base.cc revision 5807 
1/*
2 * Copyright (c) 2002-2005 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: Steve Reinhardt
29 *          Nathan Binkert
30 */
31
32#include <iostream>
33#include <string>
34#include <sstream>
35
36#include "base/cprintf.hh"
37#include "base/loader/symtab.hh"
38#include "base/misc.hh"
39#include "base/output.hh"
40#include "base/trace.hh"
41#include "cpu/base.hh"
42#include "cpu/cpuevent.hh"
43#include "cpu/thread_context.hh"
44#include "cpu/profile.hh"
45#include "params/BaseCPU.hh"
46#include "sim/sim_exit.hh"
47#include "sim/process.hh"
48#include "sim/sim_events.hh"
49#include "sim/system.hh"
50
51// Hack
52#include "sim/stat_control.hh"
53
54using namespace std;
55
56vector<BaseCPU *> BaseCPU::cpuList;
57
58// This variable reflects the max number of threads in any CPU.  Be
59// careful to only use it once all the CPUs that you care about have
60// been initialized
61int maxThreadsPerCPU = 1;
62
63CPUProgressEvent::CPUProgressEvent(BaseCPU *_cpu, Tick ival)
64    : Event(Event::Progress_Event_Pri), interval(ival), lastNumInst(0),
65      cpu(_cpu)
66{
67    if (interval)
68        cpu->schedule(this, curTick + interval);
69}
70
71void
72CPUProgressEvent::process()
73{
74    Counter temp = cpu->totalInstructions();
75#ifndef NDEBUG
76    double ipc = double(temp - lastNumInst) / (interval / cpu->ticks(1));
77
78    DPRINTFN("%s progress event, instructions committed: %lli, IPC: %0.8d\n",
79             cpu->name(), temp - lastNumInst, ipc);
80    ipc = 0.0;
81#else
82    cprintf("%lli: %s progress event, instructions committed: %lli\n",
83            curTick, cpu->name(), temp - lastNumInst);
84#endif
85    lastNumInst = temp;
86    cpu->schedule(this, curTick + interval);
87}
88
89const char *
90CPUProgressEvent::description() const
91{
92    return "CPU Progress";
93}
94
95#if FULL_SYSTEM
96BaseCPU::BaseCPU(Params *p)
97    : MemObject(p), clock(p->clock), instCnt(0), _cpuId(p->cpu_id),
98      interrupts(p->interrupts),
99      number_of_threads(p->numThreads), system(p->system),
100      phase(p->phase)
101#else
102BaseCPU::BaseCPU(Params *p)
103    : MemObject(p), clock(p->clock), _cpuId(p->cpu_id),
104      number_of_threads(p->numThreads), system(p->system),
105      phase(p->phase)
106#endif
107{
108//    currentTick = curTick;
109
110    // if Python did not provide a valid ID, do it here
111    if (_cpuId == -1 ) {
112        _cpuId = cpuList.size();
113    }
114
115    // add self to global list of CPUs
116    cpuList.push_back(this);
117
118    DPRINTF(SyscallVerbose, "Constructing CPU with id %d\n", _cpuId);
119
120    if (number_of_threads > maxThreadsPerCPU)
121        maxThreadsPerCPU = number_of_threads;
122
123    // allocate per-thread instruction-based event queues
124    comInstEventQueue = new EventQueue *[number_of_threads];
125    for (int i = 0; i < number_of_threads; ++i)
126        comInstEventQueue[i] = new EventQueue("instruction-based event queue");
127
128    //
129    // set up instruction-count-based termination events, if any
130    //
131    if (p->max_insts_any_thread != 0) {
132        const char *cause = "a thread reached the max instruction count";
133        for (int i = 0; i < number_of_threads; ++i) {
134            Event *event = new SimLoopExitEvent(cause, 0);
135            comInstEventQueue[i]->schedule(event, p->max_insts_any_thread);
136        }
137    }
138
139    if (p->max_insts_all_threads != 0) {
140        const char *cause = "all threads reached the max instruction count";
141
142        // allocate & initialize shared downcounter: each event will
143        // decrement this when triggered; simulation will terminate
144        // when counter reaches 0
145        int *counter = new int;
146        *counter = number_of_threads;
147        for (int i = 0; i < number_of_threads; ++i) {
148            Event *event = new CountedExitEvent(cause, *counter);
149            comInstEventQueue[i]->schedule(event, p->max_insts_any_thread);
150        }
151    }
152
153    // allocate per-thread load-based event queues
154    comLoadEventQueue = new EventQueue *[number_of_threads];
