1 | /**************************************************************************** |
2 | ** |
3 | ** Copyright (C) 2016 Paul Lemire <paul.lemire350@gmail.com> |
4 | ** Contact: https://www.qt.io/licensing/ |
5 | ** |
6 | ** This file is part of the Qt3D module of the Qt Toolkit. |
7 | ** |
8 | ** $QT_BEGIN_LICENSE:LGPL$ |
9 | ** Commercial License Usage |
10 | ** Licensees holding valid commercial Qt licenses may use this file in |
11 | ** accordance with the commercial license agreement provided with the |
12 | ** Software or, alternatively, in accordance with the terms contained in |
13 | ** a written agreement between you and The Qt Company. For licensing terms |
14 | ** and conditions see https://www.qt.io/terms-conditions. For further |
15 | ** information use the contact form at https://www.qt.io/contact-us. |
16 | ** |
17 | ** GNU Lesser General Public License Usage |
18 | ** Alternatively, this file may be used under the terms of the GNU Lesser |
19 | ** General Public License version 3 as published by the Free Software |
20 | ** Foundation and appearing in the file LICENSE.LGPL3 included in the |
21 | ** packaging of this file. Please review the following information to |
22 | ** ensure the GNU Lesser General Public License version 3 requirements |
23 | ** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. |
24 | ** |
25 | ** GNU General Public License Usage |
26 | ** Alternatively, this file may be used under the terms of the GNU |
27 | ** General Public License version 2.0 or (at your option) the GNU General |
28 | ** Public license version 3 or any later version approved by the KDE Free |
29 | ** Qt Foundation. The licenses are as published by the Free Software |
30 | ** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 |
31 | ** included in the packaging of this file. Please review the following |
32 | ** information to ensure the GNU General Public License requirements will |
33 | ** be met: https://www.gnu.org/licenses/gpl-2.0.html and |
34 | ** https://www.gnu.org/licenses/gpl-3.0.html. |
35 | ** |
36 | ** $QT_END_LICENSE$ |
37 | ** |
38 | ****************************************************************************/ |
39 | |
40 | #ifndef QT3DCORE_VECTOR3D_SSE_P_H |
41 | #define QT3DCORE_VECTOR3D_SSE_P_H |
42 | |
43 | // |
44 | // W A R N I N G |
45 | // ------------- |
46 | // |
47 | // This file is not part of the Qt3D API. It exists purely as an |
48 | // implementation detail. This header file may change from version to |
49 | // version without notice, or even be removed. |
50 | // |
51 | // We mean it. |
52 | // |
53 | |
54 | #include <Qt3DCore/private/qt3dcore_global_p.h> |
55 | #include <QtCore/private/qsimd_p.h> |
56 | #include <QtCore/QtGlobal> |
57 | #include <QtGui/qvector3d.h> |
58 | #include <QDebug> |
59 | #include <math.h> |
60 | |
