cpu: Make sure that a drained timing CPU isn't executing ucode
[gem5.git] / src / cpu / simple / timing.cc
1 /*
2 * Copyright (c) 2010-2012 ARM Limited
3 * All rights reserved
4 *
5 * The license below extends only to copyright in the software and shall
6 * not be construed as granting a license to any other intellectual
7 * property including but not limited to intellectual property relating
8 * to a hardware implementation of the functionality of the software
9 * licensed hereunder. You may use the software subject to the license
10 * terms below provided that you ensure that this notice is replicated
11 * unmodified and in its entirety in all distributions of the software,
12 * modified or unmodified, in source code or in binary form.
13 *
14 * Copyright (c) 2002-2005 The Regents of The University of Michigan
15 * All rights reserved.
16 *
17 * Redistribution and use in source and binary forms, with or without
18 * modification, are permitted provided that the following conditions are
19 * met: redistributions of source code must retain the above copyright
20 * notice, this list of conditions and the following disclaimer;
21 * redistributions in binary form must reproduce the above copyright
22 * notice, this list of conditions and the following disclaimer in the
23 * documentation and/or other materials provided with the distribution;
24 * neither the name of the copyright holders nor the names of its
25 * contributors may be used to endorse or promote products derived from
26 * this software without specific prior written permission.
27 *
28 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
31 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
32 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
33 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
34 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
35 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
36 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
37 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 *
40 * Authors: Steve Reinhardt
41 */
42
43 #include "arch/locked_mem.hh"
44 #include "arch/mmapped_ipr.hh"
45 #include "arch/utility.hh"
46 #include "base/bigint.hh"
47 #include "config/the_isa.hh"
48 #include "cpu/simple/timing.hh"
49 #include "cpu/exetrace.hh"
50 #include "debug/Config.hh"
51 #include "debug/Drain.hh"
52 #include "debug/ExecFaulting.hh"
53 #include "debug/SimpleCPU.hh"
54 #include "mem/packet.hh"
55 #include "mem/packet_access.hh"
56 #include "params/TimingSimpleCPU.hh"
57 #include "sim/faults.hh"
58 #include "sim/full_system.hh"
59 #include "sim/system.hh"
60
61 using namespace std;
62 using namespace TheISA;
63
64 void
65 TimingSimpleCPU::init()
66 {
67 BaseCPU::init();
68
69 if (!params()->switched_out &&
70 system->getMemoryMode() != Enums::timing) {
71 fatal("The timing CPU requires the memory system to be in "
72 "'timing' mode.\n");
73 }
74
75 // Initialise the ThreadContext's memory proxies
76 tcBase()->initMemProxies(tcBase());
77
78 if (FullSystem && !params()->switched_out) {
79 for (int i = 0; i < threadContexts.size(); ++i) {
80 ThreadContext *tc = threadContexts[i];
81 // initialize CPU, including PC
82 TheISA::initCPU(tc, _cpuId);
83 }
84 }
85 }
86
87 void
88 TimingSimpleCPU::TimingCPUPort::TickEvent::schedule(PacketPtr _pkt, Tick t)
89 {
90 pkt = _pkt;
91 cpu->schedule(this, t);
92 }
93
94 TimingSimpleCPU::TimingSimpleCPU(TimingSimpleCPUParams *p)
95 : BaseSimpleCPU(p), fetchTranslation(this), icachePort(this),
96 dcachePort(this), ifetch_pkt(NULL), dcache_pkt(NULL), previousCycle(0),
97 fetchEvent(this), drainManager(NULL)
98 {
99 _status = Idle;
100
101 system->totalNumInsts = 0;
102 }
103
104
105 TimingSimpleCPU::~TimingSimpleCPU()
106 {
107 }
108
109 unsigned int
110 TimingSimpleCPU::drain(DrainManager *drain_manager)
111 {
112 if (_status == Idle ||
113 (_status == BaseSimpleCPU::Running && isDrained()) ||
114 _status == SwitchedOut) {
115 assert(!fetchEvent.scheduled());
116 DPRINTF(Drain, "No need to drain.\n");
117 return 0;
118 } else {
119 drainManager = drain_manager;
120 DPRINTF(Drain, "Requesting drain: %s\n", pcState());
121
122 // The fetch event can become descheduled if a drain didn't
123 // succeed on the first attempt. We need to reschedule it if
124 // the CPU is waiting for a microcode routine to complete.
