Merge zizzer.eecs.umich.edu:/bk/newmem

into  doughnut.mwconnections.com:/home/gblack/newmem-o3-micro

--HG--
extra : convert_revision : 545b9e98eb1895f4b9e782224fb6615c71ed6323

22 files changed, 329 insertions, 441 deletions

diff --git a/src/cpu/o3/alpha/dyn_inst.hh b/src/cpu/o3/alpha/dyn_inst.hh
index 20759d849..a6fb7b885 100644
--- a/src/cpu/o3/alpha/dyn_inst.hh
+++ b/src/cpu/o3/alpha/dyn_inst.hh
@@ -73,8 +73,13 @@ class AlphaDynInst : public BaseDynInst<Impl>
 
   public:
     /** BaseDynInst constructor given a binary instruction. */
-    AlphaDynInst(ExtMachInst inst, Addr PC, Addr NPC,
-                 Addr Pred_PC, Addr Pred_NPC,
+    AlphaDynInst(StaticInstPtr staticInst, Addr PC, Addr NPC, Addr microPC,
+                 Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
+                 InstSeqNum seq_num, O3CPU *cpu);
+
+    /** BaseDynInst constructor given a binary instruction. */
+    AlphaDynInst(ExtMachInst inst, Addr PC, Addr NPC, Addr microPC,
+                 Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
                  InstSeqNum seq_num, O3CPU *cpu);
 
     /** BaseDynInst constructor given a static inst pointer. */
diff --git a/src/cpu/o3/alpha/dyn_inst_impl.hh b/src/cpu/o3/alpha/dyn_inst_impl.hh
index fdce1ade5..6dfe0ccdd 100644
--- a/src/cpu/o3/alpha/dyn_inst_impl.hh
+++ b/src/cpu/o3/alpha/dyn_inst_impl.hh
@@ -31,10 +31,25 @@
 #include "cpu/o3/alpha/dyn_inst.hh"
 
 template <class Impl>
-AlphaDynInst<Impl>::AlphaDynInst(ExtMachInst inst, Addr PC, Addr NPC,
+AlphaDynInst<Impl>::AlphaDynInst(StaticInstPtr staticInst,
+                                 Addr PC, Addr NPC, Addr microPC,
                                  Addr Pred_PC, Addr Pred_NPC,
+                                 Addr Pred_MicroPC,
                                  InstSeqNum seq_num, O3CPU *cpu)
-    : BaseDynInst<Impl>(inst, PC, NPC, Pred_PC, Pred_NPC, seq_num, cpu)
+    : BaseDynInst<Impl>(staticInst, PC, NPC, microPC,
+            Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu)
+{
+    initVars();
+}
+
+template <class Impl>
+AlphaDynInst<Impl>::AlphaDynInst(ExtMachInst inst,
+                                 Addr PC, Addr NPC, Addr microPC,
+                                 Addr Pred_PC, Addr Pred_NPC,
+                                 Addr Pred_MicroPC,
+                                 InstSeqNum seq_num, O3CPU *cpu)
+    : BaseDynInst<Impl>(inst, PC, NPC, microPC,
+            Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu)
 {
     initVars();
 }
diff --git a/src/cpu/o3/comm.hh b/src/cpu/o3/comm.hh
index d96919007..fb772060b 100644
--- a/src/cpu/o3/comm.hh
+++ b/src/cpu/o3/comm.hh
@@ -87,10 +87,10 @@ struct DefaultIEWDefaultCommit {
     bool squash[Impl::MaxThreads];
     bool branchMispredict[Impl::MaxThreads];
     bool branchTaken[Impl::MaxThreads];
-    bool squashDelaySlot[Impl::MaxThreads];
-    uint64_t mispredPC[Impl::MaxThreads];
-    uint64_t nextPC[Impl::MaxThreads];
-    uint64_t nextNPC[Impl::MaxThreads];
+    Addr mispredPC[Impl::MaxThreads];
+    Addr nextPC[Impl::MaxThreads];
+    Addr nextNPC[Impl::MaxThreads];
+    Addr nextMicroPC[Impl::MaxThreads];
     InstSeqNum squashedSeqNum[Impl::MaxThreads];
 
     bool includeSquashInst[Impl::MaxThreads];
@@ -114,15 +114,15 @@ struct TimeBufStruct {
         uint64_t branchAddr;
 
         InstSeqNum doneSeqNum;
-        InstSeqNum bdelayDoneSeqNum;
 
         // @todo: Might want to package this kind of branch stuff into a single
         // struct as it is used pretty frequently.
         bool branchMispredict;
         bool branchTaken;
-        uint64_t mispredPC;
-        uint64_t nextPC;
-        uint64_t nextNPC;
+        Addr mispredPC;
+        Addr nextPC;
+        Addr nextNPC;
+        Addr nextMicroPC;
 
         unsigned branchCount;
     };
@@ -160,18 +160,16 @@ struct TimeBufStruct {
 
         bool branchMispredict;
         bool branchTaken;
-        uint64_t mispredPC;
-        uint64_t nextPC;
-        uint64_t nextNPC;
+        Addr mispredPC;
+        Addr nextPC;
+        Addr nextNPC;
+        Addr nextMicroPC;
 
         // Represents the instruction that has either been retired or
         // squashed.  Similar to having a single bus that broadcasts the
         // retired or squashed sequence number.
         InstSeqNum doneSeqNum;
 
