base.cc revision 7897 
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 <string>
36#include <sstream>
37
38#include "arch/tlb.hh"
39#include "base/cprintf.hh"
40#include "base/loader/symtab.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/thread_context.hh"
47#include "cpu/profile.hh"
48#include "params/BaseCPU.hh"
49#include "sim/sim_exit.hh"
50#include "sim/process.hh"
51#include "sim/sim_events.hh"
52#include "sim/system.hh"
53
54// Hack
55#include "sim/stat_control.hh"
56
57using namespace std;
58
59vector<BaseCPU *> BaseCPU::cpuList;
60
61// This variable reflects the max number of threads in any CPU.  Be
62// careful to only use it once all the CPUs that you care about have
63// been initialized
64int maxThreadsPerCPU = 1;
65
66CPUProgressEvent::CPUProgressEvent(BaseCPU *_cpu, Tick ival)
67    : Event(Event::Progress_Event_Pri), _interval(ival), lastNumInst(0),
68      cpu(_cpu), _repeatEvent(true)
69{
70    if (_interval)
71        cpu->schedule(this, curTick() + _interval);
72}
73
74void
75CPUProgressEvent::process()
76{
77    Counter temp = cpu->totalInstructions();
78#ifndef NDEBUG
79    double ipc = double(temp - lastNumInst) / (_interval / cpu->ticks(1));
80
81    DPRINTFN("%s progress event, total committed:%i, progress insts committed: "
82             "%lli, IPC: %0.8d\n", cpu->name(), temp, temp - lastNumInst,
83             ipc);
84    ipc = 0.0;
85#else
86    cprintf("%lli: %s progress event, total committed:%i, progress insts "
87            "committed: %lli\n", curTick(), cpu->name(), temp,
88            temp - lastNumInst);
89#endif
90    lastNumInst = temp;
91
92    if (_repeatEvent)
93        cpu->schedule(this, curTick() + _interval);
94}
95
96const char *
97CPUProgressEvent::description() const
98{
99    return "CPU Progress";
100}
101
102#if FULL_SYSTEM
103BaseCPU::BaseCPU(Params *p)
104    : MemObject(p), clock(p->clock), instCnt(0), _cpuId(p->cpu_id),
105      interrupts(p->interrupts),
106      numThreads(p->numThreads), system(p->system),
107      phase(p->phase)
108#else
109BaseCPU::BaseCPU(Params *p)
110    : MemObject(p), clock(p->clock), _cpuId(p->cpu_id),
111      numThreads(p->numThreads), system(p->system),
112      phase(p->phase)
113#endif
114{
115//    currentTick = curTick();
116
117    // if Python did not provide a valid ID, do it here
118    if (_cpuId == -1 ) {
119        _cpuId = cpuList.size();
120    }
121
122    // add self to global list of CPUs
123    cpuList.push_back(this);
124
125    DPRINTF(SyscallVerbose, "Constructing CPU with id %d\n", _cpuId);
126
127    if (numThreads > maxThreadsPerCPU)
128        maxThreadsPerCPU = numThreads;
129
130    // allocate per-thread instruction-based event queues
131    comInstEventQueue = new EventQueue *[numThreads];
132    for (ThreadID tid = 0; tid < numThreads; ++tid)
133        comInstEventQueue[tid] =
134            new EventQueue("instruction-based event queue");
135
136    //
137    // set up instruction-count-based termination events, if any
138    //
139    if (p->max_insts_any_thread != 0) {
140        const char *cause = "a thread reached the max instruction count";
141        for (ThreadID tid = 0; tid < numThreads; ++tid) {
142            Event *event = new SimLoopExitEvent(cause, 0);
143            comInstEventQueue[tid]->schedule(event, p->max_insts_any_thread);
144        }
145    }
146
147    if (p->max_insts_all_threads != 0) {
148        const char *cause = "all threads reached the max instruction count";
149
150        // allocate & initialize shared downcounter: each event will
151        // decrement this when triggered; simulation will terminate
152        // when counter reaches 0
153        int *counter = new int;
154        *counter = numThreads;
155        for (ThreadID tid = 0; tid < numThreads; ++tid) {
156            Event *event = new CountedExitEvent(cause, *counter);
157            comInstEventQueue[tid]->schedule(event, p->max_insts_all_threads);
158        }
159    }
160
161    // allocate per-thread load-based event queues
162    comLoadEventQueue = new EventQueue *[numThreads];
163    for (ThreadID tid = 0; tid < numThreads; ++tid)
164        comLoadEventQueue[tid] = new EventQueue("load-based event queue");
165
166    //
167    // set up instruction-count-based termination events, if any
168    //
