1// Copyright 2009-2021 Intel Corporation
2// SPDX-License-Identifier: Apache-2.0
3
4#pragma once
5
6#include "primitive.h"
7
8namespace embree
9{
10 template<int M>
11 struct LineMi
12 {
13 /* Virtual interface to query information about the line segment type */
14 struct Type : public PrimitiveType
15 {
16 const char* name() const;
17 size_t sizeActive(const char* This) const;
18 size_t sizeTotal(const char* This) const;
19 size_t getBytes(const char* This) const;
20 };
21 static Type type;
22
23 public:
24
25 /* primitive supports multiple time segments */
26 static const bool singleTimeSegment = false;
27
28 /* Returns maximum number of stored line segments */
29 static __forceinline size_t max_size() { return M; }
30
31 /* Returns required number of primitive blocks for N line segments */
32 static __forceinline size_t blocks(size_t N) { return (N+max_size()-1)/max_size(); }
33
34 /* Returns required number of bytes for N line segments */
35 static __forceinline size_t bytes(size_t N) { return blocks(N)*sizeof(LineMi); }
36
37 public:
38
39 /* Default constructor */
40 __forceinline LineMi() { }
41
42 /* Construction from vertices and IDs */
43 __forceinline LineMi(const vuint<M>& v0, unsigned short leftExists, unsigned short rightExists, const vuint<M>& geomIDs, const vuint<M>& primIDs, Geometry::GType gtype)
44 : gtype((unsigned char)gtype), m((unsigned char)popcnt(vuint<M>(primIDs) != vuint<M>(-1))), sharedGeomID(geomIDs[0]), leftExists (leftExists), rightExists(rightExists), v0(v0), primIDs(primIDs)
45 {
46 assert(all(vuint<M>(geomID()) == geomIDs));
47 }
48
49 /* Returns a mask that tells which line segments are valid */
50 __forceinline vbool<M> valid() const { return primIDs != vuint<M>(-1); }
51
52 /* Returns if the specified line segment is valid */
53 __forceinline bool valid(const size_t i) const { assert(i<M); return primIDs[i] != -1; }
54
55 /* Returns the number of stored line segments */
56 __forceinline size_t size() const { return bsf(~movemask(valid())); }
57
58 /* Returns the geometry IDs */
59 //template<class T>
60 //static __forceinline T unmask(T &index) { return index & 0x3fffffff; }
61
62 __forceinline unsigned int geomID(unsigned int i = 0) const { return sharedGeomID; }
63 //__forceinline vuint<M> geomID() { return unmask(geomIDs); }
64 //__forceinline const vuint<M> geomID() const { return unmask(geomIDs); }
65 //__forceinline unsigned int geomID(const size_t i) const { assert(i<M); return unmask(geomIDs[i]); }
66
67 /* Returns the primitive IDs */
68 __forceinline vuint<M>& primID() { return primIDs; }
69 __forceinline const vuint<M>& primID() const { return primIDs; }
70 __forceinline unsigned int primID(const size_t i) const { assert(i<M); return primIDs[i]; }
71
72 /* gather the line segments */
73 __forceinline void gather(Vec4vf<M>& p0,
74 Vec4vf<M>& p1,
75 const LineSegments* geom) const;
76
77 __forceinline void gatheri(Vec4vf<M>& p0,
78 Vec4vf<M>& p1,
79 const LineSegments* geom,
80 const int itime) const;
81
82 __forceinline void gather(Vec4vf<M>& p0,
83 Vec4vf<M>& p1,
84 const LineSegments* geom,
85 float time) const;
86
87 /* gather the line segments with lateral info */
88 __forceinline void gather(Vec4vf<M>& p0,
89 Vec4vf<M>& p1,
90 Vec4vf<M>& pL,
91 Vec4vf<M>& pR,
92 const LineSegments* geom) const;
93
94 __forceinline void gatheri(Vec4vf<M>& p0,
95 Vec4vf<M>& p1,
96 Vec4vf<M>& pL,
97 Vec4vf<M>& pR,
98 const LineSegments* geom,
99 const int itime) const;
