xref: /freebsd/contrib/llvm-project/clang/lib/Headers/ammintrin.h (revision ebacd8013fe5f7fdf9f6a5b286f6680dd2891036)
1 /*===---- ammintrin.h - SSE4a intrinsics -----------------------------------===
2  *
3  * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4  * See https://llvm.org/LICENSE.txt for license information.
5  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6  *
7  *===-----------------------------------------------------------------------===
8  */
9 
10 #ifndef __AMMINTRIN_H
11 #define __AMMINTRIN_H
12 
13 #if !defined(__i386__) && !defined(__x86_64__)
14 #error "This header is only meant to be used on x86 and x64 architecture"
15 #endif
16 
17 #include <pmmintrin.h>
18 
19 /* Define the default attributes for the functions in this file. */
20 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("sse4a"), __min_vector_width__(128)))
21 
22 /// Extracts the specified bits from the lower 64 bits of the 128-bit
23 ///    integer vector operand at the index \a idx and of the length \a len.
24 ///
25 /// \headerfile <x86intrin.h>
26 ///
27 /// \code
28 /// __m128i _mm_extracti_si64(__m128i x, const int len, const int idx);
29 /// \endcode
30 ///
31 /// This intrinsic corresponds to the <c> EXTRQ </c> instruction.
32 ///
33 /// \param x
34 ///    The value from which bits are extracted.
35 /// \param len
36 ///    Bits [5:0] specify the length; the other bits are ignored. If bits [5:0]
37 ///    are zero, the length is interpreted as 64.
38 /// \param idx
39 ///    Bits [5:0] specify the index of the least significant bit; the other
40 ///    bits are ignored. If the sum of the index and length is greater than 64,
41 ///    the result is undefined. If the length and index are both zero, bits
42 ///    [63:0] of parameter \a x are extracted. If the length is zero but the
43 ///    index is non-zero, the result is undefined.
44 /// \returns A 128-bit integer vector whose lower 64 bits contain the bits
45 ///    extracted from the source operand.
46 #define _mm_extracti_si64(x, len, idx) \
47   ((__m128i)__builtin_ia32_extrqi((__v2di)(__m128i)(x), \
48                                   (char)(len), (char)(idx)))
49 
50 /// Extracts the specified bits from the lower 64 bits of the 128-bit
51 ///    integer vector operand at the index and of the length specified by
52 ///    \a __y.
53 ///
54 /// \headerfile <x86intrin.h>
55 ///
56 /// This intrinsic corresponds to the <c> EXTRQ </c> instruction.
57 ///
58 /// \param __x
59 ///    The value from which bits are extracted.
60 /// \param __y
61 ///    Specifies the index of the least significant bit at [13:8] and the
62 ///    length at [5:0]; all other bits are ignored. If bits [5:0] are zero, the
63 ///    length is interpreted as 64. If the sum of the index and length is
64 ///    greater than 64, the result is undefined. If the length and index are
65 ///    both zero, bits [63:0] of parameter \a __x are extracted. If the length
66 ///    is zero but the index is non-zero, the result is undefined.
67 /// \returns A 128-bit vector whose lower 64 bits contain the bits extracted
68 ///    from the source operand.
69 static __inline__ __m128i __DEFAULT_FN_ATTRS
70 _mm_extract_si64(__m128i __x, __m128i __y)
71 {
72   return (__m128i)__builtin_ia32_extrq((__v2di)__x, (__v16qi)__y);
73 }
74 
75 /// Inserts bits of a specified length from the source integer vector
76 ///    \a y into the lower 64 bits of the destination integer vector \a x at
77 ///    the index \a idx and of the length \a len.
78 ///
79 /// \headerfile <x86intrin.h>
80 ///
81 /// \code
82 /// __m128i _mm_inserti_si64(__m128i x, __m128i y, const int len,
83 /// const int idx);
84 /// \endcode
85 ///
86 /// This intrinsic corresponds to the <c> INSERTQ </c> instruction.
87 ///
88 /// \param x
89 ///    The destination operand where bits will be inserted. The inserted bits
90 ///    are defined by the length \a len and by the index \a idx specifying the
91 ///    least significant bit.
92 /// \param y
93 ///    The source operand containing the bits to be extracted. The extracted
94 ///    bits are the least significant bits of operand \a y of length \a len.
