diff options
author | Nathan Binkert <nate@binkert.org> | 2009-05-26 09:23:13 -0700 |
---|---|---|
committer | Nathan Binkert <nate@binkert.org> | 2009-05-26 09:23:13 -0700 |
commit | 47877cf2dbd6ee2f1cf9b2c609d37b0589e876ca (patch) | |
tree | 6beb00dfe7e31b9bf82f7aba4710b0c487b6543f /src/cpu/o3 | |
parent | d93392df28fc6c9a5c70fb6252a12afdc72d9344 (diff) | |
download | gem5-47877cf2dbd6ee2f1cf9b2c609d37b0589e876ca.tar.xz |
types: add a type for thread IDs and try to use it everywhere
Diffstat (limited to 'src/cpu/o3')
32 files changed, 869 insertions, 854 deletions
diff --git a/src/cpu/o3/bpred_unit.hh b/src/cpu/o3/bpred_unit.hh index 4875c03d8..49cb4233b 100644 --- a/src/cpu/o3/bpred_unit.hh +++ b/src/cpu/o3/bpred_unit.hh @@ -88,7 +88,7 @@ class BPredUnit * @param tid The thread id. * @return Returns if the branch is taken or not. */ - bool predict(DynInstPtr &inst, Addr &PC, unsigned tid); + bool predict(DynInstPtr &inst, Addr &PC, ThreadID tid); // @todo: Rename this function. void BPUncond(void * &bp_history); @@ -99,7 +99,7 @@ class BPredUnit * @param done_sn The sequence number to commit any older updates up until. * @param tid The thread id. */ - void update(const InstSeqNum &done_sn, unsigned tid); + void update(const InstSeqNum &done_sn, ThreadID tid); /** * Squashes all outstanding updates until a given sequence number. @@ -107,7 +107,7 @@ class BPredUnit * until. * @param tid The thread id. */ - void squash(const InstSeqNum &squashed_sn, unsigned tid); + void squash(const InstSeqNum &squashed_sn, ThreadID tid); /** * Squashes all outstanding updates until a given sequence number, and @@ -119,7 +119,7 @@ class BPredUnit * @param tid The thread id. */ void squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - bool actually_taken, unsigned tid); + bool actually_taken, ThreadID tid); /** * @param bp_history Pointer to the history object. The predictor @@ -179,8 +179,8 @@ class BPredUnit * information needed to update the predictor, BTB, and RAS. */ PredictorHistory(const InstSeqNum &seq_num, const Addr &inst_PC, - const bool pred_taken, void *bp_history, - const unsigned _tid) + bool pred_taken, void *bp_history, + ThreadID _tid) : seqNum(seq_num), PC(inst_PC), RASTarget(0), RASIndex(0), tid(_tid), predTaken(pred_taken), usedRAS(0), wasCall(0), bpHistory(bp_history) @@ -203,7 +203,7 @@ class BPredUnit unsigned RASIndex; /** The thread id. */ - unsigned tid; + ThreadID tid; /** Whether or not it was predicted taken. */ bool predTaken; diff --git a/src/cpu/o3/bpred_unit_impl.hh b/src/cpu/o3/bpred_unit_impl.hh index 26aba986d..1378ac135 100644 --- a/src/cpu/o3/bpred_unit_impl.hh +++ b/src/cpu/o3/bpred_unit_impl.hh @@ -144,7 +144,7 @@ BPredUnit<Impl>::takeOverFrom() template <class Impl> bool -BPredUnit<Impl>::predict(DynInstPtr &inst, Addr &PC, unsigned tid) +BPredUnit<Impl>::predict(DynInstPtr &inst, Addr &PC, ThreadID tid) { // See if branch predictor predicts taken. // If so, get its target addr either from the BTB or the RAS. @@ -254,7 +254,7 @@ BPredUnit<Impl>::predict(DynInstPtr &inst, Addr &PC, unsigned tid) template <class Impl> void -BPredUnit<Impl>::update(const InstSeqNum &done_sn, unsigned tid) +BPredUnit<Impl>::update(const InstSeqNum &done_sn, ThreadID tid) { DPRINTF(Fetch, "BranchPred: [tid:%i]: Committing branches until " "[sn:%lli].\n", tid, done_sn); @@ -272,7 +272,7 @@ BPredUnit<Impl>::update(const InstSeqNum &done_sn, unsigned tid) template <class Impl> void -BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, unsigned tid) +BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, ThreadID tid) { History &pred_hist = predHist[tid]; @@ -311,8 +311,8 @@ template <class Impl> void BPredUnit<Impl>::squash(const InstSeqNum &squashed_sn, const Addr &corr_target, - const bool actually_taken, - unsigned tid) + bool actually_taken, + ThreadID tid) { // Now that we know that a branch was mispredicted, we need to undo // all the branches that have been seen up until this branch and diff --git a/src/cpu/o3/btb.hh b/src/cpu/o3/btb.hh index 38ecabc46..6557522e0 100644 --- a/src/cpu/o3/btb.hh +++ b/src/cpu/o3/btb.hh @@ -51,7 +51,7 @@ class DefaultBTB Addr target; /** The entry's thread id. */ - unsigned tid; + ThreadID tid; /** Whether or not the entry is valid. */ bool valid; @@ -74,14 +74,14 @@ class DefaultBTB * @param tid The thread id. * @return Returns the target of the branch. */ - Addr lookup(const Addr &inst_PC, unsigned tid); + Addr lookup(const Addr &inst_PC, ThreadID tid); /** Checks if a branch is in the BTB. * @param inst_PC The address of the branch to look up. * @param tid The thread id. * @return Whether or not the branch exists in the BTB. */ - bool valid(const Addr &inst_PC, unsigned tid); + bool valid(const Addr &inst_PC, ThreadID tid); /** Updates the BTB with the target of a branch. * @param inst_PC The address of the branch being updated. @@ -89,7 +89,7 @@ class DefaultBTB * @param tid The thread id. */ void update(const Addr &inst_PC, const Addr &target_PC, - unsigned tid); + ThreadID tid); private: /** Returns the index into the BTB, based on the branch's PC. diff --git a/src/cpu/o3/commit.hh b/src/cpu/o3/commit.hh index f21c14569..d93b85984 100644 --- a/src/cpu/o3/commit.hh +++ b/src/cpu/o3/commit.hh @@ -92,10 +92,10 @@ class DefaultCommit class TrapEvent : public Event { private: DefaultCommit<Impl> *commit; - unsigned tid; + ThreadID tid; public: - TrapEvent(DefaultCommit<Impl> *_commit, unsigned _tid); + TrapEvent(DefaultCommit<Impl> *_commit, ThreadID _tid); void process(); const char *description() const; @@ -172,7 +172,7 @@ class DefaultCommit IEW *iewStage; /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Sets pointer to the commited state rename map. */ void setRenameMap(RenameMap rm_ptr[Impl::MaxThreads]); @@ -204,15 +204,15 @@ class DefaultCommit void commit(); /** Returns the number of free ROB entries for a specific thread. */ - unsigned numROBFreeEntries(unsigned tid); + size_t numROBFreeEntries(ThreadID tid); /** Generates an event to schedule a squash due to a trap. */ - void generateTrapEvent(unsigned tid); + void generateTrapEvent(ThreadID tid); /** Records that commit needs to initiate a squash due to an * external state update through the TC. */ - void generateTCEvent(unsigned tid); + void generateTCEvent(ThreadID tid); private: /** Updates the overall status of commit with the nextStatus, and @@ -237,13 +237,13 @@ class DefaultCommit bool changedROBEntries(); /** Squashes all in flight instructions. */ - void squashAll(unsigned tid); + void squashAll(ThreadID tid); /** Handles squashing due to a trap. */ - void squashFromTrap(unsigned tid); + void squashFromTrap(ThreadID tid); /** Handles squashing due to an TC write. */ - void squashFromTC(unsigned tid); + void squashFromTC(ThreadID tid); #if FULL_SYSTEM /** Handles processing an interrupt. */ @@ -268,13 +268,13 @@ class DefaultCommit void markCompletedInsts(); /** Gets the thread to commit, based on the SMT policy. */ - int getCommittingThread(); + ThreadID getCommittingThread(); /** Returns the thread ID to use based on a round robin policy. */ - int roundRobin(); + ThreadID roundRobin(); /** Returns the thread ID to use based on an oldest instruction policy. */ - int oldestReady(); + ThreadID oldestReady(); public: /** Returns the PC of the head instruction of the ROB. @@ -283,34 +283,34 @@ class DefaultCommit Addr readPC() { return PC[0]; } /** Returns the PC of a specific thread. */ - Addr readPC(unsigned tid) { return PC[tid]; } + Addr readPC(ThreadID tid) { return PC[tid]; } /** Sets the PC of a specific thread. */ - void setPC(Addr val, unsigned tid) { PC[tid] = val; } + void setPC(Addr val, ThreadID tid) { PC[tid] = val; } /** Reads the micro PC of a specific thread. */ - Addr readMicroPC(unsigned tid) { return microPC[tid]; } + Addr readMicroPC(ThreadID tid) { return microPC[tid]; } /** Sets the micro PC of a specific thread */ - void setMicroPC(Addr val, unsigned tid) { microPC[tid] = val; } + void setMicroPC(Addr val, ThreadID tid) { microPC[tid] = val; } /** Reads the next PC of a specific thread. */ - Addr readNextPC(unsigned tid) { return nextPC[tid]; } + Addr readNextPC(ThreadID tid) { return nextPC[tid]; } /** Sets the next PC of a specific thread. */ - void setNextPC(Addr val, unsigned tid) { nextPC[tid] = val; } + void setNextPC(Addr val, ThreadID tid) { nextPC[tid] = val; } /** Reads the next NPC of a specific thread. */ - Addr readNextNPC(unsigned tid) { return nextNPC[tid]; } + Addr readNextNPC(ThreadID tid) { return nextNPC[tid]; } /** Sets the next NPC of a specific thread. */ - void setNextNPC(Addr val, unsigned tid) { nextNPC[tid] = val; } + void setNextNPC(Addr val, ThreadID tid) { nextNPC[tid] = val; } /** Reads the micro PC of a specific thread. */ - Addr readNextMicroPC(unsigned tid) { return nextMicroPC[tid]; } + Addr readNextMicroPC(ThreadID tid) { return nextMicroPC[tid]; } /** Sets the micro PC of a specific thread */ - void setNextMicroPC(Addr val, unsigned tid) { nextMicroPC[tid] = val; } + void setNextMicroPC(Addr val, ThreadID tid) { nextMicroPC[tid] = val; } private: /** Time buffer interface. */ @@ -360,7 +360,7 @@ class DefaultCommit bool changedROBNumEntries[Impl::MaxThreads]; /** A counter of how many threads are currently squashing. */ - int squashCounter; + ThreadID squashCounter; /** Records if a thread has to squash this cycle due to a trap. */ bool trapSquash[Impl::MaxThreads]; @@ -369,7 +369,7 @@ class DefaultCommit bool tcSquash[Impl::MaxThreads]; /** Priority List used for Commit Policy */ - std::list<unsigned> priority_list; + std::list<ThreadID> priority_list; /** IEW to Commit delay, in ticks. */ unsigned iewToCommitDelay; @@ -394,7 +394,7 @@ class DefaultCommit unsigned numRobs; /** Number of Active Threads */ - unsigned numThreads; + ThreadID numThreads; /** Is a drain pending. */ bool drainPending; @@ -443,7 +443,7 @@ class DefaultCommit bool checkEmptyROB[Impl::MaxThreads]; /** Pointer to the list of active threads. */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Rename map interface. */ RenameMap *renameMap[Impl::MaxThreads]; diff --git a/src/cpu/o3/commit_impl.hh b/src/cpu/o3/commit_impl.hh index 7cd88b49b..7286f1b6f 100644 --- a/src/cpu/o3/commit_impl.hh +++ b/src/cpu/o3/commit_impl.hh @@ -29,9 +29,6 @@ * Korey Sewell */ -#include "config/full_system.hh" -#include "config/use_checker.hh" - #include <algorithm> #include <string> @@ -39,19 +36,22 @@ #include "base/cp_annotate.hh" #include "base/loader/symtab.hh" #include "base/timebuf.hh" +#include "config/full_system.hh" +#include "config/use_checker.hh" #include "cpu/exetrace.hh" #include "cpu/o3/commit.hh" #include "cpu/o3/thread_state.hh" +#include "params/DerivO3CPU.hh" #if USE_CHECKER #include "cpu/checker/cpu.hh" #endif -#include "params/DerivO3CPU.hh" +using namespace std; template <class Impl> DefaultCommit<Impl>::TrapEvent::TrapEvent(DefaultCommit<Impl> *_commit, - unsigned _tid) + ThreadID _tid) : Event(CPU_Tick_Pri), commit(_commit), tid(_tid) { this->setFlags(AutoDelete); @@ -105,7 +105,7 @@ DefaultCommit<Impl>::DefaultCommit(O3CPU *_cpu, DerivO3CPUParams *params) commitPolicy = RoundRobin; //Set-Up Priority List - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { priority_list.push_back(tid); } @@ -119,15 +119,19 @@ DefaultCommit<Impl>::DefaultCommit(O3CPU *_cpu, DerivO3CPUParams *params) "RoundRobin,OldestReady}"); } - for (int i=0; i < numThreads; i++) { - commitStatus[i] = Idle; - changedROBNumEntries[i] = false; - checkEmptyROB[i] = false; - trapInFlight[i] = false; - committedStores[i] = false; - trapSquash[i] = false; - tcSquash[i] = false; - microPC[i] = nextMicroPC[i] = PC[i] = nextPC[i] = nextNPC[i] = 0; + for (ThreadID tid = 0; tid < numThreads; tid++) { + commitStatus[tid] = Idle; + changedROBNumEntries[tid] = false; + checkEmptyROB[tid] = false; + trapInFlight[tid] = false; + committedStores[tid] = false; + trapSquash[tid] = false; + tcSquash[tid] = false; + microPC[tid] = 0; + nextMicroPC[tid] = 0; + PC[tid] = 0; + nextPC[tid] = 0; + nextNPC[tid] = 0; } #if FULL_SYSTEM interrupt = NoFault; @@ -175,49 +179,49 @@ DefaultCommit<Impl>::regStats() ; statComInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:count") .desc("Number of instructions committed") .flags(total) ; statComSwp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:swp_count") .desc("Number of s/w prefetches committed") .flags(total) ; statComRefs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:refs") .desc("Number of memory references committed") .flags(total) ; statComLoads - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:loads") .desc("Number of loads committed") .flags(total) ; statComMembars - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:membars") .desc("Number of memory barriers committed") .flags(total) ; statComBranches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:branches") .desc("Number of branches committed") .flags(total) ; commitEligible - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".COM:bw_limited") .desc("number of insts not committed due to BW limits") .flags(total) @@ -288,7 +292,7 @@ DefaultCommit<Impl>::setIEWStage(IEW *iew_stage) template<class Impl> void -DefaultCommit<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +DefaultCommit<Impl>::setActiveThreads(list<ThreadID> *at_ptr) { activeThreads = at_ptr; } @@ -297,9 +301,8 @@ template <class Impl> void DefaultCommit<Impl>::setRenameMap(RenameMap rm_ptr[]) { - for (int i=0; i < numThreads; i++) { - renameMap[i] = &rm_ptr[i]; - } + for (ThreadID tid = 0; tid < numThreads; tid++) + renameMap[tid] = &rm_ptr[tid]; } template <class Impl> @@ -317,10 +320,10 @@ DefaultCommit<Impl>::initStage() rob->resetEntries(); // Broadcast the number of free entries. - for (int i=0; i < numThreads; i++) { - toIEW->commitInfo[i].usedROB = true; - toIEW->commitInfo[i].freeROBEntries = rob->numFreeEntries(i); - toIEW->commitInfo[i].emptyROB = true; + for (ThreadID tid = 0; tid < numThreads; tid++) { + toIEW->commitInfo[tid].usedROB = true; + toIEW->commitInfo[tid].freeROBEntries = rob->numFreeEntries(tid); + toIEW->commitInfo[tid].emptyROB = true; } // Commit must broadcast the number of free entries it has at the @@ -363,11 +366,11 @@ DefaultCommit<Impl>::takeOverFrom() switchedOut = false; _status = Active; _nextStatus = Inactive; - for (int i=0; i < numThreads; i++) { - commitStatus[i] = Idle; - changedROBNumEntries[i] = false; - trapSquash[i] = false; - tcSquash[i] = false; + for (ThreadID tid = 0; tid < numThreads; tid++) { + commitStatus[tid] = Idle; + changedROBNumEntries[tid] = false; + trapSquash[tid] = false; + tcSquash[tid] = false; } squashCounter = 0; rob->takeOverFrom(); @@ -378,11 +381,11 @@ void DefaultCommit<Impl>::updateStatus() { // reset ROB changed variable - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; changedROBNumEntries[tid] = false; @@ -410,11 +413,11 @@ DefaultCommit<Impl>::setNextStatus() { int squashes = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (commitStatus[tid] == ROBSquashing) { squashes++; @@ -434,11 +437,11 @@ template <class Impl> bool DefaultCommit<Impl>::changedROBEntries() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (changedROBNumEntries[tid]) { return true; @@ -449,15 +452,15 @@ DefaultCommit<Impl>::changedROBEntries() } template <class Impl> -unsigned -DefaultCommit<Impl>::numROBFreeEntries(unsigned tid) +size_t +DefaultCommit<Impl>::numROBFreeEntries(ThreadID tid) { return rob->numFreeEntries(tid); } template <class Impl> void -DefaultCommit<Impl>::generateTrapEvent(unsigned tid) +DefaultCommit<Impl>::generateTrapEvent(ThreadID tid) { DPRINTF(Commit, "Generating trap event for [tid:%i]\n", tid); @@ -469,7 +472,7 @@ DefaultCommit<Impl>::generateTrapEvent(unsigned tid) template <class Impl> void -DefaultCommit<Impl>::generateTCEvent(unsigned tid) +DefaultCommit<Impl>::generateTCEvent(ThreadID tid) { assert(!trapInFlight[tid]); DPRINTF(Commit, "Generating TC squash event for [tid:%i]\n", tid); @@ -479,7 +482,7 @@ DefaultCommit<Impl>::generateTCEvent(unsigned tid) template <class Impl> void -DefaultCommit<Impl>::squashAll(unsigned tid) +DefaultCommit<Impl>::squashAll(ThreadID tid) { // If we want to include the squashing instruction in the squash, // then use one older sequence number. @@ -516,7 +519,7 @@ DefaultCommit<Impl>::squashAll(unsigned tid) template <class Impl> void -DefaultCommit<Impl>::squashFromTrap(unsigned tid) +DefaultCommit<Impl>::squashFromTrap(ThreadID tid) { squashAll(tid); @@ -534,7 +537,7 @@ DefaultCommit<Impl>::squashFromTrap(unsigned tid) template <class Impl> void -DefaultCommit<Impl>::squashFromTC(unsigned tid) +DefaultCommit<Impl>::squashFromTC(ThreadID tid) { squashAll(tid); @@ -565,13 +568,13 @@ DefaultCommit<Impl>::tick() if (activeThreads->empty()) return; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); // Check if any of the threads are done squashing. Change the // status if they are done. