1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
|
/*
* ARM AdvSIMD / SVE Vector Helpers
*
* Copyright (c) 2020 Linaro
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#ifndef TARGET_ARM_VEC_INTERNALS_H
#define TARGET_ARM_VEC_INTERNALS_H
/*
* Note that vector data is stored in host-endian 64-bit chunks,
* so addressing units smaller than that needs a host-endian fixup.
*
* The H<N> macros are used when indexing an array of elements of size N.
*
* The H1_<N> macros are used when performing byte arithmetic and then
* casting the final pointer to a type of size N.
*/
#ifdef HOST_WORDS_BIGENDIAN
#define H1(x) ((x) ^ 7)
#define H1_2(x) ((x) ^ 6)
#define H1_4(x) ((x) ^ 4)
#define H2(x) ((x) ^ 3)
#define H4(x) ((x) ^ 1)
#else
#define H1(x) (x)
#define H1_2(x) (x)
#define H1_4(x) (x)
#define H2(x) (x)
#define H4(x) (x)
#endif
/*
* Access to 64-bit elements isn't host-endian dependent; we provide H8
* and H1_8 so that when a function is being generated from a macro we
* can pass these rather than an empty macro argument, for clarity.
*/
#define H8(x) (x)
#define H1_8(x) (x)
/* Data for expanding active predicate bits to bytes, for byte elements. */
extern const uint64_t expand_pred_b_data[256];
static inline void clear_tail(void *vd, uintptr_t opr_sz, uintptr_t max_sz)
{
uint64_t *d = vd + opr_sz;
uintptr_t i;
for (i = opr_sz; i < max_sz; i += 8) {
*d++ = 0;
}
}
static inline int32_t do_sqrshl_bhs(int32_t src, int32_t shift, int bits,
bool round, uint32_t *sat)
{
if (shift <= -bits) {
/* Rounding the sign bit always produces 0. */
if (round) {
return 0;
}
return src >> 31;
} else if (shift < 0) {
if (round) {
src >>= -shift - 1;
return (src >> 1) + (src & 1);
}
return src >> -shift;
} else if (shift < bits) {
int32_t val = src << shift;
if (bits == 32) {
if (!sat || val >> shift == src) {
return val;
}
} else {
int32_t extval = sextract32(val, 0, bits);
if (!sat || val == extval) {
return extval;
}
}
} else if (!sat || src == 0) {
return 0;
}
*sat = 1;
return (1u << (bits - 1)) - (src >= 0);
}
static inline uint32_t do_uqrshl_bhs(uint32_t src, int32_t shift, int bits,
bool round, uint32_t *sat)
{
if (shift <= -(bits + round)) {
return 0;
} else if (shift < 0) {
if (round) {
src >>= -shift - 1;
return (src >> 1) + (src & 1);
}
return src >> -shift;
} else if (shift < bits) {
uint32_t val = src << shift;
if (bits == 32) {
if (!sat || val >> shift == src) {
return val;
}
} else {
uint32_t extval = extract32(val, 0, bits);
if (!sat || val == extval) {
return extval;
}
}
} else if (!sat || src == 0) {
return 0;
}
*sat = 1;
return MAKE_64BIT_MASK(0, bits);
}
static inline int32_t do_suqrshl_bhs(int32_t src, int32_t shift, int bits,
bool round, uint32_t *sat)
{
if (sat && src < 0) {
*sat = 1;
return 0;
}
return do_uqrshl_bhs(src, shift, bits, round, sat);
}
static inline int64_t do_sqrshl_d(int64_t src, int64_t shift,
bool round, uint32_t *sat)
{
if (shift <= -64) {
/* Rounding the sign bit always produces 0. */
if (round) {
return 0;
}
return src >> 63;
} else if (shift < 0) {
if (round) {
src >>= -shift - 1;
return (src >> 1) + (src & 1);
}
return src >> -shift;
} else if (shift < 64) {
int64_t val = src << shift;
if (!sat || val >> shift == src) {
return val;
}
} else if (!sat || src == 0) {
return 0;
}
*sat = 1;
return src < 0 ? INT64_MIN : INT64_MAX;
}
static inline uint64_t do_uqrshl_d(uint64_t src, int64_t shift,
bool round, uint32_t *sat)
{
if (shift <= -(64 + round)) {
return 0;
} else if (shift < 0) {
if (round) {
src >>= -shift - 1;
return (src >> 1) + (src & 1);
}
return src >> -shift;
} else if (shift < 64) {
uint64_t val = src << shift;
if (!sat || val >> shift == src) {
return val;
}
} else if (!sat || src == 0) {
return 0;
}
*sat = 1;
return UINT64_MAX;
}
static inline int64_t do_suqrshl_d(int64_t src, int64_t shift,
bool round, uint32_t *sat)
{
if (sat && src < 0) {
*sat = 1;
return 0;
}
return do_uqrshl_d(src, shift, round, sat);
}
int8_t do_sqrdmlah_b(int8_t, int8_t, int8_t, bool, bool);
int16_t do_sqrdmlah_h(int16_t, int16_t, int16_t, bool, bool, uint32_t *);
int32_t do_sqrdmlah_s(int32_t, int32_t, int32_t, bool, bool, uint32_t *);
int64_t do_sqrdmlah_d(int64_t, int64_t, int64_t, bool, bool);
/*
* 8 x 8 -> 16 vector polynomial multiply where the inputs are
* in the low 8 bits of each 16-bit element
*/
uint64_t pmull_h(uint64_t op1, uint64_t op2);
/*
* 16 x 16 -> 32 vector polynomial multiply where the inputs are
* in the low 16 bits of each 32-bit element
*/
uint64_t pmull_w(uint64_t op1, uint64_t op2);
#endif /* TARGET_ARM_VEC_INTERNALS_H */
|
