1 | /* |
2 | * Copyright (C) 2014 Apple Inc. All rights reserved. |
3 | * |
4 | * Redistribution and use in source and binary forms, with or without |
5 | * modification, are permitted provided that the following conditions |
6 | * are met: |
7 | * 1. Redistributions of source code must retain the above copyright |
8 | * notice, this list of conditions and the following disclaimer. |
9 | * 2. Redistributions in binary form must reproduce the above copyright |
10 | * notice, this list of conditions and the following disclaimer in the |
11 | * documentation and/or other materials provided with the distribution. |
12 | * |
13 | * THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY |
14 | * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
15 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
16 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR |
17 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
18 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
19 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
20 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY |
21 | * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
22 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
23 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
24 | */ |
25 | |
26 | #include "CPUCount.h" |
27 | #include "Interpreter.h" |
28 | #include <assert.h> |
29 | #include <cstddef> |
30 | #include <cstdlib> |
31 | #include <errno.h> |
32 | #include <fcntl.h> |
33 | #include <iostream> |
34 | #include <string> |
35 | #include <string.h> |
36 | #include <sys/mman.h> |
37 | #include <sys/stat.h> |
38 | #include <sys/types.h> |
39 | #include <sys/uio.h> |
40 | #include <unistd.h> |
41 | #include <vector> |
42 | |
43 | #include "mbmalloc.h" |
44 | |
45 | #define UNUSED_PARAM(variable) (void)variable |
46 | |
47 | Interpreter::Interpreter(const char* fileName, bool shouldFreeAllObjects, bool useThreadId) |
48 | : m_shouldFreeAllObjects(shouldFreeAllObjects) |
49 | , m_useThreadId(useThreadId) |
50 | , m_currentThreadId(0) |
51 | , m_ops(1024) |
52 | { |
53 | m_fd = open(fileName, O_RDONLY); |
54 | if (m_fd == -1) { |
55 | fprintf(stderr, "Failed to open op file %s: " , fileName); |
56 | perror("" ); |
57 | exit(-1); |
58 | } |
59 | |
60 | struct stat buf; |
61 | fstat(m_fd, &buf); |
62 | |
63 | m_opCount = buf.st_size / sizeof(Op); |
64 | assert(m_opCount * sizeof(Op) == static_cast<size_t>(buf.st_size)); |
65 | |
66 | size_t maxSlot = 0; |
67 | |
68 | std::vector<Op> ops(1024); |
69 | size_t remaining = m_opCount * sizeof(Op); |
70 | while (remaining) { |
71 | size_t bytes = std::min(remaining, ops.size() * sizeof(Op)); |
72 | remaining -= bytes; |
73 | auto ret = read(m_fd, ops.data(), bytes); |
74 | UNUSED_PARAM(ret); |
75 | |
76 | size_t opCount = bytes / sizeof(Op); |
77 | for (size_t i = 0; i < opCount; ++i) { |
78 | Op op = ops[i]; |
79 | if (op.slot > maxSlot) |
80 | maxSlot = op.slot; |
81 | } |
82 | } |
83 | |
84 | m_objects.resize(maxSlot + 1); |
85 | } |
86 | |
87 | Interpreter::~Interpreter() |
88 | { |
89 | int result = close(m_fd); |
90 | if (result == -1) { |
91 | perror("Failed to close op file" ); |
92 | exit(-1); |
93 | } |
94 | } |
95 | |
