base.cc (12284:b91c036913da) base.cc (12334:e0ab29a34764)
1/*
2 * Copyright (c) 2011-2012,2016-2017 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 * Copyright (c) 2011 Regents of the University of California
16 * Copyright (c) 2013 Advanced Micro Devices, Inc.
17 * Copyright (c) 2013 Mark D. Hill and David A. Wood
18 * All rights reserved.
19 *
20 * Redistribution and use in source and binary forms, with or without
21 * modification, are permitted provided that the following conditions are
22 * met: redistributions of source code must retain the above copyright
23 * notice, this list of conditions and the following disclaimer;
24 * redistributions in binary form must reproduce the above copyright
25 * notice, this list of conditions and the following disclaimer in the
26 * documentation and/or other materials provided with the distribution;
27 * neither the name of the copyright holders nor the names of its
28 * contributors may be used to endorse or promote products derived from
29 * this software without specific prior written permission.
30 *
31 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
32 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
33 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
34 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
35 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
36 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
37 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
38 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
39 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
40 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
41 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 *
43 * Authors: Steve Reinhardt
44 * Nathan Binkert
45 * Rick Strong
46 */
47
48#include "cpu/base.hh"
49
50#include <iostream>
51#include <sstream>
52#include <string>
53
54#include "arch/tlb.hh"
55#include "base/cprintf.hh"
56#include "base/loader/symtab.hh"
1/*
2 * Copyright (c) 2011-2012,2016-2017 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 * Copyright (c) 2011 Regents of the University of California
16 * Copyright (c) 2013 Advanced Micro Devices, Inc.
17 * Copyright (c) 2013 Mark D. Hill and David A. Wood
18 * All rights reserved.
19 *
20 * Redistribution and use in source and binary forms, with or without
21 * modification, are permitted provided that the following conditions are
22 * met: redistributions of source code must retain the above copyright
23 * notice, this list of conditions and the following disclaimer;
24 * redistributions in binary form must reproduce the above copyright
25 * notice, this list of conditions and the following disclaimer in the
26 * documentation and/or other materials provided with the distribution;
27 * neither the name of the copyright holders nor the names of its
28 * contributors may be used to endorse or promote products derived from
29 * this software without specific prior written permission.
30 *
31 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
32 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
33 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
34 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
35 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
36 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
37 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
38 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
39 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
40 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
41 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 *
43 * Authors: Steve Reinhardt
44 * Nathan Binkert
45 * Rick Strong
46 */
47
48#include "cpu/base.hh"
49
50#include <iostream>
51#include <sstream>
52#include <string>
53
54#include "arch/tlb.hh"
55#include "base/cprintf.hh"
56#include "base/loader/symtab.hh"
57#include "base/misc.hh"
57#include "base/logging.hh"
58#include "base/output.hh"
59#include "base/trace.hh"
60#include "cpu/checker/cpu.hh"
61#include "cpu/cpuevent.hh"
62#include "cpu/profile.hh"
63#include "cpu/thread_context.hh"
64#include "debug/Mwait.hh"
65#include "debug/SyscallVerbose.hh"
66#include "mem/page_table.hh"
67#include "params/BaseCPU.hh"
68#include "sim/clocked_object.hh"
69#include "sim/full_system.hh"
70#include "sim/process.hh"
71#include "sim/sim_events.hh"
72#include "sim/sim_exit.hh"
73#include "sim/system.hh"
74
75// Hack
76#include "sim/stat_control.hh"
77
78using namespace std;
79
80vector<BaseCPU *> BaseCPU::cpuList;
81
82// This variable reflects the max number of threads in any CPU. Be
83// careful to only use it once all the CPUs that you care about have
84// been initialized
85int maxThreadsPerCPU = 1;
86
87CPUProgressEvent::CPUProgressEvent(BaseCPU *_cpu, Tick ival)
88 : Event(Event::Progress_Event_Pri), _interval(ival), lastNumInst(0),
89 cpu(_cpu), _repeatEvent(true)
90{
91 if (_interval)
92 cpu->schedule(this, curTick() + _interval);
93}
94
95void
96CPUProgressEvent::process()
97{
98 Counter temp = cpu->totalOps();
99
100 if (_repeatEvent)
101 cpu->schedule(this, curTick() + _interval);
102
103 if (cpu->switchedOut()) {
104 return;
105 }
106
107#ifndef NDEBUG
108 double ipc = double(temp - lastNumInst) / (_interval / cpu->clockPeriod());
109
110 DPRINTFN("%s progress event, total committed:%i, progress insts committed: "
111 "%lli, IPC: %0.8d\n", cpu->name(), temp, temp - lastNumInst,
112 ipc);
113 ipc = 0.0;
114#else
115 cprintf("%lli: %s progress event, total committed:%i, progress insts "
116 "committed: %lli\n", curTick(), cpu->name(), temp,
117 temp - lastNumInst);
118#endif
119 lastNumInst = temp;
120}
121
122const char *
123CPUProgressEvent::description() const
124{
125 return "CPU Progress";
126}
127
128BaseCPU::BaseCPU(Params *p, bool is_checker)
129 : MemObject(p), instCnt(0), _cpuId(p->cpu_id), _socketId(p->socket_id),
130 _instMasterId(p->system->getMasterId(name() + ".inst")),
131 _dataMasterId(p->system->getMasterId(name() + ".data")),
132 _taskId(ContextSwitchTaskId::Unknown), _pid(invldPid),
133 _switchedOut(p->switched_out), _cacheLineSize(p->system->cacheLineSize()),
134 interrupts(p->interrupts), profileEvent(NULL),
135 numThreads(p->numThreads), system(p->system),
136 previousCycle(0), previousState(CPU_STATE_SLEEP),
137 functionTraceStream(nullptr), currentFunctionStart(0),
138 currentFunctionEnd(0), functionEntryTick(0),
139 addressMonitor(p->numThreads),
140 syscallRetryLatency(p->syscallRetryLatency),
141 pwrGatingLatency(p->pwr_gating_latency),
142 powerGatingOnIdle(p->power_gating_on_idle),
143 enterPwrGatingEvent([this]{ enterPwrGating(); }, name())
144{
145 // if Python did not provide a valid ID, do it here
146 if (_cpuId == -1 ) {
147 _cpuId = cpuList.size();
148 }
149
150 // add self to global list of CPUs
151 cpuList.push_back(this);
152
153 DPRINTF(SyscallVerbose, "Constructing CPU with id %d, socket id %d\n",
154 _cpuId, _socketId);
155
156 if (numThreads > maxThreadsPerCPU)
157 maxThreadsPerCPU = numThreads;
158
159 // allocate per-thread instruction-based event queues
160 comInstEventQueue = new EventQueue *[numThreads];
161 for (ThreadID tid = 0; tid < numThreads; ++tid)
162 comInstEventQueue[tid] =
163 new EventQueue("instruction-based event queue");
164
165 //
166 // set up instruction-count-based termination events, if any
167 //
168 if (p->max_insts_any_thread != 0) {
169 const char *cause = "a thread reached the max instruction count";
170 for (ThreadID tid = 0; tid < numThreads; ++tid)
171 scheduleInstStop(tid, p->max_insts_any_thread, cause);
172 }
173
174 // Set up instruction-count-based termination events for SimPoints
175 // Typically, there are more than one action points.
176 // Simulation.py is responsible to take the necessary actions upon
177 // exitting the simulation loop.
