cpu.cc (5702:bf84e2fa05f7) cpu.cc (5704:98224505352a)
1/*
2 * Copyright (c) 2004-2006 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: Kevin Lim
29 * Korey Sewell
30 */
31
32#include "config/full_system.hh"
33#include "config/use_checker.hh"
34
35#include "cpu/activity.hh"
36#include "cpu/simple_thread.hh"
37#include "cpu/thread_context.hh"
38#include "cpu/o3/isa_specific.hh"
39#include "cpu/o3/cpu.hh"
40#include "cpu/o3/thread_context.hh"
41#include "enums/MemoryMode.hh"
42#include "sim/core.hh"
43#include "sim/stat_control.hh"
44
45#if FULL_SYSTEM
46#include "cpu/quiesce_event.hh"
47#include "sim/system.hh"
48#else
49#include "sim/process.hh"
50#endif
51
52#if USE_CHECKER
53#include "cpu/checker/cpu.hh"
54#endif
55
56#if THE_ISA == ALPHA_ISA
57#include "arch/alpha/osfpal.hh"
58#endif
59
60class BaseCPUParams;
61
62using namespace TheISA;
63
64BaseO3CPU::BaseO3CPU(BaseCPUParams *params)
65 : BaseCPU(params), cpu_id(0)
66{
67}
68
69void
70BaseO3CPU::regStats()
71{
72 BaseCPU::regStats();
73}
74
75template <class Impl>
76FullO3CPU<Impl>::TickEvent::TickEvent(FullO3CPU<Impl> *c)
77 : Event(CPU_Tick_Pri), cpu(c)
78{
79}
80
81template <class Impl>
82void
83FullO3CPU<Impl>::TickEvent::process()
84{
85 cpu->tick();
86}
87
88template <class Impl>
89const char *
90FullO3CPU<Impl>::TickEvent::description() const
91{
92 return "FullO3CPU tick";
93}
94
95template <class Impl>
96FullO3CPU<Impl>::ActivateThreadEvent::ActivateThreadEvent()
97 : Event(CPU_Switch_Pri)
98{
99}
100
101template <class Impl>
102void
103FullO3CPU<Impl>::ActivateThreadEvent::init(int thread_num,
104 FullO3CPU<Impl> *thread_cpu)
105{
106 tid = thread_num;
107 cpu = thread_cpu;
108}
109
110template <class Impl>
111void
112FullO3CPU<Impl>::ActivateThreadEvent::process()
113{
114 cpu->activateThread(tid);
115}
116
117template <class Impl>
118const char *
119FullO3CPU<Impl>::ActivateThreadEvent::description() const
120{
121 return "FullO3CPU \"Activate Thread\"";
122}
123
124template <class Impl>
125FullO3CPU<Impl>::DeallocateContextEvent::DeallocateContextEvent()
126 : Event(CPU_Tick_Pri), tid(0), remove(false), cpu(NULL)
127{
128}
129
130template <class Impl>
131void
132FullO3CPU<Impl>::DeallocateContextEvent::init(int thread_num,
133 FullO3CPU<Impl> *thread_cpu)
134{
135 tid = thread_num;
136 cpu = thread_cpu;
137 remove = false;
138}
139
140template <class Impl>
141void
142FullO3CPU<Impl>::DeallocateContextEvent::process()
143{
144 cpu->deactivateThread(tid);
145 if (remove)
146 cpu->removeThread(tid);
147}
148
149template <class Impl>
150const char *
151FullO3CPU<Impl>::DeallocateContextEvent::description() const
152{
153 return "FullO3CPU \"Deallocate Context\"";
154}
155
156template <class Impl>
157FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params)
158 : BaseO3CPU(params),
159 itb(params->itb),
160 dtb(params->dtb),
161 tickEvent(this),
162 removeInstsThisCycle(false),
163 fetch(this, params),
164 decode(this, params),
165 rename(this, params),
166 iew(this, params),
167 commit(this, params),
168
169 regFile(this, params->numPhysIntRegs,
170 params->numPhysFloatRegs),
171
172 freeList(params->numThreads,
173 TheISA::NumIntRegs, params->numPhysIntRegs,
174 TheISA::NumFloatRegs, params->numPhysFloatRegs),
175
176 rob(this,
177 params->numROBEntries, params->squashWidth,
178 params->smtROBPolicy, params->smtROBThreshold,
179 params->numThreads),
180
181 scoreboard(params->numThreads,
182 TheISA::NumIntRegs, params->numPhysIntRegs,
183 TheISA::NumFloatRegs, params->numPhysFloatRegs,
184 TheISA::NumMiscRegs * number_of_threads,
185 TheISA::ZeroReg),
186
187 timeBuffer(params->backComSize, params->forwardComSize),
188 fetchQueue(params->backComSize, params->forwardComSize),
189 decodeQueue(params->backComSize, params->forwardComSize),
190 renameQueue(params->backComSize, params->forwardComSize),
191 iewQueue(params->backComSize, params->forwardComSize),
192 activityRec(NumStages,
193 params->backComSize + params->forwardComSize,
194 params->activity),
195
196 globalSeqNum(1),
197#if FULL_SYSTEM
198 system(params->system),
199 physmem(system->physmem),
200#endif // FULL_SYSTEM
201 drainCount(0),
202 deferRegistration(params->defer_registration),
203 numThreads(number_of_threads)
204{
205 if (!deferRegistration) {
206 _status = Running;
207 } else {
208 _status = Idle;
209 }
210
211#if USE_CHECKER
212 if (params->checker) {
213 BaseCPU *temp_checker = params->checker;
214 checker = dynamic_cast<Checker<DynInstPtr> *>(temp_checker);
215#if FULL_SYSTEM
216 checker->setSystem(params->system);
217#endif
218 } else {
219 checker = NULL;
220 }
221#endif // USE_CHECKER
222
223#if !FULL_SYSTEM
224 thread.resize(number_of_threads);
225 tids.resize(number_of_threads);
226#endif
227
228 // The stages also need their CPU pointer setup. However this
229 // must be done at the upper level CPU because they have pointers
230 // to the upper level CPU, and not this FullO3CPU.
231
232 // Set up Pointers to the activeThreads list for each stage
233 fetch.setActiveThreads(&activeThreads);
234 decode.setActiveThreads(&activeThreads);
235 rename.setActiveThreads(&activeThreads);
236 iew.setActiveThreads(&activeThreads);
237 commit.setActiveThreads(&activeThreads);
238
239 // Give each of the stages the time buffer they will use.
240 fetch.setTimeBuffer(&timeBuffer);
241 decode.setTimeBuffer(&timeBuffer);
242 rename.setTimeBuffer(&timeBuffer);
243 iew.setTimeBuffer(&timeBuffer);
244 commit.setTimeBuffer(&timeBuffer);
245
246 // Also setup each of the stages' queues.
247 fetch.setFetchQueue(&fetchQueue);
248 decode.setFetchQueue(&fetchQueue);
249 commit.setFetchQueue(&fetchQueue);
250 decode.setDecodeQueue(&decodeQueue);
251 rename.setDecodeQueue(&decodeQueue);
252 rename.setRenameQueue(&renameQueue);
253 iew.setRenameQueue(&renameQueue);
254 iew.setIEWQueue(&iewQueue);
255 commit.setIEWQueue(&iewQueue);
256 commit.setRenameQueue(&renameQueue);
257
258 commit.setIEWStage(&iew);
259 rename.setIEWStage(&iew);
260 rename.setCommitStage(&commit);
261
262#if !FULL_SYSTEM
263 int active_threads = params->workload.size();
264
265 if (active_threads > Impl::MaxThreads) {
266 panic("Workload Size too large. Increase the 'MaxThreads'"
267 "constant in your O3CPU impl. file (e.g. o3/alpha/impl.hh) or "
268 "edit your workload size.");
269 }
270#else
271 int active_threads = 1;
272#endif
273
274 //Make Sure That this a Valid Architeture
275 assert(params->numPhysIntRegs >= numThreads * TheISA::NumIntRegs);
276 assert(params->numPhysFloatRegs >= numThreads * TheISA::NumFloatRegs);
277
278 rename.setScoreboard(&scoreboard);
279 iew.setScoreboard(&scoreboard);
280
281 // Setup the rename map for whichever stages need it.
282 PhysRegIndex lreg_idx = 0;
283 PhysRegIndex freg_idx = params->numPhysIntRegs; //Index to 1 after int regs
284
285 for (int tid=0; tid < numThreads; tid++) {
286 bool bindRegs = (tid <= active_threads - 1);
287
288 commitRenameMap[tid].init(TheISA::NumIntRegs,
289 params->numPhysIntRegs,
290 lreg_idx, //Index for Logical. Regs
291
292 TheISA::NumFloatRegs,
293 params->numPhysFloatRegs,
294 freg_idx, //Index for Float Regs
295
296 TheISA::NumMiscRegs,
297
298 TheISA::ZeroReg,
299 TheISA::ZeroReg,
300
301 tid,
302 false);
303
304 renameMap[tid].init(TheISA::NumIntRegs,
305 params->numPhysIntRegs,
306 lreg_idx, //Index for Logical. Regs
307
308 TheISA::NumFloatRegs,
309 params->numPhysFloatRegs,
310 freg_idx, //Index for Float Regs
311
312 TheISA::NumMiscRegs,
313
314 TheISA::ZeroReg,
315 TheISA::ZeroReg,
316
317 tid,
318 bindRegs);
319
320 activateThreadEvent[tid].init(tid, this);
321 deallocateContextEvent[tid].init(tid, this);
322 }
323
324 rename.setRenameMap(renameMap);
325 commit.setRenameMap(commitRenameMap);
326
327 // Give renameMap & rename stage access to the freeList;
328 for (int i=0; i < numThreads; i++) {
329 renameMap[i].setFreeList(&freeList);
330 }
331 rename.setFreeList(&freeList);
332
333 // Setup the ROB for whichever stages need it.
334 commit.setROB(&rob);
335
336 lastRunningCycle = curTick;
337
338 lastActivatedCycle = -1;
339
340 // Give renameMap & rename stage access to the freeList;
341 //for (int i=0; i < numThreads; i++) {
342 //globalSeqNum[i] = 1;
343 //}
344
345 contextSwitch = false;
346 DPRINTF(O3CPU, "Creating O3CPU object.\n");
347
348 // Setup any thread state.
349 this->thread.resize(this->numThreads);
350
351 for (int i = 0; i < this->numThreads; ++i) {
352#if FULL_SYSTEM
353 // SMT is not supported in FS mode yet.
354 assert(this->numThreads == 1);
355 this->thread[i] = new Thread(this, 0);
356 this->thread[i]->setStatus(ThreadContext::Suspended);
357#else
358 if (i < params->workload.size()) {
359 DPRINTF(O3CPU, "Workload[%i] process is %#x",
360 i, this->thread[i]);
361 this->thread[i] = new typename FullO3CPU<Impl>::Thread(
362 (typename Impl::O3CPU *)(this),
363 i, params->workload[i], i);
364
365 this->thread[i]->setStatus(ThreadContext::Suspended);
366
367 //usedTids[i] = true;
368 //threadMap[i] = i;
369 } else {
370 //Allocate Empty thread so M5 can use later
371 //when scheduling threads to CPU
372 Process* dummy_proc = NULL;
373
374 this->thread[i] = new typename FullO3CPU<Impl>::Thread(
375 (typename Impl::O3CPU *)(this),
376 i, dummy_proc, i);
377 //usedTids[i] = false;
378 }
379#endif // !FULL_SYSTEM
380
381 ThreadContext *tc;
382
383 // Setup the TC that will serve as the interface to the threads/CPU.
384 O3ThreadContext<Impl> *o3_tc = new O3ThreadContext<Impl>;
385
386 tc = o3_tc;
387
388 // If we're using a checker, then the TC should be the
389 // CheckerThreadContext.
390#if USE_CHECKER
391 if (params->checker) {
392 tc = new CheckerThreadContext<O3ThreadContext<Impl> >(
393 o3_tc, this->checker);
394 }
395#endif
396
397 o3_tc->cpu = (typename Impl::O3CPU *)(this);
398 assert(o3_tc->cpu);
399 o3_tc->thread = this->thread[i];
400
401#if FULL_SYSTEM
402 // Setup quiesce event.
403 this->thread[i]->quiesceEvent = new EndQuiesceEvent(tc);
404#endif
405 // Give the thread the TC.
406 this->thread[i]->tc = tc;
407 this->thread[i]->setCpuId(params->cpu_id);
408
409 // Add the TC to the CPU's list of TC's.
410 this->threadContexts.push_back(tc);
411 }
412
413 for (int i=0; i < this->numThreads; i++) {
414 this->thread[i]->setFuncExeInst(0);
415 }
416
417 lockAddr = 0;
418 lockFlag = false;
419}
420
421#if !FULL_SYSTEM
422
423template <class Impl>
424TheISA::IntReg
425FullO3CPU<Impl>::getSyscallArg(int i, int tid)
426{
427 assert(i < TheISA::NumArgumentRegs);
428 TheISA::IntReg idx = TheISA::flattenIntIndex(this->tcBase(tid),
429 TheISA::ArgumentReg[i]);
430 TheISA::IntReg val = this->readArchIntReg(idx, tid);
431#if THE_ISA == SPARC_ISA
432 if (bits(this->readMiscRegNoEffect(SparcISA::MISCREG_PSTATE, tid), 3, 3))
433 val = bits(val, 31, 0);
434#endif
435 return val;
436}
437
438template <class Impl>
439void
440FullO3CPU<Impl>::setSyscallArg(int i, TheISA::IntReg val, int tid)
441{
442 assert(i < TheISA::NumArgumentRegs);
443 TheISA::IntReg idx = TheISA::flattenIntIndex(this->tcBase(tid),
444 TheISA::ArgumentReg[i]);
445 this->setArchIntReg(idx, val, tid);
446}
447#endif
448
449template <class Impl>
450FullO3CPU<Impl>::~FullO3CPU()
451{
452}
453
454template <class Impl>
455void
456FullO3CPU<Impl>::regStats()
457{
458 BaseO3CPU::regStats();
459
460 // Register any of the O3CPU's stats here.
