diff options
Diffstat (limited to 'src')
33 files changed, 206 insertions, 156 deletions
diff --git a/src/SConscript b/src/SConscript index 0d0cb2486..9825cafe7 100644 --- a/src/SConscript +++ b/src/SConscript @@ -89,7 +89,6 @@ base_sources = Split(''' cpu/pc_event.cc cpu/quiesce_event.cc cpu/static_inst.cc - cpu/sampler/sampler.cc cpu/simple_thread.cc cpu/thread_state.cc diff --git a/src/base/traceflags.py b/src/base/traceflags.py index 327ce6075..27c24107c 100644 --- a/src/base/traceflags.py +++ b/src/base/traceflags.py @@ -48,8 +48,10 @@ ccfilename = sys.argv[1] + '.cc' # To define a new flag, simply add it to this list. # baseFlags = [ + 'Activity', 'AlphaConsole', 'BADADDR', + 'BE', 'BPredRAS', 'Bus', 'BusAddrRanges', @@ -84,6 +86,7 @@ baseFlags = [ 'EthernetPIO', 'EthernetSM', 'Event', + 'FE', 'Fault', 'Fetch', 'Flow', @@ -97,6 +100,7 @@ baseFlags = [ 'GDBSend', 'GDBWrite', 'HWPrefetch', + 'IBE', 'IEW', 'IIC', 'IICMore', @@ -115,13 +119,8 @@ baseFlags = [ 'MSHR', 'Mbox', 'MemDepUnit', - 'BaseCPU' 'O3CPU', 'OzoneCPU', - 'FE', - 'IBE', - 'BE', - 'O3CPU', 'OzoneLSQ', 'PCEvent', 'PCIA', @@ -135,6 +134,7 @@ baseFlags = [ 'RenameMap', 'SQL', 'Sampler', + 'Scoreboard', 'ScsiCtrl', 'ScsiDisk', 'ScsiNone', @@ -158,8 +158,6 @@ baseFlags = [ 'Uart', 'VtoPhys', 'WriteBarrier', - 'Activity', - 'Scoreboard', 'Writeback', ] @@ -178,7 +176,7 @@ compoundFlagMap = { 'EthernetAll' : [ 'Ethernet', 'EthernetPIO', 'EthernetDMA', 'EthernetData' , 'EthernetDesc', 'EthernetIntr', 'EthernetSM', 'EthernetCksum' ], 'EthernetNoData' : [ 'Ethernet', 'EthernetPIO', 'EthernetDesc', 'EthernetIntr', 'EthernetSM', 'EthernetCksum' ], 'IdeAll' : [ 'IdeCtrl', 'IdeDisk' ], - 'O3CPUAll' : [ 'Fetch', 'Decode', 'Rename', 'IEW', 'Commit', 'IQ', 'ROB', 'FreeList', 'RenameMap', 'LSQ', 'LSQUnit', 'StoreSet', 'MemDepUnit', 'DynInst', 'O3CPU', 'Activity','Scoreboard','Writeback'], + 'O3CPUAll' : [ 'Fetch', 'Decode', 'Rename', 'IEW', 'Commit', 'IQ', 'ROB', 'FreeList', 'RenameMap', 'LSQ', 'LSQUnit', 'StoreSet', 'MemDepUnit', 'DynInst', 'FullCPU', 'O3CPU', 'Activity','Scoreboard','Writeback'], 'OzoneCPUAll' : [ 'BE', 'FE', 'IBE', 'OzoneLSQ', 'OzoneCPU'] } diff --git a/src/cpu/SConscript b/src/cpu/SConscript index 3dcc2f1ec..1d0a015b0 100644 --- a/src/cpu/SConscript +++ b/src/cpu/SConscript @@ -130,13 +130,13 @@ if need_simple_base: if 'FastCPU' in env['CPU_MODELS']: sources += Split('fast/cpu.cc') +need_bp_unit = False if 'O3CPU' in env['CPU_MODELS']: + need_bp_unit = True sources += SConscript('o3/SConscript', exports = 'env') sources += Split(''' - o3/2bit_local_pred.cc o3/base_dyn_inst.cc o3/bpred_unit.cc - o3/btb.cc o3/commit.cc o3/decode.cc o3/fetch.cc @@ -148,18 +148,17 @@ if 'O3CPU' in env['CPU_MODELS']: o3/lsq_unit.cc o3/lsq.cc o3/mem_dep_unit.cc - o3/ras.cc o3/rename.cc o3/rename_map.cc o3/rob.cc o3/scoreboard.cc o3/store_set.cc - o3/tournament_pred.cc ''') if env['USE_CHECKER']: sources += Split('o3/checker_builder.cc') if 'OzoneCPU' in env['CPU_MODELS']: + need_bp_unit = True sources += Split(''' ozone/base_dyn_inst.cc ozone/bpred_unit.cc @@ -174,6 +173,14 @@ if 'OzoneCPU' in env['CPU_MODELS']: if env['USE_CHECKER']: sources += Split('ozone/checker_builder.cc') +if need_bp_unit: + sources += Split(''' + o3/2bit_local_pred.cc + o3/btb.cc + o3/ras.cc + o3/tournament_pred.cc + ''') + if env['USE_CHECKER']: sources += Split('checker/cpu.cc') checker_supports = False diff --git a/src/cpu/base.cc b/src/cpu/base.cc index 40cec416b..0b9c80591 100644 --- a/src/cpu/base.cc +++ b/src/cpu/base.cc @@ -41,7 +41,6 @@ #include "cpu/cpuevent.hh" #include "cpu/thread_context.hh" #include "cpu/profile.hh" -#include "cpu/sampler/sampler.hh" #include "sim/param.hh" #include "sim/process.hh" #include "sim/sim_events.hh" diff --git a/src/cpu/base.hh b/src/cpu/base.hh index 51f3bb905..5256a411f 100644 --- a/src/cpu/base.hh +++ b/src/cpu/base.hh @@ -36,7 +36,6 @@ #include "base/statistics.hh" #include "config/full_system.hh" -#include "cpu/sampler/sampler.hh" #include "sim/eventq.hh" #include "sim/sim_object.hh" #include "arch/isa_traits.hh" diff --git a/src/cpu/checker/cpu.hh b/src/cpu/checker/cpu.hh index 785387e60..b520e1be0 100644 --- a/src/cpu/checker/cpu.hh +++ b/src/cpu/checker/cpu.hh @@ -66,7 +66,6 @@ class ThreadContext; class MemInterface; class Checkpoint; class Request; -class Sampler; /** * CheckerCPU class. Dynamically verifies instructions as they are @@ -374,7 +373,7 @@ class Checker : public CheckerCPU : CheckerCPU(p) { } - void switchOut(Sampler *s); + void switchOut(); void takeOverFrom(BaseCPU *oldCPU); void verify(DynInstPtr &inst); diff --git a/src/cpu/checker/cpu_impl.hh b/src/cpu/checker/cpu_impl.hh index 137e1c46d..81f97726c 100644 --- a/src/cpu/checker/cpu_impl.hh +++ b/src/cpu/checker/cpu_impl.hh @@ -236,9 +236,7 @@ Checker<DynInstPtr>::verify(DynInstPtr &completed_inst) willChangePC = true; newPC = thread->readPC(); DPRINTF(Checker, "Fault, PC is now %#x\n", newPC); -#else // !FULL_SYSTEM - fatal("fault (%d) detected @ PC 0x%08p", fault, thread->readPC()); -#endif // FULL_SYSTEM +#endif } else { #if THE_ISA != MIPS_ISA // go to the next instruction @@ -295,7 +293,7 @@ Checker<DynInstPtr>::verify(DynInstPtr &completed_inst) template <class DynInstPtr> void -Checker<DynInstPtr>::switchOut(Sampler *s) +Checker<DynInstPtr>::switchOut() { instList.clear(); } diff --git a/src/cpu/o3/alpha/cpu_builder.cc b/src/cpu/o3/alpha/cpu_builder.cc index 490305cbf..d61eee4b1 100644 --- a/src/cpu/o3/alpha/cpu_builder.cc +++ b/src/cpu/o3/alpha/cpu_builder.cc @@ -91,7 +91,10 @@ Param<unsigned> renameWidth; Param<unsigned> commitToIEWDelay; Param<unsigned> renameToIEWDelay; Param<unsigned> issueToExecuteDelay; +Param<unsigned> dispatchWidth; Param<unsigned> issueWidth; +Param<unsigned> wbWidth; +Param<unsigned> wbDepth; SimObjectParam<FUPool *> fuPool; Param<unsigned> iewToCommitDelay; @@ -207,7 +210,10 @@ BEGIN_INIT_SIM_OBJECT_PARAMS(DerivO3CPU) "Issue/Execute/Writeback delay"), INIT_PARAM(issueToExecuteDelay, "Issue to execute delay (internal" "to the IEW stage)"), + INIT_PARAM(dispatchWidth, "Dispatch width"), INIT_PARAM(issueWidth, "Issue width"), + INIT_PARAM(wbWidth, "Writeback width"), + INIT_PARAM(wbDepth, "Writeback depth (number of cycles it can buffer)"), INIT_PARAM_DFLT(fuPool, "Functional unit pool", NULL), INIT_PARAM(iewToCommitDelay, "Issue/Execute/Writeback to commit " @@ -333,7 +339,10 @@ CREATE_SIM_OBJECT(DerivO3CPU) params->commitToIEWDelay = commitToIEWDelay; params->renameToIEWDelay = renameToIEWDelay; params->issueToExecuteDelay = issueToExecuteDelay; + params->dispatchWidth = dispatchWidth; params->issueWidth = issueWidth; + params->wbWidth = wbWidth; + params->wbDepth = wbDepth; params->fuPool = fuPool; params->iewToCommitDelay = iewToCommitDelay; diff --git a/src/cpu/o3/alpha/params.hh b/src/cpu/o3/alpha/params.hh index b1f2a487d..c618cee08 100644 --- a/src/cpu/o3/alpha/params.hh +++ b/src/cpu/o3/alpha/params.hh @@ -54,16 +54,7 @@ class AlphaSimpleParams : public O3Params #if FULL_SYSTEM AlphaITB *itb; AlphaDTB *dtb; -#else - std::vector<Process *> workload; - Process *process; -#endif // FULL_SYSTEM - - MemObject *mem; - - BaseCPU *checker; - - unsigned decodeToFetchDelay; +#endif }; #endif // __CPU_O3_ALPHA_PARAMS_HH__ diff --git a/src/cpu/o3/cpu.cc b/src/cpu/o3/cpu.cc index 630d82cba..c88146fa6 100644 --- a/src/cpu/o3/cpu.cc +++ b/src/cpu/o3/cpu.cc @@ -709,9 +709,8 @@ FullO3CPU<Impl>::haltContext(int tid) template <class Impl> void -FullO3CPU<Impl>::switchOut(Sampler *_sampler) +FullO3CPU<Impl>::switchOut() { - sampler = _sampler; switchCount = 0; fetch.switchOut(); decode.switchOut(); @@ -740,12 +739,11 @@ FullO3CPU<Impl>::signalSwitched() #if USE_CHECKER if (checker) - checker->switchOut(sampler); + checker->switchOut(); #endif