summaryrefslogtreecommitdiff
path: root/thirdparty/embree-aarch64/kernels/geometry/curveNi_mb.h
blob: 0cd8f833fd429b405a31e594d791f3d769527ef3 (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
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
// Copyright 2009-2020 Intel Corporation
// SPDX-License-Identifier: Apache-2.0

#pragma once

#include "primitive.h"
#include "curve_intersector_precalculations.h"

namespace embree
{
  template<int M>
    struct CurveNiMB
  {
    struct Type : public PrimitiveType {
      const char* name() const;
      size_t sizeActive(const char* This) const;
      size_t sizeTotal(const char* This) const;
      size_t getBytes(const char* This) const;
    };
    static Type type;

  public:

    /* Returns maximum number of stored primitives */
    static __forceinline size_t max_size() { return M; }

    /* Returns required number of primitive blocks for N primitives */
    static __forceinline size_t blocks(size_t N) { return (N+M-1)/M; }

    static __forceinline size_t bytes(size_t N)
    {
      const size_t f = N/M, r = N%M;
      static_assert(sizeof(CurveNiMB) == 6+37*M+24, "internal data layout issue");
      return f*sizeof(CurveNiMB) + (r!=0)*(6+37*r+24);
    }

  public:

    /*! Default constructor. */
    __forceinline CurveNiMB () {}

    /*! fill curve from curve list */
    __forceinline LBBox3fa fillMB(const PrimRefMB* prims, size_t& begin, size_t _end, Scene* scene, const BBox1f time_range)
    {
      size_t end = min(begin+M,_end);
      N = (uint8_t)(end-begin);
      const unsigned int geomID0 = prims[begin].geomID();
      this->geomID(N) = geomID0;
      ty = (uint8_t) scene->get(geomID0)->getType();

      /* encode all primitives */
      LBBox3fa lbounds = empty;
      for (size_t i=0; i<N; i++)
      {
        const PrimRefMB& prim = prims[begin+i];
        const unsigned int geomID = prim.geomID(); assert(geomID == geomID0);
        const unsigned int primID = prim.primID();
        lbounds.extend(scene->get(geomID)->vlinearBounds(primID,time_range));
      }
      BBox3fa bounds = lbounds.bounds();

      /* calculate offset and scale */
      Vec3fa loffset = bounds.lower;
      float lscale = reduce_min(256.0f/(bounds.size()*sqrt(3.0f)));
      if (bounds.size() == Vec3fa(zero)) lscale = 0.0f;
      *this->offset(N) = loffset;
      *this->scale(N) = lscale;
      this->time_offset(N) = time_range.lower;
      this->time_scale(N) = 1.0f/time_range.size();
      
      /* encode all primitives */
      for (size_t i=0; i<M && begin<end; i++, begin++)
      {
        const PrimRefMB& prim = prims[begin];
        const unsigned int geomID = prim.geomID();
        const unsigned int primID = prim.primID();
        const LinearSpace3fa space2 = scene->get(geomID)->computeAlignedSpaceMB(primID,time_range);
        
        const LinearSpace3fa space3(trunc(126.0f*space2.vx),trunc(126.0f*space2.vy),trunc(126.0f*space2.vz));
        const LBBox3fa bounds = scene->get(geomID)->vlinearBounds(loffset,lscale,max(length(space3.vx),length(space3.vy),length(space3.vz)),space3.transposed(),primID,time_range);
        
        // NOTE: this weird (int8_t) (short) cast works around VS2015 Win32 compiler bug
        bounds_vx_x(N)[i] = (int8_t) (short) space3.vx.x;
        bounds_vx_y(N)[i] = (int8_t) (short) space3.vx.y;
        bounds_vx_z(N)[i] = (int8_t) (short) space3.vx.z;
        bounds_vx_lower0(N)[i] = (short) clamp(floor(bounds.bounds0.lower.x),-32767.0f,32767.0f);
        bounds_vx_upper0(N)[i] = (short) clamp(ceil (bounds.bounds0.upper.x),-32767.0f,32767.0f);
        bounds_vx_lower1(N)[i] = (short) clamp(floor(bounds.bounds1.lower.x),-32767.0f,32767.0f);
        bounds_vx_upper1(N)[i] = (short) clamp(ceil (bounds.bounds1.upper.x),-32767.0f,32767.0f);
        assert(-32767.0f <= floor(bounds.bounds0.lower.x) && floor(bounds.bounds0.lower.x) <= 32767.0f);
        assert(-32767.0f <= ceil (bounds.bounds0.upper.x) && ceil (bounds.bounds0.upper.x) <= 32767.0f);
        assert(-32767.0f <= floor(bounds.bounds1.lower.x) && floor(bounds.bounds1.lower.x) <= 32767.0f);
        assert(-32767.0f <= ceil (bounds.bounds1.upper.x) && ceil (bounds.bounds1.upper.x) <= 32767.0f);
        