461 timesIdled
462 .name(name() + ".timesIdled")
463 .desc("Number of times that the entire CPU went into an idle state and"
464 " unscheduled itself")
465 .prereq(timesIdled);
466
467 idleCycles
468 .name(name() + ".idleCycles")
469 .desc("Total number of cycles that the CPU has spent unscheduled due "
470 "to idling")
471 .prereq(idleCycles);
472
473 // Number of Instructions simulated
474 // --------------------------------
475 // Should probably be in Base CPU but need templated
476 // MaxThreads so put in here instead
477 committedInsts
478 .init(numThreads)
479 .name(name() + ".committedInsts")
480 .desc("Number of Instructions Simulated");
481
482 totalCommittedInsts
483 .name(name() + ".committedInsts_total")
484 .desc("Number of Instructions Simulated");
485
486 cpi
487 .name(name() + ".cpi")
488 .desc("CPI: Cycles Per Instruction")
489 .precision(6);
490 cpi = numCycles / committedInsts;
491
492 totalCpi
493 .name(name() + ".cpi_total")
494 .desc("CPI: Total CPI of All Threads")
495 .precision(6);
496 totalCpi = numCycles / totalCommittedInsts;
497
498 ipc
499 .name(name() + ".ipc")
500 .desc("IPC: Instructions Per Cycle")
501 .precision(6);
502 ipc = committedInsts / numCycles;
503
504 totalIpc
505 .name(name() + ".ipc_total")
506 .desc("IPC: Total IPC of All Threads")
507 .precision(6);
508 totalIpc = totalCommittedInsts / numCycles;
509
510 this->fetch.regStats();
511 this->decode.regStats();
512 this->rename.regStats();
513 this->iew.regStats();
514 this->commit.regStats();
515}
516
517template <class Impl>
518Port *
519FullO3CPU<Impl>::getPort(const std::string &if_name, int idx)
520{
521 if (if_name == "dcache_port")
522 return iew.getDcachePort();
523 else if (if_name == "icache_port")
524 return fetch.getIcachePort();
525 else
526 panic("No Such Port\n");
527}
528
529template <class Impl>
530void
531FullO3CPU<Impl>::tick()
532{
533 DPRINTF(O3CPU, "\n\nFullO3CPU: Ticking main, FullO3CPU.\n");
534
535 ++numCycles;
536
537// activity = false;
538
539 //Tick each of the stages
540 fetch.tick();
541
542 decode.tick();
543
544 rename.tick();
545
546 iew.tick();
547
548 commit.tick();
549
550#if !FULL_SYSTEM
551 doContextSwitch();
552#endif
553
554 // Now advance the time buffers
555 timeBuffer.advance();
556
557 fetchQueue.advance();
558 decodeQueue.advance();
559 renameQueue.advance();
560 iewQueue.advance();
561
562 activityRec.advance();
563
564 if (removeInstsThisCycle) {
565 cleanUpRemovedInsts();
566 }
567
568 if (!tickEvent.scheduled()) {
569 if (_status == SwitchedOut ||
570 getState() == SimObject::Drained) {
571 DPRINTF(O3CPU, "Switched out!\n");
572 // increment stat
573 lastRunningCycle = curTick;
574 } else if (!activityRec.active() || _status == Idle) {
575 DPRINTF(O3CPU, "Idle!\n");
576 lastRunningCycle = curTick;
577 timesIdled++;
578 } else {
579 schedule(tickEvent, nextCycle(curTick + ticks(1)));
580 DPRINTF(O3CPU, "Scheduling next tick!\n");
581 }
582 }
583
584#if !FULL_SYSTEM
585 updateThreadPriority();
586#endif
587}
588
589template <class Impl>
590void
591FullO3CPU<Impl>::init()
592{
593 if (!deferRegistration) {
594 registerThreadContexts();
595 }
596
597 // Set inSyscall so that the CPU doesn't squash when initially
598 // setting up registers.
599 for (int i = 0; i < number_of_threads; ++i)
600 thread[i]->inSyscall = true;
601
602 for (int tid=0; tid < number_of_threads; tid++) {
603#if FULL_SYSTEM
604 ThreadContext *src_tc = threadContexts[tid];
605#else
606 ThreadContext *src_tc = thread[tid]->getTC();
607#endif
608 // Threads start in the Suspended State
609 if (src_tc->status() != ThreadContext::Suspended) {
610 continue;
611 }
612
613#if FULL_SYSTEM
614 TheISA::initCPU(src_tc, src_tc->readCpuId());
615#endif
616 }
617
618 // Clear inSyscall.
619 for (int i = 0; i < number_of_threads; ++i)
620 thread[i]->inSyscall = false;
621
622 // Initialize stages.
623 fetch.initStage();
624 iew.initStage();
625 rename.initStage();
626 commit.initStage();
627
628 commit.setThreads(thread);
629}
630
631template <class Impl>
632void
633FullO3CPU<Impl>::activateThread(unsigned tid)
634{
635 std::list<unsigned>::iterator isActive =
636 std::find(activeThreads.begin(), activeThreads.end(), tid);
637
638 DPRINTF(O3CPU, "[tid:%i]: Calling activate thread.\n", tid);
639
640 if (isActive == activeThreads.end()) {
641 DPRINTF(O3CPU, "[tid:%i]: Adding to active threads list\n",
642 tid);
643
644 activeThreads.push_back(tid);
645 }
646}
647
648template <class Impl>
649void
650FullO3CPU<Impl>::deactivateThread(unsigned tid)
651{
652 //Remove From Active List, if Active
653 std::list<unsigned>::iterator thread_it =
654 std::find(activeThreads.begin(), activeThreads.end(), tid);
655
656 DPRINTF(O3CPU, "[tid:%i]: Calling deactivate thread.\n", tid);
657
658 if (thread_it != activeThreads.end()) {
659 DPRINTF(O3CPU,"[tid:%i]: Removing from active threads list\n",
660 tid);
661 activeThreads.erase(thread_it);
662 }
663}
664
665template <class Impl>
666void
667FullO3CPU<Impl>::activateContext(int tid, int delay)
668{
669 // Needs to set each stage to running as well.
670 if (delay){
671 DPRINTF(O3CPU, "[tid:%i]: Scheduling thread context to activate "
672 "on cycle %d\n", tid, curTick + ticks(delay));
673 scheduleActivateThreadEvent(tid, delay);
674 } else {
675 activateThread(tid);
676 }
677
678 if (lastActivatedCycle < curTick) {
679 scheduleTickEvent(delay);
680
681 // Be sure to signal that there's some activity so the CPU doesn't
682 // deschedule itself.
683 activityRec.activity();
684 fetch.wakeFromQuiesce();
685
686 lastActivatedCycle = curTick;
687
688 _status = Running;
689 }
690}
691
692template <class Impl>
693bool
694FullO3CPU<Impl>::deallocateContext(int tid, bool remove, int delay)
695{
696 // Schedule removal of thread data from CPU
697 if (delay){
698 DPRINTF(O3CPU, "[tid:%i]: Scheduling thread context to deallocate "
699 "on cycle %d\n", tid, curTick + ticks(delay));
700 scheduleDeallocateContextEvent(tid, remove, delay);
701 return false;
702 } else {
703 deactivateThread(tid);
704 if (remove)
705 removeThread(tid);
706 return true;
707 }
708}
709
710template <class Impl>
711void
712FullO3CPU<Impl>::suspendContext(int tid)
713{
714 DPRINTF(O3CPU,"[tid: %i]: Suspending Thread Context.\n", tid);
715 bool deallocated = deallocateContext(tid, false, 1);
716 // If this was the last thread then unschedule the tick event.
717 if ((activeThreads.size() == 1 && !deallocated) ||
718 activeThreads.size() == 0)
719 unscheduleTickEvent();
720 _status = Idle;
721}
722
723template <class Impl>
724void
725FullO3CPU<Impl>::haltContext(int tid)
726{
727 //For now, this is the same as deallocate
728 DPRINTF(O3CPU,"[tid:%i]: Halt Context called. Deallocating", tid);
729 deallocateContext(tid, true, 1);
730}
731
732template <class Impl>
733void
734FullO3CPU<Impl>::insertThread(unsigned tid)
735{
736 DPRINTF(O3CPU,"[tid:%i] Initializing thread into CPU");
737 // Will change now that the PC and thread state is internal to the CPU
738 // and not in the ThreadContext.
739#if FULL_SYSTEM
740 ThreadContext *src_tc = system->threadContexts[tid];
741#else
742 ThreadContext *src_tc = tcBase(tid);
743#endif
744
745 //Bind Int Regs to Rename Map
746 for (int ireg = 0; ireg < TheISA::NumIntRegs; ireg++) {
747 PhysRegIndex phys_reg = freeList.getIntReg();
748
749 renameMap[tid].setEntry(ireg,phys_reg);
750 scoreboard.setReg(phys_reg);
751 }
752
753 //Bind Float Regs to Rename Map
754 for (int freg = 0; freg < TheISA::NumFloatRegs; freg++) {
755 PhysRegIndex phys_reg = freeList.getFloatReg();
756
757 renameMap[tid].setEntry(freg,phys_reg);
758 scoreboard.setReg(phys_reg);
759 }
760
761 //Copy Thread Data Into RegFile
762 //this->copyFromTC(tid);
763
764 //Set PC/NPC/NNPC
765 setPC(src_tc->readPC(), tid);
766 setNextPC(src_tc->readNextPC(), tid);
767 setNextNPC(src_tc->readNextNPC(), tid);
768
769 src_tc->setStatus(ThreadContext::Active);
770
771 activateContext(tid,1);
772
773 //Reset ROB/IQ/LSQ Entries
774 commit.rob->resetEntries();
775 iew.resetEntries();
776}
777
778template <class Impl>
779void
780FullO3CPU<Impl>::removeThread(unsigned tid)
781{
782 DPRINTF(O3CPU,"[tid:%i] Removing thread context from CPU.\n", tid);
783
784 // Copy Thread Data From RegFile
785 // If thread is suspended, it might be re-allocated
786 // this->copyToTC(tid);
787
788
789 // @todo: 2-27-2008: Fix how we free up rename mappings
790 // here to alleviate the case for double-freeing registers
791 // in SMT workloads.
