O3CPU: Add a hack to ensure that nextPC is set correctly after syscalls.
[gem5.git] / src / cpu / base.cc
1 /*
2 * Copyright (c) 2002-2005 The Regents of The University of Michigan
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are
7 * met: redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer;
9 * 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 * neither the name of the copyright holders nor the names of its
13 * contributors may be used to endorse or promote products derived from
14 * this software without specific prior written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * Authors: Steve Reinhardt
29 * Nathan Binkert
30 */
31
32 #include <iostream>
33 #include <string>
34 #include <sstream>
35
36 #include "base/cprintf.hh"
37 #include "base/loader/symtab.hh"
38 #include "base/misc.hh"
39 #include "base/output.hh"
40 #include "base/trace.hh"
41 #include "cpu/base.hh"
42 #include "cpu/cpuevent.hh"
43 #include "cpu/thread_context.hh"
44 #include "cpu/profile.hh"
45 #include "params/BaseCPU.hh"
46 #include "sim/sim_exit.hh"
47 #include "sim/process.hh"
48 #include "sim/sim_events.hh"
49 #include "sim/system.hh"
50
51 // Hack
52 #include "sim/stat_control.hh"
53
54 using namespace std;
55
56 vector<BaseCPU *> BaseCPU::cpuList;
57
58 // This variable reflects the max number of threads in any CPU. Be
59 // careful to only use it once all the CPUs that you care about have
60 // been initialized
61 int maxThreadsPerCPU = 1;
62
63 CPUProgressEvent::CPUProgressEvent(EventQueue *q, Tick ival,
64 BaseCPU *_cpu)
65 : Event(q, Event::Progress_Event_Pri), interval(ival),
66 lastNumInst(0), cpu(_cpu)
67 {
68 if (interval)
69 schedule(curTick + interval);
70 }
71
72 void
73 CPUProgressEvent::process()
74 {
75 Counter temp = cpu->totalInstructions();
76 #ifndef NDEBUG
77 double ipc = double(temp - lastNumInst) / (interval / cpu->ticks(1));
78
79 DPRINTFN("%s progress event, instructions committed: %lli, IPC: %0.8d\n",
80 cpu->name(), temp - lastNumInst, ipc);
81 ipc = 0.0;
82 #else
83 cprintf("%lli: %s progress event, instructions committed: %lli\n",
84 curTick, cpu->name(), temp - lastNumInst);
85 #endif
86 lastNumInst = temp;
87 schedule(curTick + interval);
88 }
89
90 const char *
91 CPUProgressEvent::description() const
92 {
93 return "CPU Progress";
94 }
95
96 #if FULL_SYSTEM
97 BaseCPU::BaseCPU(Params *p)
98 : MemObject(p), clock(p->clock), instCnt(0),
99 number_of_threads(p->numThreads), system(p->system),
100 phase(p->phase)
101 #else
102 BaseCPU::BaseCPU(Params *p)
103 : MemObject(p), clock(p->clock),
104 number_of_threads(p->numThreads), system(p->system),
105 phase(p->phase)
106 #endif
107 {
108 // currentTick = curTick;
109
110 // add self to global list of CPUs
111 cpuList.push_back(this);
112
113 if (number_of_threads > maxThreadsPerCPU)
114 maxThreadsPerCPU = number_of_threads;
115
116 // allocate per-thread instruction-based event queues
117 comInstEventQueue = new EventQueue *[number_of_threads];
118 for (int i = 0; i < number_of_threads; ++i)
119 comInstEventQueue[i] = new EventQueue("instruction-based event queue");
120
121 //
122 // set up instruction-count-based termination events, if any
123 //
124 if (p->max_insts_any_thread != 0)
125 for (int i = 0; i < number_of_threads; ++i)
126 schedExitSimLoop("a thread reached the max instruction count",
127 p->max_insts_any_thread, 0,
128 comInstEventQueue[i]);
129
130 if (p->max_insts_all_threads != 0) {
131 // allocate & initialize shared downcounter: each event will
132 // decrement this when triggered; simulation will terminate
133 // when counter reaches 0
134 int *counter = new int;
135 *counter = number_of_threads;
136 for (int i = 0; i < number_of_threads; ++i)
