Clock: Rework clocks to avoid tick-to-cycle transformations

This patch introduces the notion of a clock update function that aims
to avoid costly divisions when turning the current tick into a
cycle. Each clocked object advances a private (hidden) cycle member
and a tick member and uses these to implement functions for getting
the tick of the next cycle, or the tick of a cycle some time in the
future.

In the different modules using the clocks, changes are made to avoid
counting in ticks only to later translate to cycles. There are a few
oddities in how the O3 and inorder CPU count idle cycles, as seen by a
few locations where a cycle is subtracted in the calculation. This is
done such that the regression does not change any stats, but should be
revisited in a future patch.

Another, much needed, change that is not done as part of this patch is
to introduce a new typedef uint64_t Cycle to be able to at least hint
at the unit of the variables counting Ticks vs Cycles. This will be
done as a follow-up patch.

As an additional follow up, the thread context still uses ticks for
the book keeping of last activate and last suspend and this should
probably also be changed into cycles as well.

24 files changed, 165 insertions, 110 deletions

diff --git a/src/arch/arm/table_walker.cc b/src/arch/arm/table_walker.cc
index ea71e6f1c..ffa193fbe 100644
--- a/src/arch/arm/table_walker.cc
+++ b/src/arch/arm/table_walker.cc
@@ -758,7 +758,7 @@ void
 TableWalker::nextWalk(ThreadContext *tc)
 {
     if (pendingQueue.size())
-        schedule(doProcessEvent, tc->getCpuPtr()->nextCycle(curTick()+1));
+        schedule(doProcessEvent, tc->getCpuPtr()->clockEdge(1));
 }
 
 
diff --git a/src/arch/x86/mmapped_ipr.hh b/src/arch/x86/mmapped_ipr.hh
index 4c3292388..f17b64cad 100644
--- a/src/arch/x86/mmapped_ipr.hh
+++ b/src/arch/x86/mmapped_ipr.hh
@@ -62,7 +62,7 @@ namespace X86ISA
         // Make sure we don't trot off the end of data.
         assert(offset + pkt->getSize() <= sizeof(MiscReg));
         pkt->setData(((uint8_t *)&data) + offset);
-        return xc->getCpuPtr()->ticks(1);
+        return 1;
     }
 
     inline Tick
@@ -76,7 +76,7 @@ namespace X86ISA
         assert(offset + pkt->getSize() <= sizeof(MiscReg));
         pkt->writeData(((uint8_t *)&data) + offset);
         xc->setMiscReg(index, gtoh(data));
-        return xc->getCpuPtr()->ticks(1);
+        return 1;
     }
 }
 
diff --git a/src/cpu/base.cc b/src/cpu/base.cc
index ff832f562..98ddde293 100644
--- a/src/cpu/base.cc
+++ b/src/cpu/base.cc
@@ -91,7 +91,7 @@ CPUProgressEvent::process()
 {
     Counter temp = cpu->totalOps();
 #ifndef NDEBUG
-    double ipc = double(temp - lastNumInst) / (_interval / cpu->ticks(1));
+    double ipc = double(temp - lastNumInst) / (_interval / cpu->clockPeriod());
 
     DPRINTFN("%s progress event, total committed:%i, progress insts committed: "
              "%lli, IPC: %0.8d\n", cpu->name(), temp, temp - lastNumInst,
@@ -261,9 +261,7 @@ BaseCPU::startup()
     }
 
     if (params()->progress_interval) {
-        Tick num_ticks = ticks(params()->progress_interval);
-
-        new CPUProgressEvent(this, num_ticks);
+        new CPUProgressEvent(this, params()->progress_interval);
     }
 }
 
diff --git a/src/cpu/inorder/cpu.cc b/src/cpu/inorder/cpu.cc
index 4aff0c579..ec06f19f0 100644
--- a/src/cpu/inorder/cpu.cc
+++ b/src/cpu/inorder/cpu.cc
@@ -212,7 +212,7 @@ void
 InOrderCPU::CPUEvent::scheduleEvent(int delay)
 {
     assert(!scheduled() || squashed());
-    cpu->reschedule(this, cpu->nextCycle(curTick() + cpu->ticks(delay)), true);
+    cpu->reschedule(this, cpu->clockEdge(delay), true);
 }
 
 void
@@ -401,7 +401,7 @@ InOrderCPU::InOrderCPU(Params *params)
     frontEndSked = createFrontEndSked();
     faultSked = createFaultSked();
 