-        InstSeqNum bdelayDoneSeqNum;
-        bool squashDelaySlot;
-
         //Just in case we want to do a commit/squash on a cycle
         //(necessary for multiple ROBs?)
         bool commitInsts;
diff --git a/src/cpu/o3/commit.hh b/src/cpu/o3/commit.hh
index fba618c14..27bdd20c5 100644
--- a/src/cpu/o3/commit.hh
+++ b/src/cpu/o3/commit.hh
@@ -279,25 +279,37 @@ class DefaultCommit
     /** Returns the PC of the head instruction of the ROB.
      * @todo: Probably remove this function as it returns only thread 0.
      */
-    uint64_t readPC() { return PC[0]; }
+    Addr readPC() { return PC[0]; }
 
     /** Returns the PC of a specific thread. */
-    uint64_t readPC(unsigned tid) { return PC[tid]; }
+    Addr readPC(unsigned tid) { return PC[tid]; }
 
     /** Sets the PC of a specific thread. */
-    void setPC(uint64_t val, unsigned tid) { PC[tid] = val; }
+    void setPC(Addr val, unsigned tid) { PC[tid] = val; }
+
+    /** Reads the micro PC of a specific thread. */
+    Addr readMicroPC(unsigned tid) { return microPC[tid]; }
+
+    /** Sets the micro PC of a specific thread */
+    void setMicroPC(Addr val, unsigned tid) { microPC[tid] = val; }
 
     /** Reads the next PC of a specific thread. */
-    uint64_t readNextPC(unsigned tid) { return nextPC[tid]; }
+    Addr readNextPC(unsigned tid) { return nextPC[tid]; }
 
     /** Sets the next PC of a specific thread. */
-    void setNextPC(uint64_t val, unsigned tid) { nextPC[tid] = val; }
+    void setNextPC(Addr val, unsigned tid) { nextPC[tid] = val; }
 
     /** Reads the next NPC of a specific thread. */
-    uint64_t readNextNPC(unsigned tid) { return nextNPC[tid]; }
+    Addr readNextNPC(unsigned tid) { return nextNPC[tid]; }
 
     /** Sets the next NPC of a specific thread. */
-    void setNextNPC(uint64_t val, unsigned tid) { nextNPC[tid] = val; }
+    void setNextNPC(Addr val, unsigned tid) { nextNPC[tid] = val; }
+
+    /** Reads the micro PC of a specific thread. */
+    Addr readNextMicroPC(unsigned tid) { return nextMicroPC[tid]; }
+
+    /** Sets the micro PC of a specific thread */
+    void setNextMicroPC(Addr val, unsigned tid) { nextMicroPC[tid] = val; }
 
   private:
     /** Time buffer interface. */
@@ -402,12 +414,20 @@ class DefaultCommit
      */
     Addr PC[Impl::MaxThreads];
 
+    /** The commit micro PC of each thread.  Refers to the instruction that
+     * is currently being processed/committed.
+     */
+    Addr microPC[Impl::MaxThreads];
+
     /** The next PC of each thread. */
     Addr nextPC[Impl::MaxThreads];
 
     /** The next NPC of each thread. */
     Addr nextNPC[Impl::MaxThreads];
 
+    /** The next micro PC of each thread. */
+    Addr nextMicroPC[Impl::MaxThreads];
+
     /** The sequence number of the youngest valid instruction in the ROB. */
     InstSeqNum youngestSeqNum[Impl::MaxThreads];
 
diff --git a/src/cpu/o3/commit_impl.hh b/src/cpu/o3/commit_impl.hh
index dd4c333d3..9411c6c62 100644
--- a/src/cpu/o3/commit_impl.hh
+++ b/src/cpu/o3/commit_impl.hh
@@ -124,7 +124,7 @@ DefaultCommit<Impl>::DefaultCommit(O3CPU *_cpu, Params *params)
         committedStores[i] = false;
         trapSquash[i] = false;
         tcSquash[i] = false;
-        PC[i] = nextPC[i] = nextNPC[i] = 0;
+        microPC[i] = nextMicroPC[i] = PC[i] = nextPC[i] = nextNPC[i] = 0;
     }
 #if FULL_SYSTEM
     interrupt = NoFault;
@@ -508,6 +508,7 @@ DefaultCommit<Impl>::squashAll(unsigned tid)
 
     toIEW->commitInfo[tid].nextPC = PC[tid];
     toIEW->commitInfo[tid].nextNPC = nextPC[tid];
+    toIEW->commitInfo[tid].nextMicroPC = nextMicroPC[tid];
 }
 
 template <class Impl>
@@ -738,38 +739,15 @@ DefaultCommit<Impl>::commit()
             // then use one older sequence number.
             InstSeqNum squashed_inst = fromIEW->squashedSeqNum[tid];
 
-#if ISA_HAS_DELAY_SLOT
-            InstSeqNum bdelay_done_seq_num = squashed_inst;
-            bool squash_bdelay_slot = fromIEW->squashDelaySlot[tid];
-            bool branchMispredict = fromIEW->branchMispredict[tid];
-
-            // Squashing/not squashing the branch delay slot only makes
-            // sense when you're squashing from a branch, ie from a branch
-            // mispredict.
-            if (branchMispredict && !squash_bdelay_slot) {
-                bdelay_done_seq_num++;
-            }
-#endif
-
             if (fromIEW->includeSquashInst[tid] == true) {
                 squashed_inst--;
-#if ISA_HAS_DELAY_SLOT
-                bdelay_done_seq_num--;
-#endif
             }
 