137 new CountedExitEvent(comInstEventQueue[i],
138 "all threads reached the max instruction count",
139 p->max_insts_all_threads, *counter);
140 }
141
142 // allocate per-thread load-based event queues
143 comLoadEventQueue = new EventQueue *[number_of_threads];
144 for (int i = 0; i < number_of_threads; ++i)
145 comLoadEventQueue[i] = new EventQueue("load-based event queue");
146
147 //
148 // set up instruction-count-based termination events, if any
149 //
150 if (p->max_loads_any_thread != 0)
151 for (int i = 0; i < number_of_threads; ++i)
152 schedExitSimLoop("a thread reached the max load count",
153 p->max_loads_any_thread, 0,
154 comLoadEventQueue[i]);
155
156 if (p->max_loads_all_threads != 0) {
157 // allocate & initialize shared downcounter: each event will
158 // decrement this when triggered; simulation will terminate
159 // when counter reaches 0
160 int *counter = new int;
161 *counter = number_of_threads;
162 for (int i = 0; i < number_of_threads; ++i)
163 new CountedExitEvent(comLoadEventQueue[i],
164 "all threads reached the max load count",
165 p->max_loads_all_threads, *counter);
166 }
167
168 functionTracingEnabled = false;
169 if (p->function_trace) {
170 functionTraceStream = simout.find(csprintf("ftrace.%s", name()));
171 currentFunctionStart = currentFunctionEnd = 0;
172 functionEntryTick = p->function_trace_start;
173
174 if (p->function_trace_start == 0) {
175 functionTracingEnabled = true;
176 } else {
177 new EventWrapper<BaseCPU,
178 &BaseCPU::enableFunctionTrace>(
179 this, p->function_trace_start, true);
180 }
181 }
182 #if FULL_SYSTEM
183 profileEvent = NULL;
184 if (params()->profile)
185 profileEvent = new ProfileEvent(this, params()->profile);
186 #endif
187 tracer = params()->tracer;
188 }
189
190 void
191 BaseCPU::enableFunctionTrace()
192 {
193 functionTracingEnabled = true;
194 }
195
196 BaseCPU::~BaseCPU()
197 {
198 }
199
200 void
201 BaseCPU::init()
202 {
203 if (!params()->defer_registration)
204 registerThreadContexts();
205 }
206
207 void
208 BaseCPU::startup()
209 {
210 #if FULL_SYSTEM
211 if (!params()->defer_registration && profileEvent)
212 profileEvent->schedule(curTick);
213 #endif
214
215 if (params()->progress_interval) {
216 new CPUProgressEvent(&mainEventQueue,
217 ticks(params()->progress_interval),
218 this);
219 }
220 }
221
222
223 void
224 BaseCPU::regStats()
225 {
226 using namespace Stats;
227
228 numCycles
229 .name(name() + ".numCycles")
230 .desc("number of cpu cycles simulated")
231 ;
232
233 int size = threadContexts.size();
234 if (size > 1) {
235 for (int i = 0; i < size; ++i) {
236 stringstream namestr;
237 ccprintf(namestr, "%s.ctx%d", name(), i);
238 threadContexts[i]->regStats(namestr.str());
239 }
240 } else if (size == 1)
241 threadContexts[0]->regStats(name());
242
243 #if FULL_SYSTEM
244 #endif
245 }
246
247 Tick
248 BaseCPU::nextCycle()
249 {
250 Tick next_tick = curTick - phase + clock - 1;
251 next_tick -= (next_tick % clock);
252 next_tick += phase;
253 return next_tick;
254 }
255
256 Tick
257 BaseCPU::nextCycle(Tick begin_tick)
258 {
259 Tick next_tick = begin_tick;
260 if (next_tick % clock != 0)
261 next_tick = next_tick - (next_tick % clock) + clock;
262 next_tick += phase;
263
264 assert(next_tick >= curTick);
265 return next_tick;
266 }
267
268 void
269 BaseCPU::registerThreadContexts()
270 {
271 for (int i = 0; i < threadContexts.size(); ++i) {
272 ThreadContext *tc = threadContexts[i];
273
274 #if FULL_SYSTEM
275 int id = params()->cpu_id;
276 if (id != -1)
277 id += i;
278
279 tc->setCpuId(system->registerThreadContext(tc, id));
280 #else
281 tc->setCpuId(tc->getProcessPtr()->registerThreadContext(tc));
282 #endif
283 }
284 }
285
286
287 int
288 BaseCPU::findContext(ThreadContext *tc)
