src/cpu/base.cc

   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/process.hh"
  51 #include "sim/sim_events.hh"
  52 #include "sim/sim_exit.hh"
  53 #include "sim/system.hh"
  54
  55 // Hack
  56 #include "sim/stat_control.hh"
  57
  58 using namespace std;
  59
  60 vector<BaseCPU *> BaseCPU::cpuList;
  61
  62 // This variable reflects the max number of threads in any CPU.  Be
  63 // careful to only use it once all the CPUs that you care about have
  64 // been initialized
  65 int maxThreadsPerCPU = 1;
  66
  67 CPUProgressEvent::CPUProgressEvent(BaseCPU *_cpu, Tick ival)
  68     : Event(Event::Progress_Event_Pri), _interval(ival), lastNumInst(0),
  69       cpu(_cpu), _repeatEvent(true)
  70 {
  71     if (_interval)
  72         cpu->schedule(this, curTick() + _interval);
  73 }
  74
  75 void
  76 CPUProgressEvent::process()
  77 {
  78     Counter temp = cpu->totalInstructions();
  79 #ifndef NDEBUG
  80     double ipc = double(temp - lastNumInst) / (_interval / cpu->ticks(1));
  81
  82     DPRINTFN("%s progress event, total committed:%i, progress insts committed: "
  83              "%lli, IPC: %0.8d\n", cpu->name(), temp, temp - lastNumInst,
  84              ipc);
  85     ipc = 0.0;
  86 #else
  87     cprintf("%lli: %s progress event, total committed:%i, progress insts "
  88             "committed: %lli\n", curTick(), cpu->name(), temp,
  89             temp - lastNumInst);
  90 #endif
  91     lastNumInst = temp;
  92
  93     if (_repeatEvent)
  94         cpu->schedule(this, curTick() + _interval);
  95 }
  96
  97 const char *
  98 CPUProgressEvent::description() const
  99 {
 100     return "CPU Progress";
 101 }
 102
 103 #if FULL_SYSTEM
 104 BaseCPU::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 #else
 110 BaseCPU::BaseCPU(Params *p)
 111     : MemObject(p), clock(p->clock), _cpuId(p->cpu_id),
 112       numThreads(p->numThreads), system(p->system),
 113       phase(p->phase)
 114 #endif
 115 {
 116 //    currentTick = curTick();
 117
 118     // if Python did not provide a valid ID, do it here
 119     if (_cpuId == -1 ) {
 120         _cpuId = cpuList.size();
 121     }
 122
 123     // add self to global list of CPUs
 124     cpuList.push_back(this);
 125
 126     DPRINTF(SyscallVerbose, "Constructing CPU with id %d\n", _cpuId);
 127
 128     if (numThreads > maxThreadsPerCPU)
 129         maxThreadsPerCPU = numThreads;
 130
 131     // allocate per-thread instruction-based event queues
 132     comInstEventQueue = new EventQueue *[numThreads];
 133     for (ThreadID tid = 0; tid < numThreads; ++tid)
 134         comInstEventQueue[tid] =
 135             new EventQueue("instruction-based event queue");
 136
 137     //
 138     // set up instruction-count-based termination events, if any
 139     //
 140     if (p->max_insts_any_thread != 0) {
 141         const char *cause = "a thread reached the max instruction count";
 142         for (ThreadID tid = 0; tid < numThreads; ++tid) {
 143             Event *event = new SimLoopExitEvent(cause, 0);
 144             comInstEventQueue[tid]->schedule(event, p->max_insts_any_thread);
 145         }
 146     }
 147
 148     if (p->max_insts_all_threads != 0) {
 149         const char *cause = "all threads reached the max instruction count";
 150
 151         // allocate & initialize shared downcounter: each event will
 152         // decrement this when triggered; simulation will terminate
 153         // when counter reaches 0
 154         int *counter = new int;
 155         *counter = numThreads;
