| 1 | // Copyright 2009-2021 Intel Corporation | 
| 2 | // SPDX-License-Identifier: Apache-2.0 | 
| 3 |  | 
| 4 | #pragma once | 
| 5 |  | 
| 6 | #include "subgrid.h" | 
| 7 | #include "quad_intersector_moeller.h" | 
| 8 | #include "quad_intersector_pluecker.h" | 
| 9 |  | 
| 10 | namespace embree | 
| 11 | { | 
| 12 |   namespace isa | 
| 13 |   { | 
| 14 |  | 
| 15 |     template<int M> | 
| 16 |     __forceinline void interpolateUV(PlueckerHitM<M,UVIdentity<M>> &hit,const GridMesh::Grid &g, const SubGrid& subgrid, const vint<M> &stepX, const vint<M> &stepY)  | 
| 17 |     { | 
| 18 |       /* correct U,V interpolation across the entire grid */ | 
| 19 |       const vint<M> sx((int)subgrid.x()); | 
| 20 |       const vint<M> sy((int)subgrid.y()); | 
| 21 |       const vint<M> sxM(sx + stepX); | 
| 22 |       const vint<M> syM(sy + stepY); | 
| 23 |       const float inv_resX = rcp(x: (float)((int)g.resX-1)); | 
| 24 |       const float inv_resY = rcp(x: (float)((int)g.resY-1));           | 
| 25 |       hit.U = (hit.U + vfloat<M>(sxM) * hit.UVW) * inv_resX; | 
| 26 |       hit.V = (hit.V + vfloat<M>(syM) * hit.UVW) * inv_resY; | 
| 27 |     } | 
| 28 |      | 
| 29 |     template<int M, bool filter> | 
| 30 |       struct SubGridQuadMIntersector1Pluecker; | 
| 31 |  | 
| 32 |     template<int M, bool filter> | 
| 33 |       struct SubGridQuadMIntersector1Pluecker | 
| 34 |       { | 
| 35 |         __forceinline SubGridQuadMIntersector1Pluecker() {} | 
| 36 |  | 
| 37 |         __forceinline SubGridQuadMIntersector1Pluecker(const Ray& ray, const void* ptr) {} | 
| 38 |  | 
| 39 |         __forceinline void intersect(RayHit& ray, IntersectContext* context, | 
| 40 |                                      const Vec3vf<M>& v0, const Vec3vf<M>& v1, const Vec3vf<M>& v2, const Vec3vf<M>& v3, | 
| 41 |                                      const GridMesh::Grid &g, const SubGrid& subgrid) const | 
| 42 |         { | 
| 43 |           UVIdentity<M> mapUV; | 
| 44 |           PlueckerHitM<M,UVIdentity<M>> hit(mapUV); | 
| 45 |           PlueckerIntersector1<M> intersector(ray,nullptr); | 
| 46 | 	   | 
| 47 |           Intersect1EpilogMU<M,filter> epilog(ray,context,subgrid.geomID(),subgrid.primID()); | 
| 48 |  | 
| 49 |           /* intersect first triangle */ | 
| 50 | 	  if (intersector.intersect(ray,v0,v1,v3,mapUV,hit))  | 
| 51 |           { | 
| 52 |             interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 53 |             epilog(hit.valid,hit); | 
| 54 |           } | 
| 55 |  | 
| 56 |           /* intersect second triangle */ | 
| 57 | 	  if (intersector.intersect(ray,v2,v3,v1,mapUV,hit))  | 
| 58 |           { | 
| 59 | 	    hit.U = hit.UVW - hit.U; | 
| 60 | 	    hit.V = hit.UVW - hit.V; | 
| 61 |             interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 62 |             epilog(hit.valid,hit); | 
| 63 |           } | 
| 64 |         } | 
| 65 |        | 
| 66 |         __forceinline bool occluded(Ray& ray, IntersectContext* context, | 
