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
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
|
/*
* Copyright (C) 2016-2018 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 "CPU.h"
#include "JSCJSValue.h"
WTF_ALLOW_UNSAFE_BUFFER_USAGE_BEGIN
namespace JSC {
// We use these memory operations when modifying memory that might be scanned by the concurrent collector.
// We don't call the default operations because they're not guaranteed to store to memory in eight byte aligned
// chunks. If we happened to fall into the system's normal byte copy loop, we may see a torn JSValue in the
// concurrent collector.
constexpr size_t smallCutoff = 30 * 8;
constexpr size_t mediumCutoff = 4 * 1024;
// This is a forwards loop so gcSafeMemmove can rely on the direction.
template <typename T>
ALWAYS_INLINE void gcSafeMemcpy(T* dst, const T* src, size_t bytes)
{
static_assert(sizeof(T) == sizeof(JSValue));
RELEASE_ASSERT(bytes % 8 == 0);
#if USE(JSVALUE64)
auto slowPathForwardMemcpy = [&] {
size_t count = bytes / 8;
for (unsigned i = 0; i < count; ++i)
std::bit_cast<volatile uint64_t*>(dst)[i] = std::bit_cast<volatile uint64_t*>(src)[i];
};
#if CPU(X86_64) || CPU(ARM64)
if (bytes <= smallCutoff)
slowPathForwardMemcpy();
else if (isARM64() || bytes <= mediumCutoff) {
#if CPU(X86_64)
size_t alignedBytes = (bytes / 64) * 64;
size_t tmp;
size_t offset = 0;
asm volatile(
".balign 32\t\n"
"1:\t\n"
"cmpq %q[offset], %q[alignedBytes]\t\n"
"je 2f\t\n"
"movups (%q[src], %q[offset], 1), %%xmm0\t\n"
"movups 16(%q[src], %q[offset], 1), %%xmm1\t\n"
"movups 32(%q[src], %q[offset], 1), %%xmm2\t\n"
"movups 48(%q[src], %q[offset], 1), %%xmm3\t\n"
"movups %%xmm0, (%q[dst], %q[offset], 1)\t\n"
"movups %%xmm1, 16(%q[dst], %q[offset], 1)\t\n"
"movups %%xmm2, 32(%q[dst], %q[offset], 1)\t\n"
"movups %%xmm3, 48(%q[dst], %q[offset], 1)\t\n"
"addq $64, %q[offset]\t\n"
"jmp 1b\t\n"
"2:\t\n"
"cmpq %q[offset], %q[bytes]\t\n"
"je 3f\t\n"
"movq (%q[src], %q[offset], 1), %q[tmp]\t\n"
"movq %q[tmp], (%q[dst], %q[offset], 1)\t\n"
"addq $8, %q[offset]\t\n"
"jmp 2b\t\n"
"3:\t\n"
: [alignedBytes] "+r" (alignedBytes), [bytes] "+r" (bytes), [tmp] "+r" (tmp), [offset] "+r" (offset), [dst] "+r" (dst), [src] "+r" (src)
:
: "xmm0", "xmm1", "xmm2", "xmm3", "memory", "cc"
);
#elif CPU(ARM64)
uint64_t alignedBytes = (static_cast<uint64_t>(bytes) / 32) * 32;
uint64_t dstPtr = static_cast<uint64_t>(std::bit_cast<uintptr_t>(dst));
uint64_t srcPtr = static_cast<uint64_t>(std::bit_cast<uintptr_t>(src));
uint64_t end = dstPtr + bytes;
