summaryrefslogtreecommitdiffstats
path: root/depedencies/include/glm/gtx/simd_quat.inl
blob: b84865c28149af02d9c48438121baed5d19d04ca (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
/// @ref gtx_simd_quat
/// @file glm/gtx/simd_quat.inl

namespace glm{
namespace detail{


//////////////////////////////////////
// Debugging
#if 0
void print(__m128 v)
{
    GLM_ALIGN(16) float result[4];
    _mm_store_ps(result, v);

    printf("__m128:    %f %f %f %f\n", result[0], result[1], result[2], result[3]);
}

void print(const fvec4SIMD &v)
{
    printf("fvec4SIMD: %f %f %f %f\n", v.x, v.y, v.z, v.w);
}
#endif

//////////////////////////////////////
// Implicit basic constructors

#	if !GLM_HAS_DEFAULTED_FUNCTIONS || !defined(GLM_FORCE_NO_CTOR_INIT)
	GLM_FUNC_QUALIFIER fquatSIMD::fquatSIMD()
#		ifdef GLM_FORCE_NO_CTOR_INIT
			: Data(_mm_set_ps(1.0f, 0.0f, 0.0f, 0.0f))
#		endif
	{}
#	endif

#	if !GLM_HAS_DEFAULTED_FUNCTIONS
	GLM_FUNC_QUALIFIER fquatSIMD::fquatSIMD(fquatSIMD const & q) :
		Data(q.Data)
	{}
#	endif//!GLM_HAS_DEFAULTED_FUNCTIONS

GLM_FUNC_QUALIFIER fquatSIMD::fquatSIMD(__m128 const & Data) :
	Data(Data)
{}

//////////////////////////////////////
// Explicit basic constructors

GLM_FUNC_QUALIFIER fquatSIMD::fquatSIMD(float const & w, float const & x, float const & y, float const & z) :
	Data(_mm_set_ps(w, z, y, x))
{}

GLM_FUNC_QUALIFIER fquatSIMD::fquatSIMD(quat const & q) :
	Data(_mm_set_ps(q.w, q.z, q.y, q.x))
{}

GLM_FUNC_QUALIFIER fquatSIMD::fquatSIMD(vec3 const & eulerAngles)
{
	vec3 c = glm::cos(eulerAngles * 0.5f);
	vec3 s = glm::sin(eulerAngles * 0.5f);

	Data = _mm_set_ps(
		(c.x * c.y * c.z) + (s.x * s.y * s.z),
		(c.x * c.y * s.z) - (s.x * s.y * c.z),
		(c.x * s.y * c.z) + (s.x * c.y * s.z),
		(s.x * c.y * c.z) - (c.x * s.y * s.z));
}


//////////////////////////////////////
// Unary arithmetic operators

#if !GLM_HAS_DEFAULTED_FUNCTIONS
	GLM_FUNC_QUALIFIER fquatSIMD& fquatSIMD::operator=(fquatSIMD const & q)
	{
		this->Data = q.Data;
		return *this;
	}
#endif//!GLM_HAS_DEFAULTED_FUNCTIONS

GLM_FUNC_QUALIFIER fquatSIMD& fquatSIMD::operator*=(float const & s)
{
	this->Data = _mm_mul_ps(this->Data, _mm_set_ps1(s));
	return *this;
}

GLM_FUNC_QUALIFIER fquatSIMD& fquatSIMD::operator/=(float const & s)
{
	this->Data = _mm_div_ps(Data, _mm_set1_ps(s));
	return *this;
}



// negate operator
GLM_FUNC_QUALIFIER fquatSIMD operator- (fquatSIMD const & q)
{
    return fquatSIMD(_mm_mul_ps(q.Data, _mm_set_ps(-1.0f, -1.0f, -1.0f, -1.0f)));
}

// operator+
GLM_FUNC_QUALIFIER fquatSIMD operator+ (fquatSIMD const & q1, fquatSIMD const & q2)
{
	return fquatSIMD(_mm_add_ps(q1.Data, q2.Data));
}

