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_SUB_H |
9 | #define LIBSIMDPP_SIMDPP_DETAIL_INSN_F_SUB_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_fsub(const float32<4>& a, const float32<4>& b) |
26 | { |
27 | #if SIMDPP_USE_NULL || SIMDPP_USE_NEON_NO_FLT_SP |
28 | return detail::null::sub(a, b); |
29 | #elif SIMDPP_USE_SSE2 |
30 | return _mm_sub_ps(a.native(), b.native()); |
31 | #elif SIMDPP_USE_NEON_FLT_SP |
32 | return vsubq_f32(a.native(), b.native()); |
33 | #elif SIMDPP_USE_ALTIVEC |
34 | return vec_sub(a.native(), b.native()); |
35 | #elif SIMDPP_USE_MSA |
36 | return __msa_fsub_w(a.native(), b.native()); |
37 | #endif |
38 | } |
39 | |
40 | #if SIMDPP_USE_AVX |
41 | static SIMDPP_INL |
42 | float32<8> i_fsub(const float32<8>& a, const float32<8>& b) |
43 | { |
44 | return _mm256_sub_ps(a.native(), b.native()); |
45 | } |
46 | #endif |
47 | |
48 | #if SIMDPP_USE_AVX512F |
49 | static SIMDPP_INL |
50 | float32<16> i_fsub(const float32<16>& a, const float32<16>& b) |
51 | { |
52 | return _mm512_sub_ps(a.native(), b.native()); |
53 | } |
54 | #endif |
55 | |
56 | template<class R, unsigned N, class E1, class E2> SIMDPP_INL |
57 | float32<N> i_fsub(const float32<N>& a, const float32<N>& b) |
58 | { |
59 | SIMDPP_VEC_ARRAY_IMPL2(float32<N>, sub, a, b); |
60 | } |
61 | |
62 | // ----------------------------------------------------------------------------- |
63 | |
64 | static SIMDPP_INL |
65 | float64<2> i_fsub(const float64<2>& a, const float64<2>& b) |
66 | { |
67 | #if SIMDPP_USE_SSE2 |
68 | return _mm_sub_pd(a.native(), b.native()); |
69 | #elif SIMDPP_USE_NEON64 |
70 | return vsubq_f64(a.native(), b.native()); |
71 | #elif SIMDPP_USE_VSX_206 |
72 | return vec_sub(a.native(), b.native()); |
73 | #elif SIMDPP_USE_NULL || SIMDPP_USE_NEON || SIMDPP_USE_ALTIVEC |
74 | return detail::null::sub(a, b); |
75 | #elif SIMDPP_USE_MSA |
76 | return __msa_fsub_d(a.native(), b.native()); |
77 | #endif |
78 | } |
79 | |
80 | #if SIMDPP_USE_AVX |
81 | static SIMDPP_INL |
82 | float64<4> i_fsub(const float64<4>& a, const float64<4>& b) |
83 | { |
84 | return _mm256_sub_pd(a.native(), b.native()); |
85 | } |
86 | #endif |
87 | |
88 | #if SIMDPP_USE_AVX512F |
89 | static SIMDPP_INL |
90 | float64<8> i_fsub(const float64<8>& a, const float64<8>& b) |
91 | { |
92 | return _mm512_sub_pd(a.native(), b.native()); |
93 | } |
94 | #endif |
95 | |
96 | // ----------------------------------------------------------------------------- |
97 | |
98 | template<class V> SIMDPP_INL |
99 | V i_fsub(const V& a, const V& b) |
100 | { |
101 | SIMDPP_VEC_ARRAY_IMPL2(V, i_fsub, a, b) |
102 | } |
103 | |
104 | } // namespace insn |
105 | } // namespace detail |
106 | } // namespace SIMDPP_ARCH_NAMESPACE |
107 | } // namespace simdpp |
108 | |
109 | #endif |
110 | |
111 | |