1 | /* Copyright (C) 2013-2017 Povilas Kanapickas <povilas@radix.lt> |
2 | |
3 | Distributed under the Boost Software License, Version 1.0. |
4 | (See accompanying file LICENSE_1_0.txt or copy at |
5 | http://www.boost.org/LICENSE_1_0.txt) |
6 | */ |
7 | |
8 | #ifndef LIBSIMDPP_SIMDPP_DETAIL_INSN_F_ADD_H |
9 | #define LIBSIMDPP_SIMDPP_DETAIL_INSN_F_ADD_H |
10 | |
11 | #ifndef LIBSIMDPP_SIMD_H |
12 | #error "This file must be included through simd.h" |
13 | #endif |
14 | |
15 | #include <simdpp/types.h> |
16 | #include <simdpp/detail/null/math.h> |
17 | #include <simdpp/detail/vector_array_macros.h> |
18 | |
19 | namespace simdpp { |
20 | namespace SIMDPP_ARCH_NAMESPACE { |
21 | namespace detail { |
22 | namespace insn { |
23 | |
24 | static SIMDPP_INL |
25 | float32<4> i_fadd(const float32<4>& a, const float32<4>& b) |
26 | { |
27 | #if SIMDPP_USE_NULL || SIMDPP_USE_NEON_NO_FLT_SP |
28 | return detail::null::add(a, b); |
29 | #elif SIMDPP_USE_SSE2 |
30 | return _mm_add_ps(a.native(), b.native()); |
31 | #elif SIMDPP_USE_NEON_FLT_SP |
32 | return vaddq_f32(a.native(), b.native()); |
33 | #elif SIMDPP_USE_ALTIVEC |
34 | return vec_add(a.native(), b.native()); |
35 | #elif SIMDPP_USE_MSA |
36 | return __msa_fadd_w(a.native(), b.native()); |
37 | #endif |
38 | } |
39 | |
40 | #if SIMDPP_USE_AVX |
41 | static SIMDPP_INL |
42 | float32<8> i_fadd(const float32<8>& a, const float32<8>& b) |
43 | { |
44 | return _mm256_add_ps(a.native(), b.native()); |
45 | } |
46 | #endif |
47 | |
48 | #if SIMDPP_USE_AVX512F |
49 | static SIMDPP_INL |
50 | float32<16> i_fadd(const float32<16>& a, const float32<16>& b) |
51 | { |
52 | return _mm512_add_ps(a.native(), b.native()); |
53 | } |
54 | #endif |
55 | |
56 | // ----------------------------------------------------------------------------- |
57 | |
58 | static SIMDPP_INL |
59 | float64<2> i_fadd(const float64<2>& a, const float64<2>& b) |
60 | { |
61 | #if SIMDPP_USE_SSE2 |
62 | return _mm_add_pd(a.native(), b.native()); |
63 | #elif SIMDPP_USE_NEON64 |
64 | return vaddq_f64(a.native(), b.native()); |
65 | #elif SIMDPP_USE_VSX_206 |
66 | return vec_add(a.native(), b.native()); |
67 | #elif SIMDPP_USE_MSA |
68 | return __msa_fadd_d(a.native(), b.native()); |
69 | #elif SIMDPP_USE_NULL || SIMDPP_USE_NEON32 || SIMDPP_USE_ALTIVEC |
70 | return detail::null::add(a, b); |
71 | #endif |
72 | } |
73 | |
74 | #if SIMDPP_USE_AVX |
75 | static SIMDPP_INL |
76 | float64<4> i_fadd(const float64<4>& a, const float64<4>& b) |
77 | { |
78 | return _mm256_add_pd(a.native(), b.native()); |
79 | } |
80 | #endif |
81 | |
82 | #if SIMDPP_USE_AVX512F |
83 | static SIMDPP_INL |
84 | float64<8> i_fadd(const float64<8>& a, const float64<8>& b) |
85 | { |
86 | return _mm512_add_pd(a.native(), b.native()); |
87 | } |
88 | #endif |
89 | |
90 | // ----------------------------------------------------------------------------- |
91 | |
92 | template<class V> SIMDPP_INL |
93 | V i_fadd(const V& a, const V& b) |
94 | { |
95 | SIMDPP_VEC_ARRAY_IMPL2(V, i_fadd, a, b) |
96 | } |
97 | |
98 | } // namespace insn |
99 | } // namespace detail |
100 | } // namespace SIMDPP_ARCH_NAMESPACE |
101 | } // namespace simdpp |
102 | |
103 | #endif |
104 | |
105 | |