uint64_t alignedEnd = dstPtr + alignedBytes;
asm volatile(
"1:\t\n"
"cmp %x[dstPtr], %x[alignedEnd]\t\n"
"b.eq 2f\t\n"
"ldp q0, q1, [%x[srcPtr]], #0x20\t\n"
"stp q0, q1, [%x[dstPtr]], #0x20\t\n"
"b 1b\t\n"
"2:\t\n"
"cmp %x[dstPtr], %x[end]\t\n"
"b.eq 3f\t\n"
"ldr d0, [%x[srcPtr]], #0x8\t\n"
"str d0, [%x[dstPtr]], #0x8\t\n"
"b 2b\t\n"
"3:\t\n"
: [end] "+r" (end), [alignedEnd] "+r" (alignedEnd), [dstPtr] "+r" (dstPtr), [srcPtr] "+r" (srcPtr)
:
: "d0", "d1", "memory", "cc"
);
#endif // CPU(X86_64)
} else {
RELEASE_ASSERT(isX86_64());
#if CPU(X86_64)
size_t count = bytes / 8;
asm volatile(
".balign 16\t\n"
"cld\t\n"
"rep movsq\t\n"
: "+D" (dst), "+S" (src), "+c" (count)
:
: "memory");
#endif // CPU(X86_64)
}
#else
slowPathForwardMemcpy();
#endif // CPU(X86_64) || CPU(ARM64)
#else
memcpy(dst, src, bytes);
#endif // USE(JSVALUE64)
}
template <typename T>
ALWAYS_INLINE void gcSafeMemmove(T* dst, const T* src, size_t bytes)
{
static_assert(sizeof(T) == sizeof(JSValue));
RELEASE_ASSERT(bytes % 8 == 0);
#if USE(JSVALUE64)
if (std::bit_cast<uintptr_t>(src) >= std::bit_cast<uintptr_t>(dst)) {
// This is written to do a forwards loop, so calling it is ok.
gcSafeMemcpy(dst, src, bytes);
return;
}
if ((static_cast<uint64_t>(std::bit_cast<uintptr_t>(src)) + static_cast<uint64_t>(bytes)) <= static_cast<uint64_t>(std::bit_cast<uintptr_t>(dst))) {
gcSafeMemcpy(dst, src, bytes);
return;
}
auto slowPathBackwardsMemmove = [&] {
size_t count = bytes / 8;
for (size_t i = count; i--; )
std::bit_cast<volatile uint64_t*>(dst)[i] = std::bit_cast<volatile uint64_t*>(src)[i];
};
#if CPU(X86_64) || CPU(ARM64)
if (bytes <= smallCutoff)
slowPathBackwardsMemmove();
else {
#if CPU(X86_64)
size_t alignedBytes = (bytes / 64) * 64;
size_t tail = alignedBytes;
size_t tmp;
asm volatile(
"2:\t\n"
"cmpq %q[tail], %q[bytes]\t\n"
"je 1f\t\n"
"addq $-8, %q[bytes]\t\n"
"movq (%q[src], %q[bytes], 1), %q[tmp]\t\n"
"movq %q[tmp], (%q[dst], %q[bytes], 1)\t\n"
"jmp 2b\t\n"
"1:\t\n"
"test %q[alignedBytes], %q[alignedBytes]\t\n"
"jz 3f\t\n"
".balign 32\t\n"
"100:\t\n"
"movups -64(%q[src], %q[alignedBytes], 1), %%xmm0\t\n"
"movups -48(%q[src], %q[alignedBytes], 1), %%xmm1\t\n"
"movups -32(%q[src], %q[alignedBytes], 1), %%xmm2\t\n"
"movups -16(%q[src], %q[alignedBytes], 1), %%xmm3\t\n"
"movups %%xmm0, -64(%q[dst], %q[alignedBytes], 1)\t\n"
"movups %%xmm1, -48(%q[dst], %q[alignedBytes], 1)\t\n"
"movups %%xmm2, -32(%q[dst], %q[alignedBytes], 1)\t\n"
"movups %%xmm3, -16(%q[dst], %q[alignedBytes], 1)\t\n"
"addq $-64, %q[alignedBytes]\t\n"
"jnz 100b\t\n"
