diff options
Diffstat (limited to 'cpu')
-rw-r--r-- | cpu/base_cpu.cc | 68 | ||||
-rw-r--r-- | cpu/base_cpu.hh | 38 | ||||
-rw-r--r-- | cpu/exec_context.hh | 8 | ||||
-rw-r--r-- | cpu/exetrace.cc | 2 | ||||
-rw-r--r-- | cpu/full_cpu/op_class.hh | 1 | ||||
-rw-r--r-- | cpu/memtest/memtest.cc | 49 | ||||
-rw-r--r-- | cpu/memtest/memtest.hh | 17 | ||||
-rw-r--r-- | cpu/pc_event.cc | 7 | ||||
-rw-r--r-- | cpu/simple_cpu/simple_cpu.cc | 221 | ||||
-rw-r--r-- | cpu/simple_cpu/simple_cpu.hh | 94 | ||||
-rw-r--r-- | cpu/static_inst.cc | 4 | ||||
-rw-r--r-- | cpu/static_inst.hh | 26 | ||||
-rw-r--r-- | cpu/trace/opt_cpu.cc | 2 | ||||
-rw-r--r-- | cpu/trace/opt_cpu.hh | 11 | ||||
-rw-r--r-- | cpu/trace/trace_cpu.cc | 2 | ||||
-rw-r--r-- | cpu/trace/trace_cpu.hh | 11 |
16 files changed, 276 insertions, 285 deletions
diff --git a/cpu/base_cpu.cc b/cpu/base_cpu.cc index 7605ff3c3..181b484e7 100644 --- a/cpu/base_cpu.cc +++ b/cpu/base_cpu.cc @@ -26,15 +26,17 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ +#include <iostream> #include <string> #include <sstream> -#include <iostream> #include "base/cprintf.hh" #include "base/loader/symtab.hh" #include "base/misc.hh" +#include "base/output.hh" #include "cpu/base_cpu.hh" #include "cpu/exec_context.hh" +#include "cpu/sampling_cpu/sampling_cpu.hh" #include "sim/param.hh" #include "sim/sim_events.hh" @@ -48,25 +50,12 @@ vector<BaseCPU *> BaseCPU::cpuList; int maxThreadsPerCPU = 1; #ifdef FULL_SYSTEM -BaseCPU::BaseCPU(const string &_name, int _number_of_threads, bool _def_reg, - Counter max_insts_any_thread, - Counter max_insts_all_threads, - Counter max_loads_any_thread, - Counter max_loads_all_threads, - System *_system, Tick freq, - bool _function_trace, Tick _function_trace_start) - : SimObject(_name), frequency(freq), checkInterrupts(true), - deferRegistration(_def_reg), number_of_threads(_number_of_threads), - system(_system) +BaseCPU::BaseCPU(Params *p) + : SimObject(p->name), frequency(p->freq), checkInterrupts(true), + params(p), number_of_threads(p->numberOfThreads), system(p->system) #else -BaseCPU::BaseCPU(const string &_name, int _number_of_threads, bool _def_reg, - Counter max_insts_any_thread, - Counter max_insts_all_threads, - Counter max_loads_any_thread, - Counter max_loads_all_threads, - bool _function_trace, Tick _function_trace_start) - : SimObject(_name), deferRegistration(_def_reg), - number_of_threads(_number_of_threads) +BaseCPU::BaseCPU(Params *p) + : SimObject(p->name), params(p), number_of_threads(p->numberOfThreads) #endif { // add self to global list of CPUs @@ -76,19 +65,19 @@ BaseCPU::BaseCPU(const string &_name, int _number_of_threads, bool _def_reg, maxThreadsPerCPU = number_of_threads; // allocate per-thread instruction-based event queues - comInstEventQueue = new (EventQueue *)[number_of_threads]; + comInstEventQueue = new EventQueue *[number_of_threads]; for (int i = 0; i < number_of_threads; ++i) comInstEventQueue[i] = new EventQueue("instruction-based event queue"); // // set up instruction-count-based termination events, if any // - if (max_insts_any_thread != 0) + if (p->max_insts_any_thread != 0) for (int i = 0; i < number_of_threads; ++i) - new SimExitEvent(comInstEventQueue[i], max_insts_any_thread, + new SimExitEvent(comInstEventQueue[i], p->max_insts_any_thread, "a thread reached the max instruction count"); - if (max_insts_all_threads != 0) { + if (p->max_insts_all_threads != 0) { // allocate & initialize shared downcounter: each event will // decrement this when triggered; simulation will terminate // when counter reaches 0 @@ -97,23 +86,23 @@ BaseCPU::BaseCPU(const string &_name, int _number_of_threads, bool _def_reg, for (int i = 0; i < number_of_threads; ++i) new CountedExitEvent(comInstEventQueue[i], "all threads reached the max instruction count", - max_insts_all_threads, *counter); + p->max_insts_all_threads, *counter); } // allocate per-thread load-based event queues - comLoadEventQueue = new (EventQueue *)[number_of_threads]; + comLoadEventQueue = new EventQueue *[number_of_threads]; for (int i = 0; i < number_of_threads; ++i) comLoadEventQueue[i] = new EventQueue("load-based event queue"); // // set up instruction-count-based termination events, if any // - if (max_loads_any_thread != 0) + if (p->max_loads_any_thread != 0) for (int i = 0; i < number_of_threads; ++i) - new SimExitEvent(comLoadEventQueue[i], max_loads_any_thread, + new SimExitEvent(comLoadEventQueue[i], p->max_loads_any_thread, "a thread reached the max load count"); - if (max_loads_all_threads != 0) { + if (p->max_loads_all_threads != 0) { // allocate & initialize shared downcounter: each event will // decrement this when triggered; simulation will terminate // when counter reaches 0 @@ -122,7 +111,7 @@ BaseCPU::BaseCPU(const string &_name, int _number_of_threads, bool _def_reg, for (int i = 0; i < number_of_threads; ++i) new CountedExitEvent(comLoadEventQueue[i], "all threads reached the max load count", - max_loads_all_threads, *counter); + p->max_loads_all_threads, *counter); } #ifdef FULL_SYSTEM @@ -131,19 +120,18 @@ BaseCPU::BaseCPU(const string &_name, int _number_of_threads, bool _def_reg, #endif functionTracingEnabled = false; - if (_function_trace) { - std::string filename = csprintf("ftrace.