| 67 |                                     const Vec3vf<M>& v0, const Vec3vf<M>& v1, const Vec3vf<M>& v2, const Vec3vf<M>& v3, | 
| 68 |                                     const GridMesh::Grid &g, const SubGrid& subgrid) const | 
| 69 |         { | 
| 70 |           UVIdentity<M> mapUV; | 
| 71 |           PlueckerHitM<M,UVIdentity<M>> hit(mapUV); | 
| 72 |           PlueckerIntersector1<M> intersector(ray,nullptr); | 
| 73 |           Occluded1EpilogMU<M,filter> epilog(ray,context,subgrid.geomID(),subgrid.primID()); | 
| 74 |  | 
| 75 |           /* intersect first triangle */ | 
| 76 | 	  if (intersector.intersect(ray,v0,v1,v3,mapUV,hit))  | 
| 77 |           { | 
| 78 |             interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 79 |             if (epilog(hit.valid,hit)) | 
| 80 | 	      return true; | 
| 81 |           } | 
| 82 |  | 
| 83 |           /* intersect second triangle */ | 
| 84 | 	  if (intersector.intersect(ray,v2,v3,v1,mapUV,hit))  | 
| 85 |           { | 
| 86 | 	    hit.U = hit.UVW - hit.U; | 
| 87 | 	    hit.V = hit.UVW - hit.V; | 
| 88 |             interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 89 |             if (epilog(hit.valid,hit)) | 
| 90 | 	      return true; | 
| 91 |           } | 
| 92 |           return false; | 
| 93 |         } | 
| 94 |       }; | 
| 95 |  | 
| 96 | #if defined (__AVX__) | 
| 97 |  | 
| 98 |     /*! Intersects 4 quads with 1 ray using AVX */ | 
| 99 |     template<bool filter> | 
| 100 |       struct SubGridQuadMIntersector1Pluecker<4,filter> | 
| 101 |     { | 
| 102 |       __forceinline SubGridQuadMIntersector1Pluecker() {} | 
| 103 |  | 
| 104 |       __forceinline SubGridQuadMIntersector1Pluecker(const Ray& ray, const void* ptr) {} | 
| 105 |        | 
| 106 |       template<typename Epilog> | 
| 107 |         __forceinline bool intersect(Ray& ray, const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, const GridMesh::Grid &g, const SubGrid& subgrid, const Epilog& epilog) const | 
| 108 |       { | 
| 109 |         const Vec3vf8 vtx0(vfloat8(v0.x,v2.x),vfloat8(v0.y,v2.y),vfloat8(v0.z,v2.z)); | 
| 110 | #if !defined(EMBREE_BACKFACE_CULLING) | 
| 111 |         const Vec3vf8 vtx1(vfloat8(v1.x),vfloat8(v1.y),vfloat8(v1.z)); | 
| 112 |         const Vec3vf8 vtx2(vfloat8(v3.x),vfloat8(v3.y),vfloat8(v3.z));         | 
| 113 | #else | 
| 114 |         const Vec3vf8 vtx1(vfloat8(v1.x,v3.x),vfloat8(v1.y,v3.y),vfloat8(v1.z,v3.z)); | 
| 115 |         const Vec3vf8 vtx2(vfloat8(v3.x,v1.x),vfloat8(v3.y,v1.y),vfloat8(v3.z,v1.z)); | 
| 116 | #endif | 
| 117 |  | 
| 118 |         UVIdentity<8> mapUV; | 
| 119 |         PlueckerHitM<8,UVIdentity<8>> hit(mapUV); | 
| 120 |         PlueckerIntersector1<8> intersector(ray,nullptr); | 
| 121 |         const vbool8 flags(0,0,0,0,1,1,1,1); | 
| 122 |         if (unlikely(intersector.intersect(ray,vtx0,vtx1,vtx2,mapUV,hit))) | 
| 123 |         { | 
| 124 | 	  /* correct U,V interpolation across the entire grid */ | 
| 125 | 	  const vfloat8 U = select(flags,hit.UVW - hit.V,hit.U);	   | 
| 126 | 	  const vfloat8 V = select(flags,hit.UVW - hit.U,hit.V); | 