125 if (_status == BaseSimpleCPU::Running && !isDrained() &&
126 !fetchEvent.scheduled()) {
127 schedule(fetchEvent, nextCycle());
128 }
129
130 return 1;
131 }
132 }
133
134 void
135 TimingSimpleCPU::drainResume()
136 {
137 assert(!fetchEvent.scheduled());
138
139 DPRINTF(SimpleCPU, "Resume\n");
140 if (_status != SwitchedOut && _status != Idle) {
141 if (system->getMemoryMode() != Enums::timing) {
142 fatal("The timing CPU requires the memory system to be in "
143 "'timing' mode.\n");
144 }
145
146 schedule(fetchEvent, nextCycle());
147 }
148 }
149
150 bool
151 TimingSimpleCPU::tryCompleteDrain()
152 {
153 if (!drainManager)
154 return false;
155
156 DPRINTF(Drain, "tryCompleteDrain: %s\n", pcState());
157 if (!isDrained())
158 return false;
159
160 DPRINTF(Drain, "CPU done draining, processing drain event\n");
161 drainManager->signalDrainDone();
162 drainManager = NULL;
163
164 return true;
165 }
166
167 void
168 TimingSimpleCPU::switchOut()
169 {
170 BaseSimpleCPU::switchOut();
171
172 assert(!fetchEvent.scheduled());
173 assert(_status == BaseSimpleCPU::Running || _status == Idle);
174 assert(!stayAtPC);
175 assert(microPC() == 0);
176
177 _status = SwitchedOut;
178 numCycles += curCycle() - previousCycle;
179 }
180
181
182 void
183 TimingSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
184 {
185 BaseSimpleCPU::takeOverFrom(oldCPU);
186
187 // if any of this CPU's ThreadContexts are active, mark the CPU as
188 // running and schedule its tick event.
189 for (int i = 0; i < threadContexts.size(); ++i) {
190 ThreadContext *tc = threadContexts[i];
191 if (tc->status() == ThreadContext::Active &&
192 _status != BaseSimpleCPU::Running) {
193 _status = BaseSimpleCPU::Running;
194 break;
195 }
196 }
197
198 if (_status != BaseSimpleCPU::Running) {
199 _status = Idle;
200 }
201 assert(threadContexts.size() == 1);
202 previousCycle = curCycle();
203 }
204
205
206 void
207 TimingSimpleCPU::activateContext(ThreadID thread_num, Cycles delay)
208 {
209 DPRINTF(SimpleCPU, "ActivateContext %d (%d cycles)\n", thread_num, delay);
210
211 assert(thread_num == 0);
212 assert(thread);
213
214 assert(_status == Idle);
215
216 notIdleFraction++;
217 _status = BaseSimpleCPU::Running;
218
219 // kick things off by initiating the fetch of the next instruction
220 schedule(fetchEvent, clockEdge(delay));
221 }
222
223
224 void
225 TimingSimpleCPU::suspendContext(ThreadID thread_num)
226 {
227 DPRINTF(SimpleCPU, "SuspendContext %d\n", thread_num);
228
229 assert(thread_num == 0);
230 assert(thread);
231
232 if (_status == Idle)
233 return;
234
235 assert(_status == BaseSimpleCPU::Running);
236
237 // just change status to Idle... if status != Running,
238 // completeInst() will not initiate fetch of next instruction.
