summaryrefslogtreecommitdiff
path: root/thirdparty/embree-aarch64/kernels/builders/primrefgen.cpp
blob: e23de3df28e73c1d0741e318c3a6dc5c87b19e77 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
// Copyright 2009-2020 Intel Corporation
// SPDX-License-Identifier: Apache-2.0

#include "primrefgen.h"
#include "primrefgen_presplit.h"

#include "../../common/algorithms/parallel_for_for.h"
#include "../../common/algorithms/parallel_for_for_prefix_sum.h"

namespace embree
{
  namespace isa
  {
    PrimInfo createPrimRefArray(Geometry* geometry, unsigned int geomID, mvector<PrimRef>& prims, BuildProgressMonitor& progressMonitor)
    {
      ParallelPrefixSumState<PrimInfo> pstate;
      
      /* first try */
      progressMonitor(0);
      PrimInfo pinfo = parallel_prefix_sum( pstate, size_t(0), geometry->size(), size_t(1024), PrimInfo(empty), [&](const range<size_t>& r, const PrimInfo& base) -> PrimInfo {
          return geometry->createPrimRefArray(prims,r,r.begin(),geomID);
        }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });

      /* if we need to filter out geometry, run again */
      if (pinfo.size() != prims.size())
      {
        progressMonitor(0);
        pinfo = parallel_prefix_sum( pstate, size_t(0), geometry->size(), size_t(1024), PrimInfo(empty), [&](const range<size_t>& r, const PrimInfo& base) -> PrimInfo {
          return geometry->createPrimRefArray(prims,r,base.size(),geomID);
        }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      }
      return pinfo;
    }

    PrimInfo createPrimRefArray(Scene* scene, Geometry::GTypeMask types, bool mblur, mvector<PrimRef>& prims, BuildProgressMonitor& progressMonitor)
    {
      ParallelForForPrefixSumState<PrimInfo> pstate;
      Scene::Iterator2 iter(scene,types,mblur);
      
      /* first try */
      progressMonitor(0);
      pstate.init(iter,size_t(1024));
      PrimInfo pinfo = parallel_for_for_prefix_sum0( pstate, iter, PrimInfo(empty), [&](Geometry* mesh, const range<size_t>& r, size_t k, size_t geomID) -> PrimInfo {
          return mesh->createPrimRefArray(prims,r,k,(unsigned)geomID);
        }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      
      /* if we need to filter out geometry, run again */
      if (pinfo.size() != prims.size())
      {
        progressMonitor(0);
        pinfo = parallel_for_for_prefix_sum1( pstate, iter, PrimInfo(empty), [&](Geometry* mesh, const range<size_t>& r, size_t k, size_t geomID, const PrimInfo& base) -> PrimInfo {
            return mesh->createPrimRefArray(prims,r,base.size(),(unsigned)geomID);
          }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      }
      return pinfo;
    }

    PrimInfo createPrimRefArrayMBlur(Scene* scene, Geometry::GTypeMask types, mvector<PrimRef>& prims, BuildProgressMonitor& progressMonitor, size_t itime)
    {
      ParallelForForPrefixSumState<PrimInfo> pstate;
      Scene::Iterator2 iter(scene,types,true);
      
      /* first try */
      progressMonitor(0);
      pstate.init(iter,size_t(1024));
      PrimInfo pinfo = parallel_for_for_prefix_sum0( pstate, iter, PrimInfo(empty), [&](Geometry* mesh, const range<size_t>& r, size_t k, size_t geomID) -> PrimInfo {
          return mesh->createPrimRefArrayMB(prims,itime,r,k,(unsigned)geomID);
        }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      
      /* if we need to filter out geometry, run again */
      if (pinfo.size() != prims.size())
      {
        progressMonitor(0);
        pinfo = parallel_for_for_prefix_sum1( pstate, iter, PrimInfo(empty), [&](Geometry* mesh, const range<size_t>& r, size_t k, size_t geomID, const PrimInfo& base) -> PrimInfo {
            return mesh->createPrimRefArrayMB(prims,itime,r,base.size(),(unsigned)geomID);
          }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      }
      return pinfo;
    }