178 if (!p->simpoint_start_insts.empty()) {
179 const char *cause = "simpoint starting point found";
180 for (size_t i = 0; i < p->simpoint_start_insts.size(); ++i)
181 scheduleInstStop(0, p->simpoint_start_insts[i], cause);
182 }
183
184 if (p->max_insts_all_threads != 0) {
185 const char *cause = "all threads reached the max instruction count";
186
187 // allocate & initialize shared downcounter: each event will
188 // decrement this when triggered; simulation will terminate
189 // when counter reaches 0
190 int *counter = new int;
191 *counter = numThreads;
192 for (ThreadID tid = 0; tid < numThreads; ++tid) {
193 Event *event = new CountedExitEvent(cause, *counter);
194 comInstEventQueue[tid]->schedule(event, p->max_insts_all_threads);
195 }
196 }
197
198 // allocate per-thread load-based event queues
199 comLoadEventQueue = new EventQueue *[numThreads];
200 for (ThreadID tid = 0; tid < numThreads; ++tid)
201 comLoadEventQueue[tid] = new EventQueue("load-based event queue");
202
203 //
204 // set up instruction-count-based termination events, if any
205 //
206 if (p->max_loads_any_thread != 0) {
207 const char *cause = "a thread reached the max load count";
208 for (ThreadID tid = 0; tid < numThreads; ++tid)
209 scheduleLoadStop(tid, p->max_loads_any_thread, cause);
210 }
211
212 if (p->max_loads_all_threads != 0) {
213 const char *cause = "all threads reached the max load count";
214 // allocate & initialize shared downcounter: each event will
215 // decrement this when triggered; simulation will terminate
216 // when counter reaches 0
217 int *counter = new int;
218 *counter = numThreads;
219 for (ThreadID tid = 0; tid < numThreads; ++tid) {
220 Event *event = new CountedExitEvent(cause, *counter);
221 comLoadEventQueue[tid]->schedule(event, p->max_loads_all_threads);
222 }
223 }
224
225 functionTracingEnabled = false;
226 if (p->function_trace) {
227 const string fname = csprintf("ftrace.%s", name());
228 functionTraceStream = simout.findOrCreate(fname)->stream();
229
230 currentFunctionStart = currentFunctionEnd = 0;
231 functionEntryTick = p->function_trace_start;
232
233 if (p->function_trace_start == 0) {
234 functionTracingEnabled = true;
235 } else {
236 Event *event = new EventFunctionWrapper(
237 [this]{ enableFunctionTrace(); }, name(), true);
238 schedule(event, p->function_trace_start);
239 }
240 }
241
242 // The interrupts should always be present unless this CPU is
243 // switched in later or in case it is a checker CPU
244 if (!params()->switched_out && !is_checker) {
245 fatal_if(interrupts.size() != numThreads,
246 "CPU %s has %i interrupt controllers, but is expecting one "
247 "per thread (%i)\n",
248 name(), interrupts.size(), numThreads);
249 for (ThreadID tid = 0; tid < numThreads; tid++)
250 interrupts[tid]->setCPU(this);
251 }
252
253 if (FullSystem) {
254 if (params()->profile)
255 profileEvent = new EventFunctionWrapper(
256 [this]{ processProfileEvent(); },
257 name());
258 }
259 tracer = params()->tracer;
260
261 if (params()->isa.size() != numThreads) {
262 fatal("Number of ISAs (%i) assigned to the CPU does not equal number "
263 "of threads (%i).\n", params()->isa.size(), numThreads);
264 }
265}
266
267void
268BaseCPU::enableFunctionTrace()
269{
270 functionTracingEnabled = true;
271}
272
273BaseCPU::~BaseCPU()
274{
275 delete profileEvent;
276 delete[] comLoadEventQueue;
277 delete[] comInstEventQueue;
278}
279
280void
281BaseCPU::armMonitor(ThreadID tid, Addr address)
282{
283 assert(tid < numThreads);
284 AddressMonitor &monitor = addressMonitor[tid];
285
286 monitor.armed = true;
287 monitor.vAddr = address;
288 monitor.pAddr = 0x0;
289 DPRINTF(Mwait,"[tid:%d] Armed monitor (vAddr=0x%lx)\n", tid, address);
290}
291
292bool
293BaseCPU::mwait(ThreadID tid, PacketPtr pkt)
294{
295 assert(tid < numThreads);
296 AddressMonitor &monitor = addressMonitor[tid];
297
298 if (!monitor.gotWakeup) {
299 int block_size = cacheLineSize();
300 uint64_t mask = ~((uint64_t)(block_size - 1));
301
302 assert(pkt->req->hasPaddr());
303 monitor.pAddr = pkt->getAddr() & mask;
304 monitor.waiting = true;
305
306 DPRINTF(Mwait,"[tid:%d] mwait called (vAddr=0x%lx, "
307 "line's paddr=0x%lx)\n", tid, monitor.vAddr, monitor.pAddr);
308 return true;
309 } else {
310 monitor.gotWakeup = false;
311 return false;
312 }
313}
314
315void
316BaseCPU::mwaitAtomic(ThreadID tid, ThreadContext *tc, TheISA::TLB *dtb)
317{
318 assert(tid < numThreads);
319 AddressMonitor &monitor = addressMonitor[tid];
320
321 Request req;
322 Addr addr = monitor.vAddr;
323 int block_size = cacheLineSize();
324 uint64_t mask = ~((uint64_t)(block_size - 1));
325 int size = block_size;
326
327 //The address of the next line if it crosses a cache line boundary.
