Merge zizzer:/z/m5/Bitkeeper/newmem
[gem5.git] / src / cpu / simple / timing.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 */
30
31 #include "arch/locked_mem.hh"
32 #include "arch/utility.hh"
33 #include "cpu/exetrace.hh"
34 #include "cpu/simple/timing.hh"
35 #include "mem/packet_impl.hh"
36 #include "sim/builder.hh"
37 #include "sim/system.hh"
38
39 using namespace std;
40 using namespace TheISA;
41
42 Port *
43 TimingSimpleCPU::getPort(const std::string &if_name, int idx)
44 {
45 if (if_name == "dcache_port")
46 return &dcachePort;
47 else if (if_name == "icache_port")
48 return &icachePort;
49 else
50 panic("No Such Port\n");
51 }
52
53 void
54 TimingSimpleCPU::init()
55 {
56 BaseCPU::init();
57 #if FULL_SYSTEM
58 for (int i = 0; i < threadContexts.size(); ++i) {
59 ThreadContext *tc = threadContexts[i];
60
61 // initialize CPU, including PC
62 TheISA::initCPU(tc, tc->readCpuId());
63 }
64 #endif
65 }
66
67 Tick
68 TimingSimpleCPU::CpuPort::recvAtomic(Packet *pkt)
69 {
70 panic("TimingSimpleCPU doesn't expect recvAtomic callback!");
71 return curTick;
72 }
73
74 void
75 TimingSimpleCPU::CpuPort::recvFunctional(Packet *pkt)
76 {
77 panic("TimingSimpleCPU doesn't expect recvFunctional callback!");
78 }
79
80 void
81 TimingSimpleCPU::CpuPort::recvStatusChange(Status status)
82 {
83 if (status == RangeChange)
84 return;
85
86 panic("TimingSimpleCPU doesn't expect recvStatusChange callback!");
87 }
88
89
90 void
91 TimingSimpleCPU::CpuPort::TickEvent::schedule(Packet *_pkt, Tick t)
92 {
93 pkt = _pkt;
94 Event::schedule(t);
95 }
96
97 TimingSimpleCPU::TimingSimpleCPU(Params *p)
98 : BaseSimpleCPU(p), icachePort(this, p->clock), dcachePort(this, p->clock),
99 cpu_id(p->cpu_id)
100 {
101 _status = Idle;
102 ifetch_pkt = dcache_pkt = NULL;
103 drainEvent = NULL;
104 fetchEvent = NULL;
105 changeState(SimObject::Running);
106 }
107
108
109 TimingSimpleCPU::~TimingSimpleCPU()
110 {
111 }
112
113 void
114 TimingSimpleCPU::serialize(ostream &os)
115 {
116 SimObject::State so_state = SimObject::getState();
117 SERIALIZE_ENUM(so_state);
118 BaseSimpleCPU::serialize(os);
119 }
120
121 void
122 TimingSimpleCPU::unserialize(Checkpoint *cp, const string &section)
123 {
124 SimObject::State so_state;
125 UNSERIALIZE_ENUM(so_state);
126 BaseSimpleCPU::unserialize(cp, section);
127 }
128
129 unsigned int
130 TimingSimpleCPU::drain(Event *drain_event)
131 {
132 // TimingSimpleCPU is ready to drain if it's not waiting for
133 // an access to complete.
134 if (status() == Idle || status() == Running || status() == SwitchedOut) {
135 changeState(SimObject::Drained);
136 return 0;
137 } else {
138 changeState(SimObject::Draining);
139 drainEvent = drain_event;
140 return 1;
141 }
142 }
143
144 void
145 TimingSimpleCPU::resume()
146 {
147 if (_status != SwitchedOut && _status != Idle) {
148 // Delete the old event if it existed.
149 if (fetchEvent) {
150 if (fetchEvent->scheduled())
151 fetchEvent->deschedule();
152
153 delete fetchEvent;
154 }
155
156 fetchEvent =
157 new EventWrapper<TimingSimpleCPU, &TimingSimpleCPU::fetch>(this, false);
158 fetchEvent->schedule(curTick);
159 }
160
161 assert(system->getMemoryMode() == System::Timing);
162 changeState(SimObject::Running);
163 }
164
165 void
166 TimingSimpleCPU::switchOut()
167 {
168 assert(status() == Running || status() == Idle);
169 _status = SwitchedOut;
170
171 // If we've been scheduled to resume but are then told to switch out,
172 // we'll need to cancel it.