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // Clear the bit saying if the thread has committed stores // this cycle. @@ -598,7 +601,7 @@ DefaultCommit<Impl>::tick() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!rob->isEmpty(tid) && rob->readHeadInst(tid)->readyToCommit()) { // The ROB has more instructions it can commit. Its next status @@ -704,11 +707,11 @@ DefaultCommit<Impl>::commit() //////////////////////////////////// // Check for any possible squashes, handle them first //////////////////////////////////// - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // Not sure which one takes priority. I think if we have // both, that's a bad sign. @@ -794,7 +797,7 @@ DefaultCommit<Impl>::commit() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (changedROBNumEntries[tid]) { toIEW->commitInfo[tid].usedROB = true; @@ -855,7 +858,7 @@ DefaultCommit<Impl>::commitInsts() head_inst = rob->readHeadInst(commit_thread); - int tid = head_inst->threadNumber; + ThreadID tid = head_inst->threadNumber; assert(tid == commit_thread); @@ -951,7 +954,7 @@ DefaultCommit<Impl>::commitHead(DynInstPtr &head_inst, unsigned inst_num) { assert(head_inst); - int tid = head_inst->threadNumber; + ThreadID tid = head_inst->threadNumber; // If the instruction is not executed yet, then it will need extra // handling. Signal backwards that it should be executed. @@ -1147,7 +1150,7 @@ DefaultCommit<Impl>::getInsts() DynInstPtr inst; inst = fromRename->insts[inst_num]; - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; if (!inst->isSquashed() && commitStatus[tid] != ROBSquashing && @@ -1220,11 +1223,11 @@ template <class Impl> bool DefaultCommit<Impl>::robDoneSquashing() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!rob->isDoneSquashing(tid)) return false; @@ -1237,40 +1240,40 @@ template <class Impl> void DefaultCommit<Impl>::updateComInstStats(DynInstPtr &inst) { - unsigned thread = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) { - statComSwp[thread]++; + statComSwp[tid]++; } else { - statComInst[thread]++; + statComInst[tid]++; } #else - statComInst[thread]++; + statComInst[tid]++; #endif // // Control Instructions // if (inst->isControl()) - statComBranches[thread]++; + statComBranches[tid]++; // // Memory references // if (inst->isMemRef()) { - statComRefs[thread]++; + statComRefs[tid]++; if (inst->isLoad()) { - statComLoads[thread]++; + statComLoads[tid]++; } } if (inst->isMemBarrier()) { - statComMembars[thread]++; + statComMembars[tid]++; } } @@ -1280,7 +1283,7 @@ DefaultCommit<Impl>::updateComInstStats(DynInstPtr &inst) // // //////////////////////////////////////// template <class Impl> -int +ThreadID DefaultCommit<Impl>::getCommittingThread() { if (numThreads > 1) { @@ -1299,31 +1302,31 @@ DefaultCommit<Impl>::getCommittingThread() return oldestReady(); default: - return -1; + return InvalidThreadID; } } else { assert(!activeThreads->empty()); - int tid = activeThreads->front(); + ThreadID tid = activeThreads->front(); if (commitStatus[tid] == Running || commitStatus[tid] == Idle || commitStatus[tid] == FetchTrapPending) { return tid; } else { - return -1; + return InvalidThreadID; } } } template<class Impl> -int +ThreadID DefaultCommit<Impl>::roundRobin() { - std::list<unsigned>::iterator pri_iter = priority_list.begin(); - std::list<unsigned>::iterator end = priority_list.end(); + list<ThreadID>::iterator pri_iter = priority_list.begin(); + list<ThreadID>::iterator end = priority_list.end(); while (pri_iter != end) { - unsigned tid = *pri_iter; + ThreadID tid = *pri_iter; if (commitStatus[tid] == Running || commitStatus[tid] == Idle || @@ -1340,21 +1343,21 @@ DefaultCommit<Impl>::roundRobin() pri_iter++; } - return -1; + return InvalidThreadID; } template<class Impl> -int +ThreadID DefaultCommit<Impl>::oldestReady() { unsigned oldest = 0; bool first = true; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!rob->isEmpty(tid) && (commitStatus[tid] == Running || @@ -1378,6 +1381,6 @@ DefaultCommit<Impl>::oldestReady() if (!first) { return oldest; } else { - return -1; + return InvalidThreadID; } } diff --git a/src/cpu/o3/cpu.cc b/src/cpu/o3/cpu.cc index 038e5daaa..621b6c1b9 100644 --- a/src/cpu/o3/cpu.cc +++ b/src/cpu/o3/cpu.cc @@ -60,6 +60,7 @@ class BaseCPUParams; using namespace TheISA; +using namespace std; BaseO3CPU::BaseO3CPU(BaseCPUParams *params) : BaseCPU(params) @@ -184,7 +185,7 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) scoreboard(params->numThreads, TheISA::NumIntRegs, params->numPhysIntRegs, TheISA::NumFloatRegs, params->numPhysFloatRegs, - TheISA::NumMiscRegs * number_of_threads, + TheISA::NumMiscRegs * numThreads, TheISA::ZeroReg), timeBuffer(params->backComSize, params->forwardComSize), @@ -202,8 +203,7 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) physmem(system->physmem), #endif // FULL_SYSTEM drainCount(0), - deferRegistration(params->defer_registration), - numThreads(number_of_threads) + deferRegistration(params->defer_registration) { if (!deferRegistration) { _status = Running; @@ -224,8 +224,8 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) #endif // USE_CHECKER #if !FULL_SYSTEM - thread.resize(number_of_threads); - tids.resize(number_of_threads); + thread.resize(numThreads); + tids.resize(numThreads); #endif // The stages also need their CPU pointer setup. However this @@ -263,7 +263,7 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) rename.setCommitStage(&commit); #if !FULL_SYSTEM - int active_threads = params->workload.size(); + ThreadID active_threads = params->workload.size(); if (active_threads > Impl::MaxThreads) { panic("Workload Size too large. Increase the 'MaxThreads'" @@ -271,7 +271,7 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) "edit your workload size."); } #else - int active_threads = 1; + ThreadID active_threads = 1; #endif //Make Sure That this a Valid Architeture @@ -285,7 +285,7 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) PhysRegIndex lreg_idx = 0; PhysRegIndex freg_idx = params->numPhysIntRegs; //Index to 1 after int regs - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { bool bindRegs = (tid <= active_threads - 1); commitRenameMap[tid].init(TheISA::NumIntRegs, @@ -328,9 +328,8 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) commit.setRenameMap(commitRenameMap); // Give renameMap & rename stage access to the freeList; - for (int i=0; i < numThreads; i++) { - renameMap[i].setFreeList(&freeList); - } + for (ThreadID tid = 0; tid < numThreads; tid++) + renameMap[tid].setFreeList(&freeList); rename.setFreeList(&freeList); // Setup the ROB for whichever stages need it. @@ -339,11 +338,11 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) lastRunningCycle = curTick; lastActivatedCycle = -1; - +#if 0 // Give renameMap & rename stage access to the freeList; - //for (int i=0; i < numThreads; i++) { - //globalSeqNum[i] = 1; - //} + for (ThreadID tid = 0; tid < numThreads; tid++) + globalSeqNum[tid] = 1; +#endif contextSwitch = false; DPRINTF(O3CPU, "Creating O3CPU object.\n"); @@ -351,30 +350,30 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) // Setup any thread state. this->thread.resize(this->numThreads); - for (int i = 0; i < this->numThreads; ++i) { + for (ThreadID tid = 0; tid < this->numThreads; ++tid) { #if FULL_SYSTEM // SMT is not supported in FS mode yet. assert(this->numThreads == 1); - this->thread[i] = new Thread(this, 0); + this->thread[tid] = new Thread(this, 0); #else - if (i < params->workload.size()) { + if (tid < params->workload.size()) { DPRINTF(O3CPU, "Workload[%i] process is %#x", - i, this->thread[i]); - this->thread[i] = new typename FullO3CPU<Impl>::Thread( + tid, this->thread[tid]); + this->thread[tid] = new typename FullO3CPU<Impl>::Thread( (typename Impl::O3CPU *)(this), - i, params->workload[i], i); + tid, params->workload[tid], tid); - //usedTids[i] = true; - //threadMap[i] = i; + //usedTids[tid] = true; + //threadMap[tid] = tid; } else { //Allocate Empty thread so M5 can use later //when scheduling threads to CPU Process* dummy_proc = NULL; - this->thread[i] = new typename FullO3CPU<Impl>::Thread( + this->thread[tid] = new typename FullO3CPU<Impl>::Thread( (typename Impl::O3CPU *)(this), - i, dummy_proc, i); - //usedTids[i] = false; + tid, dummy_proc, tid); + //usedTids[tid] = false; } #endif // !FULL_SYSTEM @@ -396,22 +395,21 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params) o3_tc->cpu = (typename Impl::O3CPU *)(this); assert(o3_tc->cpu); - o3_tc->thread = this->thread[i]; + o3_tc->thread = this->thread[tid]; #if FULL_SYSTEM // Setup quiesce event. - this->thread[i]->quiesceEvent = new EndQuiesceEvent(tc); + this->thread[tid]->quiesceEvent = new EndQuiesceEvent(tc); #endif // Give the thread the TC. - this->thread[i]->tc = tc; + this->thread[tid]->tc = tc; // Add the TC to the CPU's list of TC's. this->threadContexts.push_back(tc); } - for (int i=0; i < this->numThreads; i++) { - this->thread[i]->setFuncExeInst(0); - } + for (ThreadID tid = 0; tid < this->numThreads; tid++) + this->thread[tid]->setFuncExeInst(0); lockAddr = 0; lockFlag = false; @@ -565,19 +563,19 @@ FullO3CPU<Impl>::init() // Set inSyscall so that the CPU doesn't squash when initially // setting up registers. - for (int i = 0; i < number_of_threads; ++i) - thread[i]->inSyscall = true; + for (ThreadID tid = 0; tid < numThreads; ++tid) + thread[tid]->inSyscall = true; #if FULL_SYSTEM - for (int tid=0; tid < number_of_threads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { ThreadContext *src_tc = threadContexts[tid]; TheISA::initCPU(src_tc, src_tc->contextId()); } #endif // Clear inSyscall. - for (int i = 0; i < number_of_threads; ++i) - thread[i]->inSyscall = false; + for (int tid = 0; tid < numThreads; ++tid) + thread[tid]->inSyscall = false; // Initialize stages. fetch.initStage(); @@ -590,9 +588,9 @@ FullO3CPU<Impl>::init() template <class Impl> void -FullO3CPU<Impl>::activateThread(unsigned tid) +FullO3CPU<Impl>::activateThread(ThreadID tid) { - std::list<unsigned>::iterator isActive = + list<ThreadID>::iterator isActive = std::find(activeThreads.begin(), activeThreads.end(), tid); DPRINTF(O3CPU, "[tid:%i]: Calling activate thread.\n", tid); @@ -607,10 +605,10 @@ FullO3CPU<Impl>::activateThread(unsigned tid) template <class Impl> void -FullO3CPU<Impl>::deactivateThread(unsigned tid) +FullO3CPU<Impl>::deactivateThread(ThreadID tid) { //Remove From Active List, if Active - std::list<unsigned>::iterator thread_it = + list<ThreadID>::iterator thread_it = std::find(activeThreads.begin(), activeThreads.end(), tid); DPRINTF(O3CPU, "[tid:%i]: Calling deactivate thread.\n", tid); @@ -623,8 +621,21 @@ FullO3CPU<Impl>::deactivateThread(unsigned tid) } template <class Impl> +Counter +FullO3CPU<Impl>::totalInstructions() const +{ + Counter total(0); + + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; i++) + total += thread[i]->numInst; + + return total; +} + +template <class Impl> void -FullO3CPU<Impl>::activateContext(int tid, int delay) +FullO3CPU<Impl>::activateContext(ThreadID tid, int delay) { // Needs to set each stage to running as well. if (delay){ @@ -651,7 +662,7 @@ FullO3CPU<Impl>::activateContext(int tid, int delay) template <class Impl> bool -FullO3CPU<Impl>::deallocateContext(int tid, bool remove, int delay) +FullO3CPU<Impl>::deallocateContext(ThreadID tid, bool remove, int delay) { // Schedule removal of thread data from CPU if (delay){ @@ -669,7 +680,7 @@ FullO3CPU<Impl>::deallocateContext(int tid, bool remove, int delay) template <class Impl> void -FullO3CPU<Impl>::suspendContext(int tid) +FullO3CPU<Impl>::suspendContext(ThreadID tid) { DPRINTF(O3CPU,"[tid: %i]: Suspending Thread Context.\n", tid); bool deallocated = deallocateContext(tid, false, 1); @@ -682,7 +693,7 @@ FullO3CPU<Impl>::suspendContext(int tid) template <class Impl> void -FullO3CPU<Impl>::haltContext(int tid) +FullO3CPU<Impl>::haltContext(ThreadID tid) { //For now, this is the same as deallocate DPRINTF(O3CPU,"[tid:%i]: Halt Context called. Deallocating", tid); @@ -691,7 +702,7 @@ FullO3CPU<Impl>::haltContext(int tid) template <class Impl> void -FullO3CPU<Impl>::insertThread(unsigned tid) +FullO3CPU<Impl>::insertThread(ThreadID tid) { DPRINTF(O3CPU,"[tid:%i] Initializing thread into CPU"); // Will change now that the PC and thread state is internal to the CPU @@ -737,7 +748,7 @@ FullO3CPU<Impl>::insertThread(unsigned tid) template <class Impl> void -FullO3CPU<Impl>::removeThread(unsigned tid) +FullO3CPU<Impl>::removeThread(ThreadID tid) { DPRINTF(O3CPU,"[tid:%i] Removing thread context from CPU.\n", tid); @@ -796,7 +807,7 @@ FullO3CPU<Impl>::removeThread(unsigned tid) template <class Impl> void -FullO3CPU<Impl>::activateWhenReady(int tid) +FullO3CPU<Impl>::activateWhenReady(ThreadID tid) { DPRINTF(O3CPU,"[tid:%i]: Checking if resources are available for incoming" "(e.g. PhysRegs/ROB/IQ/LSQ) \n", @@ -855,7 +866,7 @@ FullO3CPU<Impl>::activateWhenReady(int tid) #if FULL_SYSTEM template <class Impl> Fault -FullO3CPU<Impl>::hwrei(unsigned tid) +FullO3CPU<Impl>::hwrei(ThreadID tid) { #if THE_ISA == ALPHA_ISA // Need to clear the lock flag upon returning from an interrupt. @@ -870,7 +881,7 @@ FullO3CPU<Impl>::hwrei(unsigned tid) template <class Impl> bool -FullO3CPU<Impl>::simPalCheck(int palFunc, unsigned tid) +FullO3CPU<Impl>::simPalCheck(int palFunc, ThreadID tid) { #if THE_ISA == ALPHA_ISA if (this->thread[tid]->kernelStats) @@ -925,14 +936,15 @@ FullO3CPU<Impl>::updateMemPorts() { // Update all ThreadContext's memory ports (Functional/Virtual // Ports) - for (int i = 0; i < thread.size(); ++i) + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; ++i) thread[i]->connectMemPorts(thread[i]->getTC()); } #endif template <class Impl> void -FullO3CPU<Impl>::trap(Fault fault, unsigned tid) +FullO3CPU<Impl>::trap(Fault fault, ThreadID tid) { // Pass the thread's TC into the invoke method. fault->invoke(this->threadContexts[tid]); @@ -942,7 +954,7 @@ FullO3CPU<Impl>::trap(Fault fault, unsigned tid) template <class Impl> void -FullO3CPU<Impl>::syscall(int64_t callnum, int tid) +FullO3CPU<Impl>::syscall(int64_t callnum, ThreadID tid) { DPRINTF(O3CPU, "[tid:%i] Executing syscall().\n\n", tid); @@ -977,7 +989,8 @@ FullO3CPU<Impl>::serialize(std::ostream &os) // get instantiated multiple times (causes a panic in statistics). static SimpleThread temp; - for (int i = 0; i < thread.size(); i++) { + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; i++) { nameOut(os, csprintf("%s.xc.%i", name(), i)); temp.copyTC(thread[i]->getTC()); temp.serialize(os); @@ -998,7 +1011,8 @@ FullO3CPU<Impl>::unserialize(Checkpoint *cp, const std::string §ion) // get instantiated multiple times (causes a panic in statistics). static SimpleThread temp; - for (int i = 0; i < thread.size(); i++) { + ThreadID size = thread.size(); + for (ThreadID i = 0; i < size; i++) { temp.copyTC(thread[i]->getTC()); temp.unserialize(cp, csprintf("%s.xc.%i", section, i)); thread[i]->getTC()->copyArchRegs(temp.getTC()); @@ -1134,9 +1148,9 @@ FullO3CPU<Impl>::takeOverFrom(BaseCPU *oldCPU) // @todo: Figure out how to properly select the tid to put onto // the active threads list. - int tid = 0; + ThreadID tid = 0; - std::list<unsigned>::iterator isActive = + list<ThreadID>::iterator isActive = std::find(activeThreads.begin(), activeThreads.end(), tid); if (isActive == activeThreads.end()) { @@ -1150,7 +1164,8 @@ FullO3CPU<Impl>::takeOverFrom(BaseCPU *oldCPU) // Set all statuses to active, schedule the CPU's tick event. // @todo: Fix up statuses so this is handled properly - for (int i = 0; i < threadContexts.size(); ++i) { + ThreadID size = threadContexts.size(); + for (ThreadID i = 0; i < size; ++i) { ThreadContext *tc = threadContexts[i]; if (tc->status() == ThreadContext::Active && _status != Running) { _status = Running; @@ -1163,14 +1178,14 @@ FullO3CPU<Impl>::takeOverFrom(BaseCPU *oldCPU) template <class Impl> TheISA::MiscReg -FullO3CPU<Impl>::readMiscRegNoEffect(int misc_reg, unsigned tid) +FullO3CPU<Impl>::readMiscRegNoEffect(int misc_reg, ThreadID tid) { return this->regFile.readMiscRegNoEffect(misc_reg, tid); } template <class Impl> TheISA::MiscReg -FullO3CPU<Impl>::readMiscReg(int misc_reg, unsigned tid) +FullO3CPU<Impl>::readMiscReg(int misc_reg, ThreadID tid) { return this->regFile.readMiscReg(misc_reg, tid); } @@ -1178,7 +1193,7 @@ FullO3CPU<Impl>::readMiscReg(int misc_reg, unsigned tid) template <class Impl> void FullO3CPU<Impl>::setMiscRegNoEffect(int misc_reg, - const TheISA::MiscReg &val, unsigned tid) + const TheISA::MiscReg &val, ThreadID tid) { this->regFile.setMiscRegNoEffect(misc_reg, val, tid); } @@ -1186,7 +1201,7 @@ FullO3CPU<Impl>::setMiscRegNoEffect(int misc_reg, template <class Impl> void FullO3CPU<Impl>::setMiscReg(int misc_reg, - const TheISA::MiscReg &val, unsigned tid) + const TheISA::MiscReg &val, ThreadID tid) { this->regFile.setMiscReg(misc_reg, val, tid); } @@ -1263,7 +1278,7 @@ FullO3CPU<Impl>::setFloatRegBits(int reg_idx, FloatRegBits val) template <class Impl> uint64_t -FullO3CPU<Impl>::readArchIntReg(int reg_idx, unsigned tid) +FullO3CPU<Impl>::readArchIntReg(int reg_idx, ThreadID tid) { PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx); @@ -1272,7 +1287,7 @@ FullO3CPU<Impl>::readArchIntReg(int reg_idx, unsigned tid) template <class Impl> float -FullO3CPU<Impl>::readArchFloatRegSingle(int reg_idx, unsigned tid) +FullO3CPU<Impl>::readArchFloatRegSingle(int reg_idx, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1282,7 +1297,7 @@ FullO3CPU<Impl>::readArchFloatRegSingle(int reg_idx, unsigned tid) template <class Impl> double -FullO3CPU<Impl>::readArchFloatRegDouble(int reg_idx, unsigned tid) +FullO3CPU<Impl>::readArchFloatRegDouble(int reg_idx, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1292,7 +1307,7 @@ FullO3CPU<Impl>::readArchFloatRegDouble(int reg_idx, unsigned tid) template <class Impl> uint64_t -FullO3CPU<Impl>::readArchFloatRegInt(int reg_idx, unsigned tid) +FullO3CPU<Impl>::readArchFloatRegInt(int reg_idx, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1302,7 +1317,7 @@ FullO3CPU<Impl>::readArchFloatRegInt(int reg_idx, unsigned tid) template <class Impl> void -FullO3CPU<Impl>::setArchIntReg(int reg_idx, uint64_t val, unsigned tid) +FullO3CPU<Impl>::setArchIntReg(int reg_idx, uint64_t val, ThreadID tid) { PhysRegIndex phys_reg = commitRenameMap[tid].lookup(reg_idx); @@ -1311,7 +1326,7 @@ FullO3CPU<Impl>::setArchIntReg(int reg_idx, uint64_t val, unsigned tid) template <class Impl> void -FullO3CPU<Impl>::setArchFloatRegSingle(int reg_idx, float val, unsigned tid) +FullO3CPU<Impl>::setArchFloatRegSingle(int reg_idx, float val, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1321,7 +1336,7 @@ FullO3CPU<Impl>::setArchFloatRegSingle(int reg_idx, float val, unsigned tid) template <class Impl> void -FullO3CPU<Impl>::setArchFloatRegDouble(int reg_idx, double val, unsigned tid) +FullO3CPU<Impl>::setArchFloatRegDouble(int reg_idx, double val, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1331,7 +1346,7 @@ FullO3CPU<Impl>::setArchFloatRegDouble(int reg_idx, double val, unsigned tid) template <class Impl> void -FullO3CPU<Impl>::setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid) +FullO3CPU<Impl>::setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid) { int idx = reg_idx + TheISA::NumIntRegs; PhysRegIndex phys_reg = commitRenameMap[tid].lookup(idx); @@ -1341,77 +1356,77 @@ FullO3CPU<Impl>::setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid) template <class Impl> uint64_t -FullO3CPU<Impl>::readPC(unsigned tid) +FullO3CPU<Impl>::readPC(ThreadID tid) { return commit.readPC(tid); } template <class Impl> void -FullO3CPU<Impl>::setPC(Addr new_PC,unsigned tid) +FullO3CPU<Impl>::setPC(Addr new_PC, ThreadID tid) { commit.setPC(new_PC, tid); } template <class Impl> uint64_t -FullO3CPU<Impl>::readMicroPC(unsigned tid) +FullO3CPU<Impl>::readMicroPC(ThreadID tid) { return commit.readMicroPC(tid); } template <class Impl> void -FullO3CPU<Impl>::setMicroPC(Addr new_PC,unsigned tid) +FullO3CPU<Impl>::setMicroPC(Addr new_PC, ThreadID tid) { commit.setMicroPC(new_PC, tid); } template <class Impl> uint64_t -FullO3CPU<Impl>::readNextPC(unsigned tid) +FullO3CPU<Impl>::readNextPC(ThreadID tid) { return commit.readNextPC(tid); } template <class Impl> void -FullO3CPU<Impl>::setNextPC(uint64_t val,unsigned tid) +FullO3CPU<Impl>::setNextPC(uint64_t val, ThreadID tid) { commit.setNextPC(val, tid); } template <class Impl> uint64_t -FullO3CPU<Impl>::readNextNPC(unsigned tid) +FullO3CPU<Impl>::readNextNPC(ThreadID tid) { return commit.readNextNPC(tid); } template <class Impl> void -FullO3CPU<Impl>::setNextNPC(uint64_t val,unsigned tid) +FullO3CPU<Impl>::setNextNPC(uint64_t val, ThreadID tid) { commit.setNextNPC(val, tid); } template <class Impl> uint64_t -FullO3CPU<Impl>::readNextMicroPC(unsigned tid) +FullO3CPU<Impl>::readNextMicroPC(ThreadID tid) { return commit.readNextMicroPC(tid); } template <class Impl> void -FullO3CPU<Impl>::setNextMicroPC(Addr new_PC,unsigned tid) +FullO3CPU<Impl>::setNextMicroPC(Addr new_PC, ThreadID tid) { commit.setNextMicroPC(new_PC, tid); } template <class Impl> void -FullO3CPU<Impl>::squashFromTC(unsigned tid) +FullO3CPU<Impl>::squashFromTC(ThreadID tid) { this->thread[tid]->inSyscall = true; this->commit.generateTCEvent(tid); @@ -1428,7 +1443,7 @@ FullO3CPU<Impl>::addInst(DynInstPtr &inst) template <class Impl> void -FullO3CPU<Impl>::instDone(unsigned tid) +FullO3CPU<Impl>::instDone(ThreadID tid) { // Keep an instruction count. thread[tid]->numInst++; @@ -1465,7 +1480,7 @@ FullO3CPU<Impl>::removeFrontInst(DynInstPtr &inst) template <class Impl> void -FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid) +FullO3CPU<Impl>::removeInstsNotInROB(ThreadID tid) { DPRINTF(O3CPU, "Thread %i: Deleting instructions from instruction" " list.