239
240 notIdleFraction--;
241 _status = Idle;
242 }
243
244 bool
245 TimingSimpleCPU::handleReadPacket(PacketPtr pkt)
246 {
247 RequestPtr req = pkt->req;
248 if (req->isMmappedIpr()) {
249 Cycles delay = TheISA::handleIprRead(thread->getTC(), pkt);
250 new IprEvent(pkt, this, clockEdge(delay));
251 _status = DcacheWaitResponse;
252 dcache_pkt = NULL;
253 } else if (!dcachePort.sendTimingReq(pkt)) {
254 _status = DcacheRetry;
255 dcache_pkt = pkt;
256 } else {
257 _status = DcacheWaitResponse;
258 // memory system takes ownership of packet
259 dcache_pkt = NULL;
260 }
261 return dcache_pkt == NULL;
262 }
263
264 void
265 TimingSimpleCPU::sendData(RequestPtr req, uint8_t *data, uint64_t *res,
266 bool read)
267 {
268 PacketPtr pkt;
269 buildPacket(pkt, req, read);
270 pkt->dataDynamicArray<uint8_t>(data);
271 if (req->getFlags().isSet(Request::NO_ACCESS)) {
272 assert(!dcache_pkt);
273 pkt->makeResponse();
274 completeDataAccess(pkt);
275 } else if (read) {
276 handleReadPacket(pkt);
277 } else {
278 bool do_access = true; // flag to suppress cache access
279
280 if (req->isLLSC()) {
281 do_access = TheISA::handleLockedWrite(thread, req);
282 } else if (req->isCondSwap()) {
283 assert(res);
284 req->setExtraData(*res);
285 }
286
287 if (do_access) {
288 dcache_pkt = pkt;
289 handleWritePacket();
290 } else {
291 _status = DcacheWaitResponse;
292 completeDataAccess(pkt);
293 }
294 }
295 }
296
297 void
298 TimingSimpleCPU::sendSplitData(RequestPtr req1, RequestPtr req2,
299 RequestPtr req, uint8_t *data, bool read)
300 {
301 PacketPtr pkt1, pkt2;
302 buildSplitPacket(pkt1, pkt2, req1, req2, req, data, read);
303 if (req->getFlags().isSet(Request::NO_ACCESS)) {
304 assert(!dcache_pkt);
305 pkt1->makeResponse();
306 completeDataAccess(pkt1);
307 } else if (read) {
308 SplitFragmentSenderState * send_state =
309 dynamic_cast<SplitFragmentSenderState *>(pkt1->senderState);
310 if (handleReadPacket(pkt1)) {
311 send_state->clearFromParent();
312 send_state = dynamic_cast<SplitFragmentSenderState *>(
313 pkt2->senderState);
314 if (handleReadPacket(pkt2)) {
315 send_state->clearFromParent();
316 }
317 }
318 } else {
319 dcache_pkt = pkt1;
320 SplitFragmentSenderState * send_state =
321 dynamic_cast<SplitFragmentSenderState *>(pkt1->senderState);
322 if (handleWritePacket()) {
323 send_state->clearFromParent();
324 dcache_pkt = pkt2;
325 send_state = dynamic_cast<SplitFragmentSenderState *>(
326 pkt2->senderState);
327 if (handleWritePacket()) {
328 send_state->clearFromParent();
329 }
330 }
331 }
332 }
333
334 void
335 TimingSimpleCPU::translationFault(Fault fault)
336 {
337 // fault may be NoFault in cases where a fault is suppressed,
338 // for instance prefetches.
339 numCycles += curCycle() - previousCycle;
340 previousCycle = curCycle();
341
342 if (traceData) {
343 // Since there was a fault, we shouldn't trace this instruction.