100
101 __forceinline void gather(Vec4vf<M>& p0,
102 Vec4vf<M>& p1,
103 Vec4vf<M>& pL,
104 Vec4vf<M>& pR,
105 const LineSegments* geom,
106 float time) const;
107
108 __forceinline void gather(Vec4vf<M>& p0,
109 Vec4vf<M>& p1,
110 vbool<M>& cL,
111 vbool<M>& cR,
112 const LineSegments* geom) const;
113
114 __forceinline void gatheri(Vec4vf<M>& p0,
115 Vec4vf<M>& p1,
116 vbool<M>& cL,
117 vbool<M>& cR,
118 const LineSegments* geom,
119 const int itime) const;
120
121 __forceinline void gather(Vec4vf<M>& p0,
122 Vec4vf<M>& p1,
123 vbool<M>& cL,
124 vbool<M>& cR,
125 const LineSegments* geom,
126 float time) const;
127
128 /* Calculate the bounds of the line segments */
129 __forceinline const BBox3fa bounds(const Scene* scene, size_t itime = 0) const
130 {
131 BBox3fa bounds = empty;
132 for (size_t i=0; i<M && valid(i); i++)
133 {
134 const LineSegments* geom = scene->get<LineSegments>(geomID(i));
135 const Vec3ff& p0 = geom->vertex(v0[i]+0,itime);
136 const Vec3ff& p1 = geom->vertex(v0[i]+1,itime);
137 BBox3fa b = merge(BBox3fa(p0),BBox3fa(p1));
138 b = enlarge(b,Vec3fa(max(p0.w,p1.w)));
139 bounds.extend(b);
140 }
141 return bounds;
142 }
143
144 /* Calculate the linear bounds of the primitive */
145 __forceinline LBBox3fa linearBounds(const Scene* scene, size_t itime) {
146 return LBBox3fa(bounds(scene,itime+0), bounds(scene,itime+1));
147 }
148
149 __forceinline LBBox3fa linearBounds(const Scene *const scene, size_t itime, size_t numTimeSteps) {
150 LBBox3fa allBounds = empty;
151 for (size_t i=0; i<M && valid(i); i++)
152 {
153 const LineSegments* geom = scene->get<LineSegments>(geomID(i));
154 allBounds.extend(geom->linearBounds(primID(i), itime, numTimeSteps));
155 }
156 return allBounds;
157 }
158
159 __forceinline LBBox3fa linearBounds(const Scene *const scene, const BBox1f time_range)
160 {
161 LBBox3fa allBounds = empty;
162 for (size_t i=0; i<M && valid(i); i++)
163 {
164 const LineSegments* geom = scene->get<LineSegments>(geomID((unsigned int)i));
165 allBounds.extend(geom->linearBounds(primID(i), time_range));
166 }
167 return allBounds;
168 }
169
170 /* Fill line segment from line segment list */
171 template<typename PrimRefT>
172 __forceinline void fill(const PrimRefT* prims, size_t& begin, size_t end, Scene* scene)
173 {
174 Geometry::GType gty = scene->get(prims[begin].geomID())->getType();
175 vuint<M> geomID, primID;
176 vuint<M> v0;
177 unsigned short leftExists = 0;
178 unsigned short rightExists = 0;
179 const PrimRefT* prim = &prims[begin];
180
181 for (size_t i=0; i<M; i++)
182 {
183 const LineSegments* geom = scene->get<LineSegments>(prim->geomID());
184 if (begin<end) {
185 geomID[i] = prim->geomID();
186 primID[i] = prim->primID();
187 v0[i] = geom->segment(prim->primID());
188 leftExists |= geom->segmentLeftExists(primID[i]) << i;
189 rightExists |= geom->segmentRightExists(primID[i]) << i;
190 begin++;
191 } else {
192 assert(i);
193 if (i>0) {
194 geomID[i] = geomID[i-1];
195 primID[i] = -1;
196 v0[i] = v0[i-1];
197 }
198 }
199 if (begin<end) prim = &prims[begin]; // FIXME: remove this line
200 }
201 new (this) LineMi(v0,leftExists,rightExists,geomID,primID,gty); // FIXME: use non temporal store
202 }
203
204 template<typename BVH, typename Allocator>
205 __forceinline static typename BVH::NodeRef createLeaf (BVH* bvh, const PrimRef* prims, const range<size_t>& set, const Allocator& alloc)
206 {
207 size_t start = set.begin();
