1 | // Copyright 2009-2021 Intel Corporation |
2 | // SPDX-License-Identifier: Apache-2.0 |
3 | |
4 | #pragma once |
5 | |
6 | #include "math.h" |
7 | |
8 | namespace embree |
9 | { |
10 | struct Vec3fa; |
11 | |
12 | //////////////////////////////////////////////////////////////////////////////// |
13 | /// Generic 3D vector Class |
14 | //////////////////////////////////////////////////////////////////////////////// |
15 | |
16 | template<typename T> struct Vec3 |
17 | { |
18 | enum { N = 3 }; |
19 | |
20 | union { |
21 | struct { |
22 | T x, y, z; |
23 | }; |
24 | #if !(defined(__WIN32__) && _MSC_VER == 1800) // workaround for older VS 2013 compiler |
25 | T components[N]; |
26 | #endif |
27 | }; |
28 | |
29 | typedef T Scalar; |
30 | |
31 | //////////////////////////////////////////////////////////////////////////////// |
32 | /// Construction |
33 | //////////////////////////////////////////////////////////////////////////////// |
34 | |
35 | __forceinline Vec3( ) {} |
36 | __forceinline explicit Vec3( const T& a ) : x(a), y(a), z(a) {} |
37 | __forceinline Vec3( const T& x, const T& y, const T& z ) : x(x), y(y), z(z) {} |
38 | |
39 | __forceinline Vec3( const Vec3& other ) { x = other.x; y = other.y; z = other.z; } |
40 | __forceinline Vec3( const Vec3fa& other ); |
41 | |
42 | template<typename T1> __forceinline Vec3( const Vec3<T1>& a ) : x(T(a.x)), y(T(a.y)), z(T(a.z)) {} |
43 | template<typename T1> __forceinline Vec3& operator =(const Vec3<T1>& other) { x = other.x; y = other.y; z = other.z; return *this; } |
44 | |
45 | __forceinline Vec3& operator =(const Vec3& other) { x = other.x; y = other.y; z = other.z; return *this; } |
46 | |
47 | //////////////////////////////////////////////////////////////////////////////// |
48 | /// Constants |
49 | //////////////////////////////////////////////////////////////////////////////// |
50 | |
51 | __forceinline Vec3( ZeroTy ) : x(zero), y(zero), z(zero) {} |
52 | __forceinline Vec3( OneTy ) : x(one), y(one), z(one) {} |
53 | __forceinline Vec3( PosInfTy ) : x(pos_inf), y(pos_inf), z(pos_inf) {} |
54 | __forceinline Vec3( NegInfTy ) : x(neg_inf), y(neg_inf), z(neg_inf) {} |
55 | |
56 | #if defined(__WIN32__) && (_MSC_VER == 1800) // workaround for older VS 2013 compiler |
57 | __forceinline const T& operator []( const size_t axis ) const { assert(axis < 3); return (&x)[axis]; } |
58 | __forceinline T& operator []( const size_t axis ) { assert(axis < 3); return (&x)[axis]; } |
59 | #else |
60 | __forceinline const T& operator [](const size_t axis) const { assert(axis < 3); return components[axis]; } |
61 | __forceinline T& operator [](const size_t axis) { assert(axis < 3); return components[axis]; } |
62 | #endif |
63 | }; |
64 | |
65 | //////////////////////////////////////////////////////////////////////////////// |
66 | /// Unary Operators |
67 | //////////////////////////////////////////////////////////////////////////////// |
68 | |
69 | template<typename T> __forceinline Vec3<T> operator +( const Vec3<T>& a ) { return Vec3<T>(+a.x, +a.y, +a.z); } |
70 | template<typename T> __forceinline Vec3<T> operator -( const Vec3<T>& a ) { return Vec3<T>(-a.x, -a.y, -a.z); } |