-    lastRunningCycle = curTick();
+    lastRunningCycle = curCycle();
 
     lockAddr = 0;
     lockFlag = false;
@@ -761,17 +761,17 @@ InOrderCPU::tick()
     if (!tickEvent.scheduled()) {
         if (_status == SwitchedOut) {
             // increment stat
-            lastRunningCycle = curTick();
+            lastRunningCycle = curCycle();
         } else if (!activityRec.active()) {
             DPRINTF(InOrderCPU, "sleeping CPU.\n");
-            lastRunningCycle = curTick();
+            lastRunningCycle = curCycle();
             timesIdled++;
         } else {
             //Tick next_tick = curTick() + cycles(1);
             //tickEvent.schedule(next_tick);
-            schedule(&tickEvent, nextCycle(curTick() + 1));
+            schedule(&tickEvent, clockEdge(1));
             DPRINTF(InOrderCPU, "Scheduled CPU for next tick @ %i.\n", 
-                    nextCycle(curTick() + 1));
+                    clockEdge(1));
         }
     }
 
@@ -959,15 +959,10 @@ InOrderCPU::scheduleCpuEvent(CPUEventType c_event, Fault fault,
     CPUEvent *cpu_event = new CPUEvent(this, c_event, fault, tid, inst,
                                        event_pri);
 
-    Tick sked_tick = nextCycle(curTick() + ticks(delay));
-    if (delay >= 0) {
-        DPRINTF(InOrderCPU, "Scheduling CPU Event (%s) for cycle %i, [tid:%i].\n",
-                eventNames[c_event], curTick() + delay, tid);
-        schedule(cpu_event, sked_tick);
-    } else {
-        cpu_event->process();
-        cpuEventRemoveList.push(cpu_event);
-    }
+    Tick sked_tick = clockEdge(delay);
+    DPRINTF(InOrderCPU, "Scheduling CPU Event (%s) for cycle %i, [tid:%i].\n",
+            eventNames[c_event], curTick() + delay, tid);
+    schedule(cpu_event, sked_tick);
 
     // Broadcast event to the Resource Pool
     // Need to reset tid just in case this is a dummy instruction
@@ -1696,7 +1691,9 @@ InOrderCPU::wakeCPU()
 
     DPRINTF(Activity, "Waking up CPU\n");
 
-    Tick extra_cycles = tickToCycles((curTick() - 1) - lastRunningCycle);
+    Tick extra_cycles = curCycle() - lastRunningCycle;
+    if (extra_cycles != 0)
+        --extra_cycles;
 
     idleCycles += extra_cycles;    
     for (int stage_num = 0; stage_num < NumStages; stage_num++) {
@@ -1705,7 +1702,7 @@ InOrderCPU::wakeCPU()
 
     numCycles += extra_cycles;
 
-    schedule(&tickEvent, nextCycle(curTick()));
+    schedule(&tickEvent, nextCycle());
 }
 
 // Lots of copied full system code...place into BaseCPU class?
diff --git a/src/cpu/inorder/cpu.hh b/src/cpu/inorder/cpu.hh
index 3103201dd..9a0a62c87 100644
--- a/src/cpu/inorder/cpu.hh
+++ b/src/cpu/inorder/cpu.hh
@@ -204,7 +204,7 @@ class InOrderCPU : public BaseCPU
     void scheduleTickEvent(int delay)
     {
         assert(!tickEvent.scheduled() || tickEvent.squashed());
-        reschedule(&tickEvent, nextCycle(curTick() + ticks(delay)), true);
+        reschedule(&tickEvent, clockEdge(delay), true);
     }
 
     /** Unschedule tick event, regardless of its current state. */
diff --git a/src/cpu/inorder/resource.cc b/src/cpu/inorder/resource.cc
index 2ddce13c3..098a4e1b4 100644
--- a/src/cpu/inorder/resource.cc
+++ b/src/cpu/inorder/resource.cc
@@ -372,7 +372,7 @@ Resource::ticks(int num_cycles)
 void
 Resource::scheduleExecution(int slot_num)
 {
-    if (latency >= 1) {
+    if (latency > 0) {
         scheduleEvent(slot_num, latency);
     } else {
         execute(slot_num);
diff --git a/src/cpu/inorder/resource_pool.cc b/src/cpu/inorder/resource_pool.cc
index 4c01165b8..31511314e 100644
--- a/src/cpu/inorder/resource_pool.cc
+++ b/src/cpu/inorder/resource_pool.cc
@@ -238,7 +238,7 @@ ResourcePool::scheduleEvent(InOrderCPU::CPUEventType e_type, DynInstPtr inst,
 {
     assert(delay >= 0);
 