208 size_t items = LineMi::blocks(set.size());
209 size_t numbytes = LineMi::bytes(set.size());
210 LineMi* accel = (LineMi*) alloc.malloc1(numbytes,M*sizeof(float));
211 for (size_t i=0; i<items; i++) {
212 accel[i].fill(prims,start,set.end(),bvh->scene);
213 }
214 return bvh->encodeLeaf((char*)accel,items);
215 };
216
217 __forceinline LBBox3fa fillMB(const PrimRef* prims, size_t& begin, size_t end, Scene* scene, size_t itime)
218 {
219 fill(prims,begin,end,scene);
220 return linearBounds(scene,itime);
221 }
222
223 __forceinline LBBox3fa fillMB(const PrimRefMB* prims, size_t& begin, size_t end, Scene* scene, const BBox1f time_range)
224 {
225 fill(prims,begin,end,scene);
226 return linearBounds(scene,time_range);
227 }
228
229 template<typename BVH, typename SetMB, typename Allocator>
230 __forceinline static typename BVH::NodeRecordMB4D createLeafMB(BVH* bvh, const SetMB& prims, const Allocator& alloc)
231 {
232 size_t start = prims.begin();
233 size_t end = prims.end();
234 size_t items = LineMi::blocks(prims.size());
235 size_t numbytes = LineMi::bytes(prims.size());
236 LineMi* accel = (LineMi*) alloc.malloc1(numbytes,M*sizeof(float));
237 const typename BVH::NodeRef node = bvh->encodeLeaf((char*)accel,items);
238
239 LBBox3fa bounds = empty;
240 for (size_t i=0; i<items; i++)
241 bounds.extend(accel[i].fillMB(prims.prims->data(),start,end,bvh->scene,prims.time_range));
242
243 return typename BVH::NodeRecordMB4D(node,bounds,prims.time_range);
244 };
245
246 /* Updates the primitive */
247 __forceinline BBox3fa update(LineSegments* geom)
248 {
249 BBox3fa bounds = empty;
250 for (size_t i=0; i<M && valid(i); i++)
251 {
252 const Vec3ff& p0 = geom->vertex(v0[i]+0);
253 const Vec3ff& p1 = geom->vertex(v0[i]+1);
254 BBox3fa b = merge(BBox3fa(p0),BBox3fa(p1));
255 b = enlarge(b,Vec3fa(max(p0.w,p1.w)));
256 bounds.extend(b);
257 }
258 return bounds;
259 }
260
261 /*! output operator */
262 friend __forceinline embree_ostream operator<<(embree_ostream cout, const LineMi& line) {
263 return cout << "Line" << M << "i {" << line.v0 << ", " << line.geomID() << ", " << line.primID() << "}";
264 }
265
266 public:
267 unsigned char gtype;
268 unsigned char m;
269 unsigned int sharedGeomID;
270 unsigned short leftExists, rightExists;
271 vuint<M> v0; // index of start vertex
272 private:
273 vuint<M> primIDs; // primitive ID
274 };
275
276 template<>
277 __forceinline void LineMi<4>::gather(Vec4vf4& p0,
278 Vec4vf4& p1,
279 const LineSegments* geom) const
280 {
281 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0]));
282 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1]));
283 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2]));
284 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3]));
285 transpose(a0,a1,a2,a3,p0.x,p0.y,p0.z,p0.w);
286
287 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1));
288 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1));
289 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1));
290 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1));
291 transpose(b0,b1,b2,b3,p1.x,p1.y,p1.z,p1.w);
292 }
293
294 template<>
295 __forceinline void LineMi<4>::gatheri(Vec4vf4& p0,
296 Vec4vf4& p1,
297 const LineSegments* geom,
298 const int itime) const
299 {
300 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0],itime));
301 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1],itime));
302 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2],itime));