344 delete traceData;
345 traceData = NULL;
346 }
347
348 postExecute();
349
350 advanceInst(fault);
351 }
352
353 void
354 TimingSimpleCPU::buildPacket(PacketPtr &pkt, RequestPtr req, bool read)
355 {
356 MemCmd cmd;
357 if (read) {
358 cmd = MemCmd::ReadReq;
359 if (req->isLLSC())
360 cmd = MemCmd::LoadLockedReq;
361 } else {
362 cmd = MemCmd::WriteReq;
363 if (req->isLLSC()) {
364 cmd = MemCmd::StoreCondReq;
365 } else if (req->isSwap()) {
366 cmd = MemCmd::SwapReq;
367 }
368 }
369 pkt = new Packet(req, cmd);
370 }
371
372 void
373 TimingSimpleCPU::buildSplitPacket(PacketPtr &pkt1, PacketPtr &pkt2,
374 RequestPtr req1, RequestPtr req2, RequestPtr req,
375 uint8_t *data, bool read)
376 {
377 pkt1 = pkt2 = NULL;
378
379 assert(!req1->isMmappedIpr() && !req2->isMmappedIpr());
380
381 if (req->getFlags().isSet(Request::NO_ACCESS)) {
382 buildPacket(pkt1, req, read);
383 return;
384 }
385
386 buildPacket(pkt1, req1, read);
387 buildPacket(pkt2, req2, read);
388
389 req->setPhys(req1->getPaddr(), req->getSize(), req1->getFlags(), dataMasterId());
390 PacketPtr pkt = new Packet(req, pkt1->cmd.responseCommand());
391
392 pkt->dataDynamicArray<uint8_t>(data);
393 pkt1->dataStatic<uint8_t>(data);
394 pkt2->dataStatic<uint8_t>(data + req1->getSize());
395
396 SplitMainSenderState * main_send_state = new SplitMainSenderState;
397 pkt->senderState = main_send_state;
398 main_send_state->fragments[0] = pkt1;
399 main_send_state->fragments[1] = pkt2;
400 main_send_state->outstanding = 2;
401 pkt1->senderState = new SplitFragmentSenderState(pkt, 0);
402 pkt2->senderState = new SplitFragmentSenderState(pkt, 1);
403 }
404
405 Fault
406 TimingSimpleCPU::readMem(Addr addr, uint8_t *data,
407 unsigned size, unsigned flags)
408 {
409 Fault fault;
410 const int asid = 0;
411 const ThreadID tid = 0;
412 const Addr pc = thread->instAddr();
413 unsigned block_size = dcachePort.peerBlockSize();
414 BaseTLB::Mode mode = BaseTLB::Read;
415
416 if (traceData) {
417 traceData->setAddr(addr);
418 }
419
420 RequestPtr req = new Request(asid, addr, size,
421 flags, dataMasterId(), pc, _cpuId, tid);
422
423 Addr split_addr = roundDown(addr + size - 1, block_size);
424 assert(split_addr <= addr || split_addr - addr < block_size);
425
426 _status = DTBWaitResponse;
427 if (split_addr > addr) {
428 RequestPtr req1, req2;
429 assert(!req->isLLSC() && !req->isSwap());
430 req->splitOnVaddr(split_addr, req1, req2);
431
432 WholeTranslationState *state =
433 new WholeTranslationState(req, req1, req2, new uint8_t[size],
434 NULL, mode);
435 DataTranslation<TimingSimpleCPU *> *trans1 =
436 new DataTranslation<TimingSimpleCPU *>(this, state, 0);
437 DataTranslation<TimingSimpleCPU *> *trans2 =
438 new DataTranslation<TimingSimpleCPU *>(this, state, 1);
439
440 thread->dtb->translateTiming(req1, tc, trans1, mode);
441 thread->dtb->translateTiming(req2, tc, trans2, mode);
442 } else {
443 WholeTranslationState *state =
444 new WholeTranslationState(req, new uint8_t[size], NULL, mode);
445 DataTranslation<TimingSimpleCPU *> *translation
446 = new DataTranslation<TimingSimpleCPU *>(this, state);
447 thread->dtb->translateTiming(req, tc, translation, mode);
448 }
449
450 return NoFault;
451 }
452
453 bool