96 | void Interpreter::run() |
97 | { |
98 | std::vector<Op> ops(1024); |
99 | lseek(m_fd, 0, SEEK_SET); |
100 | |
101 | m_remaining = m_opCount * sizeof(Op); |
102 | m_opsCursor = m_opsInBuffer = 0; |
103 | doOnSameThread(0); |
104 | |
105 | for (auto thread : m_threads) |
106 | thread->stop(); |
107 | |
108 | for (auto thread : m_threads) |
109 | delete thread; |
110 | |
111 | // A recording might not free all of its allocations. |
112 | if (!m_shouldFreeAllObjects) |
113 | return; |
114 | |
115 | for (size_t i = 0; i < m_objects.size(); ++i) { |
116 | if (!m_objects[i].object) |
117 | continue; |
118 | mbfree(m_objects[i].object, m_objects[i].size); |
119 | m_objects[i] = { 0, 0 }; |
120 | } |
121 | } |
122 | |
123 | bool Interpreter::readOps() |
124 | { |
125 | if (!m_remaining) |
126 | return false; |
127 | |
128 | size_t bytes = std::min(m_remaining, m_ops.size() * sizeof(Op)); |
129 | m_remaining -= bytes; |
130 | auto ret = read(m_fd, m_ops.data(), bytes); |
131 | UNUSED_PARAM(ret); |
132 | m_opsCursor = 0; |
133 | m_opsInBuffer = bytes / sizeof(Op); |
134 | |
135 | if (!m_opsInBuffer) |
136 | return false; |
137 | |
138 | return true; |
139 | } |
140 | |
141 | void Interpreter::doOnSameThread(ThreadId runThreadId) |
142 | { |
143 | while (true) { |
144 | if ((m_opsCursor >= m_opsInBuffer) && (!readOps())) { |
145 | if (runThreadId) |
146 | switchToThread(0); |
147 | return; |
148 | } |
149 | |
150 | for (; m_opsCursor < m_opsInBuffer; ++m_opsCursor) { |
151 | Op op = m_ops[m_opsCursor]; |
152 | ThreadId threadId = op.threadId; |
153 | if (m_useThreadId && (runThreadId != threadId)) { |
154 | switchToThread(threadId); |
155 | break; |
156 | } |
157 | |
158 | doMallocOp(op, m_currentThreadId); |
159 | } |
160 | } |
161 | } |
162 | |
163 | void Interpreter::switchToThread(ThreadId threadId) |
164 | { |
165 | if (m_currentThreadId == threadId) |
166 | return; |
167 | |
168 | for (ThreadId threadIndex = static_cast<ThreadId>(m_threads.size()); |
169 | threadIndex < threadId; ++threadIndex) |
170 | m_threads.push_back(new Thread(this, threadId)); |
171 | |
172 | ThreadId currentThreadId = m_currentThreadId; |
173 | |
174 | if (threadId == 0) { |
175 | std::unique_lock<std::mutex> lock(m_threadMutex); |
176 | m_currentThreadId = threadId; |
177 | m_shouldRun.notify_one(); |
178 | } else |
179 | m_threads[threadId - 1]->switchTo(); |
180 | |
181 | if (currentThreadId == 0) { |
182 | std::unique_lock<std::mutex> lock(m_threadMutex); |
183 | m_shouldRun.wait(lock, [this](){return m_currentThreadId == 0; }); |
184 | } else |
185 | m_threads[currentThreadId - 1]->waitToRun(); |
186 | } |
187 | |
188 | void Interpreter::detailedReport() |
189 | { |
190 | size_t totalInUse = 0; |
191 | size_t smallInUse = 0; |
192 | size_t mediumInUse = 0; |
193 | size_t largeInUse = 0; |
194 | size_t = 0; |
195 | size_t memoryAllocated = 0; |
196 | |
197 | for (size_t i = 0; i < m_objects.size(); ++i) { |
198 | if (!m_objects[i].object) |
199 | continue; |
200 | size_t objectSize = m_objects[i].size; |
201 | memoryAllocated += objectSize; |
202 | totalInUse++; |
203 | |
204 | if (objectSize <= 256) |
205 | smallInUse++; |
206 | else if (objectSize <= 1024) |