//operator*
GLM_FUNC_QUALIFIER fquatSIMD operator* (fquatSIMD const & q1, fquatSIMD const & q2)
{
    // SSE2 STATS:
    //    11 shuffle
    //    8  mul
    //    8  add
    
    // SSE4 STATS:
    //    3 shuffle
    //    4 mul
    //    4 dpps

    __m128 mul0 = _mm_mul_ps(q1.Data, _mm_shuffle_ps(q2.Data, q2.Data, _MM_SHUFFLE(0, 1, 2, 3)));
    __m128 mul1 = _mm_mul_ps(q1.Data, _mm_shuffle_ps(q2.Data, q2.Data, _MM_SHUFFLE(1, 0, 3, 2)));
    __m128 mul2 = _mm_mul_ps(q1.Data, _mm_shuffle_ps(q2.Data, q2.Data, _MM_SHUFFLE(2, 3, 0, 1)));
    __m128 mul3 = _mm_mul_ps(q1.Data, q2.Data);

#   if(GLM_ARCH & GLM_ARCH_SSE41_BIT)
    __m128 add0 = _mm_dp_ps(mul0, _mm_set_ps(1.0f, -1.0f,  1.0f,  1.0f), 0xff);
    __m128 add1 = _mm_dp_ps(mul1, _mm_set_ps(1.0f,  1.0f,  1.0f, -1.0f), 0xff);
    __m128 add2 = _mm_dp_ps(mul2, _mm_set_ps(1.0f,  1.0f, -1.0f,  1.0f), 0xff);
    __m128 add3 = _mm_dp_ps(mul3, _mm_set_ps(1.0f, -1.0f, -1.0f, -1.0f), 0xff);
#   else
           mul0 = _mm_mul_ps(mul0, _mm_set_ps(1.0f, -1.0f,  1.0f,  1.0f));
    __m128 add0 = _mm_add_ps(mul0, _mm_movehl_ps(mul0, mul0));
           add0 = _mm_add_ss(add0, _mm_shuffle_ps(add0, add0, 1));

           mul1 = _mm_mul_ps(mul1, _mm_set_ps(1.0f,  1.0f,  1.0f, -1.0f));
    __m128 add1 = _mm_add_ps(mul1, _mm_movehl_ps(mul1, mul1));
           add1 = _mm_add_ss(add1, _mm_shuffle_ps(add1, add1, 1));

           mul2 = _mm_mul_ps(mul2, _mm_set_ps(1.0f,  1.0f, -1.0f,  1.0f));
    __m128 add2 = _mm_add_ps(mul2, _mm_movehl_ps(mul2, mul2));
           add2 = _mm_add_ss(add2, _mm_shuffle_ps(add2, add2, 1));

           mul3 = _mm_mul_ps(mul3, _mm_set_ps(1.0f, -1.0f, -1.0f, -1.0f));
    __m128 add3 = _mm_add_ps(mul3, _mm_movehl_ps(mul3, mul3));
           add3 = _mm_add_ss(add3, _mm_shuffle_ps(add3, add3, 1));
#endif


    // This SIMD code is a politically correct way of doing this, but in every test I've tried it has been slower than
    // the final code below. I'll keep this here for reference - maybe somebody else can do something better...
    //
    //__m128 xxyy = _mm_shuffle_ps(add0, add1, _MM_SHUFFLE(0, 0, 0, 0));
    //__m128 zzww = _mm_shuffle_ps(add2, add3, _MM_SHUFFLE(0, 0, 0, 0));
    //
    //return _mm_shuffle_ps(xxyy, zzww, _MM_SHUFFLE(2, 0, 2, 0));
    
    float x;
    float y;
    float z;
    float w;

    _mm_store_ss(&x, add0);
    _mm_store_ss(&y, add1);
    _mm_store_ss(&z, add2);
    _mm_store_ss(&w, add3);

    return detail::fquatSIMD(w, x, y, z);
}

GLM_FUNC_QUALIFIER fvec4SIMD operator* (fquatSIMD const & q, fvec4SIMD const & v)
{
    static const __m128 two = _mm_set1_ps(2.0f);