    PrimInfoMB createPrimRefArrayMSMBlur(Scene* scene, Geometry::GTypeMask types, mvector<PrimRefMB>& prims, BuildProgressMonitor& progressMonitor, BBox1f t0t1)
    {
      ParallelForForPrefixSumState<PrimInfoMB> pstate;
      Scene::Iterator2 iter(scene,types,true);
      
      /* first try */
      progressMonitor(0);
      pstate.init(iter,size_t(1024));
      PrimInfoMB pinfo = parallel_for_for_prefix_sum0( pstate, iter, PrimInfoMB(empty), [&](Geometry* mesh, const range<size_t>& r, size_t k, size_t geomID) -> PrimInfoMB {
          return mesh->createPrimRefMBArray(prims,t0t1,r,k,(unsigned)geomID);
      }, [](const PrimInfoMB& a, const PrimInfoMB& b) -> PrimInfoMB { return PrimInfoMB::merge2(a,b); });
      
      /* if we need to filter out geometry, run again */
      if (pinfo.size() != prims.size())
      {
        progressMonitor(0);
        pinfo = parallel_for_for_prefix_sum1( pstate, iter, PrimInfoMB(empty), [&](Geometry* mesh, const range<size_t>& r, size_t k, size_t geomID, const PrimInfoMB& base) -> PrimInfoMB {
            return mesh->createPrimRefMBArray(prims,t0t1,r,base.size(),(unsigned)geomID);
        }, [](const PrimInfoMB& a, const PrimInfoMB& b) -> PrimInfoMB { return PrimInfoMB::merge2(a,b); });
      }

      /* the BVH starts with that time range, even though primitives might have smaller/larger time range */
      pinfo.time_range = t0t1;
      return pinfo;
    }

    template<typename Mesh>
    size_t createMortonCodeArray(Mesh* mesh, mvector<BVHBuilderMorton::BuildPrim>& morton, BuildProgressMonitor& progressMonitor)
    {
      size_t numPrimitives = morton.size();

      /* compute scene bounds */
      std::pair<size_t,BBox3fa> cb_empty(0,empty);
      auto cb = parallel_reduce 
        ( size_t(0), numPrimitives, size_t(1024), cb_empty, [&](const range<size_t>& r) -> std::pair<size_t,BBox3fa>
          {
            size_t num = 0;
            BBox3fa bounds = empty;
            
            for (size_t j=r.begin(); j<r.end(); j++)
            {
              BBox3fa prim_bounds = empty;
              if (unlikely(!mesh->buildBounds(j,&prim_bounds))) continue;
              bounds.extend(center2(prim_bounds));
              num++;
            }
            return std::make_pair(num,bounds);
          }, [] (const std::pair<size_t,BBox3fa>& a, const std::pair<size_t,BBox3fa>& b) {
          return std::make_pair(a.first + b.first,merge(a.second,b.second)); 
        });
      
      
      size_t numPrimitivesGen = cb.first;
      const BBox3fa centBounds = cb.second;
      
      /* compute morton codes */
      if (likely(numPrimitivesGen == numPrimitives))
      {
        /* fast path if all primitives were valid */
        BVHBuilderMorton::MortonCodeMapping mapping(centBounds);
        parallel_for( size_t(0), numPrimitives, size_t(1024), [&](const range<size_t>& r) -> void {
            BVHBuilderMorton::MortonCodeGenerator generator(mapping,&morton.data()[r.begin()]);
            for (size_t j=r.begin(); j<r.end(); j++)
              generator(mesh->bounds(j),unsigned(j));
          });
      }
      else
      {
        /* slow path, fallback in case some primitives were invalid */
        ParallelPrefixSumState<size_t> pstate;
        BVHBuilderMorton::MortonCodeMapping mapping(centBounds);
        parallel_prefix_sum( pstate, size_t(0), numPrimitives, size_t(1024), size_t(0), [&](const range<size_t>& r, const size_t base) -> size_t {
            size_t num = 0;
            BVHBuilderMorton::MortonCodeGenerator generator(mapping,&morton.data()[r.begin()]);
            for (size_t j=r.begin(); j<r.end(); j++)
            {
              BBox3fa bounds = empty;
              if (unlikely(!mesh->buildBounds(j,&bounds))) continue;
              generator(bounds,unsigned(j));
              num++;
            }
            return num;
          }, std::plus<size_t>());
        