61 | #ifdef QT_COMPILER_SUPPORTS_SSE2 |
62 | |
63 | QT_BEGIN_NAMESPACE |
64 | |
65 | namespace Qt3DCore { |
66 | |
67 | class Matrix4x4_SSE; |
68 | class Matrix4x4_AVX2; |
69 | class Vector4D_SSE; |
70 | |
71 | class Vector3D_SSE |
72 | { |
73 | public: |
74 | |
75 | Q_ALWAYS_INLINE Vector3D_SSE() |
76 | : m_xyzw(_mm_setzero_ps()) |
77 | { |
78 | } |
79 | |
80 | explicit Q_ALWAYS_INLINE Vector3D_SSE(Qt::Initialization) {} |
81 | |
82 | explicit Q_ALWAYS_INLINE Vector3D_SSE(float x, float y, float z) |
83 | : m_xyzw(_mm_set_ps(z: 0.0f, y: z, x: y, w: x)) |
84 | { |
85 | } |
86 | |
87 | explicit Q_ALWAYS_INLINE Vector3D_SSE(QVector3D v) |
88 | : m_xyzw(_mm_set_ps(z: 0.0f, y: v.z(), x: v.y(), w: v.x())) |
89 | { |
90 | } |
91 | |
92 | explicit Q_3DCORE_PRIVATE_EXPORT Vector3D_SSE(const Vector4D_SSE &v); |
93 | |
94 | Q_ALWAYS_INLINE Vector3D_SSE &operator+=(Vector3D_SSE vector) |
95 | { |
96 | m_xyzw = _mm_add_ps(a: m_xyzw, b: vector.m_xyzw); |
97 | return *this; |
98 | } |
99 | |
100 | Q_ALWAYS_INLINE Vector3D_SSE &operator-=(Vector3D_SSE vector) |
101 | { |
102 | m_xyzw = _mm_sub_ps(a: m_xyzw, b: vector.m_xyzw); |
103 | return *this; |
104 | } |
105 | |
106 | Q_ALWAYS_INLINE Vector3D_SSE &operator*=(Vector3D_SSE vector) |
107 | { |
108 | m_xyzw = _mm_mul_ps(a: m_xyzw, b: vector.m_xyzw); |
109 | return *this; |
110 | } |
111 | |
112 | Q_ALWAYS_INLINE Vector3D_SSE &operator/=(Vector3D_SSE vector) |
113 | { |
114 | m_xyzw = _mm_div_ps(a: m_xyzw, b: vector.m_xyzw); |
115 | return *this; |
116 | } |
117 | |
118 | Q_ALWAYS_INLINE Vector3D_SSE &operator*=(float factor) |
119 | { |
120 | m_xyzw = _mm_mul_ps(a: m_xyzw, b: _mm_set1_ps(w: factor)); |
121 | return *this; |
122 | } |
123 | |
124 | Q_ALWAYS_INLINE Vector3D_SSE &operator/=(float factor) |
125 | { |
126 | m_xyzw = _mm_div_ps(a: m_xyzw, b: _mm_set1_ps(w: factor)); |
127 | return *this; |
128 | } |
129 | |
130 | Q_ALWAYS_INLINE bool operator==(Vector3D_SSE other) const |
131 | { |
132 | // 0b111 == 0x7 |
133 | return ((_mm_movemask_ps(a: _mm_cmpeq_ps(a: m_xyzw, b: other.m_xyzw)) & 0x7) == 0x7); |
134 | } |
135 | |
136 | Q_ALWAYS_INLINE bool operator!=(Vector3D_SSE other) const |
137 | { |
138 | return !(*this == other); |
139 | } |
140 | |
141 | Q_ALWAYS_INLINE QVector3D toQVector3D() const |
142 | { |
143 | return QVector3D(x(), y(), z()); |
144 | } |
145 | |
146 | Q_ALWAYS_INLINE float lengthSquared() const |
147 | { |
148 | return Qt3DCore::Vector3D_SSE::dotProduct(a: *this, b: *this); |
149 | } |
150 | |
151 | Q_ALWAYS_INLINE float length() const |
152 | { |
153 | return sqrt(x: Qt3DCore::Vector3D_SSE::dotProduct(a: *this, b: *this)); |
154 | } |
155 | |
156 | Q_ALWAYS_INLINE float distanceToPoint(const Vector3D_SSE &point) const |
157 | { |
158 | return (*this - point).length(); |
159 | } |
160 | |
161 | Q_ALWAYS_INLINE void normalize() |