\n", tid); @@ -1510,8 +1525,7 @@ FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid) template <class Impl> void -FullO3CPU<Impl>::removeInstsUntil(const InstSeqNum &seq_num, - unsigned tid) +FullO3CPU<Impl>::removeInstsUntil(const InstSeqNum &seq_num, ThreadID tid) { assert(!instList.empty()); @@ -1540,7 +1554,7 @@ FullO3CPU<Impl>::removeInstsUntil(const InstSeqNum &seq_num, template <class Impl> inline void -FullO3CPU<Impl>::squashInstIt(const ListIt &instIt, const unsigned &tid) +FullO3CPU<Impl>::squashInstIt(const ListIt &instIt, ThreadID tid) { if ((*instIt)->threadNumber == tid) { DPRINTF(O3CPU, "Squashing instruction, " @@ -1646,17 +1660,17 @@ FullO3CPU<Impl>::wakeup() #endif template <class Impl> -int +ThreadID FullO3CPU<Impl>::getFreeTid() { - for (int i=0; i < numThreads; i++) { - if (!tids[i]) { - tids[i] = true; - return i; + for (ThreadID tid = 0; tid < numThreads; tid++) { + if (!tids[tid]) { + tids[tid] = true; + return tid; } } - return -1; + return InvalidThreadID; } template <class Impl> @@ -1667,7 +1681,8 @@ FullO3CPU<Impl>::doContextSwitch() //ADD CODE TO DEACTIVE THREAD HERE (???) - for (int tid=0; tid < cpuWaitList.size(); tid++) { + ThreadID size = cpuWaitList.size(); + for (ThreadID tid = 0; tid < size; tid++) { activateWhenReady(tid); } @@ -1680,12 +1695,11 @@ template <class Impl> void FullO3CPU<Impl>::updateThreadPriority() { - if (activeThreads.size() > 1) - { + if (activeThreads.size() > 1) { //DEFAULT TO ROUND ROBIN SCHEME //e.g. Move highest priority to end of thread list - std::list<unsigned>::iterator list_begin = activeThreads.begin(); - std::list<unsigned>::iterator list_end = activeThreads.end(); + list<ThreadID>::iterator list_begin = activeThreads.begin(); + list<ThreadID>::iterator list_end = activeThreads.end(); unsigned high_thread = *list_begin; diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh index 9eead4f49..5cf27df75 100644 --- a/src/cpu/o3/cpu.hh +++ b/src/cpu/o3/cpu.hh @@ -155,7 +155,7 @@ class FullO3CPU : public BaseO3CPU { private: /** Number of Thread to Activate */ - int tid; + ThreadID tid; /** Pointer to the CPU. */ FullO3CPU<Impl> *cpu; @@ -175,7 +175,8 @@ class FullO3CPU : public BaseO3CPU }; /** Schedule thread to activate , regardless of its current state. */ - void scheduleActivateThreadEvent(int tid, int delay) + void + scheduleActivateThreadEvent(ThreadID tid, int delay) { // Schedule thread to activate, regardless of its current state. if (activateThreadEvent[tid].squashed()) @@ -187,7 +188,8 @@ class FullO3CPU : public BaseO3CPU } /** Unschedule actiavte thread event, regardless of its current state. */ - void unscheduleActivateThreadEvent(int tid) + void + unscheduleActivateThreadEvent(ThreadID tid) { if (activateThreadEvent[tid].scheduled()) activateThreadEvent[tid].squash(); @@ -200,7 +202,7 @@ class FullO3CPU : public BaseO3CPU { private: /** Number of Thread to deactivate */ - int tid; + ThreadID tid; /** Should the thread be removed from the CPU? */ bool remove; @@ -226,7 +228,8 @@ class FullO3CPU : public BaseO3CPU }; /** Schedule cpu to deallocate thread context.*/ - void scheduleDeallocateContextEvent(int tid, bool remove, int delay) + void + scheduleDeallocateContextEvent(ThreadID tid, bool remove, int delay) { // Schedule thread to activate, regardless of its current state. if (deallocateContextEvent[tid].squashed()) @@ -238,7 +241,8 @@ class FullO3CPU : public BaseO3CPU } /** Unschedule thread deallocation in CPU */ - void unscheduleDeallocateContextEvent(int tid) + void + unscheduleDeallocateContextEvent(ThreadID tid) { if (deallocateContextEvent[tid].scheduled()) deallocateContextEvent[tid].squash(); @@ -288,46 +292,38 @@ class FullO3CPU : public BaseO3CPU { return activeThreads.size(); } /** Add Thread to Active Threads List */ - void activateThread(unsigned tid); + void activateThread(ThreadID tid); /** Remove Thread from Active Threads List */ - void deactivateThread(unsigned tid); + void deactivateThread(ThreadID tid); /** Setup CPU to insert a thread's context */ - void insertThread(unsigned tid); + void insertThread(ThreadID tid); /** Remove all of a thread's context from CPU */ - void removeThread(unsigned tid); + void removeThread(ThreadID tid); /** Count the Total Instructions Committed in the CPU. */ - virtual Counter totalInstructions() const - { - Counter total(0); - - for (int i=0; i < thread.size(); i++) - total += thread[i]->numInst; - - return total; - } + virtual Counter totalInstructions() const; /** Add Thread to Active Threads List. */ - void activateContext(int tid, int delay); + void activateContext(ThreadID tid, int delay); /** Remove Thread from Active Threads List */ - void suspendContext(int tid); + void suspendContext(ThreadID tid); /** Remove Thread from Active Threads List && * Possibly Remove Thread Context from CPU. */ - bool deallocateContext(int tid, bool remove, int delay = 1); + bool deallocateContext(ThreadID tid, bool remove, int delay = 1); /** Remove Thread from Active Threads List && * Remove Thread Context from CPU. */ - void haltContext(int tid); + void haltContext(ThreadID tid); /** Activate a Thread When CPU Resources are Available. */ - void activateWhenReady(int tid); + void activateWhenReady(ThreadID tid); /** Add or Remove a Thread Context in the CPU. */ void doContextSwitch(); @@ -346,7 +342,7 @@ class FullO3CPU : public BaseO3CPU /** Executes a syscall. * @todo: Determine if this needs to be virtual. */ - void syscall(int64_t callnum, int tid); + void syscall(int64_t callnum, ThreadID tid); #endif /** Starts draining the CPU's pipeline of all instructions in @@ -370,13 +366,13 @@ class FullO3CPU : public BaseO3CPU { return globalSeqNum++; } /** Traps to handle given fault. */ - void trap(Fault fault, unsigned tid); + void trap(Fault fault, ThreadID tid); #if FULL_SYSTEM /** HW return from error interrupt. */ - Fault hwrei(unsigned tid); + Fault hwrei(ThreadID tid); - bool simPalCheck(int palFunc, unsigned tid); + bool simPalCheck(int palFunc, ThreadID tid); /** Returns the Fault for any valid interrupt. */ Fault getInterrupts(); @@ -398,19 +394,19 @@ class FullO3CPU : public BaseO3CPU bool validDataAddr(Addr addr) { return true; } /** Get instruction asid. */ - int getInstAsid(unsigned tid) + int getInstAsid(ThreadID tid) { return regFile.miscRegs[tid].getInstAsid(); } /** Get data asid. */ - int getDataAsid(unsigned tid) + int getDataAsid(ThreadID tid) { return regFile.miscRegs[tid].getDataAsid(); } #else /** Get instruction asid. */ - int getInstAsid(unsigned tid) + int getInstAsid(ThreadID tid) { return thread[tid]->getInstAsid(); } /** Get data asid. */ - int getDataAsid(unsigned tid) + int getDataAsid(ThreadID tid) { return thread[tid]->getDataAsid(); } #endif @@ -418,21 +414,22 @@ class FullO3CPU : public BaseO3CPU /** Register accessors. Index refers to the physical register index. */ /** Reads a miscellaneous register. */ - TheISA::MiscReg readMiscRegNoEffect(int misc_reg, unsigned tid); + TheISA::MiscReg readMiscRegNoEffect(int misc_reg, ThreadID tid); /** Reads a misc. register, including any side effects the read * might have as defined by the architecture. */ - TheISA::MiscReg readMiscReg(int misc_reg, unsigned tid); + TheISA::MiscReg readMiscReg(int misc_reg, ThreadID tid); /** Sets a miscellaneous register. */ - void setMiscRegNoEffect(int misc_reg, const TheISA::MiscReg &val, unsigned tid); + void setMiscRegNoEffect(int misc_reg, const TheISA::MiscReg &val, + ThreadID tid); /** Sets a misc. register, including any side effects the write * might have as defined by the architecture. */ void setMiscReg(int misc_reg, const TheISA::MiscReg &val, - unsigned tid); + ThreadID tid); uint64_t readIntReg(int reg_idx); @@ -454,62 +451,62 @@ class FullO3CPU : public BaseO3CPU void setFloatRegBits(int reg_idx, TheISA::FloatRegBits val, int width); - uint64_t readArchIntReg(int reg_idx, unsigned tid); + uint64_t readArchIntReg(int reg_idx, ThreadID tid); - float readArchFloatRegSingle(int reg_idx, unsigned tid); + float readArchFloatRegSingle(int reg_idx, ThreadID tid); - double readArchFloatRegDouble(int reg_idx, unsigned tid); + double readArchFloatRegDouble(int reg_idx, ThreadID tid); - uint64_t readArchFloatRegInt(int reg_idx, unsigned tid); + uint64_t readArchFloatRegInt(int reg_idx, ThreadID tid); /** Architectural register accessors. Looks up in the commit * rename table to obtain the true physical index of the * architected register first, then accesses that physical * register. */ - void setArchIntReg(int reg_idx, uint64_t val, unsigned tid); + void setArchIntReg(int reg_idx, uint64_t val, ThreadID tid); - void setArchFloatRegSingle(int reg_idx, float val, unsigned tid); + void setArchFloatRegSingle(int reg_idx, float val, ThreadID tid); - void setArchFloatRegDouble(int reg_idx, double val, unsigned tid); + void setArchFloatRegDouble(int reg_idx, double val, ThreadID tid); - void setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid); + void setArchFloatRegInt(int reg_idx, uint64_t val, ThreadID tid); /** Reads the commit PC of a specific thread. */ - Addr readPC(unsigned tid); + Addr readPC(ThreadID tid); /** Sets the commit PC of a specific thread. */ - void setPC(Addr new_PC, unsigned tid); + void setPC(Addr new_PC, ThreadID tid); /** Reads the commit micro PC of a specific thread. */ - Addr readMicroPC(unsigned tid); + Addr readMicroPC(ThreadID tid); /** Sets the commmit micro PC of a specific thread. */ - void setMicroPC(Addr new_microPC, unsigned tid); + void setMicroPC(Addr new_microPC, ThreadID tid); /** Reads the next PC of a specific thread. */ - Addr readNextPC(unsigned tid); + Addr readNextPC(ThreadID tid); /** Sets the next PC of a specific thread. */ - void setNextPC(Addr val, unsigned tid); + void setNextPC(Addr val, ThreadID tid); /** Reads the next NPC of a specific thread. */ - Addr readNextNPC(unsigned tid); + Addr readNextNPC(ThreadID tid); /** Sets the next NPC of a specific thread. */ - void setNextNPC(Addr val, unsigned tid); + void setNextNPC(Addr val, ThreadID tid); /** Reads the commit next micro PC of a specific thread. */ - Addr readNextMicroPC(unsigned tid); + Addr readNextMicroPC(ThreadID tid); /** Sets the commit next micro PC of a specific thread. */ - void setNextMicroPC(Addr val, unsigned tid); + void setNextMicroPC(Addr val, ThreadID tid); /** Initiates a squash of all in-flight instructions for a given * thread. The source of the squash is an external update of * state through the TC. */ - void squashFromTC(unsigned tid); + void squashFromTC(ThreadID tid); /** Function to add instruction onto the head of the list of the * instructions. Used when new instructions are fetched. @@ -517,7 +514,7 @@ class FullO3CPU : public BaseO3CPU ListIt addInst(DynInstPtr &inst); /** Function to tell the CPU that an instruction has completed. */ - void instDone(unsigned tid); + void instDone(ThreadID tid); /** Add Instructions to the CPU Remove List*/ void addToRemoveList(DynInstPtr &inst); @@ -529,13 +526,13 @@ class FullO3CPU : public BaseO3CPU /** Remove all instructions that are not currently in the ROB. * There's also an option to not squash delay slot instructions.*/ - void removeInstsNotInROB(unsigned tid); + void removeInstsNotInROB(ThreadID tid); /** Remove all instructions younger than the given sequence number. */ - void removeInstsUntil(const InstSeqNum &seq_num,unsigned tid); + void removeInstsUntil(const InstSeqNum &seq_num, ThreadID tid); /** Removes the instruction pointed to by the iterator. */ - inline void squashInstIt(const ListIt &instIt, const unsigned &tid); + inline void squashInstIt(const ListIt &instIt, ThreadID tid); /** Cleans up all instructions on the remove list. */ void cleanUpRemovedInsts(); @@ -601,7 +598,7 @@ class FullO3CPU : public BaseO3CPU typename CPUPolicy::ROB rob; /** Active Threads List */ - std::list<unsigned> activeThreads; + std::list<ThreadID> activeThreads; /** Integer Register Scoreboard */ Scoreboard scoreboard; @@ -674,11 +671,12 @@ class FullO3CPU : public BaseO3CPU #endif /** Gets a free thread id. Use if thread ids change across system. */ - int getFreeTid(); + ThreadID getFreeTid(); public: /** Returns a pointer to a thread context. */ - ThreadContext *tcBase(unsigned tid) + ThreadContext * + tcBase(ThreadID tid) { return thread[tid]->getTC(); } @@ -726,14 +724,11 @@ class FullO3CPU : public BaseO3CPU /** The cycle that the CPU was last activated by a new thread*/ Tick lastActivatedCycle; - /** Number of Threads CPU can process */ - unsigned numThreads; - /** Mapping for system thread id to cpu id */ - std::map<unsigned,unsigned> threadMap; + std::map<ThreadID, unsigned> threadMap; /** Available thread ids in the cpu*/ - std::vector<unsigned> tids; + std::vector<ThreadID> tids; /** CPU read function, forwards read to LSQ. */ template <class T> diff --git a/src/cpu/o3/cpu_builder.cc b/src/cpu/o3/cpu_builder.cc index 77d7091ad..a433235a0 100644 --- a/src/cpu/o3/cpu_builder.cc +++ b/src/cpu/o3/cpu_builder.cc @@ -49,11 +49,11 @@ DerivO3CPUParams::create() { #if FULL_SYSTEM // Full-system only supports a single thread for the moment. - int actual_num_threads = 1; + ThreadID actual_num_threads = 1; #else // In non-full-system mode, we infer the number of threads from // the workload if it's not explicitly specified. - int actual_num_threads = + ThreadID actual_num_threads = (numThreads >= workload.size()) ? numThreads : workload.size(); if (workload.size() == 0) { diff --git a/src/cpu/o3/decode.hh b/src/cpu/o3/decode.hh index 294b5b623..e272b661f 100644 --- a/src/cpu/o3/decode.hh +++ b/src/cpu/o3/decode.hh @@ -105,7 +105,7 @@ class DefaultDecode void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Drains the decode stage. */ bool drain(); @@ -129,20 +129,20 @@ class DefaultDecode * change (ie switching from from blocking to unblocking). * @param tid Thread id to decode instructions from. */ - void decode(bool &status_change, unsigned tid); + void decode(bool &status_change, ThreadID tid); /** Processes instructions from fetch and passes them on to rename. * Decoding of instructions actually happens when they are created in * fetch, so this function mostly checks if PC-relative branches are * correct. */ - void decodeInsts(unsigned tid); + void decodeInsts(ThreadID tid); private: /** Inserts a thread's instructions into the skid buffer, to be decoded * once decode unblocks. */ - void skidInsert(unsigned tid); + void skidInsert(ThreadID tid); /** Returns if all of the skid buffers are empty. */ bool skidsEmpty(); @@ -156,13 +156,13 @@ class DefaultDecode void sortInsts(); /** Reads all stall signals from the backwards communication timebuffer. */ - void readStallSignals(unsigned tid); + void readStallSignals(ThreadID tid); /** Checks all input signals and updates decode's status appropriately. */ - bool checkSignalsAndUpdate(unsigned tid); + bool checkSignalsAndUpdate(ThreadID tid); /** Checks all stall signals, and returns if any are true. */ - bool checkStall(unsigned tid) const; + bool checkStall(ThreadID tid) const; /** Returns if there any instructions from fetch on this cycle. */ inline bool fetchInstsValid(); @@ -171,24 +171,24 @@ class DefaultDecode * become blocked. * @return Returns true if there is a status change. */ - bool block(unsigned tid); + bool block(ThreadID tid); /** Switches decode to unblocking if the skid buffer is empty, and * signals back that decode has unblocked. * @return Returns true if there is a status change. */ - bool unblock(unsigned tid); + bool unblock(ThreadID tid); /** Squashes if there is a PC-relative branch that was predicted * incorrectly. Sends squash information back to fetch. */ - void squash(DynInstPtr &inst, unsigned tid); + void squash(DynInstPtr &inst, ThreadID tid); public: /** Squashes due to commit signalling a squash. Changes status to * squashing and clears block/unblock signals as needed. */ - unsigned squash(unsigned tid); + unsigned squash(ThreadID tid); private: // Interfaces to objects outside of decode. @@ -263,10 +263,10 @@ class DefaultDecode unsigned toRenameIndex; /** number of Active Threads*/ - unsigned numThreads; + ThreadID numThreads; /** List of active thread ids */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Number of branches in flight. */ unsigned branchCount[Impl::MaxThreads]; diff --git a/src/cpu/o3/decode_impl.hh b/src/cpu/o3/decode_impl.hh index d92e97f61..86f87991c 100644 --- a/src/cpu/o3/decode_impl.hh +++ b/src/cpu/o3/decode_impl.hh @@ -29,9 +29,10 @@ */ #include "cpu/o3/decode.hh" - #include "params/DerivO3CPU.hh" +using namespace std; + template<class Impl> DefaultDecode<Impl>::DefaultDecode(O3CPU *_cpu, DerivO3CPUParams *params) : cpu(_cpu), @@ -45,12 +46,12 @@ DefaultDecode<Impl>::DefaultDecode(O3CPU *_cpu, DerivO3CPUParams *params) _status = Inactive; // Setup status, make sure stall signals are clear. - for (int i = 0; i < numThreads; ++i) { - decodeStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + decodeStatus[tid] = Idle; - stalls[i].rename = false; - stalls[i].iew = false; - stalls[i].commit = false; + stalls[tid].rename = false; + stalls[tid].iew = false; + stalls[tid].commit = false; } // @todo: Make into a parameter @@ -148,7 +149,7 @@ DefaultDecode<Impl>::setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr) template<class Impl> void -DefaultDecode<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +DefaultDecode<Impl>::setActiveThreads(std::list<ThreadID> *at_ptr) { activeThreads = at_ptr; } @@ -169,24 +170,24 @@ DefaultDecode<Impl>::takeOverFrom() _status = Inactive; // Be sure to reset state and clear out any old instructions. - for (int i = 0; i < numThreads; ++i) { - decodeStatus[i] = Idle; - - stalls[i].rename = false; - stalls[i].iew = false; - stalls[i].commit = false; - while (!insts[i].empty()) - insts[i].pop(); - while (!skidBuffer[i].empty()) - skidBuffer[i].pop(); - branchCount[i] = 0; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + decodeStatus[tid] = Idle; + + stalls[tid].rename = false; + stalls[tid].iew = false; + stalls[tid].commit = false; + while (!insts[tid].empty()) + insts[tid].pop(); + while (!skidBuffer[tid].empty()) + skidBuffer[tid].pop(); + branchCount[tid] = 0; } wroteToTimeBuffer = false; } template<class Impl> bool -DefaultDecode<Impl>::checkStall(unsigned tid) const +DefaultDecode<Impl>::checkStall(ThreadID tid) const { bool ret_val = false; @@ -213,7 +214,7 @@ DefaultDecode<Impl>::fetchInstsValid() template<class Impl> bool -DefaultDecode<Impl>::block(unsigned tid) +DefaultDecode<Impl>::block(ThreadID tid) { DPRINTF(Decode, "[tid:%u]: Blocking.