173 if (fetchEvent && fetchEvent->scheduled())
174 fetchEvent->deschedule();
175 }
176
177
178 void
179 TimingSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
180 {
181 BaseCPU::takeOverFrom(oldCPU);
182
183 // if any of this CPU's ThreadContexts are active, mark the CPU as
184 // running and schedule its tick event.
185 for (int i = 0; i < threadContexts.size(); ++i) {
186 ThreadContext *tc = threadContexts[i];
187 if (tc->status() == ThreadContext::Active && _status != Running) {
188 _status = Running;
189 break;
190 }
191 }
192 }
193
194
195 void
196 TimingSimpleCPU::activateContext(int thread_num, int delay)
197 {
198 assert(thread_num == 0);
199 assert(thread);
200
201 assert(_status == Idle);
202
203 notIdleFraction++;
204 _status = Running;
205 // kick things off by initiating the fetch of the next instruction
206 fetchEvent =
207 new EventWrapper<TimingSimpleCPU, &TimingSimpleCPU::fetch>(this, false);
208 fetchEvent->schedule(curTick + cycles(delay));
209 }
210
211
212 void
213 TimingSimpleCPU::suspendContext(int thread_num)
214 {
215 assert(thread_num == 0);
216 assert(thread);
217
218 assert(_status == Running);
219
220 // just change status to Idle... if status != Running,
221 // completeInst() will not initiate fetch of next instruction.
222
223 notIdleFraction--;
224 _status = Idle;
225 }
226
227
228 template <class T>
229 Fault
230 TimingSimpleCPU::read(Addr addr, T &data, unsigned flags)
231 {
232 Request *req =
233 new Request(/* asid */ 0, addr, sizeof(T), flags, thread->readPC(),
234 cpu_id, /* thread ID */ 0);
235
236 if (traceData) {
237 traceData->setAddr(req->getVaddr());
238 }
239
240 // translate to physical address
241 Fault fault = thread->translateDataReadReq(req);
242
243 // Now do the access.
244 if (fault == NoFault) {
245 Packet *pkt =
246 new Packet(req, Packet::ReadReq, Packet::Broadcast);
247 pkt->dataDynamic<T>(new T);
248
249 if (!dcachePort.sendTiming(pkt)) {
250 _status = DcacheRetry;
251 dcache_pkt = pkt;
252 } else {
253 _status = DcacheWaitResponse;
254 // memory system takes ownership of packet
255 dcache_pkt = NULL;
256 }
257 }
258
259 // This will need a new way to tell if it has a dcache attached.
260 if (req->isUncacheable())
261 recordEvent("Uncached Read");
262
263 return fault;
264 }
265
266 #ifndef DOXYGEN_SHOULD_SKIP_THIS
267
268 template
269 Fault
270 TimingSimpleCPU::read(Addr addr, uint64_t &data, unsigned flags);
271
272 template
273 Fault
274 TimingSimpleCPU::read(Addr addr, uint32_t &data, unsigned flags);
275
276 template
277 Fault
278 TimingSimpleCPU::read(Addr addr, uint16_t &data, unsigned flags);
279
280 template
281 Fault
282 TimingSimpleCPU::read(Addr addr, uint8_t &data, unsigned flags);
283
284 #endif //DOXYGEN_SHOULD_SKIP_THIS
285
286 template<>
287 Fault
288 TimingSimpleCPU::read(Addr addr, double &data, unsigned flags)
289 {
290 return read(addr, *(uint64_t*)&data, flags);
291 }
292
293 template<>
294 Fault
295 TimingSimpleCPU::read(Addr addr, float &data, unsigned flags)
296 {
297 return read(addr, *(uint32_t*)&data, flags);
298 }
299
300
301 template<>
302 Fault
303 TimingSimpleCPU::read(Addr addr, int32_t &data, unsigned flags)
304 {
305 return read(addr, (uint32_t&)data, flags);
306 }
307
308
309 template <class T>
310 Fault
311 TimingSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
312 {
313 Request *req =
314 new Request(/* asid */ 0, addr, sizeof(T), flags, thread->readPC(),
315 cpu_id, /* thread ID */ 0);
316
317 // translate to physical address
318 Fault fault = thread->translateDataWriteReq(req);
319
320 // Now do the access.