    __m128 q_wwww  = _mm_shuffle_ps(q.Data, q.Data, _MM_SHUFFLE(3, 3, 3, 3));
    __m128 q_swp0  = _mm_shuffle_ps(q.Data, q.Data, _MM_SHUFFLE(3, 0, 2, 1));
	__m128 q_swp1  = _mm_shuffle_ps(q.Data, q.Data, _MM_SHUFFLE(3, 1, 0, 2));
	__m128 v_swp0  = _mm_shuffle_ps(v.Data, v.Data, _MM_SHUFFLE(3, 0, 2, 1));
	__m128 v_swp1  = _mm_shuffle_ps(v.Data, v.Data, _MM_SHUFFLE(3, 1, 0, 2));
	
	__m128 uv      = _mm_sub_ps(_mm_mul_ps(q_swp0, v_swp1), _mm_mul_ps(q_swp1, v_swp0));
    __m128 uv_swp0 = _mm_shuffle_ps(uv, uv, _MM_SHUFFLE(3, 0, 2, 1));
    __m128 uv_swp1 = _mm_shuffle_ps(uv, uv, _MM_SHUFFLE(3, 1, 0, 2));
    __m128 uuv     = _mm_sub_ps(_mm_mul_ps(q_swp0, uv_swp1), _mm_mul_ps(q_swp1, uv_swp0));

    
    uv  = _mm_mul_ps(uv,  _mm_mul_ps(q_wwww, two));
    uuv = _mm_mul_ps(uuv, two);

    return _mm_add_ps(v.Data, _mm_add_ps(uv, uuv));
}

GLM_FUNC_QUALIFIER fvec4SIMD operator* (fvec4SIMD const & v, fquatSIMD const & q)
{
	return glm::inverse(q) * v;
}

GLM_FUNC_QUALIFIER fquatSIMD operator* (fquatSIMD const & q, float s)
{
	return fquatSIMD(_mm_mul_ps(q.Data, _mm_set1_ps(s)));
}

GLM_FUNC_QUALIFIER fquatSIMD operator* (float s, fquatSIMD const & q)
{
	return fquatSIMD(_mm_mul_ps(_mm_set1_ps(s), q.Data));
}


//operator/
GLM_FUNC_QUALIFIER fquatSIMD operator/ (fquatSIMD const & q, float s)
{
	return fquatSIMD(_mm_div_ps(q.Data, _mm_set1_ps(s)));
}


}//namespace detail


GLM_FUNC_QUALIFIER quat quat_cast
(
	detail::fquatSIMD const & x
)
{
	GLM_ALIGN(16) quat Result;
	_mm_store_ps(&Result[0], x.Data);

	return Result;
}

template <typename T>
GLM_FUNC_QUALIFIER detail::fquatSIMD quatSIMD_cast_impl(const T m0[], const T m1[], const T m2[])
{
    T trace = m0[0] + m1[1] + m2[2] + T(1.0);
    if (trace > T(0))
    {
        T s = static_cast<T>(0.5) / sqrt(trace);

        return _mm_set_ps(
            static_cast<float>(T(0.25) / s),
            static_cast<float>((m0[1] - m1[0]) * s),
            static_cast<float>((m2[0] - m0[2]) * s),
            static_cast<float>((m1[2] - m2[1]) * s));
    }
    else
    {
        if (m0[0] > m1[1])
        {
            if (m0[0] > m2[2])
            {
                // X is biggest.
                T s = sqrt(m0[0] - m1[1] - m2[2] + T(1.0)) * T(0.5);

                return _mm_set_ps(
                    static_cast<float>((m1[2] - m2[1]) * s),
                    static_cast<float>((m2[0] + m0[2]) * s),
                    static_cast<float>((m0[1] + m1[0]) * s),
                    static_cast<float>(T(0.5)          * s));
            }
        }
        else
        {
            if (m1[1] > m2[2])
            {
                // Y is biggest.
                T s = sqrt(m1[1] - m0[0] - m2[2] + T(1.0)) * T(0.5);

                return _mm_set_ps(
                    static_cast<float>((m2[0] - m0[2]) * s),
                    static_cast<float>((m1[2] + m2[1]) * s),
                    static_cast<float>(T(0.5)          * s),
                    static_cast<float>((m0[1] + m1[0]) * s));
            }
        }

