1 | // This file is part of meshoptimizer library; see meshoptimizer.h for version/license details |
2 | #include "meshoptimizer.h" |
3 | |
4 | #include <assert.h> |
5 | #include <float.h> |
6 | #include <string.h> |
7 | |
8 | // This work is based on: |
9 | // Fabian Giesen. Decoding Morton codes. 2009 |
10 | namespace meshopt |
11 | { |
12 | |
13 | // "Insert" two 0 bits after each of the 10 low bits of x |
14 | inline unsigned int part1By2(unsigned int x) |
15 | { |
16 | x &= 0x000003ff; // x = ---- ---- ---- ---- ---- --98 7654 3210 |
17 | x = (x ^ (x << 16)) & 0xff0000ff; // x = ---- --98 ---- ---- ---- ---- 7654 3210 |
18 | x = (x ^ (x << 8)) & 0x0300f00f; // x = ---- --98 ---- ---- 7654 ---- ---- 3210 |
19 | x = (x ^ (x << 4)) & 0x030c30c3; // x = ---- --98 ---- 76-- --54 ---- 32-- --10 |
20 | x = (x ^ (x << 2)) & 0x09249249; // x = ---- 9--8 --7- -6-- 5--4 --3- -2-- 1--0 |
21 | return x; |
22 | } |
23 | |
24 | static void computeOrder(unsigned int* result, const float* vertex_positions_data, size_t vertex_count, size_t vertex_positions_stride) |
25 | { |
26 | size_t vertex_stride_float = vertex_positions_stride / sizeof(float); |
27 | |
28 | float minv[3] = {FLT_MAX, FLT_MAX, FLT_MAX}; |
29 | float maxv[3] = {-FLT_MAX, -FLT_MAX, -FLT_MAX}; |
30 | |
31 | for (size_t i = 0; i < vertex_count; ++i) |
32 | { |
33 | const float* v = vertex_positions_data + i * vertex_stride_float; |
34 | |
35 | for (int j = 0; j < 3; ++j) |
36 | { |
37 | float vj = v[j]; |
38 | |
39 | minv[j] = minv[j] > vj ? vj : minv[j]; |
40 | maxv[j] = maxv[j] < vj ? vj : maxv[j]; |
41 | } |
42 | } |
43 | |
44 | float extent = 0.f; |
45 | |
46 | extent = (maxv[0] - minv[0]) < extent ? extent : (maxv[0] - minv[0]); |
47 | extent = (maxv[1] - minv[1]) < extent ? extent : (maxv[1] - minv[1]); |
48 | extent = (maxv[2] - minv[2]) < extent ? extent : (maxv[2] - minv[2]); |
49 | |
50 | float scale = extent == 0 ? 0.f : 1.f / extent; |
51 | |
52 | // generate Morton order based on the position inside a unit cube |
53 | for (size_t i = 0; i < vertex_count; ++i) |
54 | { |
55 | const float* v = vertex_positions_data + i * vertex_stride_float; |
56 | |
57 | int x = int((v[0] - minv[0]) * scale * 1023.f + 0.5f); |
58 | int y = int((v[1] - minv[1]) * scale * 1023.f + 0.5f); |
59 | int z = int((v[2] - minv[2]) * scale * 1023.f + 0.5f); |
60 | |
61 | result[i] = part1By2(x) | (part1By2(y) << 1) | (part1By2(z) << 2); |
62 | } |
63 | } |
64 | |
65 | static void computeHistogram(unsigned int (&hist)[1024][3], const unsigned int* data, size_t count) |
66 | { |
67 | memset(hist, 0, sizeof(hist)); |
68 | |
69 | // compute 3 10-bit histograms in parallel |
70 | for (size_t i = 0; i < count; ++i) |
71 | { |
72 | unsigned int id = data[i]; |
73 | |
74 | hist[(id >> 0) & 1023][0]++; |
75 | hist[(id >> 10) & 1023][1]++; |
76 | hist[(id >> 20) & 1023][2]++; |
77 | } |
78 | |
79 | unsigned int sumx = 0, sumy = 0, sumz = 0; |
80 | |
81 | // replace histogram data with prefix histogram sums in-place |
82 | for (int i = 0; i < 1024; ++i) |
83 | { |
84 | unsigned int hx = hist[i][0], hy = hist[i][1], hz = hist[i][2]; |
85 | |
86 | hist[i][0] = sumx; |
87 | hist[i][1] = sumy; |
88 | hist[i][2] = sumz; |
89 | |
90 | sumx += hx; |
91 | sumy += hy; |
92 | sumz += hz; |
93 | } |
94 | |
95 | assert(sumx == count && sumy == count && sumz == count); |
96 | } |
97 | |
98 | static void radixPass(unsigned int* destination, const unsigned int* source, const unsigned int* keys, size_t count, unsigned int (&hist)[1024][3], int pass) |
99 | { |
100 | int bitoff = pass * 10; |
101 | |
102 | for (size_t i = 0; i < count; ++i) |
103 | { |
104 | unsigned int id = (keys[source[i]] >> bitoff) & 1023; |
