Line | Branch | Exec | Source |
---|---|---|---|
1 | // Copyright Contributors to the OpenVDB Project | ||
2 | // SPDX-License-Identifier: MPL-2.0 | ||
3 | |||
4 | #ifndef OPENVDB_TOOLS_DENSESPARSETOOLS_HAS_BEEN_INCLUDED | ||
5 | #define OPENVDB_TOOLS_DENSESPARSETOOLS_HAS_BEEN_INCLUDED | ||
6 | |||
7 | #include <tbb/parallel_reduce.h> | ||
8 | #include <tbb/blocked_range3d.h> | ||
9 | #include <tbb/blocked_range2d.h> | ||
10 | #include <tbb/blocked_range.h> | ||
11 | #include <openvdb/Types.h> | ||
12 | #include <openvdb/tree/LeafManager.h> | ||
13 | #include "Dense.h" | ||
14 | #include <algorithm> // for std::min() | ||
15 | #include <vector> | ||
16 | |||
17 | |||
18 | namespace openvdb { | ||
19 | OPENVDB_USE_VERSION_NAMESPACE | ||
20 | namespace OPENVDB_VERSION_NAME { | ||
21 | namespace tools { | ||
22 | |||
23 | /// @brief Selectively extract and transform data from a dense grid, producing a | ||
24 | /// sparse tree with leaf nodes only (e.g. create a tree from the square | ||
25 | /// of values greater than a cutoff.) | ||
26 | /// @param dense A dense grid that acts as a data source | ||
27 | /// @param functor A functor that selects and transforms data for output | ||
28 | /// @param background The background value of the resulting sparse grid | ||
29 | /// @param threaded Option to use threaded or serial code path | ||
30 | /// @return @c Ptr to tree with the valuetype and configuration defined | ||
31 | /// by typedefs in the @c functor. | ||
32 | /// @note To achieve optimal sparsity consider calling the prune() | ||
33 | /// method on the result. | ||
34 | /// @note To simply copy the all the data from a Dense grid to a | ||
35 | /// OpenVDB Grid, use tools::copyFromDense() for better performance. | ||
36 | /// | ||
37 | /// The type of the sparse tree is determined by the specified OtpType | ||
38 | /// functor by means of the typedef OptType::ResultTreeType | ||
39 | /// | ||
40 | /// The OptType function is responsible for the the transformation of | ||
41 | /// dense grid data to sparse grid data on a per-voxel basis. | ||
42 | /// | ||
43 | /// Only leaf nodes with active values will be added to the sparse grid. | ||
44 | /// | ||
45 | /// The OpType must struct that defines a the minimal form | ||
46 | /// @code | ||
47 | /// struct ExampleOp | ||
48 | /// { | ||
49 | /// using ResultTreeType = DesiredTreeType; | ||
50 | /// | ||
51 | /// template<typename IndexOrCoord> | ||
52 | /// void OpType::operator() (const DenseValueType a, const IndexOrCoord& ijk, | ||
53 | /// ResultTreeType::LeafNodeType* leaf); | ||
54 | /// }; | ||
55 | /// @endcode | ||
56 | /// | ||
57 | /// For example, to generate a <ValueType, 5, 4, 3> tree with valuesOn | ||
58 | /// at locations greater than a given maskvalue | ||
59 | /// @code | ||
60 | /// template<typename ValueType> | ||
61 | /// class Rule | ||
62 | /// { | ||
63 | /// public: | ||
64 | /// // Standard tree type (e.g. MaskTree or FloatTree in openvdb.h) | ||
65 | /// using ResultTreeType = typename openvdb::tree::Tree4<ValueType, 5, 4, 3>::Type; | ||
66 | /// | ||
67 | /// using ResultLeafNodeType = typename ResultTreeType::LeafNodeType; | ||
68 | /// using ResultValueType = typename ResultTreeType::ValueType; | ||
69 | /// | ||
70 | /// using DenseValueType = float; | ||
71 | /// | ||
72 | /// using Index = openvdb::Coord::ValueType; | ||
73 | /// | ||
74 | /// Rule(const DenseValueType& value): mMaskValue(value){}; | ||
75 | /// | ||
76 | /// template<typename IndexOrCoord> | ||
77 | /// void operator()(const DenseValueType& a, const IndexOrCoord& offset, | ||
78 | /// ResultLeafNodeType* leaf) const | ||
79 | /// { | ||
80 | /// if (a > mMaskValue) { | ||
81 | /// leaf->setValueOn(offset, a); | ||
82 | /// } | ||
83 | /// } | ||
84 | /// | ||
85 | /// private: | ||
86 | /// const DenseValueType mMaskValue; | ||
87 | /// }; | ||
88 | /// @endcode | ||
89 | template<typename OpType, typename DenseType> | ||
90 | typename OpType::ResultTreeType::Ptr | ||
91 | extractSparseTree(const DenseType& dense, const OpType& functor, | ||
92 | const typename OpType::ResultValueType& background, | ||
93 | bool threaded = true); | ||
94 | |||
95 | /// This struct that aids template resolution of a new tree type | ||
96 | /// has the same configuration at TreeType, but the ValueType from | ||
97 | /// DenseType. | ||
98 | template<typename DenseType, typename TreeType> | ||
99 | struct DSConverter | ||
100 | { | ||
101 | using ValueType = typename DenseType::ValueType; | ||
102 | using Type = typename TreeType::template ValueConverter<ValueType>::Type; | ||
103 | }; | ||
104 | |||
105 | |||
106 | /// @brief Copy data from the intersection of a sparse tree and a dense input grid. | ||
107 | /// The resulting tree has the same configuration as the sparse tree, but holds | ||
108 | /// the data type specified by the dense input. | ||
109 | /// @param dense A dense grid that acts as a data source | ||
110 | /// @param mask The active voxels and tiles intersected with dense define iteration mask | ||
111 | /// @param background The background value of the resulting sparse grid | ||
112 | /// @param threaded Option to use threaded or serial code path | ||
113 | /// @return @c Ptr to tree with the same configuration as @c mask but of value type | ||
114 | /// defined by @c dense. | ||
115 | template<typename DenseType, typename MaskTreeType> | ||
116 | typename DSConverter<DenseType, MaskTreeType>::Type::Ptr | ||
117 | extractSparseTreeWithMask(const DenseType& dense, | ||
118 | const MaskTreeType& mask, | ||
119 | const typename DenseType::ValueType& background, | ||
120 | bool threaded = true); | ||
121 | |||
122 | |||
123 | /// Apply a point-wise functor to the intersection of a dense grid and a given bounding box | ||
124 | /// @param dense A dense grid to be transformed | ||
125 | /// @param bbox Index space bounding box, define region where the transformation is applied | ||
126 | /// @param op A functor that acts on the dense grid value type | ||
127 | /// @param parallel Used to select multithreaded or single threaded | ||
128 | /// Minimally, the @c op class has to support a @c operator() method, | ||
129 | /// @code | ||
130 | /// // Square values in a grid | ||
131 | /// struct Op | ||
132 | /// { | ||
133 | /// ValueT operator()(const ValueT& in) const | ||
134 | /// { | ||
135 | /// // do work | ||
136 | /// ValueT result = in * in; | ||
137 | /// | ||
138 | /// return result; | ||
139 | /// } | ||
140 | /// }; | ||
141 | /// @endcode | ||
142 | /// NB: only Dense grids with memory layout zxy are supported | ||
143 | template<typename ValueT, typename OpType> | ||
144 | void transformDense(Dense<ValueT, openvdb::tools::LayoutZYX>& dense, | ||