321 if (fault == NoFault) {
322 assert(dcache_pkt == NULL);
323 dcache_pkt = new Packet(req, Packet::WriteReq, Packet::Broadcast);
324 dcache_pkt->allocate();
325 dcache_pkt->set(data);
326
327 bool do_access = true; // flag to suppress cache access
328
329 if (req->isLocked()) {
330 do_access = TheISA::handleLockedWrite(thread, req);
331 }
332
333 if (do_access) {
334 if (!dcachePort.sendTiming(dcache_pkt)) {
335 _status = DcacheRetry;
336 } else {
337 _status = DcacheWaitResponse;
338 // memory system takes ownership of packet
339 dcache_pkt = NULL;
340 }
341 }
342 }
343
344 // This will need a new way to tell if it's hooked up to a cache or not.
345 if (req->isUncacheable())
346 recordEvent("Uncached Write");
347
348 // If the write needs to have a fault on the access, consider calling
349 // changeStatus() and changing it to "bad addr write" or something.
350 return fault;
351 }
352
353
354 #ifndef DOXYGEN_SHOULD_SKIP_THIS
355 template
356 Fault
357 TimingSimpleCPU::write(uint64_t data, Addr addr,
358 unsigned flags, uint64_t *res);
359
360 template
361 Fault
362 TimingSimpleCPU::write(uint32_t data, Addr addr,
363 unsigned flags, uint64_t *res);
364
365 template
366 Fault
367 TimingSimpleCPU::write(uint16_t data, Addr addr,
368 unsigned flags, uint64_t *res);
369
370 template
371 Fault
372 TimingSimpleCPU::write(uint8_t data, Addr addr,
373 unsigned flags, uint64_t *res);
374
375 #endif //DOXYGEN_SHOULD_SKIP_THIS
376
377 template<>
378 Fault
379 TimingSimpleCPU::write(double data, Addr addr, unsigned flags, uint64_t *res)
380 {
381 return write(*(uint64_t*)&data, addr, flags, res);
382 }
383
384 template<>
385 Fault
386 TimingSimpleCPU::write(float data, Addr addr, unsigned flags, uint64_t *res)
387 {
388 return write(*(uint32_t*)&data, addr, flags, res);
389 }
390
391
392 template<>
393 Fault
394 TimingSimpleCPU::write(int32_t data, Addr addr, unsigned flags, uint64_t *res)
395 {
396 return write((uint32_t)data, addr, flags, res);
397 }
398
399
400 void
401 TimingSimpleCPU::fetch()
402 {
403 checkForInterrupts();
404
405 Request *ifetch_req = new Request();
406 ifetch_req->setThreadContext(cpu_id, /* thread ID */ 0);
407 Fault fault = setupFetchRequest(ifetch_req);
408
409 ifetch_pkt = new Packet(ifetch_req, Packet::ReadReq, Packet::Broadcast);
410 ifetch_pkt->dataStatic(&inst);
411
412 if (fault == NoFault) {
413 if (!icachePort.sendTiming(ifetch_pkt)) {
414 // Need to wait for retry
415 _status = IcacheRetry;
416 } else {
417 // Need to wait for cache to respond
418 _status = IcacheWaitResponse;
419 // ownership of packet transferred to memory system
420 ifetch_pkt = NULL;
421 }
422 } else {
423 // fetch fault: advance directly to next instruction (fault handler)
424 advanceInst(fault);
425 }
426 }
427
428
429 void
430 TimingSimpleCPU::advanceInst(Fault fault)
431 {
432 advancePC(fault);
433
434 if (_status == Running) {
435 // kick off fetch of next instruction... callback from icache
436 // response will cause that instruction to be executed,
437 // keeping the CPU running.