303 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3],itime));
304 transpose(a0,a1,a2,a3,p0.x,p0.y,p0.z,p0.w);
305
306 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1,itime));
307 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1,itime));
308 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1,itime));
309 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1,itime));
310 transpose(b0,b1,b2,b3,p1.x,p1.y,p1.z,p1.w);
311 }
312
313 template<>
314 __forceinline void LineMi<4>::gather(Vec4vf4& p0,
315 Vec4vf4& p1,
316 const LineSegments* geom,
317 float time) const
318 {
319 float ftime;
320 const int itime = geom->timeSegment(time, ftime);
321
322 Vec4vf4 a0,a1;
323 gatheri(a0,a1,geom,itime);
324 Vec4vf4 b0,b1;
325 gatheri(b0,b1,geom,itime+1);
326 p0 = lerp(a0,b0,vfloat4(ftime));
327 p1 = lerp(a1,b1,vfloat4(ftime));
328 }
329
330 template<>
331 __forceinline void LineMi<4>::gather(Vec4vf4& p0,
332 Vec4vf4& p1,
333 vbool4& cL,
334 vbool4& cR,
335 const LineSegments* geom) const
336 {
337 gather(p0,p1,geom);
338 cL = !vbool4(leftExists);
339 cR = !vbool4(rightExists);
340 }
341
342 template<>
343 __forceinline void LineMi<4>::gatheri(Vec4vf4& p0,
344 Vec4vf4& p1,
345 vbool4& cL,
346 vbool4& cR,
347 const LineSegments* geom,
348 const int itime) const
349 {
350 gatheri(p0,p1,geom,itime);
351 cL = !vbool4(leftExists);
352 cR = !vbool4(rightExists);
353 }
354
355 template<>
356 __forceinline void LineMi<4>::gather(Vec4vf4& p0,
357 Vec4vf4& p1,
358 vbool4& cL,
359 vbool4& cR,
360 const LineSegments* geom,
361 float time) const
362 {
363 float ftime;
364 const int itime = geom->timeSegment(time, ftime);
365
366 Vec4vf4 a0,a1;
367 gatheri(a0,a1,geom,itime);
368 Vec4vf4 b0,b1;
369 gatheri(b0,b1,geom,itime+1);
370 p0 = lerp(a0,b0,vfloat4(ftime));
371 p1 = lerp(a1,b1,vfloat4(ftime));
372 cL = !vbool4(leftExists);
373 cR = !vbool4(rightExists);
374 }
375
376 template<>
377 __forceinline void LineMi<4>::gather(Vec4vf4& p0,
378 Vec4vf4& p1,
379 Vec4vf4& pL,
380 Vec4vf4& pR,
381 const LineSegments* geom) const
382 {
383 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0]));
384 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1]));
385 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2]));
386 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3]));
387 transpose(a0,a1,a2,a3,p0.x,p0.y,p0.z,p0.w);
388
389 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1));
390 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1));
391 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1));
392 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1));
393 transpose(b0,b1,b2,b3,p1.x,p1.y,p1.z,p1.w);
394
395 const vfloat4 l0 = (leftExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]-1)) : vfloat4(inf);
396 const vfloat4 l1 = (leftExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]-1)) : vfloat4(inf);
397 const vfloat4 l2 = (leftExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]-1)) : vfloat4(inf);
398 const vfloat4 l3 = (leftExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]-1)) : vfloat4(inf);
399 transpose(l0,l1,l2,l3,pL.x,pL.y,pL.z,pL.w);
400
401 const vfloat4 r0 = (rightExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]+2)) : vfloat4(inf);
402 const vfloat4 r1 = (rightExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]+2)) : vfloat4(inf);
403 const vfloat4 r2 = (rightExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]+2)) : vfloat4(inf);