             // All younger instructions will be squashed. Set the sequence
             // number as the youngest instruction in the ROB.
             youngestSeqNum[tid] = squashed_inst;
 
-#if ISA_HAS_DELAY_SLOT
-            rob->squash(bdelay_done_seq_num, tid);
-            toIEW->commitInfo[tid].squashDelaySlot = squash_bdelay_slot;
-            toIEW->commitInfo[tid].bdelayDoneSeqNum = bdelay_done_seq_num;
-#else
             rob->squash(squashed_inst, tid);
-            toIEW->commitInfo[tid].squashDelaySlot = true;
-#endif
             changedROBNumEntries[tid] = true;
 
             toIEW->commitInfo[tid].doneSeqNum = squashed_inst;
@@ -788,6 +766,7 @@ DefaultCommit<Impl>::commit()
 
             toIEW->commitInfo[tid].nextPC = fromIEW->nextPC[tid];
             toIEW->commitInfo[tid].nextNPC = fromIEW->nextNPC[tid];
+            toIEW->commitInfo[tid].nextMicroPC = fromIEW->nextMicroPC[tid];
 
             toIEW->commitInfo[tid].mispredPC = fromIEW->mispredPC[tid];
 
@@ -806,10 +785,6 @@ DefaultCommit<Impl>::commit()
 
         // Try to commit any instructions.
         commitInsts();
-    } else {
-#if ISA_HAS_DELAY_SLOT
-        skidInsert();
-#endif
     }
 
     //Check for any activity
@@ -901,6 +876,7 @@ DefaultCommit<Impl>::commitInsts()
             PC[tid] = head_inst->readPC();
             nextPC[tid] = head_inst->readNextPC();
             nextNPC[tid] = head_inst->readNextNPC();
+            nextMicroPC[tid] = head_inst->readNextMicroPC();
 
             // Increment the total number of non-speculative instructions
             // executed.
@@ -929,12 +905,10 @@ DefaultCommit<Impl>::commitInsts()
                 }
 
                 PC[tid] = nextPC[tid];
-#if ISA_HAS_DELAY_SLOT
                 nextPC[tid] = nextNPC[tid];
                 nextNPC[tid] = nextNPC[tid] + sizeof(TheISA::MachInst);
-#else
-                nextPC[tid] = nextPC[tid] + sizeof(TheISA::MachInst);
-#endif
+                microPC[tid] = nextMicroPC[tid];
+                nextMicroPC[tid] = microPC[tid] + 1;
 
 #if FULL_SYSTEM
                 int count = 0;
@@ -1161,37 +1135,13 @@ DefaultCommit<Impl>::getInsts()
 {
     DPRINTF(Commit, "Getting instructions from Rename stage.\n");
 
-#if ISA_HAS_DELAY_SLOT
-    // Read any renamed instructions and place them into the ROB.
-    int insts_to_process = std::min((int)renameWidth,
-                               (int)(fromRename->size + skidBuffer.size()));
-    int rename_idx = 0;
-
-    DPRINTF(Commit, "%i insts available to process. Rename Insts:%i "
-            "SkidBuffer Insts:%i\n", insts_to_process, fromRename->size,
-            skidBuffer.size());
-#else
     // Read any renamed instructions and place them into the ROB.
     int insts_to_process = std::min((int)renameWidth, fromRename->size);
-#endif
-
 
     for (int inst_num = 0; inst_num < insts_to_process; ++inst_num) {
         DynInstPtr inst;
 
-#if ISA_HAS_DELAY_SLOT
-        // Get insts from skidBuffer or from Rename
-        if (skidBuffer.size() > 0) {
-            DPRINTF(Commit, "Grabbing skidbuffer inst.\n");
-            inst = skidBuffer.front();
-            skidBuffer.pop();
-        } else {
-            DPRINTF(Commit, "Grabbing rename inst.\n");
-            inst = fromRename->insts[rename_idx++];
-        }
-#else
         inst = fromRename->insts[inst_num];
-#endif
         int tid = inst->threadNumber;
 
         if (!inst->isSquashed() &&
@@ -1213,30 +1163,6 @@ DefaultCommit<Impl>::getInsts()
                     inst->readPC(), inst->seqNum, tid);
         }
     }
-
-#if ISA_HAS_DELAY_SLOT
-    if (rename_idx < fromRename->size) {
-        DPRINTF(Commit,"Placing Rename Insts into skidBuffer.\n");
-
-        for (;
-             rename_idx < fromRename->size;
-             rename_idx++) {
-            DynInstPtr inst = fromRename->insts[rename_idx];
-
-            if (!inst->isSquashed()) {
-                DPRINTF(Commit, "Inserting PC %#x [sn:%i] [tid:%i] into ",
-                        "skidBuffer.\n", inst->readPC(), inst->seqNum,
-                        inst->threadNumber);
-                skidBuffer.push(inst);
-            } else {
-                DPRINTF(Commit, "Instruction PC %#x [sn:%i] [tid:%i] was "
-                        "squashed, skipping.\n",
-                        inst->readPC(), inst->seqNum, inst->threadNumber);
-            }
-        }
-    }
-#endif
-
 }
 
 template <class Impl>
diff --git a/src/cpu/o3/cpu.cc b/src/cpu/o3/cpu.cc
index a775b66d5..8e4c625df 100644
--- a/src/cpu/o3/cpu.cc
+++ b/src/cpu/o3/cpu.cc
@@ -696,7 +696,7 @@ FullO3CPU<Impl>::removeThread(unsigned tid)
 
