1// Copyright 2009-2021 Intel Corporation
2// SPDX-License-Identifier: Apache-2.0
3
4#pragma once
5
6namespace embree
7{
8 namespace isa
9 {
10 /*! Intersects a ray with a quad with backface culling
11 * enabled. The quad v0,v1,v2,v3 is split into two triangles
12 * v0,v1,v3 and v2,v3,v1. The edge v1,v2 decides which of the two
13 * triangles gets intersected. */
14 template<int N>
15 __forceinline vbool<N> intersect_quad_backface_culling(const vbool<N>& valid0,
16 const Vec3fa& ray_org,
17 const Vec3fa& ray_dir,
18 const float ray_tnear,
19 const float ray_tfar,
20 const Vec3vf<N>& quad_v0,
21 const Vec3vf<N>& quad_v1,
22 const Vec3vf<N>& quad_v2,
23 const Vec3vf<N>& quad_v3,
24 vfloat<N>& u_o,
25 vfloat<N>& v_o,
26 vfloat<N>& t_o)
27 {
28 /* calculate vertices relative to ray origin */
29 vbool<N> valid = valid0;
30 const Vec3vf<N> O = Vec3vf<N>(ray_org);
31 const Vec3vf<N> D = Vec3vf<N>(ray_dir);
32 const Vec3vf<N> va = quad_v0-O;
33 const Vec3vf<N> vb = quad_v1-O;
34 const Vec3vf<N> vc = quad_v2-O;
35 const Vec3vf<N> vd = quad_v3-O;
36
37 const Vec3vf<N> edb = vb-vd;
38 const vfloat<N> WW = dot(cross(vd,edb),D);
39 const Vec3vf<N> v0 = select(WW <= 0.0f,va,vc);
40 const Vec3vf<N> v1 = select(WW <= 0.0f,vb,vd);
41 const Vec3vf<N> v2 = select(WW <= 0.0f,vd,vb);
42
43 /* calculate edges */
44 const Vec3vf<N> e0 = v2-v0;
45 const Vec3vf<N> e1 = v0-v1;
46
47 /* perform edge tests */
48 const vfloat<N> U = dot(cross(v0,e0),D);
49 const vfloat<N> V = dot(cross(v1,e1),D);
50 valid &= max(U,V) <= 0.0f;
51 if (unlikely(none(valid))) return false;
52
53 /* calculate geometry normal and denominator */
54 const Vec3vf<N> Ng = cross(e1,e0);
55 const vfloat<N> den = dot(Ng,D);
56 const vfloat<N> rcpDen = rcp(den);
57
58 /* perform depth test */
59 const vfloat<N> t = rcpDen*dot(v0,Ng);
60 valid &= vfloat<N>(ray_tnear) <= t & t <= vfloat<N>(ray_tfar);
61 if (unlikely(none(valid))) return false;
62
63 /* avoid division by 0 */
64 valid &= den != vfloat<N>(zero);
65 if (unlikely(none(valid))) return false;
66
67 /* update hit information */
68 t_o = t;
69 u_o = U * rcpDen;
70 v_o = V * rcpDen;
71 u_o = select(WW <= 0.0f,u_o,1.0f-u_o);
72 v_o = select(WW <= 0.0f,v_o,1.0f-v_o);
73 return valid;
74 }
75 }
76}
77