-    Tick when = cpu->nextCycle(curTick() + cpu->ticks(delay));
+    Tick when = cpu->clockEdge(delay);
 
     switch ((int)e_type)
     {
@@ -460,7 +460,7 @@ ResourcePool::ResPoolEvent::scheduleEvent(int delay)
 {
     InOrderCPU *cpu = resPool->cpu;
     assert(!scheduled() || squashed());
-    cpu->reschedule(this, cpu->nextCycle(curTick() + cpu->ticks(delay)), true);
+    cpu->reschedule(this, cpu->clockEdge(delay), true);
 }
 
 /** Unschedule resource event, regardless of its current state. */
diff --git a/src/cpu/o3/O3CPU.py b/src/cpu/o3/O3CPU.py
index 3138aebbf..9306cb44e 100644
--- a/src/cpu/o3/O3CPU.py
+++ b/src/cpu/o3/O3CPU.py
@@ -76,8 +76,8 @@ class DerivO3CPU(BaseCPU):
     renameToROBDelay = Param.Unsigned(1, "Rename to reorder buffer delay")
     commitWidth = Param.Unsigned(8, "Commit width")
     squashWidth = Param.Unsigned(8, "Squash width")
-    trapLatency = Param.Tick(13, "Trap latency")
-    fetchTrapLatency = Param.Tick(1, "Fetch trap latency")
+    trapLatency = Param.Unsigned(13, "Trap latency")
+    fetchTrapLatency = Param.Unsigned(1, "Fetch trap latency")
 
     backComSize = Param.Unsigned(5, "Time buffer size for backwards communication")
     forwardComSize = Param.Unsigned(5, "Time buffer size for forward communication")
diff --git a/src/cpu/o3/commit.hh b/src/cpu/o3/commit.hh
index b5539c702..1119e8b50 100644
--- a/src/cpu/o3/commit.hh
+++ b/src/cpu/o3/commit.hh
@@ -409,7 +409,7 @@ class DefaultCommit
     /** The latency to handle a trap.  Used when scheduling trap
      * squash event.
      */
-    Tick trapLatency;
+    uint trapLatency;
 
     /** The interrupt fault. */
     Fault interrupt;
diff --git a/src/cpu/o3/commit_impl.hh b/src/cpu/o3/commit_impl.hh
index 31398c3d9..66474c05f 100644
--- a/src/cpu/o3/commit_impl.hh
+++ b/src/cpu/o3/commit_impl.hh
@@ -374,7 +374,6 @@ DefaultCommit<Impl>::initStage()
     cpu->activateStage(O3CPU::CommitIdx);
 
     cpu->activityThisCycle();
-    trapLatency = cpu->ticks(trapLatency);
 }
 
 template <class Impl>
@@ -509,7 +508,7 @@ DefaultCommit<Impl>::generateTrapEvent(ThreadID tid)
 
     TrapEvent *trap = new TrapEvent(this, tid);
 
-    cpu->schedule(trap, curTick() + trapLatency);
+    cpu->schedule(trap, cpu->clockEdge(trapLatency));
     trapInFlight[tid] = true;
     thread[tid]->trapPending = true;
 }
diff --git a/src/cpu/o3/cpu.cc b/src/cpu/o3/cpu.cc
index 71d04740c..95683a77a 100644
--- a/src/cpu/o3/cpu.cc
+++ b/src/cpu/o3/cpu.cc
@@ -386,7 +386,7 @@ FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams *params)
     // Setup the ROB for whichever stages need it.
     commit.setROB(&rob);
 
-    lastRunningCycle = curTick();
+    lastRunningCycle = curCycle();
 
     lastActivatedCycle = 0;
 #if 0
@@ -623,13 +623,13 @@ FullO3CPU<Impl>::tick()
             getState() == SimObject::Drained) {
             DPRINTF(O3CPU, "Switched out!\n");
             // increment stat
-            lastRunningCycle = curTick();
+            lastRunningCycle = curCycle();
         } else if (!activityRec.active() || _status == Idle) {
             DPRINTF(O3CPU, "Idle!\n");
-            lastRunningCycle = curTick();
+            lastRunningCycle = curCycle();
             timesIdled++;
         } else {
-            schedule(tickEvent, nextCycle(curTick() + ticks(1)));
+            schedule(tickEvent, clockEdge(1));
             DPRINTF(O3CPU, "Scheduling next tick!\n");
         }
     }
@@ -762,7 +762,10 @@ FullO3CPU<Impl>::activateContext(ThreadID tid, int delay)
         activityRec.activity();
         fetch.wakeFromQuiesce();
 