| 127 | 	  hit.U = U; | 
| 128 | 	  hit.V = V;	   | 
| 129 | 	  hit.vNg *= select(flags,vfloat8(-1.0f),vfloat8(1.0f));  | 
| 130 |           interpolateUV<8>(hit,g,subgrid,vint<8>(0,1,1,0,0,1,1,0),vint<8>(0,0,1,1,0,0,1,1)); | 
| 131 |           if (unlikely(epilog(hit.valid,hit))) | 
| 132 |             return true; | 
| 133 |         } | 
| 134 |         return false; | 
| 135 |       } | 
| 136 |        | 
| 137 |       __forceinline bool intersect(RayHit& ray, IntersectContext* context, | 
| 138 |                                    const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3,  | 
| 139 |                                    const GridMesh::Grid &g, const SubGrid& subgrid) const | 
| 140 |       { | 
| 141 |           return intersect(ray,v0,v1,v2,v3,g,subgrid,Intersect1EpilogMU<8,filter>(ray,context,subgrid.geomID(),subgrid.primID())); | 
| 142 |       } | 
| 143 |        | 
| 144 |       __forceinline bool occluded(Ray& ray, IntersectContext* context, | 
| 145 |                                   const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3,  | 
| 146 |                                   const GridMesh::Grid &g, const SubGrid& subgrid) const | 
| 147 |       { | 
| 148 |           return intersect(ray,v0,v1,v2,v3,g,subgrid,Occluded1EpilogMU<8,filter>(ray,context,subgrid.geomID(),subgrid.primID())); | 
| 149 |       } | 
| 150 |     }; | 
| 151 |  | 
| 152 | #endif | 
| 153 |  | 
| 154 |  | 
| 155 |     /* ----------------------------- */ | 
| 156 |     /* -- ray packet intersectors -- */ | 
| 157 |     /* ----------------------------- */ | 
| 158 |  | 
| 159 |     template<int K> | 
| 160 |     __forceinline void interpolateUV(const vbool<K>& valid, PlueckerHitK<K,UVIdentity<K>> &hit,const GridMesh::Grid &g, const SubGrid& subgrid, const unsigned int i)  | 
| 161 |     { | 
| 162 |       /* correct U,V interpolation across the entire grid */ | 
| 163 |       const unsigned int sx = subgrid.x() + (unsigned int)(i % 2); | 
| 164 |       const unsigned int sy = subgrid.y() + (unsigned int)(i >>1); | 
| 165 |       const float inv_resX = rcp(x: (float)(int)(g.resX-1)); | 
| 166 |       const float inv_resY = rcp(x: (float)(int)(g.resY-1));       | 
| 167 |       hit.U = select(valid,(hit.U + vfloat<K>((float)sx) * hit.UVW) * inv_resX,hit.U); | 
| 168 |       hit.V = select(valid,(hit.V + vfloat<K>((float)sy) * hit.UVW) * inv_resY,hit.V); | 
| 169 |     } | 
| 170 |      | 
| 171 |     template<int M, int K, bool filter> | 
| 172 |       struct SubGridQuadMIntersectorKPlueckerBase | 
| 173 |       { | 
| 174 |         __forceinline SubGridQuadMIntersectorKPlueckerBase(const vbool<K>& valid, const RayK<K>& ray) {} | 
| 175 |  | 
| 176 |         template<typename Epilog> | 
| 177 |         __forceinline bool intersectK(const vbool<K>& valid,  | 
| 178 |                                       RayK<K>& ray, | 
| 179 |                                       const Vec3vf<K>& v0, | 
| 180 |                                       const Vec3vf<K>& v1, | 
