base.cc revision 8780 
1/*
2 * Copyright (c) 2002-2005 The Regents of The University of Michigan
3 * Copyright (c) 2011 Regents of the University of California
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are
8 * met: redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer;
10 * redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution;
13 * neither the name of the copyright holders nor the names of its
14 * contributors may be used to endorse or promote products derived from
15 * this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 *
29 * Authors: Steve Reinhardt
30 *          Nathan Binkert
31 *          Rick Strong
32 */
33
34#include <iostream>
35#include <sstream>
36#include <string>
37
38#include "arch/tlb.hh"
39#include "base/loader/symtab.hh"
40#include "base/cprintf.hh"
41#include "base/misc.hh"
42#include "base/output.hh"
43#include "base/trace.hh"
44#include "cpu/base.hh"
45#include "cpu/cpuevent.hh"
46#include "cpu/profile.hh"
47#include "cpu/thread_context.hh"
48#include "debug/SyscallVerbose.hh"
49#include "params/BaseCPU.hh"
50#include "sim/full_system.hh"
51#include "sim/process.hh"
52#include "sim/sim_events.hh"
53#include "sim/sim_exit.hh"
54#include "sim/system.hh"
55
56// Hack
57#include "sim/stat_control.hh"
58
59using namespace std;
60
61vector<BaseCPU *> BaseCPU::cpuList;
62
63// This variable reflects the max number of threads in any CPU.  Be
64// careful to only use it once all the CPUs that you care about have
65// been initialized
66int maxThreadsPerCPU = 1;
67
68CPUProgressEvent::CPUProgressEvent(BaseCPU *_cpu, Tick ival)
69    : Event(Event::Progress_Event_Pri), _interval(ival), lastNumInst(0),
70      cpu(_cpu), _repeatEvent(true)
71{
72    if (_interval)
73        cpu->schedule(this, curTick() + _interval);
74}
75
76void
77CPUProgressEvent::process()
78{
79    Counter temp = cpu->totalInstructions();
80#ifndef NDEBUG
81    double ipc = double(temp - lastNumInst) / (_interval / cpu->ticks(1));
82
83    DPRINTFN("%s progress event, total committed:%i, progress insts committed: "
84             "%lli, IPC: %0.8d\n", cpu->name(), temp, temp - lastNumInst,
85             ipc);
86    ipc = 0.0;
87#else
88    cprintf("%lli: %s progress event, total committed:%i, progress insts "
89            "committed: %lli\n", curTick(), cpu->name(), temp,
90            temp - lastNumInst);
91#endif
92    lastNumInst = temp;
93
94    if (_repeatEvent)
95        cpu->schedule(this, curTick() + _interval);
96}
97
98const char *
99CPUProgressEvent::description() const
100{
101    return "CPU Progress";
102}
103
104BaseCPU::BaseCPU(Params *p)
105    : MemObject(p), clock(p->clock), instCnt(0), _cpuId(p->cpu_id),
106      interrupts(p->interrupts),
107      numThreads(p->numThreads), system(p->system),
108      phase(p->phase)
109{
110//    currentTick = curTick();
111
112    // if Python did not provide a valid ID, do it here
113    if (_cpuId == -1 ) {
114        _cpuId = cpuList.size();
115    }
116
117    // add self to global list of CPUs
118    cpuList.push_back(this);
119
120    DPRINTF(SyscallVerbose, "Constructing CPU with id %d\n", _cpuId);
121
122    if (numThreads > maxThreadsPerCPU)
123        maxThreadsPerCPU = numThreads;
124
125    // allocate per-thread instruction-based event queues
126    comInstEventQueue = new EventQueue *[numThreads];
127    for (ThreadID tid = 0; tid < numThreads; ++tid)
128        comInstEventQueue[tid] =
129            new EventQueue("instruction-based event queue");
130
131    //
132    // set up instruction-count-based termination events, if any
133    //
134    if (p->max_insts_any_thread != 0) {
135        const char *cause = "a thread reached the max instruction count";
136        for (ThreadID tid = 0; tid < numThreads; ++tid) {
137            Event *event = new SimLoopExitEvent(cause, 0);
138            comInstEventQueue[tid]->schedule(event, p->max_insts_any_thread);
139        }
140    }
141
142    if (p->max_insts_all_threads != 0) {
143        const char *cause = "all threads reached the max instruction count";
144
145        // allocate & initialize shared downcounter: each event will
146        // decrement this when triggered; simulation will terminate
147        // when counter reaches 0
148        int *counter = new int;
149        *counter = numThreads;
150        for (ThreadID tid = 0; tid < numThreads; ++tid) {
151            Event *event = new CountedExitEvent(cause, *counter);
152            comInstEventQueue[tid]->schedule(event, p->max_insts_all_threads);
153        }
154    }
155
156    // allocate per-thread load-based event queues
157    comLoadEventQueue = new EventQueue *[numThreads];
158    for (ThreadID tid = 0; tid < numThreads; ++tid)
159        comLoadEventQueue[tid] = new EventQueue("load-based event queue");
160
161    //
162    // set up instruction-count-based termination events, if any
163    //
164    if (p->max_loads_any_thread != 0) {
165        const char *cause = "a thread reached the max load count";
166        for (ThreadID tid = 0; tid < numThreads; ++tid) {
167            Event *event = new SimLoopExitEvent(cause, 0);
168            comLoadEventQueue[tid]->schedule(event, p->max_loads_any_thread);
169        }
170    }
171
172    if (p->max_loads_all_threads != 0) {
173        const char *cause = "all threads reached the max load count";
174        // allocate & initialize shared downcounter: each event will
175        // decrement this when triggered; simulation will terminate