if (tickEvent.scheduled()) tickEvent.squash(); - sampler->signalSwitched(); _status = SwitchedOut; } assert(switchCount <= 5); diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh index 1cff6142d..bd0451601 100644 --- a/src/cpu/o3/cpu.hh +++ b/src/cpu/o3/cpu.hh @@ -271,7 +271,7 @@ class FullO3CPU : public BaseO3CPU virtual void syscall(int tid) { panic("Unimplemented!"); } /** Switches out this CPU. */ - void switchOut(Sampler *sampler); + void switchOut(); /** Signals to this CPU that a stage has completed switching out. */ void signalSwitched(); @@ -550,9 +550,6 @@ class FullO3CPU : public BaseO3CPU /** Pointer to memory. */ MemObject *mem; - /** Pointer to the sampler */ - Sampler *sampler; - /** Counter of how many stages have completed switching out. */ int switchCount; diff --git a/src/cpu/o3/fetch.hh b/src/cpu/o3/fetch.hh index 7fcd21b7d..848ebf39e 100644 --- a/src/cpu/o3/fetch.hh +++ b/src/cpu/o3/fetch.hh @@ -40,8 +40,6 @@ #include "mem/port.hh" #include "sim/eventq.hh" -class Sampler; - /** * DefaultFetch class handles both single threaded and SMT fetch. Its * width is specified by the parameters; each cycle it tries to fetch diff --git a/src/cpu/o3/iew.hh b/src/cpu/o3/iew.hh index 2af68d8fc..9627609c2 100644 --- a/src/cpu/o3/iew.hh +++ b/src/cpu/o3/iew.hh @@ -204,6 +204,45 @@ class DefaultIEW /** Returns if the LSQ has any stores to writeback. */ bool hasStoresToWB() { return ldstQueue.hasStoresToWB(); } + void incrWb(InstSeqNum &sn) + { + if (++wbOutstanding == wbMax) + ableToIssue = false; + DPRINTF(IEW, "wbOutstanding: %i\n", wbOutstanding); +#if DEBUG + wbList.insert(sn); +#endif + } + + void decrWb(InstSeqNum &sn) + { + if (wbOutstanding-- == wbMax) + ableToIssue = true; + DPRINTF(IEW, "wbOutstanding: %i\n", wbOutstanding); +#if DEBUG + assert(wbList.find(sn) != wbList.end()); + wbList.erase(sn); +#endif + } + +#if DEBUG + std::set<InstSeqNum> wbList; + + void dumpWb() + { + std::set<InstSeqNum>::iterator wb_it = wbList.begin(); + while (wb_it != wbList.end()) { + cprintf("[sn:%lli]\n", + (*wb_it)); + wb_it++; + } + } +#endif + + bool canIssue() { return ableToIssue; } + + bool ableToIssue; + private: /** Sends commit proper information for a squash due to a branch * mispredict. @@ -384,11 +423,8 @@ class DefaultIEW */ unsigned issueToExecuteDelay; - /** Width of issue's read path, in instructions. The read path is both - * the skid buffer and the rename instruction queue. - * Note to self: is this really different than issueWidth? - */ - unsigned issueReadWidth; + /** Width of dispatch, in instructions. */ + unsigned dispatchWidth; /** Width of issue, in instructions. */ unsigned issueWidth; @@ -403,6 +439,17 @@ class DefaultIEW */ unsigned wbCycle; + /** Number of instructions in flight that will writeback. */ + unsigned wbOutstanding; + + /** Writeback width. */ + unsigned wbWidth; + + /** Writeback width * writeback depth, where writeback depth is + * the number of cycles of writing back instructions that can be + * buffered. */ + unsigned wbMax; + /** Number of active threads. */ unsigned numThreads; diff --git a/src/cpu/o3/iew_impl.hh b/src/cpu/o3/iew_impl.hh index 8e6fd46a1..118038b65 100644 --- a/src/cpu/o3/iew_impl.hh +++ b/src/cpu/o3/iew_impl.hh @@ -50,8 +50,10 @@ DefaultIEW<Impl>::DefaultIEW(Params *params) commitToIEWDelay(params->commitToIEWDelay), renameToIEWDelay(params->renameToIEWDelay), issueToExecuteDelay(params->issueToExecuteDelay), - issueReadWidth(params->issueWidth), + dispatchWidth(params->dispatchWidth), issueWidth(params->issueWidth), + wbOutstanding(0), + wbWidth(params->wbWidth), numThreads(params->numberOfThreads), switchedOut(false) { @@ -74,8 +76,12 @@ DefaultIEW<Impl>::DefaultIEW(Params *params) fetchRedirect[i] = false; } + wbMax = wbWidth * params->wbDepth; + updateLSQNextCycle = false; + ableToIssue = true; + skidBufferMax = (3 * (renameToIEWDelay * params->renameWidth)) + issueWidth; } @@ -559,12 +565,12 @@ DefaultIEW<Impl>::instToCommit(DynInstPtr &inst) // free slot. while ((*iewQueue)[wbCycle].insts[wbNumInst]) { ++wbNumInst; - if (wbNumInst == issueWidth) { + if (wbNumInst == wbWidth) { ++wbCycle; wbNumInst = 0; } - assert(wbCycle < 5); + assert((wbCycle * wbWidth + wbNumInst) < wbMax); } // Add finished instruction to queue to commit. @@ -937,7 +943,7 @@ DefaultIEW<Impl>::dispatchInsts(unsigned tid) // Loop through the instructions, putting them in the instruction // queue. for ( ; dis_num_inst < insts_to_add && - dis_num_inst < issueReadWidth; + dis_num_inst < dispatchWidth; ++dis_num_inst) { inst = insts_to_dispatch.front(); @@ -1189,6 +1195,7 @@ DefaultIEW<Impl>::executeInsts() ++iewExecSquashedInsts; + decrWb(inst->seqNum); continue; } @@ -1351,6 +1358,8 @@ DefaultIEW<Impl>::writebackInsts() } writebackCount[tid]++; } + + decrWb(inst->seqNum); } } diff --git a/src/cpu/o3/inst_queue_impl.hh b/src/cpu/o3/inst_queue_impl.hh index b99bd0900..36e0842be 100644 --- a/src/cpu/o3/inst_queue_impl.hh +++ b/src/cpu/o3/inst_queue_impl.hh @@ -687,6 +687,7 @@ InstructionQueue<Impl>::scheduleReadyInsts() int total_issued = 0; while (total_issued < totalWidth && + iewStage->canIssue() && order_it != order_end_it) { OpClass op_class = (*order_it).queueType; @@ -784,6 +785,7 @@ InstructionQueue<Impl>::scheduleReadyInsts() listOrder.erase(order_it++); statIssuedInstType[tid][op_class]++; + iewStage->incrWb(issuing_inst->seqNum); } else { statFuBusy[op_class]++; fuBusy[tid]++; diff --git a/src/cpu/o3/lsq_unit_impl.hh b/src/cpu/o3/lsq_unit_impl.hh index 714acb2ef..bb3da7eec 100644 --- a/src/cpu/o3/lsq_unit_impl.hh +++ b/src/cpu/o3/lsq_unit_impl.hh @@ -77,6 +77,7 @@ LSQUnit<Impl>::completeDataAccess(PacketPtr pkt) //iewStage->ldstQueue.removeMSHR(inst->threadNumber,inst->seqNum); if (isSwitchedOut() || inst->isSquashed()) { + iewStage->decrWb(inst->seqNum); delete state; delete pkt; return; diff --git a/src/cpu/o3/params.hh b/src/cpu/o3/params.hh index 69a1bb937..ed53fa97a 100755 --- a/src/cpu/o3/params.hh +++ b/src/cpu/o3/params.hh @@ -47,6 +47,18 @@ class O3Params : public BaseO3CPU::Params unsigned activity; // + // Pointers to key objects + // +#if !FULL_SYSTEM + std::vector<Process *> workload; + Process *process; +#endif // FULL_SYSTEM + + MemObject *mem; + + BaseCPU *checker; + + // // Caches // // MemInterface *icacheInterface; @@ -86,7 +98,10 @@ class O3Params : public BaseO3CPU::Params unsigned commitToIEWDelay; unsigned renameToIEWDelay; unsigned issueToExecuteDelay; + unsigned dispatchWidth; unsigned issueWidth; + unsigned wbWidth; + unsigned wbDepth; FUPool *fuPool; // diff --git a/src/cpu/ozone/cpu.hh b/src/cpu/ozone/cpu.hh index f726ac99b..8993781ea 100644 --- a/src/cpu/ozone/cpu.hh +++ b/src/cpu/ozone/cpu.hh @@ -55,7 +55,6 @@ class AlphaDTB; class PhysicalMemory; class MemoryController; -class Sampler; class RemoteGDB; class GDBListener; @@ -356,12 +355,10 @@ class OzoneCPU : public BaseCPU int cpuId; - void switchOut(Sampler *sampler); + void switchOut(); void signalSwitched(); void takeOverFrom(BaseCPU *oldCPU); - Sampler *sampler; - int switchCount; #if FULL_SYSTEM diff --git a/src/cpu/ozone/cpu_impl.hh b/src/cpu/ozone/cpu_impl.hh index 2cdc8a3da..ccb1c8418 100644 --- a/src/cpu/ozone/cpu_impl.hh +++ b/src/cpu/ozone/cpu_impl.hh @@ -244,9 +244,8 @@ OzoneCPU<Impl>::~OzoneCPU() template <class Impl> void -OzoneCPU<Impl>::switchOut(Sampler *_sampler) +OzoneCPU<Impl>::switchOut() { - sampler = _sampler; switchCount = 0; // Front end needs state from back end, so switch out the back end first. backEnd->switchOut(); @@ -262,13 +261,12 @@ OzoneCPU<Impl>::signalSwitched() frontEnd->doSwitchOut(); #if USE_CHECKER if (checker) - checker->switchOut(sampler); + checker->switchOut(); #endif _status = SwitchedOut; if (tickEvent.scheduled()) tickEvent.squash(); - sampler->signalSwitched(); } assert(switchCount <= 2); } diff --git a/src/cpu/simple/base.cc b/src/cpu/simple/base.cc index db5dd2acf..a50541189 100644 --- a/src/cpu/simple/base.cc +++ b/src/cpu/simple/base.cc @@ -41,7 +41,6 @@ #include "cpu/base.hh" #include "cpu/exetrace.hh" #include "cpu/profile.hh" -#include "cpu/sampler/sampler.hh" #include "cpu/simple/base.hh" #include "cpu/simple_thread.hh" #include "cpu/smt.hh" diff --git a/src/cpu/simple/base.hh b/src/cpu/simple/base.hh index 39bc86050..57cfa3c2c 100644 --- a/src/cpu/simple/base.hh +++ b/src/cpu/simple/base.hh @@ -38,7 +38,6 @@ #include "cpu/base.hh" #include "cpu/simple_thread.hh" #include "cpu/pc_event.hh" -#include "cpu/sampler/sampler.hh" #include "cpu/static_inst.hh" #include "mem/packet.hh" #include "mem/port.hh" @@ -128,11 +127,6 @@ class BaseSimpleCPU : public BaseCPU // Static data storage TheISA::IntReg dataReg; - // Pointer to the sampler that is telling us to switchover. - // Used to signal the completion of the pipe drain and schedule - // the next switchover - Sampler *sampler; - StaticInstPtr curStaticInst; void checkForInterrupts(); diff --git a/src/cpu/simple/timing.cc b/src/cpu/simple/timing.cc index 0729f9489..ad04c8d3b 100644 --- a/src/cpu/simple/timing.cc +++ b/src/cpu/simple/timing.cc @@ -88,7 +88,7 @@ TimingSimpleCPU::TimingSimpleCPU(Params *p) { _status = Idle; ifetch_pkt = dcache_pkt = NULL; - quiesceEvent = NULL; + drainEvent = NULL; state = SimObject::Timing; } @@ -112,17 +112,16 @@ TimingSimpleCPU::unserialize(Checkpoint *cp, const string §ion) } bool -TimingSimpleCPU::quiesce(Event *quiesce_event) +TimingSimpleCPU::drain(Event *drain_event) { - // TimingSimpleCPU is ready to quiesce if it's not waiting for + // TimingSimpleCPU is ready to drain if it's not waiting for // an access to complete. if (status() == Idle || status() == Running || status() == SwitchedOut) { - DPRINTF(Config, "Ready to quiesce\n"); + changeState(SimObject::DrainedTiming); return false; } else { - DPRINTF(Config, "Waiting to quiesce\n"); - changeState(SimObject::Quiescing); - quiesceEvent = quiesce_event; + changeState(SimObject::Draining); + drainEvent = drain_event; return true; } } @@ -207,7 +206,7 @@ TimingSimpleCPU::read(Addr addr, T &data, unsigned flags) { // need to fill in CPU & thread IDs here Request *data_read_req = new Request(); - + data_read_req->setThreadContext(0,0); //Need CPU/Thread IDS HERE data_read_req->setVirt(0, addr, sizeof(T), flags, thread->readPC()); if (traceData) { @@ -288,6 +287,7 @@ TimingSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res) { // need to fill in CPU & thread IDs here Request *data_write_req = new Request(); + data_write_req->setThreadContext(0,0); //Need CPU/Thread IDS HERE data_write_req->setVirt(0, addr, sizeof(T), flags, thread->readPC()); // translate to physical address @@ -371,6 +371,7 @@ TimingSimpleCPU::fetch() // need to fill in CPU & thread IDs here Request *ifetch_req = new Request(); + ifetch_req->setThreadContext(0,0); //Need CPU/Thread IDS HERE Fault fault = setupFetchRequest(ifetch_req); ifetch_pkt = new Packet(ifetch_req, Packet::ReadReq, Packet::Broadcast); @@ -420,8 +421,8 @@ TimingSimpleCPU::completeIfetch(Packet *pkt) delete pkt->req; delete pkt; - if (getState() == SimObject::Quiescing) { - completeQuiesce(); + if (getState() == SimObject::Draining) { + completeDrain(); return; } @@ -477,8 +478,8 @@ TimingSimpleCPU::completeDataAccess(Packet *pkt) assert(_status == DcacheWaitResponse); _status = Running; - if (getState() == SimObject::Quiescing) { - completeQuiesce(); + if (getState() == SimObject::Draining) { + completeDrain(); delete pkt->req; delete pkt; @@ -497,11 +498,11 @@ TimingSimpleCPU::completeDataAccess(Packet *pkt) void -TimingSimpleCPU::completeQuiesce() +TimingSimpleCPU::completeDrain() { - DPRINTF(Config, "Done