328 Addr secondAddr = roundDown(addr + size - 1, block_size);
329
330 if (secondAddr > addr)
331 size = secondAddr - addr;
332
333 req.setVirt(0, addr, size, 0x0, dataMasterId(), tc->instAddr());
334
335 // translate to physical address
336 Fault fault = dtb->translateAtomic(&req, tc, BaseTLB::Read);
337 assert(fault == NoFault);
338
339 monitor.pAddr = req.getPaddr() & mask;
340 monitor.waiting = true;
341
342 DPRINTF(Mwait,"[tid:%d] mwait called (vAddr=0x%lx, line's paddr=0x%lx)\n",
343 tid, monitor.vAddr, monitor.pAddr);
344}
345
346void
347BaseCPU::init()
348{
349 if (!params()->switched_out) {
350 registerThreadContexts();
351
352 verifyMemoryMode();
353 }
354}
355
356void
357BaseCPU::startup()
358{
359 if (FullSystem) {
360 if (!params()->switched_out && profileEvent)
361 schedule(profileEvent, curTick());
362 }
363
364 if (params()->progress_interval) {
365 new CPUProgressEvent(this, params()->progress_interval);
366 }
367
368 if (_switchedOut)
369 ClockedObject::pwrState(Enums::PwrState::OFF);
370
371 // Assumption CPU start to operate instantaneously without any latency
372 if (ClockedObject::pwrState() == Enums::PwrState::UNDEFINED)
373 ClockedObject::pwrState(Enums::PwrState::ON);
374
375}
376
377ProbePoints::PMUUPtr
378BaseCPU::pmuProbePoint(const char *name)
379{
380 ProbePoints::PMUUPtr ptr;
381 ptr.reset(new ProbePoints::PMU(getProbeManager(), name));
382
383 return ptr;
384}
385
386void
387BaseCPU::regProbePoints()
388{
389 ppAllCycles = pmuProbePoint("Cycles");
390 ppActiveCycles = pmuProbePoint("ActiveCycles");
391
392 ppRetiredInsts = pmuProbePoint("RetiredInsts");
393 ppRetiredLoads = pmuProbePoint("RetiredLoads");
394 ppRetiredStores = pmuProbePoint("RetiredStores");
395 ppRetiredBranches = pmuProbePoint("RetiredBranches");
396
397 ppSleeping = new ProbePointArg<bool>(this->getProbeManager(),
398 "Sleeping");
399}
400
401void
402BaseCPU::probeInstCommit(const StaticInstPtr &inst)
403{
404 if (!inst->isMicroop() || inst->isLastMicroop())
405 ppRetiredInsts->notify(1);
406
407
408 if (inst->isLoad())
409 ppRetiredLoads->notify(1);
410
411 if (inst->isStore())
412 ppRetiredStores->notify(1);
413
414 if (inst->isControl())
415 ppRetiredBranches->notify(1);
416}
417
418void
419BaseCPU::regStats()
420{
421 MemObject::regStats();
422
423 using namespace Stats;
424
425 numCycles
426 .name(name() + ".numCycles")
427 .desc("number of cpu cycles simulated")
428 ;
429
430 numWorkItemsStarted
431 .name(name() + ".numWorkItemsStarted")
432 .desc("number of work items this cpu started")
433 ;
434
435 numWorkItemsCompleted
436 .name(name() + ".numWorkItemsCompleted")
437 .desc("number of work items this cpu completed")
438 ;
439
440 int size = threadContexts.size();
441 if (size > 1) {
442 for (int i = 0; i < size; ++i) {
443 stringstream namestr;
444 ccprintf(namestr, "%s.ctx%d", name(), i);
445 threadContexts[i]->regStats(namestr.str());
446 }
447 } else if (size == 1)
448 threadContexts[0]->regStats(name());
449}
450
451BaseMasterPort &
452BaseCPU::getMasterPort(const string &if_name, PortID idx)
453{
454 // Get the right port based on name. This applies to all the
455 // subclasses of the base CPU and relies on their implementation
456 // of getDataPort and getInstPort. In all cases there methods
457 // return a MasterPort pointer.
458 if (if_name == "dcache_port")
459 return getDataPort();
460 else if (if_name == "icache_port")
461 return getInstPort();
462 else
463 return MemObject::getMasterPort(if_name, idx);
464}
465
466void
467BaseCPU::registerThreadContexts()
468{
469 assert(system->multiThread || numThreads == 1);
470
471 ThreadID size = threadContexts.size();
472 for (ThreadID tid = 0; tid < size; ++tid) {
473 ThreadContext *tc = threadContexts[tid];
474
475 if (system->multiThread) {
476 tc->setContextId(system->registerThreadContext(tc));
477 } else {
478 tc->setContextId(system->registerThreadContext(tc, _cpuId));
479 }
480
481 if (!FullSystem)
482 tc->getProcessPtr()->assignThreadContext(tc->contextId());
483 }
484}
485
486void
487BaseCPU::deschedulePowerGatingEvent()
488{
489 if (enterPwrGatingEvent.scheduled()){
490 deschedule(enterPwrGatingEvent);
491 }
492}
493
494void
495BaseCPU::schedulePowerGatingEvent()
496{
497 for (auto tc : threadContexts) {
498 if (tc->status() == ThreadContext::Active)
499 return;
500 }
501
502 if (ClockedObject::pwrState() == Enums::PwrState::CLK_GATED &&
503 powerGatingOnIdle) {
504 assert(!enterPwrGatingEvent.scheduled());
505 // Schedule a power gating event when clock gated for the specified
506 // amount of time
507 schedule(enterPwrGatingEvent, clockEdge(pwrGatingLatency));
508 }
509}
510
511int
512BaseCPU::findContext(ThreadContext *tc)
513{
514 ThreadID size = threadContexts.size();
515 for (ThreadID tid = 0; tid < size; ++tid) {
516 if (tc == threadContexts[tid])
517 return tid;
518 }
519 return 0;
520}
521
522void
523BaseCPU::activateContext(ThreadID thread_num)
524{
525 // Squash enter power gating event while cpu gets activated
526 if (enterPwrGatingEvent.scheduled())
527 deschedule(enterPwrGatingEvent);
528 // For any active thread running, update CPU power state to active (ON)
529 ClockedObject::pwrState(Enums::PwrState::ON);
530
531 updateCycleCounters(CPU_STATE_WAKEUP);
532}
533
534void
535BaseCPU::suspendContext(ThreadID thread_num)
536{
537 // Check if all threads are suspended
538 for (auto t : threadContexts) {
539 if (t->status() != ThreadContext::Suspended) {
540 return;
541 }
542 }
543
544 // All CPU thread are suspended, update cycle count
545 updateCycleCounters(CPU_STATE_SLEEP);
546
547 // All CPU threads suspended, enter lower power state for the CPU
548 ClockedObject::pwrState(Enums::PwrState::CLK_GATED);
549
550 // If pwrGatingLatency is set to 0 then this mechanism is disabled
551 if (powerGatingOnIdle) {
552 // Schedule power gating event when clock gated for pwrGatingLatency
553 // cycles
554 schedule(enterPwrGatingEvent, clockEdge(pwrGatingLatency));
555 }
556}
557
558void
559BaseCPU::haltContext(ThreadID thread_num)
560{
561 updateCycleCounters(BaseCPU::CPU_STATE_SLEEP);
562}
563
564void
565BaseCPU::enterPwrGating(void)
566{
567 ClockedObject::pwrState(Enums::PwrState::OFF);
568}
569
570void
571BaseCPU::switchOut()
572{
573 assert(!_switchedOut);
574 _switchedOut = true;
575 if (profileEvent && profileEvent->scheduled())
576 deschedule(profileEvent);
577
578 // Flush all TLBs in the CPU to avoid having stale translations if
579 // it gets switched in later.
580 flushTLBs();
581
582 // Go to the power gating state
583 ClockedObject::pwrState(Enums::PwrState::OFF);
584}
585
586void
587BaseCPU::takeOverFrom(BaseCPU *oldCPU)
588{
589 assert(threadContexts.size() == oldCPU->threadContexts.size());
590 assert(_cpuId == oldCPU->cpuId());
591 assert(_switchedOut);
592 assert(oldCPU != this);
593 _pid = oldCPU->getPid();
594 _taskId = oldCPU->taskId();
595 // Take over the power state of the switchedOut CPU
596 ClockedObject::pwrState(oldCPU->pwrState());
597
598 previousState = oldCPU->previousState;
599 previousCycle = oldCPU->previousCycle;
600
601 _switchedOut = false;
602
603 ThreadID size = threadContexts.size();
604 for (ThreadID i = 0; i < size; ++i) {
605 ThreadContext *newTC = threadContexts[i];
606 ThreadContext *oldTC = oldCPU->threadContexts[i];
607
608 newTC->takeOverFrom(oldTC);
609
610 CpuEvent::replaceThreadContext(oldTC, newTC);
611
612 assert(newTC->contextId() == oldTC->contextId());
613 assert(newTC->threadId() == oldTC->threadId());
614 system->replaceThreadContext(newTC, newTC->contextId());
615
616 /* This code no longer works since the zero register (e.g.,
617 * r31 on Alpha) doesn't necessarily contain zero at this
618 * point.