792
793 // Unbind Int Regs from Rename Map
794 for (int ireg = 0; ireg < TheISA::NumIntRegs; ireg++) {
795 PhysRegIndex phys_reg = renameMap[tid].lookup(ireg);
796
797 scoreboard.unsetReg(phys_reg);
798 freeList.addReg(phys_reg);
799 }
800
801 // Unbind Float Regs from Rename Map
802 for (int freg = TheISA::NumIntRegs; freg < TheISA::NumFloatRegs; freg++) {
803 PhysRegIndex phys_reg = renameMap[tid].lookup(freg);
804
805 scoreboard.unsetReg(phys_reg);
806 freeList.addReg(phys_reg);
807 }
808
809 // Squash Throughout Pipeline
810 InstSeqNum squash_seq_num = commit.rob->readHeadInst(tid)->seqNum;
811 fetch.squash(0, sizeof(TheISA::MachInst), 0, squash_seq_num, tid);
812 decode.squash(tid);
813 rename.squash(squash_seq_num, tid);
814 iew.squash(tid);
815 iew.ldstQueue.squash(squash_seq_num, tid);
816 commit.rob->squash(squash_seq_num, tid);
817
818
819 assert(iew.instQueue.getCount(tid) == 0);
820 assert(iew.ldstQueue.getCount(tid) == 0);
821
822 // Reset ROB/IQ/LSQ Entries
823
824 // Commented out for now. This should be possible to do by
825 // telling all the pipeline stages to drain first, and then
826 // checking until the drain completes. Once the pipeline is
827 // drained, call resetEntries(). - 10-09-06 ktlim
828/*
829 if (activeThreads.size() >= 1) {
830 commit.rob->resetEntries();
831 iew.resetEntries();
832 }
833*/
834}
835
836
837template <class Impl>
838void
839FullO3CPU<Impl>::activateWhenReady(int tid)
840{
841 DPRINTF(O3CPU,"[tid:%i]: Checking if resources are available for incoming"
842 "(e.g. PhysRegs/ROB/IQ/LSQ) \n",
843 tid);
844
845 bool ready = true;
846
847 if (freeList.numFreeIntRegs() >= TheISA::NumIntRegs) {
848 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
849 "Phys. Int. Regs.\n",
850 tid);
851 ready = false;
852 } else if (freeList.numFreeFloatRegs() >= TheISA::NumFloatRegs) {
853 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
854 "Phys. Float. Regs.\n",
855 tid);
856 ready = false;
857 } else if (commit.rob->numFreeEntries() >=
858 commit.rob->entryAmount(activeThreads.size() + 1)) {
859 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
860 "ROB entries.\n",
861 tid);
862 ready = false;
863 } else if (iew.instQueue.numFreeEntries() >=
864 iew.instQueue.entryAmount(activeThreads.size() + 1)) {
865 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
866 "IQ entries.\n",
867 tid);
868 ready = false;
869 } else if (iew.ldstQueue.numFreeEntries() >=
870 iew.ldstQueue.entryAmount(activeThreads.size() + 1)) {
871 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
872 "LSQ entries.\n",
873 tid);
874 ready = false;
875 }
876
877 if (ready) {
878 insertThread(tid);
879
880 contextSwitch = false;
881
882 cpuWaitList.remove(tid);
883 } else {
884 suspendContext(tid);
885
886 //blocks fetch
887 contextSwitch = true;
888
889 //@todo: dont always add to waitlist
890 //do waitlist
891 cpuWaitList.push_back(tid);
892 }
893}
894
895#if FULL_SYSTEM
896template <class Impl>
897void
1/*
2 * Copyright (c) 2004-2006 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: Kevin Lim
29 * Korey Sewell
30 */
31
32#include "config/full_system.hh"
33#include "config/use_checker.hh"
34
35#include "cpu/activity.hh"
36#include "cpu/simple_thread.hh"
37#include "cpu/thread_context.hh"
38#include "cpu/o3/isa_specific.hh"
39#include "cpu/o3/cpu.hh"
40#include "cpu/o3/thread_context.hh"
41#include "enums/MemoryMode.hh"
42#include "sim/core.hh"
43#include "sim/stat_control.hh"
44
45#if FULL_SYSTEM
46#include "cpu/quiesce_event.hh"
47#include "sim/system.hh"
48#else
49#include "sim/process.hh"
50#endif
51
52#if USE_CHECKER
53#include "cpu/checker/cpu.hh"
54#endif
55
56#if THE_ISA == ALPHA_ISA
57#include "arch/alpha/osfpal.hh"
58#endif
59
60class BaseCPUParams;
61
62using namespace TheISA;
63
64BaseO3CPU::BaseO3CPU(BaseCPUParams *params)
65 : BaseCPU(params), cpu_id(0)
66{
67}
68
69void
70BaseO3CPU::regStats()
71{
72 BaseCPU::regStats();
73}
74
75template <class Impl>
76FullO3CPU<Impl>::TickEvent::TickEvent(FullO3CPU<Impl> *c)
77 : Event(CPU_Tick_Pri), cpu(c)
78{
79}
80
81template <class Impl>
82void
83FullO3CPU<Impl>::TickEvent::process()
84{
85 cpu->tick();
86}
87
88template <class Impl>
89const char *
90FullO3CPU<Impl>::TickEvent::description() const
91{
92 return "FullO3CPU tick";
93}
94
95template <class Impl>
96FullO3CPU<Impl>::ActivateThreadEvent::ActivateThreadEvent()
97 : Event(CPU_Switch_Pri)
98{
99}
100
101template <class Impl>
102void
103FullO3CPU<Impl>::ActivateThreadEvent::init(int thread_num,
104 FullO3CPU<Impl> *thread_cpu)
105{
106 tid = thread_num;
107 cpu = thread_cpu;
108}
109
110template <class Impl>
111void
112FullO3CPU<Impl>::ActivateThreadEvent::process()
113{
114 cpu->activateThread(tid);
115}
116
117template <class Impl>
118const char *
119FullO3CPU<Impl>::ActivateThreadEvent::description() const
120{
121 return "FullO3CPU \"Activate Thread\"";
122}
123
124template <class Impl>
125FullO3CPU<Impl>::DeallocateContextEvent::DeallocateContextEvent()
126 : Event(CPU_Tick_Pri), tid(0), remove(false), cpu(NULL)
127{
128}
129
130template <class Impl>
131void
132FullO3CPU<Impl>::DeallocateContextEvent::init(int thread_num,
133 FullO3CPU<Impl> *thread_cpu)
134{
135 tid = thread_num;
136 cpu = thread_cpu;
137 remove = false;
138}
139
140template <class Impl>
141void
142FullO3CPU<Impl>::DeallocateContextEvent::process()
143{
144 cpu->deactivateThread(tid);
145 if (remove)
146 cpu->removeThread(tid);
147}
148
149template <class Impl>
150const char *
151FullO3CPU<Impl>::DeallocateContextEvent::description() const
152{
153 return "FullO3CPU \"Deallocate Context\"";
154}
155
156template <class Impl>
157FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params)
158 : BaseO3CPU(params),
159 itb(params->itb),
160 dtb(params->dtb),
161 tickEvent(this),
162 removeInstsThisCycle(false),
163 fetch(this, params),
164 decode(this, params),
165 rename(this, params),
166 iew(this, params),
167 commit(this, params),
168
169 regFile(this, params->numPhysIntRegs,
170 params->numPhysFloatRegs),
171
172 freeList(params->numThreads,
173 TheISA::NumIntRegs, params->numPhysIntRegs,
174 TheISA::NumFloatRegs, params->numPhysFloatRegs),
175
176 rob(this,
177 params->numROBEntries, params->squashWidth,
178 params->smtROBPolicy, params->smtROBThreshold,
179 params->numThreads),
180
181 scoreboard(params->numThreads,
182 TheISA::NumIntRegs, params->numPhysIntRegs,
183 TheISA::NumFloatRegs, params->numPhysFloatRegs,
184 TheISA::NumMiscRegs * number_of_threads,
185 TheISA::ZeroReg),
186
187 timeBuffer(params->backComSize, params->forwardComSize),
188 fetchQueue(params->backComSize, params->forwardComSize),
189 decodeQueue(params->backComSize, params->forwardComSize),
190 renameQueue(params->backComSize, params->forwardComSize),
191 iewQueue(params->backComSize, params->forwardComSize),
192 activityRec(NumStages,
193 params->backComSize + params->forwardComSize,
194 params->activity),
195
196 globalSeqNum(1),
197#if FULL_SYSTEM
198 system(params->system),
199 physmem(system->physmem),
200#endif // FULL_SYSTEM
201 drainCount(0),
202 deferRegistration(params->defer_registration),
203 numThreads(number_of_threads)
204{
205 if (!deferRegistration) {
206 _status = Running;
207 } else {
208 _status = Idle;
209 }
210
211#if USE_CHECKER
212 if (params->checker) {
213 BaseCPU *temp_checker = params->checker;
214 checker = dynamic_cast<Checker<DynInstPtr> *>(temp_checker);
215#if FULL_SYSTEM
216 checker->setSystem(params->system);
217#endif
218 } else {
219 checker = NULL;
220 }
221#endif // USE_CHECKER
222
223#if !FULL_SYSTEM
224 thread.resize(number_of_threads);
225 tids.resize(number_of_threads);
226#endif
227
228 // The stages also need their CPU pointer setup. However this
229 // must be done at the upper level CPU because they have pointers
230 // to the upper level CPU, and not this FullO3CPU.
231
232 // Set up Pointers to the activeThreads list for each stage
233 fetch.setActiveThreads(&activeThreads);
234 decode.setActiveThreads(&activeThreads);
235 rename.setActiveThreads(&activeThreads);
236 iew.setActiveThreads(&activeThreads);
237 commit.setActiveThreads(&activeThreads);
238
239 // Give each of the stages the time buffer they will use.
240 fetch.setTimeBuffer(&timeBuffer);
241 decode.setTimeBuffer(&timeBuffer);
242 rename.setTimeBuffer(&timeBuffer);
243 iew.setTimeBuffer(&timeBuffer);
244 commit.setTimeBuffer(&timeBuffer);
245
246 // Also setup each of the stages' queues.
247 fetch.setFetchQueue(&fetchQueue);
248 decode.setFetchQueue(&fetchQueue);
249 commit.setFetchQueue(&fetchQueue);
250 decode.setDecodeQueue(&decodeQueue);
251 rename.setDecodeQueue(&decodeQueue);
252 rename.setRenameQueue(&renameQueue);
253 iew.setRenameQueue(&renameQueue);
254 iew.setIEWQueue(&iewQueue);
255 commit.setIEWQueue(&iewQueue);
256 commit.setRenameQueue(&renameQueue);
257
258 commit.setIEWStage(&iew);
259 rename.setIEWStage(&iew);
260 rename.setCommitStage(&commit);
261
262#if !FULL_SYSTEM
263 int active_threads = params->workload.size();
264
265 if (active_threads > Impl::MaxThreads) {
266 panic("Workload Size too large. Increase the 'MaxThreads'"
267 "constant in your O3CPU impl. file (e.g. o3/alpha/impl.hh) or "
268 "edit your workload size.");
269 }
270#else
271 int active_threads = 1;
272#endif
273
274 //Make Sure That this a Valid Architeture
275 assert(params->numPhysIntRegs >= numThreads * TheISA::NumIntRegs);
276 assert(params->numPhysFloatRegs >= numThreads * TheISA::NumFloatRegs);
277
278 rename.setScoreboard(&scoreboard);
279 iew.setScoreboard(&scoreboard);
280
281 // Setup the rename map for whichever stages need it.
282 PhysRegIndex lreg_idx = 0;
283 PhysRegIndex freg_idx = params->numPhysIntRegs; //Index to 1 after int regs
284
285 for (int tid=0; tid < numThreads; tid++) {
286 bool bindRegs = (tid <= active_threads - 1);
287
288 commitRenameMap[tid].init(TheISA::NumIntRegs,
289 params->numPhysIntRegs,
290 lreg_idx, //Index for Logical. Regs
291
292 TheISA::NumFloatRegs,
293 params->numPhysFloatRegs,
294 freg_idx, //Index for Float Regs
295
296 TheISA::NumMiscRegs,
297
298 TheISA::ZeroReg,
299 TheISA::ZeroReg,
300
301 tid,
302 false);
303
304 renameMap[tid].init(TheISA::NumIntRegs,
305 params->numPhysIntRegs,
306 lreg_idx, //Index for Logical. Regs
307
308 TheISA::NumFloatRegs,
309 params->numPhysFloatRegs,
310 freg_idx, //Index for Float Regs
311
312 TheISA::NumMiscRegs,
313
314 TheISA::ZeroReg,
315 TheISA::ZeroReg,
316
317 tid,
318 bindRegs);
319
320 activateThreadEvent[tid].init(tid, this);
321 deallocateContextEvent[tid].init(tid, this);
322 }
323
324 rename.setRenameMap(renameMap);
325 commit.setRenameMap(commitRenameMap);
326
327 // Give renameMap & rename stage access to the freeList;
328 for (int i=0; i < numThreads; i++) {
329 renameMap[i].setFreeList(&freeList);
330 }
331 rename.setFreeList(&freeList);
332
333 // Setup the ROB for whichever stages need it.
334 commit.setROB(&rob);
335
336 lastRunningCycle = curTick;
337
338 lastActivatedCycle = -1;
339
340 // Give renameMap & rename stage access to the freeList;
341 //for (int i=0; i < numThreads; i++) {
342 //globalSeqNum[i] = 1;
343 //}
344
345 contextSwitch = false;
346 DPRINTF(O3CPU, "Creating O3CPU object.\n");
347
348 // Setup any thread state.
349 this->thread.resize(this->numThreads);
350
351 for (int i = 0; i < this->numThreads; ++i) {
352#if FULL_SYSTEM
353 // SMT is not supported in FS mode yet.
354 assert(this->numThreads == 1);
355 this->thread[i] = new Thread(this, 0);
356 this->thread[i]->setStatus(ThreadContext::Suspended);
357#else
358 if (i < params->workload.size()) {
359 DPRINTF(O3CPU, "Workload[%i] process is %#x",
360 i, this->thread[i]);
361 this->thread[i] = new typename FullO3CPU<Impl>::Thread(
362 (typename Impl::O3CPU *)(this),
363 i, params->workload[i], i);
364
365 this->thread[i]->setStatus(ThreadContext::Suspended);
366
367 //usedTids[i] = true;
368 //threadMap[i] = i;
369 } else {
370 //Allocate Empty thread so M5 can use later
371 //when scheduling threads to CPU
372 Process* dummy_proc = NULL;
373
374 this->thread[i] = new typename FullO3CPU<Impl>::Thread(
375 (typename Impl::O3CPU *)(this),
376 i, dummy_proc, i);
377 //usedTids[i] = false;
378 }
379#endif // !FULL_SYSTEM
380
381 ThreadContext *tc;
382
383 // Setup the TC that will serve as the interface to the threads/CPU.
384 O3ThreadContext<Impl> *o3_tc = new O3ThreadContext<Impl>;
385
386 tc = o3_tc;
387
388 // If we're using a checker, then the TC should be the
389 // CheckerThreadContext.
390#if USE_CHECKER
391 if (params->checker) {
392 tc = new CheckerThreadContext<O3ThreadContext<Impl> >(
393 o3_tc, this->checker);
394 }
395#endif
396
397 o3_tc->cpu = (typename Impl::O3CPU *)(this);
398 assert(o3_tc->cpu);
399 o3_tc->thread = this->thread[i];
400
401#if FULL_SYSTEM
402 // Setup quiesce event.
403 this->thread[i]->quiesceEvent = new EndQuiesceEvent(tc);
404#endif
405 // Give the thread the TC.
406 this->thread[i]->tc = tc;
407 this->thread[i]->setCpuId(params->cpu_id);
408
409 // Add the TC to the CPU's list of TC's.