quiescing\n"); - changeState(SimObject::QuiescedTiming); - quiesceEvent->process(); + DPRINTF(Config, "Done draining\n"); + changeState(SimObject::DrainedTiming); + drainEvent->process(); } bool diff --git a/src/cpu/simple/timing.hh b/src/cpu/simple/timing.hh index d91144e4a..c360e553e 100644 --- a/src/cpu/simple/timing.hh +++ b/src/cpu/simple/timing.hh @@ -64,7 +64,7 @@ class TimingSimpleCPU : public BaseSimpleCPU Status status() const { return _status; } - Event *quiesceEvent; + Event *drainEvent; private: @@ -133,7 +133,7 @@ class TimingSimpleCPU : public BaseSimpleCPU virtual void serialize(std::ostream &os); virtual void unserialize(Checkpoint *cp, const std::string §ion); - virtual bool quiesce(Event *quiesce_event); + virtual bool drain(Event *drain_event); virtual void resume(); virtual void setMemoryMode(State new_mode); @@ -154,7 +154,7 @@ class TimingSimpleCPU : public BaseSimpleCPU void completeDataAccess(Packet *); void advanceInst(Fault fault); private: - void completeQuiesce(); + void completeDrain(); }; #endif // __CPU_SIMPLE_TIMING_HH__ diff --git a/src/mem/cache/miss/mshr.cc b/src/mem/cache/miss/mshr.cc index 05a2fe1c5..1a85d3018 100644 --- a/src/mem/cache/miss/mshr.cc +++ b/src/mem/cache/miss/mshr.cc @@ -57,26 +57,26 @@ void MSHR::allocate(Packet::Command cmd, Addr _addr, int _asid, int size, Packet * &target) { - assert("NEED TO FIX YET\n" && 0); -#if 0 - assert(targets.empty()); - addr = _addr; - asid = _asid; - - pkt = new Packet(); // allocate new memory request - pkt->addr = addr; //picked physical address for now - pkt->cmd = cmd; - pkt->size = size; - pkt->data = new uint8_t[size]; - pkt->senderState = this; - //Set the time here for latency calculations - pkt->time = curTick; - - if (target) { - pkt->req = target->req; + if (target) + { + //Have a request, just use it + pkt = new Packet(target->req, cmd, Packet::Broadcast, size); + pkt->time = curTick; + pkt->allocate(); + pkt->senderState = (Packet::SenderState *)this; allocateTarget(target); } -#endif + else + { + //need a request first + Request * req = new Request(); + req->setPhys(addr, size, 0); + //Thread context?? + pkt = new Packet(req, cmd, Packet::Broadcast, size); + pkt->time = curTick; + pkt->allocate(); + pkt->senderState = (Packet::SenderState *)this; + } } // Since we aren't sure if data is being used, don't copy here. diff --git a/src/python/m5/__init__.py b/src/python/m5/__init__.py index 828165d15..579785a46 100644 --- a/src/python/m5/__init__.py +++ b/src/python/m5/__init__.py @@ -213,14 +213,14 @@ atexit.register(cc_main.doExitCleanup) # matter since most scripts will probably 'from m5.objects import *'. import objects -def doQuiesce(root): - quiesce = cc_main.createCountedQuiesce() - unready_objects = root.startQuiesce(quiesce, True) - # If we've got some objects that can't quiesce immediately, then simulate +def doDrain(root): + drain_event = cc_main.createCountedDrain() + unready_objects = root.startDrain(drain_event, True) + # If we've got some objects that can't drain immediately, then simulate if unready_objects > 0: - quiesce.setCount(unready_objects) + drain_event.setCount(unready_objects) simulate() - cc_main.cleanupCountedQuiesce(quiesce) + cc_main.cleanupCountedDrain(drain_event) def resume(root): root.resume() @@ -228,7 +228,7 @@ def resume(root): def checkpoint(root): if not isinstance(root, objects.Root): raise TypeError, "Object is not a root object. Checkpoint must be called on a root object." - doQuiesce(root) + doDrain(root) print "Writing checkpoint" cc_main.serializeAll() resume(root) @@ -241,7 +241,7 @@ def changeToAtomic(system): if not isinstance(system, objects.Root) and not isinstance(system, System): raise TypeError, "Object is not a root or system object. Checkpoint must be " "called on a root object." - doQuiesce(system) + doDrain(system) print "Changing memory mode to atomic" system.changeTiming(cc_main.SimObject.Atomic) resume(system) @@ -250,7 +250,7 @@ def changeToTiming(system): if not isinstance(system, objects.Root) and not isinstance(system, System): raise TypeError, "Object is not a root or system object. Checkpoint must be " "called on a root object." - doQuiesce(system) + doDrain(system) print "Changing memory mode to timing" system.changeTiming(cc_main.SimObject.Timing) resume(system) @@ -271,16 +271,16 @@ def switchCpus(cpuList): if not isinstance(cpu, objects.BaseCPU): raise TypeError, "%s is not of type BaseCPU", cpu - # Quiesce all of the individual CPUs - quiesce = cc_main.createCountedQuiesce() + # Drain all of the individual CPUs + drain_event = cc_main.createCountedDrain() unready_cpus = 0 for old_cpu in old_cpus: - unready_cpus += old_cpu.startQuiesce(quiesce, False) - # If we've got some objects that can't quiesce immediately, then simulate + unready_cpus += old_cpu.startDrain(drain_event, False) + # If we've got some objects that can't drain immediately, then simulate if unready_cpus > 0: - quiesce.setCount(unready_cpus) + drain_event.setCount(unready_cpus) simulate() - cc_main.cleanupCountedQuiesce(quiesce) + cc_main.cleanupCountedDrain(drain_event) # Now all of the CPUs are ready to be switched out for old_cpu in old_cpus: old_cpu._ccObject.switchOut() diff --git a/src/python/m5/config.py b/src/python/m5/config.py index 6f2873d40..cffe06984 100644 --- a/src/python/m5/config.py +++ b/src/python/m5/config.py @@ -543,15 +543,15 @@ class SimObject(object): for child in self._children.itervalues(): child.connectPorts() - def startQuiesce(self, quiesce_event, recursive): + def startDrain(self, drain_event, recursive): count = 0 # ParamContexts don't serialize if isinstance(self, SimObject) and not isinstance(self, ParamContext): - if self._ccObject.quiesce(quiesce_event): + if self._ccObject.drain(drain_event): count = 1 if recursive: for child in self._children.itervalues(): - count += child.startQuiesce(quiesce_event, True) + count += child.startDrain(drain_event, True) return count def resume(self): diff --git a/src/python/m5/objects/O3CPU.py b/src/python/m5/objects/O3CPU.py index 4ecfa8fbd..9ccbdcf53 100644 --- a/src/python/m5/objects/O3CPU.py +++ b/src/python/m5/objects/O3CPU.py @@ -37,12 +37,10 @@ class DerivO3CPU(BaseCPU): "Issue/Execute/Writeback delay") issueToExecuteDelay = Param.Unsigned("Issue to execute delay (internal " "to the IEW stage)") + dispatchWidth = Param.Unsigned("Dispatch width") issueWidth = Param.Unsigned("Issue width") - executeWidth = Param.Unsigned("Execute width") - executeIntWidth = Param.Unsigned("Integer execute width") - executeFloatWidth = Param.Unsigned("Floating point execute width") - executeBranchWidth = Param.Unsigned("Branch execute width") - executeMemoryWidth = Param.Unsigned("Memory execute width") + wbWidth = Param.Unsigned("Writeback width") + wbDepth = Param.Unsigned("Writeback depth") fuPool = Param.FUPool(NULL, "Functional Unit pool") iewToCommitDelay = Param.Unsigned("Issue/Execute/Writeback to commit " diff --git a/src/sim/main.cc b/src/sim/main.cc index 3eb7fa95d..e96a44930 100644 --- a/src/sim/main.cc +++ b/src/sim/main.cc @@ -523,19 +523,19 @@ simulate(Tick num_cycles = -1) } Event * -createCountedQuiesce() +createCountedDrain() { - return new CountedQuiesceEvent(); + return new CountedDrainEvent(); } void -cleanupCountedQuiesce(Event *counted_quiesce) +cleanupCountedDrain(Event *counted_drain) { - CountedQuiesceEvent *event = - dynamic_cast<CountedQuiesceEvent *>(counted_quiesce); + CountedDrainEvent *event = + dynamic_cast<CountedDrainEvent *>(counted_drain); if (event == NULL) { - fatal("Called cleanupCountedQuiesce() on an event that was not " - "a CountedQuiesceEvent."); + fatal("Called cleanupCountedDrain() on an event that was not " + "a