619 if (DTRACE(Context))
620 ThreadContext::compare(oldTC, newTC);
621 */
622
623 BaseMasterPort *old_itb_port = oldTC->getITBPtr()->getMasterPort();
624 BaseMasterPort *old_dtb_port = oldTC->getDTBPtr()->getMasterPort();
625 BaseMasterPort *new_itb_port = newTC->getITBPtr()->getMasterPort();
626 BaseMasterPort *new_dtb_port = newTC->getDTBPtr()->getMasterPort();
627
628 // Move over any table walker ports if they exist
629 if (new_itb_port) {
630 assert(!new_itb_port->isConnected());
631 assert(old_itb_port);
632 assert(old_itb_port->isConnected());
633 BaseSlavePort &slavePort = old_itb_port->getSlavePort();
634 old_itb_port->unbind();
635 new_itb_port->bind(slavePort);
636 }
637 if (new_dtb_port) {
638 assert(!new_dtb_port->isConnected());
639 assert(old_dtb_port);
640 assert(old_dtb_port->isConnected());
641 BaseSlavePort &slavePort = old_dtb_port->getSlavePort();
642 old_dtb_port->unbind();
643 new_dtb_port->bind(slavePort);
644 }
645 newTC->getITBPtr()->takeOverFrom(oldTC->getITBPtr());
646 newTC->getDTBPtr()->takeOverFrom(oldTC->getDTBPtr());
647
648 // Checker whether or not we have to transfer CheckerCPU
649 // objects over in the switch
650 CheckerCPU *oldChecker = oldTC->getCheckerCpuPtr();
651 CheckerCPU *newChecker = newTC->getCheckerCpuPtr();
652 if (oldChecker && newChecker) {
653 BaseMasterPort *old_checker_itb_port =
654 oldChecker->getITBPtr()->getMasterPort();
655 BaseMasterPort *old_checker_dtb_port =
656 oldChecker->getDTBPtr()->getMasterPort();
657 BaseMasterPort *new_checker_itb_port =
658 newChecker->getITBPtr()->getMasterPort();
659 BaseMasterPort *new_checker_dtb_port =
660 newChecker->getDTBPtr()->getMasterPort();
661
662 newChecker->getITBPtr()->takeOverFrom(oldChecker->getITBPtr());
663 newChecker->getDTBPtr()->takeOverFrom(oldChecker->getDTBPtr());
664
665 // Move over any table walker ports if they exist for checker
666 if (new_checker_itb_port) {
667 assert(!new_checker_itb_port->isConnected());
668 assert(old_checker_itb_port);
669 assert(old_checker_itb_port->isConnected());
670 BaseSlavePort &slavePort =
671 old_checker_itb_port->getSlavePort();
672 old_checker_itb_port->unbind();
673 new_checker_itb_port->bind(slavePort);
674 }
675 if (new_checker_dtb_port) {
676 assert(!new_checker_dtb_port->isConnected());
677 assert(old_checker_dtb_port);
678 assert(old_checker_dtb_port->isConnected());
679 BaseSlavePort &slavePort =
680 old_checker_dtb_port->getSlavePort();
681 old_checker_dtb_port->unbind();
682 new_checker_dtb_port->bind(slavePort);
683 }
684 }
685 }
686
687 interrupts = oldCPU->interrupts;
688 for (ThreadID tid = 0; tid < numThreads; tid++) {
689 interrupts[tid]->setCPU(this);
690 }
691 oldCPU->interrupts.clear();
692
693 if (FullSystem) {
694 for (ThreadID i = 0; i < size; ++i)
695 threadContexts[i]->profileClear();
696
697 if (profileEvent)
698 schedule(profileEvent, curTick());
699 }
700
701 // All CPUs have an instruction and a data port, and the new CPU's
702 // ports are dangling while the old CPU has its ports connected
703 // already. Unbind the old CPU and then bind the ports of the one
704 // we are switching to.
705 assert(!getInstPort().isConnected());
706 assert(oldCPU->getInstPort().isConnected());
707 BaseSlavePort &inst_peer_port = oldCPU->getInstPort().getSlavePort();
708 oldCPU->getInstPort().unbind();
709 getInstPort().bind(inst_peer_port);
710
711 assert(!getDataPort().isConnected());
712 assert(oldCPU->getDataPort().isConnected());
713 BaseSlavePort &data_peer_port = oldCPU->getDataPort().getSlavePort();
714 oldCPU->getDataPort().unbind();
715 getDataPort().bind(data_peer_port);
716}
717
718void
719BaseCPU::flushTLBs()
720{
721 for (ThreadID i = 0; i < threadContexts.size(); ++i) {
722 ThreadContext &tc(*threadContexts[i]);
723 CheckerCPU *checker(tc.getCheckerCpuPtr());
724
725 tc.getITBPtr()->flushAll();
726 tc.getDTBPtr()->flushAll();
727 if (checker) {
728 checker->getITBPtr()->flushAll();
729 checker->getDTBPtr()->flushAll();
730 }
731 }
732}
733
734void
735BaseCPU::processProfileEvent()
736{
737 ThreadID size = threadContexts.size();
738
739 for (ThreadID i = 0; i < size; ++i)
740 threadContexts[i]->profileSample();
741
742 schedule(profileEvent, curTick() + params()->profile);
743}
744
745void
746BaseCPU::serialize(CheckpointOut &cp) const
747{
748 SERIALIZE_SCALAR(instCnt);
749
750 if (!_switchedOut) {
751 /* Unlike _pid, _taskId is not serialized, as they are dynamically
752 * assigned unique ids that are only meaningful for the duration of
753 * a specific run. We will need to serialize the entire taskMap in
754 * system. */
755 SERIALIZE_SCALAR(_pid);
756
757 // Serialize the threads, this is done by the CPU implementation.
758 for (ThreadID i = 0; i < numThreads; ++i) {
759 ScopedCheckpointSection sec(cp, csprintf("xc.%i", i));
760 interrupts[i]->serialize(cp);
761 serializeThread(cp, i);
762 }
763 }
764}
765
766void
767BaseCPU::unserialize(CheckpointIn &cp)
768{
769 UNSERIALIZE_SCALAR(instCnt);
770
771 if (!_switchedOut) {
772 UNSERIALIZE_SCALAR(_pid);
773
774 // Unserialize the threads, this is done by the CPU implementation.
775 for (ThreadID i = 0; i < numThreads; ++i) {
776 ScopedCheckpointSection sec(cp, csprintf("xc.%i", i));
777 interrupts[i]->unserialize(cp);
778 unserializeThread(cp, i);
779 }
780 }
781}
782
783void
784BaseCPU::scheduleInstStop(ThreadID tid, Counter insts, const char *cause)
785{
786 const Tick now(comInstEventQueue[tid]->getCurTick());
787 Event *event(new LocalSimLoopExitEvent(cause, 0));
788
789 comInstEventQueue[tid]->schedule(event, now + insts);
790}
791
792uint64_t
793BaseCPU::getCurrentInstCount(ThreadID tid)
794{
795 return Tick(comInstEventQueue[tid]->getCurTick());
796}
797
798AddressMonitor::AddressMonitor() {
799 armed = false;
800 waiting = false;
801 gotWakeup = false;
802}
803
804bool AddressMonitor::doMonitor(PacketPtr pkt) {
805 assert(pkt->req->hasPaddr());
806 if (armed && waiting) {
807 if (pAddr == pkt->getAddr()) {
808 DPRINTF(Mwait,"pAddr=0x%lx invalidated: waking up core\n",
809 pkt->getAddr());
810 waiting = false;
811 return true;
812 }
813 }
814 return false;
815}
816
817void
818BaseCPU::scheduleLoadStop(ThreadID tid, Counter loads, const char *cause)
819{
820 const Tick now(comLoadEventQueue[tid]->getCurTick());
821 Event *event(new LocalSimLoopExitEvent(cause, 0));
822
823 comLoadEventQueue[tid]->schedule(event, now + loads);
824}
825
826
827void
828BaseCPU::traceFunctionsInternal(Addr pc)
829{
830 if (!debugSymbolTable)
831 return;
832
833 // if pc enters different function, print new function symbol and
834 // update saved range. Otherwise do nothing.