176        // when counter reaches 0
177        int *counter = new int;
178        *counter = numThreads;
179        for (ThreadID tid = 0; tid < numThreads; ++tid) {
180            Event *event = new CountedExitEvent(cause, *counter);
181            comLoadEventQueue[tid]->schedule(event, p->max_loads_all_threads);
182        }
183    }
184
185    functionTracingEnabled = false;
186    if (p->function_trace) {
187        functionTraceStream = simout.find(csprintf("ftrace.%s", name()));
188        currentFunctionStart = currentFunctionEnd = 0;
189        functionEntryTick = p->function_trace_start;
190
191        if (p->function_trace_start == 0) {
192            functionTracingEnabled = true;
193        } else {
194            typedef EventWrapper<BaseCPU, &BaseCPU::enableFunctionTrace> wrap;
195            Event *event = new wrap(this, true);
196            schedule(event, p->function_trace_start);
197        }
198    }
199    interrupts->setCPU(this);
200
201    if (FullSystem) {
202#if FULL_SYSTEM
203        profileEvent = NULL;
204        if (params()->profile)
205            profileEvent = new ProfileEvent(this, params()->profile);
206#endif
207    }
208    tracer = params()->tracer;
209}
210
211void
212BaseCPU::enableFunctionTrace()
213{
214    functionTracingEnabled = true;
215}
216
217BaseCPU::~BaseCPU()
218{
219}
220
221void
222BaseCPU::init()
223{
224    if (!params()->defer_registration)
225        registerThreadContexts();
226}
227
228void
229BaseCPU::startup()
230{
231    if (FullSystem) {
232        if (!params()->defer_registration && profileEvent)
233            schedule(profileEvent, curTick());
234    }
235
236    if (params()->progress_interval) {
237        Tick num_ticks = ticks(params()->progress_interval);
238
239        Event *event;
240        event = new CPUProgressEvent(this, num_ticks);
241    }
242}
243
244
245void
246BaseCPU::regStats()
247{
248    using namespace Stats;
249
250    numCycles
251        .name(name() + ".numCycles")
252        .desc("number of cpu cycles simulated")
253        ;
254
255    numWorkItemsStarted
256        .name(name() + ".numWorkItemsStarted")
257        .desc("number of work items this cpu started")
258        ;
259
260    numWorkItemsCompleted
261        .name(name() + ".numWorkItemsCompleted")
262        .desc("number of work items this cpu completed")
263        ;
264
265    int size = threadContexts.size();
266    if (size > 1) {
267        for (int i = 0; i < size; ++i) {
268            stringstream namestr;
269            ccprintf(namestr, "%s.ctx%d", name(), i);
270            threadContexts[i]->regStats(namestr.str());
271        }
272    } else if (size == 1)
273        threadContexts[0]->regStats(name());
274}
275
276Tick
277BaseCPU::nextCycle()
278{
279    Tick next_tick = curTick() - phase + clock - 1;
280    next_tick -= (next_tick % clock);
281    next_tick += phase;
282    return next_tick;
283}
284
285Tick
286BaseCPU::nextCycle(Tick begin_tick)
287{
288    Tick next_tick = begin_tick;
289    if (next_tick % clock != 0)
290        next_tick = next_tick - (next_tick % clock) + clock;
291    next_tick += phase;
292
293    assert(next_tick >= curTick());
294    return next_tick;
295}
296
297void
298BaseCPU::registerThreadContexts()
299{
300    ThreadID size = threadContexts.size();
301    for (ThreadID tid = 0; tid < size; ++tid) {
302        ThreadContext *tc = threadContexts[tid];
303
304        /** This is so that contextId and cpuId match where there is a
305         * 1cpu:1context relationship.  Otherwise, the order of registration
306         * could affect the assignment and cpu 1 could have context id 3, for
307         * example.  We may even want to do something like this for SMT so that
308         * cpu 0 has the lowest thread contexts and cpu N has the highest, but
309         * I'll just do this for now
310         */
311        if (numThreads == 1)
312            tc->setContextId(system->registerThreadContext(tc, _cpuId));
313        else
314            tc->setContextId(system->registerThreadContext(tc));
315
316        if (!FullSystem)
317            tc->getProcessPtr()->assignThreadContext(tc->contextId());
318    }
319}
320
321
322int
323BaseCPU::findContext(ThreadContext *tc)
324{
325    ThreadID size = threadContexts.size();
326    for (ThreadID tid = 0; tid < size; ++tid) {
327        if (tc == threadContexts[tid])
328            return tid;
329    }
330    return 0;
331}
332
333void
334BaseCPU::switchOut()
335{
336    if (profileEvent && profileEvent->scheduled())
337        deschedule(profileEvent);
338}
339
340void
341BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc)
342{
343    assert(threadContexts.size() == oldCPU->threadContexts.size());
344
345    _cpuId = oldCPU->cpuId();
346
347    ThreadID size = threadContexts.size();
348    for (ThreadID i = 0; i < size; ++i) {
349        ThreadContext *newTC = threadContexts[i];
350        ThreadContext *oldTC = oldCPU->threadContexts[i];
351
352        newTC->takeOverFrom(oldTC);
353
354        CpuEvent::replaceThreadContext(oldTC, newTC);
355
356        assert(newTC->contextId() == oldTC->contextId());
357        assert(newTC->threadId() == oldTC->threadId());
358        system->replaceThreadContext(newTC, newTC->contextId());
359
360        /* This code no longer works since the zero register (e.g.,
361         * r31 on Alpha) doesn't necessarily contain zero at this
362         * point.
