clang  6.0.0svn
ammintrin.h
Go to the documentation of this file.
1 /*===---- ammintrin.h - SSE4a intrinsics -----------------------------------===
2  *
3  * Permission is hereby granted, free of charge, to any person obtaining a copy
4  * of this software and associated documentation files (the "Software"), to deal
5  * in the Software without restriction, including without limitation the rights
6  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7  * copies of the Software, and to permit persons to whom the Software is
8  * furnished to do so, subject to the following conditions:
9  *
10  * The above copyright notice and this permission notice shall be included in
11  * all copies or substantial portions of the Software.
12  *
13  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
18  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
19  * THE SOFTWARE.
20  *
21  *===-----------------------------------------------------------------------===
22  */
23 
24 #ifndef __AMMINTRIN_H
25 #define __AMMINTRIN_H
26 
27 #include <pmmintrin.h>
28 
29 /* Define the default attributes for the functions in this file. */
30 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("sse4a")))
31 
32 /// \brief Extracts the specified bits from the lower 64 bits of the 128-bit
33 /// integer vector operand at the index \a idx and of the length \a len.
34 ///
35 /// \headerfile <x86intrin.h>
36 ///
37 /// \code
38 /// __m128i _mm_extracti_si64(__m128i x, const int len, const int idx);
39 /// \endcode
40 ///
41 /// This intrinsic corresponds to the <c> EXTRQ </c> instruction.
42 ///
43 /// \param x
44 /// The value from which bits are extracted.
45 /// \param len
46 /// Bits [5:0] specify the length; the other bits are ignored. If bits [5:0]
47 /// are zero, the length is interpreted as 64.
48 /// \param idx
49 /// Bits [5:0] specify the index of the least significant bit; the other
50 /// bits are ignored. If the sum of the index and length is greater than 64,
51 /// the result is undefined. If the length and index are both zero, bits
52 /// [63:0] of parameter \a x are extracted. If the length is zero but the
53 /// index is non-zero, the result is undefined.
54 /// \returns A 128-bit integer vector whose lower 64 bits contain the bits
55 /// extracted from the source operand.
56 #define _mm_extracti_si64(x, len, idx) \
57  ((__m128i)__builtin_ia32_extrqi((__v2di)(__m128i)(x), \
58  (char)(len), (char)(idx)))
59 
60 /// \brief Extracts the specified bits from the lower 64 bits of the 128-bit
61 /// integer vector operand at the index and of the length specified by
62 /// \a __y.
63 ///
64 /// \headerfile <x86intrin.h>
65 ///
66 /// This intrinsic corresponds to the <c> EXTRQ </c> instruction.
67 ///
68 /// \param __x
69 /// The value from which bits are extracted.
70 /// \param __y
71 /// Specifies the index of the least significant bit at [13:8] and the
72 /// length at [5:0]; all other bits are ignored. If bits [5:0] are zero, the
73 /// length is interpreted as 64. If the sum of the index and length is
74 /// greater than 64, the result is undefined. If the length and index are
75 /// both zero, bits [63:0] of parameter \a __x are extracted. If the length
76 /// is zero but the index is non-zero, the result is undefined.
77 /// \returns A 128-bit vector whose lower 64 bits contain the bits extracted
78 /// from the source operand.
79 static __inline__ __m128i __DEFAULT_FN_ATTRS
80 _mm_extract_si64(__m128i __x, __m128i __y)
81 {
82  return (__m128i)__builtin_ia32_extrq((__v2di)__x, (__v16qi)__y);
83 }
84 
85 /// \brief Inserts bits of a specified length from the source integer vector
86 /// \a y into the lower 64 bits of the destination integer vector \a x at
87 /// the index \a idx and of the length \a len.
88 ///
89 /// \headerfile <x86intrin.h>
90 ///
91 /// \code
92 /// __m128i _mm_inserti_si64(__m128i x, __m128i y, const int len,
93 /// const int idx);
94 /// \endcode
95 ///
96 /// This intrinsic corresponds to the <c> INSERTQ </c> instruction.
97 ///
98 /// \param x
99 /// The destination operand where bits will be inserted. The inserted bits
100 /// are defined by the length \a len and by the index \a idx specifying the
101 /// least significant bit.
102 /// \param y
103 /// The source operand containing the bits to be extracted. The extracted
104 /// bits are the least significant bits of operand \a y of length \a len.
105 /// \param len
106 /// Bits [5:0] specify the length; the other bits are ignored. If bits [5:0]
107 /// are zero, the length is interpreted as 64.
108 /// \param idx
109 /// Bits [5:0] specify the index of the least significant bit; the other
110 /// bits are ignored. If the sum of the index and length is greater than 64,
111 /// the result is undefined. If the length and index are both zero, bits
112 /// [63:0] of parameter \a y are inserted into parameter \a x. If the length
113 /// is zero but the index is non-zero, the result is undefined.