145 | const openvdb::CoordBBox& bbox, const OpType& op, bool parallel=true); | ||
146 | |||
147 | /// We currrently support the following operations when compositing sparse | ||
148 | /// data into a dense grid. | ||
149 | enum DSCompositeOp { | ||
150 | DS_OVER, DS_ADD, DS_SUB, DS_MIN, DS_MAX, DS_MULT, DS_SET | ||
151 | }; | ||
152 | |||
153 | /// @brief Composite data from a sparse tree into a dense array of the same value type. | ||
154 | /// @param dense Dense grid to be altered by the operation | ||
155 | /// @param source Sparse data to composite into @c dense | ||
156 | /// @param alpha Sparse Alpha mask used in compositing operations. | ||
157 | /// @param beta Constant multiplier on src | ||
158 | /// @param strength Constant multiplier on alpha | ||
159 | /// @param threaded Enable threading for this operation. | ||
160 | template<DSCompositeOp, typename TreeT> | ||
161 | void compositeToDense(Dense<typename TreeT::ValueType, LayoutZYX>& dense, | ||
162 | const TreeT& source, | ||
163 | const TreeT& alpha, | ||
164 | const typename TreeT::ValueType beta, | ||
165 | const typename TreeT::ValueType strength, | ||
166 | bool threaded = true); | ||
167 | |||
168 | |||
169 | /// @brief Functor-based class used to extract data that satisfies some | ||
170 | /// criteria defined by the embedded @c OpType functor. The @c extractSparseTree | ||
171 | /// function wraps this class. | ||
172 | template<typename OpType, typename DenseType> | ||
173 | 17 | class SparseExtractor | |
174 | { | ||
175 | public: | ||
176 | using Index = openvdb::math::Coord::ValueType; | ||
177 | |||
178 | using DenseValueType = typename DenseType::ValueType; | ||
179 | using ResultTreeType = typename OpType::ResultTreeType; | ||
180 | using ResultValueType = typename ResultTreeType::ValueType; | ||
181 | using ResultLeafNodeType = typename ResultTreeType::LeafNodeType; | ||
182 | using MaskTree = typename ResultTreeType::template ValueConverter<ValueMask>::Type; | ||
183 | |||
184 | using Range3d = tbb::blocked_range3d<Index, Index, Index>; | ||
185 | |||
186 | private: | ||
187 | const DenseType& mDense; | ||
188 | const OpType& mFunctor; | ||
189 | const ResultValueType mBackground; | ||
190 | const openvdb::math::CoordBBox mBBox; | ||
191 | const Index mWidth; | ||
192 | typename ResultTreeType::Ptr mMask; | ||
193 | openvdb::math::Coord mMin; | ||
194 | |||
195 | public: | ||
196 | 6 | SparseExtractor(const DenseType& dense, const OpType& functor, | |
197 | const ResultValueType background) : | ||
198 | mDense(dense), mFunctor(functor), | ||
199 | mBackground(background), | ||
200 | mBBox(dense.bbox()), | ||
201 | mWidth(ResultLeafNodeType::DIM), | ||
202 | 6 | mMask( new ResultTreeType(mBackground)) | |
203 | 6 | {} | |
204 | |||
205 | ✗ | SparseExtractor(const DenseType& dense, | |
206 | const openvdb::math::CoordBBox& bbox, | ||
207 | const OpType& functor, | ||
208 | const ResultValueType background) : | ||
209 | mDense(dense), mFunctor(functor), | ||
210 | mBackground(background), | ||
211 | mBBox(bbox), | ||
212 | mWidth(ResultLeafNodeType::DIM), | ||
213 | ✗ | mMask( new ResultTreeType(mBackground)) | |
214 | { | ||
215 | // mBBox must be inside the coordinate rage of the dense grid | ||
216 | if (!dense.bbox().isInside(mBBox)) { | ||
217 | ✗ | OPENVDB_THROW(ValueError, "Data extraction window out of bound"); | |
218 | } | ||
219 | } | ||
220 | |||
221 | 34 | SparseExtractor(SparseExtractor& other, tbb::split): | |
222 | 34 | mDense(other.mDense), mFunctor(other.mFunctor), | |
223 | 34 | mBackground(other.mBackground), mBBox(other.mBBox), | |
224 | 34 | mWidth(other.mWidth), | |
225 | 34 | mMask(new ResultTreeType(mBackground)), | |
226 | 34 | mMin(other.mMin) | |
227 | 34 | {} | |
228 | |||
229 | 6 | typename ResultTreeType::Ptr extract(bool threaded = true) | |
230 | { | ||
231 | // Construct 3D range of leaf nodes that | ||
232 | // intersect mBBox. | ||
233 | |||
234 | // Snap the bbox to nearest leaf nodes min and max | ||
235 | |||
236 | 6 | openvdb::math::Coord padded_min = mBBox.min(); | |
237 | 6 | openvdb::math::Coord padded_max = mBBox.max(); | |
238 | |||
239 | |||
240 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
6 | padded_min &= ~(mWidth - 1); |
241 | padded_max &= ~(mWidth - 1); | ||
242 | |||
243 | 6 | padded_max[0] += mWidth - 1; | |
244 | 6 | padded_max[1] += mWidth - 1; | |
245 | 6 | padded_max[2] += mWidth - 1; | |
246 | |||
247 | |||
248 | // number of leaf nodes in each direction | ||
249 | // division by leaf width, e.g. 8 in most cases | ||
250 | |||
251 | 6 | const Index xleafCount = ( padded_max.x() - padded_min.x() + 1 ) / mWidth; | |
252 | 6 | const Index yleafCount = ( padded_max.y() - padded_min.y() + 1 ) / mWidth; | |
253 | 6 | const Index zleafCount = ( padded_max.z() - padded_min.z() + 1 ) / mWidth; | |
254 | |||
255 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
6 | mMin = padded_min; |
256 | |||
257 | Range3d leafRange(0, xleafCount, 1, | ||
258 | 0, yleafCount, 1, | ||
259 | 0, zleafCount, 1); | ||
260 | |||
261 | // Iterate over the leafnodes applying *this as a functor. | ||
262 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
6 | if (threaded) { |
263 | 6 | tbb::parallel_reduce(leafRange, *this); | |
264 | } else { | ||
265 | ✗ | (*this)(leafRange); | |
266 | } | ||
267 | |||
268 | 6 | return mMask; | |
269 | } | ||
270 | |||
271 | 1014 | void operator()(const Range3d& range) | |
272 | { | ||
273 | ResultLeafNodeType* leaf = nullptr; | ||
274 | |||
275 | // Unpack the range3d item. | ||
276 | const Index imin = range.pages().begin(); | ||
277 | const Index imax = range.pages().end(); | ||
278 | |||
279 | const Index jmin = range.rows().begin(); | ||
280 | const Index jmax = range.rows().end(); | ||
281 | |||
282 | const Index kmin = range.cols().begin(); | ||
283 | const Index kmax = range.cols().end(); | ||
284 | |||
285 | |||
286 | // loop over all the candidate leafs. Adding only those with 'true' values | ||
287 | // to the tree | ||
288 | |||
289 |
2/2✓ Branch 0 taken 1349 times.
✓ Branch 1 taken 507 times.
|
3712 | for (Index i = imin; i < imax; ++i) { |
290 |
2/2✓ Branch 0 taken 4562 times.
✓ Branch 1 taken 1349 times.
|
11822 | for (Index j = jmin; j < jmax; ++j) { |
291 |
2/2✓ Branch 0 taken 13500 times.
✓ Branch 1 taken 4562 times.
|
36124 | for (Index k = kmin; k < kmax; ++k) { |
292 | |||
293 | // Calculate the origin of candidate leaf | ||
294 | const openvdb::math::Coord origin = | ||
295 | 27000 | mMin + openvdb::math::Coord(mWidth * i, | |
296 | mWidth * j, | ||
297 |
2/2✓ Branch 0 taken 510 times.
✓ Branch 1 taken 12990 times.
|
27000 | mWidth * k ); |
298 | |||
299 |
2/2✓ Branch 0 taken 510 times.
✓ Branch 1 taken 12990 times.
|
27000 | if (leaf == nullptr) { |
300 |
1/2✓ Branch 2 taken 307 times.