| 181 |                                       const Vec3vf<K>& v2, | 
| 182 |                                       const Vec3vf<K>& v3, | 
| 183 |                                       const GridMesh::Grid &g,  | 
| 184 |                                       const SubGrid &subgrid, | 
| 185 |                                       const unsigned int i, | 
| 186 |                                       const Epilog& epilog) const | 
| 187 |         { | 
| 188 | 	  UVIdentity<K> mapUV; | 
| 189 | 	  PlueckerHitK<K,UVIdentity<K>> hit(mapUV); | 
| 190 | 	  PlueckerIntersectorK<M,K> intersector; | 
| 191 |  | 
| 192 |           const vbool<K> valid0 = intersector.intersectK(valid,ray,v0,v1,v3,mapUV,hit); | 
| 193 | 	  if (any(valid0)) | 
| 194 | 	    { | 
| 195 | 	      interpolateUV(valid0,hit,g,subgrid,i); | 
| 196 | 	      epilog(valid0,hit); | 
| 197 | 	    } | 
| 198 |           const vbool<K> valid1 = intersector.intersectK(valid,ray,v2,v3,v1,mapUV,hit); | 
| 199 | 	  if (any(valid1)) | 
| 200 | 	    { | 
| 201 | 	      hit.U = hit.UVW - hit.U; | 
| 202 | 	      hit.V = hit.UVW - hit.V;	       | 
| 203 | 	      interpolateUV(valid1,hit,g,subgrid,i); | 
| 204 | 	      epilog(valid1,hit); | 
| 205 | 	    } | 
| 206 | 	  return any(valid0|valid1);	   | 
| 207 |         } | 
| 208 |  | 
| 209 |        template<typename Epilog> | 
| 210 |         __forceinline bool occludedK(const vbool<K>& valid,  | 
| 211 | 				     RayK<K>& ray, | 
| 212 | 				     const Vec3vf<K>& v0, | 
| 213 | 				     const Vec3vf<K>& v1, | 
| 214 | 				     const Vec3vf<K>& v2, | 
| 215 | 				     const Vec3vf<K>& v3, | 
| 216 | 				     const GridMesh::Grid &g,  | 
| 217 | 				     const SubGrid &subgrid, | 
| 218 | 				     const unsigned int i, | 
| 219 | 				     const Epilog& epilog) const | 
| 220 |         { | 
| 221 | 	  UVIdentity<K> mapUV; | 
| 222 | 	  PlueckerHitK<K,UVIdentity<K>> hit(mapUV); | 
| 223 | 	  PlueckerIntersectorK<M,K> intersector; | 
| 224 |  | 
| 225 | 	  vbool<K> valid_final = valid; | 
| 226 |           const vbool<K> valid0 = intersector.intersectK(valid,ray,v0,v1,v3,mapUV,hit); | 
| 227 | 	  if (any(valid0)) | 
| 228 | 	    { | 
| 229 | 	      interpolateUV(valid0,hit,g,subgrid,i); | 
| 230 | 	      epilog(valid0,hit); | 
| 231 | 	      valid_final &= !valid0; | 
| 232 | 	    } | 
| 233 | 	  if (none(valid_final)) return true;	      	   | 
| 234 |           const vbool<K> valid1 = intersector.intersectK(valid,ray,v2,v3,v1,mapUV,hit); | 
| 235 | 	  if (any(valid1)) | 
| 236 | 	    { | 
| 237 | 	      hit.U = hit.UVW - hit.U; | 
| 238 | 	      hit.V = hit.UVW - hit.V;	       | 
| 239 | 	      interpolateUV(valid1,hit,g,subgrid,i); | 
| 240 | 	      epilog(valid1,hit); | 
| 241 | 	      valid_final &= !valid1;	       | 
| 242 | 	    } | 
| 243 | 	  return none(valid_final); | 
| 244 |         } | 
| 245 |  | 
| 246 | 	 | 
| 247 |       }; | 
| 248 |  | 
| 249 |  | 
| 250 |      | 
| 251 |  | 
| 252 |     template<int M, int K, bool filter> | 
| 253 |       struct SubGridQuadMIntersectorKPluecker : public SubGridQuadMIntersectorKPlueckerBase<M,K,filter> | 