71 | template<typename T> __forceinline Vec3<T> abs ( const Vec3<T>& a ) { return Vec3<T>(abs (a.x), abs (a.y), abs (a.z)); } |
72 | template<typename T> __forceinline Vec3<T> rcp ( const Vec3<T>& a ) { return Vec3<T>(rcp (a.x), rcp (a.y), rcp (a.z)); } |
73 | template<typename T> __forceinline Vec3<T> rsqrt ( const Vec3<T>& a ) { return Vec3<T>(rsqrt(a.x), rsqrt(a.y), rsqrt(a.z)); } |
74 | template<typename T> __forceinline Vec3<T> sqrt ( const Vec3<T>& a ) { return Vec3<T>(sqrt (a.x), sqrt (a.y), sqrt (a.z)); } |
75 | |
76 | template<typename T> __forceinline Vec3<T> zero_fix( const Vec3<T>& a ) |
77 | { |
78 | return Vec3<T>(select(abs(a.x)<min_rcp_input,T(min_rcp_input),a.x), |
79 | select(abs(a.y)<min_rcp_input,T(min_rcp_input),a.y), |
80 | select(abs(a.z)<min_rcp_input,T(min_rcp_input),a.z)); |
81 | } |
82 | template<typename T> __forceinline Vec3<T> rcp_safe(const Vec3<T>& a) { return rcp(zero_fix(a)); } |
83 | |
84 | //////////////////////////////////////////////////////////////////////////////// |
85 | /// Binary Operators |
86 | //////////////////////////////////////////////////////////////////////////////// |
87 | |
88 | template<typename T> __forceinline Vec3<T> operator +( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<T>(a.x + b.x, a.y + b.y, a.z + b.z); } |
89 | template<typename T> __forceinline Vec3<T> operator -( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<T>(a.x - b.x, a.y - b.y, a.z - b.z); } |
90 | template<typename T> __forceinline Vec3<T> operator *( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<T>(a.x * b.x, a.y * b.y, a.z * b.z); } |
91 | template<typename T> __forceinline Vec3<T> operator *( const T& a, const Vec3<T>& b ) { return Vec3<T>(a * b.x, a * b.y, a * b.z); } |
92 | template<typename T> __forceinline Vec3<T> operator *( const Vec3<T>& a, const T& b ) { return Vec3<T>(a.x * b , a.y * b , a.z * b ); } |
93 | template<typename T> __forceinline Vec3<T> operator /( const Vec3<T>& a, const T& b ) { return Vec3<T>(a.x / b , a.y / b , a.z / b ); } |
94 | template<typename T> __forceinline Vec3<T> operator /( const T& a, const Vec3<T>& b ) { return Vec3<T>(a / b.x, a / b.y, a / b.z); } |
95 | template<typename T> __forceinline Vec3<T> operator /( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<T>(a.x / b.x, a.y / b.y, a.z / b.z); } |
96 | |
97 | template<typename T> __forceinline Vec3<T> min(const Vec3<T>& a, const Vec3<T>& b) { return Vec3<T>(min(a.x, b.x), min(a.y, b.y), min(a.z, b.z)); } |
98 | template<typename T> __forceinline Vec3<T> max(const Vec3<T>& a, const Vec3<T>& b) { return Vec3<T>(max(a.x, b.x), max(a.y, b.y), max(a.z, b.z)); } |
99 | |
100 | template<typename T> __forceinline Vec3<T> operator >>( const Vec3<T>& a, const int b ) { return Vec3<T>(a.x >> b, a.y >> b, a.z >> b); } |
101 | template<typename T> __forceinline Vec3<T> operator <<( const Vec3<T>& a, const int b ) { return Vec3<T>(a.x << b, a.y << b, a.z << b); } |
102 | |
103 | //////////////////////////////////////////////////////////////////////////////// |
104 | /// Ternary Operators |
105 | //////////////////////////////////////////////////////////////////////////////// |
106 | |
107 | template<typename T> __forceinline Vec3<T> madd ( const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>( madd(a.x,b.x,c.x), madd(a.y,b.y,c.y), madd(a.z,b.z,c.z)); } |
