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_I_ADD_SAT_H |
9 | #define LIBSIMDPP_SIMDPP_DETAIL_INSN_I_ADD_SAT_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/expr.h> |
17 | #include <simdpp/detail/null/math.h> |
18 | #include <simdpp/detail/vector_array_macros.h> |
19 | |
20 | namespace simdpp { |
21 | namespace SIMDPP_ARCH_NAMESPACE { |
22 | namespace detail { |
23 | namespace insn { |
24 | |
25 | static SIMDPP_INL |
26 | int8<16> i_add_sat(const int8<16>& a, const int8<16>& b) |
27 | { |
28 | #if SIMDPP_USE_NULL |
29 | return detail::null::add_sat(a, b); |
30 | #elif SIMDPP_USE_SSE2 |
31 | return _mm_adds_epi8(a.native(), b.native()); |
32 | #elif SIMDPP_USE_NEON |
33 | return vqaddq_s8(a.native(), b.native()); |
34 | #elif SIMDPP_USE_ALTIVEC |
35 | return vec_adds(a.native(), b.native()); |
36 | #elif SIMDPP_USE_MSA |
37 | return __msa_adds_s_b(a.native(), b.native()); |
38 | #endif |
39 | } |
40 | |
41 | #if SIMDPP_USE_AVX2 |
42 | static SIMDPP_INL |
43 | int8<32> i_add_sat(const int8<32>& a, const int8<32>& b) |
44 | { |
45 | return _mm256_adds_epi8(a.native(), b.native()); |
46 | } |
47 | #endif |
48 | |
49 | #if SIMDPP_USE_AVX512BW |
50 | static SIMDPP_INL |
51 | int8<64> i_add_sat(const int8<64>& a, const int8<64>& b) |
52 | { |
53 | return _mm512_adds_epi8(a.native(), b.native()); |
54 | } |
55 | #endif |
56 | |
57 | // ----------------------------------------------------------------------------- |
58 | |
59 | static SIMDPP_INL |
60 | int16<8> i_add_sat(const int16<8>& a, const int16<8>& b) |
61 | { |
62 | #if SIMDPP_USE_NULL |
63 | return detail::null::add_sat(a, b); |
64 | #elif SIMDPP_USE_SSE2 |
65 | return _mm_adds_epi16(a.native(), b.native()); |
66 | #elif SIMDPP_USE_NEON |
67 | return vqaddq_s16(a.native(), b.native()); |
68 | #elif SIMDPP_USE_ALTIVEC |
69 | return vec_adds(a.native(), b.native()); |
70 | #elif SIMDPP_USE_MSA |
71 | return __msa_adds_s_h(a.native(), b.native()); |
72 | #endif |
73 | } |
74 | |
75 | #if SIMDPP_USE_AVX2 |
76 | static SIMDPP_INL |
77 | int16<16> i_add_sat(const int16<16>& a, const int16<16>& b) |
78 | { |
79 | return _mm256_adds_epi16(a.native(), b.native()); |
80 | } |
81 | #endif |
82 | |
83 | #if SIMDPP_USE_AVX512BW |
84 | static SIMDPP_INL |
85 | int16<32> i_add_sat(const int16<32>& a, const int16<32>& b) |
86 | { |
87 | return _mm512_adds_epi16(a.native(), b.native()); |
88 | } |
89 | #endif |
90 | |
91 | // ----------------------------------------------------------------------------- |
92 | |
93 | static SIMDPP_INL |
94 | uint8<16> i_add_sat(const uint8<16>& a, const uint8<16>& b) |
95 | { |
96 | #if SIMDPP_USE_NULL |
97 | return detail::null::add_sat(a, b); |
98 | #elif SIMDPP_USE_SSE2 |
99 | return _mm_adds_epu8(a.native(), b.native()); |
100 | #elif SIMDPP_USE_NEON |
101 | return vqaddq_u8(a.native(), b.native()); |
102 | #elif SIMDPP_USE_ALTIVEC |
103 | return vec_adds(a.native(), b.native()); |
104 | #elif SIMDPP_USE_MSA |
105 | return __msa_adds_u_b(a.native(), b.native()); |
106 | #endif |
107 | } |
108 | |
109 | #if SIMDPP_USE_AVX2 |
110 | static SIMDPP_INL |
111 | uint8<32> i_add_sat(const uint8<32>& a, const uint8<32>& b) |
112 | { |
113 | return _mm256_adds_epu8(a.native(), b.native()); |
114 | } |
115 | #endif |
116 | |
117 | #if SIMDPP_USE_AVX512BW |
118 | static SIMDPP_INL |
119 | uint8<64> i_add_sat(const uint8<64>& a, const uint8<64>& b) |
120 | { |
121 | return _mm512_adds_epu8(a.native(), b.native()); |
122 | } |
123 | #endif |
124 | |
125 | // ----------------------------------------------------------------------------- |
126 | |
127 | static SIMDPP_INL |
128 | uint16<8> i_add_sat(const uint16<8>& a, const uint16<8>& b) |
129 | { |
130 | #if SIMDPP_USE_NULL |
131 | return detail::null::add_sat(a, b); |
132 | #elif SIMDPP_USE_SSE2 |
133 | return _mm_adds_epu16(a.native(), b.native()); |
134 | #elif SIMDPP_USE_NEON |
135 | return vqaddq_u16(a.native(), b.native()); |
136 | #elif SIMDPP_USE_ALTIVEC |
137 | return vec_adds(a.native(), b.native()); |
138 | #elif SIMDPP_USE_MSA |
139 | return __msa_adds_u_h(a.native(), b.native()); |
140 | #endif |
141 | } |
142 | |
143 | #if SIMDPP_USE_AVX2 |
144 | static SIMDPP_INL |
145 | uint16<16> i_add_sat(const uint16<16>& a, const uint16<16>& b) |
146 | { |
147 | return _mm256_adds_epu16(a.native(), b.native()); |
148 | } |
149 | #endif |
150 | |
151 | #if SIMDPP_USE_AVX512BW |
152 | static SIMDPP_INL |
153 | uint16<32> i_add_sat(const uint16<32>& a, const uint16<32>& b) |
154 | { |
155 | return _mm512_adds_epu16(a.native(), b.native()); |
156 | } |
157 | #endif |
158 | |
159 | template<class R, unsigned N, class E1, class E2> SIMDPP_INL |
160 | uint16<N> i_add_sat(const uint16<N>& a, const uint16<N>& b) |
161 | { |
162 | SIMDPP_VEC_ARRAY_IMPL2(uint16<N>, add_sat, a, b); |
163 | } |
164 | |
165 | // ----------------------------------------------------------------------------- |
166 | |
167 | template<class V> SIMDPP_INL |
168 | V i_add_sat(const V& a, const V& b) |
169 | { |
170 | SIMDPP_VEC_ARRAY_IMPL2(V, i_add_sat, a, b) |
171 | } |
172 | |
173 | } // namespace insn |
174 | } // namespace detail |
175 | } // namespace SIMDPP_ARCH_NAMESPACE |
176 | } // namespace simdpp |
177 | |
178 | #endif |
179 | |
180 | |