-        quiesceCycles += tickToCycles((curTick() - 1) - lastRunningCycle);
+        Tick cycles = curCycle() - lastRunningCycle;
+        if (cycles != 0)
+            --cycles;
+        quiesceCycles += cycles;
 
         lastActivatedCycle = curTick();
 
@@ -801,7 +804,7 @@ FullO3CPU<Impl>::suspendContext(ThreadID tid)
         unscheduleTickEvent();
 
     DPRINTF(Quiesce, "Suspending Context\n");
-    lastRunningCycle = curTick();
+    lastRunningCycle = curCycle();
     _status = Idle;
 }
 
@@ -1275,7 +1278,7 @@ FullO3CPU<Impl>::takeOverFrom(BaseCPU *oldCPU)
     if (!tickEvent.scheduled())
         schedule(tickEvent, nextCycle());
 
-    lastRunningCycle = curTick();
+    lastRunningCycle = curCycle();
 }
 
 template <class Impl>
@@ -1669,8 +1672,11 @@ FullO3CPU<Impl>::wakeCPU()
 
     DPRINTF(Activity, "Waking up CPU\n");
 
-    idleCycles += tickToCycles((curTick() - 1) - lastRunningCycle);
-    numCycles += tickToCycles((curTick() - 1) - lastRunningCycle);
+    Tick cycles = curCycle() - lastRunningCycle;
+    if (cycles != 0)
+        --cycles;
+    idleCycles += cycles;
+    numCycles += cycles;
 
     schedule(tickEvent, nextCycle());
 }
diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh
index b1fd12a2e..5910f314d 100644
--- a/src/cpu/o3/cpu.hh
+++ b/src/cpu/o3/cpu.hh
@@ -214,9 +214,9 @@ class FullO3CPU : public BaseO3CPU
     void scheduleTickEvent(int delay)
     {
         if (tickEvent.squashed())
-            reschedule(tickEvent, nextCycle(curTick() + ticks(delay)));
+            reschedule(tickEvent, clockEdge(delay));
         else if (!tickEvent.scheduled())
-            schedule(tickEvent, nextCycle(curTick() + ticks(delay)));
+            schedule(tickEvent, clockEdge(delay));
     }
 
     /** Unschedule tick event, regardless of its current state. */
@@ -256,9 +256,9 @@ class FullO3CPU : public BaseO3CPU
         // Schedule thread to activate, regardless of its current state.
         if (activateThreadEvent[tid].squashed())
             reschedule(activateThreadEvent[tid],
-                nextCycle(curTick() + ticks(delay)));
+                       clockEdge(delay));
         else if (!activateThreadEvent[tid].scheduled()) {
-            Tick when = nextCycle(curTick() + ticks(delay));
+            Tick when = clockEdge(delay);
 
             // Check if the deallocateEvent is also scheduled, and make
             // sure they do not happen at same time causing a sleep that
@@ -319,10 +319,10 @@ class FullO3CPU : public BaseO3CPU
         // Schedule thread to activate, regardless of its current state.
         if (deallocateContextEvent[tid].squashed())
             reschedule(deallocateContextEvent[tid],
-                nextCycle(curTick() + ticks(delay)));
+                       clockEdge(delay));
         else if (!deallocateContextEvent[tid].scheduled())
             schedule(deallocateContextEvent[tid],
-                nextCycle(curTick() + ticks(delay)));
+                     clockEdge(delay));
     }
 
     /** Unschedule thread deallocation in CPU */
diff --git a/src/cpu/o3/fetch_impl.hh b/src/cpu/o3/fetch_impl.hh
index caafa3fe3..9caf0c79b 100644
--- a/src/cpu/o3/fetch_impl.hh
+++ b/src/cpu/o3/fetch_impl.hh
@@ -646,7 +646,7 @@ DefaultFetch<Impl>::finishTranslation(Fault fault, RequestPtr mem_req)
             assert(!finishTranslationEvent.scheduled());
             finishTranslationEvent.setFault(fault);
             finishTranslationEvent.setReq(mem_req);
-            cpu->schedule(finishTranslationEvent, cpu->nextCycle(curTick() + cpu->ticks(1)));
+            cpu->schedule(finishTranslationEvent, cpu->clockEdge(1));
             return;
         }
         DPRINTF(Fetch, "[tid:%i] Got back req with addr %#x but expected %#x\n",
diff --git a/src/cpu/o3/inst_queue_impl.hh b/src/cpu/o3/inst_queue_impl.hh
index ae5f93c38..b6c3bd239 100644
--- a/src/cpu/o3/inst_queue_impl.hh
+++ b/src/cpu/o3/inst_queue_impl.hh
@@ -828,7 +828,7 @@ InstructionQueue<Impl>::scheduleReadyInsts()
                 FUCompletion *execution = new FUCompletion(issuing_inst,
                                                            idx, this);
 