✗ Branch 3 not taken.
|
1020 | leaf = new ResultLeafNodeType(origin, mBackground); |
301 | } else { | ||
302 | leaf->setOrigin(origin); | ||
303 | leaf->fill(mBackground); | ||
304 | leaf->setValuesOff(); | ||
305 | } | ||
306 | |||
307 | // The bounding box for this leaf | ||
308 | |||
309 | 27000 | openvdb::math::CoordBBox localBBox = leaf->getNodeBoundingBox(); | |
310 | |||
311 | // Shrink to the intersection with mBBox (i.e. the dense | ||
312 | // volume) | ||
313 | |||
314 | 27000 | localBBox.intersect(mBBox); | |
315 | |||
316 | // Early out for non-intersecting leafs | ||
317 | |||
318 | ✗ | if (localBBox.empty()) continue; | |
319 | |||
320 | |||
321 | 27000 | const openvdb::math::Coord start = localBBox.getStart(); | |
322 | const openvdb::math::Coord end = localBBox.getEnd(); | ||
323 | |||
324 | // Order the looping to respect the memory layout in | ||
325 | // the Dense source | ||
326 | |||
327 | if (mDense.memoryLayout() == openvdb::tools::LayoutZYX) { | ||
328 | |||
329 | openvdb::math::Coord ijk; | ||
330 | Index offset; | ||
331 | const DenseValueType* dp; | ||
332 |
2/2✓ Branch 0 taken 68760 times.
✓ Branch 1 taken 9000 times.
|
155520 | for (ijk[0] = start.x(); ijk[0] < end.x(); ++ijk[0] ) { |
333 |
2/2✓ Branch 0 taken 550080 times.
✓ Branch 1 taken 68760 times.
|
1237680 | for (ijk[1] = start.y(); ijk[1] < end.y(); ++ijk[1] ) { |
334 | 8159520 | for (ijk[2] = start.z(), | |
335 | 1100160 | offset = ResultLeafNodeType::coordToOffset(ijk), | |
336 | 1100160 | dp = &mDense.getValue(ijk); | |
337 |
2/2✓ Branch 0 taken 4079760 times.
✓ Branch 1 taken 550080 times.
|
9259680 | ijk[2] < end.z(); ++ijk[2], ++offset, ++dp) { |
338 | |||
339 |
2/2✓ Branch 0 taken 1 times.
✓ Branch 1 taken 2039879 times.
|
8159520 | mFunctor(*dp, offset, leaf); |
340 | } | ||
341 | } | ||
342 | } | ||
343 | |||
344 | } else { | ||
345 | |||
346 | openvdb::math::Coord ijk; | ||
347 | const DenseValueType* dp; | ||
348 |
2/2✓ Branch 0 taken 33375 times.
✓ Branch 1 taken 4500 times.
|
75750 | for (ijk[2] = start.z(); ijk[2] < end.z(); ++ijk[2]) { |
349 |
2/2✓ Branch 0 taken 267000 times.
✓ Branch 1 taken 33375 times.
|
600750 | for (ijk[1] = start.y(); ijk[1] < end.y(); ++ijk[1]) { |
350 | 4079760 | for (ijk[0] = start.x(), | |
351 | 534000 | dp = &mDense.getValue(ijk); | |
352 |
2/2✓ Branch 0 taken 2039880 times.
✓ Branch 1 taken 267000 times.
|
4613760 | ijk[0] < end.x(); ++ijk[0], ++dp) { |
353 | |||
354 |
2/2✓ Branch 0 taken 1 times.
✓ Branch 1 taken 2039879 times.
|
4079760 | mFunctor(*dp, ijk, leaf); |
355 | |||
356 | } | ||
357 | } | ||
358 | } | ||
359 | } | ||
360 | |||
361 | // Only add non-empty leafs (empty is defined as all inactive) | ||
362 | |||
363 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 13497 times.
|
27000 | if (!leaf->isEmpty()) { |
364 | 6 | mMask->addLeaf(leaf); | |
365 | leaf = nullptr; | ||
366 | } | ||
367 | |||
368 | } | ||
369 | } | ||
370 | } | ||
371 | |||
372 | // Clean up an unused leaf. | ||
373 | |||
374 |
1/2✓ Branch 0 taken 507 times.
✗ Branch 1 not taken.
|
1624 | if (leaf != nullptr) delete leaf; |
375 | } | ||
376 | |||
377 | void join(SparseExtractor& rhs) { | ||
378 | 17 | mMask->merge(*rhs.mMask); | |
379 | 17 | } | |
380 | }; // class SparseExtractor | ||
381 | |||
382 | |||
383 | template<typename OpType, typename DenseType> | ||
384 | typename OpType::ResultTreeType::Ptr | ||
385 | 6 | extractSparseTree(const DenseType& dense, const OpType& functor, | |
386 | const typename OpType::ResultValueType& background, | ||
387 | bool threaded) | ||
388 | { | ||
389 | // Construct the mask using a parallel reduce pattern. | ||
390 | // Each thread computes disjoint mask-trees. The join merges | ||
391 | // into a single tree. | ||
392 | |||
393 | 6 | SparseExtractor<OpType, DenseType> extractor(dense, functor, background); | |
394 | |||
395 |
1/2✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
|
12 | return extractor.extract(threaded); |
396 | } | ||
397 | |||
398 | |||
399 | /// @brief Functor-based class used to extract data from a dense grid, at | ||
400 | /// the index-space intersection with a supplied mask in the form of a sparse tree. | ||
401 | /// The @c extractSparseTreeWithMask function wraps this class. | ||
402 | template<typename DenseType, typename MaskTreeType> | ||
403 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | class SparseMaskedExtractor |
404 | { | ||
405 | public: | ||
406 | using _ResultTreeType = typename DSConverter<DenseType, MaskTreeType>::Type; | ||
407 | using ResultTreeType = _ResultTreeType; | ||
408 | using ResultLeafNodeType = typename ResultTreeType::LeafNodeType; | ||
409 | using ResultValueType = typename ResultTreeType::ValueType; | ||
410 | using DenseValueType = ResultValueType; | ||
411 | |||
412 | using MaskTree = typename ResultTreeType::template ValueConverter<ValueMask>::Type; | ||
413 | using MaskLeafCIter = typename MaskTree::LeafCIter; | ||
414 | using MaskLeafVec = std::vector<const typename MaskTree::LeafNodeType*>; | ||
415 | |||
416 | |||
417 | 1 | SparseMaskedExtractor(const DenseType& dense, | |
418 | const ResultValueType& background, | ||
419 | const MaskLeafVec& leafVec | ||
420 | ): | ||
421 | mDense(dense), mBackground(background), mBBox(dense.bbox()), | ||
422 | mLeafVec(leafVec), | ||
423 | 1 | mResult(new ResultTreeType(mBackground)) | |
424 | 1 | {} | |
425 | |||
426 | ✗ | SparseMaskedExtractor(const SparseMaskedExtractor& other, tbb::split): | |
427 | ✗ | mDense(other.mDense), mBackground(other.mBackground), mBBox(other.mBBox), | |
428 | ✗ | mLeafVec(other.mLeafVec), mResult( new ResultTreeType(mBackground)) | |
429 | ✗ | {} | |
430 | |||
431 | 1 | typename ResultTreeType::Ptr extract(bool threaded = true) | |
432 | { | ||
433 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | tbb::blocked_range<size_t> range(0, mLeafVec.size()); |
434 | |||
435 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (threaded) { |
436 | 1 | tbb::parallel_reduce(range, *this); | |
437 | } else { | ||
438 | ✗ | (*this)(range); | |
439 | } | ||
440 | |||
441 | 1 | return mResult; | |
442 | } | ||
443 | |||
444 | // Used in looping over leaf nodes in the masked grid | ||
445 | // and using the active mask to select data to | ||
446 | 2 | void operator()(const tbb::blocked_range<size_t>& range) | |
447 | { | ||
448 | ResultLeafNodeType* leaf = nullptr; | ||
449 | |||
450 | // loop over all the candidate leafs. Adding only those with 'true' values | ||
451 | // to the tree | ||
452 | |||
453 |
2/2✓ Branch 0 taken 2 times.