162 | { |
163 | const float len = length(); |
164 | m_xyzw = _mm_div_ps(a: m_xyzw, b: _mm_set_ps1(w: len)); |
165 | } |
166 | |
167 | Q_ALWAYS_INLINE Vector3D_SSE normalized() const |
168 | { |
169 | Vector3D_SSE v = *this; |
170 | v.normalize(); |
171 | return v; |
172 | } |
173 | |
174 | Q_ALWAYS_INLINE bool isNull() const |
175 | { |
176 | // Ignore last bit |
177 | // 0b111 = 0x7 |
178 | return ((_mm_movemask_ps(a: _mm_cmpeq_ps(a: m_xyzw, b: _mm_set_ps1(w: 0.0f))) & 0x7) == 0x7); |
179 | } |
180 | |
181 | #if defined(__AVX2__) && defined(QT_COMPILER_SUPPORTS_AVX2) |
182 | Q_3DCORE_PRIVATE_EXPORT Vector3D_SSE unproject(const Matrix4x4_AVX2 &modelView, const Matrix4x4_AVX2 &projection, const QRect &viewport) const; |
183 | Q_3DCORE_PRIVATE_EXPORT Vector3D_SSE project(const Matrix4x4_AVX2 &modelView, const Matrix4x4_AVX2 &projection, const QRect &viewport) const; |
184 | #else |
185 | Q_3DCORE_PRIVATE_EXPORT Vector3D_SSE unproject(const Matrix4x4_SSE &modelView, const Matrix4x4_SSE &projection, const QRect &viewport) const; |
186 | Q_3DCORE_PRIVATE_EXPORT Vector3D_SSE project(const Matrix4x4_SSE &modelView, const Matrix4x4_SSE &projection, const QRect &viewport) const; |
187 | #endif |
188 | |
189 | Q_ALWAYS_INLINE float x() const { return _mm_cvtss_f32(a: m_xyzw); } |
190 | |
191 | Q_ALWAYS_INLINE float y() const |
192 | { |
193 | // 0b01010101 = 0x55 |
194 | return _mm_cvtss_f32(_mm_shuffle_ps(m_xyzw, m_xyzw, 0x55)); |
195 | } |
196 | |
197 | Q_ALWAYS_INLINE float z() const |
198 | { |
199 | // 0b10101010 = 0xaa |
200 | return _mm_cvtss_f32(a: _mm_unpackhi_ps(a: m_xyzw, b: m_xyzw)); |
201 | } |
202 | |
203 | Q_ALWAYS_INLINE void setX(float x) |
204 | { |
205 | m_xyzw = _mm_move_ss(a: m_xyzw, b: _mm_set_ss(w: x)); |
206 | } |
207 | |
208 | Q_ALWAYS_INLINE void setY(float y) |
209 | { |
210 | // m_xyzw = a, b, c, d |
211 | |
212 | // y, y, y, y |
213 | const __m128 yVec = _mm_set_ps1(w: y); |
214 | |
215 | // y, y, a, a |
216 | // 0b00000000 == 0x0 |
217 | const __m128 yaVec = _mm_shuffle_ps(yVec, m_xyzw, 0x0); |
218 | |
219 | // a, y, c, d |
220 | // 0b11100010 == 0xe2 |
221 | m_xyzw = _mm_shuffle_ps(yaVec, m_xyzw, 0xe2); |
222 | } |
223 | |
224 | Q_ALWAYS_INLINE void setZ(float z) |
225 | { |
226 | // m_xyzw = a, b, c, d |
227 | |
228 | // z, z, z, z |
229 | const __m128 zVec = _mm_set_ps1(w: z); |
230 | |
231 | // z, z, d, d |
232 | // 0b11110000 == 0xf0 |
233 | const __m128 zdVec = _mm_shuffle_ps(zVec, m_xyzw, 0xf0); |
234 | |
235 | // a, b, z, d |
236 | // 0b10000100 == 0x84 |
237 | m_xyzw = _mm_shuffle_ps(m_xyzw, zdVec, 0x84); |
238 | } |
239 | |
240 | Q_ALWAYS_INLINE float operator[](int idx) const |
241 | { |
242 | switch (idx) { |
243 | case 0: |
244 | return x(); |
245 | case 1: |
246 | return y(); |
247 | case 2: |
248 | return z(); |
249 | default: |