454 TimingSimpleCPU::handleWritePacket()
455 {
456 RequestPtr req = dcache_pkt->req;
457 if (req->isMmappedIpr()) {
458 Cycles delay = TheISA::handleIprWrite(thread->getTC(), dcache_pkt);
459 new IprEvent(dcache_pkt, this, clockEdge(delay));
460 _status = DcacheWaitResponse;
461 dcache_pkt = NULL;
462 } else if (!dcachePort.sendTimingReq(dcache_pkt)) {
463 _status = DcacheRetry;
464 } else {
465 _status = DcacheWaitResponse;
466 // memory system takes ownership of packet
467 dcache_pkt = NULL;
468 }
469 return dcache_pkt == NULL;
470 }
471
472 Fault
473 TimingSimpleCPU::writeMem(uint8_t *data, unsigned size,
474 Addr addr, unsigned flags, uint64_t *res)
475 {
476 uint8_t *newData = new uint8_t[size];
477 memcpy(newData, data, size);
478
479 const int asid = 0;
480 const ThreadID tid = 0;
481 const Addr pc = thread->instAddr();
482 unsigned block_size = dcachePort.peerBlockSize();
483 BaseTLB::Mode mode = BaseTLB::Write;
484
485 if (traceData) {
486 traceData->setAddr(addr);
487 }
488
489 RequestPtr req = new Request(asid, addr, size,
490 flags, dataMasterId(), pc, _cpuId, tid);
491
492 Addr split_addr = roundDown(addr + size - 1, block_size);
493 assert(split_addr <= addr || split_addr - addr < block_size);
494
495 _status = DTBWaitResponse;
496 if (split_addr > addr) {
497 RequestPtr req1, req2;
498 assert(!req->isLLSC() && !req->isSwap());
499 req->splitOnVaddr(split_addr, req1, req2);
500
501 WholeTranslationState *state =
502 new WholeTranslationState(req, req1, req2, newData, res, mode);
503 DataTranslation<TimingSimpleCPU *> *trans1 =
504 new DataTranslation<TimingSimpleCPU *>(this, state, 0);
505 DataTranslation<TimingSimpleCPU *> *trans2 =
506 new DataTranslation<TimingSimpleCPU *>(this, state, 1);
507
508 thread->dtb->translateTiming(req1, tc, trans1, mode);
509 thread->dtb->translateTiming(req2, tc, trans2, mode);
510 } else {
511 WholeTranslationState *state =
512 new WholeTranslationState(req, newData, res, mode);
513 DataTranslation<TimingSimpleCPU *> *translation =
514 new DataTranslation<TimingSimpleCPU *>(this, state);
515 thread->dtb->translateTiming(req, tc, translation, mode);
516 }
517
518 // Translation faults will be returned via finishTranslation()
519 return NoFault;
520 }
521
522
523 void
524 TimingSimpleCPU::finishTranslation(WholeTranslationState *state)
525 {
526 _status = BaseSimpleCPU::Running;
527
528 if (state->getFault() != NoFault) {
529 if (state->isPrefetch()) {
530 state->setNoFault();
531 }
532 delete [] state->data;
533 state->deleteReqs();
534 translationFault(state->getFault());
535 } else {
536 if (!state->isSplit) {
537 sendData(state->mainReq, state->data, state->res,
538 state->mode == BaseTLB::Read);
539 } else {
540 sendSplitData(state->sreqLow, state->sreqHigh, state->mainReq,
541 state->data, state->mode == BaseTLB::Read);
542 }
543 }
544
545 delete state;
546 }
547
548
549 void
550 TimingSimpleCPU::fetch()
551 {
552 DPRINTF(SimpleCPU, "Fetch\n");
553
554 if (!curStaticInst || !curStaticInst->isDelayedCommit())
555 checkForInterrupts();
556
557 checkPcEventQueue();
558
559 // We must have just got suspended by a PC event
560 if (_status == Idle)
561 return;
562
563 TheISA::PCState pcState = thread->pcState();
