1 | /* Copyright (C) 2013-2014 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_CORE_CMP_GT_H |
9 | #define LIBSIMDPP_SIMDPP_CORE_CMP_GT_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/insn/cmp_gt.h> |
17 | #include <simdpp/core/detail/scalar_arg_impl.h> |
18 | |
19 | namespace simdpp { |
20 | namespace SIMDPP_ARCH_NAMESPACE { |
21 | |
22 | /** Compares the values of two signed int16x8 vectors for greater-than |
23 | |
24 | @code |
25 | r0 = (a0 > b0) ? 0xff : 0x0 |
26 | ... |
27 | rN = (aN > bN) ? 0xff : 0x0 |
28 | @endcode |
29 | |
30 | @par 256-bit version: |
31 | @icost{SSE2-AVX, NEON, ALTIVEC, 2} |
32 | */ |
33 | template<unsigned N, class E1, class E2> SIMDPP_INL |
34 | mask_int8<N,expr_empty> cmp_gt(const int8<N,E1>& a, |
35 | const int8<N,E2>& b) |
36 | { |
37 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
38 | } |
39 | |
40 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int8, int8) |
41 | |
42 | |
43 | /** Compares the values of two unsigned int16x8 vectors for greater-than |
44 | |
45 | @code |
46 | r0 = (a0 > b0) ? 0xff : 0x0 |
47 | ... |
48 | rN = (aN > bN) ? 0xff : 0x0 |
49 | @endcode |
50 | |
51 | @par 128-bit version: |
52 | @icost{SSE2-AVX2, 3-4} |
53 | @icost{XOP, 1} |
54 | |
55 | @par 256-bit version: |
56 | @icost{SSE2-AVX, 6-7} |
57 | @icost{AVX2, 3-4} |
58 | @icost{XOP, 2} |
59 | @icost{NEON, ALTIVEC, 2} |
60 | */ |
61 | template<unsigned N, class E1, class E2> SIMDPP_INL |
62 | mask_int8<N,expr_empty> cmp_gt(const uint8<N,E1>& a, |
63 | const uint8<N,E2>& b) |
64 | { |
65 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
66 | } |
67 | |
68 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int8, uint8) |
69 | |
70 | /** Compares the values of two signed int16x8 vectors for greater-than |
71 | |
72 | @code |
73 | r0 = (a0 > b0) ? 0xffff : 0x0 |
74 | ... |
75 | rN = (aN > bN) ? 0xffff : 0x0 |
76 | @endcode |
77 | |
78 | @par 256-bit version: |
79 | @icost{SSE2-AVX, NEON, ALTIVEC, 2} |
80 | */ |
81 | template<unsigned N, class E1, class E2> SIMDPP_INL |
82 | mask_int16<N,expr_empty> cmp_gt(const int16<N,E1>& a, |
83 | const int16<N,E2>& b) |
84 | { |
85 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
86 | } |
87 | |
88 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int16, int16) |
89 | |
90 | /** Compares the values of two unsigned int16x8 vectors for greater-than |
91 | |
92 | @code |
93 | r0 = (a0 > b0) ? 0xffff : 0x0 |
94 | ... |
95 | rN = (aN > bN) ? 0xffff : 0x0 |
96 | @endcode |
97 | |
98 | @par 128-bit version: |
99 | @icost{SSE2-AVX2, 3-4} |
100 | @icost{XOP, 1} |
101 | |
102 | @par 256-bit version: |
103 | @icost{SSE2-AVX, 6-7} |
104 | @icost{AVX2, 3-4} |
105 | @icost{XOP, NEON, ALTIVEC, 2} |
106 | */ |
107 | template<unsigned N, class E1, class E2> SIMDPP_INL |
108 | mask_int16<N,expr_empty> cmp_gt(const uint16<N,E1>& a, |
109 | const uint16<N,E2>& b) |
110 | { |
111 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
112 | } |
113 | |
114 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int16, uint16) |
115 | |
116 | /** Compares the values of two signed int32x4 vectors for greater-than |
117 | |
118 | @code |
119 | r0 = (a0 > b0) ? 0xffffffff : 0x0 |
120 | ... |
121 | rN = (aN > bN) ? 0xffffffff : 0x0 |
122 | @endcode |
123 | |
124 | @par 256-bit version: |
125 | @icost{SSE2-AVX, NEON, ALTIVEC, 2} |
126 | */ |
127 | template<unsigned N, class E1, class E2> SIMDPP_INL |
128 | mask_int32<N,expr_empty> cmp_gt(const int32<N,E1>& a, |
129 | const int32<N,E2>& b) |
130 | { |
131 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
132 | } |
