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
|
/*
* Copyright (C) 2023 Apple Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include "SIMDInfo.h"
WTF_ALLOW_UNSAFE_BUFFER_USAGE_BEGIN
namespace JSC {
class SIMDShuffle {
public:
static std::optional<unsigned> isOnlyOneSideMask(v128_t pattern)
{
unsigned first = pattern.u8x16[0];
if (first < 16) {
for (unsigned i = 1; i < 16; ++i) {
if (pattern.u8x16[i] >= 16)
return std::nullopt;
}
return 0;
}
if (first >= 32)
return std::nullopt;
for (unsigned i = 1; i < 16; ++i) {
if (pattern.u8x16[i] < 16)
return std::nullopt;
if (pattern.u8x16[i] >= 32)
return std::nullopt;
}
return 1;
}
static std::optional<uint8_t> isI8x16DupElement(v128_t pattern)
{
constexpr unsigned numberOfElements = 16 / sizeof(uint8_t);
if (std::all_of(pattern.u8x16, pattern.u8x16 + numberOfElements, [&](auto value) { return value == pattern.u8x16[0]; })) {
uint8_t lane = pattern.u8x16[0] / sizeof(uint8_t);
if (lane < numberOfElements)
return lane;
}
return std::nullopt;
}
static std::optional<uint8_t> isI16x8DupElement(v128_t pattern)
{
if (!isI16x8Shuffle(pattern))
return std::nullopt;
constexpr unsigned numberOfElements = 16 / sizeof(uint16_t);
if (std::all_of(pattern.u16x8, pattern.u16x8 + numberOfElements, [&](auto value) { return value == pattern.u16x8[0]; })) {
uint8_t lane = pattern.u8x16[0] / sizeof(uint16_t);
if (lane < numberOfElements)
return lane;
}
return std::nullopt;
}
static std::optional<uint8_t> isI32x4DupElement(v128_t pattern)
{
if (!isI32x4Shuffle(pattern))
return std::nullopt;
constexpr unsigned numberOfElements = 16 / sizeof(uint32_t);
if (std::all_of(pattern.u32x4, pattern.u32x4 + numberOfElements, [&](auto value) { return value == pattern.u32x4[0]; })) {
uint8_t lane = pattern.u8x16[0] / sizeof(uint32_t);
if (lane < numberOfElements)
return lane;
}
return std::nullopt;
}
static std::optional<uint8_t> isI64x2DupElement(v128_t pattern)
{
if (!isI64x2Shuffle(pattern))
return std::nullopt;
constexpr unsigned numberOfElements = 16 / sizeof(uint64_t);
if (std::all_of(pattern.u64x2, pattern.u64x2 + numberOfElements, [&](auto value) { return value == pattern.u64x2[0]; })) {
uint8_t lane = pattern.u8x16[0] / sizeof(uint64_t);
if (lane < numberOfElements)
return lane;
}
return std::nullopt;
}
static bool isI16x8Shuffle(v128_t pattern)
{
return isLargerElementShuffle(pattern, 2);
}
static bool isI32x4Shuffle(v128_t pattern)
{
return isLargerElementShuffle(pattern, 4);
}
static bool isI64x2Shuffle(v128_t pattern)
{
return isLargerElementShuffle(pattern, 8);
}
static bool isIdentity(v128_t pattern)
{
return isLargerElementShuffle(pattern, 16);
}
static bool isAllOutOfBoundsForUnaryShuffle(v128_t pattern)
{
for (unsigned i = 0; i < 16; ++i) {
if constexpr (isX86()) {
// https://www.felixcloutier.com/x86/pshufb
// On x64, OOB index means that highest bit is set.
// The acutal index is extracted by masking with 0b1111.
// So, for example, 0x11 index (17) will be converted to 0x1 access (not OOB).
if (!(pattern.u8x16[i] & 0x80))
return false;
} else if constexpr (isARM64()) {
// https://developer.arm.com/documentation/dui0801/g/A64-SIMD-Vector-Instructions/TBL--vector-
// On ARM64, OOB index means out of 0..15 range for unary TBL.
if (pattern.u8x16[i] < 16)
return false;
} else
return false;
}
return true;
}
static bool isAllOutOfBoundsForBinaryShuffle(v128_t pattern)
{
ASSERT(isARM64()); // Binary Shuffle is only supported by ARM64.
for (unsigned i = 0; i < 16; ++i) {
if (pattern.u8x16[i] < 32)
return false;
}
return true;
}
private:
static bool isLargerElementShuffle(v128_t pattern, unsigned size)
{
unsigned numberOfElements = 16 / size;
for (unsigned i = 0; i < numberOfElements; ++i) {
unsigned firstIndex = i * size;
unsigned first = pattern.u8x16[firstIndex];
if (first % size != 0)
return false;
for (unsigned j = 1; j < size; ++j) {
unsigned index = j + firstIndex;
if (pattern.u8x16[index] != (first + j))
return false;
}
}
return true;
}
};
} // namespace JSC
WTF_ALLOW_UNSAFE_BUFFER_USAGE_END
|