-                cpu->schedule(execution, curTick() + cpu->ticks(op_latency - 1));
+                cpu->schedule(execution, cpu->clockEdge(op_latency - 1));
 
                 // @todo: Enforce that issue_latency == 1 or op_latency
                 if (issue_latency > 1) {
diff --git a/src/cpu/o3/lsq_unit.hh b/src/cpu/o3/lsq_unit.hh
index b886a2259..c567341c7 100644
--- a/src/cpu/o3/lsq_unit.hh
+++ b/src/cpu/o3/lsq_unit.hh
@@ -632,7 +632,7 @@ LSQUnit<Impl>::read(Request *req, Request *sreqLow, Request *sreqHigh,
             delete snd_data_pkt;
         }
         WritebackEvent *wb = new WritebackEvent(load_inst, data_pkt, this);
-        cpu->schedule(wb, curTick() + delay);
+        cpu->schedule(wb, cpu->clockEdge(delay));
         return NoFault;
     }
 
diff --git a/src/cpu/simple/atomic.cc b/src/cpu/simple/atomic.cc
index fc6724939..d1b0391fc 100644
--- a/src/cpu/simple/atomic.cc
+++ b/src/cpu/simple/atomic.cc
@@ -208,10 +208,10 @@ AtomicSimpleCPU::activateContext(ThreadID thread_num, int delay)
     assert(!tickEvent.scheduled());
 
     notIdleFraction++;
-    numCycles += tickToCycles(thread->lastActivate - thread->lastSuspend);
+    numCycles += tickToCycle(thread->lastActivate - thread->lastSuspend);
 
     //Make sure ticks are still on multiples of cycles
-    schedule(tickEvent, nextCycle(curTick() + ticks(delay)));
+    schedule(tickEvent, clockEdge(delay));
     _status = Running;
 }
 
@@ -518,7 +518,7 @@ AtomicSimpleCPU::tick()
                 stall_ticks += dcache_latency;
 
             if (stall_ticks) {
-                Tick stall_cycles = stall_ticks / ticks(1);
+                Tick stall_cycles = stall_ticks / clockPeriod();
                 Tick aligned_stall_ticks = ticks(stall_cycles);
 
                 if (aligned_stall_ticks < stall_ticks)
@@ -533,8 +533,8 @@ AtomicSimpleCPU::tick()
     }
 
     // instruction takes at least one cycle
-    if (latency < ticks(1))
-        latency = ticks(1);
+    if (latency < clockPeriod())
+        latency = clockPeriod();
 
     if (_status != Idle)
         schedule(tickEvent, curTick() + latency);
diff --git a/src/cpu/simple/timing.cc b/src/cpu/simple/timing.cc
index 9022845ce..5437e77aa 100644
--- a/src/cpu/simple/timing.cc
+++ b/src/cpu/simple/timing.cc
@@ -87,13 +87,11 @@ TimingSimpleCPU::TimingCPUPort::TickEvent::schedule(PacketPtr _pkt, Tick t)
 
 TimingSimpleCPU::TimingSimpleCPU(TimingSimpleCPUParams *p)
     : BaseSimpleCPU(p), fetchTranslation(this), icachePort(this),
-    dcachePort(this), fetchEvent(this)
+      dcachePort(this), ifetch_pkt(NULL), dcache_pkt(NULL), previousCycle(0),
+      fetchEvent(this)
 {
     _status = Idle;
 
-    ifetch_pkt = dcache_pkt = NULL;
-    drainEvent = NULL;
-    previousTick = 0;
     changeState(SimObject::Running);
     system->totalNumInsts = 0;
 }
@@ -156,7 +154,7 @@ TimingSimpleCPU::switchOut()
 {
     assert(_status == Running || _status == Idle);
     _status = SwitchedOut;
-    numCycles += tickToCycles(curTick() - previousTick);
+    numCycles += curCycle() - previousCycle;
 
     // If we've been scheduled to resume but are then told to switch out,
     // we'll need to cancel it.
@@ -184,7 +182,7 @@ TimingSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
         _status = Idle;
     }
     assert(threadContexts.size() == 1);
-    previousTick = curTick();
+    previousCycle = curCycle();
 }
 
 
@@ -202,7 +200,7 @@ TimingSimpleCPU::activateContext(ThreadID thread_num, int delay)
     _status = Running;
 