133 | |
134 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int32, int32) |
135 | |
136 | /** Compares the values of two unsigned int32x4 vectors for greater-than |
137 | |
138 | @code |
139 | r0 = (a0 > b0) ? 0xffffffff : 0x0 |
140 | ... |
141 | rN = (aN > bN) ? 0xffffffff : 0x0 |
142 | @endcode |
143 | |
144 | @par 128-bit version: |
145 | @icost{SSE2-AVX2, 3-4} |
146 | @icost{XOP, 1} |
147 | |
148 | @par 256-bit version: |
149 | @icost{SSE2-AVX, 6-7} |
150 | @icost{AVX2, 3-4} |
151 | @icost{XOP, NEON, ALTIVEC, 2} |
152 | */ |
153 | template<unsigned N, class E1, class E2> SIMDPP_INL |
154 | mask_int32<N,expr_empty> cmp_gt(const uint32<N,E1>& a, |
155 | const uint32<N,E2>& b) |
156 | { |
157 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
158 | } |
159 | |
160 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int32, uint32) |
161 | |
162 | /** Compares the values of two signed int64 vectors for greater-than |
163 | |
164 | @code |
165 | r0 = (a0 > b0) ? 0xffffffffffff : 0x0 |
166 | ... |
167 | rN = (aN > bN) ? 0xffffffffffff : 0x0 |
168 | @endcode |
169 | |
170 | Supported since AVX2, NEON64. Not supported on ALTIVEC. |
171 | */ |
172 | template<unsigned N, class E1, class E2> SIMDPP_INL |
173 | mask_int64<N,expr_empty> cmp_gt(const int64<N,E1>& a, |
174 | const int64<N,E2>& b) |
175 | { |
176 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
177 | } |
178 | |
179 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int64, int64) |
180 | |
181 | /** Compares the values of two unsigned int64 vectors for greater-than |
182 | |
183 | @code |
184 | r0 = (a0 > b0) ? 0xffffffffffff : 0x0 |
185 | ... |
186 | rN = (aN > bN) ? 0xffffffffffff : 0x0 |
187 | @endcode |
188 | |
189 | Supported since AVX2, NEON64. Not supported on ALTIVEC. |
190 | */ |
191 | template<unsigned N, class E1, class E2> SIMDPP_INL |
192 | mask_int64<N,expr_empty> cmp_gt(const uint64<N,E1>& a, |
193 | const uint64<N,E2>& b) |
194 | { |
195 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
196 | } |
197 | |
198 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_int64, uint64) |
199 | |
200 | /** Compares the values of two float32x4 vectors for greater-than |
201 | |
202 | @code |
203 | r0 = (a0 > b0) ? 0xffffffff : 0x0 |
204 | ... |
205 | rN = (aN > bN) ? 0xffffffff : 0x0 |
206 | @endcode |
207 | |
208 | @par 256-bit version: |
209 | @icost{SSE2-SSE4.1, NEON, ALTIVEC, 2} |
210 | */ |
211 | template<unsigned N, class E1, class E2> |
212 | mask_float32<N,expr_empty> cmp_gt(const float32<N,E1>& a, |
213 | const float32<N,E2>& b) |
214 | { |
215 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
216 | } |
217 | |
218 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_float32, float32) |
219 | |
220 | /** Compares the values of two float64x2 vectors for greater-than |
221 | |
222 | @code |
223 | r0 = (a0 > b0) ? 0xffffffffffffffff : 0x0 |
224 | ... |
225 | rN = (aN > bN) ? 0xffffffffffffffff : 0x0 |
226 | @endcode |
227 | |
228 | @par 128-bit version: |
229 | @novec{NEON, ALTIVEC} |
230 | |
231 | @par 256-bit version: |
232 | @novec{NEON, ALTIVEC} |
233 | @icost{SSE2-SSE4.1, 2} |
234 | */ |
235 | template<unsigned N, class E1, class E2> SIMDPP_INL |
236 | mask_float64<N,expr_empty> cmp_gt(const float64<N,E1>& a, |
237 | const float64<N,E2>& b) |
238 | { |
239 | return detail::insn::i_cmp_gt(a.eval(), b.eval()); |
240 | } |
241 | |
242 | SIMDPP_SCALAR_ARG_IMPL_VEC(cmp_gt, mask_float64, float64) |
243 | |
244 | } // namespace SIMDPP_ARCH_NAMESPACE |
245 | } // namespace simdpp |
246 | |
247 | #endif |
248 | |
249 | |