1 | /* Copyright (C) 2016 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_EXPR_TEST_BITS_H |
9 | #define LIBSIMDPP_SIMDPP_DETAIL_EXPR_TEST_BITS_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/insn/test_bits.h> |
18 | |
19 | namespace simdpp { |
20 | namespace SIMDPP_ARCH_NAMESPACE { |
21 | namespace detail { |
22 | |
23 | template<unsigned N, class V> SIMDPP_INL |
24 | bool e_test_bits_any(const any_vec<N,V>& a) |
25 | { |
26 | typename detail::get_expr_nosign<V>::type ra; |
27 | ra = a.wrapped().eval(); |
28 | return insn::i_test_bits_any(ra); |
29 | } |
30 | |
31 | #if SIMDPP_USE_SSE4_1 |
32 | template<unsigned N, template<unsigned, typename> class V, |
33 | class V1, class V2> SIMDPP_INL |
34 | bool e_test_bits_any(const any_vec<16, V<N, expr_bit_and<V1,V2>>>& e) |
35 | { |
36 | uint8<16> a, b; |
37 | a = e.wrapped().e.a.eval(); |
38 | b = e.wrapped().e.b.eval(); |
39 | return !_mm_testz_si128(b.native(), a.native()); |
40 | } |
41 | |
42 | template<unsigned N, template<unsigned, typename> class V, |
43 | class V1, class V2> SIMDPP_INL |
44 | bool e_test_bits_any(const any_vec<16, V<N, expr_bit_andnot<V1,V2>>>& e) |
45 | { |
46 | uint8<16> a, b; |
47 | a = e.wrapped().e.a.eval(); |
48 | b = e.wrapped().e.b.eval(); |
49 | return !_mm_testc_si128(b.native(), a.native()); |
50 | } |
51 | #endif |
52 | |
53 | #if SIMDPP_USE_AVX2 |
54 | template<unsigned N, template<unsigned, typename> class V, |
55 | class V1, class V2> SIMDPP_INL |
56 | bool e_test_bits_any(const any_vec<32, V<N, expr_bit_and<V1,V2>>>& e) |
57 | { |
58 | uint8<32> a, b; |
59 | a = e.wrapped().e.a.eval(); |
60 | b = e.wrapped().e.b.eval(); |
61 | return !_mm256_testz_si256(b.native(), a.native()); |
62 | } |
63 | |
64 | template<unsigned N, template<unsigned, typename> class V, |
65 | class V1, class V2> |
66 | bool e_test_bits_any(const any_vec<32, V<N, expr_bit_andnot<V1,V2>>>& e) |
67 | { |
68 | uint8<32> a, b; |
69 | a = e.wrapped().e.a.eval(); |
70 | b = e.wrapped().e.b.eval(); |
71 | return !_mm256_testc_si256(b.native(), a.native()); |
72 | } |
73 | #endif |
74 | |
75 | #if SIMDPP_USE_AVX512F |
76 | template<class V1, class V2> SIMDPP_INL |
77 | bool e_test_bits_any(const uint32<16, expr_bit_and<V1,V2>>& e) |
78 | { |
79 | uint32<16> a, b; |
80 | a = e.wrapped().e.a.eval(); |
81 | b = e.wrapped().e.b.eval(); |
82 | return _mm512_test_epi64_mask(a.native(), b.native()) != 0; |
83 | } |
84 | |
85 | template<class V1, class V2> SIMDPP_INL |
86 | bool e_test_bits_any(const uint64<8, expr_bit_and<V1,V2>>& e) |
87 | { |
88 | uint64<8> a, b; |
89 | a = e.wrapped().e.a.eval(); |
90 | b = e.wrapped().e.b.eval(); |
91 | return _mm512_test_epi64_mask(a.native(), b.native()) != 0; |
92 | } |
93 | #endif |
94 | |
95 | |
96 | } // namespace detail |
97 | } // namespace SIMDPP_ARCH_NAMESPACE |
98 | } // namespace simdpp |
99 | |
100 | #endif |
101 | |