404 const vfloat4 r3 = (rightExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]+2)) : vfloat4(inf);
405 transpose(r0,r1,r2,r3,pR.x,pR.y,pR.z,pR.w);
406 }
407
408 template<>
409 __forceinline void LineMi<4>::gatheri(Vec4vf4& p0,
410 Vec4vf4& p1,
411 Vec4vf4& pL,
412 Vec4vf4& pR,
413 const LineSegments* geom,
414 const int itime) const
415 {
416 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0],itime));
417 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1],itime));
418 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2],itime));
419 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3],itime));
420 transpose(a0,a1,a2,a3,p0.x,p0.y,p0.z,p0.w);
421
422 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1,itime));
423 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1,itime));
424 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1,itime));
425 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1,itime));
426 transpose(b0,b1,b2,b3,p1.x,p1.y,p1.z,p1.w);
427
428 const vfloat4 l0 = (leftExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]-1,itime)) : vfloat4(inf);
429 const vfloat4 l1 = (leftExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]-1,itime)) : vfloat4(inf);
430 const vfloat4 l2 = (leftExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]-1,itime)) : vfloat4(inf);
431 const vfloat4 l3 = (leftExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]-1,itime)) : vfloat4(inf);
432 transpose(l0,l1,l2,l3,pL.x,pL.y,pL.z,pL.w);
433
434 const vfloat4 r0 = (rightExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]+2,itime)) : vfloat4(inf);
435 const vfloat4 r1 = (rightExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]+2,itime)) : vfloat4(inf);
436 const vfloat4 r2 = (rightExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]+2,itime)) : vfloat4(inf);
437 const vfloat4 r3 = (rightExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]+2,itime)) : vfloat4(inf);
438 transpose(r0,r1,r2,r3,pR.x,pR.y,pR.z,pR.w);
439 }
440
441 template<>
442 __forceinline void LineMi<4>::gather(Vec4vf4& p0,
443 Vec4vf4& p1,
444 Vec4vf4& pL,
445 Vec4vf4& pR,
446 const LineSegments* geom,
447 float time) const
448 {
449 float ftime;
450 const int itime = geom->timeSegment(time, ftime);
451
452 Vec4vf4 a0,a1,aL,aR;
453 gatheri(a0,a1,aL,aR,geom,itime);
454 Vec4vf4 b0,b1,bL,bR;
455 gatheri(b0,b1,bL,bR,geom,itime+1);
456 p0 = lerp(a0,b0,vfloat4(ftime));
457 p1 = lerp(a1,b1,vfloat4(ftime));
458 pL = lerp(aL,bL,vfloat4(ftime));
459 pR = lerp(aR,bR,vfloat4(ftime));
460 }
461
462#if defined(__AVX__)
463
464 template<>
465 __forceinline void LineMi<8>::gather(Vec4vf8& p0,
466 Vec4vf8& p1,
467 const LineSegments* geom) const
468 {
469 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0]));
470 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1]));
471 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2]));
472 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3]));
473 const vfloat4 a4 = vfloat4::loadu(geom->vertexPtr(v0[4]));
474 const vfloat4 a5 = vfloat4::loadu(geom->vertexPtr(v0[5]));
475 const vfloat4 a6 = vfloat4::loadu(geom->vertexPtr(v0[6]));
476 const vfloat4 a7 = vfloat4::loadu(geom->vertexPtr(v0[7]));
477 transpose(a0,a1,a2,a3,a4,a5,a6,a7,p0.x,p0.y,p0.z,p0.w);
478
479 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1));
480 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1));
481 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1));