105 | |
106 | destination[hist[id][pass]++] = source[i]; |
107 | } |
108 | } |
109 | |
110 | } // namespace meshopt |
111 | |
112 | void meshopt_spatialSortRemap(unsigned int* destination, const float* vertex_positions, size_t vertex_count, size_t vertex_positions_stride) |
113 | { |
114 | using namespace meshopt; |
115 | |
116 | assert(vertex_positions_stride >= 12 && vertex_positions_stride <= 256); |
117 | assert(vertex_positions_stride % sizeof(float) == 0); |
118 | |
119 | meshopt_Allocator allocator; |
120 | |
121 | unsigned int* keys = allocator.allocate<unsigned int>(vertex_count); |
122 | computeOrder(keys, vertex_positions, vertex_count, vertex_positions_stride); |
123 | |
124 | unsigned int hist[1024][3]; |
125 | computeHistogram(hist, keys, vertex_count); |
126 | |
127 | unsigned int* scratch = allocator.allocate<unsigned int>(vertex_count); |
128 | |
129 | for (size_t i = 0; i < vertex_count; ++i) |
130 | destination[i] = unsigned(i); |
131 | |
132 | // 3-pass radix sort computes the resulting order into scratch |
133 | radixPass(scratch, destination, keys, vertex_count, hist, 0); |
134 | radixPass(destination, scratch, keys, vertex_count, hist, 1); |
135 | radixPass(scratch, destination, keys, vertex_count, hist, 2); |
136 | |
137 | // since our remap table is mapping old=>new, we need to reverse it |
138 | for (size_t i = 0; i < vertex_count; ++i) |
139 | destination[scratch[i]] = unsigned(i); |
140 | } |
141 | |
142 | void meshopt_spatialSortTriangles(unsigned int* destination, const unsigned int* indices, size_t index_count, const float* vertex_positions, size_t vertex_count, size_t vertex_positions_stride) |
143 | { |
144 | using namespace meshopt; |
145 | |
146 | assert(index_count % 3 == 0); |
147 | assert(vertex_positions_stride >= 12 && vertex_positions_stride <= 256); |
148 | assert(vertex_positions_stride % sizeof(float) == 0); |
149 | |
150 | (void)vertex_count; |
151 | |
152 | size_t face_count = index_count / 3; |
153 | size_t vertex_stride_float = vertex_positions_stride / sizeof(float); |
154 | |
155 | meshopt_Allocator allocator; |
156 | |
157 | float* centroids = allocator.allocate<float>(face_count * 3); |
158 | |
159 | for (size_t i = 0; i < face_count; ++i) |
160 | { |
161 | unsigned int a = indices[i * 3 + 0], b = indices[i * 3 + 1], c = indices[i * 3 + 2]; |
162 | assert(a < vertex_count && b < vertex_count && c < vertex_count); |
163 | |
164 | const float* va = vertex_positions + a * vertex_stride_float; |
165 | const float* vb = vertex_positions + b * vertex_stride_float; |
166 | const float* vc = vertex_positions + c * vertex_stride_float; |
167 | |
168 | centroids[i * 3 + 0] = (va[0] + vb[0] + vc[0]) / 3.f; |
169 | centroids[i * 3 + 1] = (va[1] + vb[1] + vc[1]) / 3.f; |
170 | centroids[i * 3 + 2] = (va[2] + vb[2] + vc[2]) / 3.f; |
171 | } |
172 | |
173 | unsigned int* remap = allocator.allocate<unsigned int>(face_count); |
174 | |
175 | meshopt_spatialSortRemap(remap, centroids, face_count, sizeof(float) * 3); |
176 | |
177 | // support in-order remap |
178 | if (destination == indices) |
179 | { |
180 | unsigned int* indices_copy = allocator.allocate<unsigned int>(index_count); |
181 | memcpy(indices_copy, indices, index_count * sizeof(unsigned int)); |
182 | indices = indices_copy; |
183 | } |
184 | |
185 | for (size_t i = 0; i < face_count; ++i) |
186 | { |
187 | unsigned int a = indices[i * 3 + 0], b = indices[i * 3 + 1], c = indices[i * 3 + 2]; |
188 | unsigned int r = remap[i]; |
189 | |
190 | destination[r * 3 + 0] = a; |
191 | destination[r * 3 + 1] = b; |
192 | destination[r * 3 + 2] = c; |
193 | } |
194 | } |
195 | |