108 | template<typename T> __forceinline Vec3<T> msub ( const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>( msub(a.x,b.x,c.x), msub(a.y,b.y,c.y), msub(a.z,b.z,c.z)); } |
109 | template<typename T> __forceinline Vec3<T> nmadd ( const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>(nmadd(a.x,b.x,c.x),nmadd(a.y,b.y,c.y),nmadd(a.z,b.z,c.z));} |
110 | template<typename T> __forceinline Vec3<T> nmsub ( const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>(nmsub(a.x,b.x,c.x),nmsub(a.y,b.y,c.y),nmsub(a.z,b.z,c.z)); } |
111 | |
112 | template<typename T> __forceinline Vec3<T> madd ( const T& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>( madd(a,b.x,c.x), madd(a,b.y,c.y), madd(a,b.z,c.z)); } |
113 | template<typename T> __forceinline Vec3<T> msub ( const T& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>( msub(a,b.x,c.x), msub(a,b.y,c.y), msub(a,b.z,c.z)); } |
114 | template<typename T> __forceinline Vec3<T> nmadd ( const T& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>(nmadd(a,b.x,c.x),nmadd(a,b.y,c.y),nmadd(a,b.z,c.z));} |
115 | template<typename T> __forceinline Vec3<T> nmsub ( const T& a, const Vec3<T>& b, const Vec3<T>& c) { return Vec3<T>(nmsub(a,b.x,c.x),nmsub(a,b.y,c.y),nmsub(a,b.z,c.z)); } |
116 | |
117 | //////////////////////////////////////////////////////////////////////////////// |
118 | /// Assignment Operators |
119 | //////////////////////////////////////////////////////////////////////////////// |
120 | |
121 | template<typename T> __forceinline Vec3<T>& operator +=( Vec3<T>& a, const T b ) { a.x += b; a.y += b; a.z += b; return a; } |
122 | template<typename T> __forceinline Vec3<T>& operator +=( Vec3<T>& a, const Vec3<T>& b ) { a.x += b.x; a.y += b.y; a.z += b.z; return a; } |
123 | template<typename T> __forceinline Vec3<T>& operator -=( Vec3<T>& a, const Vec3<T>& b ) { a.x -= b.x; a.y -= b.y; a.z -= b.z; return a; } |
124 | template<typename T> __forceinline Vec3<T>& operator *=( Vec3<T>& a, const T& b ) { a.x *= b ; a.y *= b ; a.z *= b ; return a; } |
125 | template<typename T> __forceinline Vec3<T>& operator /=( Vec3<T>& a, const T& b ) { a.x /= b ; a.y /= b ; a.z /= b ; return a; } |
126 | |
127 | //////////////////////////////////////////////////////////////////////////////// |
128 | /// Reduction Operators |
129 | //////////////////////////////////////////////////////////////////////////////// |
130 | |
131 | template<typename T> __forceinline T reduce_add( const Vec3<T>& a ) { return a.x + a.y + a.z; } |
132 | template<typename T> __forceinline T reduce_mul( const Vec3<T>& a ) { return a.x * a.y * a.z; } |
133 | template<typename T> __forceinline T reduce_min( const Vec3<T>& a ) { return min(a.x, a.y, a.z); } |
134 | template<typename T> __forceinline T reduce_max( const Vec3<T>& a ) { return max(a.x, a.y, a.z); } |
135 | |
136 | //////////////////////////////////////////////////////////////////////////////// |
137 | /// Comparison Operators |
138 | //////////////////////////////////////////////////////////////////////////////// |
139 | |
140 | template<typename T> __forceinline bool operator ==( const Vec3<T>& a, const Vec3<T>& b ) { return a.x == b.x && a.y == b.y && a.z == b.z; } |