169    if (p->max_loads_any_thread != 0) {
170        const char *cause = "a thread reached the max load count";
171        for (ThreadID tid = 0; tid < numThreads; ++tid) {
172            Event *event = new SimLoopExitEvent(cause, 0);
173            comLoadEventQueue[tid]->schedule(event, p->max_loads_any_thread);
174        }
175    }
176
177    if (p->max_loads_all_threads != 0) {
178        const char *cause = "all threads reached the max load count";
179        // allocate & initialize shared downcounter: each event will
180        // decrement this when triggered; simulation will terminate
181        // when counter reaches 0
182        int *counter = new int;
183        *counter = numThreads;
184        for (ThreadID tid = 0; tid < numThreads; ++tid) {
185            Event *event = new CountedExitEvent(cause, *counter);
186            comLoadEventQueue[tid]->schedule(event, p->max_loads_all_threads);
187        }
188    }
189
190    functionTracingEnabled = false;
191    if (p->function_trace) {
192        functionTraceStream = simout.find(csprintf("ftrace.%s", name()));
193        currentFunctionStart = currentFunctionEnd = 0;
194        functionEntryTick = p->function_trace_start;
195
196        if (p->function_trace_start == 0) {
197            functionTracingEnabled = true;
198        } else {
199            typedef EventWrapper<BaseCPU, &BaseCPU::enableFunctionTrace> wrap;
200            Event *event = new wrap(this, true);
201            schedule(event, p->function_trace_start);
202        }
203    }
204#if FULL_SYSTEM
205    interrupts->setCPU(this);
206
207    profileEvent = NULL;
208    if (params()->profile)
209        profileEvent = new ProfileEvent(this, params()->profile);
210#endif
211    tracer = params()->tracer;
212}
213
214void
215BaseCPU::enableFunctionTrace()
216{
217    functionTracingEnabled = true;
218}
219
220BaseCPU::~BaseCPU()
221{
222}
223
224void
225BaseCPU::init()
226{
227    if (!params()->defer_registration)
228        registerThreadContexts();
229}
230
231void
232BaseCPU::startup()
233{
234#if FULL_SYSTEM
235    if (!params()->defer_registration && profileEvent)
236        schedule(profileEvent, curTick());
237#endif
238
239    if (params()->progress_interval) {
240        Tick num_ticks = ticks(params()->progress_interval);
241
242        Event *event;
243        event = new CPUProgressEvent(this, num_ticks);
244    }
245}
246
247
248void
249BaseCPU::regStats()
250{
251    using namespace Stats;
252
253    numCycles
254        .name(name() + ".numCycles")
255        .desc("number of cpu cycles simulated")
256        ;
257
258    int size = threadContexts.size();
259    if (size > 1) {
260        for (int i = 0; i < size; ++i) {
261            stringstream namestr;
262            ccprintf(namestr, "%s.ctx%d", name(), i);
263            threadContexts[i]->regStats(namestr.str());
264        }
265    } else if (size == 1)
266        threadContexts[0]->regStats(name());
267
268#if FULL_SYSTEM
269#endif
270}
271
272Tick
273BaseCPU::nextCycle()
274{
275    Tick next_tick = curTick() - phase + clock - 1;
276    next_tick -= (next_tick % clock);
277    next_tick += phase;
278    return next_tick;
279}
280
281Tick
282BaseCPU::nextCycle(Tick begin_tick)
283{
284    Tick next_tick = begin_tick;
285    if (next_tick % clock != 0)
286        next_tick = next_tick - (next_tick % clock) + clock;
287    next_tick += phase;
288
289    assert(next_tick >= curTick());
290    return next_tick;
291}
292
293void
294BaseCPU::registerThreadContexts()
295{
296    ThreadID size = threadContexts.size();
297    for (ThreadID tid = 0; tid < size; ++tid) {
298        ThreadContext *tc = threadContexts[tid];
299
300        /** This is so that contextId and cpuId match where there is a
301         * 1cpu:1context relationship.  Otherwise, the order of registration
302         * could affect the assignment and cpu 1 could have context id 3, for
303         * example.  We may even want to do something like this for SMT so that
304         * cpu 0 has the lowest thread contexts and cpu N has the highest, but
305         * I'll just do this for now
306         */
307        if (numThreads == 1)
308            tc->setContextId(system->registerThreadContext(tc, _cpuId));
309        else
310            tc->setContextId(system->registerThreadContext(tc));