     // kick things off by initiating the fetch of the next instruction
-    schedule(fetchEvent, nextCycle(curTick() + ticks(delay)));
+    schedule(fetchEvent, clockEdge(delay));
 }
 
 
@@ -231,9 +229,8 @@ TimingSimpleCPU::handleReadPacket(PacketPtr pkt)
 {
     RequestPtr req = pkt->req;
     if (req->isMmappedIpr()) {
-        Tick delay;
-        delay = TheISA::handleIprRead(thread->getTC(), pkt);
-        new IprEvent(pkt, this, nextCycle(curTick() + delay));
+        Tick delay = TheISA::handleIprRead(thread->getTC(), pkt);
+        new IprEvent(pkt, this, clockEdge(delay));
         _status = DcacheWaitResponse;
         dcache_pkt = NULL;
     } else if (!dcachePort.sendTimingReq(pkt)) {
@@ -322,8 +319,8 @@ TimingSimpleCPU::translationFault(Fault fault)
 {
     // fault may be NoFault in cases where a fault is suppressed,
     // for instance prefetches.
-    numCycles += tickToCycles(curTick() - previousTick);
-    previousTick = curTick();
+    numCycles += curCycle() - previousCycle;
+    previousCycle = curCycle();
 
     if (traceData) {
         // Since there was a fault, we shouldn't trace this instruction.
@@ -446,9 +443,8 @@ TimingSimpleCPU::handleWritePacket()
 {
     RequestPtr req = dcache_pkt->req;
     if (req->isMmappedIpr()) {
-        Tick delay;
-        delay = TheISA::handleIprWrite(thread->getTC(), dcache_pkt);
-        new IprEvent(dcache_pkt, this, nextCycle(curTick() + delay));
+        Tick delay = TheISA::handleIprWrite(thread->getTC(), dcache_pkt);
+        new IprEvent(dcache_pkt, this, clockEdge(delay));
         _status = DcacheWaitResponse;
         dcache_pkt = NULL;
     } else if (!dcachePort.sendTimingReq(dcache_pkt)) {
@@ -567,8 +563,8 @@ TimingSimpleCPU::fetch()
         _status = IcacheWaitResponse;
         completeIfetch(NULL);
 
-        numCycles += tickToCycles(curTick() - previousTick);
-        previousTick = curTick();
+        numCycles += curCycle() - previousCycle;
+        previousCycle = curCycle();
     }
 }
 
@@ -600,8 +596,8 @@ TimingSimpleCPU::sendFetch(Fault fault, RequestPtr req, ThreadContext *tc)
         advanceInst(fault);
     }
 
-    numCycles += tickToCycles(curTick() - previousTick);
-    previousTick = curTick();
+    numCycles += curCycle() - previousCycle;
+    previousCycle = curCycle();
 }
 
 
@@ -647,8 +643,8 @@ TimingSimpleCPU::completeIfetch(PacketPtr pkt)
 
     _status = Running;
 
-    numCycles += tickToCycles(curTick() - previousTick);
-    previousTick = curTick();
+    numCycles += curCycle() - previousCycle;
+    previousCycle = curCycle();
 
     if (getState() == SimObject::Draining) {
         if (pkt) {
@@ -754,8 +750,8 @@ TimingSimpleCPU::completeDataAccess(PacketPtr pkt)
     assert(_status == DcacheWaitResponse || _status == DTBWaitResponse ||
            pkt->req->getFlags().isSet(Request::NO_ACCESS));
 
-    numCycles += tickToCycles(curTick() - previousTick);
-    previousTick = curTick();
+    numCycles += curCycle() - previousCycle;
+    previousCycle = curCycle();
 
     if (pkt->senderState) {
         SplitFragmentSenderState * send_state =
diff --git a/src/cpu/simple/timing.hh b/src/cpu/simple/timing.hh
index b6b78c5db..c4d1573af 100644
--- a/src/cpu/simple/timing.hh
+++ b/src/cpu/simple/timing.hh
@@ -234,7 +234,7 @@ class TimingSimpleCPU : public BaseSimpleCPU
     PacketPtr ifetch_pkt;
     PacketPtr dcache_pkt;
 