%s", name()); - functionTraceStream = makeOutputStream(filename); + if (p->functionTrace) { + functionTraceStream = simout.find(csprintf("ftrace.%s", name())); currentFunctionStart = currentFunctionEnd = 0; - functionEntryTick = _function_trace_start; + functionEntryTick = p->functionTraceStart; - if (_function_trace_start == 0) { + if (p->functionTraceStart == 0) { functionTracingEnabled = true; } else { Event *e = new EventWrapper<BaseCPU, &BaseCPU::enableFunctionTrace>(this, true); - e->schedule(_function_trace_start); + e->schedule(p->functionTraceStart); } } } @@ -157,15 +145,12 @@ BaseCPU::enableFunctionTrace() BaseCPU::~BaseCPU() { - if (functionTracingEnabled) - closeOutputStream(functionTraceStream); } - void BaseCPU::init() { - if (!deferRegistration) + if (!params->deferRegistration) registerExecContexts(); } @@ -210,9 +195,10 @@ BaseCPU::registerExecContexts() void -BaseCPU::switchOut() +BaseCPU::switchOut(SamplingCPU *sampler) { - // default: do nothing + // default: do nothing, signal done + sampler->signalSwitched(); } void diff --git a/cpu/base_cpu.hh b/cpu/base_cpu.hh index baa956aa8..f346f4ec5 100644 --- a/cpu/base_cpu.hh +++ b/cpu/base_cpu.hh @@ -32,6 +32,7 @@ #include <vector> #include "base/statistics.hh" +#include "cpu/sampling_cpu/sampling_cpu.hh" #include "sim/eventq.hh" #include "sim/sim_object.hh" #include "targetarch/isa_traits.hh" @@ -90,33 +91,36 @@ class BaseCPU : public SimObject virtual void haltContext(int thread_num) {} public: - + struct Params + { + std::string name; + int numberOfThreads; + bool deferRegistration; + Counter max_insts_any_thread; + Counter max_insts_all_threads; + Counter max_loads_any_thread; + Counter max_loads_all_threads; + Tick freq; + bool functionTrace; + Tick functionTraceStart; #ifdef FULL_SYSTEM - BaseCPU(const std::string &_name, int _number_of_threads, bool _def_reg, - Counter max_insts_any_thread, Counter max_insts_all_threads, - Counter max_loads_any_thread, Counter max_loads_all_threads, - System *_system, Tick freq, - bool _function_trace = false, Tick _function_trace_start = 0); -#else - BaseCPU(const std::string &_name, int _number_of_threads, bool _def_reg, - Counter max_insts_any_thread = 0, - Counter max_insts_all_threads = 0, - Counter max_loads_any_thread = 0, - Counter max_loads_all_threads = 0, - bool _function_trace = false, Tick _function_trace_start = 0); + System *system; #endif + }; + + const Params *params; + BaseCPU(Params *params); virtual ~BaseCPU(); virtual void init(); virtual void regStats(); - bool deferRegistration; void registerExecContexts(); - /// Prepare for another CPU to take over execution. Called by - /// takeOverFrom() on its argument. - virtual void switchOut(); + /// Prepare for another CPU to take over execution. When it is + /// is ready (drained pipe) it signals the sampler. + virtual void switchOut(SamplingCPU *); /// Take over execution from the given CPU. Used for warm-up and /// sampling. diff --git a/cpu/exec_context.hh b/cpu/exec_context.hh index 8437a5585..d6140d52f 100644 --- a/cpu/exec_context.hh +++ b/cpu/exec_context.hh @@ -435,20 +435,20 @@ class ExecContext regs.intRegFile[ArgumentReg0 + i] = val; } - void setSyscallReturn(int64_t return_value) + void setSyscallReturn(SyscallReturn return_value) { // check for error condition. Alpha syscall convention is to // indicate success/failure in reg a3 (r19) and put the // return value itself in the standard return value reg (v0). const int RegA3 = 19; // only place this is used - if (return_value >= 0) { + if (return_value.successful()) { // no error regs.intRegFile[RegA3] = 0; - regs.intRegFile[ReturnValueReg] = return_value; + regs.intRegFile[ReturnValueReg] = return_value.value(); } else { // got an error, return details regs.intRegFile[RegA3] = (IntReg) -1; - regs.intRegFile[ReturnValueReg] = -return_value; + regs.intRegFile[ReturnValueReg] = -return_value.value(); } } diff --git a/cpu/exetrace.cc b/cpu/exetrace.cc index ff7e90c9e..048a7d283 100644 --- a/cpu/exetrace.cc +++ b/cpu/exetrace.cc @@ -154,7 +154,7 @@ class ExecutionTraceParamContext : public ParamContext ExecutionTraceParamContext exeTraceParams("exetrace"); Param<bool> exe_trace_spec(&exeTraceParams, "speculative", - "capture speculative instructions", false); + "capture