311#if !FULL_SYSTEM
312        tc->getProcessPtr()->assignThreadContext(tc->contextId());
313#endif
314    }
315}
316
317
318int
319BaseCPU::findContext(ThreadContext *tc)
320{
321    ThreadID size = threadContexts.size();
322    for (ThreadID tid = 0; tid < size; ++tid) {
323        if (tc == threadContexts[tid])
324            return tid;
325    }
326    return 0;
327}
328
329void
330BaseCPU::switchOut()
331{
332//    panic("This CPU doesn't support sampling!");
333#if FULL_SYSTEM
334    if (profileEvent && profileEvent->scheduled())
335        deschedule(profileEvent);
336#endif
337}
338
339void
340BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc)
341{
342    assert(threadContexts.size() == oldCPU->threadContexts.size());
343
344    _cpuId = oldCPU->cpuId();
345
346    ThreadID size = threadContexts.size();
347    for (ThreadID i = 0; i < size; ++i) {
348        ThreadContext *newTC = threadContexts[i];
349        ThreadContext *oldTC = oldCPU->threadContexts[i];
350
351        newTC->takeOverFrom(oldTC);
352
353        CpuEvent::replaceThreadContext(oldTC, newTC);
354
355        assert(newTC->contextId() == oldTC->contextId());
356        assert(newTC->threadId() == oldTC->threadId());
357        system->replaceThreadContext(newTC, newTC->contextId());
358
359        /* This code no longer works since the zero register (e.g.,
360         * r31 on Alpha) doesn't necessarily contain zero at this
361         * point.
362           if (DTRACE(Context))
363            ThreadContext::compare(oldTC, newTC);
364        */
365
366        Port  *old_itb_port, *old_dtb_port, *new_itb_port, *new_dtb_port;
367        old_itb_port = oldTC->getITBPtr()->getPort();
368        old_dtb_port = oldTC->getDTBPtr()->getPort();
369        new_itb_port = newTC->getITBPtr()->getPort();
370        new_dtb_port = newTC->getDTBPtr()->getPort();
371
372        // Move over any table walker ports if they exist
373        if (new_itb_port && !new_itb_port->isConnected()) {
374            assert(old_itb_port);
375            Port *peer = old_itb_port->getPeer();;
376            new_itb_port->setPeer(peer);
377            peer->setPeer(new_itb_port);
378        }
379        if (new_dtb_port && !new_dtb_port->isConnected()) {
380            assert(old_dtb_port);
381            Port *peer = old_dtb_port->getPeer();;
382            new_dtb_port->setPeer(peer);
383            peer->setPeer(new_dtb_port);
384        }
385    }
386
387#if FULL_SYSTEM
388    interrupts = oldCPU->interrupts;
389    interrupts->setCPU(this);
390
391    for (ThreadID i = 0; i < size; ++i)
392        threadContexts[i]->profileClear();
393
394    if (profileEvent)
395        schedule(profileEvent, curTick());
396#endif
397
398    // Connect new CPU to old CPU's memory only if new CPU isn't
399    // connected to anything.  Also connect old CPU's memory to new
400    // CPU.
401    if (!ic->isConnected()) {
402        Port *peer = oldCPU->getPort("icache_port")->getPeer();
403        ic->setPeer(peer);
404        peer->setPeer(ic);
405    }
406
407    if (!dc->isConnected()) {
408        Port *peer = oldCPU->getPort("dcache_port")->getPeer();
409        dc->setPeer(peer);
410        peer->setPeer(dc);
411    }
412}
413
414
415#if FULL_SYSTEM
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
446#endif // FULL_SYSTEM
447
448void
449BaseCPU::traceFunctionsInternal(Addr pc)
450{
451    if (!debugSymbolTable)
452        return;
453
454    // if pc enters different function, print new function symbol and
455    // update saved range.  Otherwise do nothing.
456    if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
457        string sym_str;
458        bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
459                                                         currentFunctionStart,
460                                                         currentFunctionEnd);
461
462        if (!found) {
463            // no symbol found: use addr as label
464            sym_str = csprintf("0x%x", pc);
465            currentFunctionStart = pc;
466            currentFunctionEnd = pc + 1;
467        }
468
469        ccprintf(*functionTraceStream, " (%d)\n%d: %s",
470                 curTick() - functionEntryTick, curTick(), sym_str);
471        functionEntryTick = curTick();
472    }
473}
474