-    Tick previousTick;
+    Tick previousCycle;
 
   protected:
 
diff --git a/src/cpu/testers/memtest/memtest.cc b/src/cpu/testers/memtest/memtest.cc
index 642af4677..f8a8b66bd 100644
--- a/src/cpu/testers/memtest/memtest.cc
+++ b/src/cpu/testers/memtest/memtest.cc
@@ -246,7 +246,7 @@ void
 MemTest::tick()
 {
     if (!tickEvent.scheduled())
-        schedule(tickEvent, curTick() + ticks(1));
+        schedule(tickEvent, clockEdge(1));
 
     if (++noResponseCycles >= 500000) {
         if (issueDmas) {
diff --git a/src/cpu/testers/networktest/networktest.cc b/src/cpu/testers/networktest/networktest.cc
index 5d0e8e0c9..05568b3c0 100644
--- a/src/cpu/testers/networktest/networktest.cc
+++ b/src/cpu/testers/networktest/networktest.cc
@@ -165,7 +165,7 @@ NetworkTest::tick()
         exitSimLoop("Network Tester completed simCycles");
     else {
         if (!tickEvent.scheduled())
-            schedule(tickEvent, curTick() + ticks(1));
+            schedule(tickEvent, clockEdge(1));
     }
 }
 
diff --git a/src/dev/arm/pl111.cc b/src/dev/arm/pl111.cc
index d79a1cf39..7990e02ed 100644
--- a/src/dev/arm/pl111.cc
+++ b/src/dev/arm/pl111.cc
@@ -440,7 +440,7 @@ Pl111::readFramebuffer()
         schedule(intEvent, nextCycle());
 
     curAddr = 0;
-    startTime = curTick();
+    startTime = curCycle();
 
     maxAddr = static_cast<Addr>(length * bytesPerPixel);
 
@@ -475,12 +475,12 @@ Pl111::fillFifo()
 void
 Pl111::dmaDone()
 {
-    Tick maxFrameTime = lcdTiming2.cpl * height * clock;
+    Tick maxFrameTime = lcdTiming2.cpl * height;
 
     --dmaPendingNum;
 
     if (maxAddr == curAddr && !dmaPendingNum) {
-        if ((curTick() - startTime) > maxFrameTime) {
+        if ((curCycle() - startTime) > maxFrameTime) {
             warn("CLCD controller buffer underrun, took %d cycles when should"
                  " have taken %d\n", curTick() - startTime, maxFrameTime);
             lcdRis.underflow = 1;
@@ -498,11 +498,13 @@ Pl111::dmaDone()
         pic->seekp(0);
         bmp->write(pic);
 
-        DPRINTF(PL111, "-- schedule next dma read event at %d tick \n",
-                maxFrameTime + curTick());
-
+        // schedule the next read based on when the last frame started
+        // and the desired fps (i.e. maxFrameTime), we turn the
+        // argument into a relative number of cycles in the future by
+        // subtracting curCycle()
         if (lcdControl.lcden)
-            schedule(readEvent, nextCycle(startTime + maxFrameTime));
+            schedule(readEvent, clockEdge(startTime + maxFrameTime -
+                                          curCycle()));
     }
 
     if (dmaPendingNum > (maxOutstandingDma - waterMark))
diff --git a/src/dev/i8254xGBe.cc b/src/dev/i8254xGBe.cc
index f7f6a1178..3ba140bce 100644
--- a/src/dev/i8254xGBe.cc
+++ b/src/dev/i8254xGBe.cc
@@ -2052,7 +2052,7 @@ IGbE::restartClock()
 {
     if (!tickEvent.scheduled() && (rxTick || txTick || txFifoTick) &&
         getState() == SimObject::Running)
-        schedule(tickEvent, (curTick() / ticks(1)) * ticks(1) + ticks(1));
+        schedule(tickEvent, clockEdge(1));
 }
 
 unsigned int
@@ -2434,7 +2434,7 @@ IGbE::tick()
 
 
     if (rxTick || txTick || txFifoTick)
-        schedule(tickEvent, curTick() + ticks(1));
+        schedule(tickEvent, curTick() + clockPeriod());
 }
 
 void
diff --git a/src/dev/ns_gige.cc b/src/dev/ns_gige.cc
index 6a55516c5..583cdb140 100644
--- a/src/dev/ns_gige.cc
+++ b/src/dev/ns_gige.cc
@@ -1147,7 +1147,7 @@ NSGigE::rxKick()
         }
 
         // Go to the next state machine clock tick.
-        rxKickTick = curTick() + ticks(1);
+        rxKickTick = curTick() + clockPeriod();
     }
 
     switch(rxDmaState) {
@@ -1594,7 +1594,7 @@ NSGigE::txKick()
         }
 
         // Go to the next state machine clock tick.
-        txKickTick = curTick() + ticks(1);
+        txKickTick = curTick() + clockPeriod();
     }
 
     switch(txDmaState) {
@@ -2015,7 +2015,7 @@ NSGigE::transferDone()
 