\n", tid); @@ -241,7 +242,7 @@ DefaultDecode<Impl>::block(unsigned tid) template<class Impl> bool -DefaultDecode<Impl>::unblock(unsigned tid) +DefaultDecode<Impl>::unblock(ThreadID tid) { // Decode is done unblocking only if the skid buffer is empty. if (skidBuffer[tid].empty()) { @@ -260,7 +261,7 @@ DefaultDecode<Impl>::unblock(unsigned tid) template<class Impl> void -DefaultDecode<Impl>::squash(DynInstPtr &inst, unsigned tid) +DefaultDecode<Impl>::squash(DynInstPtr &inst, ThreadID tid) { DPRINTF(Decode, "[tid:%i]: [sn:%i] Squashing due to incorrect branch prediction " "detected at decode.\n", tid, inst->seqNum); @@ -320,7 +321,7 @@ DefaultDecode<Impl>::squash(DynInstPtr &inst, unsigned tid) template<class Impl> unsigned -DefaultDecode<Impl>::squash(unsigned tid) +DefaultDecode<Impl>::squash(ThreadID tid) { DPRINTF(Decode, "[tid:%i]: Squashing.\n",tid); @@ -369,7 +370,7 @@ DefaultDecode<Impl>::squash(unsigned tid) template<class Impl> void -DefaultDecode<Impl>::skidInsert(unsigned tid) +DefaultDecode<Impl>::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -395,11 +396,11 @@ template<class Impl> bool DefaultDecode<Impl>::skidsEmpty() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!skidBuffer[tid].empty()) return false; } @@ -413,11 +414,11 @@ DefaultDecode<Impl>::updateStatus() { bool any_unblocking = false; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (decodeStatus[tid] == Unblocking) { any_unblocking = true; @@ -452,8 +453,8 @@ DefaultDecode<Impl>::sortInsts() { int insts_from_fetch = fromFetch->size; #ifdef DEBUG - for (int i=0; i < numThreads; i++) - assert(insts[i].empty()); + for (ThreadID tid = 0; tid < numThreads; tid++) + assert(insts[tid].empty()); #endif for (int i = 0; i < insts_from_fetch; ++i) { insts[fromFetch->insts[i]->threadNumber].push(fromFetch->insts[i]); @@ -462,7 +463,7 @@ DefaultDecode<Impl>::sortInsts() template<class Impl> void -DefaultDecode<Impl>::readStallSignals(unsigned tid) +DefaultDecode<Impl>::readStallSignals(ThreadID tid) { if (fromRename->renameBlock[tid]) { stalls[tid].rename = true; @@ -494,7 +495,7 @@ DefaultDecode<Impl>::readStallSignals(unsigned tid) template <class Impl> bool -DefaultDecode<Impl>::checkSignalsAndUpdate(unsigned tid) +DefaultDecode<Impl>::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is. // Check stall signals, block if necessary. @@ -569,14 +570,14 @@ DefaultDecode<Impl>::tick() toRenameIndex = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); sortInsts(); //Check stall and squash signals. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; DPRINTF(Decode,"Processing [tid:%i]\n",tid); status_change = checkSignalsAndUpdate(tid) || status_change; @@ -597,7 +598,7 @@ DefaultDecode<Impl>::tick() template<class Impl> void -DefaultDecode<Impl>::decode(bool &status_change, unsigned tid) +DefaultDecode<Impl>::decode(bool &status_change, ThreadID tid) { // If status is Running or idle, // call decodeInsts() @@ -642,7 +643,7 @@ DefaultDecode<Impl>::decode(bool &status_change, unsigned tid) template <class Impl> void -DefaultDecode<Impl>::decodeInsts(unsigned tid) +DefaultDecode<Impl>::decodeInsts(ThreadID tid) { // Instructions can come either from the skid buffer or the list of // instructions coming from fetch, depending on decode's status. diff --git a/src/cpu/o3/fetch.hh b/src/cpu/o3/fetch.hh index 08ccb094b..9cbc50899 100644 --- a/src/cpu/o3/fetch.hh +++ b/src/cpu/o3/fetch.hh @@ -157,7 +157,7 @@ class DefaultFetch FetchPriority fetchPolicy; /** List that has the threads organized by priority. */ - std::list<unsigned> priorityList; + std::list<ThreadID> priorityList; public: /** DefaultFetch constructor. */ @@ -176,7 +176,7 @@ class DefaultFetch void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Sets pointer to time buffer used to communicate to the next stage. */ void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr); @@ -240,21 +240,21 @@ class DefaultFetch * @param tid Thread id. * @return Any fault that occured. */ - bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid); + bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid); /** Squashes a specific thread and resets the PC. */ inline void doSquash(const Addr &new_PC, const Addr &new_NPC, - const Addr &new_MicroPC, unsigned tid); + const Addr &new_MicroPC, ThreadID tid); /** Squashes a specific thread and resets the PC. Also tells the CPU to * remove any instructions between fetch and decode that should be sqaushed. */ void squashFromDecode(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid); + const InstSeqNum &seq_num, ThreadID tid); /** Checks if a thread is stalled. */ - bool checkStall(unsigned tid) const; + bool checkStall(ThreadID tid) const; /** Updates overall fetch stage status; to be called at the end of each * cycle. */ @@ -267,7 +267,7 @@ class DefaultFetch */ void squash(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid); + const InstSeqNum &seq_num, ThreadID tid); /** Ticks the fetch stage, processing all inputs signals and fetching * as many instructions as possible. @@ -277,7 +277,7 @@ class DefaultFetch /** Checks all input signals and updates the status as necessary. * @return: Returns if the status has changed due to input signals. */ - bool checkSignalsAndUpdate(unsigned tid); + bool checkSignalsAndUpdate(ThreadID tid); /** Does the actual fetching of instructions and passing them on to the * next stage. @@ -298,19 +298,20 @@ class DefaultFetch void recvRetry(); /** Returns the appropriate thread to fetch, given the fetch policy. */ - int getFetchingThread(FetchPriority &fetch_priority); + ThreadID getFetchingThread(FetchPriority &fetch_priority); /** Returns the appropriate thread to fetch using a round robin policy. */ - int roundRobin(); + ThreadID roundRobin(); /** Returns the appropriate thread to fetch using the IQ count policy. */ - int iqCount(); + ThreadID iqCount(); /** Returns the appropriate thread to fetch using the LSQ count policy. */ - int lsqCount(); + ThreadID lsqCount(); - /** Returns the appropriate thread to fetch using the branch count policy. */ - int branchCount(); + /** Returns the appropriate thread to fetch using the branch count + * policy. */ + ThreadID branchCount(); private: /** Pointer to the O3CPU. */ @@ -400,7 +401,7 @@ class DefaultFetch PacketPtr retryPkt; /** The thread that is waiting on the cache to tell fetch to retry. */ - int retryTid; + ThreadID retryTid; /** Cache block size. */ int cacheBlkSize; @@ -424,16 +425,16 @@ class DefaultFetch Counter lastIcacheStall[Impl::MaxThreads]; /** List of Active Threads */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Number of threads. */ - unsigned numThreads; + ThreadID numThreads; /** Number of threads that are actively fetching. */ - unsigned numFetchingThreads; + ThreadID numFetchingThreads; /** Thread ID being fetched. */ - int threadFetched; + ThreadID threadFetched; /** Checks if there is an interrupt pending. If there is, fetch * must stop once it is not fetching PAL instructions. diff --git a/src/cpu/o3/fetch_impl.hh b/src/cpu/o3/fetch_impl.hh index 4d8033a8c..a76e07576 100644 --- a/src/cpu/o3/fetch_impl.hh +++ b/src/cpu/o3/fetch_impl.hh @@ -51,6 +51,8 @@ #include "sim/system.hh" #endif // FULL_SYSTEM +using namespace std; + template<class Impl> void DefaultFetch<Impl>::IcachePort::setPeer(Port *port) @@ -122,7 +124,7 @@ DefaultFetch<Impl>::DefaultFetch(O3CPU *_cpu, DerivO3CPUParams *params) fetchWidth(params->fetchWidth), cacheBlocked(false), retryPkt(NULL), - retryTid(-1), + retryTid(InvalidThreadID), numThreads(params->numThreads), numFetchingThreads(params->smtNumFetchingThreads), interruptPending(false), @@ -292,7 +294,7 @@ DefaultFetch<Impl>::setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer) template<class Impl> void -DefaultFetch<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +DefaultFetch<Impl>::setActiveThreads(std::list<ThreadID> *at_ptr) { activeThreads = at_ptr; } @@ -312,13 +314,13 @@ void DefaultFetch<Impl>::initStage() { // Setup PC and nextPC with initial state. - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { PC[tid] = cpu->readPC(tid); nextPC[tid] = cpu->readNextPC(tid); microPC[tid] = cpu->readMicroPC(tid); } - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { fetchStatus[tid] = Running; @@ -350,7 +352,7 @@ DefaultFetch<Impl>::setIcache() // Create mask to get rid of offset bits. cacheBlkMask = (cacheBlkSize - 1); - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { // Create space to store a cache line. cacheData[tid] = new uint8_t[cacheBlkSize]; cacheDataPC[tid] = 0; @@ -362,7 +364,7 @@ template<class Impl> void DefaultFetch<Impl>::processCacheCompletion(PacketPtr pkt) { - unsigned tid = pkt->req->threadId(); + ThreadID tid = pkt->req->threadId(); DPRINTF(Fetch, "[tid:%u] Waking up from cache miss.\n",tid); @@ -437,7 +439,7 @@ void DefaultFetch<Impl>::takeOverFrom() { // Reset all state - for (int i = 0; i < Impl::MaxThreads; ++i) { + for (ThreadID i = 0; i < Impl::MaxThreads; ++i) { stalls[i].decode = 0; stalls[i].rename = 0; stalls[i].iew = 0; @@ -518,7 +520,7 @@ DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, //would reset the micro pc to 0. next_MicroPC = 0; - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; Addr pred_PC = next_PC; predict_taken = branchPred.predict(inst, pred_PC, tid); @@ -560,7 +562,7 @@ DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, template <class Impl> bool -DefaultFetch<Impl>::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid) +DefaultFetch<Impl>::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid) { Fault fault = NoFault; @@ -637,7 +639,7 @@ DefaultFetch<Impl>::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid // exists within the cache. if (!icachePort->sendTiming(data_pkt)) { assert(retryPkt == NULL); - assert(retryTid == -1); + assert(retryTid == InvalidThreadID); DPRINTF(Fetch, "[tid:%i] Out of MSHRs!\n", tid); fetchStatus[tid] = IcacheWaitRetry; retryPkt = data_pkt; @@ -666,7 +668,7 @@ DefaultFetch<Impl>::fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid template <class Impl> inline void DefaultFetch<Impl>::doSquash(const Addr &new_PC, - const Addr &new_NPC, const Addr &new_microPC, unsigned tid) + const Addr &new_NPC, const Addr &new_microPC, ThreadID tid) { DPRINTF(Fetch, "[tid:%i]: Squashing, setting PC to: %#x, NPC to: %#x.\n", tid, new_PC, new_NPC); @@ -690,7 +692,7 @@ DefaultFetch<Impl>::doSquash(const Addr &new_PC, delete retryPkt; } retryPkt = NULL; - retryTid = -1; + retryTid = InvalidThreadID; } fetchStatus[tid] = Squashing; @@ -702,7 +704,7 @@ template<class Impl> void DefaultFetch<Impl>::squashFromDecode(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid) + const InstSeqNum &seq_num, ThreadID tid) { DPRINTF(Fetch, "[tid:%i]: Squashing from decode.\n",tid); @@ -715,7 +717,7 @@ DefaultFetch<Impl>::squashFromDecode(const Addr &new_PC, const Addr &new_NPC, template<class Impl> bool -DefaultFetch<Impl>::checkStall(unsigned tid) const +DefaultFetch<Impl>::checkStall(ThreadID tid) const { bool ret_val = false; @@ -744,11 +746,11 @@ typename DefaultFetch<Impl>::FetchStatus DefaultFetch<Impl>::updateFetchStatus() { //Check Running - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (fetchStatus[tid] == Running || fetchStatus[tid] == Squashing || @@ -783,7 +785,7 @@ template <class Impl> void DefaultFetch<Impl>::squash(const Addr &new_PC, const Addr &new_NPC, const Addr &new_MicroPC, - const InstSeqNum &seq_num, unsigned tid) + const InstSeqNum &seq_num, ThreadID tid) { DPRINTF(Fetch, "[tid:%u]: Squash from commit.\n",tid); @@ -797,14 +799,14 @@ template <class Impl> void DefaultFetch<Impl>::tick() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); bool status_change = false; wroteToTimeBuffer = false; while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // Check the signals for each thread to determine the proper status // for each thread. @@ -851,7 +853,7 @@ DefaultFetch<Impl>::tick() template <class Impl> bool -DefaultFetch<Impl>::checkSignalsAndUpdate(unsigned tid) +DefaultFetch<Impl>::checkSignalsAndUpdate(ThreadID tid) { // Update the per thread stall statuses. if (fromDecode->decodeBlock[tid]) { @@ -1000,9 +1002,9 @@ DefaultFetch<Impl>::fetch(bool &status_change) ////////////////////////////////////////// // Start actual fetch ////////////////////////////////////////// - int tid = getFetchingThread(fetchPolicy); + ThreadID tid = getFetchingThread(fetchPolicy); - if (tid == -1 || drainPending) { + if (tid == InvalidThreadID || drainPending) { DPRINTF(Fetch,"There are no more threads available to fetch from.\n"); // Breaks looping condition in tick() @@ -1277,17 +1279,17 @@ DefaultFetch<Impl>::recvRetry() { if (retryPkt != NULL) { assert(cacheBlocked); - assert(retryTid != -1); + assert(retryTid != InvalidThreadID); assert(fetchStatus[retryTid] == IcacheWaitRetry); if (icachePort->sendTiming(retryPkt)) { fetchStatus[retryTid] = IcacheWaitResponse; retryPkt = NULL; - retryTid = -1; + retryTid = InvalidThreadID; cacheBlocked = false; } } else { - assert(retryTid == -1); + assert(retryTid == InvalidThreadID); // Access has been squashed since it was sent out. Just clear // the cache being blocked. cacheBlocked = false; @@ -1300,7 +1302,7 @@ DefaultFetch<Impl>::recvRetry() // // /////////////////////////////////////// template<class Impl> -int +ThreadID DefaultFetch<Impl>::getFetchingThread(FetchPriority &fetch_priority) { if (numThreads > 1) { @@ -1322,36 +1324,35 @@ DefaultFetch<Impl>::getFetchingThread(FetchPriority &fetch_priority) return branchCount(); default: - return -1; + return InvalidThreadID; } } else { - std::list<unsigned>::iterator thread = activeThreads->begin(); + list<ThreadID>::iterator thread = activeThreads->begin(); if (thread == activeThreads->end()) { - return -1; + return InvalidThreadID; } - int tid = *thread; + ThreadID tid = *thread; if (fetchStatus[tid] == Running || fetchStatus[tid] == IcacheAccessComplete || fetchStatus[tid] == Idle) { return tid; } else { - return -1; + return InvalidThreadID; } } - } template<class Impl> -int +ThreadID DefaultFetch<Impl>::roundRobin() { - std::list<unsigned>::iterator pri_iter = priorityList.begin(); - std::list<unsigned>::iterator end = priorityList.end(); + list<ThreadID>::iterator pri_iter = priorityList.begin(); + list<ThreadID>::iterator end = priorityList.end(); - int high_pri; + ThreadID high_pri; while (pri_iter != end) { high_pri = *pri_iter; @@ -1371,27 +1372,26 @@ DefaultFetch<Impl>::roundRobin() pri_iter++; } - return -1; + return InvalidThreadID; } template<class Impl> -int +ThreadID DefaultFetch<Impl>::iqCount() { - std::priority_queue<unsigned> PQ; + std::priority_queue<ThreadID> PQ; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; PQ.push(fromIEW->iewInfo[tid].iqCount); } while (!PQ.empty()) { - - unsigned high_pri = PQ.top(); + ThreadID high_pri = PQ.top(); if (fetchStatus[high_pri] == Running || fetchStatus[high_pri] == IcacheAccessComplete || @@ -1402,27 +1402,26 @@ DefaultFetch<Impl>::iqCount() } - return -1; + return InvalidThreadID; } template<class Impl> -int +ThreadID DefaultFetch<Impl>::lsqCount() { - std::priority_queue<unsigned> PQ; + std::priority_queue<ThreadID> PQ; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; PQ.push(fromIEW->iewInfo[tid].ldstqCount); } while (!PQ.empty()) { - - unsigned high_pri = PQ.top(); + ThreadID high_pri = PQ.top(); if (fetchStatus[high_pri] == Running || fetchStatus[high_pri] == IcacheAccessComplete || @@ -1430,20 +1429,21 @@ DefaultFetch<Impl>::lsqCount() return high_pri; else PQ.pop(); - } - return -1; + return InvalidThreadID; } template<class Impl> -int +ThreadID DefaultFetch<Impl>::branchCount() { - std::list<unsigned>::iterator thread = activeThreads->begin(); +#if 0 + list<ThreadID>::iterator thread = activeThreads->begin(); assert(thread != activeThreads->end()); - unsigned tid = *thread; + ThreadID tid = *thread; +#endif panic("Branch Count Fetch policy unimplemented\n"); - return 0 * tid; + return InvalidThreadID; } diff --git a/src/cpu/o3/free_list.cc b/src/cpu/o3/free_list.cc index ae651398b..1144238f4 100644 --- a/src/cpu/o3/free_list.cc +++ b/src/cpu/o3/free_list.cc @@ -32,7 +32,7 @@ #include "cpu/o3/free_list.hh" -SimpleFreeList::SimpleFreeList(unsigned activeThreads, +SimpleFreeList::SimpleFreeList(ThreadID activeThreads, unsigned _numLogicalIntRegs, unsigned _numPhysicalIntRegs, unsigned _numLogicalFloatRegs, diff --git a/src/cpu/o3/free_list.hh b/src/cpu/o3/free_list.hh index 97b56909e..79e10524b 100644 --- a/src/cpu/o3/free_list.hh +++ b/src/cpu/o3/free_list.hh @@ -85,7 +85,7 @@ class SimpleFreeList * @param _numLogicalFloatRegs Number of logical fp registers. * @param _numPhysicalFloatRegs Number of physical fp registers. */ - SimpleFreeList(unsigned activeThreads, + SimpleFreeList(ThreadID activeThreads, unsigned _numLogicalIntRegs, unsigned _numPhysicalIntRegs, unsigned _numLogicalFloatRegs, diff --git a/src/cpu/o3/iew.hh b/src/cpu/o3/iew.hh index 3458f09d6..6797410d8 100644 --- a/src/cpu/o3/iew.hh +++ b/src/cpu/o3/iew.hh @@ -31,12 +31,11 @@ #ifndef __CPU_O3_IEW_HH__ #define __CPU_O3_IEW_HH__ -#include "config/full_system.hh" - #include <queue> #include "base/statistics.hh" #include "base/timebuf.hh" +#include "config/full_system.hh" #include "cpu/o3/comm.hh" #include "cpu/o3/scoreboard.hh" #include "cpu/o3/lsq.hh" @@ -139,7 +138,7 @@ class DefaultIEW void