438 fetch();
439 }
440 }
441
442
443 void
444 TimingSimpleCPU::completeIfetch(Packet *pkt)
445 {
446 // received a response from the icache: execute the received
447 // instruction
448 assert(pkt->result == Packet::Success);
449 assert(_status == IcacheWaitResponse);
450
451 _status = Running;
452
453 delete pkt->req;
454 delete pkt;
455
456 if (getState() == SimObject::Draining) {
457 completeDrain();
458 return;
459 }
460
461 preExecute();
462 if (curStaticInst->isMemRef() && !curStaticInst->isDataPrefetch()) {
463 // load or store: just send to dcache
464 Fault fault = curStaticInst->initiateAcc(this, traceData);
465 if (_status != Running) {
466 // instruction will complete in dcache response callback
467 assert(_status == DcacheWaitResponse || _status == DcacheRetry);
468 assert(fault == NoFault);
469 } else {
470 if (fault == NoFault) {
471 // early fail on store conditional: complete now
472 assert(dcache_pkt != NULL);
473 fault = curStaticInst->completeAcc(dcache_pkt, this,
474 traceData);
475 delete dcache_pkt->req;
476 delete dcache_pkt;
477 dcache_pkt = NULL;
478 }
479 postExecute();
480 advanceInst(fault);
481 }
482 } else {
483 // non-memory instruction: execute completely now
484 Fault fault = curStaticInst->execute(this, traceData);
485 postExecute();
486 advanceInst(fault);
487 }
488 }
489
490 void
491 TimingSimpleCPU::IcachePort::ITickEvent::process()
492 {
493 cpu->completeIfetch(pkt);
494 }
495
496 bool
497 TimingSimpleCPU::IcachePort::recvTiming(Packet *pkt)
498 {
499 // delay processing of returned data until next CPU clock edge
500 Tick time = pkt->req->getTime();
501 while (time < curTick)
502 time += lat;
503
504 if (time == curTick)
505 cpu->completeIfetch(pkt);
506 else
507 tickEvent.schedule(pkt, time);
508
509 return true;
510 }
511
512 void
513 TimingSimpleCPU::IcachePort::recvRetry()
514 {
515 // we shouldn't get a retry unless we have a packet that we're
516 // waiting to transmit
517 assert(cpu->ifetch_pkt != NULL);
518 assert(cpu->_status == IcacheRetry);
519 Packet *tmp = cpu->ifetch_pkt;
520 if (sendTiming(tmp)) {
521 cpu->_status = IcacheWaitResponse;
522 cpu->ifetch_pkt = NULL;
523 }
524 }
525
526 void
527 TimingSimpleCPU::completeDataAccess(Packet *pkt)
528 {
529 // received a response from the dcache: complete the load or store
530 // instruction
531 assert(pkt->result == Packet::Success);
532 assert(_status == DcacheWaitResponse);
533 _status = Running;
534
535 if (getState() == SimObject::Draining) {
536 completeDrain();
537
538 delete pkt->req;
539 delete pkt;
540
541 return;
542 }
543
544 Fault fault = curStaticInst->completeAcc(pkt, this, traceData);
545
546 if (pkt->isRead() && pkt->req->isLocked()) {
547 TheISA::handleLockedRead(thread, pkt->req);
548 }
549
550 delete pkt->req;
551 delete pkt;
552
553 postExecute();
554 advanceInst(fault);
555 }
556
557
558 void
559 TimingSimpleCPU::completeDrain()
560 {
561 DPRINTF(Config, "Done draining\n");
562 changeState(SimObject::Drained);
563 drainEvent->process();
564 }
565
566 bool
567 TimingSimpleCPU::DcachePort::recvTiming(Packet *pkt)
568 {
569 // delay processing of returned data until next CPU clock edge
570 Tick time = pkt->req->getTime();
571 while (time < curTick)
572 time += lat;
573
574 if (time == curTick)
575 cpu->completeDataAccess(pkt);
576 else
577 tickEvent.schedule(pkt, time);
578
579 return true;
580 }
581
582 void
583 TimingSimpleCPU::DcachePort::DTickEvent::process()
584 {
585 cpu->completeDataAccess(pkt);
586 }
587
588 void
589 TimingSimpleCPU::DcachePort::recvRetry()
590 {
