.position = point; } else { rect.expand_to(point); } points[i] = point; } for (uint32_t i = 0; i < point_count; i++) { points[i] = (points[i] - rect.position) / rect.size; } float delta_max = Math::sqrt(2.0) * 20.0; Vector3 center = Vector3(0.5, 0.5, 0.5); // any simplex that contains everything is good points[point_count + 0] = center + Vector3(0, 1, 0) * delta_max; points[point_count + 1] = center + Vector3(0, -1, 1) * delta_max; points[point_count + 2] = center + Vector3(1, -1, -1) * delta_max; points[point_count + 3] = center + Vector3(-1, -1, -1) * delta_max; } List acceleration_grid[ACCEL_GRID_SIZE][ACCEL_GRID_SIZE][ACCEL_GRID_SIZE]; List simplex_list; { //create root simplex Simplex *root = memnew(Simplex(point_count + 0, point_count + 1, point_count + 2, point_count + 3)); root->SE = simplex_list.push_back(root); for (uint32_t i = 0; i < ACCEL_GRID_SIZE; i++) { for (uint32_t j = 0; j < ACCEL_GRID_SIZE; j++) { for (uint32_t k = 0; k < ACCEL_GRID_SIZE; k++) { GridPos gp; gp.E = acceleration_grid[i][j][k].push_back(root); gp.pos = Vector3i(i, j, k); root->grid_positions.push_back(gp); } } } circum_sphere_compute(points, root); } HashMap triangles_inserted; LocalVector triangles; for (uint32_t i = 0; i < point_count; i++) { bool unique = true; for (uint32_t j = i + 1; j < point_count; j++) { if (points[i].is_equal_approx(points[j])) { unique = false; break; } } if (!unique) { continue; } Vector3i grid_pos = Vector3i(points[i] * ACCEL_GRID_SIZE); grid_pos.x = CLAMP(grid_pos.x, 0, ACCEL_GRID_SIZE - 1); grid_pos.y = CLAMP(grid_pos.y, 0, ACCEL_GRID_SIZE - 1); grid_pos.z = CLAMP(grid_pos.z, 0, ACCEL_GRID_SIZE - 1); for (List::Element *E = acceleration_grid[grid_pos.x][grid_pos.y][grid_pos.z].front(); E;) { List::Element *N = E->next(); //may be deleted Simplex *simplex = E->get(); if (simplex_contains(points, *simplex, i)) { static const uint32_t triangle_order[4][3] = { { 0, 1, 2 }, { 0, 1, 3 }, { 0, 2, 3 }, { 1, 2, 3 }, }; for (uint32_t k = 0; k < 4; k++) { Triangle t = Triangle(simplex->points[triangle_order[k][0]], simplex->points[triangle_order[k][1]], simplex->points[triangle_order[k][2]]); uint32_t *p = triangles_inserted.getptr(t); if (p) { triangles[*p].bad = true; } else { triangles_inserted.set(t, triangles.size()); triangles.push_back(t); } } //remove simplex and continue simplex_list.erase(simplex->SE); for (uint32_t k = 0; k < simplex->grid_positions.size(); k++) { Vector3i p = simplex->grid_positions[k].pos; acceleration_grid[p.x][p.y][p.z].erase(simplex->grid_positions[k].E); } memdelete(simplex); } E = N; } //uint32_t good_triangles = 0; for (uint32_t j = 0; j < triangles.size(); j++) { if (triangles[j].bad) { continue; } Simplex *new_simplex = memnew(Simplex(triangles[j].triangle[0], triangles[j].triangle[1], triangles[j].triangle[2], i)); circum_sphere_compute(points, new_simplex); new_simplex->SE = simplex_list.push_back(new_simplex); { Vector3 center; center.x = double(new_simplex->circum_center_x); center.y = double(new_simplex->circum_center_y); center.z = double(new_simplex->circum_center_z); float radius2 = Math::sqrt(double(new_simplex->circum_r2)); radius2 += 0.0001; // Vector3 extents = Vector3(radius2, radius2, radius2); Vector3i from = Vector3i((center - extents) * ACCEL_GRID_SIZE); Vector3i to = Vector3i((center + extents) * ACCEL_GRID_SIZE); from.x = CLAMP(from.x, 0, ACCEL_GRID_SIZE - 1); from.y = CLAMP(from.y, 0, ACCEL_GRID_SIZE - 1); from.z = CLAMP(from.z, 0, ACCEL_GRID_SIZE - 1); to.x = CLAMP(to.x, 0, ACCEL_GRID_SIZE - 1); to.y = CLAMP(to.y, 0, ACCEL_GRID_SIZE - 1); to.z = CLAMP(to.z, 0, ACCEL_GRID_SIZE - 1); for (int32_t x = from.x; x <= to.x; x++) { for (int32_t y = from.y; y <= to.y; y++) { for (int32_t z = from.z; z <= to.z; z++) { GridPos gp; gp.pos = Vector3(x, y, z); gp.E = acceleration_grid[x][y][z].push_back(new_simplex); new_simplex->grid_positions.push_back(gp); } } } } //good_triangles++; } //print_line("at point " + itos(i) + "/" + itos(point_count) + " simplices added " + itos(good_triangles) + "/" + itos(simplex_list.size()) + " - triangles: " + itos(triangles.size())); triangles.clear(); triangles_inserted.clear(); } //print_line("end with simplices: " + itos(simplex_list.size())); Vector ret_simplices; ret_simplices.resize(simplex_list.size()); OutputSimplex *ret_simplicesw = ret_simplices.ptrw(); uint32_t simplices_written = 0; //List simplex_list; //for (Simplex *simplex : simplex_list) { for (List::Element *E = simplex_list.front(); E; E = E->next()) { Simplex *simplex = E->get(); bool invalid = false; for (int j = 0; j < 4; j++) { if (simplex->points[j] >= point_count) { invalid = true; break; } } if (invalid || simplex_is_coplanar(points, *simplex)) { memdelete(simplex); continue; } ret_simplicesw[simplices_written].points[0] = simplex->points[0]; ret_simplicesw[simplices_written].points[1] = simplex->points[1]; ret_simplicesw[simplices_written].points[2] = simplex->points[2]; ret_simplicesw[simplices_written].points[3] = simplex->points[3]; simplices_written++; memdelete(simplex); } ret_simplices.resize(simplices_written); memfree(points); return ret_simplices; } }; #endif // DELAUNAY_3D_H