        parallel_prefix_sum( pstate, size_t(0), numPrimitives, size_t(1024), size_t(0), [&](const range<size_t>& r, const size_t base) -> size_t {
            size_t num = 0;
            BVHBuilderMorton::MortonCodeGenerator generator(mapping,&morton.data()[base]);
            for (size_t j=r.begin(); j<r.end(); j++)
            {
              BBox3fa bounds = empty;
              if (!mesh->buildBounds(j,&bounds)) continue;
              generator(bounds,unsigned(j));
              num++;
            }
            return num;
          }, std::plus<size_t>());          
      }
      return numPrimitivesGen;
    }

    // ====================================================================================================
    // ====================================================================================================
    // ====================================================================================================

    // template for grid meshes

#if 0
    template<>
    PrimInfo createPrimRefArray<GridMesh,false>(Scene* scene, mvector<PrimRef>& prims, BuildProgressMonitor& progressMonitor)
    {
      PING;
      ParallelForForPrefixSumState<PrimInfo> pstate;
      Scene::Iterator<GridMesh,false> iter(scene);
      
      /* first try */
      progressMonitor(0);
      pstate.init(iter,size_t(1024));
      PrimInfo pinfo = parallel_for_for_prefix_sum0( pstate, iter, PrimInfo(empty), [&](GridMesh* mesh, const range<size_t>& r, size_t k) -> PrimInfo
      {
        PrimInfo pinfo(empty);
        for (size_t j=r.begin(); j<r.end(); j++)
        {
          BBox3fa bounds = empty;
          if (!mesh->buildBounds(j,&bounds)) continue;
          const PrimRef prim(bounds,mesh->geomID,unsigned(j));
          pinfo.add_center2(prim);
          prims[k++] = prim;
        }
        return pinfo;
      }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      
      /* if we need to filter out geometry, run again */
      if (pinfo.size() != prims.size())
      {
        progressMonitor(0);
        pinfo = parallel_for_for_prefix_sum1( pstate, iter, PrimInfo(empty), [&](GridMesh* mesh, const range<size_t>& r, size_t k, const PrimInfo& base) -> PrimInfo
        {
          k = base.size();
          PrimInfo pinfo(empty);
          for (size_t j=r.begin(); j<r.end(); j++)
          {
            BBox3fa bounds = empty;
            if (!mesh->buildBounds(j,&bounds)) continue;
            const PrimRef prim(bounds,mesh->geomID,unsigned(j));
            pinfo.add_center2(prim);
            prims[k++] = prim;
          }
          return pinfo;
        }, [](const PrimInfo& a, const PrimInfo& b) -> PrimInfo { return PrimInfo::merge(a,b); });
      }
      return pinfo;
    }
#endif

    // ====================================================================================================
    // ====================================================================================================
    // ====================================================================================================

    IF_ENABLED_TRIS (template size_t createMortonCodeArray<TriangleMesh>(TriangleMesh* mesh COMMA mvector<BVHBuilderMorton::BuildPrim>& morton COMMA BuildProgressMonitor& progressMonitor));
    IF_ENABLED_QUADS(template size_t createMortonCodeArray<QuadMesh>(QuadMesh* mesh COMMA mvector<BVHBuilderMorton::BuildPrim>& morton COMMA BuildProgressMonitor& progressMonitor));
    IF_ENABLED_USER (template size_t createMortonCodeArray<UserGeometry>(UserGeometry* mesh COMMA mvector<BVHBuilderMorton::BuildPrim>& morton COMMA BuildProgressMonitor& progressMonitor));
    IF_ENABLED_INSTANCE (template size_t createMortonCodeArray<Instance>(Instance* mesh COMMA mvector<BVHBuilderMorton::BuildPrim>& morton COMMA BuildProgressMonitor& progressMonitor));
  }
}