835 if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
836 string sym_str;
837 bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
838 currentFunctionStart,
839 currentFunctionEnd);
840
841 if (!found) {
842 // no symbol found: use addr as label
843 sym_str = csprintf("0x%x", pc);
844 currentFunctionStart = pc;
845 currentFunctionEnd = pc + 1;
846 }
847
848 ccprintf(*functionTraceStream, " (%d)\n%d: %s",
849 curTick() - functionEntryTick, curTick(), sym_str);
850 functionEntryTick = curTick();
851 }
852}
853
854bool
855BaseCPU::waitForRemoteGDB() const
856{
857 return params()->wait_for_remote_gdb;
858}
58#include "base/output.hh"
59#include "base/trace.hh"
60#include "cpu/checker/cpu.hh"
61#include "cpu/cpuevent.hh"
62#include "cpu/profile.hh"
63#include "cpu/thread_context.hh"
64#include "debug/Mwait.hh"
65#include "debug/SyscallVerbose.hh"
66#include "mem/page_table.hh"
67#include "params/BaseCPU.hh"
68#include "sim/clocked_object.hh"
69#include "sim/full_system.hh"
70#include "sim/process.hh"
71#include "sim/sim_events.hh"
72#include "sim/sim_exit.hh"
73#include "sim/system.hh"
74
75// Hack
76#include "sim/stat_control.hh"
77
78using namespace std;
79
80vector<BaseCPU *> BaseCPU::cpuList;
81
82// This variable reflects the max number of threads in any CPU. Be
83// careful to only use it once all the CPUs that you care about have
84// been initialized
85int maxThreadsPerCPU = 1;
86
87CPUProgressEvent::CPUProgressEvent(BaseCPU *_cpu, Tick ival)
88 : Event(Event::Progress_Event_Pri), _interval(ival), lastNumInst(0),
89 cpu(_cpu), _repeatEvent(true)
90{
91 if (_interval)
92 cpu->schedule(this, curTick() + _interval);
93}
94
95void
96CPUProgressEvent::process()
97{
98 Counter temp = cpu->totalOps();
99
100 if (_repeatEvent)
101 cpu->schedule(this, curTick() + _interval);
102
103 if (cpu->switchedOut()) {
104 return;
105 }
106
107#ifndef NDEBUG
108 double ipc = double(temp - lastNumInst) / (_interval / cpu->clockPeriod());
109
110 DPRINTFN("%s progress event, total committed:%i, progress insts committed: "
111 "%lli, IPC: %0.8d\n", cpu->name(), temp, temp - lastNumInst,
112 ipc);
113 ipc = 0.0;
114#else
115 cprintf("%lli: %s progress event, total committed:%i, progress insts "
116 "committed: %lli\n", curTick(), cpu->name(), temp,
117 temp - lastNumInst);
118#endif
119 lastNumInst = temp;
120}
121
122const char *
123CPUProgressEvent::description() const
124{
125 return "CPU Progress";
126}
127
128BaseCPU::BaseCPU(Params *p, bool is_checker)
129 : MemObject(p), instCnt(0), _cpuId(p->cpu_id), _socketId(p->socket_id),
130 _instMasterId(p->system->getMasterId(name() + ".inst")),
131 _dataMasterId(p->system->getMasterId(name() + ".data")),
132 _taskId(ContextSwitchTaskId::Unknown), _pid(invldPid),
133 _switchedOut(p->switched_out), _cacheLineSize(p->system->cacheLineSize()),
134 interrupts(p->interrupts), profileEvent(NULL),
135 numThreads(p->numThreads), system(p->system),
136 previousCycle(0), previousState(CPU_STATE_SLEEP),
137 functionTraceStream(nullptr), currentFunctionStart(0),
138 currentFunctionEnd(0), functionEntryTick(0),
139 addressMonitor(p->numThreads),
140 syscallRetryLatency(p->syscallRetryLatency),
141 pwrGatingLatency(p->pwr_gating_latency),
142 powerGatingOnIdle(p->power_gating_on_idle),
143 enterPwrGatingEvent([this]{ enterPwrGating(); }, name())
144{
145 // if Python did not provide a valid ID, do it here
146 if (_cpuId == -1 ) {
147 _cpuId = cpuList.size();
148 }
149
150 // add self to global list of CPUs
151 cpuList.push_back(this);
152
153 DPRINTF(SyscallVerbose, "Constructing CPU with id %d, socket id %d\n",
154 _cpuId, _socketId);
155
156 if (numThreads > maxThreadsPerCPU)
157 maxThreadsPerCPU = numThreads;
158
159 // allocate per-thread instruction-based event queues
160 comInstEventQueue = new EventQueue *[numThreads];
161 for (ThreadID tid = 0; tid < numThreads; ++tid)
162 comInstEventQueue[tid] =
163 new EventQueue("instruction-based event queue");
164
165 //
166 // set up instruction-count-based termination events, if any
167 //
168 if (p->max_insts_any_thread != 0) {
169 const char *cause = "a thread reached the max instruction count";
170 for (ThreadID tid = 0; tid < numThreads; ++tid)
171 scheduleInstStop(tid, p->max_insts_any_thread, cause);
172 }
173
174 // Set up instruction-count-based termination events for SimPoints
175 // Typically, there are more than one action points.
176 // Simulation.py is responsible to take the necessary actions upon
177 // exitting the simulation loop.