410 this->threadContexts.push_back(tc);
411 }
412
413 for (int i=0; i < this->numThreads; i++) {
414 this->thread[i]->setFuncExeInst(0);
415 }
416
417 lockAddr = 0;
418 lockFlag = false;
419}
420
421#if !FULL_SYSTEM
422
423template <class Impl>
424TheISA::IntReg
425FullO3CPU<Impl>::getSyscallArg(int i, int tid)
426{
427 assert(i < TheISA::NumArgumentRegs);
428 TheISA::IntReg idx = TheISA::flattenIntIndex(this->tcBase(tid),
429 TheISA::ArgumentReg[i]);
430 TheISA::IntReg val = this->readArchIntReg(idx, tid);
431#if THE_ISA == SPARC_ISA
432 if (bits(this->readMiscRegNoEffect(SparcISA::MISCREG_PSTATE, tid), 3, 3))
433 val = bits(val, 31, 0);
434#endif
435 return val;
436}
437
438template <class Impl>
439void
440FullO3CPU<Impl>::setSyscallArg(int i, TheISA::IntReg val, int tid)
441{
442 assert(i < TheISA::NumArgumentRegs);
443 TheISA::IntReg idx = TheISA::flattenIntIndex(this->tcBase(tid),
444 TheISA::ArgumentReg[i]);
445 this->setArchIntReg(idx, val, tid);
446}
447#endif
448
449template <class Impl>
450FullO3CPU<Impl>::~FullO3CPU()
451{
452}
453
454template <class Impl>
455void
456FullO3CPU<Impl>::regStats()
457{
458 BaseO3CPU::regStats();
459
460 // Register any of the O3CPU's stats here.
461 timesIdled
462 .name(name() + ".timesIdled")
463 .desc("Number of times that the entire CPU went into an idle state and"
464 " unscheduled itself")
465 .prereq(timesIdled);
466
467 idleCycles
468 .name(name() + ".idleCycles")
469 .desc("Total number of cycles that the CPU has spent unscheduled due "
470 "to idling")
471 .prereq(idleCycles);
472
473 // Number of Instructions simulated
474 // --------------------------------
475 // Should probably be in Base CPU but need templated
476 // MaxThreads so put in here instead
477 committedInsts
478 .init(numThreads)
479 .name(name() + ".committedInsts")
480 .desc("Number of Instructions Simulated");
481
482 totalCommittedInsts
483 .name(name() + ".committedInsts_total")
484 .desc("Number of Instructions Simulated");
485
486 cpi
487 .name(name() + ".cpi")
488 .desc("CPI: Cycles Per Instruction")
489 .precision(6);
490 cpi = numCycles / committedInsts;
491
492 totalCpi
493 .name(name() + ".cpi_total")
494 .desc("CPI: Total CPI of All Threads")
495 .precision(6);
496 totalCpi = numCycles / totalCommittedInsts;
497
498 ipc
499 .name(name() + ".ipc")
500 .desc("IPC: Instructions Per Cycle")
501 .precision(6);
502 ipc = committedInsts / numCycles;
503
504 totalIpc
505 .name(name() + ".ipc_total")
506 .desc("IPC: Total IPC of All Threads")
507 .precision(6);
508 totalIpc = totalCommittedInsts / numCycles;
509
510 this->fetch.regStats();
511 this->decode.regStats();
512 this->rename.regStats();
513 this->iew.regStats();
514 this->commit.regStats();
515}
516
517template <class Impl>
518Port *
519FullO3CPU<Impl>::getPort(const std::string &if_name, int idx)
520{
521 if (if_name == "dcache_port")
522 return iew.getDcachePort();
523 else if (if_name == "icache_port")
524 return fetch.getIcachePort();
525 else
526 panic("No Such Port\n");
527}
528
529template <class Impl>
530void
531FullO3CPU<Impl>::tick()
532{
533 DPRINTF(O3CPU, "\n\nFullO3CPU: Ticking main, FullO3CPU.\n");
534
535 ++numCycles;
536
537// activity = false;
538
539 //Tick each of the stages
540 fetch.tick();
541
542 decode.tick();
543
544 rename.tick();
545
546 iew.tick();
547
548 commit.tick();
549
550#if !FULL_SYSTEM
551 doContextSwitch();
552#endif
553
554 // Now advance the time buffers
555 timeBuffer.advance();
556
557 fetchQueue.advance();
558 decodeQueue.advance();
559 renameQueue.advance();
560 iewQueue.advance();
561
562 activityRec.advance();
563
564 if (removeInstsThisCycle) {
565 cleanUpRemovedInsts();
566 }
567
568 if (!tickEvent.scheduled()) {
569 if (_status == SwitchedOut ||
570 getState() == SimObject::Drained) {
571 DPRINTF(O3CPU, "Switched out!\n");
572 // increment stat
573 lastRunningCycle = curTick;
574 } else if (!activityRec.active() || _status == Idle) {
575 DPRINTF(O3CPU, "Idle!\n");
576 lastRunningCycle = curTick;
577 timesIdled++;
578 } else {
579 schedule(tickEvent, nextCycle(curTick + ticks(1)));
580 DPRINTF(O3CPU, "Scheduling next tick!\n");
581 }
582 }
583
584#if !FULL_SYSTEM
585 updateThreadPriority();
586#endif
587}
588
589template <class Impl>
590void
591FullO3CPU<Impl>::init()
592{
593 if (!deferRegistration) {
594 registerThreadContexts();
595 }
596
597 // Set inSyscall so that the CPU doesn't squash when initially
598 // setting up registers.
599 for (int i = 0; i < number_of_threads; ++i)
600 thread[i]->inSyscall = true;
601
602 for (int tid=0; tid < number_of_threads; tid++) {
603#if FULL_SYSTEM
604 ThreadContext *src_tc = threadContexts[tid];
605#else
606 ThreadContext *src_tc = thread[tid]->getTC();
607#endif
608 // Threads start in the Suspended State
609 if (src_tc->status() != ThreadContext::Suspended) {
610 continue;
611 }
612
613#if FULL_SYSTEM
614 TheISA::initCPU(src_tc, src_tc->readCpuId());
615#endif
616 }
617
618 // Clear inSyscall.
619 for (int i = 0; i < number_of_threads; ++i)
620 thread[i]->inSyscall = false;
621
622 // Initialize stages.
623 fetch.initStage();
624 iew.initStage();
625 rename.initStage();
626 commit.initStage();
627
628 commit.setThreads(thread);
629}
630
631template <class Impl>
632void
633FullO3CPU<Impl>::activateThread(unsigned tid)
634{
635 std::list<unsigned>::iterator isActive =
636 std::find(activeThreads.begin(), activeThreads.end(), tid);
637
638 DPRINTF(O3CPU, "[tid:%i]: Calling activate thread.\n", tid);
639
640 if (isActive == activeThreads.end()) {
641 DPRINTF(O3CPU, "[tid:%i]: Adding to active threads list\n",
642 tid);
643
644 activeThreads.push_back(tid);
645 }
646}
647
648template <class Impl>
649void
650FullO3CPU<Impl>::deactivateThread(unsigned tid)
651{
652 //Remove From Active List, if Active
653 std::list<unsigned>::iterator thread_it =
654 std::find(activeThreads.begin(), activeThreads.end(), tid);
655
656 DPRINTF(O3CPU, "[tid:%i]: Calling deactivate thread.\n", tid);
657
658 if (thread_it != activeThreads.end()) {
659 DPRINTF(O3CPU,"[tid:%i]: Removing from active threads list\n",
660 tid);
661 activeThreads.erase(thread_it);
662 }
663}
664
665template <class Impl>
666void
667FullO3CPU<Impl>::activateContext(int tid, int delay)
668{
669 // Needs to set each stage to running as well.
670 if (delay){
671 DPRINTF(O3CPU, "[tid:%i]: Scheduling thread context to activate "
672 "on cycle %d\n", tid, curTick + ticks(delay));
673 scheduleActivateThreadEvent(tid, delay);
674 } else {
675 activateThread(tid);
676 }
677
678 if (lastActivatedCycle < curTick) {
679 scheduleTickEvent(delay);
680
681 // Be sure to signal that there's some activity so the CPU doesn't
682 // deschedule itself.
683 activityRec.activity();
684 fetch.wakeFromQuiesce();
685
686 lastActivatedCycle = curTick;
687
688 _status = Running;
689 }
690}
691
692template <class Impl>
693bool
694FullO3CPU<Impl>::deallocateContext(int tid, bool remove, int delay)
695{
696 // Schedule removal of thread data from CPU
697 if (delay){
698 DPRINTF(O3CPU, "[tid:%i]: Scheduling thread context to deallocate "
699 "on cycle %d\n", tid, curTick + ticks(delay));
700 scheduleDeallocateContextEvent(tid, remove, delay);
701 return false;
702 } else {
703 deactivateThread(tid);
704 if (remove)
705 removeThread(tid);
706 return true;
707 }
708}
709
710template <class Impl>
711void
712FullO3CPU<Impl>::suspendContext(int tid)
713{
714 DPRINTF(O3CPU,"[tid: %i]: Suspending Thread Context.\n", tid);
715 bool deallocated = deallocateContext(tid, false, 1);
716 // If this was the last thread then unschedule the tick event.
717 if ((activeThreads.size() == 1 && !deallocated) ||
718 activeThreads.size() == 0)
719 unscheduleTickEvent();
720 _status = Idle;
721}
722
723template <class Impl>
724void
725FullO3CPU<Impl>::haltContext(int tid)
726{
727 //For now, this is the same as deallocate
728 DPRINTF(O3CPU,"[tid:%i]: Halt Context called. Deallocating", tid);
729 deallocateContext(tid, true, 1);
730}
731
732template <class Impl>
733void
734FullO3CPU<Impl>::insertThread(unsigned tid)
735{
736 DPRINTF(O3CPU,"[tid:%i] Initializing thread into CPU");
737 // Will change now that the PC and thread state is internal to the CPU
738 // and not in the ThreadContext.
739#if FULL_SYSTEM
740 ThreadContext *src_tc = system->threadContexts[tid];
741#else
742 ThreadContext *src_tc = tcBase(tid);
743#endif
744
745 //Bind Int Regs to Rename Map
746 for (int ireg = 0; ireg < TheISA::NumIntRegs; ireg++) {
747 PhysRegIndex phys_reg = freeList.getIntReg();
748
749 renameMap[tid].setEntry(ireg,phys_reg);
750 scoreboard.setReg(phys_reg);
751 }
752
753 //Bind Float Regs to Rename Map
754 for (int freg = 0; freg < TheISA::NumFloatRegs; freg++) {
755 PhysRegIndex phys_reg = freeList.getFloatReg();
756
757 renameMap[tid].setEntry(freg,phys_reg);
758 scoreboard.setReg(phys_reg);
759 }
760
761 //Copy Thread Data Into RegFile
762 //this->copyFromTC(tid);
763
764 //Set PC/NPC/NNPC
765 setPC(src_tc->readPC(), tid);
766 setNextPC(src_tc->readNextPC(), tid);
767 setNextNPC(src_tc->readNextNPC(), tid);
768
769 src_tc->setStatus(ThreadContext::Active);
770
771 activateContext(tid,1);
772
773 //Reset ROB/IQ/LSQ Entries
774 commit.rob->resetEntries();
775 iew.resetEntries();
776}
777
778template <class Impl>
779void
780FullO3CPU<Impl>::removeThread(unsigned tid)
781{
782 DPRINTF(O3CPU,"[tid:%i] Removing thread context from CPU.\n", tid);
783
784 // Copy Thread Data From RegFile
785 // If thread is suspended, it might be re-allocated
786 // this->copyToTC(tid);
787
788
789 // @todo: 2-27-2008: Fix how we free up rename mappings
790 // here to alleviate the case for double-freeing registers
791 // in SMT workloads.