setIEWQueue(TimeBuffer<IEWStruct> *iq_ptr); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Sets pointer to the scoreboard. */ void setScoreboard(Scoreboard *sb_ptr); @@ -160,7 +159,7 @@ class DefaultIEW bool isSwitchedOut() { return switchedOut; } /** Squashes instructions in IEW for a specific thread. */ - void squash(unsigned tid); + void squash(ThreadID tid); /** Wakes all dependents of a completed instruction. */ void wakeDependents(DynInstPtr &inst); @@ -177,7 +176,7 @@ class DefaultIEW void instToCommit(DynInstPtr &inst); /** Inserts unused instructions of a thread into the skid buffer. */ - void skidInsert(unsigned tid); + void skidInsert(ThreadID tid); /** Returns the max of the number of entries in all of the skid buffers. */ int skidCount(); @@ -209,7 +208,7 @@ class DefaultIEW bool hasStoresToWB() { return ldstQueue.hasStoresToWB(); } /** Returns if the LSQ has any stores to writeback. */ - bool hasStoresToWB(unsigned tid) { return ldstQueue.hasStoresToWB(tid); } + bool hasStoresToWB(ThreadID tid) { return ldstQueue.hasStoresToWB(tid); } void incrWb(InstSeqNum &sn) { @@ -256,31 +255,31 @@ class DefaultIEW /** Sends commit proper information for a squash due to a branch * mispredict. */ - void squashDueToBranch(DynInstPtr &inst, unsigned thread_id); + void squashDueToBranch(DynInstPtr &inst, ThreadID tid); /** Sends commit proper information for a squash due to a memory order * violation. */ - void squashDueToMemOrder(DynInstPtr &inst, unsigned thread_id); + void squashDueToMemOrder(DynInstPtr &inst, ThreadID tid); /** Sends commit proper information for a squash due to memory becoming * blocked (younger issued instructions must be retried). */ - void squashDueToMemBlocked(DynInstPtr &inst, unsigned thread_id); + void squashDueToMemBlocked(DynInstPtr &inst, ThreadID tid); /** Sets Dispatch to blocked, and signals back to other stages to block. */ - void block(unsigned thread_id); + void block(ThreadID tid); /** Unblocks Dispatch if the skid buffer is empty, and signals back to * other stages to unblock. */ - void unblock(unsigned thread_id); + void unblock(ThreadID tid); /** Determines proper actions to take given Dispatch's status. */ - void dispatch(unsigned tid); + void dispatch(ThreadID tid); /** Dispatches instructions to IQ and LSQ. */ - void dispatchInsts(unsigned tid); + void dispatchInsts(ThreadID tid); /** Executes instructions. In the case of memory operations, it informs the * LSQ to execute the instructions. Also handles any redirects that occur @@ -301,16 +300,16 @@ class DefaultIEW unsigned validInstsFromRename(); /** Reads the stall signals. */ - void readStallSignals(unsigned tid); + void readStallSignals(ThreadID tid); /** Checks if any of the stall conditions are currently true. */ - bool checkStall(unsigned tid); + bool checkStall(ThreadID tid); /** Processes inputs and changes state accordingly. */ - void checkSignalsAndUpdate(unsigned tid); + void checkSignalsAndUpdate(ThreadID tid); /** Removes instructions from rename from a thread's instruction list. */ - void emptyRenameInsts(unsigned tid); + void emptyRenameInsts(ThreadID tid); /** Sorts instructions coming from rename into lists separated by thread. */ void sortInsts(); @@ -453,10 +452,10 @@ class DefaultIEW unsigned wbMax; /** Number of active threads. */ - unsigned numThreads; + ThreadID numThreads; /** Pointer to list of active threads. */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Maximum size of the skid buffer. */ unsigned skidBufferMax; diff --git a/src/cpu/o3/iew_impl.hh b/src/cpu/o3/iew_impl.hh index 778ff444f..ba29df196 100644 --- a/src/cpu/o3/iew_impl.hh +++ b/src/cpu/o3/iew_impl.hh @@ -39,6 +39,8 @@ #include "cpu/o3/iew.hh" #include "params/DerivO3CPU.hh" +using namespace std; + template<class Impl> DefaultIEW<Impl>::DefaultIEW(O3CPU *_cpu, DerivO3CPUParams *params) : issueToExecQueue(params->backComSize, params->forwardComSize), @@ -66,10 +68,10 @@ DefaultIEW<Impl>::DefaultIEW(O3CPU *_cpu, DerivO3CPUParams *params) // Instruction queue needs the queue between issue and execute. instQueue.setIssueToExecuteQueue(&issueToExecQueue); - for (int i=0; i < numThreads; i++) { - dispatchStatus[i] = Running; - stalls[i].commit = false; - fetchRedirect[i] = false; + for (ThreadID tid = 0; tid < numThreads; tid++) { + dispatchStatus[tid] = Running; + stalls[tid].commit = false; + fetchRedirect[tid] = false; } wbMax = wbWidth * params->wbDepth; @@ -164,7 +166,7 @@ DefaultIEW<Impl>::regStats() .desc("Number of executed instructions"); iewExecLoadInsts - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".iewExecLoadInsts") .desc("Number of load instructions executed") .flags(total); @@ -174,25 +176,25 @@ DefaultIEW<Impl>::regStats() .desc("Number of squashed instructions skipped in execute"); iewExecutedSwp - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:swp") .desc("number of swp insts executed") .flags(total); iewExecutedNop - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:nop") .desc("number of nop insts executed") .flags(total); iewExecutedRefs - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:refs") .desc("number of memory reference insts executed") .flags(total); iewExecutedBranches - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".EXEC:branches") .desc("Number of branches executed") .flags(total); @@ -211,31 +213,31 @@ DefaultIEW<Impl>::regStats() iewExecRate = iewExecutedInsts / cpu->numCycles; iewInstsToCommit - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:sent") .desc("cumulative count of insts sent to commit") .flags(total); writebackCount - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:count") .desc("cumulative count of insts written-back") .flags(total); producerInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:producers") .desc("num instructions producing a value") .flags(total); consumerInst - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:consumers") .desc("num instructions consuming a value") .flags(total); wbPenalized - .init(cpu->number_of_threads) + .init(cpu->numThreads) .name(name() + ".WB:penalized") .desc("number of instrctions required to write to 'other' IQ") .flags(total); @@ -265,7 +267,7 @@ template<class Impl> void DefaultIEW<Impl>::initStage() { - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { toRename->iewInfo[tid].usedIQ = true; toRename->iewInfo[tid].freeIQEntries = instQueue.numFreeEntries(tid); @@ -318,7 +320,7 @@ DefaultIEW<Impl>::setIEWQueue(TimeBuffer<IEWStruct> *iq_ptr) template<class Impl> void -DefaultIEW<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +DefaultIEW<Impl>::setActiveThreads(list<ThreadID> *at_ptr) { activeThreads = at_ptr; @@ -361,11 +363,11 @@ DefaultIEW<Impl>::switchOut() ldstQueue.switchOut(); fuPool->switchOut(); - for (int i = 0; i < numThreads; i++) { - while (!insts[i].empty()) - insts[i].pop(); - while (!skidBuffer[i].empty()) - skidBuffer[i].pop(); + for (ThreadID tid = 0; tid < numThreads; tid++) { + while (!insts[tid].empty()) + insts[tid].pop(); + while (!skidBuffer[tid].empty()) + skidBuffer[tid].pop(); } } @@ -386,10 +388,10 @@ DefaultIEW<Impl>::takeOverFrom() initStage(); cpu->activityThisCycle(); - for (int i=0; i < numThreads; i++) { - dispatchStatus[i] = Running; - stalls[i].commit = false; - fetchRedirect[i] = false; + for (ThreadID tid = 0; tid < numThreads; tid++) { + dispatchStatus[tid] = Running; + stalls[tid].commit = false; + fetchRedirect[tid] = false; } updateLSQNextCycle = false; @@ -401,10 +403,9 @@ DefaultIEW<Impl>::takeOverFrom() template<class Impl> void -DefaultIEW<Impl>::squash(unsigned tid) +DefaultIEW<Impl>::squash(ThreadID tid) { - DPRINTF(IEW, "[tid:%i]: Squashing all instructions.\n", - tid); + DPRINTF(IEW, "[tid:%i]: Squashing all instructions.\n", tid); // Tell the IQ to start squashing. instQueue.squash(tid); @@ -433,7 +434,7 @@ DefaultIEW<Impl>::squash(unsigned tid) template<class Impl> void -DefaultIEW<Impl>::squashDueToBranch(DynInstPtr &inst, unsigned tid) +DefaultIEW<Impl>::squashDueToBranch(DynInstPtr &inst, ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Squashing from a specific instruction, PC: %#x " "[sn:%i].\n", tid, inst->readPC(), inst->seqNum); @@ -464,7 +465,7 @@ DefaultIEW<Impl>::squashDueToBranch(DynInstPtr &inst, unsigned tid) template<class Impl> void -DefaultIEW<Impl>::squashDueToMemOrder(DynInstPtr &inst, unsigned tid) +DefaultIEW<Impl>::squashDueToMemOrder(DynInstPtr &inst, ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Squashing from a specific instruction, " "PC: %#x [sn:%i].\n", tid, inst->readPC(), inst->seqNum); @@ -482,7 +483,7 @@ DefaultIEW<Impl>::squashDueToMemOrder(DynInstPtr &inst, unsigned tid) template<class Impl> void -DefaultIEW<Impl>::squashDueToMemBlocked(DynInstPtr &inst, unsigned tid) +DefaultIEW<Impl>::squashDueToMemBlocked(DynInstPtr &inst, ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Memory blocked, squashing load and younger insts, " "PC: %#x [sn:%i].\n", tid, inst->readPC(), inst->seqNum); @@ -503,7 +504,7 @@ DefaultIEW<Impl>::squashDueToMemBlocked(DynInstPtr &inst, unsigned tid) template<class Impl> void -DefaultIEW<Impl>::block(unsigned tid) +DefaultIEW<Impl>::block(ThreadID tid) { DPRINTF(IEW, "[tid:%u]: Blocking.\n", tid); @@ -522,7 +523,7 @@ DefaultIEW<Impl>::block(unsigned tid) template<class Impl> void -DefaultIEW<Impl>::unblock(unsigned tid) +DefaultIEW<Impl>::unblock(ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Reading instructions out of the skid " "buffer %u.\n",tid, tid); @@ -605,7 +606,7 @@ DefaultIEW<Impl>::validInstsFromRename() template<class Impl> void -DefaultIEW<Impl>::skidInsert(unsigned tid) +DefaultIEW<Impl>::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -631,11 +632,11 @@ DefaultIEW<Impl>::skidCount() { int max=0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; unsigned thread_count = skidBuffer[tid].size(); if (max < thread_count) max = thread_count; @@ -648,11 +649,11 @@ template<class Impl> bool DefaultIEW<Impl>::skidsEmpty() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!skidBuffer[tid].empty()) return false; @@ -667,11 +668,11 @@ DefaultIEW<Impl>::updateStatus() { bool any_unblocking = false; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (dispatchStatus[tid] == Unblocking) { any_unblocking = true; @@ -711,7 +712,7 @@ DefaultIEW<Impl>::resetEntries() template <class Impl> void -DefaultIEW<Impl>::readStallSignals(unsigned tid) +DefaultIEW<Impl>::readStallSignals(ThreadID tid) { if (fromCommit->commitBlock[tid]) { stalls[tid].commit = true; @@ -725,7 +726,7 @@ DefaultIEW<Impl>::readStallSignals(unsigned tid) template <class Impl> bool -DefaultIEW<Impl>::checkStall(unsigned tid) +DefaultIEW<Impl>::checkStall(ThreadID tid) { bool ret_val(false); @@ -761,7 +762,7 @@ DefaultIEW<Impl>::checkStall(unsigned tid) template <class Impl> void -DefaultIEW<Impl>::checkSignalsAndUpdate(unsigned tid) +DefaultIEW<Impl>::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is // Check stall signals, block if there is. @@ -834,8 +835,8 @@ DefaultIEW<Impl>::sortInsts() { int insts_from_rename = fromRename->size; #ifdef DEBUG - for (int i = 0; i < numThreads; i++) - assert(insts[i].empty()); + for (ThreadID tid = 0; tid < numThreads; tid++) + assert(insts[tid].empty()); #endif for (int i = 0; i < insts_from_rename; ++i) { insts[fromRename->insts[i]->threadNumber].push(fromRename->insts[i]); @@ -844,7 +845,7 @@ DefaultIEW<Impl>::sortInsts() template <class Impl> void -DefaultIEW<Impl>::emptyRenameInsts(unsigned tid) +DefaultIEW<Impl>::emptyRenameInsts(ThreadID tid) { DPRINTF(IEW, "[tid:%i]: Removing incoming rename instructions\n", tid); @@ -894,7 +895,7 @@ DefaultIEW<Impl>::deactivateStage() template<class Impl> void -DefaultIEW<Impl>::dispatch(unsigned tid) +DefaultIEW<Impl>::dispatch(ThreadID tid) { // If status is Running or idle, // call dispatchInsts() @@ -942,7 +943,7 @@ DefaultIEW<Impl>::dispatch(unsigned tid) template <class Impl> void -DefaultIEW<Impl>::dispatchInsts(unsigned tid) +DefaultIEW<Impl>::dispatchInsts(ThreadID tid) { // Obtain instructions from skid buffer if unblocking, or queue from rename // otherwise. @@ -1169,11 +1170,11 @@ DefaultIEW<Impl>::executeInsts() wbNumInst = 0; wbCycle = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; fetchRedirect[tid] = false; } @@ -1273,7 +1274,7 @@ DefaultIEW<Impl>::executeInsts() // This probably needs to prioritize the redirects if a different // scheduler is used. Currently the scheduler schedules the oldest // instruction first, so the branch resolution order will be correct. - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; if (!fetchRedirect[tid] || toCommit->squashedSeqNum[tid] > inst->seqNum) { @@ -1391,7 +1392,7 @@ DefaultIEW<Impl>::writebackInsts() for (int inst_num = 0; inst_num < wbWidth && toCommit->insts[inst_num]; inst_num++) { DynInstPtr inst = toCommit->insts[inst_num]; - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; DPRINTF(IEW, "Sending instructions to commit, [sn:%lli] PC %#x.\n", inst->seqNum, inst->readPC()); @@ -1439,12 +1440,12 @@ DefaultIEW<Impl>::tick() // Free function units marked as being freed this cycle. fuPool->processFreeUnits(); - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); // Check stall and squash signals, dispatch any instructions. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; DPRINTF(IEW,"Issue: Processing [tid:%i]\n",tid); @@ -1486,7 +1487,7 @@ DefaultIEW<Impl>::tick() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = (*threads++); + ThreadID tid = (*threads++); DPRINTF(IEW,"Processing [tid:%i]\n",tid); @@ -1552,14 +1553,14 @@ template <class Impl> void DefaultIEW<Impl>::updateExeInstStats(DynInstPtr &inst) { - int thread_number = inst->threadNumber; + ThreadID tid = inst->threadNumber; // // Pick off the software prefetches // #ifdef TARGET_ALPHA if (inst->isDataPrefetch()) - iewExecutedSwp[thread_number]++; + iewExecutedSwp[tid]++; else iewIewExecutedcutedInsts++; #else @@ -1570,16 +1571,16 @@ DefaultIEW<Impl>::updateExeInstStats(DynInstPtr &inst) // Control operations // if (inst->isControl()) - iewExecutedBranches[thread_number]++; + iewExecutedBranches[tid]++; // // Memory operations // if (inst->isMemRef()) { - iewExecutedRefs[thread_number]++; + iewExecutedRefs[tid]++; if (inst->isLoad()) { - iewExecLoadInsts[thread_number]++; + iewExecLoadInsts[tid]++; } } } diff --git a/src/cpu/o3/inst_queue.hh b/src/cpu/o3/inst_queue.hh index 0b814ccb4..bccb1aa0e 100644 --- a/src/cpu/o3/inst_queue.hh +++ b/src/cpu/o3/inst_queue.hh @@ -126,7 +126,7 @@ class InstructionQueue void resetState(); /** Sets active threads list. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Sets the timer buffer between issue and execute. */ void setIssueToExecuteQueue(TimeBuffer<IssueStruct> *i2eQueue); @@ -144,7 +144,7 @@ class InstructionQueue bool isSwitchedOut() { return switchedOut; } /** Number of entries needed for given amount of threads. */ - int entryAmount(int num_threads); + int entryAmount(ThreadID num_threads); /** Resets max entries for all threads. */ void resetEntries(); @@ -153,13 +153,13 @@ class InstructionQueue unsigned numFreeEntries(); /** Returns number of free entries for a thread. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns whether or not the IQ is full. */ bool isFull(); /** Returns whether or not the IQ is full for a specific thread. */ - bool isFull(unsigned tid); + bool isFull(ThreadID tid); /** Returns if there are any ready instructions in the IQ. */ bool hasReadyInsts(); @@ -203,7 +203,7 @@ class InstructionQueue * Commits all instructions up to and including the given sequence number, * for a specific thread. */ - void commit(const InstSeqNum &inst, unsigned tid = 0); + void commit(const InstSeqNum &inst, ThreadID tid = 0); /** Wakes all dependents of a completed instruction. */ int wakeDependents(DynInstPtr &completed_inst); @@ -230,17 +230,17 @@ class InstructionQueue * Squashes instructions for a thread. Squashing information is obtained * from the time buffer. */ - void squash(unsigned tid); + void squash(ThreadID tid); /** Returns the number of used entries for a thread. */ - unsigned getCount(unsigned tid) { return count[tid]; }; + unsigned getCount(ThreadID tid) { return count[tid]; }; /** Debug function to print all instructions. */ void printInsts(); private: /** Does the actual squashing. */ - void doSquash(unsigned tid); + void doSquash(ThreadID tid); ///////////////////////// // Various pointers @@ -368,10 +368,10 @@ class InstructionQueue IQPolicy iqPolicy; /** Number of Total Threads*/ - unsigned numThreads; + ThreadID numThreads; /** Pointer to list of active threads. */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Per Thread IQ count */ unsigned count[Impl::MaxThreads]; diff --git a/src/cpu/o3/inst_queue_impl.hh b/src/cpu/o3/inst_queue_impl.hh index 1d0f4b9f6..b6d1ec8b0 100644 --- a/src/cpu/o3/inst_queue_impl.hh +++ b/src/cpu/o3/inst_queue_impl.hh @@ -35,9 +35,10 @@ #include "cpu/o3/fu_pool.hh" #include "cpu/o3/inst_queue.hh" #include "enums/OpClass.hh" +#include "params/DerivO3CPU.hh" #include "sim/core.hh" -#include "params/DerivO3CPU.hh" +using namespace std; template <class Impl> InstructionQueue<Impl>::FUCompletion::FUCompletion(DynInstPtr &_inst, @@ -93,9 +94,9 @@ InstructionQueue<Impl>::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, regScoreboard.resize(numPhysRegs); //Initialize Mem Dependence Units - for (int i = 0; i < numThreads; i++) { - memDepUnit[i].init(params,i); - memDepUnit[i].setIQ(this); + for (ThreadID tid = 0; tid < numThreads; tid++) { + memDepUnit[tid].init(params, tid); + memDepUnit[tid].setIQ(this); } resetState(); @@ -111,8 +112,8 @@ InstructionQueue<Impl>::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, iqPolicy = Dynamic; //Set Max Entries to Total ROB Capacity - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = numEntries; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = numEntries; } } else if (policy == "partitioned") { @@ -122,8 +123,8 @@ InstructionQueue<Impl>::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, int part_amt = numEntries / numThreads; //Divide ROB up evenly - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = part_amt; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = part_amt; } DPRINTF(IQ, "IQ sharing policy set to Partitioned:" @@ -136,8 +137,8 @@ InstructionQueue<Impl>::InstructionQueue(O3CPU *cpu_ptr, IEW *iew_ptr, int thresholdIQ = (int)((double)threshold * numEntries); //Divide up by threshold amount - for (int i = 0; i < numThreads; i++) { - maxEntries[i] = thresholdIQ; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = thresholdIQ; } DPRINTF(IQ, "IQ sharing policy set to Threshold:" @@ -315,9 +316,9 @@ InstructionQueue<Impl>::regStats() ; fuBusyRate = fuBusy / iqInstsIssued; - for ( int i=0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { // Tell mem dependence unit to reg stats as well. - memDepUnit[i].regStats(); + memDepUnit[tid].regStats(); } } @@ -326,9 +327,9 @@ void InstructionQueue<Impl>::resetState() { //Initialize thread IQ counts - for (int i = 0; i <numThreads; i++) { - count[i] = 0; - instList[i].clear(); + for (ThreadID tid = 0; tid <numThreads; tid++) { + count[tid] = 0; + instList[tid].clear(); } // Initialize the number of free IQ entries. @@ -343,8 +344,8 @@ InstructionQueue<Impl>::resetState() regScoreboard[i] = false; } - for (int i = 0; i < numThreads; ++i) { - squashedSeqNum[i] = 0; + for (ThreadID tid = 0; tid < numThreads; ++tid) { + squashedSeqNum[tid] = 0; } for (int i = 0; i < Num_OpClasses; ++i) { @@ -359,7 +360,7 @@ InstructionQueue<Impl>::resetState() template <class Impl> void -InstructionQueue<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +InstructionQueue<Impl>::setActiveThreads(list<ThreadID> *at_ptr) { activeThreads = at_ptr; } @@ -395,8 +396,8 @@ InstructionQueue<Impl>::switchOut() dependGraph.reset(); instsToExecute.clear(); switchedOut = true; - for (int i = 0; i < numThreads; ++i) { - memDepUnit[i].switchOut(); + for (ThreadID tid = 0; tid < numThreads; ++tid) { + memDepUnit[tid].switchOut(); } } @@ -409,7 +410,7 @@ InstructionQueue<Impl>::takeOverFrom() template <class Impl> int -InstructionQueue<Impl>::entryAmount(int num_threads) +InstructionQueue<Impl>::entryAmount(ThreadID num_threads) { if (iqPolicy == Partitioned) { return numEntries / num_threads; @@ -426,11 +427,11 @@ InstructionQueue<Impl>::resetEntries() if (iqPolicy != Dynamic || numThreads > 1) { int active_threads = activeThreads->size(); - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (iqPolicy == Partitioned) { maxEntries[tid] = numEntries / active_threads; @@ -450,7 +451,7 @@ InstructionQueue<Impl>::numFreeEntries() template <class Impl> unsigned -InstructionQueue<Impl>::numFreeEntries(unsigned tid) +InstructionQueue<Impl>::numFreeEntries(ThreadID tid) { return maxEntries[tid] - count[tid]; } @@ -470,7 +471,7 @@ InstructionQueue<Impl>::isFull() template <class Impl> bool -InstructionQueue<Impl>::isFull(unsigned tid) +InstructionQueue<Impl>::isFull(ThreadID tid) { if (numFreeEntries(tid) == 0) { return(true); @@ -726,7 +727,7 @@ InstructionQueue<Impl>::scheduleReadyInsts() int idx = -2; int op_latency = 1; - int tid = issuing_inst->threadNumber; + ThreadID tid = issuing_inst->threadNumber; if (op_class != No_OpClass) { idx = fuPool->getUnit(op_class); @@ -825,7 +826,7 @@ InstructionQueue<Impl>::scheduleNonSpec(const InstSeqNum &inst) assert(inst_it != nonSpecInsts.end()); - unsigned tid = (*inst_it).second->threadNumber; + ThreadID tid = (*inst_it).second->threadNumber; (*inst_it).second->setAtCommit(); @@ -844,7 +845,7 @@ InstructionQueue<Impl>::scheduleNonSpec(const InstSeqNum &inst) template <class Impl> void -InstructionQueue<Impl>::commit(const InstSeqNum &inst, unsigned tid) +InstructionQueue<Impl>::commit(const InstSeqNum &inst, ThreadID tid) { DPRINTF(IQ, "[tid:%i]: Committing instructions older than [sn:%i]\n", tid,inst); @@ -976,7 +977,7 @@ template <class Impl> void InstructionQueue<Impl>::completeMemInst(DynInstPtr &completed_inst) { - int tid = completed_inst->threadNumber; + ThreadID tid = completed_inst->threadNumber; DPRINTF(IQ, "Completing mem instruction PC:%#x [sn:%lli]\n", completed_inst->readPC(), completed_inst->seqNum); @@ -999,7 +1000,7 @@ InstructionQueue<Impl>::violation(DynInstPtr &store, template <class Impl> void -InstructionQueue<Impl>::squash(unsigned tid) +InstructionQueue<Impl>::squash(ThreadID tid) { DPRINTF(IQ, "[tid:%i]: Starting to squash instructions in " "the IQ.\n", tid); @@ -1019,7 +1020,7 @@ InstructionQueue<Impl>::squash(unsigned tid) template <class Impl> void -InstructionQueue<Impl>::doSquash(unsigned tid) +InstructionQueue<Impl>::doSquash(ThreadID tid) { // Start at the tail. ListIt squash_it = instList[tid].end(); @@ -1253,10 +1254,10 @@ InstructionQueue<Impl>::countInsts() // Change the #if if you want to use this method. int total_insts = 0; - for (int i = 0; i < numThreads; ++i) { - ListIt count_it = instList[i].begin(); + for (ThreadID tid = 0; tid < numThreads; ++tid) { + ListIt count_it = instList[tid].begin(); - while (count_it != instList[i].end()) { + while (count_it != instList[tid].end()) { if (!(*count_it)->isSquashed() && !(*count_it)->isSquashedInIQ()) { if (!(*count_it)->isIssued()) { ++total_insts; @@ -1325,15 +1326,13 @@ template <class Impl> void InstructionQueue<Impl>::dumpInsts() { - for (int i = 0; i < numThreads; ++i) { + for (ThreadID tid = 0; tid < numThreads; ++tid) { int num = 0; int valid_num = 0; - ListIt inst_list_it = instList[i].begin(); + ListIt inst_list_it = instList[tid].begin(); - while (inst_list_it != instList[i].end()) - { - cprintf("Instruction:%i\n", - num); + while (inst_list_it != instList[tid].end()) { + cprintf("Instruction:%i\n", num); if (!(*inst_list_it)->isSquashed()) { if (!(*inst_list_it)->isIssued()) { ++valid_num; diff --git a/src/cpu/o3/lsq.hh b/src/cpu/o3/lsq.hh index cf27552d4..a0bae058c 100644 --- a/src/cpu/o3/lsq.hh +++ b/src/cpu/o3/lsq.hh @@ -74,24 +74,24 @@ class LSQ { Port *getDcachePort() { return &dcachePort; } /** Sets the pointer to the list of active threads. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Switches out the LSQ. */ void switchOut(); /** Takes over execution from another CPU's thread. */ void takeOverFrom(); /** Number of entries needed for the given amount of threads.*/ - int entryAmount(int num_threads); - void removeEntries(unsigned tid); + int entryAmount(ThreadID num_threads); + void removeEntries(ThreadID tid); /** Reset the max entries for each thread. */ void resetEntries(); /** Resize the max entries for a thread. */ - void resizeEntries(unsigned size, unsigned tid); + void resizeEntries(unsigned size, ThreadID tid); /** Ticks the LSQ. */ void tick(); /** Ticks a specific LSQ Unit. */ - void tick(unsigned tid) + void tick(ThreadID tid) { thread[tid].tick(); } /** Inserts a load into the LSQ. */ @@ -108,13 +108,13 @@ class LSQ { /** * Commits loads up until the given sequence number for a specific thread. */ - void commitLoads(InstSeqNum &youngest_inst, unsigned tid) + void commitLoads(InstSeqNum &youngest_inst, ThreadID tid) { thread[tid].commitLoads(youngest_inst); } /** * Commits stores up until the given sequence number for a specific thread. */ - void commitStores(InstSeqNum &youngest_inst, unsigned tid) + void commitStores(InstSeqNum &youngest_inst, ThreadID tid) { thread[tid].commitStores(youngest_inst); } /** @@ -123,12 +123,12 @@ class LSQ { */ void writebackStores(); /** Same as above, but only for one thread. */ - void writebackStores(unsigned tid); + void writebackStores(ThreadID tid); /** * Squash instructions from a thread until the specified sequence number. */ - void squash(const InstSeqNum &squashed_num, unsigned tid) + void squash(const InstSeqNum &squashed_num, ThreadID tid) { thread[tid].squash(squashed_num); } /** Returns whether or not there was a memory ordering violation. */ @@ -137,41 +137,41 @@ class LSQ { * Returns whether or not there was a memory ordering violation for a * specific thread. */ - bool violation(unsigned tid) + bool violation(ThreadID tid) { return thread[tid].violation(); } /** Returns if a load is blocked due to the memory system for a specific * thread. */ - bool loadBlocked(unsigned tid) + bool loadBlocked(ThreadID tid) { return thread[tid].loadBlocked(); } - bool isLoadBlockedHandled(unsigned tid) + bool isLoadBlockedHandled(ThreadID tid) { return thread[tid].isLoadBlockedHandled(); } - void setLoadBlockedHandled(unsigned tid) + void setLoadBlockedHandled(ThreadID tid) { thread[tid].setLoadBlockedHandled(); } /** Gets the instruction that caused the memory ordering violation. */ - DynInstPtr getMemDepViolator(unsigned tid) + DynInstPtr getMemDepViolator(ThreadID tid) { return thread[tid].getMemDepViolator(); } /** Returns the head index of the load queue for a specific thread. */ - int getLoadHead(unsigned tid) + int getLoadHead(ThreadID tid) { return thread[tid].getLoadHead(); } /** Returns the sequence number of the head of the load queue. */ - InstSeqNum getLoadHeadSeqNum(unsigned tid) + InstSeqNum getLoadHeadSeqNum(ThreadID tid) { return thread[tid].getLoadHeadSeqNum(); } /** Returns the head index of the store queue. */ - int getStoreHead(unsigned tid) + int getStoreHead(ThreadID tid) { return thread[tid].getStoreHead(); } /** Returns the sequence number of the head of the store queue. */ - InstSeqNum getStoreHeadSeqNum(unsigned tid) + InstSeqNum getStoreHeadSeqNum(ThreadID tid) { return thread[tid].getStoreHeadSeqNum(); } @@ -179,31 +179,31 @@ class LSQ { /** Returns the number of instructions in all of the queues. */ int getCount(); /** Returns the number of instructions in the queues of one thread. */ - int getCount(unsigned tid) + int getCount(ThreadID tid) { return thread[tid].getCount(); } /** Returns the total number of loads in the load queue. */ int numLoads(); /** Returns the total number of loads for a single thread. */ - int numLoads(unsigned tid) + int numLoads(ThreadID tid) { return thread[tid].numLoads(); } /** Returns the total number of stores in the store queue. */ int numStores(); /** Returns the total number of stores for a single thread. */ - int numStores(unsigned tid) + int numStores(ThreadID tid) { return thread[tid].numStores(); } /** Returns the total number of loads that are ready. */ int numLoadsReady(); /** Returns the number of loads that are ready for a single thread. */ - int numLoadsReady(unsigned tid) + int numLoadsReady(ThreadID tid) { return thread[tid].numLoadsReady(); } /** Returns the number of free entries. */ unsigned numFreeEntries(); /** Returns the number of free entries for a specific thread. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns if the LSQ is full (either LQ or SQ is full). */ bool isFull(); @@ -211,17 +211,17 @@ class LSQ { * Returns if the LSQ is full for a specific thread (either LQ or SQ is * full). */ - bool isFull(unsigned tid); + bool isFull(ThreadID tid); /** Returns if any of the LQs are full. */ bool lqFull(); /** Returns if the LQ of a given thread is full. */ - bool lqFull(unsigned tid); + bool lqFull(ThreadID tid); /** Returns if any of the SQs are full. */ bool sqFull(); /** Returns if the SQ of a given thread is full. */ - bool sqFull(unsigned tid); + bool sqFull(ThreadID tid); /** * Returns if the LSQ is stalled due to a memory operation that must be @@ -232,7 +232,7 @@ class LSQ { * Returns if the LSQ of a specific thread is stalled due to a memory * operation that must be replayed. */ - bool isStalled(unsigned tid); + bool isStalled(ThreadID tid); /** Returns whether or not there are any stores to write back to memory. */ bool hasStoresToWB(); @@ -240,11 +240,11 @@ class LSQ { /** Returns whether or not a specific thread has any stores to write back * to memory. */ - bool hasStoresToWB(unsigned tid) + bool hasStoresToWB(ThreadID tid) { return thread[tid].hasStoresToWB(); } /** Returns the number of stores a specific thread has to write back. */ - int numStoresToWB(unsigned tid) + int numStoresToWB(ThreadID tid) { return thread[tid].numStoresToWB(); } /** Returns if the LSQ will write back to memory this cycle. */ @@ -252,22 +252,22 @@ class LSQ { /** Returns if the LSQ of a specific thread will write back to memory this * cycle. */ - bool willWB(unsigned tid) + bool willWB(ThreadID tid) { return thread[tid].willWB(); } /** Returns if the cache is currently blocked. */ bool cacheBlocked() - { return retryTid != -1; } + { return retryTid != InvalidThreadID; } /** Sets the retry thread id, indicating that one of the LSQUnits * tried to access the cache but the cache was blocked. */ - void setRetryTid(int tid) + void setRetryTid(ThreadID tid) { retryTid = tid; } /** Debugging function to print out all instructions. */ void dumpInsts(); /** Debugging function to print out instructions from a specific thread. */ - void dumpInsts(unsigned tid) + void dumpInsts(ThreadID tid) { thread[tid].dumpInsts(); } /** Executes a read operation, using the load specified at the load index. */ @@ -345,7 +345,7 @@ class LSQ { LSQUnit thread[Impl::MaxThreads]; /** List of Active Threads in System. */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Total Size of LQ Entries. */ unsigned LQEntries; @@ -359,11 +359,11 @@ class LSQ { unsigned maxSQEntries; /** Number of Threads. */ - unsigned numThreads; + ThreadID numThreads; /** The thread id of the LSQ Unit that is currently waiting for a * retry. */ - int retryTid; + ThreadID retryTid; }; template <class Impl> @@ -371,7 +371,7 @@ template <class T> Fault LSQ<Impl>::read(RequestPtr req, T &data, int load_idx) { - unsigned tid = req->threadId(); + ThreadID tid = req->threadId(); return thread[tid].read(req, data, load_idx); } @@ -381,7 +381,7 @@ template <class T> Fault LSQ<Impl>::write(RequestPtr req, T &data, int store_idx) { - unsigned tid = req->threadId(); + ThreadID tid = req->threadId(); return thread[tid].write(req, data, store_idx); } diff --git a/src/cpu/o3/lsq_impl.hh b/src/cpu/o3/lsq_impl.hh index 8f9f63081..e780b14e4 100644 --- a/src/cpu/o3/lsq_impl.hh +++ b/src/cpu/o3/lsq_impl.hh @@ -33,9 +33,10 @@ #include <string> #include "cpu/o3/lsq.hh" - #include "params/DerivO3CPU.hh" +using namespace std; + template<class Impl> void LSQ<Impl>::DcachePort::setPeer(Port *port) @@ -170,7 +171,7 @@ LSQ<Impl>::LSQ(O3CPU *cpu_ptr, IEW *iew_ptr, DerivO3CPUParams *params) } //Initialize LSQs - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].init(cpu, iew_ptr, params, this, maxLQEntries, maxSQEntries, tid); thread[tid].setDcachePort(&dcachePort); @@ -190,14 +191,14 @@ void LSQ<Impl>::regStats() { //Initialize LSQs - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].regStats(); } } template<class Impl> void -LSQ<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +LSQ<Impl>::setActiveThreads(list<ThreadID> *at_ptr) { activeThreads = at_ptr; assert(activeThreads != 0); @@ -207,7 +208,7 @@ template <class Impl> void LSQ<Impl>::switchOut() { - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].switchOut(); } } @@ -216,14 +217,14 @@ template <class Impl> void LSQ<Impl>::takeOverFrom() { - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { thread[tid].takeOverFrom(); } } template <class Impl> int -LSQ<Impl>::entryAmount(int num_threads) +LSQ<Impl>::entryAmount(ThreadID num_threads) { if (lsqPolicy == Partitioned) { return LQEntries / num_threads; @@ -249,11 +250,11 @@ LSQ<Impl>::resetEntries() maxEntries = LQEntries; } - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; resizeEntries(maxEntries, tid); } @@ -262,7 +263,7 @@ LSQ<Impl>::resetEntries() template<class Impl> void -LSQ<Impl>::removeEntries(unsigned tid) +LSQ<Impl>::removeEntries(ThreadID tid) { thread[tid].clearLQ(); thread[tid].clearSQ(); @@ -270,7 +271,7 @@ LSQ<Impl>::removeEntries(unsigned tid) template<class Impl> void -LSQ<Impl>::resizeEntries(unsigned size,unsigned tid) +LSQ<Impl>::resizeEntries(unsigned size, ThreadID tid) { thread[tid].resizeLQ(size); thread[tid].resizeSQ(size); @@ -280,11 +281,11 @@ template<class Impl> void LSQ<Impl>::tick() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; thread[tid].tick(); } @@ -294,7 +295,7 @@ template<class Impl> void LSQ<Impl>::insertLoad(DynInstPtr &load_inst) { - unsigned tid = load_inst->threadNumber; + ThreadID tid = load_inst->threadNumber; thread[tid].insertLoad(load_inst); } @@ -303,7 +304,7 @@ template<class Impl> void LSQ<Impl>::insertStore(DynInstPtr &store_inst) { - unsigned tid = store_inst->threadNumber; + ThreadID tid = store_inst->threadNumber; thread[tid].insertStore(store_inst); } @@ -312,7 +313,7 @@ template<class Impl> Fault LSQ<Impl>::executeLoad(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; return thread[tid].executeLoad(inst); } @@ -321,7 +322,7 @@ template<class Impl> Fault LSQ<Impl>::executeStore(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; return thread[tid].executeStore(inst); } @@ -330,11 +331,11 @@ template<class Impl> void LSQ<Impl>::writebackStores() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (numStoresToWB(tid) > 0) { DPRINTF(Writeback,"[tid:%i] Writing back stores. %i stores " @@ -350,11 +351,11 @@ bool LSQ<Impl>::violation() { /* Answers: Does Anybody Have a Violation?*/ - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (thread[tid].violation()) return true; @@ -369,11 +370,11 @@ LSQ<Impl>::getCount() { unsigned total = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += getCount(tid); } @@ -387,11 +388,11 @@ LSQ<Impl>::numLoads() { unsigned total = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += numLoads(tid); } @@ -405,11 +406,11 @@ LSQ<Impl>::numStores() { unsigned total = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += thread[tid].numStores(); } @@ -423,11 +424,11 @@ LSQ<Impl>::numLoadsReady() { unsigned total = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += thread[tid].numLoadsReady(); } @@ -441,11 +442,11 @@ LSQ<Impl>::numFreeEntries() { unsigned total = 0; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; total += thread[tid].numFreeEntries(); } @@ -455,7 +456,7 @@ LSQ<Impl>::numFreeEntries() template<class Impl> unsigned -LSQ<Impl>::numFreeEntries(unsigned tid) +LSQ<Impl>::numFreeEntries(ThreadID tid) { //if (lsqPolicy == Dynamic) //return numFreeEntries(); @@ -467,11 +468,11 @@ template<class Impl> bool LSQ<Impl>::isFull() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!