speculative instructions", true); Param<bool> exe_trace_print_cycle(&exeTraceParams, "print_cycle", "print cycle number", true); diff --git a/cpu/full_cpu/op_class.hh b/cpu/full_cpu/op_class.hh index a14ccfaed..8e85e8d8a 100644 --- a/cpu/full_cpu/op_class.hh +++ b/cpu/full_cpu/op_class.hh @@ -51,6 +51,7 @@ enum OpClass { FloatSqrtOp, /* floating point square root */ MemReadOp, /* memory read port */ MemWriteOp, /* memory write port */ + IprAccessOp, /* Internal Processor Register read/write port */ InstPrefetchOp, /* instruction prefetch port (on I-cache) */ Num_OpClasses /* total functional unit classes */ }; diff --git a/cpu/memtest/memtest.cc b/cpu/memtest/memtest.cc index e967c79da..14b119880 100644 --- a/cpu/memtest/memtest.cc +++ b/cpu/memtest/memtest.cc @@ -36,6 +36,7 @@ #include "base/misc.hh" #include "base/statistics.hh" +#include "cpu/exec_context.hh" #include "cpu/memtest/memtest.hh" #include "mem/cache/base_cache.hh" #include "mem/functional_mem/main_memory.hh" @@ -59,10 +60,8 @@ MemTest::MemTest(const string &name, unsigned _percentSourceUnaligned, unsigned _percentDestUnaligned, Addr _traceAddr, - Counter max_loads_any_thread, - Counter max_loads_all_threads) - : BaseCPU(name, 1, true, 0, 0, max_loads_any_thread, - max_loads_all_threads), + Counter _max_loads) + : SimObject(name), tickEvent(this), cacheInterface(_cache_interface), mainMem(main_mem), @@ -74,12 +73,13 @@ MemTest::MemTest(const string &name, progressInterval(_progressInterval), nextProgressMessage(_progressInterval), percentSourceUnaligned(_percentSourceUnaligned), - percentDestUnaligned(percentDestUnaligned) + percentDestUnaligned(percentDestUnaligned), + maxLoads(_max_loads) { vector<string> cmd; cmd.push_back("/bin/ls"); vector<string> null_vec; - xc = new ExecContext(this ,0,mainMem,0); + xc = new ExecContext(NULL, 0, mainMem, 0); blockSize = cacheInterface->getBlockSize(); blockAddrMask = blockSize - 1; @@ -160,7 +160,8 @@ MemTest::completeRequest(MemReqPtr &req, uint8_t *data) nextProgressMessage += progressInterval; } - comLoadEventQueue[0]->serviceEvents(numReads); + if (numReads >= maxLoads) + SimExit(curTick, "Maximum number of loads reached!"); break; case Write: @@ -402,8 +403,7 @@ BEGIN_DECLARE_SIM_OBJECT_PARAMS(MemTest) Param<unsigned> percent_source_unaligned; Param<unsigned> percent_dest_unaligned; Param<Addr> trace_addr; - Param<Counter> max_loads_any_thread; - Param<Counter> max_loads_all_threads; + Param<Counter> max_loads; END_DECLARE_SIM_OBJECT_PARAMS(MemTest) @@ -413,23 +413,17 @@ BEGIN_INIT_SIM_OBJECT_PARAMS(MemTest) INIT_PARAM(cache, "L1 cache"), INIT_PARAM(main_mem, "hierarchical memory"), INIT_PARAM(check_mem, "check memory"), - INIT_PARAM_DFLT(memory_size, "memory size", 65536), - INIT_PARAM_DFLT(percent_reads, "target read percentage", 65), - INIT_PARAM_DFLT(percent_copies, "target copy percentage", 0), - INIT_PARAM_DFLT(percent_uncacheable, "target uncacheable percentage", 10), - INIT_PARAM_DFLT(progress_interval, - "progress report interval (in accesses)", 1000000), - INIT_PARAM_DFLT(percent_source_unaligned, "percent of copy source address " - "that are unaligned", 50), - INIT_PARAM_DFLT(percent_dest_unaligned, "percent of copy dest address " - "that are unaligned", 50), - INIT_PARAM_DFLT(trace_addr, "address to trace", 0), - INIT_PARAM_DFLT(max_loads_any_thread, - "terminate when any thread reaches this load count", - 0), - INIT_PARAM_DFLT(max_loads_all_threads, - "terminate when all threads have reached this load count", - 0) + INIT_PARAM(memory_size, "memory size"), + INIT_PARAM(percent_reads, "target read percentage"), + INIT_PARAM(percent_copies, "target copy percentage"), + INIT_PARAM(percent_uncacheable, "target uncacheable percentage"), + INIT_PARAM(progress_interval, "progress report interval (in accesses)"), + INIT_PARAM(percent_source_unaligned, + "percent of copy source address that are unaligned"), + INIT_PARAM(percent_dest_unaligned, + "percent of copy dest address that are unaligned"), + INIT_PARAM(trace_addr, "address to trace"), + INIT_PARAM(max_loads, "terminate when we have reached this load count") END_INIT_SIM_OBJECT_PARAMS(MemTest) @@ -440,8 +434,7 @@ CREATE_SIM_OBJECT(MemTest) check_mem, memory_size, percent_reads, percent_copies, percent_uncacheable, progress_interval, percent_source_unaligned, percent_dest_unaligned, - trace_addr, max_loads_any_thread, - max_loads_all_threads); + trace_addr, max_loads); } REGISTER_SIM_OBJECT("MemTest", MemTest) diff --git a/cpu/memtest/memtest.hh b/cpu/memtest/memtest.hh index 43b17a713..45b2d24e8 100644 --- a/cpu/memtest/memtest.hh +++ b/cpu/memtest/memtest.hh @@ -26,20 +26,21 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -#ifndef __MEMTEST_HH__ -#define __MEMTEST_HH__ +#ifndef __CPU_MEMTEST_MEMTEST_HH__ +#define __CPU_MEMTEST_MEMTEST_HH__ #include <set> #include "base/statistics.hh" -#include "cpu/base_cpu.hh" -#include "cpu/exec_context.hh" #include "mem/functional_mem/functional_memory.hh" #include "mem/mem_interface.hh" +#include "sim/eventq.hh" +#include "sim/sim_exit.hh" #include "sim/sim_object.hh" #include "sim/stats.hh" -class MemTest : public BaseCPU +class ExecContext; +class MemTest : public SimObject { public: @@ -55,8 +56,7 @@ class MemTest : public BaseCPU unsigned _percentSourceUnaligned, unsigned _percentDestUnaligned, Addr _traceAddr, - Counter max_loads_any_thread, - Counter max_loads_all_threads); + Counter _max_loads); // register statistics virtual void regStats(); @@ -116,6 +116,7 @@ class MemTest : public BaseCPU Tick noResponseCycles; uint64_t numReads; + uint64_t maxLoads; Stats::Scalar<> numReadsStat; Stats::Scalar<> numWritesStat; Stats::Scalar<> numCopiesStat; @@ -146,7 +147,7 @@ class MemCompleteEvent : public Event virtual const char *description(); }; -#endif // __MEMTEST_HH__ +#endif // __CPU_MEMTEST_MEMTEST_HH__ diff --git a/cpu/pc_event.cc b/cpu/pc_event.cc index 8f046a7a4..21eef677a 100644 --- a/cpu/pc_event.cc +++ b/cpu/pc_event.cc @@ -31,10 +31,11 @@ #include <string> #include <utility> -#include "sim/debug.hh" +#include "base/trace.hh" +#include "cpu/base_cpu.hh" #include "cpu/exec_context.hh" #include "cpu/pc_event.hh" -#include "base/trace.hh" +#include "sim/debug.hh" #include "sim/universe.hh" using namespace std; @@ -123,6 +124,8 @@ BreakPCEvent::BreakPCEvent(PCEventQueue *q, const std::string &desc, bool del) void BreakPCEvent::process(ExecContext *xc) { + StringWrap name(xc->cpu->name() + ".break_event"); + DPRINTFN("break event %s triggered", descr()); debug_break(); if (remove) delete this; diff --git a/cpu/simple_cpu/simple_cpu.cc b/cpu/simple_cpu/simple_cpu.cc index d48f93663..86aeab7d7 100644 --- a/cpu/simple_cpu/simple_cpu.cc +++ b/cpu/simple_cpu/simple_cpu.cc @@ -47,6 +47,7 @@ #include "cpu/exec_context.hh" #include "cpu/exetrace.hh" #include "cpu/full_cpu/smt.hh" +#include "cpu/sampling_cpu/sampling_cpu.hh" #include "cpu/simple_cpu/simple_cpu.hh" #include "cpu/static_inst.hh" #include "mem/base_mem.hh" @@ -74,15 +75,16 @@ using namespace std; -SimpleCPU::TickEvent::TickEvent(SimpleCPU *c) - : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c), multiplier(1) + +SimpleCPU::TickEvent::TickEvent(SimpleCPU *c, int w) + : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c), width(w) { } void SimpleCPU::TickEvent::process() { - int count = multiplier; + int count = width; do { cpu->tick(); } while (--count > 0 && cpu->status() == Running); @@ -96,8 +98,7 @@ SimpleCPU::TickEvent::description() SimpleCPU::CacheCompletionEvent::CacheCompletionEvent(SimpleCPU *_cpu) - : Event(&mainEventQueue), - cpu(_cpu) + : Event(&mainEventQueue), cpu(_cpu) { } @@ -112,52 +113,22 @@ SimpleCPU::CacheCompletionEvent::description() return "SimpleCPU cache completion event"; } -#ifdef FULL_SYSTEM -SimpleCPU::SimpleCPU(const string &_name, - System *_system, - Counter max_insts_any_thread, - Counter max_insts_all_threads, - Counter max_loads_any_thread, - Counter max_loads_all_threads, - AlphaITB *itb, AlphaDTB *dtb, - FunctionalMemory *mem, - MemInterface *icache_interface, - MemInterface *dcache_interface, - bool _def_reg, Tick freq, - bool _function_trace, Tick _function_trace_start) - : BaseCPU(_name, /* number_of_threads */ 1, _def_reg, - max_insts_any_thread, max_insts_all_threads, - max_loads_any_thread, max_loads_all_threads, - _system, freq, _function_trace, _function_trace_start), -#else -SimpleCPU::SimpleCPU(const string &_name, Process *_process, - Counter max_insts_any_thread, - Counter max_insts_all_threads, - Counter max_loads_any_thread, - Counter max_loads_all_threads, - MemInterface *icache_interface, - MemInterface *dcache_interface, - bool _def_reg, - bool _function_trace, Tick _function_trace_start) - : BaseCPU(_name, /* number_of_threads */ 1, _def_reg, - max_insts_any_thread, max_insts_all_threads, - max_loads_any_thread, max_loads_all_threads, - _function_trace, _function_trace_start), -#endif - tickEvent(this), xc(NULL), cacheCompletionEvent(this) +SimpleCPU::SimpleCPU(Params *p) + : BaseCPU(p), tickEvent(this, p->width), xc(NULL), + cacheCompletionEvent(this) { _status = Idle; #ifdef FULL_SYSTEM - xc = new ExecContext(this, 0, system, itb, dtb, mem); + xc = new ExecContext(this, 0, p->system, p->itb, p->dtb, p->mem); // initialize CPU, including PC TheISA::initCPU(&xc->regs); #else - xc = new ExecContext(this, /* thread_num */ 0, _process, /* asid */ 0); + xc = new ExecContext(this, /* thread_num */ 0, p->process, /* asid */ 0); #endif // !FULL_SYSTEM - icacheInterface = icache_interface; - dcacheInterface = dcache_interface; + icacheInterface = p->icache_interface; + dcacheInterface = p->dcache_interface; memReq = new MemReq(); memReq->xc = xc; @@ -179,11 +150,21 @@ SimpleCPU::~SimpleCPU() } void -SimpleCPU::switchOut() +SimpleCPU::switchOut(SamplingCPU *s) { - _status = SwitchedOut; - if (tickEvent.scheduled()) - tickEvent.squash(); + sampler = s; + if (status() == DcacheMissStall) { + DPRINTF(Sampler,"Outstanding dcache access, waiting for completion\n"); + _status = DcacheMissSwitch; + } + else { + _status = SwitchedOut; + + if (tickEvent.scheduled()) + tickEvent.squash(); + + sampler->signalSwitched(); + } } @@ -203,8 +184,6 @@ SimpleCPU::takeOverFrom(BaseCPU *oldCPU) tickEvent.schedule(curTick); } } - - oldCPU->switchOut(); } @@ -414,21 +393,22 @@ template <class T> Fault SimpleCPU::read(Addr addr, T &data, unsigned flags) { + if (status() == DcacheMissStall) { + Fault fault = xc->read(memReq,data); + + if (traceData) { + traceData->setAddr(addr); + if (fault == No_Fault) + traceData->setData(data); + } + return fault; + } + memReq->reset(addr, sizeof(T), flags); // translate to physical address Fault fault = xc->translateDataReadReq(memReq); - // do functional access - if (fault == No_Fault) - fault = xc->read(memReq, data); - - if (traceData) { - traceData->setAddr(addr); - if (fault == No_Fault) - traceData->setData(data); - } - // if we have a cache, do cache access too if (fault == No_Fault && dcacheInterface) { memReq->cmd = Read; @@ -444,6 +424,24 @@ SimpleCPU::read(Addr addr, T &data, unsigned flags) lastDcacheStall = curTick; unscheduleTickEvent(); _status = DcacheMissStall; + } else { + // do functional access + fault = xc->read(memReq, data); + + if (traceData) { + traceData->setAddr(addr); + if (fault == No_Fault) + traceData->setData(data); + } + } + } else if(fault == No_Fault) { + // do functional access + fault = xc->read(memReq, data); + + if (traceData) { + traceData->setAddr(addr); + if (fault == No_Fault) + traceData->setData(data); } } @@ -605,10 +603,19 @@ SimpleCPU::processCacheCompletion() scheduleTickEvent(1); break; case DcacheMissStall: + if (memReq->cmd.isRead()) { + curStaticInst->execute(this,traceData); + } dcacheStallCycles += curTick - lastDcacheStall; _status = Running; scheduleTickEvent(1); break; + case DcacheMissSwitch: + if (memReq->cmd.isRead()) { + curStaticInst->execute(this,traceData); + } + _status = SwitchedOut; + sampler->signalSwitched(); case SwitchedOut: // If this CPU has been switched out due to sampling/warm-up, // ignore any further status changes (e.g., due to cache @@ -750,10 +757,10 @@ SimpleCPU::tick() comInstEventQueue[0]->serviceEvents(numInst); // decode the instruction - inst = htoa(inst); - StaticInstPtr<TheISA> si(inst); + inst = htoa(inst); + curStaticInst = StaticInst<TheISA>::decode(inst); - traceData = Trace::getInstRecord(curTick, xc, this, si, + traceData = Trace::getInstRecord(curTick, xc, this, curStaticInst, xc->regs.pc); #ifdef FULL_SYSTEM @@ -762,18 +769,18 @@ SimpleCPU::tick() xc->func_exe_inst++; - fault = si->execute(this, traceData); + fault = curStaticInst->execute(this, traceData); #ifdef FULL_SYSTEM if (xc->fnbin) - xc->execute(si.get()); + xc->execute(curStaticInst.get()); #endif - if (si->isMemRef()) { + if (curStaticInst->isMemRef()) { numMemRefs++; } - if (si->isLoad()) { + if (curStaticInst->isLoad()) { ++numLoad; comLoadEventQueue[0]->serviceEvents(numLoad); } @@ -840,7 +847,7 @@ BEGIN_DECLARE_SIM_OBJECT_PARAMS(SimpleCPU) SimObjectParam<BaseMem *> dcache; Param<bool> defer_registration; - Param<int> multiplier; + Param<int> width; Param<bool> function_trace; Param<Tick> function_trace_start; @@ -848,71 +855,67 @@ END_DECLARE_SIM_OBJECT_PARAMS(SimpleCPU) BEGIN_INIT_SIM_OBJECT_PARAMS(SimpleCPU) - INIT_PARAM_DFLT(max_insts_any_thread, - "terminate when any thread reaches this inst count", - 0), - INIT_PARAM_DFLT(max_insts_all_threads, - "terminate when all threads have reached this inst count", - 0), - INIT_PARAM_DFLT(max_loads_any_thread, - "terminate when any thread reaches this load count", - 0), - INIT_PARAM_DFLT(max_loads_all_threads, - "terminate when all threads have reached this load count", - 0), + INIT_PARAM(max_insts_any_thread, + "terminate when any thread reaches this inst count"), + INIT_PARAM(max_insts_all_threads, + "terminate when all threads have reached this inst count"), + INIT_PARAM(max_loads_any_thread, + "terminate when any thread reaches this load count"), + INIT_PARAM(max_loads_all_threads, + "terminate when all threads have reached this load count"), #ifdef FULL_SYSTEM INIT_PARAM(itb, "Instruction TLB"), INIT_PARAM(dtb, "Data TLB"), INIT_PARAM(mem, "memory"), INIT_PARAM(system, "system object"), - INIT_PARAM_DFLT(mult, "system clock multiplier", 1), + INIT_PARAM(mult, "system clock