     // Squash Throughout Pipeline
     InstSeqNum squash_seq_num = commit.rob->readHeadInst(tid)->seqNum;
-    fetch.squash(0, sizeof(TheISA::MachInst), squash_seq_num, true, tid);
+    fetch.squash(0, sizeof(TheISA::MachInst), 0, squash_seq_num, tid);
     decode.squash(tid);
     rename.squash(squash_seq_num, tid);
     iew.squash(tid);
@@ -1152,6 +1152,20 @@ FullO3CPU<Impl>::setPC(Addr new_PC,unsigned tid)
 
 template <class Impl>
 uint64_t
+FullO3CPU<Impl>::readMicroPC(unsigned tid)
+{
+    return commit.readMicroPC(tid);
+}
+
+template <class Impl>
+void
+FullO3CPU<Impl>::setMicroPC(Addr new_PC,unsigned tid)
+{
+    commit.setMicroPC(new_PC, tid);
+}
+
+template <class Impl>
+uint64_t
 FullO3CPU<Impl>::readNextPC(unsigned tid)
 {
     return commit.readNextPC(tid);
@@ -1179,6 +1193,20 @@ FullO3CPU<Impl>::setNextNPC(uint64_t val,unsigned tid)
 }
 
 template <class Impl>
+uint64_t
+FullO3CPU<Impl>::readNextMicroPC(unsigned tid)
+{
+    return commit.readNextMicroPC(tid);
+}
+
+template <class Impl>
+void
+FullO3CPU<Impl>::setNextMicroPC(Addr new_PC,unsigned tid)
+{
+    commit.setNextMicroPC(new_PC, tid);
+}
+
+template <class Impl>
 typename FullO3CPU<Impl>::ListIt
 FullO3CPU<Impl>::addInst(DynInstPtr &inst)
 {
@@ -1226,9 +1254,7 @@ FullO3CPU<Impl>::removeFrontInst(DynInstPtr &inst)
 
 template <class Impl>
 void
-FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid,
-                                     bool squash_delay_slot,
-                                     const InstSeqNum &delay_slot_seq_num)
+FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid)
 {
     DPRINTF(O3CPU, "Thread %i: Deleting instructions from instruction"
             " list.\n", tid);
@@ -1259,12 +1285,6 @@ FullO3CPU<Impl>::removeInstsNotInROB(unsigned tid,
     while (inst_it != end_it) {
         assert(!instList.empty());
 
-#if ISA_HAS_DELAY_SLOT
-        if(!squash_delay_slot &&
-           delay_slot_seq_num >= (*inst_it)->seqNum) {
-            break;
-        }
-#endif
         squashInstIt(inst_it, tid);
 
         inst_it--;
diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh
index e71d05c8e..bff78bf9e 100644
--- a/src/cpu/o3/cpu.hh
+++ b/src/cpu/o3/cpu.hh
@@ -433,22 +433,34 @@ class FullO3CPU : public BaseO3CPU
     void setArchFloatRegInt(int reg_idx, uint64_t val, unsigned tid);
 
     /** Reads the commit PC of a specific thread. */
-    uint64_t readPC(unsigned tid);
+    Addr readPC(unsigned tid);
 
     /** Sets the commit PC of a specific thread. */
     void setPC(Addr new_PC, unsigned tid);
 
+    /** Reads the commit micro PC of a specific thread. */
+    Addr readMicroPC(unsigned tid);
+
+    /** Sets the commmit micro PC of a specific thread. */
+    void setMicroPC(Addr new_microPC, unsigned tid);
+
     /** Reads the next PC of a specific thread. */
-    uint64_t readNextPC(unsigned tid);
+    Addr readNextPC(unsigned tid);
 
     /** Sets the next PC of a specific thread. */
-    void setNextPC(uint64_t val, unsigned tid);
+    void setNextPC(Addr val, unsigned tid);
 
     /** Reads the next NPC of a specific thread. */
-    uint64_t readNextNPC(unsigned tid);
+    Addr readNextNPC(unsigned tid);
 
     /** Sets the next NPC of a specific thread. */
-    void setNextNPC(uint64_t val, unsigned tid);
+    void setNextNPC(Addr val, unsigned tid);
+
+    /** Reads the commit next micro PC of a specific thread. */
+    Addr readNextMicroPC(unsigned tid);
+
+    /** Sets the commit next micro PC of a specific thread. */
+    void setNextMicroPC(Addr val, unsigned tid);
 
     /** Function to add instruction onto the head of the list of the
      *  instructions.  Used when new instructions are fetched.
@@ -468,8 +480,7 @@ 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, bool squash_delay_slot,
-                             const InstSeqNum &delay_slot_seq_num);
+    void removeInstsNotInROB(unsigned tid);
 
     /** Remove all instructions younger than the given sequence number. */
     void removeInstsUntil(const InstSeqNum &seq_num,unsigned tid);
diff --git a/src/cpu/o3/decode_impl.hh b/src/cpu/o3/decode_impl.hh
index 314864f94..ce6738456 100644
--- a/src/cpu/o3/decode_impl.hh
+++ b/src/cpu/o3/decode_impl.hh
@@ -49,8 +49,6 @@ DefaultDecode<Impl>::DefaultDecode(O3CPU *_cpu, Params *params)
         stalls[i].rename = false;
         stalls[i].iew = false;
         stalls[i].commit = false;
-
-        squashAfterDelaySlot[i] = false;
     }
 
