OLD | NEW |
1 /* | 1 /* |
2 * Copyright 2015 Google Inc. | 2 * Copyright 2015 Google Inc. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license that can be | 4 * Use of this source code is governed by a BSD-style license that can be |
5 * found in the LICENSE file. | 5 * found in the LICENSE file. |
6 */ | 6 */ |
7 | 7 |
8 // It is important _not_ to put header guards here. | 8 // It is important _not_ to put header guards here. |
9 // This file will be intentionally included three times. | 9 // This file will be intentionally included three times. |
10 | 10 |
(...skipping 20 matching lines...) Expand all Loading... |
31 M(Sk2f&) operator=(const Sk2f& o) { fVec = o.fVec; return *this; } | 31 M(Sk2f&) operator=(const Sk2f& o) { fVec = o.fVec; return *this; } |
32 | 32 |
33 M(Sk2f) Load(const float vals[2]) { | 33 M(Sk2f) Load(const float vals[2]) { |
34 return _mm_castsi128_ps(_mm_loadl_epi64((const __m128i*)vals)); | 34 return _mm_castsi128_ps(_mm_loadl_epi64((const __m128i*)vals)); |
35 } | 35 } |
36 M(void) store(float vals[2]) const { _mm_storel_pi((__m64*)vals, fVec); } | 36 M(void) store(float vals[2]) const { _mm_storel_pi((__m64*)vals, fVec); } |
37 | 37 |
38 M(Sk2f) add(const Sk2f& o) const { return _mm_add_ps(fVec, o.fVec); } | 38 M(Sk2f) add(const Sk2f& o) const { return _mm_add_ps(fVec, o.fVec); } |
39 M(Sk2f) subtract(const Sk2f& o) const { return _mm_sub_ps(fVec, o.fVec); } | 39 M(Sk2f) subtract(const Sk2f& o) const { return _mm_sub_ps(fVec, o.fVec); } |
40 M(Sk2f) multiply(const Sk2f& o) const { return _mm_mul_ps(fVec, o.fVec); } | 40 M(Sk2f) multiply(const Sk2f& o) const { return _mm_mul_ps(fVec, o.fVec); } |
| 41 M(Sk2f) divide(const Sk2f& o) const { return _mm_div_ps(fVec, o.fVec); } |
41 | 42 |
42 M(Sk2f) Min(const Sk2f& a, const Sk2f& b) { return _mm_min_ps(a.fVec, b.fVec); } | 43 M(Sk2f) Min(const Sk2f& a, const Sk2f& b) { return _mm_min_ps(a.fVec, b.fVec); } |
43 M(Sk2f) Max(const Sk2f& a, const Sk2f& b) { return _mm_max_ps(a.fVec, b.fVec); } | 44 M(Sk2f) Max(const Sk2f& a, const Sk2f& b) { return _mm_max_ps(a.fVec, b.fVec); } |
44 | 45 |
45 M(Sk2f) rsqrt() const { return _mm_rsqrt_ps(fVec); } | 46 M(Sk2f) rsqrt() const { return _mm_rsqrt_ps(fVec); } |
46 M(Sk2f) sqrt() const { return _mm_sqrt_ps (fVec); } | 47 M(Sk2f) sqrt() const { return _mm_sqrt_ps (fVec); } |
47 | 48 |
48 #undef M | 49 #undef M |
49 | 50 |
50 #define M(...) template <> inline __VA_ARGS__ Sk2x<double>:: | 51 #define M(...) template <> inline __VA_ARGS__ Sk2x<double>:: |
51 | 52 |
52 M() Sk2x() {} | 53 M() Sk2x() {} |
53 M() Sk2x(double val) { fVec = _mm_set1_pd(val); } | 54 M() Sk2x(double val) { fVec = _mm_set1_pd(val); } |
54 M() Sk2x(double a, double b) { fVec = _mm_set_pd(b, a); } | 55 M() Sk2x(double a, double b) { fVec = _mm_set_pd(b, a); } |
55 M(Sk2d&) operator=(const Sk2d& o) { fVec = o.fVec; return *this; } | 56 M(Sk2d&) operator=(const Sk2d& o) { fVec = o.fVec; return *this; } |
56 | 57 |
57 M(Sk2d) Load(const double vals[2]) { return _mm_loadu_pd(vals); } | 58 M(Sk2d) Load(const double vals[2]) { return _mm_loadu_pd(vals); } |
58 M(void) store(double vals[2]) const { _mm_storeu_pd(vals, fVec); } | 59 M(void) store(double vals[2]) const { _mm_storeu_pd(vals, fVec); } |
59 | 60 |
60 M(Sk2d) add(const Sk2d& o) const { return _mm_add_pd(fVec, o.fVec); } | 61 M(Sk2d) add(const Sk2d& o) const { return _mm_add_pd(fVec, o.fVec); } |
61 M(Sk2d) subtract(const Sk2d& o) const { return _mm_sub_pd(fVec, o.fVec); } | 62 M(Sk2d) subtract(const Sk2d& o) const { return _mm_sub_pd(fVec, o.fVec); } |
62 M(Sk2d) multiply(const Sk2d& o) const { return _mm_mul_pd(fVec, o.fVec); } | 63 M(Sk2d) multiply(const Sk2d& o) const { return _mm_mul_pd(fVec, o.fVec); } |
| 64 M(Sk2d) divide(const Sk2d& o) const { return _mm_div_pd(fVec, o.fVec); } |
63 | 65 |
64 M(Sk2d) Min(const Sk2d& a, const Sk2d& b) { return _mm_min_pd(a.fVec, b.fVec); } | 66 M(Sk2d) Min(const Sk2d& a, const Sk2d& b) { return _mm_min_pd(a.fVec, b.fVec); } |
65 M(Sk2d) Max(const Sk2d& a, const Sk2d& b) { return _mm_max_pd(a.fVec, b.fVec); } | 67 M(Sk2d) Max(const Sk2d& a, const Sk2d& b) { return _mm_max_pd(a.fVec, b.fVec); } |
66 | 68 |
67 // There is no _mm_rsqrt_pd, so we do Sk2d::rsqrt() in floats. | 69 // There is no _mm_rsqrt_pd, so we do Sk2d::rsqrt() in floats. |
68 M(Sk2d) rsqrt() const { return _mm_cvtps_pd(_mm_rsqrt_ps(_mm_cvtpd_ps(fVec))); } | 70 M(Sk2d) rsqrt() const { return _mm_cvtps_pd(_mm_rsqrt_ps(_mm_cvtpd_ps(fVec))); } |
69 M(Sk2d) sqrt() const { return _mm_sqrt_pd(fVec); } | 71 M(Sk2d) sqrt() const { return _mm_sqrt_pd(fVec); } |
70 | 72 |
71 #undef M | 73 #undef M |
72 | 74 |
73 #endif | 75 #endif |
OLD | NEW |