207 | mediumInUse++; |
208 | else if (objectSize <= 1032192) |
209 | largeInUse++; |
210 | else |
211 | extraLargeInUse++; |
212 | } |
213 | |
214 | std::cout << "0B-256B objects in use: " << smallInUse << std::endl; |
215 | std::cout << "257B-1K objects in use: " << mediumInUse << std::endl; |
216 | std::cout << " 1K-1M objects in use: " << largeInUse << std::endl; |
217 | std::cout << " 1M+ objects in use: " << extraLargeInUse << std::endl; |
218 | std::cout << " Total objects in use: " << totalInUse << std::endl; |
219 | std::cout << "Total allocated memory: " << memoryAllocated / 1024 << "kB" << std::endl; |
220 | } |
221 | static size_t compute2toPower(unsigned log2n) |
222 | { |
223 | // Check for bad alignment log2 value and return a bad alignment. |
224 | if (log2n > 64) |
225 | return 0xff00; |
226 | |
227 | size_t result = 1; |
228 | while (log2n--) |
229 | result <<= 1; |
230 | |
231 | return result; |
232 | } |
233 | |
234 | void Interpreter::doMallocOp(Op op, ThreadId) |
235 | { |
236 | switch (op.opcode) { |
237 | case op_malloc: { |
238 | m_objects[op.slot] = { mbmalloc(op.size), op.size }; |
239 | assert(m_objects[op.slot].object); |
240 | bzero(m_objects[op.slot].object, op.size); |
241 | break; |
242 | } |
243 | case op_free: { |
244 | if (!m_objects[op.slot].object) |
245 | return; |
246 | mbfree(m_objects[op.slot].object, m_objects[op.slot].size); |
247 | m_objects[op.slot] = { 0, 0 }; |
248 | break; |
249 | } |
250 | case op_realloc: { |
251 | if (!m_objects[op.slot].object) |
252 | return; |
253 | m_objects[op.slot] = { mbrealloc(m_objects[op.slot].object, m_objects[op.slot].size, op.size), op.size }; |
254 | break; |
255 | } |
256 | case op_align_malloc: { |
257 | size_t alignment = compute2toPower(op.alignLog2); |
258 | m_objects[op.slot] = { mbmemalign(alignment, op.size), op.size }; |
259 | assert(m_objects[op.slot].object); |
260 | bzero(m_objects[op.slot].object, op.size); |
261 | break; |
262 | } |
263 | default: { |
264 | fprintf(stderr, "bad opcode: %d\n" , op.opcode); |
265 | abort(); |
266 | break; |
267 | } |
268 | } |
269 | } |
270 | |
271 | Interpreter::Thread::Thread(Interpreter* myInterpreter, ThreadId threadId) |
272 | : m_threadId(threadId) |
273 | , m_myInterpreter(myInterpreter) |
274 | { |
275 | m_thread = std::thread(&Thread::runThread, this); |
276 | } |
277 | |
278 | void Interpreter::Thread::stop() |
279 | { |
280 | m_myInterpreter->switchToThread(m_threadId); |
281 | } |
282 | |
283 | Interpreter::Thread::~Thread() |
284 | { |
285 | switchTo(); |
286 | m_thread.join(); |
287 | } |
288 | |
289 | void Interpreter::Thread::runThread() |
290 | { |
291 | waitToRun(); |
292 | m_myInterpreter->doOnSameThread(m_threadId); |
293 | } |
294 | |
295 | void Interpreter::Thread::waitToRun() |
296 | { |
297 | std::unique_lock<std::mutex> lock(m_myInterpreter->m_threadMutex); |
298 | m_shouldRun.wait(lock, [this](){return m_myInterpreter->m_currentThreadId == m_threadId; }); |
299 | } |
300 | |
301 | void Interpreter::Thread::switchTo() |
302 | { |
303 | std::unique_lock<std::mutex> lock(m_myInterpreter->m_threadMutex); |
304 | m_myInterpreter->m_currentThreadId = m_threadId; |
305 | m_shouldRun.notify_one(); |
306 | } |
307 | |