     DPRINTF(Ethernet, "transfer complete: data in txFifo...schedule xmit\n");
 
-    reschedule(txEvent, curTick() + ticks(1), true);
+    reschedule(txEvent, curTick() + clockPeriod(), true);
 }
 
 bool
diff --git a/src/sim/clocked_object.hh b/src/sim/clocked_object.hh
index 8ab637772..050a15a74 100644
--- a/src/sim/clocked_object.hh
+++ b/src/sim/clocked_object.hh
@@ -58,6 +58,14 @@ class ClockedObject : public SimObject
 
   private:
 
+    // the tick value of the next clock edge (>= curTick()) at the
+    // time of the last call to update()
+    mutable Tick tick;
+
+    // The cycle counter value corresponding to the current value of
+    // 'tick'
+    mutable Tick cycle;
+
     /**
      * Prevent inadvertent use of the copy constructor and assignment
      * operator by making them private.
@@ -65,6 +73,34 @@ class ClockedObject : public SimObject
     ClockedObject(ClockedObject&);
     ClockedObject& operator=(ClockedObject&);
 
+    /**
+     *  Align cycle and tick to the next clock edge if not already done.
+     */
+    void update() const
+    {
+        // both tick and cycle are up-to-date and we are done, note
+        // that the >= is important as it captures cases where tick
+        // has already passed curTick()
+        if (tick >= curTick())
+            return;
+
+        // optimise for the common case and see if the tick should be
+        // advanced by a single clock period
+        tick += clock;
+        ++cycle;
+
+        // see if we are done at this point
+        if (tick >= curTick())
+            return;
+
+        // if not, we have to recalculate the cycle and tick, we
+        // perform the calculations in terms of relative cycles to
+        // allow changes to the clock period in the future
+        Tick elapsedCycles = divCeil(curTick() - tick, clock);
+        cycle += elapsedCycles;
+        tick += elapsedCycles * clock;
+    }
+
   protected:
 
     // Clock period in ticks
@@ -74,7 +110,8 @@ class ClockedObject : public SimObject
      * Create a clocked object and set the clock based on the
      * parameters.
      */
-    ClockedObject(const ClockedObjectParams* p) : SimObject(p), clock(p->clock)
+    ClockedObject(const ClockedObjectParams* p) :
+        SimObject(p), tick(0), cycle(0), clock(p->clock)
     { }
 
     /**
@@ -85,33 +122,53 @@ class ClockedObject : public SimObject
   public:
 
     /**
-     * Based on the clock of the object, determine the tick when the
-     * next cycle begins, in other words, round the curTick() to the
-     * next tick that is a multiple of the clock.
+     * Determine the tick when a cycle begins, by default the current
+     * one, but the argument also enables the caller to determine a
+     * future cycle.
      *
-     * @return The tick when the next cycle starts
+     * @param cycles The number of cycles into the future
+     *
+     * @return The tick when the clock edge occurs
      */
-    Tick nextCycle() const
-    { return divCeil(curTick(), clock) * clock; }
+    inline Tick clockEdge(int cycles = 0) const
+    {
+        // align tick to the next clock edge
+        update();
+
+        // figure out when this future cycle is
+        return tick + ticks(cycles);
+    }
 
     /**
-     * Determine the next cycle starting from a given tick instead of
-     * curTick().
+     * Determine the current cycle, corresponding to a tick aligned to
+     * a clock edge.
      *
-     * @param begin_tick The tick to round to a clock edge
+     * @return The current cycle
+     */
+    inline Tick curCycle() const
+    {
+        // align cycle to the next clock edge.
+        update();
+
+        return cycle;
+    }
+
+    /**
+     * Based on the clock of the object, determine the tick when the
+     * next cycle begins, in other words, return the next clock edge.
      *
-     * @return The tick when the cycle after or on begin_tick starts
+     * @return The tick when the next cycle starts
      */
-    Tick nextCycle(Tick begin_tick) const
-    { return divCeil(begin_tick, clock) * clock; }
+    Tick nextCycle() const
+    { return clockEdge(); }
 
     inline Tick frequency() const { return SimClock::Frequency / clock; }
 
-    inline Tick ticks(int numCycles) const { return clock * numCycles; }
+    inline Tick ticks(int cycles) const { return clock * cycles; }
 
-    inline Tick curCycle() const { return curTick() / clock; }
+    inline Tick clockPeriod() const { return clock; }
 
-    inline Tick tickToCycles(Tick val) const { return val / clock; }
+    inline Tick tickToCycle(Tick tick) const { return tick / clock; }
 
 };