482 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1));
483 const vfloat4 b4 = vfloat4::loadu(geom->vertexPtr(v0[4]+1));
484 const vfloat4 b5 = vfloat4::loadu(geom->vertexPtr(v0[5]+1));
485 const vfloat4 b6 = vfloat4::loadu(geom->vertexPtr(v0[6]+1));
486 const vfloat4 b7 = vfloat4::loadu(geom->vertexPtr(v0[7]+1));
487 transpose(b0,b1,b2,b3,b4,b5,b6,b7,p1.x,p1.y,p1.z,p1.w);
488 }
489
490 template<>
491 __forceinline void LineMi<8>::gatheri(Vec4vf8& p0,
492 Vec4vf8& p1,
493 const LineSegments* geom,
494 const int itime) const
495 {
496 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0],itime));
497 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1],itime));
498 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2],itime));
499 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3],itime));
500 const vfloat4 a4 = vfloat4::loadu(geom->vertexPtr(v0[4],itime));
501 const vfloat4 a5 = vfloat4::loadu(geom->vertexPtr(v0[5],itime));
502 const vfloat4 a6 = vfloat4::loadu(geom->vertexPtr(v0[6],itime));
503 const vfloat4 a7 = vfloat4::loadu(geom->vertexPtr(v0[7],itime));
504 transpose(a0,a1,a2,a3,a4,a5,a6,a7,p0.x,p0.y,p0.z,p0.w);
505
506 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1,itime));
507 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1,itime));
508 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1,itime));
509 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1,itime));
510 const vfloat4 b4 = vfloat4::loadu(geom->vertexPtr(v0[4]+1,itime));
511 const vfloat4 b5 = vfloat4::loadu(geom->vertexPtr(v0[5]+1,itime));
512 const vfloat4 b6 = vfloat4::loadu(geom->vertexPtr(v0[6]+1,itime));
513 const vfloat4 b7 = vfloat4::loadu(geom->vertexPtr(v0[7]+1,itime));
514 transpose(b0,b1,b2,b3,b4,b5,b6,b7,p1.x,p1.y,p1.z,p1.w);
515 }
516
517 template<>
518 __forceinline void LineMi<8>::gather(Vec4vf8& p0,
519 Vec4vf8& p1,
520 const LineSegments* geom,
521 float time) const
522 {
523 float ftime;
524 const int itime = geom->timeSegment(time, ftime);
525
526 Vec4vf8 a0,a1;
527 gatheri(a0,a1,geom,itime);
528 Vec4vf8 b0,b1;
529 gatheri(b0,b1,geom,itime+1);
530 p0 = lerp(a0,b0,vfloat8(ftime));
531 p1 = lerp(a1,b1,vfloat8(ftime));
532 }
533
534 template<>
535 __forceinline void LineMi<8>::gather(Vec4vf8& p0,
536 Vec4vf8& p1,
537 Vec4vf8& pL,
538 Vec4vf8& pR,
539 const LineSegments* geom) const
540 {
541 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0]));
542 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1]));
543 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2]));
544 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3]));
545 const vfloat4 a4 = vfloat4::loadu(geom->vertexPtr(v0[4]));
546 const vfloat4 a5 = vfloat4::loadu(geom->vertexPtr(v0[5]));
547 const vfloat4 a6 = vfloat4::loadu(geom->vertexPtr(v0[6]));
548 const vfloat4 a7 = vfloat4::loadu(geom->vertexPtr(v0[7]));
549 transpose(a0,a1,a2,a3,a4,a5,a6,a7,p0.x,p0.y,p0.z,p0.w);
550
551 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1));
552 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1));
553 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1));
554 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1));
555 const vfloat4 b4 = vfloat4::loadu(geom->vertexPtr(v0[4]+1));
556 const vfloat4 b5 = vfloat4::loadu(geom->vertexPtr(v0[5]+1));
557 const vfloat4 b6 = vfloat4::loadu(geom->vertexPtr(v0[6]+1));
558 const vfloat4 b7 = vfloat4::loadu(geom->vertexPtr(v0[7]+1));
559 transpose(b0,b1,b2,b3,b4,b5,b6,b7,p1.x,p1.y,p1.z,p1.w);