✓ Branch 1 taken 2 times.
|
4 | for (size_t idx = range.begin(); idx < range.end(); ++ idx) { |
454 | |||
455 | 2 | const typename MaskTree::LeafNodeType* maskLeaf = mLeafVec[idx]; | |
456 | |||
457 | // The bounding box for this leaf | ||
458 | |||
459 | 2 | openvdb::math::CoordBBox localBBox = maskLeaf->getNodeBoundingBox(); | |
460 | |||
461 | // Shrink to the intersection with the dense volume | ||
462 | |||
463 | 2 | localBBox.intersect(mBBox); | |
464 | |||
465 | // Early out if there was no intersection | ||
466 | |||
467 | 1 | if (localBBox.empty()) continue; | |
468 | |||
469 | // Reset or allocate the target leaf | ||
470 | |||
471 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (leaf == nullptr) { |
472 |
1/2✓ Branch 2 taken 1 times.
✗ Branch 3 not taken.
|
1 | leaf = new ResultLeafNodeType(maskLeaf->origin(), mBackground); |
473 | } else { | ||
474 | leaf->setOrigin(maskLeaf->origin()); | ||
475 | leaf->fill(mBackground); | ||
476 | leaf->setValuesOff(); | ||
477 | } | ||
478 | |||
479 | // Iterate over the intersecting bounding box | ||
480 | // copying active values to the result tree | ||
481 | |||
482 | 1 | const openvdb::math::Coord start = localBBox.getStart(); | |
483 | const openvdb::math::Coord end = localBBox.getEnd(); | ||
484 | |||
485 | openvdb::math::Coord ijk; | ||
486 | |||
487 | if (mDense.memoryLayout() == openvdb::tools::LayoutZYX | ||
488 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | && maskLeaf->isDense()) { |
489 | |||
490 | Index offset; | ||
491 | const DenseValueType* src; | ||
492 | ✗ | for (ijk[0] = start.x(); ijk[0] < end.x(); ++ijk[0] ) { | |
493 | ✗ | for (ijk[1] = start.y(); ijk[1] < end.y(); ++ijk[1] ) { | |
494 | ✗ | for (ijk[2] = start.z(), | |
495 | ✗ | offset = ResultLeafNodeType::coordToOffset(ijk), | |
496 | ✗ | src = &mDense.getValue(ijk); | |
497 | ✗ | ijk[2] < end.z(); ++ijk[2], ++offset, ++src) { | |
498 | |||
499 | // copy into leaf | ||
500 | leaf->setValueOn(offset, *src); | ||
501 | } | ||
502 | |||
503 | } | ||
504 | } | ||
505 | |||
506 | } else { | ||
507 | |||
508 | Index offset; | ||
509 |
2/2✓ Branch 0 taken 8 times.
✓ Branch 1 taken 1 times.
|
9 | for (ijk[0] = start.x(); ijk[0] < end.x(); ++ijk[0] ) { |
510 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 8 times.
|
72 | for (ijk[1] = start.y(); ijk[1] < end.y(); ++ijk[1] ) { |
511 | 320 | for (ijk[2] = start.z(), | |
512 | offset = ResultLeafNodeType::coordToOffset(ijk); | ||
513 |
2/2✓ Branch 0 taken 256 times.
✓ Branch 1 taken 64 times.
|
320 | ijk[2] < end.z(); ++ijk[2], ++offset) { |
514 | |||
515 |
2/2✓ Branch 1 taken 1 times.
✓ Branch 2 taken 255 times.
|
256 | if (maskLeaf->isValueOn(offset)) { |
516 | 1 | const ResultValueType denseValue = mDense.getValue(ijk); | |
517 | leaf->setValueOn(offset, denseValue); | ||
518 | } | ||
519 | } | ||
520 | } | ||
521 | } | ||
522 | } | ||
523 | // Only add non-empty leafs (empty is defined as all inactive) | ||
524 | |||
525 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (!leaf->isEmpty()) { |
526 | 1 | mResult->addLeaf(leaf); | |
527 | leaf = nullptr; | ||
528 | } | ||
529 | } | ||
530 | |||
531 | // Clean up an unused leaf. | ||
532 | |||
533 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 2 times.
|
2 | if (leaf != nullptr) delete leaf; |
534 | 2 | } | |
535 | |||
536 | void join(SparseMaskedExtractor& rhs) { | ||
537 | ✗ | mResult->merge(*rhs.mResult); | |
538 | } | ||
539 | |||
540 | |||
541 | private: | ||
542 | const DenseType& mDense; | ||
543 | const ResultValueType mBackground; | ||
544 | const openvdb::math::CoordBBox& mBBox; | ||
545 | const MaskLeafVec& mLeafVec; | ||
546 | |||
547 | typename ResultTreeType::Ptr mResult; | ||
548 | |||
549 | }; // class SparseMaskedExtractor | ||
550 | |||
551 | |||
552 | /// @brief a simple utility class used by @c extractSparseTreeWithMask | ||
553 | template<typename _ResultTreeType, typename DenseValueType> | ||
554 | struct ExtractAll | ||
555 | { | ||
556 | using ResultTreeType = _ResultTreeType; | ||
557 | using ResultLeafNodeType = typename ResultTreeType::LeafNodeType; | ||
558 | |||
559 | template<typename CoordOrIndex> inline void | ||
560 | operator()(const DenseValueType& a, const CoordOrIndex& offset, ResultLeafNodeType* leaf) const | ||
561 | { | ||
562 | ✗ | leaf->setValueOn(offset, a); | |
563 | } | ||
564 | }; | ||
565 | |||
566 | |||
567 | template<typename DenseType, typename MaskTreeType> | ||
568 | typename DSConverter<DenseType, MaskTreeType>::Type::Ptr | ||
569 | 1 | extractSparseTreeWithMask(const DenseType& dense, | |
570 | const MaskTreeType& maskProxy, | ||
571 | const typename DenseType::ValueType& background, | ||
572 | bool threaded) | ||
573 | { | ||
574 | using LeafExtractor = SparseMaskedExtractor<DenseType, MaskTreeType>; | ||
575 | using DenseValueType = typename LeafExtractor::DenseValueType; | ||
576 | using ResultTreeType = typename LeafExtractor::ResultTreeType; | ||
577 | using MaskLeafVec = typename LeafExtractor::MaskLeafVec; | ||
578 | using MaskTree = typename LeafExtractor::MaskTree; | ||
579 | using MaskLeafCIter = typename LeafExtractor::MaskLeafCIter; | ||
580 | using ExtractionRule = ExtractAll<ResultTreeType, DenseValueType>; | ||
581 | |||
582 | // Use Mask tree to hold the topology | ||
583 | |||
584 |
1/2✓ Branch 2 taken 1 times.
✗ Branch 3 not taken.
|
2 | MaskTree maskTree(maskProxy, false, TopologyCopy()); |
585 | |||
586 | // Construct an array of pointers to the mask leafs. | ||
587 | |||
588 | 1 | const size_t leafCount = maskTree.leafCount(); | |
589 |
2/4✓ Branch 1 taken 1 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 1 times.
✗ Branch 5 not taken.
|
1 | MaskLeafVec leafarray(leafCount); |
590 | MaskLeafCIter leafiter = maskTree.cbeginLeaf(); | ||
591 |
2/2✓ Branch 0 taken 2 times.
✓ Branch 1 taken 1 times.
|
3 | for (size_t n = 0; n != leafCount; ++n, ++leafiter) { |
592 |
1/2✓ Branch 1 taken 2 times.
✗ Branch 2 not taken.
|
2 | leafarray[n] = leafiter.getLeaf(); |
593 | } | ||
594 | |||
595 | |||
596 | // Extract the data that is masked leaf nodes in the mask. | ||
597 | |||
598 |
1/2✓ Branch 1 taken 1 times.