792
793 // Unbind Int Regs from Rename Map
794 for (int ireg = 0; ireg < TheISA::NumIntRegs; ireg++) {
795 PhysRegIndex phys_reg = renameMap[tid].lookup(ireg);
796
797 scoreboard.unsetReg(phys_reg);
798 freeList.addReg(phys_reg);
799 }
800
801 // Unbind Float Regs from Rename Map
802 for (int freg = TheISA::NumIntRegs; freg < TheISA::NumFloatRegs; freg++) {
803 PhysRegIndex phys_reg = renameMap[tid].lookup(freg);
804
805 scoreboard.unsetReg(phys_reg);
806 freeList.addReg(phys_reg);
807 }
808
809 // Squash Throughout Pipeline
810 InstSeqNum squash_seq_num = commit.rob->readHeadInst(tid)->seqNum;
811 fetch.squash(0, sizeof(TheISA::MachInst), 0, squash_seq_num, tid);
812 decode.squash(tid);
813 rename.squash(squash_seq_num, tid);
814 iew.squash(tid);
815 iew.ldstQueue.squash(squash_seq_num, tid);
816 commit.rob->squash(squash_seq_num, tid);
817
818
819 assert(iew.instQueue.getCount(tid) == 0);
820 assert(iew.ldstQueue.getCount(tid) == 0);
821
822 // Reset ROB/IQ/LSQ Entries
823
824 // Commented out for now. This should be possible to do by
825 // telling all the pipeline stages to drain first, and then
826 // checking until the drain completes. Once the pipeline is
827 // drained, call resetEntries(). - 10-09-06 ktlim
828/*
829 if (activeThreads.size() >= 1) {
830 commit.rob->resetEntries();
831 iew.resetEntries();
832 }
833*/
834}
835
836
837template <class Impl>
838void
839FullO3CPU<Impl>::activateWhenReady(int tid)
840{
841 DPRINTF(O3CPU,"[tid:%i]: Checking if resources are available for incoming"
842 "(e.g. PhysRegs/ROB/IQ/LSQ) \n",
843 tid);
844
845 bool ready = true;
846
847 if (freeList.numFreeIntRegs() >= TheISA::NumIntRegs) {
848 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
849 "Phys. Int. Regs.\n",
850 tid);
851 ready = false;
852 } else if (freeList.numFreeFloatRegs() >= TheISA::NumFloatRegs) {
853 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
854 "Phys. Float. Regs.\n",
855 tid);
856 ready = false;
857 } else if (commit.rob->numFreeEntries() >=
858 commit.rob->entryAmount(activeThreads.size() + 1)) {
859 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
860 "ROB entries.\n",
861 tid);
862 ready = false;
863 } else if (iew.instQueue.numFreeEntries() >=
864 iew.instQueue.entryAmount(activeThreads.size() + 1)) {
865 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
866 "IQ entries.\n",
867 tid);
868 ready = false;
869 } else if (iew.ldstQueue.numFreeEntries() >=
870 iew.ldstQueue.entryAmount(activeThreads.size() + 1)) {
871 DPRINTF(O3CPU,"[tid:%i] Suspending thread due to not enough "
872 "LSQ entries.\n",
873 tid);
874 ready = false;
875 }
876
877 if (ready) {
878 insertThread(tid);
879
880 contextSwitch = false;
881
882 cpuWaitList.remove(tid);
883 } else {
884 suspendContext(tid);
885
886 //blocks fetch
887 contextSwitch = true;
888
889 //@todo: dont always add to waitlist
890 //do waitlist
891 cpuWaitList.push_back(tid);
892 }
893}
894
895#if FULL_SYSTEM
896template <class Impl>
897void
898FullO3CPU<Impl>::post_interrupt(int int_num, int index)
898FullO3CPU<Impl>::postInterrupt(int int_num, int index)
899{
899{
900 BaseCPU::post_interrupt(int_num, index);
900 BaseCPU::postInterrupt(int_num, index);
901
902 if (this->thread[0]->status() == ThreadContext::Suspended) {
903 DPRINTF(IPI,"Suspended Processor awoke\n");
904 this->threadContexts[0]->activate();
905 }
906}
907
908template <class Impl>
909Fault
910FullO3CPU<Impl>::hwrei(unsigned tid)
911{
912#if THE_ISA == ALPHA_ISA
913 // Need to clear the lock flag upon returning from an interrupt.
914 this->setMiscRegNoEffect(AlphaISA::MISCREG_LOCKFLAG, false, tid);
915
916 this->thread[tid]->kernelStats->hwrei();
917
918 // FIXME: XXX check for interrupts? XXX
919#endif
920 return NoFault;
921}
922
923template <class Impl>
924bool
925FullO3CPU<Impl>::simPalCheck(int palFunc, unsigned tid)
926{
927#if THE_ISA == ALPHA_ISA
928 if (this->thread[tid]->kernelStats)
929 this->thread[tid]->kernelStats->callpal(palFunc,
930 this->threadContexts[tid]);
931
932 switch (palFunc) {
933 case PAL::halt:
934 halt();
935 if (--System::numSystemsRunning == 0)
936 exitSimLoop("all cpus halted");
937 break;
938
939 case PAL::bpt:
940 case PAL::bugchk:
941 if (this->system->breakpoint())
942 return false;
943 break;
944 }
945#endif
946 return true;
947}
948
949template <class Impl>
950Fault
951FullO3CPU<Impl>::getInterrupts()
952{
953 // Check if there are any outstanding interrupts
954 return this->interrupts->getInterrupt(this->threadContexts[0]);
955}
956
957template <class Impl>
958void
959FullO3CPU<Impl>::processInterrupts(Fault interrupt)
960{
961 // Check for interrupts here. For now can copy the code that
962 // exists within isa_fullsys_traits.hh. Also assume that thread 0
963 // is the one that handles the interrupts.
964 // @todo: Possibly consolidate the interrupt checking code.
965 // @todo: Allow other threads to handle interrupts.
966
967 assert(interrupt != NoFault);
968 this->interrupts->updateIntrInfo(this->threadContexts[0]);
969
970 DPRINTF(O3CPU, "Interrupt %s being handled\n", interrupt->name());
971 this->trap(interrupt, 0);
972}
973
974template <class Impl>
975void
976FullO3CPU<Impl>::updateMemPorts()
977{
978 // Update all ThreadContext's memory ports (Functional/Virtual
979 // Ports)
980 for (int i = 0; i < thread.size(); ++i)
981 thread[i]->connectMemPorts(thread[i]->getTC());
982}
983#endif
984
985template <class Impl>
986void
987FullO3CPU<Impl>::trap(Fault fault, unsigned tid)
988{
989 // Pass the thread's TC into the invoke method.
990 fault->invoke(this->threadContexts[tid]);
991}
992
993#if !FULL_SYSTEM
994
995template <class Impl>
996void
997FullO3CPU<Impl>::syscall(int64_t callnum, int tid)
998{
999 DPRINTF(O3CPU, "[tid:%i] Executing syscall().\n\n", tid);
1000
1001 DPRINTF(Activity,"Activity: syscall() called.\n");
1002
1003 // Temporarily increase this by one to account for the syscall
1004 // instruction.
1005 ++(this->thread[tid]->funcExeInst);
1006
1007 // Execute the actual syscall.
1008 this->thread[tid]->syscall(callnum);
1009
1010 // Decrease funcExeInst by one as the normal commit will handle
1011 // incrementing it.
1012 --(this->thread[tid]->funcExeInst);
1013}
1014
1015template <class Impl>
1016void
1017FullO3CPU<Impl>::setSyscallReturn(SyscallReturn return_value, int tid)
1018{
1019 TheISA::setSyscallReturn(return_value, this->tcBase(tid));
1020}
1021
1022#endif
1023
1024template <class Impl>
1025void
1026FullO3CPU<Impl>::serialize(std::ostream &os)
1027{
1028 SimObject::State so_state = SimObject::getState();
1029 SERIALIZE_ENUM(so_state);
1030 BaseCPU::serialize(os);
1031 nameOut(os, csprintf("%s.tickEvent", name()));
1032 tickEvent.serialize(os);
1033
1034 // Use SimpleThread's ability to checkpoint to make it easier to
1035 // write out the registers. Also make this static so it doesn't
1036 // get instantiated multiple times (causes a panic in statistics).
1037 static SimpleThread temp;
1038
1039 for (int i = 0; i < thread.size(); i++) {
1040 nameOut(os, csprintf("%s.xc.%i", name(), i));
1041 temp.copyTC(thread[i]->getTC());
1042 temp.serialize(os);
1043 }
1044}
1045
1046template <class Impl>
1047void
1048FullO3CPU<Impl>::unserialize(Checkpoint *cp, const std::string &section)
1049{
1050 SimObject::State so_state;
1051 UNSERIALIZE_ENUM(so_state);
1052 BaseCPU::unserialize(cp, section);
1053 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
1054
1055 // Use SimpleThread's ability to checkpoint to make it easier to
1056 // read in the registers. Also make this static so it doesn't
1057 // get instantiated multiple times (causes a panic in statistics).
1058 static SimpleThread temp;
1059
1060 for (int i = 0; i < thread.size(); i++) {
1061 temp.copyTC(thread[i]->getTC());
1062 temp.unserialize(cp, csprintf("%s.xc.%i", section, i));
1063 thread[i]->getTC()->copyArchRegs(temp.getTC());
1064 }
1065}
1066
1067template <class Impl>
1068unsigned int
1069FullO3CPU<Impl>::drain(Event *drain_event)
1070{
1071 DPRINTF(O3CPU, "Switching out\n");
1072
1073 // If the CPU isn't doing anything, then return immediately.
1074 if (_status == Idle || _status == SwitchedOut) {
1075 return 0;
1076 }
1077
1078 drainCount = 0;
1079 fetch.drain();
1080 decode.drain();
1081 rename.drain();
1082 iew.drain();
1083 commit.drain();
1084
1085 // Wake the CPU and record activity so everything can drain out if
1086 // the CPU was not able to immediately drain.
1087 if (getState() != SimObject::Drained) {
1088 // A bit of a hack...set the drainEvent after all the drain()
1089 // calls have been made, that way if all of the stages drain
1090 // immediately, the signalDrained() function knows not to call
1091 // process on the drain event.
1092 drainEvent = drain_event;
1093
1094 wakeCPU();
1095 activityRec.activity();
1096
1097 return 1;
1098 } else {
1099 return 0;
1100 }
1101}
1102
1103template <class Impl>
1104void
1105FullO3CPU<Impl>::resume()
1106{
1107 fetch.resume();
1108 decode.resume();
1109 rename.resume();
1110 iew.resume();
1111 commit.resume();
1112
1113 changeState(SimObject::Running);
1114
1115 if (_status == SwitchedOut || _status == Idle)
1116 return;
1117
1118#if FULL_SYSTEM
1119 assert(system->getMemoryMode() == Enums::timing);
1120#endif
1121
1122 if (!tickEvent.scheduled())
1123 schedule(tickEvent, nextCycle());
1124 _status = Running;
1125}
1126
1127template <class Impl>
1128void
1129FullO3CPU<Impl>::signalDrained()
1130{
1131 if (++drainCount == NumStages) {
1132 if (tickEvent.scheduled())
1133 tickEvent.squash();
1134
1135 changeState(SimObject::Drained);
1136
1137 BaseCPU::switchOut();
1138
1139 if (drainEvent) {
1140 drainEvent->process();
1141 drainEvent = NULL;
1142 }
1143 }
1144 assert(drainCount <= 5);
1145}
1146
1147template <class Impl>
1148void
1149FullO3CPU<Impl>::switchOut()
1150{
1151 fetch.switchOut();
1152 rename.switchOut();
1153 iew.switchOut();
1154 commit.switchOut();
1155 instList.clear();
1156 while (!removeList.empty()) {
1157 removeList.pop();
1158 }
1159
1160 _status = SwitchedOut;
1161#if USE_CHECKER
1162 if (checker)
1163 checker->switchOut();
1164#endif
1165 if (tickEvent.scheduled())
1166 tickEvent.squash();
1167}
1168
1169template <class Impl>
1170void
1171FullO3CPU<Impl>::takeOverFrom(BaseCPU *oldCPU)
1172{
1173 // Flush out any old data from the time buffers.
1174 for (int i = 0; i < timeBuffer.getSize(); ++i) {
1175 timeBuffer.advance();
1176 fetchQueue.advance();
1177 decodeQueue.advance();
1178 renameQueue.advance();
1179 iewQueue.advance();
1180 }
1181
1182 activityRec.reset();
1183
1184 BaseCPU::takeOverFrom(oldCPU, fetch.getIcachePort(), iew.getDcachePort());
1185
1186 fetch.takeOverFrom();
1187 decode.takeOverFrom();
1188 rename.takeOverFrom();
1189 iew.takeOverFrom();
1190 commit.takeOverFrom();
1191
1192 assert(!tickEvent.scheduled());
1193
1194 // @todo: Figure out how to properly select the tid to put onto
1195 // the active threads list.
1196 int tid = 0;
1197
1198 std::list<unsigned>::iterator isActive =
1199 std::find(activeThreads.begin(), activeThreads.end(), tid);
1200
1201 if (isActive == activeThreads.end()) {
1202 //May Need to Re-code this if the delay variable is the delay
1203 //needed for thread to activate
1204 DPRINTF(O3CPU, "Adding Thread %i to active threads list\n",
1205 tid);
1206
1207 activeThreads.push_back(tid);
1208 }
1209
1210 // Set all statuses to active, schedule the CPU's tick event.