 156         for (ThreadID tid = 0; tid < numThreads; ++tid) {
 157             Event *event = new CountedExitEvent(cause, *counter);
 158             comInstEventQueue[tid]->schedule(event, p->max_insts_all_threads);
 159         }
 160     }
 161
 162     // allocate per-thread load-based event queues
 163     comLoadEventQueue = new EventQueue *[numThreads];
 164     for (ThreadID tid = 0; tid < numThreads; ++tid)
 165         comLoadEventQueue[tid] = new EventQueue("load-based event queue");
 166
 167     //
 168     // set up instruction-count-based termination events, if any
 169     //
 170     if (p->max_loads_any_thread != 0) {
 171         const char *cause = "a thread reached the max load count";
 172         for (ThreadID tid = 0; tid < numThreads; ++tid) {
 173             Event *event = new SimLoopExitEvent(cause, 0);
 174             comLoadEventQueue[tid]->schedule(event, p->max_loads_any_thread);
 175         }
 176     }
 177
 178     if (p->max_loads_all_threads != 0) {
 179         const char *cause = "all threads reached the max load count";
 180         // allocate & initialize shared downcounter: each event will
 181         // decrement this when triggered; simulation will terminate
 182         // when counter reaches 0
 183         int *counter = new int;
 184         *counter = numThreads;
 185         for (ThreadID tid = 0; tid < numThreads; ++tid) {
 186             Event *event = new CountedExitEvent(cause, *counter);
 187             comLoadEventQueue[tid]->schedule(event, p->max_loads_all_threads);
 188         }
 189     }
 190
 191     functionTracingEnabled = false;
 192     if (p->function_trace) {
 193         const string fname = csprintf("ftrace.%s", name());
 194         functionTraceStream = simout.find(fname);
 195         if (!functionTraceStream)
 196             functionTraceStream = simout.create(fname);
 197
 198         currentFunctionStart = currentFunctionEnd = 0;
 199         functionEntryTick = p->function_trace_start;
 200
 201         if (p->function_trace_start == 0) {
 202             functionTracingEnabled = true;
 203         } else {
 204             typedef EventWrapper<BaseCPU, &BaseCPU::enableFunctionTrace> wrap;
 205             Event *event = new wrap(this, true);
 206             schedule(event, p->function_trace_start);
 207         }
 208     }
 209 #if FULL_SYSTEM
 210     interrupts->setCPU(this);
 211
 212     profileEvent = NULL;
 213     if (params()->profile)
 214         profileEvent = new ProfileEvent(this, params()->profile);
 215 #endif
 216     tracer = params()->tracer;
 217 }
 218
 219 void
 220 BaseCPU::enableFunctionTrace()
 221 {
 222     functionTracingEnabled = true;
 223 }
 224
 225 BaseCPU::~BaseCPU()
 226 {
 227 }
 228
 229 void
 230 BaseCPU::init()
 231 {
 232     if (!params()->defer_registration)
 233         registerThreadContexts();
 234 }
 235
 236 void
 237 BaseCPU::startup()
 238 {
 239 #if FULL_SYSTEM
 240     if (!params()->defer_registration && profileEvent)
 241         schedule(profileEvent, curTick());
 242 #endif
 243
 244     if (params()->progress_interval) {
 245         Tick num_ticks = ticks(params()->progress_interval);
 246
 247         new CPUProgressEvent(this, num_ticks);
 248     }
 249 }
 250
 251
 252 void
 253 BaseCPU::regStats()
 254 {
 255     using namespace Stats;
 256
 257     numCycles
 258         .name(name() + ".numCycles")
 259         .desc("number of cpu cycles simulated")
 260         ;
 261
 262     numWorkItemsStarted
 263         .name(name() + ".numWorkItemsStarted")
 264         .desc("number of work items this cpu started")