564 bool needToFetch = !isRomMicroPC(pcState.microPC()) && !curMacroStaticInst;
565
566 if (needToFetch) {
567 _status = BaseSimpleCPU::Running;
568 Request *ifetch_req = new Request();
569 ifetch_req->setThreadContext(_cpuId, /* thread ID */ 0);
570 setupFetchRequest(ifetch_req);
571 DPRINTF(SimpleCPU, "Translating address %#x\n", ifetch_req->getVaddr());
572 thread->itb->translateTiming(ifetch_req, tc, &fetchTranslation,
573 BaseTLB::Execute);
574 } else {
575 _status = IcacheWaitResponse;
576 completeIfetch(NULL);
577
578 numCycles += curCycle() - previousCycle;
579 previousCycle = curCycle();
580 }
581 }
582
583
584 void
585 TimingSimpleCPU::sendFetch(Fault fault, RequestPtr req, ThreadContext *tc)
586 {
587 if (fault == NoFault) {
588 DPRINTF(SimpleCPU, "Sending fetch for addr %#x(pa: %#x)\n",
589 req->getVaddr(), req->getPaddr());
590 ifetch_pkt = new Packet(req, MemCmd::ReadReq);
591 ifetch_pkt->dataStatic(&inst);
592 DPRINTF(SimpleCPU, " -- pkt addr: %#x\n", ifetch_pkt->getAddr());
593
594 if (!icachePort.sendTimingReq(ifetch_pkt)) {
595 // Need to wait for retry
596 _status = IcacheRetry;
597 } else {
598 // Need to wait for cache to respond
599 _status = IcacheWaitResponse;
600 // ownership of packet transferred to memory system
601 ifetch_pkt = NULL;
602 }
603 } else {
604 DPRINTF(SimpleCPU, "Translation of addr %#x faulted\n", req->getVaddr());
605 delete req;
606 // fetch fault: advance directly to next instruction (fault handler)
607 _status = BaseSimpleCPU::Running;
608 advanceInst(fault);
609 }
610
611 numCycles += curCycle() - previousCycle;
612 previousCycle = curCycle();
613 }
614
615
616 void
617 TimingSimpleCPU::advanceInst(Fault fault)
618 {
619 if (_status == Faulting)
620 return;
621
622 if (fault != NoFault) {
623 advancePC(fault);
624 DPRINTF(SimpleCPU, "Fault occured, scheduling fetch event\n");
625 reschedule(fetchEvent, nextCycle(), true);
626 _status = Faulting;
627 return;
628 }
629
630
631 if (!stayAtPC)
632 advancePC(fault);
633
634 if (tryCompleteDrain())
635 return;
636
637 if (_status == BaseSimpleCPU::Running) {
638 // kick off fetch of next instruction... callback from icache
639 // response will cause that instruction to be executed,
640 // keeping the CPU running.
641 fetch();
642 }
643 }
644
645
646 void
647 TimingSimpleCPU::completeIfetch(PacketPtr pkt)
648 {
649 DPRINTF(SimpleCPU, "Complete ICache Fetch for addr %#x\n", pkt ?
650 pkt->getAddr() : 0);
651
652 // received a response from the icache: execute the received
653 // instruction
654
655 assert(!pkt || !pkt->isError());
656 assert(_status == IcacheWaitResponse);
657
658 _status = BaseSimpleCPU::Running;
659
660 numCycles += curCycle() - previousCycle;
661 previousCycle = curCycle();
662
663 preExecute();
664 if (curStaticInst && curStaticInst->isMemRef()) {
665 // load or store: just send to dcache
666 Fault fault = curStaticInst->initiateAcc(this, traceData);
667
668 // If we're not running now the instruction will complete in a dcache
669 // response callback or the instruction faulted and has started an
670 // ifetch
671 if (_status == BaseSimpleCPU::Running) {
672 if (fault != NoFault && traceData) {
673 // If there was a fault, we shouldn't trace this instruction.