CountedDrainEvent."); } assert(event->getCount() == 0); delete event; diff --git a/src/sim/pseudo_inst.cc b/src/sim/pseudo_inst.cc index b2854e491..869805f5c 100644 --- a/src/sim/pseudo_inst.cc +++ b/src/sim/pseudo_inst.cc @@ -52,8 +52,6 @@ using namespace std; -extern Sampler *SampCPU; - using namespace Stats; using namespace TheISA; @@ -209,6 +207,7 @@ namespace AlphaPseudo { if (!doCheckpointInsts) return; + exitSimLoop("checkpoint"); } uint64_t @@ -280,7 +279,6 @@ namespace AlphaPseudo void switchcpu(ThreadContext *tc) { - if (SampCPU) - SampCPU->switchCPUs(); + exitSimLoop("switchcpu"); } } diff --git a/src/sim/sim_events.cc b/src/sim/sim_events.cc index 97f7ae03c..d9e8bdeaa 100644 --- a/src/sim/sim_events.cc +++ b/src/sim/sim_events.cc @@ -79,10 +79,10 @@ exitSimLoop(const std::string &message, int exit_code) } void -CountedQuiesceEvent::process() +CountedDrainEvent::process() { if (--count == 0) { - exitSimLoop("Finished quiesce"); + exitSimLoop("Finished drain"); } } diff --git a/src/sim/sim_events.hh b/src/sim/sim_events.hh index 50368f258..3c4a9dd05 100644 --- a/src/sim/sim_events.hh +++ b/src/sim/sim_events.hh @@ -67,13 +67,13 @@ class SimLoopExitEvent : public Event virtual const char *description(); }; -class CountedQuiesceEvent : public SimLoopExitEvent +class CountedDrainEvent : public SimLoopExitEvent { private: - // Count down to quiescing + // Count of how many objects have not yet drained int count; public: - CountedQuiesceEvent() + CountedDrainEvent() : count(0) { } void process(); diff --git a/src/sim/sim_object.cc b/src/sim/sim_object.cc index 551555b25..655bdcf4e 100644 --- a/src/sim/sim_object.cc +++ b/src/sim/sim_object.cc @@ -37,7 +37,6 @@ #include "base/misc.hh" #include "base/trace.hh" #include "base/stats/events.hh" -#include "base/serializer.hh" #include "sim/host.hh" #include "sim/sim_object.hh" #include "sim/stats.hh" @@ -271,22 +270,22 @@ SimObject::recordEvent(const std::string &stat) } bool -SimObject::quiesce(Event *quiesce_event) +SimObject::drain(Event *drain_event) { - if (state != QuiescedAtomic && state != Atomic) { - panic("Must implement your own quiesce function if it is to be used " + if (state != DrainedAtomic && state != Atomic) { + panic("Must implement your own drain function if it is to be used " "in timing mode!"); } - state = QuiescedAtomic; + state = DrainedAtomic; return false; } void SimObject::resume() { - if (state == QuiescedAtomic) { + if (state == DrainedAtomic) { state = Atomic; - } else if (state == QuiescedTiming) { + } else if (state == DrainedTiming) { state = Timing; } } @@ -295,10 +294,10 @@ void SimObject::setMemoryMode(State new_mode) { assert(new_mode == Timing || new_mode == Atomic); - if (state == QuiescedAtomic && new_mode == Timing) { - state = QuiescedTiming; - } else if (state == QuiescedTiming && new_mode == Atomic) { - state = QuiescedAtomic; + if (state == DrainedAtomic && new_mode == Timing) { + state = DrainedTiming; + } else if (state == DrainedTiming && new_mode == Atomic) { + state = DrainedAtomic; } else { state = new_mode; } diff --git a/src/sim/sim_object.hh b/src/sim/sim_object.hh index e0b21782f..4833192d6 100644 --- a/src/sim/sim_object.hh +++ b/src/sim/sim_object.hh @@ -62,9 +62,9 @@ class SimObject : public Serializable, protected StartupCallback enum State { Atomic, Timing, - Quiescing, - QuiescedAtomic, - QuiescedTiming + Draining, + DrainedAtomic, + DrainedTiming }; protected: @@ -117,7 +117,7 @@ class SimObject : public Serializable, protected StartupCallback // Methods to drain objects in order to take checkpoints // Or switch from timing -> atomic memory model // Quiesce returns true if the SimObject cannot quiesce immediately. - virtual bool quiesce(Event *quiesce_event); + virtual bool drain(Event *drain_event); virtual void resume(); virtual void setMemoryMode(State new_mode); virtual void switchOut(); |