250 | Q_UNREACHABLE(); |
251 | return 0.0f; |
252 | } |
253 | } |
254 | |
255 | struct DigitWrapper |
256 | { |
257 | explicit DigitWrapper(int idx, Vector3D_SSE *vec) |
258 | : m_vec(vec) |
259 | , m_idx(idx) |
260 | {} |
261 | |
262 | operator float() const |
263 | { |
264 | switch (m_idx) { |
265 | case 0: |
266 | return m_vec->x(); |
267 | case 1: |
268 | return m_vec->y(); |
269 | case 2: |
270 | return m_vec->z(); |
271 | default: |
272 | Q_UNREACHABLE(); |
273 | return 0.0f; |
274 | } |
275 | } |
276 | |
277 | void operator =(float value) |
278 | { |
279 | switch (m_idx) { |
280 | case 0: |
281 | m_vec->setX(value); |
282 | break; |
283 | case 1: |
284 | m_vec->setY(value); |
285 | break; |
286 | case 2: |
287 | m_vec->setZ(value); |
288 | break; |
289 | default: |
290 | Q_UNREACHABLE(); |
291 | } |
292 | } |
293 | |
294 | private: |
295 | Vector3D_SSE *m_vec; |
296 | const int m_idx; |
297 | }; |
298 | |
299 | Q_ALWAYS_INLINE DigitWrapper operator[](int idx) |
300 | { |
301 | return DigitWrapper(idx, this); |
302 | } |
303 | |
304 | static Q_ALWAYS_INLINE float dotProduct(Vector3D_SSE a, Vector3D_SSE b) |
305 | { |
306 | #if defined(__SSE4_1__) |
307 | // 0b01111111 = 0x7f |
308 | return _mm_cvtss_f32(_mm_dp_ps(a.m_xyzw, b.m_xyzw, 0x7f)); |
309 | #elif defined(__SSE3__) |
310 | const __m128 mult = _mm_mul_ps(a.m_xyzw, b.m_xyzw); |
311 | // a + b, c + d, a + d, c + d |
312 | const __m128 partialSum = _mm_hadd_ps(mult, mult); |
313 | // c + d, ...... |
314 | // 0x00000001 = |
315 | const __m128 partialSumShuffle = _mm_shuffle_ps(partialSum, partialSum, 0x1); |
316 | return _mm_cvtss_f32(_mm_hadd_ps(partialSum, partialSumShuffle)); |
317 | #else |
318 | const __m128 mult = _mm_mul_ps(a: a.m_xyzw, b: b.m_xyzw); |
319 | |
320 | // (multX, 0, 0, 0) + (multY, 0, 0, 0) -> (multX + multY, 0, 0, 0) |
321 | // 0b11111101 == 0xfd |
322 | const __m128 shuffled = _mm_shuffle_ps(mult, mult, 0xfd); |
323 | // (multX + multY, 0, 0, 0) + (multZ, 0, 0, 0); |
324 | // 0b11111110 == 0xfe |
325 | const __m128 shuffled2 = _mm_shuffle_ps(mult, mult, 0xfe); |
326 | const __m128 result = _mm_add_ps(a: _mm_add_ps(a: shuffled, b: mult), b: shuffled2); |
327 | return _mm_cvtss_f32(a: result); |
328 | #endif |
329 | } |
330 | |
331 | static Q_ALWAYS_INLINE Vector3D_SSE crossProduct(Vector3D_SSE a, Vector3D_SSE b) |
332 | { |
333 | // a.y * b.z - a.z * b.y, a.z * b.x - a.x * b.z, a.x * b.y - a.y * b.x |
334 | // (a.y, a.z, a.z, a.x, a.x, a.y) (b.z, b.y, b.x, b.z, b.y, b.x) |
335 | // (a.y, a.z, a.x) * (b.z, b.x, b.y) - (a.z, a.x, a.y) (b.y, b.z, b.x) |
336 | |
337 | // 0b11001001 == 0xc9 |
338 | const __m128 a1 = _mm_shuffle_ps(a.m_xyzw, a.m_xyzw, 0xc9); |
339 | const __m128 b2 = _mm_shuffle_ps(b.m_xyzw, b.m_xyzw, 0xc9); |
340 | // 0b11010010 == 0xd2 |
341 | const __m128 a2 = _mm_shuffle_ps(a.m_xyzw, a.m_xyzw, 0xd2); |