289 {
290 for (int i = 0; i < threadContexts.size(); ++i) {
291 if (tc == threadContexts[i])
292 return i;
293 }
294 return 0;
295 }
296
297 void
298 BaseCPU::switchOut()
299 {
300 // panic("This CPU doesn't support sampling!");
301 #if FULL_SYSTEM
302 if (profileEvent && profileEvent->scheduled())
303 profileEvent->deschedule();
304 #endif
305 }
306
307 void
308 BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc)
309 {
310 assert(threadContexts.size() == oldCPU->threadContexts.size());
311
312 for (int i = 0; i < threadContexts.size(); ++i) {
313 ThreadContext *newTC = threadContexts[i];
314 ThreadContext *oldTC = oldCPU->threadContexts[i];
315
316 newTC->takeOverFrom(oldTC);
317
318 CpuEvent::replaceThreadContext(oldTC, newTC);
319
320 assert(newTC->readCpuId() == oldTC->readCpuId());
321 #if FULL_SYSTEM
322 system->replaceThreadContext(newTC, newTC->readCpuId());
323 #else
324 assert(newTC->getProcessPtr() == oldTC->getProcessPtr());
325 newTC->getProcessPtr()->replaceThreadContext(newTC, newTC->readCpuId());
326 #endif
327
328 if (DTRACE(Context))
329 ThreadContext::compare(oldTC, newTC);
330 }
331
332 #if FULL_SYSTEM
333 interrupts = oldCPU->interrupts;
334
335 for (int i = 0; i < threadContexts.size(); ++i)
336 threadContexts[i]->profileClear();
337
338 if (profileEvent)
339 profileEvent->schedule(curTick);
340 #endif
341
342 // Connect new CPU to old CPU's memory only if new CPU isn't
343 // connected to anything. Also connect old CPU's memory to new
344 // CPU.
345 if (!ic->isConnected()) {
346 Port *peer = oldCPU->getPort("icache_port")->getPeer();
347 ic->setPeer(peer);
348 peer->setPeer(ic);
349 }
350
351 if (!dc->isConnected()) {
352 Port *peer = oldCPU->getPort("dcache_port")->getPeer();
353 dc->setPeer(peer);
354 peer->setPeer(dc);
355 }
356 }
357
358
359 #if FULL_SYSTEM
360 BaseCPU::ProfileEvent::ProfileEvent(BaseCPU *_cpu, Tick _interval)
361 : Event(&mainEventQueue), cpu(_cpu), interval(_interval)
362 { }
363
364 void
365 BaseCPU::ProfileEvent::process()
366 {
367 for (int i = 0, size = cpu->threadContexts.size(); i < size; ++i) {
368 ThreadContext *tc = cpu->threadContexts[i];
369 tc->profileSample();
370 }
371
372 schedule(curTick + interval);
373 }
374
375 void
376 BaseCPU::post_interrupt(int int_num, int index)
377 {
378 interrupts.post(int_num, index);
379 }
380
381 void
382 BaseCPU::clear_interrupt(int int_num, int index)
383 {
384 interrupts.clear(int_num, index);
385 }
386
387 void
388 BaseCPU::clear_interrupts()
389 {
390 interrupts.clear_all();
391 }
392
393 uint64_t
394 BaseCPU::get_interrupts(int int_num)
395 {
396 return interrupts.get_vec(int_num);
397 }
398
399 void
400 BaseCPU::serialize(std::ostream &os)
401 {
402 SERIALIZE_SCALAR(instCnt);
403 interrupts.serialize(os);
404 }
405
406 void
407 BaseCPU::unserialize(Checkpoint *cp, const std::string &section)
408 {
409 UNSERIALIZE_SCALAR(instCnt);
410 interrupts.unserialize(cp, section);
411 }
412
413 #endif // FULL_SYSTEM
414
415 void
416 BaseCPU::traceFunctionsInternal(Addr pc)
417 {
418 if (!debugSymbolTable)
419 return;
420
421 // if pc enters different function, print new function symbol and
422 // update saved range. Otherwise do nothing.
423 if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
424 string sym_str;
425 bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
426 currentFunctionStart,
427 currentFunctionEnd);
428
429 if (!found) {
430 // no symbol found: use addr as label
431 sym_str = csprintf("0x%x", pc);
432 currentFunctionStart = pc;
433 currentFunctionEnd = pc + 1;
434 }
435
436 ccprintf(*functionTraceStream, " (%d)\n%d: %s",
437 curTick - functionEntryTick, curTick, sym_str);
438 functionEntryTick = curTick;
439 }
440 }