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
|
//===-- Implementation of memset and bzero --------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_LIBC_SRC_STRING_MEMORY_UTILS_MEMSET_IMPLEMENTATIONS_H
#define LLVM_LIBC_SRC_STRING_MEMORY_UTILS_MEMSET_IMPLEMENTATIONS_H
#include "src/__support/common.h"
#include "src/__support/macros/optimization.h"
#include "src/__support/macros/properties/architectures.h"
#include "src/string/memory_utils/op_aarch64.h"
#include "src/string/memory_utils/op_builtin.h"
#include "src/string/memory_utils/op_generic.h"
#include "src/string/memory_utils/op_x86.h"
#include "src/string/memory_utils/utils.h"
#include <stddef.h> // size_t
namespace __llvm_libc {
[[maybe_unused]] LIBC_INLINE static void
inline_memset_byte_per_byte(Ptr dst, size_t offset, uint8_t value,
size_t count) {
LIBC_LOOP_NOUNROLL
for (; offset < count; ++offset)
generic::Memset<uint8_t>::block(dst + offset, value);
}
[[maybe_unused]] LIBC_INLINE static void
inline_memset_aligned_access_32bit(Ptr dst, uint8_t value, size_t count) {
constexpr size_t kAlign = sizeof(uint32_t);
if (count <= 2 * kAlign)
return inline_memset_byte_per_byte(dst, 0, value, count);
size_t bytes_to_dst_align = distance_to_align_up<kAlign>(dst);
inline_memset_byte_per_byte(dst, 0, value, bytes_to_dst_align);
size_t offset = bytes_to_dst_align;
for (; offset < count - kAlign; offset += kAlign)
store32_aligned<uint32_t>(generic::splat<uint32_t>(value), dst, offset);
inline_memset_byte_per_byte(dst, offset, value, count);
}
[[maybe_unused]] LIBC_INLINE static void
inline_memset_aligned_access_64bit(Ptr dst, uint8_t value, size_t count) {
constexpr size_t kAlign = sizeof(uint64_t);
if (count <= 2 * kAlign)
return inline_memset_byte_per_byte(dst, 0, value, count);
size_t bytes_to_dst_align = distance_to_align_up<kAlign>(dst);
inline_memset_byte_per_byte(dst, 0, value, bytes_to_dst_align);
size_t offset = bytes_to_dst_align;
for (; offset < count - kAlign; offset += kAlign)
store64_aligned<uint64_t>(generic::splat<uint64_t>(value), dst, offset);
inline_memset_byte_per_byte(dst, offset, value, count);
}
#if defined(LIBC_TARGET_ARCH_IS_X86)
[[maybe_unused]] LIBC_INLINE static void
inline_memset_x86(Ptr dst, uint8_t value, size_t count) {
#if defined(__AVX512F__)
using uint128_t = uint8x16_t;
using uint256_t = uint8x32_t;
using uint512_t = uint8x64_t;
#elif defined(__AVX__)
using uint128_t = uint8x16_t;
using uint256_t = uint8x32_t;
using uint512_t = cpp::array<uint8x32_t, 2>;
#elif defined(__SSE2__)
using uint128_t = uint8x16_t;
using uint256_t = cpp::array<uint8x16_t, 2>;
using uint512_t = cpp::array<uint8x16_t, 4>;
#else
using uint128_t = cpp::array<uint64_t, 2>;
using uint256_t = cpp::array<uint64_t, 4>;
using uint512_t = cpp::array<uint64_t, 8>;
#endif
if (count == 0)
return;
if (count == 1)
return generic::Memset<uint8_t>::block(dst, value);
if (count == 2)
return generic::Memset<uint16_t>::block(dst, value);
if (count == 3)
return generic::MemsetSequence<uint16_t, uint8_t>::block(dst, value);
if (count <= 8)
return generic::Memset<uint32_t>::head_tail(dst, value, count);
if (count <= 16)
return generic::Memset<uint64_t>::head_tail(dst, value, count);
if (count <= 32)
return generic::Memset<uint128_t>::head_tail(dst, value, count);
if (count <= 64)
return generic::Memset<uint256_t>::head_tail(dst, value, count);
if (count <= 128)
return generic::Memset<uint512_t>::head_tail(dst, value, count);
// Aligned loop
generic::Memset<uint256_t>::block(dst, value);
align_to_next_boundary<32>(dst, count);
return generic::Memset<uint256_t>::loop_and_tail(dst, value, count);
}
#endif // defined(LIBC_TARGET_ARCH_IS_X86)
#if defined(LIBC_TARGET_ARCH_IS_AARCH64)
[[maybe_unused]] LIBC_INLINE static void
inline_memset_aarch64(Ptr dst, uint8_t value, size_t count) {
static_assert(aarch64::kNeon, "aarch64 supports vector types");
using uint128_t = uint8x16_t;
using uint256_t = uint8x32_t;
using uint512_t = uint8x64_t;
if (count == 0)
return;
if (count <= 3) {
generic::Memset<uint8_t>::block(dst, value);
if (count > 1)
generic::Memset<uint16_t>::tail(dst, value, count);
return;
}
if (count <= 8)
return generic::Memset<uint32_t>::head_tail(dst, value, count);
if (count <= 16)
return generic::Memset<uint64_t>::head_tail(dst, value, count);
if (count <= 32)
return generic::Memset<uint128_t>::head_tail(dst, value, count);
if (count <= (32 + 64)) {
generic::Memset<uint256_t>::block(dst, value);
if (count <= 64)
return generic::Memset<uint256_t>::tail(dst, value, count);
generic::Memset<uint256_t>::block(dst + 32, value);
generic::Memset<uint256_t>::tail(dst, value, count);
return;
}
if (count >= 448 && value == 0 && aarch64::neon::hasZva()) {
generic::Memset<uint512_t>::block(dst, 0);
align_to_next_boundary<64>(dst, count);
return aarch64::neon::BzeroCacheLine::loop_and_tail(dst, 0, count);
} else {
generic::Memset<uint128_t>::block(dst, value);
align_to_next_boundary<16>(dst, count);
return generic::Memset<uint512_t>::loop_and_tail(dst, value, count);
}
}
#endif // defined(LIBC_TARGET_ARCH_IS_AARCH64)
LIBC_INLINE static void inline_memset(Ptr dst, uint8_t value, size_t count) {
#if defined(LIBC_TARGET_ARCH_IS_X86)
return inline_memset_x86(dst, value, count);
#elif defined(LIBC_TARGET_ARCH_IS_AARCH64)
return inline_memset_aarch64(dst, value, count);
#elif defined(LIBC_TARGET_ARCH_IS_RISCV64)
return inline_memset_aligned_access_64bit(dst, value, count);
#elif defined(LIBC_TARGET_ARCH_IS_RISCV32)
return inline_memset_aligned_access_32bit(dst, value, count);
#else
return inline_memset_byte_per_byte(dst, 0, value, count);
#endif
}
LIBC_INLINE static void inline_memset(void *dst, uint8_t value, size_t count) {
inline_memset(reinterpret_cast<Ptr>(dst), value, count);
}
} // namespace __llvm_libc
#endif // LLVM_LIBC_SRC_STRING_MEMORY_UTILS_MEMSET_IMPLEMENTATIONS_H
|