multiplier"), #else INIT_PARAM(workload, "processes to run"), #endif // FULL_SYSTEM - INIT_PARAM_DFLT(icache, "L1 instruction cache object", NULL), - INIT_PARAM_DFLT(dcache, "L1 data cache object", NULL), - INIT_PARAM_DFLT(defer_registration, "defer registration with system " - "(for sampling)", false), - - INIT_PARAM_DFLT(multiplier, "clock multiplier", 1), - INIT_PARAM_DFLT(function_trace, "Enable function trace", false), - INIT_PARAM_DFLT(function_trace_start, "Cycle to start function trace", 0) + INIT_PARAM(icache, "L1 instruction cache object"), + INIT_PARAM(dcache, "L1 data cache object"), + INIT_PARAM(defer_registration, "defer system registration (for sampling)"), + INIT_PARAM(width, "cpu width"), + INIT_PARAM(function_trace, "Enable function trace"), + INIT_PARAM(function_trace_start, "Cycle to start function trace") END_INIT_SIM_OBJECT_PARAMS(SimpleCPU) CREATE_SIM_OBJECT(SimpleCPU) { - SimpleCPU *cpu; #ifdef FULL_SYSTEM if (mult != 1) panic("processor clock multiplier must be 1\n"); +#endif - cpu = new SimpleCPU(getInstanceName(), system, - max_insts_any_thread, max_insts_all_threads, - max_loads_any_thread, max_loads_all_threads, - itb, dtb, mem, - (icache) ? icache->getInterface() : NULL, - (dcache) ? dcache->getInterface() : NULL, - defer_registration, - ticksPerSecond * mult, - function_trace, function_trace_start); -#else - - cpu = new SimpleCPU(getInstanceName(), workload, - max_insts_any_thread, max_insts_all_threads, - max_loads_any_thread, max_loads_all_threads, - (icache) ? icache->getInterface() : NULL, - (dcache) ? dcache->getInterface() : NULL, - defer_registration, - function_trace, function_trace_start); + SimpleCPU::Params *params = new SimpleCPU::Params(); + params->name = getInstanceName(); + params->numberOfThreads = 1; + params->max_insts_any_thread = max_insts_any_thread; + params->max_insts_all_threads = max_insts_all_threads; + params->max_loads_any_thread = max_loads_any_thread; + params->max_loads_all_threads = max_loads_all_threads; + params->deferRegistration = defer_registration; + params->freq = ticksPerSecond; + params->functionTrace = function_trace; + params->functionTraceStart = function_trace_start; + params->icache_interface = (icache) ? icache->getInterface() : NULL; + params->dcache_interface = (dcache) ? dcache->getInterface() : NULL; + params->width = width; -#endif // FULL_SYSTEM - - cpu->setTickMultiplier(multiplier); +#ifdef FULL_SYSTEM + params->itb = itb; + params->dtb = dtb; + params->mem = mem; + params->system = system; +#else + params->process = workload; +#endif + SimpleCPU *cpu = new SimpleCPU(params); return cpu; } diff --git a/cpu/simple_cpu/simple_cpu.hh b/cpu/simple_cpu/simple_cpu.hh index 341a0da23..f245a7bba 100644 --- a/cpu/simple_cpu/simple_cpu.hh +++ b/cpu/simple_cpu/simple_cpu.hh @@ -26,15 +26,16 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -#ifndef __SIMPLE_CPU_HH__ -#define __SIMPLE_CPU_HH__ +#ifndef __CPU_SIMPLE_CPU_SIMPLE_CPU_HH__ +#define __CPU_SIMPLE_CPU_SIMPLE_CPU_HH__ -#include "cpu/base_cpu.hh" -#include "sim/eventq.hh" -#include "cpu/pc_event.hh" #include "base/statistics.hh" +#include "cpu/base_cpu.hh" #include "cpu/exec_context.hh" +#include "cpu/pc_event.hh" +#include "cpu/sampling_cpu/sampling_cpu.hh" #include "cpu/static_inst.hh" +#include "sim/eventq.hh" // forward declarations #ifdef FULL_SYSTEM @@ -69,9 +70,9 @@ class SimpleCPU : public BaseCPU struct TickEvent : public Event { SimpleCPU *cpu; - int multiplier; + int width; - TickEvent(SimpleCPU *c); + TickEvent(SimpleCPU *c, int w); void process(); const char *description(); }; @@ -94,20 +95,8 @@ class SimpleCPU : public BaseCPU tickEvent.squash(); } - public: - void setTickMultiplier(int multiplier) - { - tickEvent.multiplier = multiplier; - } - private: Trace::InstRecord *traceData; - template<typename T> - void trace_data(T data) { - if (traceData) { - traceData->setData(data); - } - }; public: // @@ -117,6 +106,7 @@ class SimpleCPU : public BaseCPU IcacheMissStall, IcacheMissComplete, DcacheMissStall, + DcacheMissSwitch, SwitchedOut }; @@ -134,36 +124,28 @@ class SimpleCPU : public BaseCPU } }; + public: + struct Params : public BaseCPU::Params + { + MemInterface *icache_interface; + MemInterface *dcache_interface; + int width; #ifdef FULL_SYSTEM - - SimpleCPU(const std::string &_name, - System *_system, - Counter max_insts_any_thread, Counter max_insts_all_threads, - Counter max_loads_any_thread, Counter max_loads_all_threads, - AlphaITB *itb, AlphaDTB *dtb, FunctionalMemory *mem, - MemInterface *icache_interface, MemInterface *dcache_interface, - bool _def_reg, Tick freq, - bool _function_trace, Tick _function_trace_start); - + AlphaITB *itb; + AlphaDTB *dtb; + FunctionalMemory *mem; #else - - SimpleCPU(const std::string &_name, Process *_process, - Counter max_insts_any_thread, - Counter max_insts_all_threads, - Counter max_loads_any_thread, - Counter max_loads_all_threads, - MemInterface *icache_interface, MemInterface *dcache_interface, - bool _def_reg, - bool _function_trace, Tick _function_trace_start); - + Process *process; #endif - + }; + SimpleCPU(Params *params); virtual ~SimpleCPU(); + public: // execution context ExecContext *xc; - void switchOut(); + void switchOut(SamplingCPU *s); void takeOverFrom(BaseCPU *oldCPU); #ifdef FULL_SYSTEM @@ -184,6 +166,13 @@ class SimpleCPU : public BaseCPU // Refcounted pointer to the one memory request. MemReqPtr memReq; + // Pointer to the sampler that is telling us to switchover. + // Used to signal the completion of the pipe drain and schedule + // the next switchover + SamplingCPU *sampler; + + StaticInstPtr<TheISA> curStaticInst; + class CacheCompletionEvent : public Event { private: @@ -249,6 +238,11 @@ class SimpleCPU : public BaseCPU template <class T> Fault write(T data, Addr addr, unsigned flags, uint64_t *res); + // These functions are only used in CPU models that split + // effective address computation from the actual memory access. + void setEA(Addr EA) { panic("SimpleCPU::setEA() not implemented\n"); } + Addr getEA() { panic("SimpleCPU::getEA() not implemented\n"); } + void prefetch(Addr addr, unsigned flags) { // need to do this... @@ -274,47 +268,47 @@ class SimpleCPU : public BaseCPU // storage (which is pretty hard to imagine they would have reason // to do). - uint64_t readIntReg(StaticInst<TheISA> *si, int idx) + uint64_t readIntReg(const StaticInst<TheISA> *si, int idx) { return xc->readIntReg(si->srcRegIdx(idx)); } - float readFloatRegSingle(StaticInst<TheISA> *si, int idx) + float readFloatRegSingle(const StaticInst<TheISA> *si, int idx) { int reg_idx = si->srcRegIdx(idx) - TheISA::FP_Base_DepTag; return xc->readFloatRegSingle(reg_idx); } - double readFloatRegDouble(StaticInst<TheISA> *si, int idx) + double readFloatRegDouble(const StaticInst<TheISA> *si, int idx) { int reg_idx = si->srcRegIdx(idx) - TheISA::FP_Base_DepTag; return xc->readFloatRegDouble(reg_idx); } - uint64_t readFloatRegInt(StaticInst<TheISA> *si, int idx) + uint64_t readFloatRegInt(const StaticInst<TheISA> *si, int idx) { int reg_idx = si->srcRegIdx(idx) - TheISA::FP_Base_DepTag; return xc->readFloatRegInt(reg_idx); } - void setIntReg(StaticInst<TheISA> *si, int idx, uint64_t val) + void setIntReg(const StaticInst<TheISA> *si, int idx, uint64_t val) { xc->setIntReg(si->destRegIdx(idx), val); } - void setFloatRegSingle(StaticInst<TheISA> *si, int idx, float val) + void setFloatRegSingle(const StaticInst<TheISA> *si, int idx, float val) { int reg_idx = si->destRegIdx(idx) - TheISA::FP_Base_DepTag; xc->setFloatRegSingle(reg_idx, val); } - void setFloatRegDouble(StaticInst<TheISA> *si, int idx, double val) + void setFloatRegDouble(const StaticInst<TheISA> *si, int idx, double val) { int reg_idx = si->destRegIdx(idx) - TheISA::FP_Base_DepTag; xc->setFloatRegDouble(reg_idx, val); } - void setFloatRegInt(StaticInst<TheISA> *si, int idx, uint64_t val) + void setFloatRegInt(const StaticInst<TheISA> *si, int idx, uint64_t val) { int reg_idx = si->destRegIdx(idx) - TheISA::FP_Base_DepTag; xc->setFloatRegInt(reg_idx, val); @@ -346,4 +340,4 @@ class SimpleCPU : public BaseCPU ExecContext *xcBase() { return xc; } }; -#endif // __SIMPLE_CPU_HH__ +#endif // __CPU_SIMPLE_CPU_SIMPLE_CPU_HH__ diff --git a/cpu/static_inst.cc b/cpu/static_inst.cc index 4cb45a818..d522dbf5a 100644 --- a/cpu/static_inst.cc +++ b/cpu/static_inst.cc @@ -68,7 +68,7 @@ StaticInst<AlphaISA>::nullStaticInstPtr; template <class ISA> bool -StaticInst<ISA>::hasBranchTarget(Addr pc, ExecContext *xc, Addr &tgt) +StaticInst<ISA>::hasBranchTarget(Addr pc, ExecContext *xc, Addr &tgt) const { if (isDirectCtrl()) { tgt = branchTarget(pc); @@ -85,4 +85,4 @@ StaticInst<ISA>::hasBranchTarget(Addr pc, ExecContext *xc, Addr &tgt) // force instantiation of template function(s) above -template StaticInst<AlphaISA>; +template class StaticInst<AlphaISA>; diff --git a/cpu/static_inst.hh b/cpu/static_inst.hh index 46b2e4b19..75bdcc286 100644 --- a/cpu/static_inst.hh +++ b/cpu/static_inst.hh @@ -285,13 +285,13 @@ class StaticInst : public StaticInstBase * String representation of disassembly (lazily evaluated via * disassemble()). */ - std::string *cachedDisassembly; + mutable std::string *cachedDisassembly; /** * Internal function to generate disassembly string. */ - virtual std::string generateDisassembly(Addr pc, - const SymbolTable *symtab) = 0; + virtual std::string + generateDisassembly(Addr pc, const SymbolTable *symtab) const = 0; /// Constructor. StaticInst(const char *_mnemonic, MachInst _machInst, OpClass __opClass) @@ -311,23 +311,27 @@ class StaticInst : public StaticInstBase /** * Execute this instruction under SimpleCPU model. */ - virtual Fault execute(SimpleCPU *xc, Trace::InstRecord *traceData) = 0; + virtual Fault execute(SimpleCPU *xc, + Trace::InstRecord *traceData) const = 0; - /** + /** * Execute this instruction under InorderCPU model. */ - virtual Fault execute(InorderCPU *xc, Trace::InstRecord *traceData) = 0; + virtual Fault execute(InorderCPU *xc, + Trace::InstRecord *traceData) const = 0; /** * Execute this instruction under FastCPU model. */ - virtual Fault execute(FastCPU *xc, Trace::InstRecord *traceData) = 0; + virtual Fault execute(FastCPU *xc, + Trace::InstRecord *traceData) const = 0; /** * Execute this instruction under detailed FullCPU model. */ - virtual Fault execute(DynInst *xc, Trace::InstRecord *traceData) = 0; + virtual Fault execute(DynInst *xc, + Trace::InstRecord *traceData) const = 0; /** * Return the target address for a PC-relative branch. @@ -357,7 +361,7 @@ class StaticInst : public StaticInstBase * Return true if the instruction is a control transfer, and if so, * return the target address as well. */ - bool hasBranchTarget(Addr pc, ExecContext *xc, Addr &tgt); + bool hasBranchTarget(Addr pc, ExecContext *xc, Addr &tgt) const; /** * Return string representation of disassembled instruction. @@ -367,7 +371,7 @@ class StaticInst : public StaticInstBase * should not be cached, this function should be overridden directly. */ virtual const std::string &disassemble(Addr pc, - const SymbolTable *symtab = 0) + const SymbolTable *symtab = 0) const { if (!cachedDisassembly) cachedDisassembly = @@ -456,7 +460,7 @@ class StaticInstPtr : public RefCountingPtr<StaticInst<ISA> > /// Convert to pointer to StaticInstBase class. operator const StaticInstBasePtr() { - return get(); + return this->get(); } }; diff --git a/cpu/trace/opt_cpu.cc b/cpu/trace/opt_cpu.cc index 77211e382..15b53d46e 100644 --- a/cpu/trace/opt_cpu.cc +++ b/cpu/trace/opt_cpu.cc @@ -48,7 +48,7 @@ OptCPU::OptCPU(const string &name, int block_size, int cache_size, int _assoc) - : BaseCPU(name, 1, true), tickEvent(this), trace(_trace), + : SimObject(name), tickEvent(this), trace(_trace), numBlks(cache_size/block_size), assoc(_assoc), numSets(numBlks/assoc), setMask(numSets - 1) { diff --git a/cpu/trace/opt_cpu.hh b/cpu/trace/opt_cpu.hh index 847147b3c..7f6aa3e18 100644 --- a/cpu/trace/opt_cpu.hh +++ b/cpu/trace/opt_cpu.hh @@ -32,14 +32,14 @@ * trace to access a fully associative cache with optimal replacement. */ -#ifndef __OPT_CPU_HH__ -#define __OPT_CPU_HH__ +#ifndef __CPU_TRACE_OPT_CPU_HH__ +#define __CPU_TRACE_OPT_CPU_HH__ #include <vector> -#include "cpu/base_cpu.hh" #include "mem/mem_req.hh" // for MemReqPtr #include "sim/eventq.hh" // for Event +#include "sim/sim_object.hh" // Forward Declaration class MemTraceReader; @@ -47,8 +47,9 @@ class MemTraceReader; /** * A CPU object to simulate a fully-associative cache with optimal replacement. */ -class OptCPU : public BaseCPU +class OptCPU : public SimObject { + private: typedef int RefIndex; typedef std::vector<RefIndex> L3Table; @@ -219,4 +220,4 @@ class OptCPU : public BaseCPU void tick(); }; -#endif +#endif // __CPU_TRACE_OPT_CPU_HH__ diff --git a/cpu/trace/trace_cpu.cc b/cpu/trace/trace_cpu.cc index f1160337a..1902d0be4 100644 --- a/cpu/trace/trace_cpu.cc +++ b/cpu/trace/trace_cpu.cc @@ -47,7 +47,7 @@ TraceCPU::TraceCPU(const string &name, MemInterface *icache_interface, MemInterface *dcache_interface, MemTraceReader *data_trace) - : BaseCPU(name, 4, true), icacheInterface(icache_interface), + : SimObject(name), icacheInterface(icache_interface), dcacheInterface(dcache_interface), dataTrace(data_trace), outstandingRequests(0), tickEvent(this) { diff --git a/cpu/trace/trace_cpu.hh b/cpu/trace/trace_cpu.hh index 1711646a8..cdac4bb4f 100644 --- a/cpu/trace/trace_cpu.hh +++ b/cpu/trace/trace_cpu.hh @@ -32,14 +32,14 @@ * provided memory hierarchy. */ -#ifndef __TRACE_CPU_HH__ -#define __TRACE_CPU_HH__ +#ifndef __CPU_TRACE_TRACE_CPU_HH__ +#define __CPU_TRACE_TRACE_CPU_HH__ #include <string> -#include "cpu/base_cpu.hh" #include "mem/mem_req.hh" // for MemReqPtr #include "sim/eventq.hh" // for Event +#include "sim/sim_object.hh" // Forward declaration. class MemInterface; @@ -48,8 +48,9 @@ class MemTraceReader; /** * A cpu object for running memory traces through a memory hierarchy. */ -class TraceCPU : public BaseCPU +class TraceCPU : public SimObject { + private: /** Interface for instruction trace requests, if any. */ MemInterface *icacheInterface; /** Interface for data trace requests, if any. */ @@ -133,5 +134,5 @@ class TraceCompleteEvent : public Event virtual const char *description(); }; -#endif //__TRACE_CPU_HH__ +#endif // __CPU_TRACE_TRACE_CPU_HH__ |