     // @todo: Make into a parameter
@@ -275,20 +273,16 @@ DefaultDecode<Impl>::squash(DynInstPtr &inst, unsigned tid)
     ///explicitly for ISAs with delay slots.
     toFetch->decodeInfo[tid].nextNPC =
         inst->branchTarget() + sizeof(TheISA::MachInst);
+    toFetch->decodeInfo[tid].nextMicroPC = inst->readMicroPC();
 #if ISA_HAS_DELAY_SLOT
     toFetch->decodeInfo[tid].branchTaken = inst->readNextNPC() !=
         (inst->readNextPC() + sizeof(TheISA::MachInst));
-
-    toFetch->decodeInfo[tid].bdelayDoneSeqNum = bdelayDoneSeqNum[tid];
-    squashAfterDelaySlot[tid] = false;
-
-    InstSeqNum squash_seq_num = bdelayDoneSeqNum[tid];
 #else
     toFetch->decodeInfo[tid].branchTaken =
         inst->readNextPC() != (inst->readPC() + sizeof(TheISA::MachInst));
+#endif
 
     InstSeqNum squash_seq_num = inst->seqNum;
-#endif
 
     // Might have to tell fetch to unblock.
     if (decodeStatus[tid] == Blocked ||
@@ -309,30 +303,10 @@ DefaultDecode<Impl>::squash(DynInstPtr &inst, unsigned tid)
     // Clear the instruction list and skid buffer in case they have any
     // insts in them.
     while (!insts[tid].empty()) {
-
-#if ISA_HAS_DELAY_SLOT
-        if (insts[tid].front()->seqNum <= squash_seq_num) {
-            DPRINTF(Decode, "[tid:%i]: Cannot remove incoming decode "
-                    "instructions before delay slot [sn:%i]. %i insts"
-                    "left in decode.\n", tid, squash_seq_num,
-                    insts[tid].size());
-            break;
-        }
-#endif
         insts[tid].pop();
     }
 
     while (!skidBuffer[tid].empty()) {
-
-#if ISA_HAS_DELAY_SLOT
-        if (skidBuffer[tid].front()->seqNum <= squash_seq_num) {
-            DPRINTF(Decode, "[tid:%i]: Cannot remove skidBuffer "
-                    "instructions before delay slot [sn:%i]. %i insts"
-                    "left in decode.\n", tid, squash_seq_num,
-                    insts[tid].size());
-            break;
-        }
-#endif
         skidBuffer[tid].pop();
     }
 
@@ -760,48 +734,13 @@ DefaultDecode<Impl>::decodeInsts(unsigned tid)
 
                 // Might want to set some sort of boolean and just do
                 // a check at the end
-#if !ISA_HAS_DELAY_SLOT
                 squash(inst, inst->threadNumber);
                 Addr target = inst->branchTarget();
-                inst->setPredTarg(target, target + sizeof(TheISA::MachInst));
+                //The micro pc after an instruction level branch should be 0
+                inst->setPredTarg(target, target + sizeof(TheISA::MachInst), 0);
                 break;
-#else
-                // If mispredicted as taken, then ignore delay slot
-                // instruction... else keep delay slot and squash
-                // after it is sent to rename
-                if (inst->readPredTaken() && inst->isCondDelaySlot()) {
-                    DPRINTF(Decode, "[tid:%i]: Conditional delay slot inst."
-                            "[sn:%i] PC %#x mispredicted as taken.\n", tid,
-                            inst->seqNum, inst->PC);
-                    bdelayDoneSeqNum[tid] = inst->seqNum;
-                    squash(inst, inst->threadNumber);
-                    Addr target = inst->branchTarget();
-                    inst->setPredTarg(target,
-                            target + sizeof(TheISA::MachInst));
-                    break;
-                } else {
-                    DPRINTF(Decode, "[tid:%i]: Misprediction detected at "
-                            "[sn:%i] PC %#x, will squash after delay slot "
-                            "inst. is sent to Rename\n",
-                            tid, inst->seqNum, inst->PC);
-                    bdelayDoneSeqNum[tid] = inst->seqNum + 1;
-                    squashAfterDelaySlot[tid] = true;
-                    squashInst[tid] = inst;
-                    continue;
-                }
-#endif
             }
         }
-
-        if (squashAfterDelaySlot[tid]) {
-            assert(!inst->isSquashed());
-            squash(squashInst[tid], squashInst[tid]->threadNumber);
-            Addr target = squashInst[tid]->branchTarget();
-            squashInst[tid]->setPredTarg(target,
-                    target + sizeof(TheISA::MachInst));
-            assert(!inst->isSquashed());
-            break;
-        }
     }
 
     // If we didn't process all instructions, then we will need to block
diff --git a/src/cpu/o3/fetch.hh b/src/cpu/o3/fetch.hh
index 241935416..7645a226c 100644
--- a/src/cpu/o3/fetch.hh
+++ b/src/cpu/o3/fetch.hh
@@ -227,7 +227,7 @@ class DefaultFetch
      * @param next_NPC Used for ISAs which use delay slots.
      * @return Whether or not a branch was predicted as taken.
      */
-    bool lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, Addr &next_NPC);
+    bool lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, Addr &next_NPC, Addr &next_MicroPC);
 
     /**
      * Fetches the cache line that contains fetch_PC.  Returns any
@@ -242,12 +242,14 @@ class DefaultFetch
     bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid);
 
     /** Squashes a specific thread and resets the PC. */
-    inline void doSquash(const Addr &new_PC, const Addr &new_NPC, unsigned tid);
+    inline void doSquash(const Addr &new_PC, const Addr &new_NPC,
+                         const Addr &new_MicroPC, unsigned 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);
 