178 if (!p->simpoint_start_insts.empty()) {
179 const char *cause = "simpoint starting point found";
180 for (size_t i = 0; i < p->simpoint_start_insts.size(); ++i)
181 scheduleInstStop(0, p->simpoint_start_insts[i], cause);
182 }
183
184 if (p->max_insts_all_threads != 0) {
185 const char *cause = "all threads reached the max instruction count";
186
187 // allocate & initialize shared downcounter: each event will
188 // decrement this when triggered; simulation will terminate
189 // when counter reaches 0
190 int *counter = new int;
191 *counter = numThreads;
192 for (ThreadID tid = 0; tid < numThreads; ++tid) {
193 Event *event = new CountedExitEvent(cause, *counter);
194 comInstEventQueue[tid]->schedule(event, p->max_insts_all_threads);
195 }
196 }
197
198 // allocate per-thread load-based event queues
199 comLoadEventQueue = new EventQueue *[numThreads];
200 for (ThreadID tid = 0; tid < numThreads; ++tid)
201 comLoadEventQueue[tid] = new EventQueue("load-based event queue");
202
203 //
204 // set up instruction-count-based termination events, if any
205 //
206 if (p->max_loads_any_thread != 0) {
207 const char *cause = "a thread reached the max load count";
208 for (ThreadID tid = 0; tid < numThreads; ++tid)
209 scheduleLoadStop(tid, p->max_loads_any_thread, cause);
210 }
211
212 if (p->max_loads_all_threads != 0) {
213 const char *cause = "all threads reached the max load count";
214 // allocate & initialize shared downcounter: each event will
215 // decrement this when triggered; simulation will terminate
216 // when counter reaches 0
217 int *counter = new int;
218 *counter = numThreads;
219 for (ThreadID tid = 0; tid < numThreads; ++tid) {
220 Event *event = new CountedExitEvent(cause, *counter);
221 comLoadEventQueue[tid]->schedule(event, p->max_loads_all_threads);
222 }
223 }
224
225 functionTracingEnabled = false;
226 if (p->function_trace) {
227 const string fname = csprintf("ftrace.%s", name());
228 functionTraceStream = simout.findOrCreate(fname)->stream();
229
230 currentFunctionStart = currentFunctionEnd = 0;
231 functionEntryTick = p->function_trace_start;
232
233 if (p->function_trace_start == 0) {
234 functionTracingEnabled = true;
235 } else {
236 Event *event = new EventFunctionWrapper(
237 [this]{ enableFunctionTrace(); }, name(), true);
238 schedule(event, p->function_trace_start);
239 }
240 }
241
242 // The interrupts should always be present unless this CPU is
243 // switched in later or in case it is a checker CPU
244 if (!params()->switched_out && !is_checker) {
245 fatal_if(interrupts.size() != numThreads,
246 "CPU %s has %i interrupt controllers, but is expecting one "
247 "per thread (%i)\n",
248 name(), interrupts.size(), numThreads);
249 for (ThreadID tid = 0; tid < numThreads; tid++)
250 interrupts[tid]->setCPU(this);
251 }
252
253 if (FullSystem) {
254 if (params()->profile)
255 profileEvent = new EventFunctionWrapper(
256 [this]{ processProfileEvent(); },
257 name());
258 }
259 tracer = params()->tracer;
260
261 if (params()->isa.size() != numThreads) {
262 fatal("Number of ISAs (%i) assigned to the CPU does not equal number "
263 "of threads (%i).\n", params()->isa.size(), numThreads);
264 }
265}
266
267void
268BaseCPU::enableFunctionTrace()
269{
270 functionTracingEnabled = true;
271}
272
273BaseCPU::~BaseCPU()
274{
275 delete profileEvent;
276 delete[] comLoadEventQueue;
277 delete[] comInstEventQueue;
278}
279
280void
281BaseCPU::armMonitor(ThreadID tid, Addr address)
282{
283 assert(tid < numThreads);
284 AddressMonitor &monitor = addressMonitor[tid];
285
286 monitor.armed = true;
287 monitor.vAddr = address;
288 monitor.pAddr = 0x0;
289 DPRINTF(Mwait,"[tid:%d] Armed monitor (vAddr=0x%lx)\n", tid, address);
290}
291
292bool
293BaseCPU::mwait(ThreadID tid, PacketPtr pkt)
294{
295 assert(tid < numThreads);
296 AddressMonitor &monitor = addressMonitor[tid];
297
298 if (!monitor.gotWakeup) {
299 int block_size = cacheLineSize();
300 uint64_t mask = ~((uint64_t)(block_size - 1));
301
302 assert(pkt->req->hasPaddr());
303 monitor.pAddr = pkt->getAddr() & mask;
304 monitor.waiting = true;
305
306 DPRINTF(Mwait,"[tid:%d] mwait called (vAddr=0x%lx, "
307 "line's paddr=0x%lx)\n", tid, monitor.vAddr, monitor.pAddr);
308 return true;
309 } else {
310 monitor.gotWakeup = false;
311 return false;
312 }
313}
314
315void
316BaseCPU::mwaitAtomic(ThreadID tid, ThreadContext *tc, TheISA::TLB *dtb)
317{
318 assert(tid < numThreads);
319 AddressMonitor &monitor = addressMonitor[tid];
320
321 Request req;
322 Addr addr = monitor.vAddr;
323 int block_size = cacheLineSize();
324 uint64_t mask = ~((uint64_t)(block_size - 1));
325 int size = block_size;
326
327 //The address of the next line if it crosses a cache line boundary.
328 Addr secondAddr = roundDown(addr + size - 1, block_size);
329
330 if (secondAddr > addr)
331 size = secondAddr - addr;
332
333 req.setVirt(0, addr, size, 0x0, dataMasterId(), tc->instAddr());
334
335 // translate to physical address
336 Fault fault = dtb->translateAtomic(&req, tc, BaseTLB::Read);
337 assert(fault == NoFault);
338
339 monitor.pAddr = req.getPaddr() & mask;
340 monitor.waiting = true;
341
342 DPRINTF(Mwait,"[tid:%d] mwait called (vAddr=0x%lx, line's paddr=0x%lx)\n",
343 tid, monitor.vAddr, monitor.pAddr);
344}
345
346void
347BaseCPU::init()
348{
349 if (!params()->switched_out) {
350 registerThreadContexts();
351
352 verifyMemoryMode();
353 }
354}
355
356void
357BaseCPU::startup()
358{
359 if (FullSystem) {
360 if (!params()->switched_out && profileEvent)
361 schedule(profileEvent, curTick());
362 }
363
364 if (params()->progress_interval) {
365 new CPUProgressEvent(this, params()->progress_interval);
366 }
367
368 if (_switchedOut)
369 ClockedObject::pwrState(Enums::PwrState::OFF);
370
371 // Assumption CPU start to operate instantaneously without any latency
372 if (ClockedObject::pwrState() == Enums::PwrState::UNDEFINED)
373 ClockedObject::pwrState(Enums::PwrState::ON);
374
375}
376
377ProbePoints::PMUUPtr
378BaseCPU::pmuProbePoint(const char *name)
379{
380 ProbePoints::PMUUPtr ptr;
381 ptr.reset(new ProbePoints::PMU(getProbeManager(), name));
382
383 return ptr;
384}
385
386void
387BaseCPU::regProbePoints()
388{
389 ppAllCycles = pmuProbePoint("Cycles");
390 ppActiveCycles = pmuProbePoint("ActiveCycles");
391
392 ppRetiredInsts = pmuProbePoint("RetiredInsts");
393 ppRetiredLoads = pmuProbePoint("RetiredLoads");
394 ppRetiredStores = pmuProbePoint("RetiredStores");
395 ppRetiredBranches = pmuProbePoint("RetiredBranches");
396
397 ppSleeping = new ProbePointArg<bool>(this->getProbeManager(),
398 "Sleeping");
399}
400
401void
402BaseCPU::probeInstCommit(const StaticInstPtr &inst)
403{
404 if (!inst->isMicroop() || inst->isLastMicroop())
405 ppRetiredInsts->notify(1);
406
407
408 if (inst->isLoad())
409 ppRetiredLoads->notify(1);
410
411 if (inst->isStore())
412 ppRetiredStores->notify(1);
413
414 if (inst->isControl())
415 ppRetiredBranches->notify(1);
416}
417
418void
419BaseCPU::regStats()
420{
421 MemObject::regStats();
422
423 using namespace Stats;
424
425 numCycles
426 .name(name() + ".numCycles")
427 .desc("number of cpu cycles simulated")
428 ;
429
430 numWorkItemsStarted
431 .name(name() + ".numWorkItemsStarted")
432 .desc("number of work items this cpu started")
433 ;
434
435 numWorkItemsCompleted
436 .name(name() + ".numWorkItemsCompleted")
437 .desc("number of work items this cpu completed")
438 ;
439
440 int size = threadContexts.size();
441 if (size > 1) {
442 for (int i = 0; i < size; ++i) {
443 stringstream namestr;
444 ccprintf(namestr, "%s.ctx%d", name(), i);
445 threadContexts[i]->regStats(namestr.str());
446 }
447 } else if (size == 1)
448 threadContexts[0]->regStats(name());
449}
450
451BaseMasterPort &
452BaseCPU::getMasterPort(const string &if_name, PortID idx)
453{
454 // Get the right port based on name. This applies to all the
455 // subclasses of the base CPU and relies on their implementation
456 // of getDataPort and getInstPort. In all cases there methods
457 // return a MasterPort pointer.