674 delete traceData;
675 traceData = NULL;
676 }
677
678 postExecute();
679 // @todo remove me after debugging with legion done
680 if (curStaticInst && (!curStaticInst->isMicroop() ||
681 curStaticInst->isFirstMicroop()))
682 instCnt++;
683 advanceInst(fault);
684 }
685 } else if (curStaticInst) {
686 // non-memory instruction: execute completely now
687 Fault fault = curStaticInst->execute(this, traceData);
688
689 // keep an instruction count
690 if (fault == NoFault)
691 countInst();
692 else if (traceData && !DTRACE(ExecFaulting)) {
693 delete traceData;
694 traceData = NULL;
695 }
696
697 postExecute();
698 // @todo remove me after debugging with legion done
699 if (curStaticInst && (!curStaticInst->isMicroop() ||
700 curStaticInst->isFirstMicroop()))
701 instCnt++;
702 advanceInst(fault);
703 } else {
704 advanceInst(NoFault);
705 }
706
707 if (pkt) {
708 delete pkt->req;
709 delete pkt;
710 }
711 }
712
713 void
714 TimingSimpleCPU::IcachePort::ITickEvent::process()
715 {
716 cpu->completeIfetch(pkt);
717 }
718
719 bool
720 TimingSimpleCPU::IcachePort::recvTimingResp(PacketPtr pkt)
721 {
722 DPRINTF(SimpleCPU, "Received timing response %#x\n", pkt->getAddr());
723 // delay processing of returned data until next CPU clock edge
724 Tick next_tick = cpu->nextCycle();
725
726 if (next_tick == curTick())
727 cpu->completeIfetch(pkt);
728 else
729 tickEvent.schedule(pkt, next_tick);
730
731 return true;
732 }
733
734 void
735 TimingSimpleCPU::IcachePort::recvRetry()
736 {
737 // we shouldn't get a retry unless we have a packet that we're
738 // waiting to transmit
739 assert(cpu->ifetch_pkt != NULL);
740 assert(cpu->_status == IcacheRetry);
741 PacketPtr tmp = cpu->ifetch_pkt;
742 if (sendTimingReq(tmp)) {
743 cpu->_status = IcacheWaitResponse;
744 cpu->ifetch_pkt = NULL;
745 }
746 }
747
748 void
749 TimingSimpleCPU::completeDataAccess(PacketPtr pkt)
750 {
751 // received a response from the dcache: complete the load or store
752 // instruction
753 assert(!pkt->isError());
754 assert(_status == DcacheWaitResponse || _status == DTBWaitResponse ||
755 pkt->req->getFlags().isSet(Request::NO_ACCESS));
756
757 numCycles += curCycle() - previousCycle;
758 previousCycle = curCycle();
759
760 if (pkt->senderState) {
761 SplitFragmentSenderState * send_state =
762 dynamic_cast<SplitFragmentSenderState *>(pkt->senderState);
763 assert(send_state);
764 delete pkt->req;
765 delete pkt;
766 PacketPtr big_pkt = send_state->bigPkt;
767 delete send_state;
768
769 SplitMainSenderState * main_send_state =
770 dynamic_cast<SplitMainSenderState *>(big_pkt->senderState);
771 assert(main_send_state);
772 // Record the fact that this packet is no longer outstanding.
773 assert(main_send_state->outstanding != 0);
774 main_send_state->outstanding--;
775
776 if (main_send_state->outstanding) {
777 return;
778 } else {
779 delete main_send_state;
780 big_pkt->senderState = NULL;
781 pkt = big_pkt;
782 }
783 }
784
785 _status = BaseSimpleCPU::Running;
786
787 Fault fault = curStaticInst->completeAcc(pkt, this, traceData);
788
789 // keep an instruction count
790 if (fault == NoFault)
791 countInst();
792 else if (traceData) {
793 // If there was a fault, we shouldn't trace this instruction.