        bounds_vy_x(N)[i] = (int8_t) (short) space3.vy.x;
        bounds_vy_y(N)[i] = (int8_t) (short) space3.vy.y;
        bounds_vy_z(N)[i] = (int8_t) (short) space3.vy.z;
        bounds_vy_lower0(N)[i] = (short) clamp(floor(bounds.bounds0.lower.y),-32767.0f,32767.0f);
        bounds_vy_upper0(N)[i] = (short) clamp(ceil (bounds.bounds0.upper.y),-32767.0f,32767.0f);
        bounds_vy_lower1(N)[i] = (short) clamp(floor(bounds.bounds1.lower.y),-32767.0f,32767.0f);
        bounds_vy_upper1(N)[i] = (short) clamp(ceil (bounds.bounds1.upper.y),-32767.0f,32767.0f);
        assert(-32767.0f <= floor(bounds.bounds0.lower.y) && floor(bounds.bounds0.lower.y) <= 32767.0f);
        assert(-32767.0f <= ceil (bounds.bounds0.upper.y) && ceil (bounds.bounds0.upper.y) <= 32767.0f);
        assert(-32767.0f <= floor(bounds.bounds1.lower.y) && floor(bounds.bounds1.lower.y) <= 32767.0f);
        assert(-32767.0f <= ceil (bounds.bounds1.upper.y) && ceil (bounds.bounds1.upper.y) <= 32767.0f);

        bounds_vz_x(N)[i] = (int8_t) (short) space3.vz.x;
        bounds_vz_y(N)[i] = (int8_t) (short) space3.vz.y;
        bounds_vz_z(N)[i] = (int8_t) (short) space3.vz.z;
        bounds_vz_lower0(N)[i] = (short) clamp(floor(bounds.bounds0.lower.z),-32767.0f,32767.0f);
        bounds_vz_upper0(N)[i] = (short) clamp(ceil (bounds.bounds0.upper.z),-32767.0f,32767.0f);
        bounds_vz_lower1(N)[i] = (short) clamp(floor(bounds.bounds1.lower.z),-32767.0f,32767.0f);
        bounds_vz_upper1(N)[i] = (short) clamp(ceil (bounds.bounds1.upper.z),-32767.0f,32767.0f);
        assert(-32767.0f <= floor(bounds.bounds0.lower.z) && floor(bounds.bounds0.lower.z) <= 32767.0f);
        assert(-32767.0f <= ceil (bounds.bounds0.upper.z) && ceil (bounds.bounds0.upper.z) <= 32767.0f);
        assert(-32767.0f <= floor(bounds.bounds1.lower.z) && floor(bounds.bounds1.lower.z) <= 32767.0f);
        assert(-32767.0f <= ceil (bounds.bounds1.upper.z) && ceil (bounds.bounds1.upper.z) <= 32767.0f);
               
        this->primID(N)[i] = primID;
      }
      
      return lbounds;
    }

    template<typename BVH, typename SetMB, typename Allocator>
    __forceinline static typename BVH::NodeRecordMB4D createLeafMB(BVH* bvh, const SetMB& prims, const Allocator& alloc)
    {
      size_t start = prims.begin();
      size_t end   = prims.end();
      size_t items = CurveNiMB::blocks(prims.size());
      size_t numbytes = CurveNiMB::bytes(prims.size());
      CurveNiMB* accel = (CurveNiMB*) alloc.malloc1(numbytes,BVH::byteAlignment);
      const typename BVH::NodeRef node = bvh->encodeLeaf((int8_t*)accel,items);
      
      LBBox3fa bounds = empty;
      for (size_t i=0; i<items; i++)
        bounds.extend(accel[i].fillMB(prims.prims->data(),start,end,bvh->scene,prims.time_range));
      
      return typename BVH::NodeRecordMB4D(node,bounds,prims.time_range);
    };

    
  public:
    