        // Z is biggest.
        T s = sqrt(m2[2] - m0[0] - m1[1] + T(1.0)) * T(0.5);

        return _mm_set_ps(
            static_cast<float>((m0[1] - m1[0]) * s),
            static_cast<float>(T(0.5)          * s),
            static_cast<float>((m1[2] + m2[1]) * s),
            static_cast<float>((m2[0] + m0[2]) * s));
    }
}

GLM_FUNC_QUALIFIER detail::fquatSIMD quatSIMD_cast
(
	detail::fmat4x4SIMD const & m
)
{
    // Scalar implementation for now.
    GLM_ALIGN(16) float m0[4];
    GLM_ALIGN(16) float m1[4];
    GLM_ALIGN(16) float m2[4];

    _mm_store_ps(m0, m[0].Data);
    _mm_store_ps(m1, m[1].Data);
    _mm_store_ps(m2, m[2].Data);

    return quatSIMD_cast_impl(m0, m1, m2);
}

template <typename T, precision P>
GLM_FUNC_QUALIFIER detail::fquatSIMD quatSIMD_cast
(
    tmat4x4<T, P> const & m
)
{
    return quatSIMD_cast_impl(&m[0][0], &m[1][0], &m[2][0]);
}

template <typename T, precision P>
GLM_FUNC_QUALIFIER detail::fquatSIMD quatSIMD_cast
(
    tmat3x3<T, P> const & m
)
{
    return quatSIMD_cast_impl(&m[0][0], &m[1][0], &m[2][0]);
}


GLM_FUNC_QUALIFIER detail::fmat4x4SIMD mat4SIMD_cast
(
	detail::fquatSIMD const & q
)
{
    detail::fmat4x4SIMD result;

    __m128 _wwww  = _mm_shuffle_ps(q.Data, q.Data, _MM_SHUFFLE(3, 3, 3, 3));
    __m128 _xyzw  = q.Data;
    __m128 _zxyw  = _mm_shuffle_ps(q.Data, q.Data, _MM_SHUFFLE(3, 1, 0, 2));
    __m128 _yzxw  = _mm_shuffle_ps(q.Data, q.Data, _MM_SHUFFLE(3, 0, 2, 1));

    __m128 _xyzw2 = _mm_add_ps(_xyzw, _xyzw);
    __m128 _zxyw2 = _mm_shuffle_ps(_xyzw2, _xyzw2, _MM_SHUFFLE(3, 1, 0, 2));
    __m128 _yzxw2 = _mm_shuffle_ps(_xyzw2, _xyzw2, _MM_SHUFFLE(3, 0, 2, 1));
    
    __m128 _tmp0  = _mm_sub_ps(_mm_set1_ps(1.0f), _mm_mul_ps(_yzxw2, _yzxw));
           _tmp0  = _mm_sub_ps(_tmp0, _mm_mul_ps(_zxyw2, _zxyw));

    __m128 _tmp1  = _mm_mul_ps(_yzxw2, _xyzw);
           _tmp1  = _mm_add_ps(_tmp1, _mm_mul_ps(_zxyw2, _wwww));

    __m128 _tmp2  = _mm_mul_ps(_zxyw2, _xyzw);
           _tmp2  = _mm_sub_ps(_tmp2, _mm_mul_ps(_yzxw2, _wwww));


    // There's probably a better, more politically correct way of doing this...
    result[0].Data = _mm_set_ps(
        0.0f,
        reinterpret_cast<float*>(&_tmp2)[0],
        reinterpret_cast<float*>(&_tmp1)[0],
        reinterpret_cast<float*>(&_tmp0)[0]);

    result[1].Data = _mm_set_ps(
        0.0f,
        reinterpret_cast<float*>(&_tmp1)[1],
        reinterpret_cast<float*>(&_tmp0)[1],
        reinterpret_cast<float*>(&_tmp2)[1]);

    result[2].Data = _mm_set_ps(
        0.0f,
        reinterpret_cast<float*>(&_tmp0)[2],
        reinterpret_cast<float*>(&_tmp2)[2],
        reinterpret_cast<float*>(&_tmp1)[2]);

   result[3].Data = _mm_set_ps(
        1.0f,
        0.0f,
        0.0f,
        0.0f);


    return result;
}

GLM_FUNC_QUALIFIER mat4 mat4_cast
(
	detail::fquatSIMD const & q
)
{
    return mat4_cast(mat4SIMD_cast(q));
}