(thread[tid].lqFull() || thread[tid].sqFull())) return false; @@ -482,7 +483,7 @@ LSQ<Impl>::isFull() template<class Impl> bool -LSQ<Impl>::isFull(unsigned tid) +LSQ<Impl>::isFull(ThreadID tid) { //@todo: Change to Calculate All Entries for //Dynamic Policy @@ -496,11 +497,11 @@ template<class Impl> bool LSQ<Impl>::lqFull() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!thread[tid].lqFull()) return false; @@ -511,7 +512,7 @@ LSQ<Impl>::lqFull() template<class Impl> bool -LSQ<Impl>::lqFull(unsigned tid) +LSQ<Impl>::lqFull(ThreadID tid) { //@todo: Change to Calculate All Entries for //Dynamic Policy @@ -525,11 +526,11 @@ template<class Impl> bool LSQ<Impl>::sqFull() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!sqFull(tid)) return false; @@ -540,7 +541,7 @@ LSQ<Impl>::sqFull() template<class Impl> bool -LSQ<Impl>::sqFull(unsigned tid) +LSQ<Impl>::sqFull(ThreadID tid) { //@todo: Change to Calculate All Entries for //Dynamic Policy @@ -554,11 +555,11 @@ template<class Impl> bool LSQ<Impl>::isStalled() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!thread[tid].isStalled()) return false; @@ -569,7 +570,7 @@ LSQ<Impl>::isStalled() template<class Impl> bool -LSQ<Impl>::isStalled(unsigned tid) +LSQ<Impl>::isStalled(ThreadID tid) { if (lsqPolicy == Dynamic) return isStalled(); @@ -581,11 +582,11 @@ template<class Impl> bool LSQ<Impl>::hasStoresToWB() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (hasStoresToWB(tid)) return true; @@ -598,11 +599,11 @@ template<class Impl> bool LSQ<Impl>::willWB() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (willWB(tid)) return true; @@ -615,11 +616,11 @@ template<class Impl> void LSQ<Impl>::dumpInsts() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; thread[tid].dumpInsts(); } diff --git a/src/cpu/o3/lsq_unit.hh b/src/cpu/o3/lsq_unit.hh index 28e6f4506..a917caef3 100644 --- a/src/cpu/o3/lsq_unit.hh +++ b/src/cpu/o3/lsq_unit.hh @@ -325,7 +325,7 @@ class LSQUnit { private: /** The LSQUnit thread id. */ - unsigned lsqID; + ThreadID lsqID; /** The store queue. */ std::vector<SQEntry> storeQueue; diff --git a/src/cpu/o3/lsq_unit_impl.hh b/src/cpu/o3/lsq_unit_impl.hh index afc9faf9c..edc8c9b3f 100644 --- a/src/cpu/o3/lsq_unit_impl.hh +++ b/src/cpu/o3/lsq_unit_impl.hh @@ -938,7 +938,7 @@ LSQUnit<Impl>::recvRetry() storePostSend(retryPkt); retryPkt = NULL; isStoreBlocked = false; - lsq->setRetryTid(-1); + lsq->setRetryTid(InvalidThreadID); } else { // Still blocked! ++lsqCacheBlocked; diff --git a/src/cpu/o3/mem_dep_unit.hh b/src/cpu/o3/mem_dep_unit.hh index 4f9e7c9f7..a9560f446 100644 --- a/src/cpu/o3/mem_dep_unit.hh +++ b/src/cpu/o3/mem_dep_unit.hh @@ -86,7 +86,7 @@ class MemDepUnit std::string name() const { return _name; } /** Initializes the unit with parameters and a thread id. */ - void init(DerivO3CPUParams *params, int tid); + void init(DerivO3CPUParams *params, ThreadID tid); /** Registers statistics. */ void regStats(); @@ -135,7 +135,7 @@ class MemDepUnit /** Squashes all instructions up until a given sequence number for a * specific thread. */ - void squash(const InstSeqNum &squashed_num, unsigned tid); + void squash(const InstSeqNum &squashed_num, ThreadID tid); /** Indicates an ordering violation between a store and a younger load. */ void violation(DynInstPtr &store_inst, DynInstPtr &violating_load); diff --git a/src/cpu/o3/mem_dep_unit_impl.hh b/src/cpu/o3/mem_dep_unit_impl.hh index 8754539f9..59344d9f7 100644 --- a/src/cpu/o3/mem_dep_unit_impl.hh +++ b/src/cpu/o3/mem_dep_unit_impl.hh @@ -54,7 +54,7 @@ MemDepUnit<MemDepPred, Impl>::MemDepUnit(DerivO3CPUParams *params) template <class MemDepPred, class Impl> MemDepUnit<MemDepPred, Impl>::~MemDepUnit() { - for (int tid=0; tid < Impl::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < Impl::MaxThreads; tid++) { ListIt inst_list_it = instList[tid].begin(); @@ -78,7 +78,7 @@ MemDepUnit<MemDepPred, Impl>::~MemDepUnit() template <class MemDepPred, class Impl> void -MemDepUnit<MemDepPred, Impl>::init(DerivO3CPUParams *params, int tid) +MemDepUnit<MemDepPred, Impl>::init(DerivO3CPUParams *params, ThreadID tid) { DPRINTF(MemDepUnit, "Creating MemDepUnit %i object.\n",tid); @@ -145,7 +145,7 @@ template <class MemDepPred, class Impl> void MemDepUnit<MemDepPred, Impl>::insert(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; MemDepEntryPtr inst_entry = new MemDepEntry(inst); @@ -242,7 +242,7 @@ template <class MemDepPred, class Impl> void MemDepUnit<MemDepPred, Impl>::insertNonSpec(DynInstPtr &inst) { - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; MemDepEntryPtr inst_entry = new MemDepEntry(inst); @@ -292,7 +292,7 @@ MemDepUnit<MemDepPred, Impl>::insertBarrier(DynInstPtr &barr_inst) DPRINTF(MemDepUnit, "Inserted a write barrier\n"); } - unsigned tid = barr_inst->threadNumber; + ThreadID tid = barr_inst->threadNumber; MemDepEntryPtr inst_entry = new MemDepEntry(barr_inst); @@ -382,7 +382,7 @@ MemDepUnit<MemDepPred, Impl>::completed(DynInstPtr &inst) "[sn:%lli].\n", inst->readPC(), inst->seqNum); - unsigned tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; // Remove the instruction from the hash and the list. MemDepHashIt hash_it = memDepHash.find(inst->seqNum); @@ -457,7 +457,7 @@ MemDepUnit<MemDepPred, Impl>::wakeDependents(DynInstPtr &inst) template <class MemDepPred, class Impl> void MemDepUnit<MemDepPred, Impl>::squash(const InstSeqNum &squashed_num, - unsigned tid) + ThreadID tid) { if (!instsToReplay.empty()) { ListIt replay_it = instsToReplay.begin(); @@ -552,7 +552,7 @@ template <class MemDepPred, class Impl> void MemDepUnit<MemDepPred, Impl>::dumpLists() { - for (unsigned tid=0; tid < Impl::MaxThreads; tid++) { + for (ThreadID tid = 0; tid < Impl::MaxThreads; tid++) { cprintf("Instruction list %i size: %i\n", tid, instList[tid].size()); diff --git a/src/cpu/o3/regfile.hh b/src/cpu/o3/regfile.hh index 53ac2d683..07f8d487b 100644 --- a/src/cpu/o3/regfile.hh +++ b/src/cpu/o3/regfile.hh @@ -230,27 +230,28 @@ class PhysRegFile floatRegFile[reg_idx].q = val; } - MiscReg readMiscRegNoEffect(int misc_reg, unsigned thread_id) + MiscReg + readMiscRegNoEffect(int misc_reg, ThreadID tid) { - return miscRegs[thread_id].readRegNoEffect(misc_reg); + return miscRegs[tid].readRegNoEffect(misc_reg); } - MiscReg readMiscReg(int misc_reg, unsigned thread_id) + MiscReg + readMiscReg(int misc_reg, ThreadID tid) { - return miscRegs[thread_id].readReg(misc_reg, cpu->tcBase(thread_id)); + return miscRegs[tid].readReg(misc_reg, cpu->tcBase(tid)); } - void setMiscRegNoEffect(int misc_reg, - const MiscReg &val, unsigned thread_id) + void + setMiscRegNoEffect(int misc_reg, const MiscReg &val, ThreadID tid) { - miscRegs[thread_id].setRegNoEffect(misc_reg, val); + miscRegs[tid].setRegNoEffect(misc_reg, val); } - void setMiscReg(int misc_reg, const MiscReg &val, - unsigned thread_id) + void + setMiscReg(int misc_reg, const MiscReg &val, ThreadID tid) { - miscRegs[thread_id].setReg(misc_reg, val, - cpu->tcBase(thread_id)); + miscRegs[tid].setReg(misc_reg, val, cpu->tcBase(tid)); } public: diff --git a/src/cpu/o3/rename.hh b/src/cpu/o3/rename.hh index 0fdf28b19..734b63105 100644 --- a/src/cpu/o3/rename.hh +++ b/src/cpu/o3/rename.hh @@ -145,7 +145,7 @@ class DefaultRename void initStage(); /** Sets pointer to list of active threads. */ - void setActiveThreads(std::list<unsigned> *at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Sets pointer to rename maps (per-thread structures). */ void setRenameMap(RenameMap rm_ptr[Impl::MaxThreads]); @@ -169,7 +169,7 @@ class DefaultRename void takeOverFrom(); /** Squashes all instructions in a thread. */ - void squash(const InstSeqNum &squash_seq_num, unsigned tid); + void squash(const InstSeqNum &squash_seq_num, ThreadID tid); /** Ticks rename, which processes all input signals and attempts to rename * as many instructions as possible. @@ -185,17 +185,17 @@ class DefaultRename * change (ie switching from blocking to unblocking). * @param tid Thread id to rename instructions from. */ - void rename(bool &status_change, unsigned tid); + void rename(bool &status_change, ThreadID tid); /** Renames instructions for the given thread. Also handles serializing * instructions. */ - void renameInsts(unsigned tid); + void renameInsts(ThreadID tid); /** Inserts unused instructions from a given thread into the skid buffer, * to be renamed once rename unblocks. */ - void skidInsert(unsigned tid); + void skidInsert(ThreadID tid); /** Separates instructions from decode into individual lists of instructions * sorted by thread. @@ -212,49 +212,49 @@ class DefaultRename * blocked. * @return Returns true if there is a status change. */ - bool block(unsigned tid); + bool block(ThreadID tid); /** Switches rename to unblocking if the skid buffer is empty, and signals * back that rename has unblocked. * @return Returns true if there is a status change. */ - bool unblock(unsigned tid); + bool unblock(ThreadID tid); /** Executes actual squash, removing squashed instructions. */ - void doSquash(const InstSeqNum &squash_seq_num, unsigned tid); + void doSquash(const InstSeqNum &squash_seq_num, ThreadID tid); /** Removes a committed instruction's rename history. */ - void removeFromHistory(InstSeqNum inst_seq_num, unsigned tid); + void removeFromHistory(InstSeqNum inst_seq_num, ThreadID tid); /** Renames the source registers of an instruction. */ - inline void renameSrcRegs(DynInstPtr &inst, unsigned tid); + inline void renameSrcRegs(DynInstPtr &inst, ThreadID tid); /** Renames the destination registers of an instruction. */ - inline void renameDestRegs(DynInstPtr &inst, unsigned tid); + inline void renameDestRegs(DynInstPtr &inst, ThreadID tid); /** Calculates the number of free ROB entries for a specific thread. */ - inline int calcFreeROBEntries(unsigned tid); + inline int calcFreeROBEntries(ThreadID tid); /** Calculates the number of free IQ entries for a specific thread. */ - inline int calcFreeIQEntries(unsigned tid); + inline int calcFreeIQEntries(ThreadID tid); /** Calculates the number of free LSQ entries for a specific thread. */ - inline int calcFreeLSQEntries(unsigned tid); + inline int calcFreeLSQEntries(ThreadID tid); /** Returns the number of valid instructions coming from decode. */ unsigned validInsts(); /** Reads signals telling rename to block/unblock. */ - void readStallSignals(unsigned tid); + void readStallSignals(ThreadID tid); /** Checks if any stages are telling rename to block. */ - bool checkStall(unsigned tid); + bool checkStall(ThreadID tid); /** Gets the number of free entries for a specific thread. */ - void readFreeEntries(unsigned tid); + void readFreeEntries(ThreadID tid); /** Checks the signals and updates the status. */ - bool checkSignalsAndUpdate(unsigned tid); + bool checkSignalsAndUpdate(ThreadID tid); /** Either serializes on the next instruction available in the InstQueue, * or records that it must serialize on the next instruction to enter @@ -263,7 +263,7 @@ class DefaultRename * thread that has the serializeAfter instruction. * @param tid The thread id. */ - void serializeAfter(InstQueue &inst_list, unsigned tid); + void serializeAfter(InstQueue &inst_list, ThreadID tid); /** Holds the information for each destination register rename. It holds * the instruction's sequence number, the arch register, the old physical @@ -332,7 +332,7 @@ class DefaultRename FreeList *freeList; /** Pointer to the list of active threads. */ - std::list<unsigned> *activeThreads; + std::list<ThreadID> *activeThreads; /** Pointer to the scoreboard. */ Scoreboard *scoreboard; @@ -418,7 +418,7 @@ class DefaultRename bool resumeUnblocking; /** The number of threads active in rename. */ - unsigned numThreads; + ThreadID numThreads; /** The maximum skid buffer size. */ unsigned skidBufferMax; diff --git a/src/cpu/o3/rename_impl.hh b/src/cpu/o3/rename_impl.hh index 81647b133..2bca6f81c 100644 --- a/src/cpu/o3/rename_impl.hh +++ b/src/cpu/o3/rename_impl.hh @@ -37,6 +37,8 @@ #include "cpu/o3/rename.hh" #include "params/DerivO3CPU.hh" +using namespace std; + template <class Impl> DefaultRename<Impl>::DefaultRename(O3CPU *_cpu, DerivO3CPUParams *params) : cpu(_cpu), @@ -52,22 +54,22 @@ DefaultRename<Impl>::DefaultRename(O3CPU *_cpu, DerivO3CPUParams *params) { _status = Inactive; - for (int i=0; i< numThreads; i++) { - renameStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; tid++) { + renameStatus[tid] = Idle; - freeEntries[i].iqEntries = 0; - freeEntries[i].lsqEntries = 0; - freeEntries[i].robEntries = 0; + freeEntries[tid].iqEntries = 0; + freeEntries[tid].lsqEntries = 0; + freeEntries[tid].robEntries = 0; - stalls[i].iew = false; - stalls[i].commit = false; - serializeInst[i] = NULL; + stalls[tid].iew = false; + stalls[tid].commit = false; + serializeInst[tid] = NULL; - instsInProgress[i] = 0; + instsInProgress[tid] = 0; - emptyROB[i] = true; + emptyROB[tid] = true; - serializeOnNextInst[i] = false; + serializeOnNextInst[tid] = false; } // @todo: Make into a parameter. @@ -207,7 +209,7 @@ void DefaultRename<Impl>::initStage() { // Grab the number of free entries directly from the stages. - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { freeEntries[tid].iqEntries = iew_ptr->instQueue.numFreeEntries(tid); freeEntries[tid].lsqEntries = iew_ptr->ldstQueue.numFreeEntries(tid); freeEntries[tid].robEntries = commit_ptr->numROBFreeEntries(tid); @@ -217,7 +219,7 @@ DefaultRename<Impl>::initStage() template<class Impl> void -DefaultRename<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +DefaultRename<Impl>::setActiveThreads(list<ThreadID> *at_ptr) { activeThreads = at_ptr; } @@ -227,9 +229,8 @@ template <class Impl> void DefaultRename<Impl>::setRenameMap(RenameMap rm_ptr[]) { - for (int i=0; i<numThreads; i++) { - renameMap[i] = &rm_ptr[i]; - } + for (ThreadID tid = 0; tid < numThreads; tid++) + renameMap[tid] = &rm_ptr[tid]; } template <class Impl> @@ -260,19 +261,19 @@ void DefaultRename<Impl>::switchOut() { // Clear any state, fix up the rename map. - for (int i = 0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { typename std::list<RenameHistory>::iterator hb_it = - historyBuffer[i].begin(); + historyBuffer[tid].begin(); - while (!historyBuffer[i].empty()) { - assert(hb_it != historyBuffer[i].end()); + while (!historyBuffer[tid].empty()) { + assert(hb_it != historyBuffer[tid].end()); DPRINTF(Rename, "[tid:%u]: Removing history entry with sequence " - "number %i.\n", i, (*hb_it).instSeqNum); + "number %i.\n", tid, (*hb_it).instSeqNum); // Tell the rename map to set the architected register to the // previous physical register that it was renamed to. - renameMap[i]->setEntry(hb_it->archReg, hb_it->prevPhysReg); + renameMap[tid]->setEntry(hb_it->archReg, hb_it->prevPhysReg); // Put the renamed physical register back on the free list. freeList->addReg(hb_it->newPhysReg); @@ -282,10 +283,10 @@ DefaultRename<Impl>::switchOut() scoreboard->setReg(hb_it->newPhysReg); } - historyBuffer[i].erase(hb_it++); + historyBuffer[tid].erase(hb_it++); } - insts[i].clear(); - skidBuffer[i].clear(); + insts[tid].clear(); + skidBuffer[tid].clear(); } } @@ -297,24 +298,24 @@ DefaultRename<Impl>::takeOverFrom() initStage(); // Reset all state prior to taking over from the other CPU. - for (int i=0; i< numThreads; i++) { - renameStatus[i] = Idle; + for (ThreadID tid = 0; tid < numThreads; tid++) { + renameStatus[tid] = Idle; - stalls[i].iew = false; - stalls[i].commit = false; - serializeInst[i] = NULL; + stalls[tid].iew = false; + stalls[tid].commit = false; + serializeInst[tid] = NULL; - instsInProgress[i] = 0; + instsInProgress[tid] = 0; - emptyROB[i] = true; + emptyROB[tid] = true; - serializeOnNextInst[i] = false; + serializeOnNextInst[tid] = false; } } template <class Impl> void -DefaultRename<Impl>::squash(const InstSeqNum &squash_seq_num, unsigned tid) +DefaultRename<Impl>::squash(const InstSeqNum &squash_seq_num, ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Squashing instructions.\n",tid); @@ -380,12 +381,12 @@ DefaultRename<Impl>::tick() sortInsts(); - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); // Check stall and squash signals. while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; DPRINTF(Rename, "Processing [tid:%i]\n", tid); @@ -406,7 +407,7 @@ DefaultRename<Impl>::tick() threads = activeThreads->begin(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; // If we committed this cycle then doneSeqNum will be > 0 if (fromCommit->commitInfo[tid].doneSeqNum != 0 && @@ -419,7 +420,7 @@ DefaultRename<Impl>::tick() } // @todo: make into updateProgress function - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { instsInProgress[tid] -= fromIEW->iewInfo[tid].dispatched; assert(instsInProgress[tid] >=0); @@ -429,7 +430,7 @@ DefaultRename<Impl>::tick() template<class Impl> void -DefaultRename<Impl>::rename(bool &status_change, unsigned tid) +DefaultRename<Impl>::rename(bool &status_change, ThreadID tid) { // If status is Running or idle, // call renameInsts() @@ -483,7 +484,7 @@ DefaultRename<Impl>::rename(bool &status_change, unsigned tid) template <class Impl> void -DefaultRename<Impl>::renameInsts(unsigned tid) +DefaultRename<Impl>::renameInsts(ThreadID tid) { // Instructions can be either in the skid buffer or the queue of // instructions coming from decode, depending on the status. @@ -703,7 +704,7 @@ DefaultRename<Impl>::renameInsts(unsigned tid) template<class Impl> void -DefaultRename<Impl>::skidInsert(unsigned tid) +DefaultRename<Impl>::skidInsert(ThreadID tid) { DynInstPtr inst = NULL; @@ -742,8 +743,8 @@ DefaultRename<Impl>::sortInsts() { int insts_from_decode = fromDecode->size; #ifdef DEBUG - for (int i=0; i < numThreads; i++) - assert(insts[i].empty()); + for (ThreadID tid = 0; tid < numThreads; tid++) + assert(insts[tid].empty()); #endif for (int i = 0; i < insts_from_decode; ++i) { DynInstPtr inst = fromDecode->insts[i]; @@ -755,11 +756,11 @@ template<class Impl> bool DefaultRename<Impl>::skidsEmpty() { - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (!skidBuffer[tid].empty()) return false; @@ -774,11 +775,11 @@ DefaultRename<Impl>::updateStatus() { bool any_unblocking = false; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (renameStatus[tid] == Unblocking) { any_unblocking = true; @@ -809,7 +810,7 @@ DefaultRename<Impl>::updateStatus() template <class Impl> bool -DefaultRename<Impl>::block(unsigned tid) +DefaultRename<Impl>::block(ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Blocking.\n", tid); @@ -843,7 +844,7 @@ DefaultRename<Impl>::block(unsigned tid) template <class Impl> bool -DefaultRename<Impl>::unblock(unsigned tid) +DefaultRename<Impl>::unblock(ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Trying to unblock.\n", tid); @@ -864,7 +865,7 @@ DefaultRename<Impl>::unblock(unsigned tid) template <class Impl> void -DefaultRename<Impl>::doSquash(const InstSeqNum &squashed_seq_num, unsigned tid) +DefaultRename<Impl>::doSquash(const InstSeqNum &squashed_seq_num, ThreadID tid) { typename std::list<RenameHistory>::iterator hb_it = historyBuffer[tid].begin(); @@ -904,7 +905,7 @@ DefaultRename<Impl>::doSquash(const InstSeqNum &squashed_seq_num, unsigned tid) template<class Impl> void -DefaultRename<Impl>::removeFromHistory(InstSeqNum inst_seq_num, unsigned tid) +DefaultRename<Impl>::removeFromHistory(InstSeqNum inst_seq_num, ThreadID tid) { DPRINTF(Rename, "[tid:%u]: Removing a committed instruction from the " "history buffer %u (size=%i), until [sn:%lli].