458 if (if_name == "dcache_port")
459 return getDataPort();
460 else if (if_name == "icache_port")
461 return getInstPort();
462 else
463 return MemObject::getMasterPort(if_name, idx);
464}
465
466void
467BaseCPU::registerThreadContexts()
468{
469 assert(system->multiThread || numThreads == 1);
470
471 ThreadID size = threadContexts.size();
472 for (ThreadID tid = 0; tid < size; ++tid) {
473 ThreadContext *tc = threadContexts[tid];
474
475 if (system->multiThread) {
476 tc->setContextId(system->registerThreadContext(tc));
477 } else {
478 tc->setContextId(system->registerThreadContext(tc, _cpuId));
479 }
480
481 if (!FullSystem)
482 tc->getProcessPtr()->assignThreadContext(tc->contextId());
483 }
484}
485
486void
487BaseCPU::deschedulePowerGatingEvent()
488{
489 if (enterPwrGatingEvent.scheduled()){
490 deschedule(enterPwrGatingEvent);
491 }
492}
493
494void
495BaseCPU::schedulePowerGatingEvent()
496{
497 for (auto tc : threadContexts) {
498 if (tc->status() == ThreadContext::Active)
499 return;
500 }
501
502 if (ClockedObject::pwrState() == Enums::PwrState::CLK_GATED &&
503 powerGatingOnIdle) {
504 assert(!enterPwrGatingEvent.scheduled());
505 // Schedule a power gating event when clock gated for the specified
506 // amount of time
507 schedule(enterPwrGatingEvent, clockEdge(pwrGatingLatency));
508 }
509}
510
511int
512BaseCPU::findContext(ThreadContext *tc)
513{
514 ThreadID size = threadContexts.size();
515 for (ThreadID tid = 0; tid < size; ++tid) {
516 if (tc == threadContexts[tid])
517 return tid;
518 }
519 return 0;
520}
521
522void
523BaseCPU::activateContext(ThreadID thread_num)
524{
525 // Squash enter power gating event while cpu gets activated
526 if (enterPwrGatingEvent.scheduled())
527 deschedule(enterPwrGatingEvent);
528 // For any active thread running, update CPU power state to active (ON)
529 ClockedObject::pwrState(Enums::PwrState::ON);
530
531 updateCycleCounters(CPU_STATE_WAKEUP);
532}
533
534void
535BaseCPU::suspendContext(ThreadID thread_num)
536{
537 // Check if all threads are suspended
538 for (auto t : threadContexts) {
539 if (t->status() != ThreadContext::Suspended) {
540 return;
541 }
542 }
543
544 // All CPU thread are suspended, update cycle count
545 updateCycleCounters(CPU_STATE_SLEEP);
546
547 // All CPU threads suspended, enter lower power state for the CPU
548 ClockedObject::pwrState(Enums::PwrState::CLK_GATED);
549
550 // If pwrGatingLatency is set to 0 then this mechanism is disabled
551 if (powerGatingOnIdle) {
552 // Schedule power gating event when clock gated for pwrGatingLatency
553 // cycles
554 schedule(enterPwrGatingEvent, clockEdge(pwrGatingLatency));
555 }
556}
557
558void
559BaseCPU::haltContext(ThreadID thread_num)
560{
561 updateCycleCounters(BaseCPU::CPU_STATE_SLEEP);
562}
563
564void
565BaseCPU::enterPwrGating(void)
566{
567 ClockedObject::pwrState(Enums::PwrState::OFF);
568}
569
570void
571BaseCPU::switchOut()
572{
573 assert(!_switchedOut);
574 _switchedOut = true;
575 if (profileEvent && profileEvent->scheduled())
576 deschedule(profileEvent);
577
578 // Flush all TLBs in the CPU to avoid having stale translations if
579 // it gets switched in later.
580 flushTLBs();
581
582 // Go to the power gating state
583 ClockedObject::pwrState(Enums::PwrState::OFF);
584}
585
586void
587BaseCPU::takeOverFrom(BaseCPU *oldCPU)
588{
589 assert(threadContexts.size() == oldCPU->threadContexts.size());
590 assert(_cpuId == oldCPU->cpuId());
591 assert(_switchedOut);
592 assert(oldCPU != this);
593 _pid = oldCPU->getPid();
594 _taskId = oldCPU->taskId();
595 // Take over the power state of the switchedOut CPU
596 ClockedObject::pwrState(oldCPU->pwrState());
597
598 previousState = oldCPU->previousState;
599 previousCycle = oldCPU->previousCycle;
600
601 _switchedOut = false;
602
603 ThreadID size = threadContexts.size();
604 for (ThreadID i = 0; i < size; ++i) {
605 ThreadContext *newTC = threadContexts[i];
606 ThreadContext *oldTC = oldCPU->threadContexts[i];
607
608 newTC->takeOverFrom(oldTC);
609
610 CpuEvent::replaceThreadContext(oldTC, newTC);
611
612 assert(newTC->contextId() == oldTC->contextId());
613 assert(newTC->threadId() == oldTC->threadId());
614 system->replaceThreadContext(newTC, newTC->contextId());
615
616 /* This code no longer works since the zero register (e.g.,
617 * r31 on Alpha) doesn't necessarily contain zero at this
618 * point.