1211 // @todo: Fix up statuses so this is handled properly
1212 for (int i = 0; i < threadContexts.size(); ++i) {
1213 ThreadContext *tc = threadContexts[i];
1214 if (tc->status() == ThreadContext::Active && _status != Running) {
1215 _status = Running;
1216 schedule(tickEvent, nextCycle());
1217 }
1218 }
1219 if (!tickEvent.scheduled())
1220 schedule(tickEvent, nextCycle());
1221}
1222
1223template <class Impl>
1224TheISA::MiscReg
1225FullO3CPU<Impl>::readMiscRegNoEffect(int misc_reg, unsigned tid)
1226{
1227 return this->regFile.readMiscRegNoEffect(misc_reg, tid);
1228}
1229
1230template <class Impl>
1231TheISA::MiscReg
1232FullO3CPU<Impl>::readMiscReg(int misc_reg, unsigned tid)
1233{
1234 return this->regFile.readMiscReg(misc_reg, tid);
1235}
1236
1237template <class Impl>
1238void
1239FullO3CPU<Impl>::setMiscRegNoEffect(int misc_reg,
1240 const TheISA::MiscReg &val, unsigned tid)
1241{
1242 this->regFile.setMiscRegNoEffect(misc_reg, val, tid);
1243}
1244
1245template <class Impl>
1246void
1247FullO3CPU<Impl>::setMiscReg(int misc_reg,
1248 const TheISA::MiscReg &val, unsigned tid)
1249{
1250 this->regFile.setMiscReg(misc_reg, val, tid);
1251}
1252
1253template <class Impl>
1254uint64_t
1255FullO3CPU<Impl>::readIntReg(int reg_idx)
1256{
1257 return regFile.readIntReg(reg_idx);
1258}
1259
1260template <class Impl>
1261FloatReg
1262FullO3CPU<Impl>::readFloatReg(int reg_idx, int width)
1263{
1264 return regFile.readFloatReg(reg_idx, width);
1265}
1266
1267template <class Impl>
1268FloatReg
1269FullO3CPU<Impl>::readFloatReg(int reg_idx)
1270{
1271 return regFile.readFloatReg(reg_idx);
1272}
1273
1274template <class Impl>
1275FloatRegBits
1276FullO3CPU<Impl>::readFloatRegBits(int reg_idx, int width)
1277{
1278 return regFile.readFloatRegBits(reg_idx, width);
1279}
1280
1281template <class Impl>
1282FloatRegBits
1283FullO3CPU<Impl>::readFloatRegBits(int reg_idx)
1284{
1285 return regFile.readFloatRegBits(reg_idx);
1286}
1287
1288template <class Impl>
1289void
1290FullO3CPU<Impl>::setIntReg(int reg_idx, uint64_t val)
1291{
1292 regFile.setIntReg(reg_idx, val);
1293}
1294
1295template <class Impl>
1296void
1297FullO3CPU<Impl>::setFloatReg(int reg_idx, FloatReg val, int width)
1298{
1299 regFile.setFloatReg(reg_idx, val, width);
1300}
1301
1302template <class Impl>
1303void
1304FullO3CPU<Impl>::setFloatReg(int reg_idx, FloatReg val)
1305{
1306 regFile.setFloatReg(reg_idx, val);
1307}
1308
1309template <class Impl>
1310void
1311FullO3CPU<Impl>::setFloatRegBits(int reg_idx, FloatRegBits val, int width)
1312{
1313 regFile.setFloatRegBits(reg_idx, val, width);
1314}
1315
1316template <class Impl>
1317void
1318FullO3CPU<Impl>::setFloatRegBits(int reg_idx, FloatRegBits val)
1319{
1320 regFile.setFloatRegBits(reg_idx, val);
1321}
1322
1323template <class Impl>
1324uint64_t
1325FullO3CPU<Impl>::readArchIntReg(int reg_idx, unsigned tid)
1326{
1327 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx);
1328
1329 return regFile.readIntReg(phys_reg);
1330}
1331
1332template <class Impl>
1333float
1334FullO3CPU<Impl>::readArchFloatRegSingle(int reg_idx, unsigned tid)
1335{
1336 int idx = reg_idx + TheISA::FP_Base_DepTag;
1337 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1338
1339 return regFile.readFloatReg(phys_reg);
1340}
1341
1342template <class Impl>
1343double
1344FullO3CPU<Impl>::readArchFloatRegDouble(int reg_idx, unsigned tid)
1345{
1346 int idx = reg_idx + TheISA::FP_Base_DepTag;
1347 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1348
1349 return regFile.readFloatReg(phys_reg, 64);
1350}
1351
1352template <class Impl>
1353uint64_t
1354FullO3CPU<Impl>::readArchFloatRegInt(int reg_idx, unsigned tid)
1355{
1356 int idx = reg_idx + TheISA::FP_Base_DepTag;
1357 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1358
1359 return regFile.readFloatRegBits(phys_reg);
1360}
1361
1362template <class Impl>
1363void
1364FullO3CPU<Impl>::setArchIntReg(int reg_idx, uint64_t val, unsigned tid)
1365{
1366 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx);
1367
1368 regFile.setIntReg(phys_reg, val);
1369}
1370
1371template <class Impl>
1372void
1373FullO3CPU<Impl>::setArchFloatRegSingle(int reg_idx, float val, unsigned tid)
1374{
1375 int idx = reg_idx + TheISA::FP_Base_DepTag;
1376 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1377
1378 regFile.setFloatReg(phys_reg, val);
1379}
1380
1381template <class Impl>
1382void
1383FullO3CPU<Impl>::setArchFloatRegDouble(int reg_idx, double val, unsigned tid)
1384{
1385 int idx = reg_idx + TheISA::FP_Base_DepTag;
1386 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1387
1388 regFile.setFloatReg(phys_reg, val, 64);
1389}
1390
1391template <class Impl>
1392void
1393FullO3CPU<Impl>::setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid)
1394{
1395 int idx = reg_idx + TheISA::FP_Base_DepTag;
1396 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1397
1398 regFile.setFloatRegBits(phys_reg, val);
1399}
1400
1401template <class Impl>
1402uint64_t
1403FullO3CPU<Impl>::readPC(unsigned tid)
1404{
1405 return commit.readPC(tid);
1406}
1407
1408template <class Impl>
1409void
1410FullO3CPU<Impl>::setPC(Addr new_PC,unsigned tid)
1411{
1412 commit.setPC(new_PC, tid);
1413}
1414
1415template <class Impl>
1416uint64_t
1417FullO3CPU<Impl>::readMicroPC(unsigned tid)
1418{
1419 return commit.readMicroPC(tid);
1420}
1421
1422template <class Impl>
1423void
1424FullO3CPU<Impl>::setMicroPC(Addr new_PC,unsigned tid)
1425{
1426 commit.setMicroPC(new_PC, tid);
1427}
1428
1429template <class Impl>
1430uint64_t
1431FullO3CPU<Impl>::readNextPC(unsigned tid)
1432{
1433 return commit.readNextPC(tid);
1434}
1435
1436template <class Impl>
1437void
1438FullO3CPU<Impl>::setNextPC(uint64_t val,unsigned tid)
1439{
1440 commit.setNextPC(val, tid);
1441}
1442
1443template <class Impl>
1444uint64_t
1445FullO3CPU<Impl>::readNextNPC(unsigned tid)
1446{
1447 return commit.readNextNPC(tid);
1448}
1449
1450template <class Impl>
1451void
1452FullO3CPU<Impl>::setNextNPC(uint64_t val,unsigned tid)
1453{
1454 commit.setNextNPC(val, tid);
1455}
1456
1457template <class Impl>
1458uint64_t
1459FullO3CPU<Impl>::readNextMicroPC(unsigned tid)
1460{
1461 return commit.readNextMicroPC(tid);
1462}
1463
1464template <class Impl>
1465void
1466FullO3CPU<Impl>::setNextMicroPC(Addr new_PC,unsigned tid)
1467{
1468 commit.setNextMicroPC(new_PC, tid);
1469}
1470
1471template <class Impl>
1472void
1473FullO3CPU<Impl>::squashFromTC(unsigned tid)
1474{
1475 this->thread[tid]->inSyscall = true;
1476 this->commit.generateTCEvent(tid);
1477}
1478
1479template <class Impl>
1480typename FullO3CPU<Impl>::ListIt
1481FullO3CPU<Impl>::addInst(DynInstPtr &inst)
1482{
1483 instList.push_back(inst);
1484
1485 return --(instList.end());
1486}
1487
1488template <class Impl>
1489void
1490FullO3CPU<Impl>::instDone(unsigned tid)
1491{
1492 // Keep an instruction count.
1493 thread[tid]->numInst++;
1494 thread[tid]->numInsts++;
1495 committedInsts[tid]++;
1496 totalCommittedInsts++;
1497
1498 // Check for instruction-count-based events.
1499 comInstEventQueue[tid]->serviceEvents(thread[tid]->numInst);
1500}
1501
1502template <class Impl>
1503void
1504FullO3CPU<Impl>::addToRemoveList(DynInstPtr &inst)
1505{
1506 removeInstsThisCycle = true;
1507
1508 removeList.push(inst->getInstListIt());
1509}
1510
1511template <class Impl>
1512void
1513FullO3CPU<Impl>::removeFrontInst(DynInstPtr &inst)
1514{
1515 DPRINTF(O3CPU, "Removing committed instruction [tid:%i] PC %#x "
1516 "[sn:%lli]\n",
1517 inst->threadNumber, inst->readPC(), inst->seqNum);
1518
1519 removeInstsThisCycle = true;
1520
1521 // Remove the front instruction.
1522 removeList.push(inst->getInstListIt());
1523}
1524
1525template <class Impl>
1526void
1527FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid)
1528{
1529 DPRINTF(O3CPU, "Thread %i: Deleting instructions from instruction"
1530 " list.\n", tid);
1531
1532 ListIt end_it;
1533
1534 bool rob_empty = false;
1535
1536 if (instList.empty()) {
1537 return;
1538 } else if (rob.isEmpty(/*tid*/)) {
1539 DPRINTF(O3CPU, "ROB is empty, squashing all insts.\n");
1540 end_it = instList.begin();
1541 rob_empty = true;
1542 } else {
1543 end_it = (rob.readTailInst(tid))->getInstListIt();
1544 DPRINTF(O3CPU, "ROB is not empty, squashing insts not in ROB.\n");
1545 }
1546
1547 removeInstsThisCycle = true;
1548
1549 ListIt inst_it = instList.end();
1550
1551 inst_it--;
1552
1553 // Walk through the instruction list, removing any instructions
1554 // that were inserted after the given instruction iterator, end_it.
1555 while (inst_it != end_it) {
1556 assert(!instList.empty());
1557
1558 squashInstIt(inst_it, tid);
1559
1560 inst_it--;
1561 }
1562
1563 // If the ROB was empty, then we actually need to remove the first
1564 // instruction as well.
1565 if (rob_empty) {
1566 squashInstIt(inst_it, tid);
1567 }
1568}
1569
1570template <class Impl>
1571void
1572FullO3CPU<Impl>::removeInstsUntil(const InstSeqNum &seq_num,
1573 unsigned tid)
1574{
1575 assert(!instList.empty());
1576
1577 removeInstsThisCycle = true;
1578
1579 ListIt inst_iter = instList.end();
1580
1581 inst_iter--;
1582
1583 DPRINTF(O3CPU, "Deleting instructions from instruction "
1584 "list that are from [tid:%i] and above [sn:%lli] (end=%lli).\n",
1585 tid, seq_num, (*inst_iter)->seqNum);
1586
1587 while ((*inst_iter)->seqNum > seq_num) {
1588
1589 bool break_loop = (inst_iter == instList.begin());
1590
1591 squashInstIt(inst_iter, tid);
1592
1593 inst_iter--;
1594
1595 if (break_loop)
1596 break;
1597 }
1598}
1599
1600template <class Impl>
1601inline void
1602FullO3CPU<Impl>::squashInstIt(const ListIt &instIt, const unsigned &tid)
1603{
1604 if ((*instIt)->threadNumber == tid) {
1605 DPRINTF(O3CPU, "Squashing instruction, "
1606 "[tid:%i] [sn:%lli] PC %#x\n",
1607 (*instIt)->threadNumber,
1608 (*instIt)->seqNum,
1609 (*instIt)->readPC());
1610
1611 // Mark it as squashed.
1612 (*instIt)->setSquashed();
1613
1614 // @todo: Formulate a consistent method for deleting
1615 // instructions from the instruction list
1616 // Remove the instruction from the list.
1617 removeList.push(instIt);
1618 }
1619}
1620
1621template <class Impl>
1622void
1623FullO3CPU<Impl>::cleanUpRemovedInsts()
1624{
1625 while (!removeList.empty()) {
1626 DPRINTF(O3CPU, "Removing instruction, "
1627 "[tid:%i] [sn:%lli] PC %#x\n",
1628 (*removeList.front())->threadNumber,
1629 (*removeList.front())->seqNum,
1630 (*removeList.front())->readPC());
1631
1632 instList.erase(removeList.front());
1633
1634 removeList.pop();
1635 }
1636
1637 removeInstsThisCycle = false;
1638}
1639/*
1640template <class Impl>
1641void
1642FullO3CPU<Impl>::removeAllInsts()
1643{
1644 instList.clear();
1645}
1646*/
1647template <class Impl>
1648void
1649FullO3CPU<Impl>::dumpInsts()
1650{
1651 int num = 0;
1652
1653 ListIt inst_list_it = instList.begin();
1654
1655 cprintf("Dumping Instruction List\n");
1656
1657 while (inst_list_it != instList.end()) {
1658 cprintf("Instruction:%i\nPC:%#x\n[tid:%i]\n[sn:%lli]\nIssued:%i\n"
1659 "Squashed:%i\n\n",
1660 num, (*inst_list_it)->readPC(), (*inst_list_it)->threadNumber,
1661 (*inst_list_it)->seqNum, (*inst_list_it)->isIssued(),
1662 (*inst_list_it)->isSquashed());
1663 inst_list_it++;
1664 ++num;
1665 }
1666}
1667/*
1668template <class Impl>
1669void
1670FullO3CPU<Impl>::wakeDependents(DynInstPtr &inst)
1671{
1672 iew.wakeDependents(inst);
1673}
1674*/
1675template <class Impl>
1676void
1677FullO3CPU<Impl>::wakeCPU()
1678{
1679 if (activityRec.active() || tickEvent.scheduled()) {
1680 DPRINTF(Activity, "CPU already running.\n");
1681 return;
1682 }
1683
1684 DPRINTF(Activity, "Waking up CPU\n");
1685
1686 idleCycles += tickToCycles((curTick - 1) - lastRunningCycle);
1687 numCycles += tickToCycles((curTick - 1) - lastRunningCycle);
1688
1689 schedule(tickEvent, nextCycle());
1690}
1691
1692template <class Impl>
1693int
1694FullO3CPU<Impl>::getFreeTid()
1695{
1696 for (int i=0; i < numThreads; i++) {
1697 if (!tids[i]) {
1698 tids[i] = true;
1699 return i;
1700 }
1701 }
1702
1703 return -1;
1704}
1705
1706template <class Impl>
1707void
1708FullO3CPU<Impl>::doContextSwitch()
1709{
1710 if (contextSwitch) {
1711
1712 //ADD CODE TO DEACTIVE THREAD HERE (???)