| 254 |     { | 
| 255 |       __forceinline SubGridQuadMIntersectorKPluecker(const vbool<K>& valid, const RayK<K>& ray) | 
| 256 |         : SubGridQuadMIntersectorKPlueckerBase<M,K,filter>(valid,ray) {} | 
| 257 |  | 
| 258 |       __forceinline void intersect1(RayHitK<K>& ray, size_t k, IntersectContext* context, | 
| 259 |                                     const Vec3vf<M>& v0, const Vec3vf<M>& v1, const Vec3vf<M>& v2, const Vec3vf<M>& v3, const GridMesh::Grid &g, const SubGrid &subgrid) const | 
| 260 |       { | 
| 261 | 	UVIdentity<M> mapUV; | 
| 262 | 	PlueckerHitM<M,UVIdentity<M>> hit(mapUV); | 
| 263 |         Intersect1KEpilogMU<M,K,filter> epilog(ray,k,context,subgrid.geomID(),subgrid.primID()); | 
| 264 | 	PlueckerIntersectorK<M,K> intersector; | 
| 265 | 	 | 
| 266 | 	/* intersect first triangle */ | 
| 267 | 	if (intersector.intersect(ray,k,v0,v1,v3,mapUV,hit))  | 
| 268 |           { | 
| 269 |             interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 270 |             epilog(hit.valid,hit); | 
| 271 |           } | 
| 272 |  | 
| 273 | 	/* intersect second triangle */ | 
| 274 | 	if (intersector.intersect(ray,k,v2,v3,v1,mapUV,hit))  | 
| 275 |           { | 
| 276 | 	    hit.U = hit.UVW - hit.U; | 
| 277 | 	    hit.V = hit.UVW - hit.V; | 
| 278 |             interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 279 |             epilog(hit.valid,hit); | 
| 280 |           } | 
| 281 |       } | 
| 282 |        | 
| 283 |       __forceinline bool occluded1(RayK<K>& ray, size_t k, IntersectContext* context, | 
| 284 |                                    const Vec3vf<M>& v0, const Vec3vf<M>& v1, const Vec3vf<M>& v2, const Vec3vf<M>& v3, const GridMesh::Grid &g, const SubGrid &subgrid) const | 
| 285 |       { | 
| 286 | 	UVIdentity<M> mapUV; | 
| 287 | 	PlueckerHitM<M,UVIdentity<M>> hit(mapUV); | 
| 288 |         Occluded1KEpilogMU<M,K,filter> epilog(ray,k,context,subgrid.geomID(),subgrid.primID());	 | 
| 289 | 	PlueckerIntersectorK<M,K> intersector; | 
| 290 | 	 | 
| 291 | 	/* intersect first triangle */ | 
| 292 | 	if (intersector.intersect(ray,k,v0,v1,v3,mapUV,hit))  | 
| 293 |         { | 
| 294 |           interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 295 |           if (epilog(hit.valid,hit)) return true; | 
| 296 |         } | 
| 297 |  | 
| 298 | 	/* intersect second triangle */ | 
| 299 | 	if (intersector.intersect(ray,k,v2,v3,v1,mapUV,hit))  | 
| 300 |         { | 
| 301 | 	  hit.U = hit.UVW - hit.U; | 
| 302 | 	  hit.V = hit.UVW - hit.V;	   | 
| 303 |           interpolateUV<M>(hit,g,subgrid,vint<M>(0,1,1,0),vint<M>(0,0,1,1)); | 
| 304 |           if (epilog(hit.valid,hit)) return true; | 
| 305 |         } | 
| 306 |         return false; | 
| 307 |       } | 
| 308 |     }; | 
| 309 |  | 
| 310 |  | 
| 311 | #if defined (__AVX__) | 
| 312 |  | 
| 313 |     /*! Intersects 4 quads with 1 ray using AVX */ | 
| 314 |     template<int K, bool filter> | 
| 315 |       struct SubGridQuadMIntersectorKPluecker<4,K,filter> : public SubGridQuadMIntersectorKPlueckerBase<4,K,filter> | 