155    for (int i = 0; i < number_of_threads; ++i)
156        comLoadEventQueue[i] = new EventQueue("load-based event queue");
157
158    //
159    // set up instruction-count-based termination events, if any
160    //
161    if (p->max_loads_any_thread != 0) {
162        const char *cause = "a thread reached the max load count";
163        for (int i = 0; i < number_of_threads; ++i) {
164            Event *event = new SimLoopExitEvent(cause, 0);
165            comLoadEventQueue[i]->schedule(event, p->max_loads_any_thread);
166        }
167    }
168
169    if (p->max_loads_all_threads != 0) {
170        const char *cause = "all threads reached the max load count";
171        // allocate & initialize shared downcounter: each event will
172        // decrement this when triggered; simulation will terminate
173        // when counter reaches 0
174        int *counter = new int;
175        *counter = number_of_threads;
176        for (int i = 0; i < number_of_threads; ++i) {
177            Event *event = new CountedExitEvent(cause, *counter);
178            comLoadEventQueue[i]->schedule(event, p->max_loads_all_threads);
179        }
180    }
181
182    functionTracingEnabled = false;
183    if (p->function_trace) {
184        functionTraceStream = simout.find(csprintf("ftrace.%s", name()));
185        currentFunctionStart = currentFunctionEnd = 0;
186        functionEntryTick = p->function_trace_start;
187
188        if (p->function_trace_start == 0) {
189            functionTracingEnabled = true;
190        } else {
191            typedef EventWrapper<BaseCPU, &BaseCPU::enableFunctionTrace> wrap;
192            Event *event = new wrap(this, true);
193            schedule(event, p->function_trace_start);
194        }
195    }
196#if FULL_SYSTEM
197    profileEvent = NULL;
198    if (params()->profile)
199        profileEvent = new ProfileEvent(this, params()->profile);
200#endif
201    tracer = params()->tracer;
202}
203
204void
205BaseCPU::enableFunctionTrace()
206{
207    functionTracingEnabled = true;
208}
209
210BaseCPU::~BaseCPU()
211{
212}
213
214void
215BaseCPU::init()
216{
217    if (!params()->defer_registration)
218        registerThreadContexts();
219}
220
221void
222BaseCPU::startup()
223{
224#if FULL_SYSTEM
225    if (!params()->defer_registration && profileEvent)
226        schedule(profileEvent, curTick);
227#endif
228
229    if (params()->progress_interval) {
230        Tick num_ticks = ticks(params()->progress_interval);
231        Event *event = new CPUProgressEvent(this, num_ticks);
232        schedule(event, curTick + num_ticks);
233    }
234}
235
236
237void
238BaseCPU::regStats()
239{
240    using namespace Stats;
241
242    numCycles
243        .name(name() + ".numCycles")
244        .desc("number of cpu cycles simulated")
245        ;
246
247    int size = threadContexts.size();
248    if (size > 1) {
249        for (int i = 0; i < size; ++i) {
250            stringstream namestr;
251            ccprintf(namestr, "%s.ctx%d", name(), i);
252            threadContexts[i]->regStats(namestr.str());
253        }
254    } else if (size == 1)
255        threadContexts[0]->regStats(name());
256
257#if FULL_SYSTEM
258#endif
259}
260
261Tick
262BaseCPU::nextCycle()
263{
264    Tick next_tick = curTick - phase + clock - 1;
265    next_tick -= (next_tick % clock);
266    next_tick += phase;
267    return next_tick;
268}
269
270Tick
271BaseCPU::nextCycle(Tick begin_tick)
272{
273    Tick next_tick = begin_tick;
274    if (next_tick % clock != 0)
275        next_tick = next_tick - (next_tick % clock) + clock;
276    next_tick += phase;
277
278    assert(next_tick >= curTick);
279    return next_tick;
280}
281
282void
283BaseCPU::registerThreadContexts()
284{
285    for (int i = 0; i < threadContexts.size(); ++i) {
286        ThreadContext *tc = threadContexts[i];
287
288        /** This is so that contextId and cpuId match where there is a
289         * 1cpu:1context relationship.  Otherwise, the order of registration
290         * could affect the assignment and cpu 1 could have context id 3, for
291         * example.  We may even want to do something like this for SMT so that