\n", @@ -945,7 +946,7 @@ DefaultRename<Impl>::removeFromHistory(InstSeqNum inst_seq_num, unsigned tid) template <class Impl> inline void -DefaultRename<Impl>::renameSrcRegs(DynInstPtr &inst,unsigned tid) +DefaultRename<Impl>::renameSrcRegs(DynInstPtr &inst, ThreadID tid) { assert(renameMap[tid] != 0); @@ -996,7 +997,7 @@ DefaultRename<Impl>::renameSrcRegs(DynInstPtr &inst,unsigned tid) template <class Impl> inline void -DefaultRename<Impl>::renameDestRegs(DynInstPtr &inst,unsigned tid) +DefaultRename<Impl>::renameDestRegs(DynInstPtr &inst, ThreadID tid) { typename RenameMap::RenameInfo rename_result; @@ -1057,7 +1058,7 @@ DefaultRename<Impl>::renameDestRegs(DynInstPtr &inst,unsigned tid) template <class Impl> inline int -DefaultRename<Impl>::calcFreeROBEntries(unsigned tid) +DefaultRename<Impl>::calcFreeROBEntries(ThreadID tid) { int num_free = freeEntries[tid].robEntries - (instsInProgress[tid] - fromIEW->iewInfo[tid].dispatched); @@ -1069,7 +1070,7 @@ DefaultRename<Impl>::calcFreeROBEntries(unsigned tid) template <class Impl> inline int -DefaultRename<Impl>::calcFreeIQEntries(unsigned tid) +DefaultRename<Impl>::calcFreeIQEntries(ThreadID tid) { int num_free = freeEntries[tid].iqEntries - (instsInProgress[tid] - fromIEW->iewInfo[tid].dispatched); @@ -1081,7 +1082,7 @@ DefaultRename<Impl>::calcFreeIQEntries(unsigned tid) template <class Impl> inline int -DefaultRename<Impl>::calcFreeLSQEntries(unsigned tid) +DefaultRename<Impl>::calcFreeLSQEntries(ThreadID tid) { int num_free = freeEntries[tid].lsqEntries - (instsInProgress[tid] - fromIEW->iewInfo[tid].dispatchedToLSQ); @@ -1107,7 +1108,7 @@ DefaultRename<Impl>::validInsts() template <class Impl> void -DefaultRename<Impl>::readStallSignals(unsigned tid) +DefaultRename<Impl>::readStallSignals(ThreadID tid) { if (fromIEW->iewBlock[tid]) { stalls[tid].iew = true; @@ -1130,7 +1131,7 @@ DefaultRename<Impl>::readStallSignals(unsigned tid) template <class Impl> bool -DefaultRename<Impl>::checkStall(unsigned tid) +DefaultRename<Impl>::checkStall(ThreadID tid) { bool ret_val = false; @@ -1165,7 +1166,7 @@ DefaultRename<Impl>::checkStall(unsigned tid) template <class Impl> void -DefaultRename<Impl>::readFreeEntries(unsigned tid) +DefaultRename<Impl>::readFreeEntries(ThreadID tid) { bool updated = false; if (fromIEW->iewInfo[tid].usedIQ) { @@ -1199,7 +1200,7 @@ DefaultRename<Impl>::readFreeEntries(unsigned tid) template <class Impl> bool -DefaultRename<Impl>::checkSignalsAndUpdate(unsigned tid) +DefaultRename<Impl>::checkSignalsAndUpdate(ThreadID tid) { // Check if there's a squash signal, squash if there is // Check stall signals, block if necessary. @@ -1308,8 +1309,7 @@ DefaultRename<Impl>::checkSignalsAndUpdate(unsigned tid) template<class Impl> void -DefaultRename<Impl>::serializeAfter(InstQueue &inst_list, - unsigned tid) +DefaultRename<Impl>::serializeAfter(InstQueue &inst_list, ThreadID tid) { if (inst_list.empty()) { // Mark a bit to say that I must serialize on the next instruction. @@ -1347,11 +1347,11 @@ DefaultRename<Impl>::dumpHistory() { typename std::list<RenameHistory>::iterator buf_it; - for (int i = 0; i < numThreads; i++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { - buf_it = historyBuffer[i].begin(); + buf_it = historyBuffer[tid].begin(); - while (buf_it != historyBuffer[i].end()) { + while (buf_it != historyBuffer[tid].end()) { cprintf("Seq num: %i\nArch reg: %i New phys reg: %i Old phys " "reg: %i\n", (*buf_it).instSeqNum, (int)(*buf_it).archReg, (int)(*buf_it).newPhysReg, (int)(*buf_it).prevPhysReg); diff --git a/src/cpu/o3/rob.hh b/src/cpu/o3/rob.hh index 00329abb0..657bc8d06 100644 --- a/src/cpu/o3/rob.hh +++ b/src/cpu/o3/rob.hh @@ -84,14 +84,14 @@ class ROB */ ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, std::string smtROBPolicy, unsigned _smtROBThreshold, - unsigned _numThreads); + ThreadID _numThreads); std::string name() const; /** Sets pointer to the list of active threads. * @param at_ptr Pointer to the list of active threads. */ - void setActiveThreads(std::list<unsigned>* at_ptr); + void setActiveThreads(std::list<ThreadID> *at_ptr); /** Switches out the ROB. */ void switchOut(); @@ -116,7 +116,7 @@ class ROB * the ROB. * @return Pointer to the DynInst that is at the head of the ROB. */ - DynInstPtr readHeadInst(unsigned tid); + DynInstPtr readHeadInst(ThreadID tid); /** Returns pointer to the tail instruction within the ROB. There is * no guarantee as to the return value if the ROB is empty. @@ -128,7 +128,7 @@ class ROB * the ROB. * @return Pointer to the DynInst that is at the tail of the ROB. */ - DynInstPtr readTailInst(unsigned tid); + DynInstPtr readTailInst(ThreadID tid); /** Retires the head instruction, removing it from the ROB. */ // void retireHead(); @@ -136,13 +136,13 @@ class ROB /** Retires the head instruction of a specific thread, removing it from the * ROB. */ - void retireHead(unsigned tid); + void retireHead(ThreadID tid); /** Is the oldest instruction across all threads ready. */ // bool isHeadReady(); /** Is the oldest instruction across a particular thread ready. */ - bool isHeadReady(unsigned tid); + bool isHeadReady(ThreadID tid); /** Is there any commitable head instruction across all threads ready. */ bool canCommit(); @@ -151,20 +151,20 @@ class ROB void resetEntries(); /** Number of entries needed For 'num_threads' amount of threads. */ - int entryAmount(int num_threads); + int entryAmount(ThreadID num_threads); /** Returns the number of total free entries in the ROB. */ unsigned numFreeEntries(); /** Returns the number of free entries in a specific ROB paritition. */ - unsigned numFreeEntries(unsigned tid); + unsigned numFreeEntries(ThreadID tid); /** Returns the maximum number of entries for a specific thread. */ - unsigned getMaxEntries(unsigned tid) + unsigned getMaxEntries(ThreadID tid) { return maxEntries[tid]; } /** Returns the number of entries being used by a specific thread. */ - unsigned getThreadEntries(unsigned tid) + unsigned getThreadEntries(ThreadID tid) { return threadEntries[tid]; } /** Returns if the ROB is full. */ @@ -172,7 +172,7 @@ class ROB { return numInstsInROB == numEntries; } /** Returns if a specific thread's partition is full. */ - bool isFull(unsigned tid) + bool isFull(ThreadID tid) { return threadEntries[tid] == numEntries; } /** Returns if the ROB is empty. */ @@ -180,16 +180,16 @@ class ROB { return numInstsInROB == 0; } /** Returns if a specific thread's partition is empty. */ - bool isEmpty(unsigned tid) + bool isEmpty(ThreadID tid) { return threadEntries[tid] == 0; } /** Executes the squash, marking squashed instructions. */ - void doSquash(unsigned tid); + void doSquash(ThreadID tid); /** Squashes all instructions younger than the given sequence number for * the specific thread. */ - void squash(InstSeqNum squash_num, unsigned tid); + void squash(InstSeqNum squash_num, ThreadID tid); /** Updates the head instruction with the new oldest instruction. */ void updateHead(); @@ -201,37 +201,37 @@ class ROB // uint64_t readHeadPC(); /** Reads the PC of the head instruction of a specific thread. */ -// uint64_t readHeadPC(unsigned tid); +// uint64_t readHeadPC(ThreadID tid); /** Reads the next PC of the oldest head instruction. */ // uint64_t readHeadNextPC(); /** Reads the next PC of the head instruction of a specific thread. */ -// uint64_t readHeadNextPC(unsigned tid); +// uint64_t readHeadNextPC(ThreadID tid); /** Reads the sequence number of the oldest head instruction. */ // InstSeqNum readHeadSeqNum(); /** Reads the sequence number of the head instruction of a specific thread. */ -// InstSeqNum readHeadSeqNum(unsigned tid); +// InstSeqNum readHeadSeqNum(ThreadID tid); /** Reads the PC of the youngest tail instruction. */ // uint64_t readTailPC(); /** Reads the PC of the tail instruction of a specific thread. */ -// uint64_t readTailPC(unsigned tid); +// uint64_t readTailPC(ThreadID tid); /** Reads the sequence number of the youngest tail instruction. */ // InstSeqNum readTailSeqNum(); /** Reads the sequence number of tail instruction of a specific thread. */ -// InstSeqNum readTailSeqNum(unsigned tid); +// InstSeqNum readTailSeqNum(ThreadID tid); /** Checks if the ROB is still in the process of squashing instructions. * @retval Whether or not the ROB is done squashing. */ - bool isDoneSquashing(unsigned tid) const + bool isDoneSquashing(ThreadID tid) const { return doneSquashing[tid]; } /** Checks if the ROB is still in the process of squashing instructions for @@ -249,14 +249,14 @@ class ROB * threadEntries to get the instructions in the ROB unless you are * double checking that variable. */ - int countInsts(unsigned tid); + int countInsts(ThreadID tid); private: /** Pointer to the CPU. */ O3CPU *cpu; /** Active Threads in CPU */ - std::list<unsigned>* activeThreads; + std::list<ThreadID> *activeThreads; /** Number of instructions in the ROB. */ unsigned numEntries; @@ -309,7 +309,7 @@ class ROB bool doneSquashing[Impl::MaxThreads]; /** Number of active threads. */ - unsigned numThreads; + ThreadID numThreads; }; #endif //__CPU_O3_ROB_HH__ diff --git a/src/cpu/o3/rob_impl.hh b/src/cpu/o3/rob_impl.hh index 7ff3aa274..8acaa8ecb 100644 --- a/src/cpu/o3/rob_impl.hh +++ b/src/cpu/o3/rob_impl.hh @@ -29,22 +29,24 @@ * Korey Sewell */ +#include <list> + #include "config/full_system.hh" #include "cpu/o3/rob.hh" -#include <list> +using namespace std; template <class Impl> ROB<Impl>::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, std::string _smtROBPolicy, unsigned _smtROBThreshold, - unsigned _numThreads) + ThreadID _numThreads) : cpu(_cpu), numEntries(_numEntries), squashWidth(_squashWidth), numInstsInROB(0), numThreads(_numThreads) { - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { squashedSeqNum[tid] = 0; doneSquashing[tid] = true; threadEntries[tid] = 0; @@ -61,8 +63,8 @@ ROB<Impl>::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, robPolicy = Dynamic; //Set Max Entries to Total ROB Capacity - for (int i = 0; i < numThreads; i++) { - maxEntries[i]=numEntries; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = numEntries; } } else if (policy == "partitioned") { @@ -73,8 +75,8 @@ ROB<Impl>::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, int part_amt = numEntries / numThreads; //Divide ROB up evenly - for (int i = 0; i < numThreads; i++) { - maxEntries[i]=part_amt; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = part_amt; } } else if (policy == "threshold") { @@ -84,8 +86,8 @@ ROB<Impl>::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, int threshold = _smtROBThreshold;; //Divide up by threshold amount - for (int i = 0; i < numThreads; i++) { - maxEntries[i]=threshold; + for (ThreadID tid = 0; tid < numThreads; tid++) { + maxEntries[tid] = threshold; } } else { assert(0 && "Invalid ROB Sharing Policy.Options Are:{Dynamic," @@ -93,8 +95,8 @@ ROB<Impl>::ROB(O3CPU *_cpu, unsigned _numEntries, unsigned _squashWidth, } // Set the per-thread iterators to the end of the instruction list. - for (int i=0; i < numThreads;i++) { - squashIt[i] = instList[i].end(); + for (ThreadID tid = 0; tid < numThreads; tid++) { + squashIt[tid] = instList[tid].end(); } // Initialize the "universal" ROB head & tail point to invalid @@ -112,7 +114,7 @@ ROB<Impl>::name() const template <class Impl> void -ROB<Impl>::setActiveThreads(std::list<unsigned> *at_ptr) +ROB<Impl>::setActiveThreads(list<ThreadID> *at_ptr) { DPRINTF(ROB, "Setting active threads list pointer.\n"); activeThreads = at_ptr; @@ -122,7 +124,7 @@ template <class Impl> void ROB<Impl>::switchOut() { - for (int tid = 0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { instList[tid].clear(); } } @@ -131,7 +133,7 @@ template <class Impl> void ROB<Impl>::takeOverFrom() { - for (int tid=0; tid < numThreads; tid++) { + for (ThreadID tid = 0; tid < numThreads; tid++) { doneSquashing[tid] = true; threadEntries[tid] = 0; squashIt[tid] = instList[tid].end(); @@ -151,11 +153,11 @@ ROB<Impl>::resetEntries() if (robPolicy != Dynamic || numThreads > 1) { int active_threads = activeThreads->size(); - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (robPolicy == Partitioned) { maxEntries[tid] = numEntries / active_threads; @@ -168,7 +170,7 @@ ROB<Impl>::resetEntries() template <class Impl> int -ROB<Impl>::entryAmount(int num_threads) +ROB<Impl>::entryAmount(ThreadID num_threads) { if (robPolicy == Partitioned) { return numEntries / num_threads; @@ -181,17 +183,17 @@ template <class Impl> int ROB<Impl>::countInsts() { - int total=0; + int total = 0; - for (int i=0;i < numThreads;i++) - total += countInsts(i); + for (ThreadID tid = 0; tid < numThreads; tid++) + total += countInsts(tid); return total; } template <class Impl> int -ROB<Impl>::countInsts(unsigned tid) +ROB<Impl>::countInsts(ThreadID tid) { return instList[tid].size(); } @@ -207,7 +209,7 @@ ROB<Impl>::insertInst(DynInstPtr &inst) assert(numInstsInROB != numEntries); - int tid = inst->threadNumber; + ThreadID tid = inst->threadNumber; instList[tid].push_back(inst); @@ -242,7 +244,7 @@ ROB<Impl>::retireHead() //assert(numInstsInROB == countInsts()); assert(numInstsInROB > 0); - int tid = (*head)->threadNumber; + ThreadID tid = (*head)->threadNumber; retireHead(tid); @@ -254,7 +256,7 @@ ROB<Impl>::retireHead() template <class Impl> void -ROB<Impl>::retireHead(unsigned tid) +ROB<Impl>::retireHead(ThreadID tid) { //assert(numInstsInROB == countInsts()); assert(numInstsInROB > 0); @@ -300,7 +302,7 @@ ROB<Impl>::isHeadReady() */ template <class Impl> bool -ROB<Impl>::isHeadReady(unsigned tid) +ROB<Impl>::isHeadReady(ThreadID tid) { if (threadEntries[tid] != 0) { return instList[tid].front()->readyToCommit(); @@ -314,11 +316,11 @@ bool ROB<Impl>::canCommit() { //@todo: set ActiveThreads through ROB or CPU - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (isHeadReady(tid)) { return true; @@ -339,14 +341,14 @@ ROB<Impl>::numFreeEntries() template <class Impl> unsigned -ROB<Impl>::numFreeEntries(unsigned tid) +ROB<Impl>::numFreeEntries(ThreadID tid) { return maxEntries[tid] - threadEntries[tid]; } template <class Impl> void -ROB<Impl>::doSquash(unsigned tid) +ROB<Impl>::doSquash(ThreadID tid) { DPRINTF(ROB, "[tid:%u]: Squashing instructions until [sn:%i].\n", tid, squashedSeqNum[tid]); @@ -429,11 +431,11 @@ ROB<Impl>::updateHead() bool first_valid = true; // @todo: set ActiveThreads through ROB or CPU - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (instList[tid].empty()) continue; @@ -470,11 +472,11 @@ ROB<Impl>::updateTail() tail = instList[0].end(); bool first_valid = true; - std::list<unsigned>::iterator threads = activeThreads->begin(); - std::list<unsigned>::iterator end = activeThreads->end(); + list<ThreadID>::iterator threads = activeThreads->begin(); + list<ThreadID>::iterator end = activeThreads->end(); while (threads != end) { - unsigned tid = *threads++; + ThreadID tid = *threads++; if (instList[tid].empty()) { continue; @@ -503,7 +505,7 @@ ROB<Impl>::updateTail() template <class Impl> void -ROB<Impl>::squash(InstSeqNum squash_num,unsigned tid) +ROB<Impl>::squash(InstSeqNum squash_num, ThreadID tid) { if (isEmpty()) { DPRINTF(ROB, "Does not need to squash due to being empty " @@ -546,7 +548,7 @@ ROB<Impl>::readHeadInst() template <class Impl> typename Impl::DynInstPtr -ROB<Impl>::readHeadInst(unsigned tid) +ROB<Impl>::readHeadInst(ThreadID tid) { if (threadEntries[tid] != 0) { InstIt head_thread = instList[tid].begin(); @@ -573,7 +575,7 @@ ROB<Impl>::readHeadPC() template <class Impl> uint64_t -ROB<Impl>::readHeadPC(unsigned tid) +ROB<Impl>::readHeadPC(ThreadID tid) { //assert(numInstsInROB == countInsts()); InstIt head_thread = instList[tid].begin(); @@ -595,7 +597,7 @@ ROB<Impl>::readHeadNextPC() template <class Impl> uint64_t -ROB<Impl>::readHeadNextPC(unsigned tid) +ROB<Impl>::readHeadNextPC(ThreadID tid) { //assert(numInstsInROB == countInsts()); InstIt head_thread = instList[tid].begin(); @@ -615,7 +617,7 @@ ROB<Impl>::readHeadSeqNum() template <class Impl> InstSeqNum -ROB<Impl>::readHeadSeqNum(unsigned tid) +ROB<Impl>::readHeadSeqNum(ThreadID tid) { InstIt head_thread = instList[tid].begin(); @@ -634,7 +636,7 @@ ROB<Impl>::readTailInst() */ template <class Impl> typename Impl::DynInstPtr -ROB<Impl>::readTailInst(unsigned tid) +ROB<Impl>::readTailInst(ThreadID tid) { //assert(tail_thread[tid] != instList[tid].end()); @@ -658,7 +660,7 @@ ROB<Impl>::readTailPC() template <class Impl> uint64_t -ROB<Impl>::readTailPC(unsigned tid) +ROB<Impl>::readTailPC(ThreadID tid) { //assert(tail_thread[tid] != instList[tid].end()); @@ -680,7 +682,7 @@ ROB<Impl>::readTailSeqNum() template <class Impl> InstSeqNum -ROB<Impl>::readTailSeqNum(unsigned tid) +ROB<Impl>::readTailSeqNum(ThreadID tid) { // Return the last sequence number that has not been squashed. Other // stages can use it to squash any instructions younger than the current diff --git a/src/cpu/o3/store_set.cc b/src/cpu/o3/store_set.cc index 2d28b617f..df4ee00ad 100644 --- a/src/cpu/o3/store_set.cc +++ b/src/cpu/o3/store_set.cc @@ -186,8 +186,7 @@ StoreSet::insertLoad(Addr load_PC, InstSeqNum load_seq_num) } void -StoreSet::insertStore(Addr store_PC, InstSeqNum store_seq_num, - unsigned tid) +StoreSet::insertStore(Addr store_PC, InstSeqNum store_seq_num, ThreadID tid) { int index = calcIndex(store_PC); @@ -288,7 +287,7 @@ StoreSet::issued(Addr issued_PC, InstSeqNum issued_seq_num, bool is_store) } void -StoreSet::squash(InstSeqNum squashed_num, unsigned tid) +StoreSet::squash(InstSeqNum squashed_num, ThreadID tid) { DPRINTF(StoreSet, "StoreSet: Squashing until inum %i\n", squashed_num); diff --git a/src/cpu/o3/store_set.hh b/src/cpu/o3/store_set.hh index 57cd2a197..ce4591f68 100644 --- a/src/cpu/o3/store_set.hh +++ b/src/cpu/o3/store_set.hh @@ -82,8 +82,7 @@ class StoreSet /** Inserts a store into the store set predictor. Updates the * LFST if the store has a valid SSID. */ - void insertStore(Addr store_PC, InstSeqNum store_seq_num, - unsigned tid); + void insertStore(Addr store_PC, InstSeqNum store_seq_num, ThreadID tid); /** Checks if the instruction with the given PC is dependent upon * any store. @return Returns the sequence number of the store @@ -95,7 +94,7 @@ class StoreSet void issued(Addr issued_PC, InstSeqNum issued_seq_num, bool is_store); /** Squashes for a specific thread until the given sequence number. */ - void squash(InstSeqNum squashed_num, unsigned tid); + void squash(InstSeqNum squashed_num, ThreadID tid); /** Resets all tables. */ void clear(); diff --git a/src/cpu/o3/thread_context_impl.hh b/src/cpu/o3/thread_context_impl.hh index 07140a19f..bce334dc4 100755 --- a/src/cpu/o3/thread_context_impl.hh +++ b/src/cpu/o3/thread_context_impl.hh @@ -226,7 +226,7 @@ O3ThreadContext<Impl>::copyArchRegs(ThreadContext *tc) { // This function will mess things up unless the ROB is empty and // there are no instructions in the pipeline. - unsigned tid = thread->threadId(); + ThreadID tid = thread->threadId(); PhysRegIndex renamed_reg; // First loop through the integer registers. |