| 316 |     { | 
| 317 |       __forceinline SubGridQuadMIntersectorKPluecker(const vbool<K>& valid, const RayK<K>& ray) | 
| 318 |         : SubGridQuadMIntersectorKPlueckerBase<4,K,filter>(valid,ray) {} | 
| 319 |        | 
| 320 |       template<typename Epilog> | 
| 321 |         __forceinline bool intersect1(RayK<K>& ray, size_t k,const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3,  | 
| 322 |                                       const GridMesh::Grid &g, const SubGrid &subgrid, const Epilog& epilog) const | 
| 323 |       { | 
| 324 |         const Vec3vf8 vtx0(vfloat8(v0.x,v2.x),vfloat8(v0.y,v2.y),vfloat8(v0.z,v2.z)); | 
| 325 | #if !defined(EMBREE_BACKFACE_CULLING) | 
| 326 |         const Vec3vf8 vtx1(vfloat8(v1.x),vfloat8(v1.y),vfloat8(v1.z)); | 
| 327 |         const Vec3vf8 vtx2(vfloat8(v3.x),vfloat8(v3.y),vfloat8(v3.z)); | 
| 328 | #else | 
| 329 |         const Vec3vf8 vtx1(vfloat8(v1.x,v3.x),vfloat8(v1.y,v3.y),vfloat8(v1.z,v3.z)); | 
| 330 |         const Vec3vf8 vtx2(vfloat8(v3.x,v1.x),vfloat8(v3.y,v1.y),vfloat8(v3.z,v1.z)); | 
| 331 | #endif | 
| 332 | 	UVIdentity<8> mapUV; | 
| 333 | 	PlueckerHitM<8,UVIdentity<8>> hit(mapUV); | 
| 334 | 	PlueckerIntersectorK<8,K> intersector; | 
| 335 | 	const vbool8 flags(0,0,0,0,1,1,1,1); | 
| 336 |         if (unlikely(intersector.intersect(ray,k,vtx0,vtx1,vtx2,mapUV,hit)))	 | 
| 337 |         { | 
| 338 |           /* correct U,V interpolation across the entire grid */ | 
| 339 | 	  const vfloat8 U = select(flags,hit.UVW - hit.V,hit.U);	   | 
| 340 | 	  const vfloat8 V = select(flags,hit.UVW - hit.U,hit.V); | 
| 341 | 	  hit.U = U; | 
| 342 | 	  hit.V = V;	   | 
| 343 | 	  hit.vNg *= select(flags,vfloat8(-1.0f),vfloat8(1.0f));  | 
| 344 |           interpolateUV<8>(hit,g,subgrid,vint<8>(0,1,1,0,0,1,1,0),vint<8>(0,0,1,1,0,0,1,1)); | 
| 345 |           if (unlikely(epilog(hit.valid,hit))) | 
| 346 |             return true; | 
| 347 |         } | 
| 348 |         return false; | 
| 349 |       } | 
| 350 |        | 
| 351 |       __forceinline bool intersect1(RayHitK<K>& ray, size_t k, IntersectContext* context, | 
| 352 |                                     const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, const GridMesh::Grid &g, const SubGrid &subgrid) const | 
| 353 |       { | 
| 354 |         return intersect1(ray,k,v0,v1,v2,v3,g,subgrid,Intersect1KEpilogMU<8,K,filter>(ray,k,context,subgrid.geomID(),subgrid.primID())); | 
| 355 |       } | 
| 356 |        | 
| 357 |       __forceinline bool occluded1(RayK<K>& ray, size_t k, IntersectContext* context, | 
| 358 |                                    const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, const GridMesh::Grid &g, const SubGrid &subgrid) const | 
| 359 |       { | 
| 360 |         return intersect1(ray,k,v0,v1,v2,v3,g,subgrid,Occluded1KEpilogMU<8,K,filter>(ray,k,context,subgrid.geomID(),subgrid.primID())); | 
| 361 |       } | 
| 362 |     }; | 
| 363 | #endif | 
| 364 |  | 
| 365 |      | 
| 366 |   } | 
| 367 | } | 
| 368 |  |