"3:\t\n"
: [alignedBytes] "+r" (alignedBytes), [tail] "+r" (tail), [bytes] "+r" (bytes), [tmp] "+r" (tmp), [dst] "+r" (dst), [src] "+r" (src)
:
: "xmm0", "xmm1", "xmm2", "xmm3", "memory", "cc"
);
#elif CPU(ARM64)
uint64_t alignedBytes = (static_cast<uint64_t>(bytes) / 32) * 32;
uint64_t dstPtr = static_cast<uint64_t>(std::bit_cast<uintptr_t>(dst) + static_cast<uint64_t>(bytes));
uint64_t srcPtr = static_cast<uint64_t>(std::bit_cast<uintptr_t>(src) + static_cast<uint64_t>(bytes));
uint64_t alignedEnd = std::bit_cast<uintptr_t>(dst) + alignedBytes;
uint64_t end = std::bit_cast<uintptr_t>(dst);
asm volatile(
"1:\t\n"
"cmp %x[dstPtr], %x[alignedEnd]\t\n"
"b.eq 2f\t\n"
"ldr d0, [%x[srcPtr], #-0x8]!\t\n"
"str d0, [%x[dstPtr], #-0x8]!\t\n"
"b 1b\t\n"
"2:\t\n"
"cmp %x[dstPtr], %x[end]\t\n"
"b.eq 3f\t\n"
"ldp q0, q1, [%x[srcPtr], #-0x20]!\t\n"
"stp q0, q1, [%x[dstPtr], #-0x20]!\t\n"
"b 2b\t\n"
"3:\t\n"
: [alignedEnd] "+r" (alignedEnd), [end] "+r" (end), [dstPtr] "+r" (dstPtr), [srcPtr] "+r" (srcPtr)
:
: "d0", "d1", "memory", "cc"
);
#endif // CPU(X86_64)
}
#else
slowPathBackwardsMemmove();
#endif // CPU(X86_64) || CPU(ARM64)
#else
memmove(dst, src, bytes);
#endif // USE(JSVALUE64)
}
template <typename T>
ALWAYS_INLINE void gcSafeZeroMemory(T* dst, size_t bytes)
{
static_assert(sizeof(T) == sizeof(JSValue));
RELEASE_ASSERT(bytes % 8 == 0);
#if USE(JSVALUE64)
#if CPU(X86_64)
uint64_t zero = 0;
size_t count = bytes / 8;
asm volatile (
"rep stosq\n\t"
: "+D"(dst), "+c"(count)
: "a"(zero)
: "memory"
);
#elif CPU(ARM64)
uint64_t alignedBytes = (static_cast<uint64_t>(bytes) / 64) * 64;
uint64_t dstPtr = static_cast<uint64_t>(std::bit_cast<uintptr_t>(dst));
uint64_t end = dstPtr + bytes;
uint64_t alignedEnd = dstPtr + alignedBytes;
asm volatile(
"movi d0, #0\t\n"
"movi d1, #0\t\n"
".p2align 4\t\n"
"2:\t\n"
"cmp %x[dstPtr], %x[alignedEnd]\t\n"
"b.eq 4f\t\n"
"stnp q0, q0, [%x[dstPtr]]\t\n"
"stnp q0, q0, [%x[dstPtr], #0x20]\t\n"
"add %x[dstPtr], %x[dstPtr], #0x40\t\n"
"b 2b\t\n"
"4:\t\n"
"cmp %x[dstPtr], %x[end]\t\n"
"b.eq 5f\t\n"
"str d0, [%x[dstPtr]], #0x8\t\n"
"b 4b\t\n"
"5:\t\n"
: [alignedBytes] "+r" (alignedBytes), [bytes] "+r" (bytes), [dstPtr] "+r" (dstPtr), [end] "+r" (end), [alignedEnd] "+r" (alignedEnd)
:
: "d0", "d1", "memory", "cc"
);
#else
size_t count = bytes / 8;
for (size_t i = 0; i < count; ++i)
std::bit_cast<volatile uint64_t*>(dst)[i] = 0;
#endif
#else
memset(reinterpret_cast<char*>(dst), 0, bytes);
#endif // USE(JSVALUE64)
}
} // namespace JSC
WTF_ALLOW_UNSAFE_BUFFER_USAGE_END
|