GLM_FUNC_QUALIFIER float length
(
	detail::fquatSIMD const & q
)
{
    return glm::sqrt(dot(q, q));
}

GLM_FUNC_QUALIFIER detail::fquatSIMD normalize
(
	detail::fquatSIMD const & q
)
{
    return _mm_mul_ps(q.Data, _mm_set1_ps(1.0f / length(q)));
}

GLM_FUNC_QUALIFIER float dot
(
	detail::fquatSIMD const & q1,
	detail::fquatSIMD const & q2
)
{
    float result;
    _mm_store_ss(&result, detail::sse_dot_ps(q1.Data, q2.Data));

    return result;
}

GLM_FUNC_QUALIFIER detail::fquatSIMD mix
(
	detail::fquatSIMD const & x, 
	detail::fquatSIMD const & y, 
	float const & a
)
{
	float cosTheta = dot(x, y);

    if (cosTheta > 1.0f - glm::epsilon<float>())
    {
	    return _mm_add_ps(x.Data, _mm_mul_ps(_mm_set1_ps(a), _mm_sub_ps(y.Data, x.Data)));
    }
    else
    {
        float angle = glm::acos(cosTheta);
        
        
        float s0 = glm::sin((1.0f - a) * angle);
        float s1 = glm::sin(a * angle);
        float d  = 1.0f / glm::sin(angle);

        return (s0 * x + s1 * y) * d;
    }
}

GLM_FUNC_QUALIFIER detail::fquatSIMD lerp
(
	detail::fquatSIMD const & x, 
	detail::fquatSIMD const & y, 
	float const & a
)
{
	// Lerp is only defined in [0, 1]
	assert(a >= 0.0f);
	assert(a <= 1.0f);

    return _mm_add_ps(x.Data, _mm_mul_ps(_mm_set1_ps(a), _mm_sub_ps(y.Data, x.Data)));
}

GLM_FUNC_QUALIFIER detail::fquatSIMD slerp
(
	detail::fquatSIMD const & x, 
	detail::fquatSIMD const & y, 
	float const & a
)
{
	detail::fquatSIMD z = y;

	float cosTheta = dot(x, y);

	// If cosTheta < 0, the interpolation will take the long way around the sphere. 
	// To fix this, one quat must be negated.
	if (cosTheta < 0.0f)
	{
		z        = -y;
		cosTheta = -cosTheta;
	}

	// Perform a linear interpolation when cosTheta is close to 1 to avoid side effect of sin(angle) becoming a zero denominator
	if(cosTheta > 1.0f - epsilon<float>())
	{
		return _mm_add_ps(x.Data, _mm_mul_ps(_mm_set1_ps(a), _mm_sub_ps(y.Data, x.Data)));
	}
	else
	{
        float angle = glm::acos(cosTheta);


		float s0 = glm::sin((1.0f - a) * angle);
        float s1 = glm::sin(a * angle);
        float d  = 1.0f / glm::sin(angle);

        return (s0 * x + s1 * y) * d;
	}
}


GLM_FUNC_QUALIFIER detail::fquatSIMD fastMix
(
	detail::fquatSIMD const & x, 
	detail::fquatSIMD const & y, 
	float const & a
)
{
	float cosTheta = dot(x, y);

    if (cosTheta > 1.0f - glm::epsilon<float>())
    {
	    return _mm_add_ps(x.Data, _mm_mul_ps(_mm_set1_ps(a), _mm_sub_ps(y.Data, x.Data)));
    }
    else
    {
        float angle = glm::fastAcos(cosTheta);