114 /// \returns A 128-bit integer vector containing the original lower 64-bits of
115 /// destination operand \a x with the specified bitfields replaced by the
116 /// lower bits of source operand \a y. The upper 64 bits of the return value
117 /// are undefined.
118 #define _mm_inserti_si64(x, y, len, idx) \
119  ((__m128i)__builtin_ia32_insertqi((__v2di)(__m128i)(x), \
120  (__v2di)(__m128i)(y), \
121  (char)(len), (char)(idx)))
122 
123 /// \brief Inserts bits of a specified length from the source integer vector
124 /// \a __y into the lower 64 bits of the destination integer vector \a __x
125 /// at the index and of the length specified by \a __y.
126 ///
127 /// \headerfile <x86intrin.h>
128 ///
129 /// This intrinsic corresponds to the <c> INSERTQ </c> instruction.
130 ///
131 /// \param __x
132 /// The destination operand where bits will be inserted. The inserted bits
133 /// are defined by the length and by the index of the least significant bit
134 /// specified by operand \a __y.
135 /// \param __y
136 /// The source operand containing the bits to be extracted. The extracted
137 /// bits are the least significant bits of operand \a __y with length
138 /// specified by bits [69:64]. These are inserted into the destination at the
139 /// index specified by bits [77:72]; all other bits are ignored. If bits
140 /// [69:64] are zero, the length is interpreted as 64. If the sum of the
141 /// index and length is greater than 64, the result is undefined. If the
142 /// length and index are both zero, bits [63:0] of parameter \a __y are
143 /// inserted into parameter \a __x. If the length is zero but the index is
144 /// non-zero, the result is undefined.
145 /// \returns A 128-bit integer vector containing the original lower 64-bits of
146 /// destination operand \a __x with the specified bitfields replaced by the
147 /// lower bits of source operand \a __y. The upper 64 bits of the return
148 /// value are undefined.
149 static __inline__ __m128i __DEFAULT_FN_ATTRS
150 _mm_insert_si64(__m128i __x, __m128i __y)
151 {
152  return (__m128i)__builtin_ia32_insertq((__v2di)__x, (__v2di)__y);
153 }
154 
155 /// \brief Stores a 64-bit double-precision value in a 64-bit memory location.
156 /// To minimize caching, the data is flagged as non-temporal (unlikely to be
157 /// used again soon).
158 ///
159 /// \headerfile <x86intrin.h>
160 ///
161 /// This intrinsic corresponds to the <c> MOVNTSD </c> instruction.
162 ///
163 /// \param __p
164 /// The 64-bit memory location used to store the register value.
165 /// \param __a
166 /// The 64-bit double-precision floating-point register value to be stored.
167 static __inline__ void __DEFAULT_FN_ATTRS
168 _mm_stream_sd(double *__p, __m128d __a)
169 {
170  __builtin_ia32_movntsd(__p, (__v2df)__a);
171 }
172 
173 /// \brief Stores a 32-bit single-precision floating-point value in a 32-bit
174 /// memory location. To minimize caching, the data is flagged as
175 /// non-temporal (unlikely to be used again soon).
176 ///
177 /// \headerfile <x86intrin.h>
178 ///
179 /// This intrinsic corresponds to the <c> MOVNTSS </c> instruction.
180 ///
181 /// \param __p
182 /// The 32-bit memory location used to store the register value.
183 /// \param __a
184 /// The 32-bit single-precision floating-point register value to be stored.
185 static __inline__ void __DEFAULT_FN_ATTRS
186 _mm_stream_ss(float *__p, __m128 __a)
187 {
188  __builtin_ia32_movntss(__p, (__v4sf)__a);
189 }
190 
191 #undef __DEFAULT_FN_ATTRS
192 
193 #endif /* __AMMINTRIN_H */
static __inline unsigned char unsigned int unsigned int __y
Definition: adxintrin.h:36
static __inline__ void __DEFAULT_FN_ATTRS _mm_stream_ss(float *__p, __m128 __a)
Stores a 32-bit single-precision floating-point value in a 32-bit memory location.
Definition: ammintrin.h:186
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_extract_si64(__m128i __x, __m128i __y)
Extracts the specified bits from the lower 64 bits of the 128-bit integer vector operand at the index...
Definition: ammintrin.h:80
static __inline unsigned char unsigned int unsigned int unsigned int * __p
Definition: adxintrin.h:38
static __inline unsigned char unsigned int __x
Definition: adxintrin.h:36
static __inline__ void __DEFAULT_FN_ATTRS _mm_stream_sd(double *__p, __m128d __a)
Stores a 64-bit double-precision value in a 64-bit memory location.
Definition: ammintrin.h:168
#define __DEFAULT_FN_ATTRS
Definition: ammintrin.h:30
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_insert_si64(__m128i __x, __m128i __y)
Inserts bits of a specified length from the source integer vector __y into the lower 64 bits of the d...
Definition: ammintrin.h:150