godot/thirdparty/embree/kernels/geometry/quad_intersector_pluecker.h

// Copyright 2009-2021 Intel Corporation
// SPDX-License-Identifier: Apache-2.0

#pragma once

#include "quad_intersector_moeller.h"

/*! Modified Pluecker ray/triangle intersector. The test first shifts
 *  the ray origin into the origin of the coordinate system and then
 *  uses Pluecker coordinates for the intersection. Due to the shift,
 *  the Pluecker coordinate calculation simplifies and the tests get
 *  numerically stable. The edge equations are watertight along the
 *  edge for neighboring triangles. */

namespace embree
{
  namespace isa
  {
    template<int M>
    struct QuadHitPlueckerM
    { … };

    template<int K>
    struct QuadHitPlueckerK
    { … };

    struct PlueckerIntersectorTriangle1
    { … };

    /*! Intersects M quads with 1 ray */
    template<int M, bool filter>
    struct QuadMIntersector1Pluecker
    { … };

#if defined(__AVX__)

    /*! Intersects 4 quads with 1 ray using AVX */
    template<bool filter>
    struct QuadMIntersector1Pluecker<4,filter>
    {
      __forceinline QuadMIntersector1Pluecker() {}

      __forceinline QuadMIntersector1Pluecker(const Ray& ray, const void* ptr) {}
      
      template<typename Epilog>
      __forceinline bool intersect(Ray& ray, const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, const Epilog& epilog) const
      {
        const Vec3vf8 vtx0(vfloat8(v0.x,v2.x),vfloat8(v0.y,v2.y),vfloat8(v0.z,v2.z));
#if !defined(EMBREE_BACKFACE_CULLING)
        const Vec3vf8 vtx1(vfloat8(v1.x),vfloat8(v1.y),vfloat8(v1.z));
        const Vec3vf8 vtx2(vfloat8(v3.x),vfloat8(v3.y),vfloat8(v3.z));
#else
        const Vec3vf8 vtx1(vfloat8(v1.x,v3.x),vfloat8(v1.y,v3.y),vfloat8(v1.z,v3.z));
        const Vec3vf8 vtx2(vfloat8(v3.x,v1.x),vfloat8(v3.y,v1.y),vfloat8(v3.z,v1.z));
#endif
        const vbool8 flags(0,0,0,0,1,1,1,1);
        return PlueckerIntersectorTriangle1::intersect<8>(ray,vtx0,vtx1,vtx2,flags,epilog); 
      }
      
      __forceinline bool intersect(RayHit& ray, RayQueryContext* context, const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, 
                                   const vuint4& geomID, const vuint4& primID) const
      {
        return intersect(ray,v0,v1,v2,v3,Intersect1EpilogM<8,filter>(ray,context,vuint8(geomID),vuint8(primID)));
      }
      
      __forceinline bool occluded(Ray& ray, RayQueryContext* context, const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3,
                                  const vuint4& geomID, const vuint4& primID) const
      {
        return intersect(ray,v0,v1,v2,v3,Occluded1EpilogM<8,filter>(ray,context,vuint8(geomID),vuint8(primID)));
      }
    };

#endif


    /* ----------------------------- */
    /* -- ray packet intersectors -- */
    /* ----------------------------- */

    struct PlueckerIntersector1KTriangleM
    { … };

    template<int M, int K, bool filter>
    struct QuadMIntersectorKPlueckerBase
    { … };

    template<int M, int K, bool filter>
      struct QuadMIntersectorKPluecker : public QuadMIntersectorKPlueckerBase<M,K,filter>
    { … };

#if defined(__AVX__)

    /*! Intersects 4 quads with 1 ray using AVX */
    template<int K, bool filter>
    struct QuadMIntersectorKPluecker<4,K,filter> : public QuadMIntersectorKPlueckerBase<4,K,filter>
    {
      __forceinline QuadMIntersectorKPluecker(const vbool<K>& valid, const RayK<K>& ray)
        : QuadMIntersectorKPlueckerBase<4,K,filter>(valid,ray) {}
      
      template<typename Epilog>
      __forceinline bool intersect1(RayK<K>& ray, size_t k, const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, const Epilog& epilog) const
      {
        const Vec3vf8 vtx0(vfloat8(v0.x,v2.x),vfloat8(v0.y,v2.y),vfloat8(v0.z,v2.z));
        const vbool8 flags(0,0,0,0,1,1,1,1);
#if !defined(EMBREE_BACKFACE_CULLING)
        const Vec3vf8 vtx1(vfloat8(v1.x),vfloat8(v1.y),vfloat8(v1.z));
        const Vec3vf8 vtx2(vfloat8(v3.x),vfloat8(v3.y),vfloat8(v3.z));
#else
        const Vec3vf8 vtx1(vfloat8(v1.x,v3.x),vfloat8(v1.y,v3.y),vfloat8(v1.z,v3.z));
        const Vec3vf8 vtx2(vfloat8(v3.x,v1.x),vfloat8(v3.y,v1.y),vfloat8(v3.z,v1.z));
#endif
        return PlueckerIntersector1KTriangleM::intersect1<8,K>(ray,k,vtx0,vtx1,vtx2,flags,epilog); 
      }
      
      __forceinline bool intersect1(RayHitK<K>& ray, size_t k, RayQueryContext* context,
                                    const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, 
                                    const vuint4& geomID, const vuint4& primID) const
      {
        return intersect1(ray,k,v0,v1,v2,v3,Intersect1KEpilogM<8,K,filter>(ray,k,context,vuint8(geomID),vuint8(primID)));
      }
      
      __forceinline bool occluded1(RayK<K>& ray, size_t k, RayQueryContext* context,
                                   const Vec3vf4& v0, const Vec3vf4& v1, const Vec3vf4& v2, const Vec3vf4& v3, 
                                   const vuint4& geomID, const vuint4& primID) const
      {
        return intersect1(ray,k,v0,v1,v2,v3,Occluded1KEpilogM<8,K,filter>(ray,k,context,vuint8(geomID),vuint8(primID)));
      }
    };

#endif
  }
}