     /** Checks if a thread is stalled. */
@@ -263,8 +265,8 @@ class DefaultFetch
      * squash should be the commit stage.
      */
     void squash(const Addr &new_PC, const Addr &new_NPC,
-                const InstSeqNum &seq_num,
-                bool squash_delay_slot, unsigned tid);
+                const Addr &new_MicroPC,
+                const InstSeqNum &seq_num, unsigned tid);
 
     /** Ticks the fetch stage, processing all inputs signals and fetching
      * as many instructions as possible.
@@ -347,16 +349,12 @@ class DefaultFetch
     /** Per-thread fetch PC. */
     Addr PC[Impl::MaxThreads];
 
+    /** Per-thread fetch micro PC. */
+    Addr microPC[Impl::MaxThreads];
+
     /** Per-thread next PC. */
     Addr nextPC[Impl::MaxThreads];
 
-    /** Per-thread next Next PC.
-     *  This is not a real register but is used for
-     *  architectures that use a branch-delay slot.
-     *  (such as MIPS or Sparc)
-     */
-    Addr nextNPC[Impl::MaxThreads];
-
     /** Memory request used to access cache. */
     RequestPtr memReq[Impl::MaxThreads];
 
diff --git a/src/cpu/o3/fetch_impl.hh b/src/cpu/o3/fetch_impl.hh
index e16f97558..3ae7bc402 100644
--- a/src/cpu/o3/fetch_impl.hh
+++ b/src/cpu/o3/fetch_impl.hh
@@ -312,7 +312,7 @@ DefaultFetch<Impl>::initStage()
     for (int tid = 0; tid < numThreads; tid++) {
         PC[tid] = cpu->readPC(tid);
         nextPC[tid] = cpu->readNextPC(tid);
-        nextNPC[tid] = cpu->readNextNPC(tid);
+        microPC[tid] = cpu->readMicroPC(tid);
     }
 
     for (int tid=0; tid < numThreads; tid++) {
@@ -439,11 +439,7 @@ DefaultFetch<Impl>::takeOverFrom()
         stalls[i].commit = 0;
         PC[i] = cpu->readPC(i);
         nextPC[i] = cpu->readNextPC(i);
-#if ISA_HAS_DELAY_SLOT
-        nextNPC[i] = cpu->readNextNPC(i);
-#else
-        nextNPC[i] = nextPC[i] + sizeof(TheISA::MachInst);
-#endif
+        microPC[i] = cpu->readMicroPC(i);
         fetchStatus[i] = Running;
     }
     numInst = 0;
@@ -493,7 +489,7 @@ DefaultFetch<Impl>::switchToInactive()
 template <class Impl>
 bool
 DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC,
-                                          Addr &next_NPC)
+                                          Addr &next_NPC, Addr &next_MicroPC)
 {
     // Do branch prediction check here.
     // A bit of a misnomer...next_PC is actually the current PC until
@@ -501,13 +497,22 @@ DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC,
     bool predict_taken;
 
     if (!inst->isControl()) {
-        next_PC  = next_NPC;
-        next_NPC = next_NPC + instSize;
-        inst->setPredTarg(next_PC, next_NPC);
+        if (inst->isMicroOp() && !inst->isLastMicroOp()) {
+            next_MicroPC++;
+        } else {
+            next_PC  = next_NPC;
+            next_NPC = next_NPC + instSize;
+            next_MicroPC = 0;
+        }
+        inst->setPredTarg(next_PC, next_NPC, next_MicroPC);
         inst->setPredTaken(false);
         return false;
     }
 
+    //Assume for now that all control flow is to a different macroop which
+    //would reset the micro pc to 0.
+    next_MicroPC = 0;
+
     int tid = inst->threadNumber;
     Addr pred_PC = next_PC;
     predict_taken = branchPred.predict(inst, pred_PC, tid);
@@ -534,7 +539,7 @@ DefaultFetch<Impl>::lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC,
 #endif
 /*    DPRINTF(Fetch, "[tid:%i]: Branch predicted to go to %#x and then %#x.\n",
             tid, next_PC, next_NPC);*/
-    inst->setPredTarg(next_PC, next_NPC);
+    inst->setPredTarg(next_PC, next_NPC, next_MicroPC);
     inst->setPredTaken(predict_taken);
 
     ++fetchedBranches;
@@ -658,14 +663,14 @@ 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, unsigned tid)
+        const Addr &new_NPC, const Addr &new_microPC, unsigned tid)
 {
     DPRINTF(Fetch, "[tid:%i]: Squashing, setting PC to: %#x, NPC to: %#x.\n",
             tid, new_PC, new_NPC);
 
     PC[tid] = new_PC;
     nextPC[tid] = new_NPC;
-    nextNPC[tid] = new_NPC + instSize;
+    microPC[tid] = new_microPC;
 
     // Clear the icache miss if it's outstanding.
     if (fetchStatus[tid] == IcacheWaitResponse) {
@@ -693,12 +698,12 @@ DefaultFetch<Impl>::doSquash(const Addr &new_PC,
 template<class Impl>
 void
 DefaultFetch<Impl>::squashFromDecode(const Addr &new_PC, const Addr &new_NPC,
-                                     const InstSeqNum &seq_num,
-                                     unsigned tid)
+                                     const Addr &new_MicroPC,
+                                     const InstSeqNum &seq_num, unsigned tid)
 {
     DPRINTF(Fetch, "[tid:%i]: Squashing from decode.\n",tid);
 