141 | template<typename T> __forceinline bool operator !=( const Vec3<T>& a, const Vec3<T>& b ) { return a.x != b.x || a.y != b.y || a.z != b.z; } |
142 | template<typename T> __forceinline bool operator < ( const Vec3<T>& a, const Vec3<T>& b ) { |
143 | if (a.x != b.x) return a.x < b.x; |
144 | if (a.y != b.y) return a.y < b.y; |
145 | if (a.z != b.z) return a.z < b.z; |
146 | return false; |
147 | } |
148 | |
149 | //////////////////////////////////////////////////////////////////////////////// |
150 | /// Shift Operators |
151 | //////////////////////////////////////////////////////////////////////////////// |
152 | |
153 | template<typename T> __forceinline Vec3<T> shift_right_1( const Vec3<T>& a ) { |
154 | return Vec3<T>(shift_right_1(a.x),shift_right_1(a.y),shift_right_1(a.z)); |
155 | } |
156 | |
157 | //////////////////////////////////////////////////////////////////////////////// |
158 | /// Select |
159 | //////////////////////////////////////////////////////////////////////////////// |
160 | |
161 | template<typename T> __forceinline Vec3<T> select ( bool s, const Vec3<T>& t, const Vec3<T>& f ) { |
162 | return Vec3<T>(select(s,t.x,f.x),select(s,t.y,f.y),select(s,t.z,f.z)); |
163 | } |
164 | |
165 | template<typename T> __forceinline Vec3<T> select ( const Vec3<bool>& s, const Vec3<T>& t, const Vec3<T>& f ) { |
166 | return Vec3<T>(select(s.x,t.x,f.x),select(s.y,t.y,f.y),select(s.z,t.z,f.z)); |
167 | } |
168 | |
169 | template<typename T> __forceinline Vec3<T> select ( const typename T::Bool& s, const Vec3<T>& t, const Vec3<T>& f ) { |
170 | return Vec3<T>(select(s,t.x,f.x),select(s,t.y,f.y),select(s,t.z,f.z)); |
171 | } |
172 | |
173 | template<typename T> |
174 | __forceinline Vec3<T> lerp(const Vec3<T>& v0, const Vec3<T>& v1, const T& t) { |
175 | return madd(Vec3<T>(T(1.0f)-t),v0,t*v1); |
176 | } |
177 | |
178 | template<typename T> __forceinline int maxDim ( const Vec3<T>& a ) |
179 | { |
180 | const Vec3<T> b = abs(a); |
181 | if (b.x > b.y) { |
182 | if (b.x > b.z) return 0; else return 2; |
183 | } else { |
184 | if (b.y > b.z) return 1; else return 2; |
185 | } |
186 | } |
187 | |
188 | //////////////////////////////////////////////////////////////////////////////// |
189 | /// Comparison Operators |
190 | //////////////////////////////////////////////////////////////////////////////// |
191 | |
192 | template<typename T> __forceinline Vec3<bool> eq_mask( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<bool>(a.x==b.x,a.y==b.y,a.z==b.z); } |
193 | template<typename T> __forceinline Vec3<bool> neq_mask(const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<bool>(a.x!=b.x,a.y!=b.y,a.z!=b.z); } |
194 | template<typename T> __forceinline Vec3<bool> lt_mask( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<bool>(a.x< b.x,a.y< b.y,a.z< b.z); } |
195 | template<typename T> __forceinline Vec3<bool> le_mask( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<bool>(a.x<=b.x,a.y<=b.y,a.z<=b.z); } |
196 | template<typename T> __forceinline Vec3<bool> gt_mask( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<bool>(a.x> b.x,a.y> b.y,a.z> b.z); } |
197 | template<typename T> __forceinline Vec3<bool> ge_mask( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<bool>(a.x>=b.x,a.y>=b.y,a.z>=b.z); } |