1713
1714 for (int tid=0; tid < cpuWaitList.size(); tid++) {
1715 activateWhenReady(tid);
1716 }
1717
1718 if (cpuWaitList.size() == 0)
1719 contextSwitch = true;
1720 }
1721}
1722
1723template <class Impl>
1724void
1725FullO3CPU<Impl>::updateThreadPriority()
1726{
1727 if (activeThreads.size() > 1)
1728 {
1729 //DEFAULT TO ROUND ROBIN SCHEME
1730 //e.g. Move highest priority to end of thread list
1731 std::list<unsigned>::iterator list_begin = activeThreads.begin();
1732 std::list<unsigned>::iterator list_end = activeThreads.end();
1733
1734 unsigned high_thread = *list_begin;
1735
1736 activeThreads.erase(list_begin);
1737
1738 activeThreads.push_back(high_thread);
1739 }
1740}
1741
1742// Forward declaration of FullO3CPU.
1743template class FullO3CPU<O3CPUImpl>;
901
902 if (this->thread[0]->status() == ThreadContext::Suspended) {
903 DPRINTF(IPI,"Suspended Processor awoke\n");
904 this->threadContexts[0]->activate();
905 }
906}
907
908template <class Impl>
909Fault
910FullO3CPU<Impl>::hwrei(unsigned tid)
911{
912#if THE_ISA == ALPHA_ISA
913 // Need to clear the lock flag upon returning from an interrupt.
914 this->setMiscRegNoEffect(AlphaISA::MISCREG_LOCKFLAG, false, tid);
915
916 this->thread[tid]->kernelStats->hwrei();
917
918 // FIXME: XXX check for interrupts? XXX
919#endif
920 return NoFault;
921}
922
923template <class Impl>
924bool
925FullO3CPU<Impl>::simPalCheck(int palFunc, unsigned tid)
926{
927#if THE_ISA == ALPHA_ISA
928 if (this->thread[tid]->kernelStats)
929 this->thread[tid]->kernelStats->callpal(palFunc,
930 this->threadContexts[tid]);
931
932 switch (palFunc) {
933 case PAL::halt:
934 halt();
935 if (--System::numSystemsRunning == 0)
936 exitSimLoop("all cpus halted");
937 break;
938
939 case PAL::bpt:
940 case PAL::bugchk:
941 if (this->system->breakpoint())
942 return false;
943 break;
944 }
945#endif
946 return true;
947}
948
949template <class Impl>
950Fault
951FullO3CPU<Impl>::getInterrupts()
952{
953 // Check if there are any outstanding interrupts
954 return this->interrupts->getInterrupt(this->threadContexts[0]);
955}
956
957template <class Impl>
958void
959FullO3CPU<Impl>::processInterrupts(Fault interrupt)
960{
961 // Check for interrupts here. For now can copy the code that
962 // exists within isa_fullsys_traits.hh. Also assume that thread 0
963 // is the one that handles the interrupts.
964 // @todo: Possibly consolidate the interrupt checking code.
965 // @todo: Allow other threads to handle interrupts.
966
967 assert(interrupt != NoFault);
968 this->interrupts->updateIntrInfo(this->threadContexts[0]);
969
970 DPRINTF(O3CPU, "Interrupt %s being handled\n", interrupt->name());
971 this->trap(interrupt, 0);
972}
973
974template <class Impl>
975void
976FullO3CPU<Impl>::updateMemPorts()
977{
978 // Update all ThreadContext's memory ports (Functional/Virtual
979 // Ports)
980 for (int i = 0; i < thread.size(); ++i)
981 thread[i]->connectMemPorts(thread[i]->getTC());
982}
983#endif
984
985template <class Impl>
986void
987FullO3CPU<Impl>::trap(Fault fault, unsigned tid)
988{
989 // Pass the thread's TC into the invoke method.
990 fault->invoke(this->threadContexts[tid]);
991}
992
993#if !FULL_SYSTEM
994
995template <class Impl>
996void
997FullO3CPU<Impl>::syscall(int64_t callnum, int tid)
998{
999 DPRINTF(O3CPU, "[tid:%i] Executing syscall().\n\n", tid);
1000
1001 DPRINTF(Activity,"Activity: syscall() called.\n");
1002
1003 // Temporarily increase this by one to account for the syscall
1004 // instruction.
1005 ++(this->thread[tid]->funcExeInst);
1006
1007 // Execute the actual syscall.
1008 this->thread[tid]->syscall(callnum);
1009
1010 // Decrease funcExeInst by one as the normal commit will handle
1011 // incrementing it.
1012 --(this->thread[tid]->funcExeInst);
1013}
1014
1015template <class Impl>
1016void
1017FullO3CPU<Impl>::setSyscallReturn(SyscallReturn return_value, int tid)
1018{
1019 TheISA::setSyscallReturn(return_value, this->tcBase(tid));
1020}
1021
1022#endif
1023
1024template <class Impl>
1025void
1026FullO3CPU<Impl>::serialize(std::ostream &os)
1027{
1028 SimObject::State so_state = SimObject::getState();
1029 SERIALIZE_ENUM(so_state);
1030 BaseCPU::serialize(os);
1031 nameOut(os, csprintf("%s.tickEvent", name()));
1032 tickEvent.serialize(os);
1033
1034 // Use SimpleThread's ability to checkpoint to make it easier to
1035 // write out the registers. Also make this static so it doesn't
1036 // get instantiated multiple times (causes a panic in statistics).
1037 static SimpleThread temp;
1038
1039 for (int i = 0; i < thread.size(); i++) {
1040 nameOut(os, csprintf("%s.xc.%i", name(), i));
1041 temp.copyTC(thread[i]->getTC());
1042 temp.serialize(os);
1043 }
1044}
1045
1046template <class Impl>
1047void
1048FullO3CPU<Impl>::unserialize(Checkpoint *cp, const std::string &section)
1049{
1050 SimObject::State so_state;
1051 UNSERIALIZE_ENUM(so_state);
1052 BaseCPU::unserialize(cp, section);
1053 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
1054
1055 // Use SimpleThread's ability to checkpoint to make it easier to
1056 // read in the registers. Also make this static so it doesn't
1057 // get instantiated multiple times (causes a panic in statistics).
1058 static SimpleThread temp;
1059
1060 for (int i = 0; i < thread.size(); i++) {
1061 temp.copyTC(thread[i]->getTC());
1062 temp.unserialize(cp, csprintf("%s.xc.%i", section, i));
1063 thread[i]->getTC()->copyArchRegs(temp.getTC());
1064 }
1065}
1066
1067template <class Impl>
1068unsigned int
1069FullO3CPU<Impl>::drain(Event *drain_event)
1070{
1071 DPRINTF(O3CPU, "Switching out\n");
1072
1073 // If the CPU isn't doing anything, then return immediately.
1074 if (_status == Idle || _status == SwitchedOut) {
1075 return 0;
1076 }
1077
1078 drainCount = 0;
1079 fetch.drain();
1080 decode.drain();
1081 rename.drain();
1082 iew.drain();
1083 commit.drain();
1084
1085 // Wake the CPU and record activity so everything can drain out if
1086 // the CPU was not able to immediately drain.
1087 if (getState() != SimObject::Drained) {
1088 // A bit of a hack...set the drainEvent after all the drain()
1089 // calls have been made, that way if all of the stages drain
1090 // immediately, the signalDrained() function knows not to call
1091 // process on the drain event.
1092 drainEvent = drain_event;
1093
1094 wakeCPU();
1095 activityRec.activity();
1096
1097 return 1;
1098 } else {
1099 return 0;
1100 }
1101}
1102
1103template <class Impl>
1104void
1105FullO3CPU<Impl>::resume()
1106{
1107 fetch.resume();
1108 decode.resume();
1109 rename.resume();
1110 iew.resume();
1111 commit.resume();
1112
1113 changeState(SimObject::Running);
1114
1115 if (_status == SwitchedOut || _status == Idle)
1116 return;
1117
1118#if FULL_SYSTEM
1119 assert(system->getMemoryMode() == Enums::timing);
1120#endif
1121
1122 if (!tickEvent.scheduled())
1123 schedule(tickEvent, nextCycle());
1124 _status = Running;
1125}
1126
1127template <class Impl>
1128void
1129FullO3CPU<Impl>::signalDrained()
1130{
1131 if (++drainCount == NumStages) {
1132 if (tickEvent.scheduled())
1133 tickEvent.squash();
1134
1135 changeState(SimObject::Drained);
1136
1137 BaseCPU::switchOut();
1138
1139 if (drainEvent) {
1140 drainEvent->process();
1141 drainEvent = NULL;
1142 }
1143 }
1144 assert(drainCount <= 5);
1145}
1146
1147template <class Impl>
1148void
1149FullO3CPU<Impl>::switchOut()
1150{
1151 fetch.switchOut();
1152 rename.switchOut();
1153 iew.switchOut();
1154 commit.switchOut();
1155 instList.clear();
1156 while (!removeList.empty()) {
1157 removeList.pop();
1158 }
1159
1160 _status = SwitchedOut;
1161#if USE_CHECKER
1162 if (checker)
1163 checker->switchOut();
1164#endif
1165 if (tickEvent.scheduled())
1166 tickEvent.squash();
1167}
1168
1169template <class Impl>
1170void
1171FullO3CPU<Impl>::takeOverFrom(BaseCPU *oldCPU)
1172{
1173 // Flush out any old data from the time buffers.
1174 for (int i = 0; i < timeBuffer.getSize(); ++i) {
1175 timeBuffer.advance();
1176 fetchQueue.advance();
1177 decodeQueue.advance();
1178 renameQueue.advance();
1179 iewQueue.advance();
1180 }
1181
1182 activityRec.reset();
1183
1184 BaseCPU::takeOverFrom(oldCPU, fetch.getIcachePort(), iew.getDcachePort());
1185
1186 fetch.takeOverFrom();
1187 decode.takeOverFrom();
1188 rename.takeOverFrom();
1189 iew.takeOverFrom();
1190 commit.takeOverFrom();
1191
1192 assert(!tickEvent.scheduled());
1193
1194 // @todo: Figure out how to properly select the tid to put onto
1195 // the active threads list.
1196 int tid = 0;
1197
1198 std::list<unsigned>::iterator isActive =
1199 std::find(activeThreads.begin(), activeThreads.end(), tid);
1200
1201 if (isActive == activeThreads.end()) {
1202 //May Need to Re-code this if the delay variable is the delay
1203 //needed for thread to activate
1204 DPRINTF(O3CPU, "Adding Thread %i to active threads list\n",
1205 tid);
1206
1207 activeThreads.push_back(tid);
1208 }
1209
1210 // Set all statuses to active, schedule the CPU's tick event.