95 /// \param len
96 ///    Bits [5:0] specify the length; the other bits are ignored. If bits [5:0]
97 ///    are zero, the length is interpreted as 64.
98 /// \param idx
99 ///    Bits [5:0] specify the index of the least significant bit; the other
100 ///    bits are ignored. If the sum of the index and length is greater than 64,
101 ///    the result is undefined. If the length and index are both zero, bits
102 ///    [63:0] of parameter \a y are inserted into parameter \a x. If the length
103 ///    is zero but the index is non-zero, the result is undefined.
104 /// \returns A 128-bit integer vector containing the original lower 64-bits of
105 ///    destination operand \a x with the specified bitfields replaced by the
106 ///    lower bits of source operand \a y. The upper 64 bits of the return value
107 ///    are undefined.
108 #define _mm_inserti_si64(x, y, len, idx) \
109   ((__m128i)__builtin_ia32_insertqi((__v2di)(__m128i)(x), \
110                                     (__v2di)(__m128i)(y), \
111                                     (char)(len), (char)(idx)))
112 
113 /// Inserts bits of a specified length from the source integer vector
114 ///    \a __y into the lower 64 bits of the destination integer vector \a __x
115 ///    at the index and of the length specified by \a __y.
116 ///
117 /// \headerfile <x86intrin.h>
118 ///
119 /// This intrinsic corresponds to the <c> INSERTQ </c> instruction.
120 ///
121 /// \param __x
122 ///    The destination operand where bits will be inserted. The inserted bits
123 ///    are defined by the length and by the index of the least significant bit
124 ///    specified by operand \a __y.
125 /// \param __y
126 ///    The source operand containing the bits to be extracted. The extracted
127 ///    bits are the least significant bits of operand \a __y with length
128 ///    specified by bits [69:64]. These are inserted into the destination at the
129 ///    index specified by bits [77:72]; all other bits are ignored. If bits
130 ///    [69:64] are zero, the length is interpreted as 64. If the sum of the
131 ///    index and length is greater than 64, the result is undefined. If the
132 ///    length and index are both zero, bits [63:0] of parameter \a __y are
133 ///    inserted into parameter \a __x. If the length is zero but the index is
134 ///    non-zero, the result is undefined.
135 /// \returns A 128-bit integer vector containing the original lower 64-bits of
136 ///    destination operand \a __x with the specified bitfields replaced by the
137 ///    lower bits of source operand \a __y. The upper 64 bits of the return
138 ///    value are undefined.
139 static __inline__ __m128i __DEFAULT_FN_ATTRS
140 _mm_insert_si64(__m128i __x, __m128i __y)
141 {
142   return (__m128i)__builtin_ia32_insertq((__v2di)__x, (__v2di)__y);
143 }
144 
145 /// Stores a 64-bit double-precision value in a 64-bit memory location.
146 ///    To minimize caching, the data is flagged as non-temporal (unlikely to be
147 ///    used again soon).
148 ///
149 /// \headerfile <x86intrin.h>
150 ///
151 /// This intrinsic corresponds to the <c> MOVNTSD </c> instruction.
152 ///
153 /// \param __p
154 ///    The 64-bit memory location used to store the register value.
155 /// \param __a
156 ///    The 64-bit double-precision floating-point register value to be stored.
157 static __inline__ void __DEFAULT_FN_ATTRS
158 _mm_stream_sd(double *__p, __m128d __a)
159 {
160   __builtin_ia32_movntsd(__p, (__v2df)__a);
161 }
162 
163 /// Stores a 32-bit single-precision floating-point value in a 32-bit
164 ///    memory location. To minimize caching, the data is flagged as
165 ///    non-temporal (unlikely to be used again soon).
166 ///
167 /// \headerfile <x86intrin.h>
168 ///
169 /// This intrinsic corresponds to the <c> MOVNTSS </c> instruction.
170 ///
171 /// \param __p
172 ///    The 32-bit memory location used to store the register value.
173 /// \param __a
174 ///    The 32-bit single-precision floating-point register value to be stored.
175 static __inline__ void __DEFAULT_FN_ATTRS
176 _mm_stream_ss(float *__p, __m128 __a)
177 {
178   __builtin_ia32_movntss(__p, (__v4sf)__a);
179 }
180 
181 #undef __DEFAULT_FN_ATTRS
182 
183 #endif /* __AMMINTRIN_H */
184