    // 27.6 - 46 bytes per primitive
    uint8_t ty;
    uint8_t N;
    uint8_t data[4+37*M+24];

    /*
    struct Layout
    {
      unsigned int geomID;
      unsigned int primID[N];
      
      int8_t bounds_vx_x[N];
      int8_t bounds_vx_y[N];
      int8_t bounds_vx_z[N];
      short bounds_vx_lower0[N];
      short bounds_vx_upper0[N];
      short bounds_vx_lower1[N];
      short bounds_vx_upper1[N];
      
      int8_t bounds_vy_x[N];
      int8_t bounds_vy_y[N];
      int8_t bounds_vy_z[N];
      short bounds_vy_lower0[N];
      short bounds_vy_upper0[N];
      short bounds_vy_lower1[N];
      short bounds_vy_upper1[N];
      
      int8_t bounds_vz_x[N];
      int8_t bounds_vz_y[N];
      int8_t bounds_vz_z[N];
      short bounds_vz_lower0[N];
      short bounds_vz_upper0[N];
      short bounds_vz_lower1[N];
      short bounds_vz_upper1[N];
      
      Vec3f offset;
      float scale;

      float time_offset;
      float time_scale;
    };
    */
    
    __forceinline       unsigned int& geomID(size_t N)       { return *(unsigned int*)((int8_t*)this+2); }
    __forceinline const unsigned int& geomID(size_t N) const { return *(unsigned int*)((int8_t*)this+2); }
    
    __forceinline       unsigned int* primID(size_t N)       { return (unsigned int*)((int8_t*)this+6); }
    __forceinline const unsigned int* primID(size_t N) const { return (unsigned int*)((int8_t*)this+6); }
    
    __forceinline       int8_t* bounds_vx_x(size_t N)       { return (int8_t*)((int8_t*)this+6+4*N); }
    __forceinline const int8_t* bounds_vx_x(size_t N) const { return (int8_t*)((int8_t*)this+6+4*N); }
    
    __forceinline       int8_t* bounds_vx_y(size_t N)       { return (int8_t*)((int8_t*)this+6+5*N); }
    __forceinline const int8_t* bounds_vx_y(size_t N) const { return (int8_t*)((int8_t*)this+6+5*N); }
    
    __forceinline       int8_t* bounds_vx_z(size_t N)       { return (int8_t*)((int8_t*)this+6+6*N); }
    __forceinline const int8_t* bounds_vx_z(size_t N) const { return (int8_t*)((int8_t*)this+6+6*N); }
    
    __forceinline       short* bounds_vx_lower0(size_t N)       { return (short*)((int8_t*)this+6+7*N); }
    __forceinline const short* bounds_vx_lower0(size_t N) const { return (short*)((int8_t*)this+6+7*N); }
    
    __forceinline       short* bounds_vx_upper0(size_t N)       { return (short*)((int8_t*)this+6+9*N); }
    __forceinline const short* bounds_vx_upper0(size_t N) const { return (short*)((int8_t*)this+6+9*N); }

    __forceinline       short* bounds_vx_lower1(size_t N)       { return (short*)((int8_t*)this+6+11*N); }
    __forceinline const short* bounds_vx_lower1(size_t N) const { return (short*)((int8_t*)this+6+11*N); }
    
    __forceinline       short* bounds_vx_upper1(size_t N)       { return (short*)((int8_t*)this+6+13*N); }
    __forceinline const short* bounds_vx_upper1(size_t N) const { return (short*)((int8_t*)this+6+13*N); }

    __forceinline       int8_t* bounds_vy_x(size_t N)       { return (int8_t*)((int8_t*)this+6+15*N); }
    __forceinline const int8_t* bounds_vy_x(size_t N) const { return (int8_t*)((int8_t*)this+6+15*N); }
    
    __forceinline       int8_t* bounds_vy_y(size_t N)       { return (int8_t*)((int8_t*)this+6+16*N); }
    __forceinline const int8_t* bounds_vy_y(size_t N) const { return (int8_t*)((int8_t*)this+6+16*N); }
    