1211 // @todo: Fix up statuses so this is handled properly
1212 for (int i = 0; i < threadContexts.size(); ++i) {
1213 ThreadContext *tc = threadContexts[i];
1214 if (tc->status() == ThreadContext::Active && _status != Running) {
1215 _status = Running;
1216 schedule(tickEvent, nextCycle());
1217 }
1218 }
1219 if (!tickEvent.scheduled())
1220 schedule(tickEvent, nextCycle());
1221}
1222
1223template <class Impl>
1224TheISA::MiscReg
1225FullO3CPU<Impl>::readMiscRegNoEffect(int misc_reg, unsigned tid)
1226{
1227 return this->regFile.readMiscRegNoEffect(misc_reg, tid);
1228}
1229
1230template <class Impl>
1231TheISA::MiscReg
1232FullO3CPU<Impl>::readMiscReg(int misc_reg, unsigned tid)
1233{
1234 return this->regFile.readMiscReg(misc_reg, tid);
1235}
1236
1237template <class Impl>
1238void
1239FullO3CPU<Impl>::setMiscRegNoEffect(int misc_reg,
1240 const TheISA::MiscReg &val, unsigned tid)
1241{
1242 this->regFile.setMiscRegNoEffect(misc_reg, val, tid);
1243}
1244
1245template <class Impl>
1246void
1247FullO3CPU<Impl>::setMiscReg(int misc_reg,
1248 const TheISA::MiscReg &val, unsigned tid)
1249{
1250 this->regFile.setMiscReg(misc_reg, val, tid);
1251}
1252
1253template <class Impl>
1254uint64_t
1255FullO3CPU<Impl>::readIntReg(int reg_idx)
1256{
1257 return regFile.readIntReg(reg_idx);
1258}
1259
1260template <class Impl>
1261FloatReg
1262FullO3CPU<Impl>::readFloatReg(int reg_idx, int width)
1263{
1264 return regFile.readFloatReg(reg_idx, width);
1265}
1266
1267template <class Impl>
1268FloatReg
1269FullO3CPU<Impl>::readFloatReg(int reg_idx)
1270{
1271 return regFile.readFloatReg(reg_idx);
1272}
1273
1274template <class Impl>
1275FloatRegBits
1276FullO3CPU<Impl>::readFloatRegBits(int reg_idx, int width)
1277{
1278 return regFile.readFloatRegBits(reg_idx, width);
1279}
1280
1281template <class Impl>
1282FloatRegBits
1283FullO3CPU<Impl>::readFloatRegBits(int reg_idx)
1284{
1285 return regFile.readFloatRegBits(reg_idx);
1286}
1287
1288template <class Impl>
1289void
1290FullO3CPU<Impl>::setIntReg(int reg_idx, uint64_t val)
1291{
1292 regFile.setIntReg(reg_idx, val);
1293}
1294
1295template <class Impl>
1296void
1297FullO3CPU<Impl>::setFloatReg(int reg_idx, FloatReg val, int width)
1298{
1299 regFile.setFloatReg(reg_idx, val, width);
1300}
1301
1302template <class Impl>
1303void
1304FullO3CPU<Impl>::setFloatReg(int reg_idx, FloatReg val)
1305{
1306 regFile.setFloatReg(reg_idx, val);
1307}
1308
1309template <class Impl>
1310void
1311FullO3CPU<Impl>::setFloatRegBits(int reg_idx, FloatRegBits val, int width)
1312{
1313 regFile.setFloatRegBits(reg_idx, val, width);
1314}
1315
1316template <class Impl>
1317void
1318FullO3CPU<Impl>::setFloatRegBits(int reg_idx, FloatRegBits val)
1319{
1320 regFile.setFloatRegBits(reg_idx, val);
1321}
1322
1323template <class Impl>
1324uint64_t
1325FullO3CPU<Impl>::readArchIntReg(int reg_idx, unsigned tid)
1326{
1327 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx);
1328
1329 return regFile.readIntReg(phys_reg);
1330}
1331
1332template <class Impl>
1333float
1334FullO3CPU<Impl>::readArchFloatRegSingle(int reg_idx, unsigned tid)
1335{
1336 int idx = reg_idx + TheISA::FP_Base_DepTag;
1337 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1338
1339 return regFile.readFloatReg(phys_reg);
1340}
1341
1342template <class Impl>
1343double
1344FullO3CPU<Impl>::readArchFloatRegDouble(int reg_idx, unsigned tid)
1345{
1346 int idx = reg_idx + TheISA::FP_Base_DepTag;
1347 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1348
1349 return regFile.readFloatReg(phys_reg, 64);
1350}
1351
1352template <class Impl>
1353uint64_t
1354FullO3CPU<Impl>::readArchFloatRegInt(int reg_idx, unsigned tid)
1355{
1356 int idx = reg_idx + TheISA::FP_Base_DepTag;
1357 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1358
1359 return regFile.readFloatRegBits(phys_reg);
1360}
1361
1362template <class Impl>
1363void
1364FullO3CPU<Impl>::setArchIntReg(int reg_idx, uint64_t val, unsigned tid)
1365{
1366 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx);
1367
1368 regFile.setIntReg(phys_reg, val);
1369}
1370
1371template <class Impl>
1372void
1373FullO3CPU<Impl>::setArchFloatRegSingle(int reg_idx, float val, unsigned tid)
1374{
1375 int idx = reg_idx + TheISA::FP_Base_DepTag;
1376 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1377
1378 regFile.setFloatReg(phys_reg, val);
1379}
1380
1381template <class Impl>
1382void
1383FullO3CPU<Impl>::setArchFloatRegDouble(int reg_idx, double val, unsigned tid)
1384{
1385 int idx = reg_idx + TheISA::FP_Base_DepTag;
1386 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1387
1388 regFile.setFloatReg(phys_reg, val, 64);
1389}
1390
1391template <class Impl>
1392void
1393FullO3CPU<Impl>::setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid)
1394{
1395 int idx = reg_idx + TheISA::FP_Base_DepTag;
1396 PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx);
1397
1398 regFile.setFloatRegBits(phys_reg, val);
1399}
1400
1401template <class Impl>
1402uint64_t
1403FullO3CPU<Impl>::readPC(unsigned tid)
1404{
1405 return commit.readPC(tid);
1406}
1407
1408template <class Impl>
1409void
1410FullO3CPU<Impl>::setPC(Addr new_PC,unsigned tid)
1411{
1412 commit.setPC(new_PC, tid);
1413}
1414
1415template <class Impl>
1416uint64_t
1417FullO3CPU<Impl>::readMicroPC(unsigned tid)
1418{
1419 return commit.readMicroPC(tid);
1420}
1421
1422template <class Impl>
1423void
1424FullO3CPU<Impl>::setMicroPC(Addr new_PC,unsigned tid)
1425{
1426 commit.setMicroPC(new_PC, tid);
1427}
1428
1429template <class Impl>
1430uint64_t
1431FullO3CPU<Impl>::readNextPC(unsigned tid)
1432{
1433 return commit.readNextPC(tid);
1434}
1435
1436template <class Impl>
1437void
1438FullO3CPU<Impl>::setNextPC(uint64_t val,unsigned tid)
1439{
1440 commit.setNextPC(val, tid);
1441}
1442
1443template <class Impl>
1444uint64_t
1445FullO3CPU<Impl>::readNextNPC(unsigned tid)
1446{
1447 return commit.readNextNPC(tid);
1448}
1449
1450template <class Impl>
1451void
1452FullO3CPU<Impl>::setNextNPC(uint64_t val,unsigned tid)
1453{
1454 commit.setNextNPC(val, tid);
1455}
1456
1457template <class Impl>
1458uint64_t
1459FullO3CPU<Impl>::readNextMicroPC(unsigned tid)
1460{
1461 return commit.readNextMicroPC(tid);
1462}
1463
1464template <class Impl>
1465void
1466FullO3CPU<Impl>::setNextMicroPC(Addr new_PC,unsigned tid)
1467{
1468 commit.setNextMicroPC(new_PC, tid);
1469}
1470
1471template <class Impl>
1472void
1473FullO3CPU<Impl>::squashFromTC(unsigned tid)
1474{
1475 this->thread[tid]->inSyscall = true;
1476 this->commit.generateTCEvent(tid);
1477}
1478
1479template <class Impl>
1480typename FullO3CPU<Impl>::ListIt
1481FullO3CPU<Impl>::addInst(DynInstPtr &inst)
1482{
1483 instList.push_back(inst);
1484
1485 return --(instList.end());
1486}
1487
1488template <class Impl>
1489void
1490FullO3CPU<Impl>::instDone(unsigned tid)
1491{
1492 // Keep an instruction count.
1493 thread[tid]->numInst++;
1494 thread[tid]->numInsts++;
1495 committedInsts[tid]++;
1496 totalCommittedInsts++;
1497
1498 // Check for instruction-count-based events.
1499 comInstEventQueue[tid]->serviceEvents(thread[tid]->numInst);
1500}
1501
1502template <class Impl>
1503void
1504FullO3CPU<Impl>::addToRemoveList(DynInstPtr &inst)
1505{
1506 removeInstsThisCycle = true;
1507
1508 removeList.push(inst->getInstListIt());
1509}
1510
1511template <class Impl>
1512void
1513FullO3CPU<Impl>::removeFrontInst(DynInstPtr &inst)
1514{
1515 DPRINTF(O3CPU, "Removing committed instruction [tid:%i] PC %#x "
1516 "[sn:%lli]\n",
1517 inst->threadNumber, inst->readPC(), inst->seqNum);
1518
1519 removeInstsThisCycle = true;
1520
1521 // Remove the front instruction.
1522 removeList.push(inst->getInstListIt());
1523}
1524
1525template <class Impl>
1526void
1527FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid)
1528{
1529 DPRINTF(O3CPU, "Thread %i: Deleting instructions from instruction"
1530 " list.\n", tid);
1531
1532 ListIt end_it;
1533
1534 bool rob_empty = false;
1535
1536 if (instList.empty()) {
1537 return;
1538 } else if (rob.isEmpty(/*tid*/)) {
1539 DPRINTF(O3CPU, "ROB is empty, squashing all insts.\n");
1540 end_it = instList.begin();
1541 rob_empty = true;
1542 } else {
1543 end_it = (rob.readTailInst(tid))->getInstListIt();
1544 DPRINTF(O3CPU, "ROB is not empty, squashing insts not in ROB.\n");
1545 }
1546
1547 removeInstsThisCycle = true;
1548
1549 ListIt inst_it = instList.end();
1550
1551 inst_it--;
1552
1553 // Walk through the instruction list, removing any instructions
1554 // that were inserted after the given instruction iterator, end_it.
1555 while (inst_it != end_it) {
1556 assert(!instList.empty());
1557
1558 squashInstIt(inst_it, tid);
1559
1560 inst_it--;
1561 }
1562
1563 // If the ROB was empty, then we actually need to remove the first
1564 // instruction as well.
1565 if (rob_empty) {
1566 squashInstIt(inst_it, tid);
1567 }
1568}
1569
1570template <class Impl>
1571void
1572FullO3CPU<Impl>::removeInstsUntil(const InstSeqNum &seq_num,
1573 unsigned tid)
1574{
1575 assert(!instList.empty());
1576
1577 removeInstsThisCycle = true;
1578
1579 ListIt inst_iter = instList.end();
1580
1581 inst_iter--;
1582
1583 DPRINTF(O3CPU, "Deleting instructions from instruction "
1584 "list that are from [tid:%i] and above [sn:%lli] (end=%lli).\n",
1585 tid, seq_num, (*inst_iter)->seqNum);
1586
1587 while ((*inst_iter)->seqNum > seq_num) {
1588
1589 bool break_loop = (inst_iter == instList.begin());
1590
1591 squashInstIt(inst_iter, tid);
1592
1593 inst_iter--;
1594
1595 if (break_loop)
1596 break;
1597 }
1598}
1599
1600template <class Impl>
1601inline void
1602FullO3CPU<Impl>::squashInstIt(const ListIt &instIt, const unsigned &tid)
1603{
1604 if ((*instIt)->threadNumber == tid) {
1605 DPRINTF(O3CPU, "Squashing instruction, "
1606 "[tid:%i] [sn:%lli] PC %#x\n",
1607 (*instIt)->threadNumber,
1608 (*instIt)->seqNum,
1609 (*instIt)->readPC());
1610
1611 // Mark it as squashed.
1612 (*instIt)->setSquashed();
1613
1614 // @todo: Formulate a consistent method for deleting
1615 // instructions from the instruction list
1616 // Remove the instruction from the list.
1617 removeList.push(instIt);
1618 }
1619}
1620
1621template <class Impl>
1622void
1623FullO3CPU<Impl>::cleanUpRemovedInsts()
1624{
1625 while (!removeList.empty()) {
1626 DPRINTF(O3CPU, "Removing instruction, "
1627 "[tid:%i] [sn:%lli] PC %#x\n",
1628 (*removeList.front())->threadNumber,
1629 (*removeList.front())->seqNum,
1630 (*removeList.front())->readPC());
1631
1632 instList.erase(removeList.front());
1633
1634 removeList.pop();
1635 }
1636
1637 removeInstsThisCycle = false;
1638}
1639/*
1640template <class Impl>
1641void
1642FullO3CPU<Impl>::removeAllInsts()
1643{
1644 instList.clear();
1645}
1646*/
1647template <class Impl>
1648void
1649FullO3CPU<Impl>::dumpInsts()
1650{
1651 int num = 0;
1652
1653 ListIt inst_list_it = instList.begin();
1654
1655 cprintf("Dumping Instruction List\n");
1656
1657 while (inst_list_it != instList.end()) {
1658 cprintf("Instruction:%i\nPC:%#x\n[tid:%i]\n[sn:%lli]\nIssued:%i\n"
1659 "Squashed:%i\n\n",
1660 num, (*inst_list_it)->readPC(), (*inst_list_it)->threadNumber,
1661 (*inst_list_it)->seqNum, (*inst_list_it)->isIssued(),
1662 (*inst_list_it)->isSquashed());
1663 inst_list_it++;
1664 ++num;
1665 }
1666}
1667/*
1668template <class Impl>
1669void
1670FullO3CPU<Impl>::wakeDependents(DynInstPtr &inst)
1671{
1672 iew.wakeDependents(inst);
1673}
1674*/
1675template <class Impl>
1676void
1677FullO3CPU<Impl>::wakeCPU()
1678{
1679 if (activityRec.active() || tickEvent.scheduled()) {
1680 DPRINTF(Activity, "CPU already running.\n");
1681 return;
1682 }
1683
1684 DPRINTF(Activity, "Waking up CPU\n");
1685
1686 idleCycles += tickToCycles((curTick - 1) - lastRunningCycle);
1687 numCycles += tickToCycles((curTick - 1) - lastRunningCycle);
1688
1689 schedule(tickEvent, nextCycle());
1690}
1691
1692template <class Impl>
1693int
1694FullO3CPU<Impl>::getFreeTid()
1695{
1696 for (int i=0; i < numThreads; i++) {
1697 if (!tids[i]) {
1698 tids[i] = true;
1699 return i;
1700 }
1701 }
1702
1703 return -1;
1704}
1705
1706template <class Impl>
1707void
1708FullO3CPU<Impl>::doContextSwitch()
1709{
1710 if (contextSwitch) {
1711
1712 //ADD CODE TO DEACTIVE THREAD HERE (???)
1713
1714 for (int tid=0; tid < cpuWaitList.size(); tid++) {
1715 activateWhenReady(tid);
1716 }
1717
1718 if (cpuWaitList.size() == 0)
1719 contextSwitch = true;
1720 }
1721}
1722
1723template <class Impl>
1724void
1725FullO3CPU<Impl>::updateThreadPriority()
1726{
1727 if (activeThreads.size() > 1)
1728 {
1729 //DEFAULT TO ROUND ROBIN SCHEME
1730 //e.g. Move highest priority to end of thread list
1731 std::list<unsigned>::iterator list_begin = activeThreads.begin();
1732 std::list<unsigned>::iterator list_end = activeThreads.end();
1733
1734 unsigned high_thread = *list_begin;
1735
1736 activeThreads.erase(list_begin);
1737
1738 activeThreads.push_back(high_thread);
1739 }
1740}
1741
1742// Forward declaration of FullO3CPU.
1743template class FullO3CPU<O3CPUImpl>;