-    doSquash(new_PC, new_NPC, tid);
+    doSquash(new_PC, new_NPC, new_MicroPC, tid);
 
     // Tell the CPU to remove any instructions that are in flight between
     // fetch and decode.
@@ -774,20 +779,15 @@ DefaultFetch<Impl>::updateFetchStatus()
 template <class Impl>
 void
 DefaultFetch<Impl>::squash(const Addr &new_PC, const Addr &new_NPC,
-                           const InstSeqNum &seq_num,
-                           bool squash_delay_slot, unsigned tid)
+                           const Addr &new_MicroPC,
+                           const InstSeqNum &seq_num, unsigned tid)
 {
     DPRINTF(Fetch, "[tid:%u]: Squash from commit.\n",tid);
 
-    doSquash(new_PC, new_NPC, tid);
+    doSquash(new_PC, new_NPC, new_MicroPC, tid);
 
-#if ISA_HAS_DELAY_SLOT
     // Tell the CPU to remove any instructions that are not in the ROB.
-    cpu->removeInstsNotInROB(tid, squash_delay_slot, seq_num);
-#else
-    // Tell the CPU to remove any instructions that are not in the ROB.
-    cpu->removeInstsNotInROB(tid, true, 0);
-#endif
+    cpu->removeInstsNotInROB(tid);
 }
 
 template <class Impl>
@@ -896,17 +896,11 @@ DefaultFetch<Impl>::checkSignalsAndUpdate(unsigned tid)
 
         DPRINTF(Fetch, "[tid:%u]: Squashing instructions due to squash "
                 "from commit.\n",tid);
-
-#if ISA_HAS_DELAY_SLOT
-    InstSeqNum doneSeqNum = fromCommit->commitInfo[tid].bdelayDoneSeqNum;
-#else
-    InstSeqNum doneSeqNum = fromCommit->commitInfo[tid].doneSeqNum;
-#endif
         // In any case, squash.
         squash(fromCommit->commitInfo[tid].nextPC,
                fromCommit->commitInfo[tid].nextNPC,
-               doneSeqNum,
-               fromCommit->commitInfo[tid].squashDelaySlot,
+               fromCommit->commitInfo[tid].nextMicroPC,
+               fromCommit->commitInfo[tid].doneSeqNum,
                tid);
 
         // Also check if there's a mispredict that happened.
@@ -955,18 +949,14 @@ DefaultFetch<Impl>::checkSignalsAndUpdate(unsigned tid)
 
         if (fetchStatus[tid] != Squashing) {
 
-#if ISA_HAS_DELAY_SLOT
-            InstSeqNum doneSeqNum = fromDecode->decodeInfo[tid].bdelayDoneSeqNum;
-#else
-            InstSeqNum doneSeqNum = fromDecode->decodeInfo[tid].doneSeqNum;
-#endif
             DPRINTF(Fetch, "Squashing from decode with PC = %#x, NPC = %#x\n",
                     fromDecode->decodeInfo[tid].nextPC,
                     fromDecode->decodeInfo[tid].nextNPC);
             // Squash unless we're already squashing
             squashFromDecode(fromDecode->decodeInfo[tid].nextPC,
                              fromDecode->decodeInfo[tid].nextNPC,
-                             doneSeqNum,
+                             fromDecode->decodeInfo[tid].nextMicroPC,
+                             fromDecode->decodeInfo[tid].doneSeqNum,
                              tid);
 
             return true;
@@ -1020,9 +1010,9 @@ DefaultFetch<Impl>::fetch(bool &status_change)
     DPRINTF(Fetch, "Attempting to fetch from [tid:%i]\n", tid);
 
     // The current PC.
-    Addr &fetch_PC = PC[tid];
-
-    Addr &fetch_NPC = nextPC[tid];
+    Addr fetch_PC = PC[tid];
+    Addr fetch_NPC = nextPC[tid];
+    Addr fetch_MicroPC = microPC[tid];
 
     // Fault code for memory access.
     Fault fault = NoFault;
@@ -1081,6 +1071,7 @@ DefaultFetch<Impl>::fetch(bool &status_change)
 
     Addr next_PC = fetch_PC;
     Addr next_NPC = fetch_NPC;
+    Addr next_MicroPC = fetch_MicroPC;
 
     InstSeqNum inst_seq;
     MachInst inst;
@@ -1088,6 +1079,9 @@ DefaultFetch<Impl>::fetch(bool &status_change)
     // @todo: Fix this hack.
     unsigned offset = (fetch_PC & cacheBlkMask) & ~3;
 
+    StaticInstPtr staticInst = NULL;
+    StaticInstPtr macroop = NULL;
+
     if (fault == NoFault) {
         // If the read of the first instruction was successful, then grab the
         // instructions from the rest of the cache line and put them into the
@@ -1100,11 +1094,9 @@ DefaultFetch<Impl>::fetch(bool &status_change)
         // ended this fetch block.
         bool predicted_branch = false;
 
-        for (;
-             offset < cacheBlkSize &&
-                 numInst < fetchWidth &&
-                 !predicted_branch;
-             ++numInst) {
+        while (offset < cacheBlkSize &&
+               numInst < fetchWidth &&
+               !predicted_branch) {
 
             // If we're branching after this instruction, quite fetching
             // from the same block then.
@@ -1115,91 +1107,103 @@ DefaultFetch<Impl>::fetch(bool &status_change)
                         fetch_PC, fetch_NPC);
             }
 