    __forceinline       int8_t* bounds_vy_z(size_t N)       { return (int8_t*)((int8_t*)this+6+17*N); }
    __forceinline const int8_t* bounds_vy_z(size_t N) const { return (int8_t*)((int8_t*)this+6+17*N); }
    
    __forceinline       short* bounds_vy_lower0(size_t N)       { return (short*)((int8_t*)this+6+18*N); }
    __forceinline const short* bounds_vy_lower0(size_t N) const { return (short*)((int8_t*)this+6+18*N); }
    
    __forceinline       short* bounds_vy_upper0(size_t N)       { return (short*)((int8_t*)this+6+20*N); }
    __forceinline const short* bounds_vy_upper0(size_t N) const { return (short*)((int8_t*)this+6+20*N); }

    __forceinline       short* bounds_vy_lower1(size_t N)       { return (short*)((int8_t*)this+6+22*N); }
    __forceinline const short* bounds_vy_lower1(size_t N) const { return (short*)((int8_t*)this+6+22*N); }
    
    __forceinline       short* bounds_vy_upper1(size_t N)       { return (short*)((int8_t*)this+6+24*N); }
    __forceinline const short* bounds_vy_upper1(size_t N) const { return (short*)((int8_t*)this+6+24*N); }
    
    __forceinline       int8_t* bounds_vz_x(size_t N)       { return (int8_t*)((int8_t*)this+6+26*N); }
    __forceinline const int8_t* bounds_vz_x(size_t N) const { return (int8_t*)((int8_t*)this+6+26*N); }
    
    __forceinline       int8_t* bounds_vz_y(size_t N)       { return (int8_t*)((int8_t*)this+6+27*N); }
    __forceinline const int8_t* bounds_vz_y(size_t N) const { return (int8_t*)((int8_t*)this+6+27*N); }
    
    __forceinline       int8_t* bounds_vz_z(size_t N)       { return (int8_t*)((int8_t*)this+6+28*N); }
    __forceinline const int8_t* bounds_vz_z(size_t N) const { return (int8_t*)((int8_t*)this+6+28*N); }
    
    __forceinline       short* bounds_vz_lower0(size_t N)       { return (short*)((int8_t*)this+6+29*N); }
    __forceinline const short* bounds_vz_lower0(size_t N) const { return (short*)((int8_t*)this+6+29*N); }
    
    __forceinline       short* bounds_vz_upper0(size_t N)       { return (short*)((int8_t*)this+6+31*N); }
    __forceinline const short* bounds_vz_upper0(size_t N) const { return (short*)((int8_t*)this+6+31*N); }

    __forceinline       short* bounds_vz_lower1(size_t N)       { return (short*)((int8_t*)this+6+33*N); }
    __forceinline const short* bounds_vz_lower1(size_t N) const { return (short*)((int8_t*)this+6+33*N); }
    
    __forceinline       short* bounds_vz_upper1(size_t N)       { return (short*)((int8_t*)this+6+35*N); }
    __forceinline const short* bounds_vz_upper1(size_t N) const { return (short*)((int8_t*)this+6+35*N); }

    __forceinline       Vec3f* offset(size_t N)       { return (Vec3f*)((int8_t*)this+6+37*N); }
    __forceinline const Vec3f* offset(size_t N) const { return (Vec3f*)((int8_t*)this+6+37*N); }
    
    __forceinline       float* scale(size_t N)       { return (float*)((int8_t*)this+6+37*N+12); }
    __forceinline const float* scale(size_t N) const { return (float*)((int8_t*)this+6+37*N+12); }

    __forceinline       float& time_offset(size_t N)       { return *(float*)((int8_t*)this+6+37*N+16); }
    __forceinline const float& time_offset(size_t N) const { return *(float*)((int8_t*)this+6+37*N+16); }
    
    __forceinline       float& time_scale(size_t N)       { return *(float*)((int8_t*)this+6+37*N+20); }
    __forceinline const float& time_scale(size_t N) const { return *(float*)((int8_t*)this+6+37*N+20); }

    __forceinline       int8_t* end(size_t N)       { return (int8_t*)this+6+37*N+24; }
    __forceinline const int8_t* end(size_t N) const { return (int8_t*)this+6+37*N+24; }
  };

  template<int M>
    typename CurveNiMB<M>::Type CurveNiMB<M>::type;

  typedef CurveNiMB<4> Curve4iMB;
  typedef CurveNiMB<8> Curve8iMB;
}