342 | const __m128 b1 = _mm_shuffle_ps(b.m_xyzw, b.m_xyzw, 0xd2); |
343 | |
344 | Vector3D_SSE v(Qt::Uninitialized); |
345 | v.m_xyzw = _mm_sub_ps(a: _mm_mul_ps(a: a1, b: b1), b: _mm_mul_ps(a: a2, b: b2)); |
346 | return v; |
347 | } |
348 | |
349 | friend class Vector4D_SSE; |
350 | |
351 | #if defined(__AVX2__) && defined(QT_COMPILER_SUPPORTS_AVX2) |
352 | friend class Matrix4x4_AVX2; |
353 | friend Vector3D_SSE operator*(const Vector3D_SSE &vector, const Matrix4x4_AVX2 &matrix); |
354 | friend Vector3D_SSE operator*(const Matrix4x4_AVX2 &matrix, const Vector3D_SSE &vector); |
355 | #endif |
356 | |
357 | friend class Matrix4x4_SSE; |
358 | friend Vector3D_SSE operator*(const Vector3D_SSE &vector, const Matrix4x4_SSE &matrix); |
359 | friend Vector3D_SSE operator*(const Matrix4x4_SSE &matrix, const Vector3D_SSE &vector); |
360 | |
361 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator+(Vector3D_SSE v1, Vector3D_SSE v2) { return v1 += v2; } |
362 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator-(Vector3D_SSE v1, Vector3D_SSE v2) { return v1 -= v2; } |
363 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator*(float factor, Vector3D_SSE vector) { return vector *= factor; } |
364 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator*(Vector3D_SSE vector, float factor) { return vector *= factor; } |
365 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator*(Vector3D_SSE v1, Vector3D_SSE v2) { return v1 *= v2; } |
366 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator-(Vector3D_SSE vector) |
367 | { |
368 | Vector3D_SSE c(Qt::Uninitialized); |
369 | |
370 | c.m_xyzw = _mm_xor_ps(a: vector.m_xyzw, b: _mm_set1_ps(w: -0.0f)); |
371 | |
372 | return c; |
373 | } |
374 | |
375 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator/(Vector3D_SSE vector, float divisor) { return vector /= divisor; } |
376 | friend Q_ALWAYS_INLINE const Vector3D_SSE operator/(Vector3D_SSE vector, Vector3D_SSE divisor) { return vector /= divisor; } |
377 | |
378 | friend Q_3DCORE_PRIVATE_EXPORT QDebug operator<<(QDebug dbg, const Vector3D_SSE &v); |
379 | friend Q_ALWAYS_INLINE bool qFuzzyCompare(const Vector3D_SSE& v1, const Vector3D_SSE& v2) |
380 | { |
381 | return ::qFuzzyCompare(p1: v1.x(), p2: v2.x()) && |
382 | ::qFuzzyCompare(p1: v1.y(), p2: v2.y()) && |
383 | ::qFuzzyCompare(p1: v1.z(), p2: v2.z()); |
384 | } |
385 | |
386 | private: |
387 | // Q_DECL_ALIGN(16) float m[4];// for SSE support |
388 | __m128 m_xyzw; |
389 | }; |
390 | |
391 | } // Qt3DCore |
392 | |
393 | Q_DECLARE_TYPEINFO(Qt3DCore::Vector3D_SSE, Q_PRIMITIVE_TYPE); |
394 | |
395 | QT_END_NAMESPACE |
396 | |
397 | Q_DECLARE_METATYPE(Qt3DCore::Vector3D_SSE) |
398 | |
399 | #endif // QT_COMPILER_SUPPORTS_SSE2 |
400 | |
401 | #endif // QT3DCORE_VECTOR3D_SSE_P_H |
402 | |