 265         ;
 266
 267     numWorkItemsCompleted
 268         .name(name() + ".numWorkItemsCompleted")
 269         .desc("number of work items this cpu completed")
 270         ;
 271
 272     int size = threadContexts.size();
 273     if (size > 1) {
 274         for (int i = 0; i < size; ++i) {
 275             stringstream namestr;
 276             ccprintf(namestr, "%s.ctx%d", name(), i);
 277             threadContexts[i]->regStats(namestr.str());
 278         }
 279     } else if (size == 1)
 280         threadContexts[0]->regStats(name());
 281
 282 #if FULL_SYSTEM
 283 #endif
 284 }
 285
 286 Tick
 287 BaseCPU::nextCycle()
 288 {
 289     Tick next_tick = curTick() - phase + clock - 1;
 290     next_tick -= (next_tick % clock);
 291     next_tick += phase;
 292     return next_tick;
 293 }
 294
 295 Tick
 296 BaseCPU::nextCycle(Tick begin_tick)
 297 {
 298     Tick next_tick = begin_tick;
 299     if (next_tick % clock != 0)
 300         next_tick = next_tick - (next_tick % clock) + clock;
 301     next_tick += phase;
 302
 303     assert(next_tick >= curTick());
 304     return next_tick;
 305 }
 306
 307 void
 308 BaseCPU::registerThreadContexts()
 309 {
 310     ThreadID size = threadContexts.size();
 311     for (ThreadID tid = 0; tid < size; ++tid) {
 312         ThreadContext *tc = threadContexts[tid];
 313
 314         /** This is so that contextId and cpuId match where there is a
 315          * 1cpu:1context relationship.  Otherwise, the order of registration
 316          * could affect the assignment and cpu 1 could have context id 3, for
 317          * example.  We may even want to do something like this for SMT so that
 318          * cpu 0 has the lowest thread contexts and cpu N has the highest, but
 319          * I'll just do this for now
 320          */
 321         if (numThreads == 1)
 322             tc->setContextId(system->registerThreadContext(tc, _cpuId));
 323         else
 324             tc->setContextId(system->registerThreadContext(tc));
 325 #if !FULL_SYSTEM
 326         tc->getProcessPtr()->assignThreadContext(tc->contextId());
 327 #endif
 328     }
 329 }
 330
 331
 332 int
 333 BaseCPU::findContext(ThreadContext *tc)
 334 {
 335     ThreadID size = threadContexts.size();
 336     for (ThreadID tid = 0; tid < size; ++tid) {
 337         if (tc == threadContexts[tid])
 338             return tid;
 339     }
 340     return 0;
 341 }
 342
 343 void
 344 BaseCPU::switchOut()
 345 {
 346 //    panic("This CPU doesn't support sampling!");
 347 #if FULL_SYSTEM
 348     if (profileEvent && profileEvent->scheduled())
 349         deschedule(profileEvent);
 350 #endif
 351 }
 352
 353 void
 354 BaseCPU::takeOverFrom(BaseCPU *oldCPU, Port *ic, Port *dc)
 355 {
 356     assert(threadContexts.size() == oldCPU->threadContexts.size());
 357
 358     _cpuId = oldCPU->cpuId();
 359
 360     ThreadID size = threadContexts.size();
 361     for (ThreadID i = 0; i < size; ++i) {
 362         ThreadContext *newTC = threadContexts[i];
 363         ThreadContext *oldTC = oldCPU->threadContexts[i];
 364
 365         newTC->takeOverFrom(oldTC);
 366
 367         CpuEvent::replaceThreadContext(oldTC, newTC);
 368
 369         assert(newTC->contextId() == oldTC->contextId());
 370         assert(newTC->threadId() == oldTC->threadId());
 371         system->replaceThreadContext(newTC, newTC->contextId());
 372
 373         /* This code no longer works since the zero register (e.g.,
 374          * r31 on Alpha) doesn't necessarily contain zero at this
 375          * point.