✗ Branch 2 not taken.
|
1 | LeafExtractor leafextractor(dense, background, leafarray); |
599 |
1/2✓ Branch 1 taken 1 times.
✗ Branch 2 not taken.
|
1 | typename ResultTreeType::Ptr resultTree = leafextractor.extract(threaded); |
600 | |||
601 | |||
602 | // Extract data that is masked by tiles in the mask. | ||
603 | |||
604 | |||
605 | // Loop over the mask tiles, extracting the data into new trees. | ||
606 | // These trees will be leaf-orthogonal to the leafTree (i.e. no leaf | ||
607 | // nodes will overlap). Merge these trees into the result. | ||
608 | |||
609 |
1/2✓ Branch 1 taken 1 times.
✗ Branch 2 not taken.
|
1 | typename MaskTreeType::ValueOnCIter tileIter(maskProxy); |
610 | 1 | tileIter.setMaxDepth(MaskTreeType::ValueOnCIter::LEAF_DEPTH - 1); | |
611 | |||
612 | // Return the leaf tree if the mask had no tiles | ||
613 | |||
614 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1 times.
|
1 | if (!tileIter) return resultTree; |
615 | |||
616 | ExtractionRule allrule; | ||
617 | |||
618 | // Loop over the tiles in series, but the actual data extraction | ||
619 | // is in parallel. | ||
620 | |||
621 | ✗ | CoordBBox bbox; | |
622 | ✗ | for ( ; tileIter; ++tileIter) { | |
623 | |||
624 | // Find the intersection of the tile with the dense grid. | ||
625 | |||
626 | ✗ | tileIter.getBoundingBox(bbox); | |
627 | ✗ | bbox.intersect(dense.bbox()); | |
628 | |||
629 | ✗ | if (bbox.empty()) continue; | |
630 | |||
631 | ✗ | SparseExtractor<ExtractionRule, DenseType> copyData(dense, bbox, allrule, background); | |
632 | ✗ | typename ResultTreeType::Ptr fromTileTree = copyData.extract(threaded); | |
633 | ✗ | resultTree->merge(*fromTileTree); | |
634 | } | ||
635 | |||
636 | return resultTree; | ||
637 | } | ||
638 | |||
639 | |||
640 | /// @brief Class that applies a functor to the index space intersection | ||
641 | /// of a prescribed bounding box and the dense grid. | ||
642 | /// NB: This class only supports DenseGrids with ZYX memory layout. | ||
643 | template<typename _ValueT, typename OpType> | ||
644 | class DenseTransformer | ||
645 | { | ||
646 | public: | ||
647 | using ValueT = _ValueT; | ||
648 | using DenseT = Dense<ValueT, openvdb::tools::LayoutZYX>; | ||
649 | using IntType = openvdb::math::Coord::ValueType; | ||
650 | using RangeType = tbb::blocked_range2d<IntType, IntType>; | ||
651 | |||
652 | private: | ||
653 | DenseT& mDense; | ||
654 | const OpType& mOp; | ||
655 | openvdb::math::CoordBBox mBBox; | ||
656 | |||
657 | public: | ||
658 | 1 | DenseTransformer(DenseT& dense, const openvdb::math::CoordBBox& bbox, const OpType& functor): | |
659 | 1 | mDense(dense), mOp(functor), mBBox(dense.bbox()) | |
660 | { | ||
661 | // The iteration space is the intersection of the | ||
662 | // input bbox and the index-space covered by the dense grid | ||
663 | 1 | mBBox.intersect(bbox); | |
664 | } | ||
665 | |||
666 | 20 | DenseTransformer(const DenseTransformer& other) : | |
667 | 2 | mDense(other.mDense), mOp(other.mOp), mBBox(other.mBBox) {} | |
668 | |||
669 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
2 | void apply(bool threaded = true) { |
670 | |||
671 | // Early out if the iteration space is empty | ||
672 | |||
673 | ✗ | if (mBBox.empty()) return; | |
674 | |||
675 | |||
676 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
2 | const openvdb::math::Coord start = mBBox.getStart(); |
677 | const openvdb::math::Coord end = mBBox.getEnd(); | ||
678 | |||
679 | // The iteration range only the slower two directions. | ||
680 | const RangeType range(start.x(), end.x(), 1, | ||
681 | start.y(), end.y(), 1); | ||
682 | |||
683 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
2 | if (threaded) { |
684 | 2 | tbb::parallel_for(range, *this); | |
685 | } else { | ||
686 | ✗ | (*this)(range); | |
687 | } | ||
688 | } | ||
689 | |||
690 | 70 | void operator()(const RangeType& range) const { | |
691 | |||
692 | // The stride in the z-direction. | ||
693 | // Note: the bbox is [inclusive, inclusive] | ||
694 | |||
695 | 70 | const size_t zlength = size_t(mBBox.max().z() - mBBox.min().z() + 1); | |
696 | |||
697 | const IntType imin = range.rows().begin(); | ||
698 | const IntType imax = range.rows().end(); | ||
699 | const IntType jmin = range.cols().begin(); | ||
700 | const IntType jmax = range.cols().end(); | ||
701 | |||
702 | |||
703 | openvdb::math::Coord xyz(imin, jmin, mBBox.min().z()); | ||
704 |
2/2✓ Branch 0 taken 35 times.
✓ Branch 1 taken 35 times.
|
140 | for (xyz[0] = imin; xyz[0] != imax; ++xyz[0]) { |
705 |
2/2✓ Branch 0 taken 35 times.
✓ Branch 1 taken 35 times.
|
140 | for (xyz[1] = jmin; xyz[1] != jmax; ++xyz[1]) { |
706 | |||
707 | 70 | mOp.transform(mDense, xyz, zlength); | |
708 | } | ||
709 | } | ||
710 | } | ||
711 | }; // class DenseTransformer | ||
712 | |||
713 | |||
714 | /// @brief a wrapper struct used to avoid unnecessary computation of | ||
715 | /// memory access from @c Coord when all offsets are guaranteed to be | ||
716 | /// within the dense grid. | ||
717 | template<typename ValueT, typename PointWiseOp> | ||
718 | struct ContiguousOp | ||
719 | { | ||
720 | 1 | ContiguousOp(const PointWiseOp& op) : mOp(op){} | |
721 | |||
722 | using DenseT = Dense<ValueT, openvdb::tools::LayoutZYX>; | ||
723 | 70 | inline void transform(DenseT& dense, openvdb::math::Coord& ijk, size_t size) const | |
724 | { | ||
725 | ValueT* dp = const_cast<ValueT*>(&dense.getValue(ijk)); | ||
726 | |||
727 |
2/2✓ Branch 0 taken 105 times.
✓ Branch 1 taken 35 times.
|
280 | for (size_t offset = 0; offset < size; ++offset) { |
728 | 210 | dp[offset] = mOp(dp[offset]); | |
729 | } | ||
730 | } | ||
731 | |||
732 | const PointWiseOp mOp; | ||
733 | }; | ||
734 | |||
735 | |||
736 | /// Apply a point-wise functor to the intersection of a dense grid and a given bounding box | ||
737 | template<typename ValueT, typename PointwiseOpT> | ||
738 | void | ||
739 | 2 | transformDense(Dense<ValueT, openvdb::tools::LayoutZYX>& dense, | |
740 | const openvdb::CoordBBox& bbox, | ||
741 | const PointwiseOpT& functor, bool parallel) | ||
742 | { | ||
743 | using OpT = ContiguousOp<ValueT, PointwiseOpT>; | ||
744 | |||
745 | // Convert the Op so it operates on a contiguous line in memory | ||
746 | |||
747 | OpT op(functor); | ||
748 | |||
749 | // Apply to the index space intersection in the dense grid | ||
750 | DenseTransformer<ValueT, OpT> transformer(dense, bbox, op); | ||
751 | 2 | transformer.apply(parallel); | |
752 | } | ||
753 | |||
754 | |||
755 | template<typename CompositeMethod, typename _TreeT> | ||
756 | class SparseToDenseCompositor | ||
757 | { | ||
758 | public: | ||
759 | using TreeT = _TreeT; | ||
760 | using ValueT = typename TreeT::ValueType; | ||
761 | using LeafT = typename TreeT::LeafNodeType; | ||
762 | using MaskTreeT = typename TreeT::template ValueConverter<ValueMask>::Type; | ||
763 | using MaskLeafT = typename MaskTreeT::LeafNodeType; | ||
764 | using DenseT = Dense<ValueT, openvdb::tools::LayoutZYX>; | ||
765 | using Index = openvdb::math::Coord::ValueType; | ||
766 | using Range3d = tbb::blocked_range3d<Index, Index, Index>; | ||
767 | |||
768 | 3 | SparseToDenseCompositor(DenseT& dense, const TreeT& source, const TreeT& alpha, | |
769 | const ValueT beta, const ValueT strength) : | ||
770 |
3/6✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
✓ Branch 3 taken 1 times.