560
561 const vfloat4 l0 = (leftExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]-1)) : vfloat4(inf);
562 const vfloat4 l1 = (leftExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]-1)) : vfloat4(inf);
563 const vfloat4 l2 = (leftExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]-1)) : vfloat4(inf);
564 const vfloat4 l3 = (leftExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]-1)) : vfloat4(inf);
565 const vfloat4 l4 = (leftExists & (1<<4)) ? vfloat4::loadu(geom->vertexPtr(v0[4]-1)) : vfloat4(inf);
566 const vfloat4 l5 = (leftExists & (1<<5)) ? vfloat4::loadu(geom->vertexPtr(v0[5]-1)) : vfloat4(inf);
567 const vfloat4 l6 = (leftExists & (1<<6)) ? vfloat4::loadu(geom->vertexPtr(v0[6]-1)) : vfloat4(inf);
568 const vfloat4 l7 = (leftExists & (1<<7)) ? vfloat4::loadu(geom->vertexPtr(v0[7]-1)) : vfloat4(inf);
569 transpose(l0,l1,l2,l3,l4,l5,l6,l7,pL.x,pL.y,pL.z,pL.w);
570
571 const vfloat4 r0 = (rightExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]+2)) : vfloat4(inf);
572 const vfloat4 r1 = (rightExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]+2)) : vfloat4(inf);
573 const vfloat4 r2 = (rightExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]+2)) : vfloat4(inf);
574 const vfloat4 r3 = (rightExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]+2)) : vfloat4(inf);
575 const vfloat4 r4 = (rightExists & (1<<4)) ? vfloat4::loadu(geom->vertexPtr(v0[4]+2)) : vfloat4(inf);
576 const vfloat4 r5 = (rightExists & (1<<5)) ? vfloat4::loadu(geom->vertexPtr(v0[5]+2)) : vfloat4(inf);
577 const vfloat4 r6 = (rightExists & (1<<6)) ? vfloat4::loadu(geom->vertexPtr(v0[6]+2)) : vfloat4(inf);
578 const vfloat4 r7 = (rightExists & (1<<7)) ? vfloat4::loadu(geom->vertexPtr(v0[7]+2)) : vfloat4(inf);
579 transpose(r0,r1,r2,r3,r4,r5,r6,r7,pR.x,pR.y,pR.z,pR.w);
580 }
581
582 template<>
583 __forceinline void LineMi<8>::gatheri(Vec4vf8& p0,
584 Vec4vf8& p1,
585 Vec4vf8& pL,
586 Vec4vf8& pR,
587 const LineSegments* geom,
588 const int itime) const
589 {
590 const vfloat4 a0 = vfloat4::loadu(geom->vertexPtr(v0[0],itime));
591 const vfloat4 a1 = vfloat4::loadu(geom->vertexPtr(v0[1],itime));
592 const vfloat4 a2 = vfloat4::loadu(geom->vertexPtr(v0[2],itime));
593 const vfloat4 a3 = vfloat4::loadu(geom->vertexPtr(v0[3],itime));
594 const vfloat4 a4 = vfloat4::loadu(geom->vertexPtr(v0[4],itime));
595 const vfloat4 a5 = vfloat4::loadu(geom->vertexPtr(v0[5],itime));
596 const vfloat4 a6 = vfloat4::loadu(geom->vertexPtr(v0[6],itime));
597 const vfloat4 a7 = vfloat4::loadu(geom->vertexPtr(v0[7],itime));
598 transpose(a0,a1,a2,a3,a4,a5,a6,a7,p0.x,p0.y,p0.z,p0.w);
599
600 const vfloat4 b0 = vfloat4::loadu(geom->vertexPtr(v0[0]+1,itime));
601 const vfloat4 b1 = vfloat4::loadu(geom->vertexPtr(v0[1]+1,itime));
602 const vfloat4 b2 = vfloat4::loadu(geom->vertexPtr(v0[2]+1,itime));
603 const vfloat4 b3 = vfloat4::loadu(geom->vertexPtr(v0[3]+1,itime));
604 const vfloat4 b4 = vfloat4::loadu(geom->vertexPtr(v0[4]+1,itime));
605 const vfloat4 b5 = vfloat4::loadu(geom->vertexPtr(v0[5]+1,itime));
606 const vfloat4 b6 = vfloat4::loadu(geom->vertexPtr(v0[6]+1,itime));
607 const vfloat4 b7 = vfloat4::loadu(geom->vertexPtr(v0[7]+1,itime));
608 transpose(b0,b1,b2,b3,b4,b5,b6,b7,p1.x,p1.y,p1.z,p1.w);
609
610 const vfloat4 l0 = (leftExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]-1,itime)) : vfloat4(inf);
611 const vfloat4 l1 = (leftExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]-1,itime)) : vfloat4(inf);