198 | |
199 | //////////////////////////////////////////////////////////////////////////////// |
200 | /// Euclidian Space Operators |
201 | //////////////////////////////////////////////////////////////////////////////// |
202 | |
203 | template<typename T> __forceinline T sqr ( const Vec3<T>& a ) { return dot(a,a); } |
204 | template<typename T> __forceinline T dot ( const Vec3<T>& a, const Vec3<T>& b ) { return madd(a.x,b.x,madd(a.y,b.y,a.z*b.z)); } |
205 | template<typename T> __forceinline T length ( const Vec3<T>& a ) { return sqrt(sqr(a)); } |
206 | template<typename T> __forceinline T rcp_length( const Vec3<T>& a ) { return rsqrt(sqr(a)); } |
207 | template<typename T> __forceinline Vec3<T> normalize( const Vec3<T>& a ) { return a*rsqrt(sqr(a)); } |
208 | template<typename T> __forceinline T distance ( const Vec3<T>& a, const Vec3<T>& b ) { return length(a-b); } |
209 | template<typename T> __forceinline Vec3<T> cross ( const Vec3<T>& a, const Vec3<T>& b ) { return Vec3<T>(msub(a.y,b.z,a.z*b.y), msub(a.z,b.x,a.x*b.z), msub(a.x,b.y,a.y*b.x)); } |
210 | |
211 | template<typename T> __forceinline Vec3<T> stable_triangle_normal( const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c ) |
212 | { |
213 | const T ab_x = a.z*b.y, ab_y = a.x*b.z, ab_z = a.y*b.x; |
214 | const T bc_x = b.z*c.y, bc_y = b.x*c.z, bc_z = b.y*c.x; |
215 | const Vec3<T> cross_ab(msub(a.y,b.z,ab_x), msub(a.z,b.x,ab_y), msub(a.x,b.y,ab_z)); |
216 | const Vec3<T> cross_bc(msub(b.y,c.z,bc_x), msub(b.z,c.x,bc_y), msub(b.x,c.y,bc_z)); |
217 | const auto sx = abs(ab_x) < abs(bc_x); |
218 | const auto sy = abs(ab_y) < abs(bc_y); |
219 | const auto sz = abs(ab_z) < abs(bc_z); |
220 | return Vec3<T>(select(sx,cross_ab.x,cross_bc.x), |
221 | select(sy,cross_ab.y,cross_bc.y), |
222 | select(sz,cross_ab.z,cross_bc.z)); |
223 | } |
224 | |
225 | template<typename T> __forceinline T sum ( const Vec3<T>& a ) { return a.x+a.y+a.z; } |
226 | |
227 | template<typename T> __forceinline T halfArea ( const Vec3<T>& d ) { return madd(d.x,(d.y+d.z),d.y*d.z); } |
228 | template<typename T> __forceinline T area ( const Vec3<T>& d ) { return 2.0f*halfArea(d); } |
229 | |
230 | template<typename T> __forceinline Vec3<T> normalize_safe( const Vec3<T>& a ) { |
231 | const T d = dot(a,a); return select(d == T( zero ), a , a*rsqrt(d) ); |
232 | } |
233 | |
234 | template<typename T> __forceinline T sqr_point_to_line_distance(const Vec3<T>& P, const Vec3<T>& Q0, const Vec3<T>& Q1) |
235 | { |
236 | const Vec3<T> N = cross(P-Q0,Q1-Q0); |
237 | const Vec3<T> D = Q1-Q0; |
238 | return dot(N,N)*rcp(dot(D,D)); |
239 | } |
240 | |
241 | template<typename T> __forceinline T sqr_point_to_line_distance(const Vec3<T>& PmQ0, const Vec3<T>& Q1mQ0) |
242 | { |
243 | const Vec3<T> N = cross(PmQ0,Q1mQ0); |
244 | const Vec3<T> D = Q1mQ0; |
245 | return dot(N,N)*rcp(dot(D,D)); |
246 | } |
247 | |
248 | //////////////////////////////////////////////////////////////////////////////// |
249 | /// Output Operators |
250 | //////////////////////////////////////////////////////////////////////////////// |
251 | |
252 | template<typename T> __forceinline embree_ostream operator<<(embree_ostream cout, const Vec3<T>& a) { |