        __m128 s  = glm::fastSin(_mm_set_ps((1.0f - a) * angle, a * angle, angle, 0.0f));

        __m128 s0 =                               _mm_shuffle_ps(s, s, _MM_SHUFFLE(3, 3, 3, 3));
        __m128 s1 =                               _mm_shuffle_ps(s, s, _MM_SHUFFLE(2, 2, 2, 2));
        __m128 d  = _mm_div_ps(_mm_set1_ps(1.0f), _mm_shuffle_ps(s, s, _MM_SHUFFLE(1, 1, 1, 1)));
        
        return _mm_mul_ps(_mm_add_ps(_mm_mul_ps(s0, x.Data), _mm_mul_ps(s1, y.Data)), d);
    }
}

GLM_FUNC_QUALIFIER detail::fquatSIMD fastSlerp
(
	detail::fquatSIMD const & x, 
	detail::fquatSIMD const & y, 
	float const & a
)
{
	detail::fquatSIMD z = y;

	float cosTheta = dot(x, y);
	if (cosTheta < 0.0f)
	{
		z        = -y;
		cosTheta = -cosTheta;
	}


	if(cosTheta > 1.0f - epsilon<float>())
	{
		return _mm_add_ps(x.Data, _mm_mul_ps(_mm_set1_ps(a), _mm_sub_ps(y.Data, x.Data)));
	}
	else
	{
        float angle = glm::fastAcos(cosTheta);


        __m128 s  = glm::fastSin(_mm_set_ps((1.0f - a) * angle, a * angle, angle, 0.0f));

        __m128 s0 =                               _mm_shuffle_ps(s, s, _MM_SHUFFLE(3, 3, 3, 3));
        __m128 s1 =                               _mm_shuffle_ps(s, s, _MM_SHUFFLE(2, 2, 2, 2));
        __m128 d  = _mm_div_ps(_mm_set1_ps(1.0f), _mm_shuffle_ps(s, s, _MM_SHUFFLE(1, 1, 1, 1)));
        
        return _mm_mul_ps(_mm_add_ps(_mm_mul_ps(s0, x.Data), _mm_mul_ps(s1, y.Data)), d);
	}
}



GLM_FUNC_QUALIFIER detail::fquatSIMD conjugate
(
	detail::fquatSIMD const & q
)
{
	return detail::fquatSIMD(_mm_mul_ps(q.Data, _mm_set_ps(1.0f, -1.0f, -1.0f, -1.0f)));
}

GLM_FUNC_QUALIFIER detail::fquatSIMD inverse
(
	detail::fquatSIMD const & q
)
{
	return conjugate(q) / dot(q, q);
}


GLM_FUNC_QUALIFIER detail::fquatSIMD angleAxisSIMD
(
	float const & angle,
	vec3 const & v
)
{
	float s = glm::sin(angle * 0.5f);

	return _mm_set_ps(
		glm::cos(angle * 0.5f),
		v.z * s,
		v.y * s,
		v.x * s);
}

GLM_FUNC_QUALIFIER detail::fquatSIMD angleAxisSIMD
(
	float const & angle, 
	float const & x, 
	float const & y, 
	float const & z
)
{
	return angleAxisSIMD(angle, vec3(x, y, z));
}


GLM_FUNC_QUALIFIER __m128 fastSin(__m128 x)
{
	static const __m128 c0 = _mm_set1_ps(0.16666666666666666666666666666667f);
	static const __m128 c1 = _mm_set1_ps(0.00833333333333333333333333333333f);
	static const __m128 c2 = _mm_set1_ps(0.00019841269841269841269841269841f);

	__m128 x3 = _mm_mul_ps(x,  _mm_mul_ps(x, x));
	__m128 x5 = _mm_mul_ps(x3, _mm_mul_ps(x, x));
	__m128 x7 = _mm_mul_ps(x5, _mm_mul_ps(x, x));

	__m128 y0 = _mm_mul_ps(x3, c0);
	__m128 y1 = _mm_mul_ps(x5, c1);
	__m128 y2 = _mm_mul_ps(x7, c2);

	return _mm_sub_ps(_mm_add_ps(_mm_sub_ps(x, y0), y1), y2);
}


}//namespace glm