292         * cpu 0 has the lowest thread contexts and cpu N has the highest, but
293         * I'll just do this for now
294         */
295        if (number_of_threads == 1)
296            tc->setContextId(system->registerThreadContext(tc, _cpuId));
297        else
298            tc->setContextId(system->registerThreadContext(tc));
299#if !FULL_SYSTEM
300        tc->getProcessPtr()->assignThreadContext(tc->contextId());
301#endif
302    }
303}
304
305
306int
307BaseCPU::findContext(ThreadContext *tc)
308{
309    for (int i = 0; i < threadContexts.size(); ++i) {
310        if (tc == threadContexts[i])
311            return i;
312    }
313    return 0;
314}
315
316void
317BaseCPU::switchOut()
318{
319//    panic("This CPU doesn't support sampling!");
320#if FULL_SYSTEM
321    if (profileEvent && profileEvent->scheduled())
322        deschedule(profileEvent);
323#endif
324}
325
326void
327BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc)
328{
329    assert(threadContexts.size() == oldCPU->threadContexts.size());
330
331    _cpuId = oldCPU->cpuId();
332
333    for (int i = 0; i < threadContexts.size(); ++i) {
334        ThreadContext *newTC = threadContexts[i];
335        ThreadContext *oldTC = oldCPU->threadContexts[i];
336
337        newTC->takeOverFrom(oldTC);
338
339        CpuEvent::replaceThreadContext(oldTC, newTC);
340
341        assert(newTC->contextId() == oldTC->contextId());
342        assert(newTC->threadId() == oldTC->threadId());
343        system->replaceThreadContext(newTC, newTC->contextId());
344
345        if (DTRACE(Context))
346            ThreadContext::compare(oldTC, newTC);
347    }
348
349#if FULL_SYSTEM
350    interrupts = oldCPU->interrupts;
351
352    for (int i = 0; i < threadContexts.size(); ++i)
353        threadContexts[i]->profileClear();
354
355    if (profileEvent)
356        schedule(profileEvent, curTick);
357#endif
358
359    // Connect new CPU to old CPU's memory only if new CPU isn't
360    // connected to anything.  Also connect old CPU's memory to new
361    // CPU.
362    if (!ic->isConnected()) {
363        Port *peer = oldCPU->getPort("icache_port")->getPeer();
364        ic->setPeer(peer);
365        peer->setPeer(ic);
366    }
367
368    if (!dc->isConnected()) {
369        Port *peer = oldCPU->getPort("dcache_port")->getPeer();
370        dc->setPeer(peer);
371        peer->setPeer(dc);
372    }
373}
374
375
376#if FULL_SYSTEM
377BaseCPU::ProfileEvent::ProfileEvent(BaseCPU *_cpu, Tick _interval)
378    : cpu(_cpu), interval(_interval)
379{ }
380
381void
382BaseCPU::ProfileEvent::process()
383{
384    for (int i = 0, size = cpu->threadContexts.size(); i < size; ++i) {
385        ThreadContext *tc = cpu->threadContexts[i];
386        tc->profileSample();
387    }
388
389    cpu->schedule(this, curTick + interval);
390}
391
392void
393BaseCPU::serialize(std::ostream &os)
394{
395    SERIALIZE_SCALAR(instCnt);
396    interrupts->serialize(os);
397}
398
399void
400BaseCPU::unserialize(Checkpoint *cp, const std::string &section)
401{
402    UNSERIALIZE_SCALAR(instCnt);
403    interrupts->unserialize(cp, section);
404}
405
406#endif // FULL_SYSTEM
407
408void
409BaseCPU::traceFunctionsInternal(Addr pc)
410{
411    if (!debugSymbolTable)
412        return;
413
414    // if pc enters different function, print new function symbol and
415    // update saved range.  Otherwise do nothing.
416    if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
417        string sym_str;
418        bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
419                                                         currentFunctionStart,
420                                                         currentFunctionEnd);
421
422        if (!found) {
423            // no symbol found: use addr as label
424            sym_str = csprintf("0x%x", pc);
425            currentFunctionStart = pc;
426            currentFunctionEnd = pc + 1;
427        }
428
429        ccprintf(*functionTraceStream, " (%d)\n%d: %s",
430                 curTick - functionEntryTick, curTick, sym_str);
431        functionEntryTick = curTick;
432    }
433}
434