612 const vfloat4 l2 = (leftExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]-1,itime)) : vfloat4(inf);
613 const vfloat4 l3 = (leftExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]-1,itime)) : vfloat4(inf);
614 const vfloat4 l4 = (leftExists & (1<<4)) ? vfloat4::loadu(geom->vertexPtr(v0[4]-1,itime)) : vfloat4(inf);
615 const vfloat4 l5 = (leftExists & (1<<5)) ? vfloat4::loadu(geom->vertexPtr(v0[5]-1,itime)) : vfloat4(inf);
616 const vfloat4 l6 = (leftExists & (1<<6)) ? vfloat4::loadu(geom->vertexPtr(v0[6]-1,itime)) : vfloat4(inf);
617 const vfloat4 l7 = (leftExists & (1<<7)) ? vfloat4::loadu(geom->vertexPtr(v0[7]-1,itime)) : vfloat4(inf);
618 transpose(l0,l1,l2,l3,l4,l5,l6,l7,pL.x,pL.y,pL.z,pL.w);
619
620 const vfloat4 r0 = (rightExists & (1<<0)) ? vfloat4::loadu(geom->vertexPtr(v0[0]+2,itime)) : vfloat4(inf);
621 const vfloat4 r1 = (rightExists & (1<<1)) ? vfloat4::loadu(geom->vertexPtr(v0[1]+2,itime)) : vfloat4(inf);
622 const vfloat4 r2 = (rightExists & (1<<2)) ? vfloat4::loadu(geom->vertexPtr(v0[2]+2,itime)) : vfloat4(inf);
623 const vfloat4 r3 = (rightExists & (1<<3)) ? vfloat4::loadu(geom->vertexPtr(v0[3]+2,itime)) : vfloat4(inf);
624 const vfloat4 r4 = (rightExists & (1<<4)) ? vfloat4::loadu(geom->vertexPtr(v0[4]+2,itime)) : vfloat4(inf);
625 const vfloat4 r5 = (rightExists & (1<<5)) ? vfloat4::loadu(geom->vertexPtr(v0[5]+2,itime)) : vfloat4(inf);
626 const vfloat4 r6 = (rightExists & (1<<6)) ? vfloat4::loadu(geom->vertexPtr(v0[6]+2,itime)) : vfloat4(inf);
627 const vfloat4 r7 = (rightExists & (1<<7)) ? vfloat4::loadu(geom->vertexPtr(v0[7]+2,itime)) : vfloat4(inf);
628 transpose(r0,r1,r2,r3,r4,r5,r6,r7,pR.x,pR.y,pR.z,pR.w);
629 }
630
631 template<>
632 __forceinline void LineMi<8>::gather(Vec4vf8& p0,
633 Vec4vf8& p1,
634 Vec4vf8& pL,
635 Vec4vf8& pR,
636 const LineSegments* geom,
637 float time) const
638 {
639 float ftime;
640 const int itime = geom->timeSegment(time, ftime);
641
642 Vec4vf8 a0,a1,aL,aR;
643 gatheri(a0,a1,aL,aR,geom,itime);
644 Vec4vf8 b0,b1,bL,bR;
645 gatheri(b0,b1,bL,bR,geom,itime+1);
646 p0 = lerp(a0,b0,vfloat8(ftime));
647 p1 = lerp(a1,b1,vfloat8(ftime));
648 pL = lerp(aL,bL,vfloat8(ftime));
649 pR = lerp(aR,bR,vfloat8(ftime));
650 }
651
652 template<>
653 __forceinline void LineMi<8>::gather(Vec4vf8& p0,
654 Vec4vf8& p1,
655 vbool8& cL,
656 vbool8& cR,
657 const LineSegments* geom) const
658 {
659 gather(p0,p1,geom);
660 cL = !vbool8(leftExists);
661 cR = !vbool8(rightExists);
662 }
663
664 template<>
665 __forceinline void LineMi<8>::gatheri(Vec4vf8& p0,
666 Vec4vf8& p1,
667 vbool8& cL,
668 vbool8& cR,
669 const LineSegments* geom,
670 const int itime) const
671 {
672 gatheri(p0,p1,geom,itime);
673 cL = !vbool8(leftExists);
674 cR = !vbool8(rightExists);
675 }
676
677 template<>
678 __forceinline void LineMi<8>::gather(Vec4vf8& p0,
679 Vec4vf8& p1,
680 vbool8& cL,
681 vbool8& cR,
682 const LineSegments* geom,
683 float time) const
684 {
685 float ftime;
686 const int itime = geom->timeSegment(time, ftime);
687
688 Vec4vf8 a0,a1;
689 gatheri(a0,a1,geom,itime);
690 Vec4vf8 b0,b1;
691 gatheri(b0,b1,geom,itime+1);
692 p0 = lerp(a0,b0,vfloat8(ftime));
693 p1 = lerp(a1,b1,vfloat8(ftime));
694 cL = !vbool8(leftExists);
695 cR = !vbool8(rightExists);
696 }
697
698#endif
699
700 template<int M>
701 typename LineMi<M>::Type LineMi<M>::type;
702
703 typedef LineMi<4> Line4i;
704 typedef LineMi<8> Line8i;
705}
706