363           if (DTRACE(Context))
364            ThreadContext::compare(oldTC, newTC);
365        */
366
367        Port  *old_itb_port, *old_dtb_port, *new_itb_port, *new_dtb_port;
368        old_itb_port = oldTC->getITBPtr()->getPort();
369        old_dtb_port = oldTC->getDTBPtr()->getPort();
370        new_itb_port = newTC->getITBPtr()->getPort();
371        new_dtb_port = newTC->getDTBPtr()->getPort();
372
373        // Move over any table walker ports if they exist
374        if (new_itb_port && !new_itb_port->isConnected()) {
375            assert(old_itb_port);
376            Port *peer = old_itb_port->getPeer();;
377            new_itb_port->setPeer(peer);
378            peer->setPeer(new_itb_port);
379        }
380        if (new_dtb_port && !new_dtb_port->isConnected()) {
381            assert(old_dtb_port);
382            Port *peer = old_dtb_port->getPeer();;
383            new_dtb_port->setPeer(peer);
384            peer->setPeer(new_dtb_port);
385        }
386    }
387
388    interrupts = oldCPU->interrupts;
389    interrupts->setCPU(this);
390
391    if (FullSystem) {
392        for (ThreadID i = 0; i < size; ++i)
393            threadContexts[i]->profileClear();
394
395        if (profileEvent)
396            schedule(profileEvent, curTick());
397    }
398
399    // Connect new CPU to old CPU's memory only if new CPU isn't
400    // connected to anything.  Also connect old CPU's memory to new
401    // CPU.
402    if (!ic->isConnected()) {
403        Port *peer = oldCPU->getPort("icache_port")->getPeer();
404        ic->setPeer(peer);
405        peer->setPeer(ic);
406    }
407
408    if (!dc->isConnected()) {
409        Port *peer = oldCPU->getPort("dcache_port")->getPeer();
410        dc->setPeer(peer);
411        peer->setPeer(dc);
412    }
413}
414
415
416BaseCPU::ProfileEvent::ProfileEvent(BaseCPU *_cpu, Tick _interval)
417    : cpu(_cpu), interval(_interval)
418{ }
419
420void
421BaseCPU::ProfileEvent::process()
422{
423    ThreadID size = cpu->threadContexts.size();
424    for (ThreadID i = 0; i < size; ++i) {
425        ThreadContext *tc = cpu->threadContexts[i];
426        tc->profileSample();
427    }
428
429    cpu->schedule(this, curTick() + interval);
430}
431
432void
433BaseCPU::serialize(std::ostream &os)
434{
435    SERIALIZE_SCALAR(instCnt);
436    interrupts->serialize(os);
437}
438
439void
440BaseCPU::unserialize(Checkpoint *cp, const std::string &section)
441{
442    UNSERIALIZE_SCALAR(instCnt);
443    interrupts->unserialize(cp, section);
444}
445
446void
447BaseCPU::traceFunctionsInternal(Addr pc)
448{
449    if (!debugSymbolTable)
450        return;
451
452    // if pc enters different function, print new function symbol and
453    // update saved range.  Otherwise do nothing.
454    if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
455        string sym_str;
456        bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
457                                                         currentFunctionStart,
458                                                         currentFunctionEnd);
459
460        if (!found) {
461            // no symbol found: use addr as label
462            sym_str = csprintf("0x%x", pc);
463            currentFunctionStart = pc;
464            currentFunctionEnd = pc + 1;
465        }
466
467        ccprintf(*functionTraceStream, " (%d)\n%d: %s",
468                 curTick() - functionEntryTick, curTick(), sym_str);
469        functionEntryTick = curTick();
470    }
471}
472