794 delete traceData;
795 traceData = NULL;
796 }
797
798 // the locked flag may be cleared on the response packet, so check
799 // pkt->req and not pkt to see if it was a load-locked
800 if (pkt->isRead() && pkt->req->isLLSC()) {
801 TheISA::handleLockedRead(thread, pkt->req);
802 }
803
804 delete pkt->req;
805 delete pkt;
806
807 postExecute();
808
809 advanceInst(fault);
810 }
811
812 bool
813 TimingSimpleCPU::DcachePort::recvTimingResp(PacketPtr pkt)
814 {
815 // delay processing of returned data until next CPU clock edge
816 Tick next_tick = cpu->nextCycle();
817
818 if (next_tick == curTick()) {
819 cpu->completeDataAccess(pkt);
820 } else {
821 if (!tickEvent.scheduled()) {
822 tickEvent.schedule(pkt, next_tick);
823 } else {
824 // In the case of a split transaction and a cache that is
825 // faster than a CPU we could get two responses before
826 // next_tick expires
827 if (!retryEvent.scheduled())
828 cpu->schedule(retryEvent, next_tick);
829 return false;
830 }
831 }
832
833 return true;
834 }
835
836 void
837 TimingSimpleCPU::DcachePort::DTickEvent::process()
838 {
839 cpu->completeDataAccess(pkt);
840 }
841
842 void
843 TimingSimpleCPU::DcachePort::recvRetry()
844 {
845 // we shouldn't get a retry unless we have a packet that we're
846 // waiting to transmit
847 assert(cpu->dcache_pkt != NULL);
848 assert(cpu->_status == DcacheRetry);
849 PacketPtr tmp = cpu->dcache_pkt;
850 if (tmp->senderState) {
851 // This is a packet from a split access.
852 SplitFragmentSenderState * send_state =
853 dynamic_cast<SplitFragmentSenderState *>(tmp->senderState);
854 assert(send_state);
855 PacketPtr big_pkt = send_state->bigPkt;
856
857 SplitMainSenderState * main_send_state =
858 dynamic_cast<SplitMainSenderState *>(big_pkt->senderState);
859 assert(main_send_state);
860
861 if (sendTimingReq(tmp)) {
862 // If we were able to send without retrying, record that fact
863 // and try sending the other fragment.
864 send_state->clearFromParent();
865 int other_index = main_send_state->getPendingFragment();
866 if (other_index > 0) {
867 tmp = main_send_state->fragments[other_index];
868 cpu->dcache_pkt = tmp;
869 if ((big_pkt->isRead() && cpu->handleReadPacket(tmp)) ||
870 (big_pkt->isWrite() && cpu->handleWritePacket())) {
871 main_send_state->fragments[other_index] = NULL;
872 }
873 } else {
874 cpu->_status = DcacheWaitResponse;
875 // memory system takes ownership of packet
876 cpu->dcache_pkt = NULL;
877 }
878 }
879 } else if (sendTimingReq(tmp)) {
880 cpu->_status = DcacheWaitResponse;
881 // memory system takes ownership of packet
882 cpu->dcache_pkt = NULL;
883 }
884 }
885
886 TimingSimpleCPU::IprEvent::IprEvent(Packet *_pkt, TimingSimpleCPU *_cpu,
887 Tick t)
888 : pkt(_pkt), cpu(_cpu)
889 {
890 cpu->schedule(this, t);
891 }
892
893 void
894 TimingSimpleCPU::IprEvent::process()
895 {
896 cpu->completeDataAccess(pkt);
897 }
898
899 const char *
900 TimingSimpleCPU::IprEvent::description() const
901 {
902 return "Timing Simple CPU Delay IPR event";
903 }
904
905
906 void
907 TimingSimpleCPU::printAddr(Addr a)
908 {
909 dcachePort.printAddr(a);
910 }
911
912
913 ////////////////////////////////////////////////////////////////////////
914 //
915 // TimingSimpleCPU Simulation Object
916 //
917 TimingSimpleCPU *
918 TimingSimpleCPUParams::create()
919 {
920 numThreads = 1;
921 if (!FullSystem && workload.size() != 1)
922 panic("only one workload allowed");
923 return new TimingSimpleCPU(this);
924 }