253 | return cout << "(" << a.x << ", " << a.y << ", " << a.z << ")" ; |
254 | } |
255 | |
256 | typedef Vec3<bool > Vec3b; |
257 | typedef Vec3<int > Vec3i; |
258 | typedef Vec3<float> Vec3f; |
259 | } |
260 | |
261 | #include "vec3ba.h" |
262 | #include "vec3ia.h" |
263 | #include "vec3fa.h" |
264 | |
265 | //////////////////////////////////////////////////////////////////////////////// |
266 | /// SSE / AVX / MIC specializations |
267 | //////////////////////////////////////////////////////////////////////////////// |
268 | |
269 | #if defined __SSE__ |
270 | #include "../simd/sse.h" |
271 | #endif |
272 | |
273 | #if defined __AVX__ |
274 | #include "../simd/avx.h" |
275 | #endif |
276 | |
277 | #if defined(__AVX512F__) |
278 | #include "../simd/avx512.h" |
279 | #endif |
280 | |
281 | namespace embree |
282 | { |
283 | template<typename Out, typename In> |
284 | __forceinline Vec3<Out> broadcast(const Vec3<In>& a, const size_t k) { |
285 | return Vec3<Out>(Out(a.x[k]), Out(a.y[k]), Out(a.z[k])); |
286 | } |
287 | |
288 | template<> __forceinline Vec3<float>::Vec3(const Vec3fa& a) { x = a.x; y = a.y; z = a.z; } |
289 | |
290 | #if defined(__AVX__) |
291 | template<> __forceinline Vec3<vfloat4>::Vec3(const Vec3fa& a) { |
292 | x = a.x; y = a.y; z = a.z; |
293 | } |
294 | #elif defined(__SSE__) |
295 | template<> |
296 | __forceinline Vec3<vfloat4>::Vec3(const Vec3fa& a) { |
297 | const vfloat4 v = vfloat4(a.m128); x = shuffle<0,0,0,0>(v); y = shuffle<1,1,1,1>(v); z = shuffle<2,2,2,2>(v); |
298 | } |
299 | #endif |
300 | |
301 | #if defined(__SSE__) |
302 | template<> |
303 | __forceinline Vec3<vfloat4> broadcast<vfloat4,vfloat4>(const Vec3<vfloat4>& a, const size_t k) { |
304 | return Vec3<vfloat4>(vfloat4::broadcast(a: &a.x[k]), vfloat4::broadcast(a: &a.y[k]), vfloat4::broadcast(a: &a.z[k])); |
305 | } |
306 | |
307 | template<int i0, int i1, int i2, int i3> |
308 | __forceinline Vec3<vfloat4> shuffle(const Vec3<vfloat4>& b) { |
309 | return Vec3<vfloat4>(shuffle<i0,i1,i2,i3>(b.x), shuffle<i0,i1,i2,i3>(b.y), shuffle<i0,i1,i2,i3>(b.z)); |
310 | } |
311 | #endif |
312 | |
313 | #if defined(__AVX__) |
314 | template<> |
315 | __forceinline Vec3<vfloat8>::Vec3(const Vec3fa& a) { |
316 | x = a.x; y = a.y; z = a.z; |
317 | } |
318 | |
319 | template<> |
320 | __forceinline Vec3<vfloat8> broadcast<vfloat8,vfloat4>(const Vec3<vfloat4>& a, const size_t k) { |
321 | return Vec3<vfloat8>(vfloat8::broadcast(&a.x[k]), vfloat8::broadcast(&a.y[k]), vfloat8::broadcast(&a.z[k])); |
322 | } |
323 | template<> |
324 | __forceinline Vec3<vfloat8> broadcast<vfloat8,vfloat8>(const Vec3<vfloat8>& a, const size_t k) { |
325 | return Vec3<vfloat8>(vfloat8::broadcast(&a.x[k]), vfloat8::broadcast(&a.y[k]), vfloat8::broadcast(&a.z[k])); |
326 | } |
327 | |
328 | template<int i0, int i1, int i2, int i3> |
329 | __forceinline Vec3<vfloat8> shuffle(const Vec3<vfloat8>& b) { |
330 | return Vec3<vfloat8>(shuffle<i0,i1,i2,i3>(b.x), shuffle<i0,i1,i2,i3>(b.y), shuffle<i0,i1,i2,i3>(b.z)); |
331 | } |
332 | #endif |
333 | |
334 | #if defined(__AVX512F__) |
335 | template<> __forceinline Vec3<vfloat16>::Vec3(const Vec3fa& a) : x(a.x), y(a.y), z(a.z) {} |
336 | #endif |
337 | } |
338 | |