591 // we shouldn't get a retry unless we have a packet that we're
592 // waiting to transmit
593 assert(cpu->dcache_pkt != NULL);
594 assert(cpu->_status == DcacheRetry);
595 Packet *tmp = cpu->dcache_pkt;
596 if (sendTiming(tmp)) {
597 cpu->_status = DcacheWaitResponse;
598 // memory system takes ownership of packet
599 cpu->dcache_pkt = NULL;
600 }
601 }
602
603
604 ////////////////////////////////////////////////////////////////////////
605 //
606 // TimingSimpleCPU Simulation Object
607 //
608 BEGIN_DECLARE_SIM_OBJECT_PARAMS(TimingSimpleCPU)
609
610 Param<Counter> max_insts_any_thread;
611 Param<Counter> max_insts_all_threads;
612 Param<Counter> max_loads_any_thread;
613 Param<Counter> max_loads_all_threads;
614 Param<Tick> progress_interval;
615 SimObjectParam<MemObject *> mem;
616 SimObjectParam<System *> system;
617 Param<int> cpu_id;
618
619 #if FULL_SYSTEM
620 SimObjectParam<AlphaITB *> itb;
621 SimObjectParam<AlphaDTB *> dtb;
622 Param<Tick> profile;
623 #else
624 SimObjectParam<Process *> workload;
625 #endif // FULL_SYSTEM
626
627 Param<int> clock;
628
629 Param<bool> defer_registration;
630 Param<int> width;
631 Param<bool> function_trace;
632 Param<Tick> function_trace_start;
633 Param<bool> simulate_stalls;
634
635 END_DECLARE_SIM_OBJECT_PARAMS(TimingSimpleCPU)
636
637 BEGIN_INIT_SIM_OBJECT_PARAMS(TimingSimpleCPU)
638
639 INIT_PARAM(max_insts_any_thread,
640 "terminate when any thread reaches this inst count"),
641 INIT_PARAM(max_insts_all_threads,
642 "terminate when all threads have reached this inst count"),
643 INIT_PARAM(max_loads_any_thread,
644 "terminate when any thread reaches this load count"),
645 INIT_PARAM(max_loads_all_threads,
646 "terminate when all threads have reached this load count"),
647 INIT_PARAM(progress_interval, "Progress interval"),
648 INIT_PARAM(mem, "memory"),
649 INIT_PARAM(system, "system object"),
650 INIT_PARAM(cpu_id, "processor ID"),
651
652 #if FULL_SYSTEM
653 INIT_PARAM(itb, "Instruction TLB"),
654 INIT_PARAM(dtb, "Data TLB"),
655 INIT_PARAM(profile, ""),
656 #else
657 INIT_PARAM(workload, "processes to run"),
658 #endif // FULL_SYSTEM
659
660 INIT_PARAM(clock, "clock speed"),
661 INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
662 INIT_PARAM(width, "cpu width"),
663 INIT_PARAM(function_trace, "Enable function trace"),
664 INIT_PARAM(function_trace_start, "Cycle to start function trace"),
665 INIT_PARAM(simulate_stalls, "Simulate cache stall cycles")
666
667 END_INIT_SIM_OBJECT_PARAMS(TimingSimpleCPU)
668
669
670 CREATE_SIM_OBJECT(TimingSimpleCPU)
671 {
672 TimingSimpleCPU::Params *params = new TimingSimpleCPU::Params();
673 params->name = getInstanceName();
674 params->numberOfThreads = 1;
675 params->max_insts_any_thread = max_insts_any_thread;
676 params->max_insts_all_threads = max_insts_all_threads;
677 params->max_loads_any_thread = max_loads_any_thread;
678 params->max_loads_all_threads = max_loads_all_threads;
679 params->progress_interval = progress_interval;
680 params->deferRegistration = defer_registration;
681 params->clock = clock;
682 params->functionTrace = function_trace;
683 params->functionTraceStart = function_trace_start;
684 params->mem = mem;
685 params->system = system;
686 params->cpu_id = cpu_id;
687
688 #if FULL_SYSTEM
689 params->itb = itb;
690 params->dtb = dtb;
691 params->profile = profile;
692 #else
693 params->process = workload;
694 #endif
695
696 TimingSimpleCPU *cpu = new TimingSimpleCPU(params);
697 return cpu;
698 }
699
700 REGISTER_SIM_OBJECT("TimingSimpleCPU", TimingSimpleCPU)
701