 376            if (DTRACE(Context))
 377             ThreadContext::compare(oldTC, newTC);
 378         */
 379
 380         Port  *old_itb_port, *old_dtb_port, *new_itb_port, *new_dtb_port;
 381         old_itb_port = oldTC->getITBPtr()->getPort();
 382         old_dtb_port = oldTC->getDTBPtr()->getPort();
 383         new_itb_port = newTC->getITBPtr()->getPort();
 384         new_dtb_port = newTC->getDTBPtr()->getPort();
 385
 386         // Move over any table walker ports if they exist
 387         if (new_itb_port && !new_itb_port->isConnected()) {
 388             assert(old_itb_port);
 389             Port *peer = old_itb_port->getPeer();;
 390             new_itb_port->setPeer(peer);
 391             peer->setPeer(new_itb_port);
 392         }
 393         if (new_dtb_port && !new_dtb_port->isConnected()) {
 394             assert(old_dtb_port);
 395             Port *peer = old_dtb_port->getPeer();;
 396             new_dtb_port->setPeer(peer);
 397             peer->setPeer(new_dtb_port);
 398         }
 399     }
 400
 401 #if FULL_SYSTEM
 402     interrupts = oldCPU->interrupts;
 403     interrupts->setCPU(this);
 404
 405     for (ThreadID i = 0; i < size; ++i)
 406         threadContexts[i]->profileClear();
 407
 408     if (profileEvent)
 409         schedule(profileEvent, curTick());
 410 #endif
 411
 412     // Connect new CPU to old CPU's memory only if new CPU isn't
 413     // connected to anything.  Also connect old CPU's memory to new
 414     // CPU.
 415     if (!ic->isConnected()) {
 416         Port *peer = oldCPU->getPort("icache_port")->getPeer();
 417         ic->setPeer(peer);
 418         peer->setPeer(ic);
 419     }
 420
 421     if (!dc->isConnected()) {
 422         Port *peer = oldCPU->getPort("dcache_port")->getPeer();
 423         dc->setPeer(peer);
 424         peer->setPeer(dc);
 425     }
 426 }
 427
 428
 429 #if FULL_SYSTEM
 430 BaseCPU::ProfileEvent::ProfileEvent(BaseCPU *_cpu, Tick _interval)
 431     : cpu(_cpu), interval(_interval)
 432 { }
 433
 434 void
 435 BaseCPU::ProfileEvent::process()
 436 {
 437     ThreadID size = cpu->threadContexts.size();
 438     for (ThreadID i = 0; i < size; ++i) {
 439         ThreadContext *tc = cpu->threadContexts[i];
 440         tc->profileSample();
 441     }
 442
 443     cpu->schedule(this, curTick() + interval);
 444 }
 445
 446 void
 447 BaseCPU::serialize(std::ostream &os)
 448 {
 449     SERIALIZE_SCALAR(instCnt);
 450     interrupts->serialize(os);
 451 }
 452
 453 void
 454 BaseCPU::unserialize(Checkpoint *cp, const std::string &section)
 455 {
 456     UNSERIALIZE_SCALAR(instCnt);
 457     interrupts->unserialize(cp, section);
 458 }
 459
 460 #endif // FULL_SYSTEM
 461
 462 void
 463 BaseCPU::traceFunctionsInternal(Addr pc)
 464 {
 465     if (!debugSymbolTable)
 466         return;
 467
 468     // if pc enters different function, print new function symbol and
 469     // update saved range.  Otherwise do nothing.
 470     if (pc < currentFunctionStart || pc >= currentFunctionEnd) {
 471         string sym_str;
 472         bool found = debugSymbolTable->findNearestSymbol(pc, sym_str,
 473                                                          currentFunctionStart,
 474                                                          currentFunctionEnd);
 475
 476         if (!found) {
 477             // no symbol found: use addr as label
 478             sym_str = csprintf("0x%x", pc);
 479             currentFunctionStart = pc;
 480             currentFunctionEnd = pc + 1;
 481         }
 482
 483         ccprintf(*functionTraceStream, " (%d)\n%d: %s",
 484                  curTick() - functionEntryTick, curTick(), sym_str);
 485         functionEntryTick = curTick();
 486     }
 487 }