✗ Branch 4 not taken.
✓ Branch 6 taken 1 times.
✗ Branch 7 not taken.
|
3 | mDense(dense), mSource(source), mAlpha(alpha), mBeta(beta), mStrength(strength) |
771 | {} | ||
772 | |||
773 | 8 | SparseToDenseCompositor(const SparseToDenseCompositor& other): | |
774 | 8 | mDense(other.mDense), mSource(other.mSource), mAlpha(other.mAlpha), | |
775 |
1/2✓ Branch 4 taken 2 times.
✗ Branch 5 not taken.
|
5 | mBeta(other.mBeta), mStrength(other.mStrength) {} |
776 | |||
777 | |||
778 | 2 | void sparseComposite(bool threaded) | |
779 | { | ||
780 | 2 | const ValueT beta = mBeta; | |
781 | 2 | const ValueT strength = mStrength; | |
782 | |||
783 | // construct a tree that defines the iteration space | ||
784 | |||
785 | 2 | MaskTreeT maskTree(mSource, false /*background*/, openvdb::TopologyCopy()); | |
786 |
1/2✓ Branch 1 taken 2 times.
✗ Branch 2 not taken.
|
2 | maskTree.topologyUnion(mAlpha); |
787 | |||
788 | // Composite regions that are represented by leafnodes in either mAlpha or mSource | ||
789 | // Parallelize over bool-leafs | ||
790 | |||
791 |
1/2✓ Branch 1 taken 2 times.
✗ Branch 2 not taken.
|
2 | openvdb::tree::LeafManager<const MaskTreeT> maskLeafs(maskTree); |
792 |
1/2✓ Branch 1 taken 2 times.
✗ Branch 2 not taken.
|
2 | maskLeafs.foreach(*this, threaded); |
793 | |||
794 | // Composite regions that are represented by tiles | ||
795 | // Parallelize within each tile. | ||
796 | |||
797 | typename MaskTreeT::ValueOnCIter citer = maskTree.cbeginValueOn(); | ||
798 | 2 | citer.setMaxDepth(MaskTreeT::ValueOnCIter::LEAF_DEPTH - 1); | |
799 | |||
800 |
1/2✓ Branch 0 taken 2 times.
✗ Branch 1 not taken.
|
2 | if (!citer) return; |
801 | |||
802 | ✗ | typename tree::ValueAccessor<const TreeT> alphaAccessor(mAlpha); | |
803 | ✗ | typename tree::ValueAccessor<const TreeT> sourceAccessor(mSource); | |
804 | |||
805 | ✗ | for (; citer; ++citer) { | |
806 | |||
807 | ✗ | const openvdb::math::Coord org = citer.getCoord(); | |
808 | |||
809 | // Early out if both alpha and source are zero in this tile. | ||
810 | |||
811 | ✗ | const ValueT alphaValue = alphaAccessor.getValue(org); | |
812 | ✗ | const ValueT sourceValue = sourceAccessor.getValue(org); | |
813 | |||
814 | ✗ | if (openvdb::math::isZero(alphaValue) && | |
815 | ✗ | openvdb::math::isZero(sourceValue)) continue; | |
816 | |||
817 | // Compute overlap of tile with the dense grid | ||
818 | |||
819 | openvdb::math::CoordBBox localBBox = citer.getBoundingBox(); | ||
820 | ✗ | localBBox.intersect(mDense.bbox()); | |
821 | |||
822 | // Early out if there is no intersection | ||
823 | |||
824 | ✗ | if (localBBox.empty()) continue; | |
825 | |||
826 | // Composite the tile-uniform values into the dense grid. | ||
827 | ✗ | compositeFromTile(mDense, localBBox, sourceValue, | |
828 | alphaValue, beta, strength, threaded); | ||
829 | } | ||
830 | } | ||
831 | |||
832 | // Composites leaf values where the alpha values are active. | ||
833 | // Used in sparseComposite | ||
834 | 2 | void inline operator()(const MaskLeafT& maskLeaf, size_t /*i*/) const | |
835 | { | ||
836 | using ULeaf = UniformLeaf; | ||
837 | 2 | openvdb::math::CoordBBox localBBox = maskLeaf.getNodeBoundingBox(); | |
838 | 2 | localBBox.intersect(mDense.bbox()); | |
839 | |||
840 | // Early out for non-overlapping leafs | ||
841 | |||
842 | ✗ | if (localBBox.empty()) return; | |
843 | |||
844 | 2 | const openvdb::math::Coord org = maskLeaf.origin(); | |
845 | 2 | const LeafT* alphaLeaf = mAlpha.probeLeaf(org); | |
846 | 2 | const LeafT* sourceLeaf = mSource.probeLeaf(org); | |
847 | |||
848 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 2 times.
|
2 | if (!sourceLeaf) { |
849 | |||
850 | // Create a source leaf proxy with the correct value | ||
851 | ✗ | ULeaf uniformSource(mSource.getValue(org)); | |
852 | |||
853 | ✗ | if (!alphaLeaf) { | |
854 | |||
855 | // Create an alpha leaf proxy with the correct value | ||
856 | ✗ | ULeaf uniformAlpha(mAlpha.getValue(org)); | |
857 | |||
858 | ✗ | compositeFromLeaf(mDense, localBBox, uniformSource, uniformAlpha, | |
859 | ✗ | mBeta, mStrength); | |
860 | } else { | ||
861 | |||
862 | ✗ | compositeFromLeaf(mDense, localBBox, uniformSource, *alphaLeaf, | |
863 | ✗ | mBeta, mStrength); | |
864 | } | ||
865 | } else { | ||
866 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 2 times.
|
2 | if (!alphaLeaf) { |
867 | |||
868 | // Create an alpha leaf proxy with the correct value | ||
869 | ✗ | ULeaf uniformAlpha(mAlpha.getValue(org)); | |
870 | |||
871 | ✗ | compositeFromLeaf(mDense, localBBox, *sourceLeaf, uniformAlpha, | |
872 | ✗ | mBeta, mStrength); | |
873 | } else { | ||
874 | |||
875 | 2 | compositeFromLeaf(mDense, localBBox, *sourceLeaf, *alphaLeaf, | |
876 | 2 | mBeta, mStrength); | |
877 | } | ||
878 | } | ||
879 | } | ||
880 | // i.e. it assumes that all valueOff Alpha voxels have value 0. | ||
881 | |||
882 | template<typename LeafT1, typename LeafT2> | ||
883 | 4 | inline static void compositeFromLeaf(DenseT& dense, const openvdb::math::CoordBBox& bbox, | |
884 | const LeafT1& source, const LeafT2& alpha, | ||
885 | const ValueT beta, const ValueT strength) | ||
886 | { | ||
887 | using IntType = openvdb::math::Coord::ValueType; | ||
888 | |||
889 | const ValueT sbeta = strength * beta; | ||
890 | 4 | openvdb::math::Coord ijk = bbox.min(); | |
891 | |||
892 | |||
893 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 2 times.