619 if (DTRACE(Context))
620 ThreadContext::compare(oldTC, newTC);
621 */
622
623 BaseMasterPort *old_itb_port = oldTC->getITBPtr()->getMasterPort();
624 BaseMasterPort *old_dtb_port = oldTC->getDTBPtr()->getMasterPort();
625 BaseMasterPort *new_itb_port = newTC->getITBPtr()->getMasterPort();
626 BaseMasterPort *new_dtb_port = newTC->getDTBPtr()->getMasterPort();
627
628 // Move over any table walker ports if they exist
629 if (new_itb_port) {
630 assert(!new_itb_port->isConnected());
631 assert(old_itb_port);
632 assert(old_itb_port->isConnected());
633 BaseSlavePort &slavePort = old_itb_port->getSlavePort();
634 old_itb_port->unbind();
635 new_itb_port->bind(slavePort);
636 }
637 if (new_dtb_port) {
638 assert(!new_dtb_port->isConnected());
639 assert(old_dtb_port);
640 assert(old_dtb_port->isConnected());
641 BaseSlavePort &slavePort = old_dtb_port->getSlavePort();
642 old_dtb_port->unbind();
643 new_dtb_port->bind(slavePort);
644 }
645 newTC->getITBPtr()->takeOverFrom(oldTC->getITBPtr());
646 newTC->getDTBPtr()->takeOverFrom(oldTC->getDTBPtr());
647
648 // Checker whether or not we have to transfer CheckerCPU
649 // objects over in the switch
650 CheckerCPU *oldChecker = oldTC->getCheckerCpuPtr();
651 CheckerCPU *newChecker = newTC->getCheckerCpuPtr();
652 if (oldChecker && newChecker) {
653 BaseMasterPort *old_checker_itb_port =
654 oldChecker->getITBPtr()->getMasterPort();
655 BaseMasterPort *old_checker_dtb_port =
656 oldChecker->getDTBPtr()->getMasterPort();
657 BaseMasterPort *new_checker_itb_port =
658 newChecker->getITBPtr()->getMasterPort();
659 BaseMasterPort *new_checker_dtb_port =
660 newChecker->getDTBPtr()->getMasterPort();
661
662 newChecker->getITBPtr()->takeOverFrom(oldChecker->getITBPtr());
663 newChecker->getDTBPtr()->takeOverFrom(oldChecker->getDTBPtr());
664
665 // Move over any table walker ports if they exist for checker
666 if (new_checker_itb_port) {
667 assert(!new_checker_itb_port->isConnected());
668 assert(old_checker_itb_port);
669 assert(old_checker_itb_port->isConnected());
670 BaseSlavePort &slavePort =
671 old_checker_itb_port->getSlavePort();
672 old_checker_itb_port->unbind();
673 new_checker_itb_port->bind(slavePort);
674 }
675 if (new_checker_dtb_port) {
676 assert(!new_checker_dtb_port->isConnected());
677 assert(old_checker_dtb_port);
678 assert(old_checker_dtb_port->isConnected());
679 BaseSlavePort &slavePort =
680 old_checker_dtb_port->getSlavePort();
681 old_checker_dtb_port->unbind();
682 new_checker_dtb_port->bind(slavePort);
683 }
684 }
685 }
686
687 interrupts = oldCPU->interrupts;
688 for (ThreadID tid = 0; tid < numThreads; tid++) {
689 interrupts[tid]->setCPU(this);
690 }
691 oldCPU->interrupts.clear();
692
693 if (FullSystem) {
694 for (ThreadID i = 0; i < size; ++i)
695 threadContexts[i]->profileClear();
696
697 if (profileEvent)
698 schedule(profileEvent, curTick());
699 }
700
701 // All CPUs have an instruction and a data port, and the new CPU's
702 // ports are dangling while the old CPU has its ports connected
703 // already. Unbind the old CPU and then bind the ports of the one
704 // we are switching to.
705 assert(!getInstPort().isConnected());
706 assert(oldCPU->getInstPort().isConnected());
707 BaseSlavePort &inst_peer_port = oldCPU->getInstPort().getSlavePort();
708 oldCPU->getInstPort().unbind();
709 getInstPort().bind(inst_peer_port);
710
711 assert(!getDataPort().isConnected());
712 assert(oldCPU->getDataPort().isConnected());
713 BaseSlavePort &data_peer_port = oldCPU->getDataPort().getSlavePort();
714 oldCPU->getDataPort().unbind();
715 getDataPort().bind(data_peer_port);
716}
717
718void
719BaseCPU::flushTLBs()
720{
721 for (ThreadID i = 0; i < threadContexts.size(); ++i) {
722 ThreadContext &tc(*threadContexts[i]);
723 CheckerCPU *checker(tc.getCheckerCpuPtr());
724
725 tc.getITBPtr()->flushAll();
726 tc.getDTBPtr()->flushAll();
727 if (checker) {
728 checker->getITBPtr()->flushAll();
729 checker->getDTBPtr()->flushAll();
730 }
731 }
732}
733
734void
735BaseCPU::processProfileEvent()
736{
737 ThreadID size = threadContexts.size();
738
739 for (ThreadID i = 0; i < size; ++i)
740 threadContexts[i]->profileSample();
741
742 schedule(profileEvent, curTick() + params()->profile);
743}
744
745void
746BaseCPU::serialize(CheckpointOut &cp) const
747{
748 SERIALIZE_SCALAR(instCnt);
749
750 if (!_switchedOut) {
751 /* Unlike _pid, _taskId is not serialized, as they are dynamically
752 * assigned unique ids that are only meaningful for the duration of
753 * a specific run. We will need to serialize the entire taskMap in
754 * system. */
755 SERIALIZE_SCALAR(_pid);
756
757 // Serialize the threads, this is done by the CPU implementation.
758 for (ThreadID i = 0; i < numThreads; ++i) {
759 ScopedCheckpointSection sec(cp, csprintf("xc.%i", i));
760 interrupts[i]->serialize(cp);
761 serializeThread(cp, i);
762 }
763 }
764}
765
766void
767BaseCPU::unserialize(CheckpointIn &cp)
768{
769 UNSERIALIZE_SCALAR(instCnt);
770
771 if (!_switchedOut) {
772 UNSERIALIZE_SCALAR(_pid);
773
774 // Unserialize the threads, this is done by the CPU implementation.
775 for (ThreadID i = 0; i < numThreads; ++i) {
776 ScopedCheckpointSection sec(cp, csprintf("xc.%i", i));
777 interrupts[i]->unserialize(cp);
778 unserializeThread(cp, i);
779 }
780 }
781}
782
783void
784BaseCPU::scheduleInstStop(ThreadID tid, Counter insts, const char *cause)
785{
786 const Tick now(comInstEventQueue[tid]->getCurTick());
787 Event *event(new LocalSimLoopExitEvent(cause, 0));
788
789 comInstEventQueue[tid]->schedule(event, now + insts);
790}
791
792uint64_t
793BaseCPU::getCurrentInstCount(ThreadID tid)
794{
795 return Tick(comInstEventQueue[tid]->getCurTick());
796}
797
798AddressMonitor::AddressMonitor() {
799 armed = false;
800 waiting = false;
801 gotWakeup = false;
802}
803
804bool AddressMonitor::doMonitor(PacketPtr pkt) {
805 assert(pkt->req->hasPaddr());
806 if (armed && waiting) {
807 if (pAddr == pkt->getAddr()) {
808 DPRINTF(Mwait,"pAddr=0x%lx invalidated: waking up core\n",
809 pkt->getAddr());
810 waiting = false;
811 return true;
812 }
813 }
814 return false;
815}
816
817void
818BaseCPU::scheduleLoadStop(ThreadID tid, Counter loads, const char *cause)
819{
820 const Tick now(comLoadEventQueue[tid]->getCurTick());
821 Event *event(new LocalSimLoopExitEvent(cause, 0));
822
823 comLoadEventQueue[tid]->schedule(event, now + loads);
824}
825
826
827void
828BaseCPU::traceFunctionsInternal(Addr pc)
829{
830 if (!debugSymbolTable)
831 return;
832
833 // if pc enters different function, print new function symbol and
834 // update saved range. Otherwise do nothing.
835 if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
836 string sym_str;
837 bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
838 currentFunctionStart,
839 currentFunctionEnd);
840
841 if (!found) {
842 // no symbol found: use addr as label
843 sym_str = csprintf("0x%x", pc);
844 currentFunctionStart = pc;
845 currentFunctionEnd = pc + 1;
846 }
847
848 ccprintf(*functionTraceStream, " (%d)\n%d: %s",
849 curTick() - functionEntryTick, curTick(), sym_str);
850 functionEntryTick = curTick();
851 }
852}
853
854bool
855BaseCPU::waitForRemoteGDB() const
856{
857 return params()->wait_for_remote_gdb;
858}