|
4 | if (alpha.isDense() /*all active values*/) { |
894 | |||
895 | // Optimal path for dense alphaLeaf | ||
896 | ✗ | const IntType size = bbox.max().z() + 1 - bbox.min().z(); | |
897 | |||
898 | ✗ | for (ijk[0] = bbox.min().x(); ijk[0] < bbox.max().x() + 1; ++ijk[0]) { | |
899 | ✗ | for (ijk[1] = bbox.min().y(); ijk[1] < bbox.max().y() + 1; ++ijk[1]) { | |
900 | |||
901 | ValueT* d = const_cast<ValueT*>(&dense.getValue(ijk)); | ||
902 | ✗ | const ValueT* a = &alpha.getValue(ijk); | |
903 | ✗ | const ValueT* s = &source.getValue(ijk); | |
904 | |||
905 | ✗ | for (IntType idx = 0; idx < size; ++idx) { | |
906 | ✗ | d[idx] = CompositeMethod::apply(d[idx], a[idx], s[idx], | |
907 | strength, beta, sbeta); | ||
908 | } | ||
909 | } | ||
910 | } | ||
911 | } else { | ||
912 | |||
913 | // AlphaLeaf has non-active cells. | ||
914 | |||
915 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 2 times.
|
12 | for (ijk[0] = bbox.min().x(); ijk[0] < bbox.max().x() + 1; ++ijk[0]) { |
916 |
2/2✓ Branch 0 taken 24 times.
✓ Branch 1 taken 4 times.
|
56 | for (ijk[1] = bbox.min().y(); ijk[1] < bbox.max().y() + 1; ++ijk[1]) { |
917 |
2/2✓ Branch 0 taken 72 times.
✓ Branch 1 taken 24 times.
|
192 | for (ijk[2] = bbox.min().z(); ijk[2] < bbox.max().z() + 1; ++ijk[2]) { |
918 | |||
919 |
2/2✓ Branch 1 taken 2 times.
✓ Branch 2 taken 70 times.
|
144 | if (alpha.isValueOn(ijk)) { |
920 | 4 | dense.setValue(ijk, CompositeMethod::apply(dense.getValue(ijk), | |
921 | alpha.getValue(ijk), source.getValue(ijk), strength, beta, sbeta)); | ||
922 | } | ||
923 | } | ||
924 | } | ||
925 | } | ||
926 | } | ||
927 | } | ||
928 | |||
929 | inline static void compositeFromTile(DenseT& dense, openvdb::math::CoordBBox& bbox, | ||
930 | const ValueT& sourceValue, const ValueT& alphaValue, | ||
931 | const ValueT& beta, const ValueT& strength, | ||
932 | bool threaded) | ||
933 | { | ||
934 | using TileTransformer = UniformTransformer; | ||
935 | TileTransformer functor(sourceValue, alphaValue, beta, strength); | ||
936 | |||
937 | // Transform the data inside the bbox according to the TileTranformer. | ||
938 | |||
939 | ✗ | transformDense(dense, bbox, functor, threaded); | |
940 | } | ||
941 | |||
942 | 1 | void denseComposite(bool threaded) | |
943 | { | ||
944 | /// Construct a range that corresponds to the | ||
945 | /// bounding box of the dense volume | ||
946 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | const openvdb::math::CoordBBox& bbox = mDense.bbox(); |
947 | |||
948 | Range3d range(bbox.min().x(), bbox.max().x(), LeafT::DIM, | ||
949 | bbox.min().y(), bbox.max().y(), LeafT::DIM, | ||
950 | bbox.min().z(), bbox.max().z(), LeafT::DIM); | ||
951 | |||
952 | // Iterate over the range, compositing into | ||
953 | // the dense grid using value accessors for | ||
954 | // sparse the grids. | ||
955 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (threaded) { |
956 | 1 | tbb::parallel_for(range, *this); | |
957 | } else { | ||
958 | ✗ | (*this)(range); | |
959 | } | ||
960 | 1 | } | |
961 | |||
962 | // Composites a dense region using value accessors | ||
963 | // into a dense grid | ||
964 | 4 | void operator()(const Range3d& range) const | |
965 | { | ||
966 | // Use value accessors to alpha and source | ||
967 | |||
968 | 4 | typename tree::ValueAccessor<const TreeT> alphaAccessor(mAlpha); | |
969 |
1/2✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
|
4 | typename tree::ValueAccessor<const TreeT> sourceAccessor(mSource); |
970 | |||
971 | 4 | const ValueT strength = mStrength; | |
972 | 4 | const ValueT beta = mBeta; | |
973 | const ValueT sbeta = strength * beta; | ||
974 | |||
975 | // Unpack the range3d item. | ||
976 | const Index imin = range.pages().begin(); | ||
977 | const Index imax = range.pages().end(); | ||
978 | |||
979 | const Index jmin = range.rows().begin(); | ||
980 | const Index jmax = range.rows().end(); | ||
981 | |||
982 | const Index kmin = range.cols().begin(); | ||
983 | const Index kmax = range.cols().end(); | ||
984 | |||
985 | openvdb::Coord ijk; | ||
986 |
2/2✓ Branch 0 taken 20 times.
✓ Branch 1 taken 4 times.
|
24 | for (ijk[0] = imin; ijk[0] < imax; ++ijk[0]) { |
987 |
2/2✓ Branch 0 taken 100 times.
✓ Branch 1 taken 20 times.
|
120 | for (ijk[1] = jmin; ijk[1] < jmax; ++ijk[1]) { |
988 |
2/2✓ Branch 0 taken 500 times.
✓ Branch 1 taken 100 times.
|
600 | for (ijk[2] = kmin; ijk[2] < kmax; ++ijk[2]) { |
989 | 500 | const ValueT d_old = mDense.getValue(ijk); | |
990 |
1/2✓ Branch 1 taken 500 times.
✗ Branch 2 not taken.
|
500 | const ValueT& alpha = alphaAccessor.getValue(ijk); |
991 |
1/2✓ Branch 1 taken 500 times.
✗ Branch 2 not taken.
|
500 | const ValueT& src = sourceAccessor.getValue(ijk); |
992 | |||
993 | 500 | mDense.setValue(ijk, | |
994 | 500 | CompositeMethod::apply(d_old, alpha, src, strength, beta, sbeta)); | |
995 | } | ||
996 | } | ||
997 | } | ||
998 | 4 | } | |
999 | |||
1000 | private: | ||
1001 | // Internal class that wraps the templated composite method | ||
1002 | // for use when both alpha and source are uniform over | ||
1003 | // a prescribed bbox (e.g. a tile). | ||
1004 | class UniformTransformer | ||
1005 | { | ||
1006 | public: | ||
1007 | ✗ | UniformTransformer(const ValueT& source, const ValueT& alpha, const ValueT& _beta, | |
1008 | const ValueT& _strength) : | ||
1009 | mSource(source), mAlpha(alpha), mBeta(_beta), | ||
1010 | ✗ | mStrength(_strength), mSBeta(_strength * _beta) | |
1011 | {} | ||
1012 | |||
1013 | ValueT operator()(const ValueT& input) const | ||
1014 | { | ||
1015 | ✗ | return CompositeMethod::apply(input, mAlpha, mSource, mStrength, mBeta, mSBeta); | |
1016 | } | ||
1017 | |||
1018 | private: | ||
1019 | const ValueT mSource; | ||
1020 | const ValueT mAlpha; | ||
1021 | const ValueT mBeta; | ||
1022 | const ValueT mStrength; | ||
1023 | const ValueT mSBeta; | ||
1024 | }; | ||
1025 | |||
1026 | |||
1027 | // Simple Class structure that mimics a leaf | ||
1028 | // with uniform values. Holds LeafT::DIM copies | ||
1029 | // of a value in an array. | ||
1030 | struct Line { ValueT mValues[LeafT::DIM]; }; | ||
1031 | class UniformLeaf : private Line | ||
1032 | { | ||
1033 | public: | ||
1034 | using ValueT = typename LeafT::ValueType; | ||
1035 | |||
1036 | using BaseT = Line; | ||
1037 | ✗ | UniformLeaf(const ValueT& value) : BaseT(init(value)) {} | |
1038 | |||
1039 | static const BaseT init(const ValueT& value) { | ||
1040 | BaseT tmp; | ||
1041 | ✗ | for (openvdb::Index i = 0; i < LeafT::DIM; ++i) { | |
1042 | ✗ | tmp.mValues[i] = value; | |
1043 | } | ||
1044 | return tmp; | ||
1045 | } | ||
1046 | |||
1047 | bool isDense() const { return true; } | ||
1048 | bool isValueOn(openvdb::math::Coord&) const { return true; } | ||
1049 | |||
1050 | ✗ | const ValueT& getValue(const openvdb::math::Coord&) const { return BaseT::mValues[0]; } | |
1051 | }; | ||
1052 | |||
1053 | private: | ||
1054 | DenseT& mDense; | ||
1055 | const TreeT& mSource; | ||
1056 | const TreeT& mAlpha; | ||
1057 | ValueT mBeta; | ||
1058 | ValueT mStrength; | ||
1059 | }; // class SparseToDenseCompositor | ||
1060 | |||
1061 | |||
1062 | namespace ds | ||
1063 | { | ||
1064 | //@{ | ||
1065 | /// @brief Point wise methods used to apply various compositing operations. | ||
1066 | template<typename ValueT> | ||
1067 | struct OpOver | ||
1068 | { | ||
1069 | static inline ValueT apply(const ValueT u, const ValueT alpha, | ||
1070 | const ValueT v, | ||
1071 | const ValueT strength, | ||
1072 | const ValueT beta, | ||
1073 | const ValueT /*sbeta*/) | ||
1074 |
1/2✓ Branch 2 taken 1 times.
✗ Branch 3 not taken.
|
503 | { return (u + strength * alpha * (beta * v - u)); } |
1075 | }; | ||
1076 | |||
1077 | template<typename ValueT> | ||
1078 | struct OpAdd | ||
1079 | { | ||
1080 | static inline ValueT apply(const ValueT u, const ValueT alpha, | ||
1081 | const ValueT v, | ||
1082 | const ValueT /*strength*/, | ||
1083 | const ValueT /*beta*/, | ||
1084 | const ValueT sbeta) | ||
1085 | { return (u + sbeta * alpha * v); } | ||
1086 | }; | ||
1087 | |||
1088 | template<typename ValueT> | ||
1089 | struct OpSub | ||
1090 | { | ||
1091 | static inline ValueT apply(const ValueT u, const ValueT alpha, | ||
1092 | const ValueT v, | ||
1093 | const ValueT /*strength*/, | ||
1094 | const ValueT /*beta*/, | ||
1095 | const ValueT sbeta) | ||
1096 | { return (u - sbeta * alpha * v); } | ||
1097 | }; | ||
1098 | |||
1099 | template<typename ValueT> | ||
1100 | struct OpMin | ||
1101 | { | ||
1102 | static inline ValueT apply(const ValueT u, const ValueT alpha, | ||
1103 | const ValueT v, | ||
1104 | const ValueT s /*trength*/, | ||
1105 | const ValueT beta, | ||
1106 | const ValueT /*sbeta*/) | ||
1107 | { return ( ( 1 - s * alpha) * u + s * alpha * std::min(u, beta * v) ); } | ||
1108 | }; | ||
1109 | |||
1110 | template<typename ValueT> | ||
1111 | struct OpMax | ||
1112 | { | ||
1113 | static inline ValueT apply(const ValueT u, const ValueT alpha, | ||
1114 | const ValueT v, | ||
1115 | const ValueT s/*trength*/, | ||
1116 | const ValueT beta, | ||
1117 | const ValueT /*sbeta*/) | ||
1118 | { return ( ( 1 - s * alpha ) * u + s * alpha * std::min(u, beta * v) ); } | ||
1119 | }; | ||
1120 | |||
1121 | template<typename ValueT> | ||
1122 | struct OpMult | ||
1123 | { | ||
1124 | static inline ValueT apply(const ValueT u, const ValueT alpha, | ||
1125 | const ValueT v, | ||
1126 | const ValueT s/*trength*/, | ||
1127 | const ValueT /*beta*/, | ||
1128 | const ValueT sbeta) | ||
1129 | { return ( ( 1 + alpha * (sbeta * v - s)) * u ); } | ||
1130 | }; | ||
1131 | //@} | ||
1132 | |||
1133 | //@{ | ||
1134 | /// Translator that converts an enum to compositing functor types | ||
1135 | template<DSCompositeOp OP, typename ValueT> | ||
1136 | struct CompositeFunctorTranslator{}; | ||
1137 | |||
1138 | template<typename ValueT> | ||
1139 | struct CompositeFunctorTranslator<DS_OVER, ValueT>{ using OpT = OpOver<ValueT>; }; | ||
1140 | |||
1141 | template<typename ValueT> | ||
1142 | struct CompositeFunctorTranslator<DS_ADD, ValueT>{ using OpT = OpAdd<ValueT>; }; | ||
1143 | |||
1144 | template<typename ValueT> | ||
1145 | struct CompositeFunctorTranslator<DS_SUB, ValueT>{ using OpT = OpSub<ValueT>; }; | ||
1146 | |||
1147 | template<typename ValueT> | ||
1148 | struct CompositeFunctorTranslator<DS_MIN, ValueT>{ using OpT = OpMin<ValueT>; }; | ||
1149 | |||
1150 | template<typename ValueT> | ||
1151 | struct CompositeFunctorTranslator<DS_MAX, ValueT>{ using OpT = OpMax<ValueT>; }; | ||
1152 | |||
1153 | template<typename ValueT> | ||
1154 | struct CompositeFunctorTranslator<DS_MULT, ValueT>{ using OpT = OpMult<ValueT>; }; | ||
1155 | //@} | ||
1156 | |||
1157 | } // namespace ds | ||
1158 | |||
1159 | |||
1160 | template<DSCompositeOp OpT, typename TreeT> | ||
1161 | inline void | ||
1162 | 1 | compositeToDense( | |
1163 | Dense<typename TreeT::ValueType, LayoutZYX>& dense, | ||
1164 | const TreeT& source, const TreeT& alpha, | ||
1165 | const typename TreeT::ValueType beta, | ||
1166 | const typename TreeT::ValueType strength, | ||
1167 | bool threaded) | ||
1168 | { | ||
1169 | using ValueT = typename TreeT::ValueType; | ||
1170 | using Translator = ds::CompositeFunctorTranslator<OpT, ValueT>; | ||
1171 | using Method = typename Translator::OpT; | ||
1172 | |||
1173 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1 times.
|
1 | if (openvdb::math::isZero(strength)) return; |
1174 | |||
1175 | SparseToDenseCompositor<Method, TreeT> tool(dense, source, alpha, beta, strength); | ||
1176 | |||
1177 |
2/4✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 1 times.
✗ Branch 3 not taken.
|
1 | if (openvdb::math::isZero(alpha.background()) && |
1178 | openvdb::math::isZero(source.background())) | ||
1179 | { | ||
1180 | // Use the sparsity of (alpha U source) as the iteration space. | ||
1181 | 1 | tool.sparseComposite(threaded); | |
1182 | } else { | ||
1183 | // Use the bounding box of dense as the iteration space. | ||
1184 | ✗ | tool.denseComposite(threaded); | |
1185 | } | ||
1186 | } | ||
1187 | |||
1188 | } // namespace tools | ||
1189 | } // namespace OPENVDB_VERSION_NAME | ||
1190 | } // namespace openvdb | ||
1191 | |||
1192 | #endif //OPENVDB_TOOLS_DENSESPARSETOOLS_HAS_BEEN_INCLUDED | ||
1193 |