-
-            // Get a sequence number.
-            inst_seq = cpu->getAndIncrementInstSeq();
-
             // Make sure this is a valid index.
             assert(offset <= cacheBlkSize - instSize);
 
-            // Get the instruction from the array of the cache line.
-            inst = TheISA::gtoh(*reinterpret_cast<TheISA::MachInst *>
-                        (&cacheData[tid][offset]));
+            if (!macroop) {
+                // Get the instruction from the array of the cache line.
+                inst = TheISA::gtoh(*reinterpret_cast<TheISA::MachInst *>
+                            (&cacheData[tid][offset]));
 
-            predecoder.setTC(cpu->thread[tid]->getTC());
-            predecoder.moreBytes(fetch_PC, 0, inst);
+                predecoder.setTC(cpu->thread[tid]->getTC());
+                predecoder.moreBytes(fetch_PC, 0, inst);
 
-            ext_inst = predecoder.getExtMachInst();
+                ext_inst = predecoder.getExtMachInst();
+                staticInst = StaticInstPtr(ext_inst);
+                if (staticInst->isMacroOp())
+                    macroop = staticInst;
+            }
+            do {
+                if (macroop) {
+                    staticInst = macroop->fetchMicroOp(fetch_MicroPC);
+                    if (staticInst->isLastMicroOp())
+                        macroop = NULL;
+                }
 
-            // Create a new DynInst from the instruction fetched.
-            DynInstPtr instruction = new DynInst(ext_inst,
-                                                 fetch_PC, fetch_NPC,
-                                                 next_PC, next_NPC,
-                                                 inst_seq, cpu);
-            instruction->setTid(tid);
+                // Get a sequence number.
+                inst_seq = cpu->getAndIncrementInstSeq();
 
-            instruction->setASID(tid);
+                // Create a new DynInst from the instruction fetched.
+                DynInstPtr instruction = new DynInst(staticInst,
+                                                     fetch_PC, fetch_NPC, fetch_MicroPC,
+                                                     next_PC, next_NPC, next_MicroPC,
+                                                     inst_seq, cpu);
+                instruction->setTid(tid);
 
-            instruction->setThreadState(cpu->thread[tid]);
+                instruction->setASID(tid);
 
-            DPRINTF(Fetch, "[tid:%i]: Instruction PC %#x created "
-                    "[sn:%lli]\n",
-                    tid, instruction->readPC(), inst_seq);
+                instruction->setThreadState(cpu->thread[tid]);
 
-            //DPRINTF(Fetch, "[tid:%i]: MachInst is %#x\n", tid, ext_inst);
+                DPRINTF(Fetch, "[tid:%i]: Instruction PC %#x created "
+                        "[sn:%lli]\n",
+                        tid, instruction->readPC(), inst_seq);
 
-            DPRINTF(Fetch, "[tid:%i]: Instruction is: %s\n",
-                    tid, instruction->staticInst->disassemble(fetch_PC));
+                //DPRINTF(Fetch, "[tid:%i]: MachInst is %#x\n", tid, ext_inst);
 
-            instruction->traceData =
-                Trace::getInstRecord(curTick, cpu->tcBase(tid),
-                                     instruction->staticInst,
-                                     instruction->readPC());
+                DPRINTF(Fetch, "[tid:%i]: Instruction is: %s\n",
+                        tid, instruction->staticInst->disassemble(fetch_PC));
 
-            ///FIXME This needs to be more robust in dealing with delay slots
-#if !ISA_HAS_DELAY_SLOT
-//	    predicted_branch |=
-#endif
-            lookupAndUpdateNextPC(instruction, next_PC, next_NPC);
-            predicted_branch |= (next_PC != fetch_NPC);
+                instruction->traceData =
+                    Trace::getInstRecord(curTick, cpu->tcBase(tid),
+                                         instruction->staticInst,
+                                         instruction->readPC());
 
-            // Add instruction to the CPU's list of instructions.
-            instruction->setInstListIt(cpu->addInst(instruction));
+                ///FIXME This needs to be more robust in dealing with delay slots
+                predicted_branch |=
+                    lookupAndUpdateNextPC(instruction, next_PC, next_NPC, next_MicroPC);
 
-            // Write the instruction to the first slot in the queue
-            // that heads to decode.
-            toDecode->insts[numInst] = instruction;
+                // Add instruction to the CPU's list of instructions.
+                instruction->setInstListIt(cpu->addInst(instruction));
 
-            toDecode->size++;
+                // Write the instruction to the first slot in the queue
+                // that heads to decode.
+                toDecode->insts[numInst] = instruction;
 
-            // Increment stat of fetched instructions.
-            ++fetchedInsts;
+                toDecode->size++;
 
-            // Move to the next instruction, unless we have a branch.
-            fetch_PC = next_PC;
-            fetch_NPC = next_NPC;
+                // Increment stat of fetched instructions.
+                ++fetchedInsts;
 
-            if (instruction->isQuiesce()) {
-                DPRINTF(Fetch, "Quiesce instruction encountered, halting fetch!",
-                        curTick);
-                fetchStatus[tid] = QuiescePending;
-                ++numInst;
-                status_change = true;
-                break;
-            }
+                // Move to the next instruction, unless we have a branch.
+                fetch_PC = next_PC;
+                fetch_NPC = next_NPC;
+                fetch_MicroPC = next_MicroPC;
+
+                if (instruction->isQuiesce()) {
+                    DPRINTF(Fetch, "Quiesce instruction encountered, halting fetch!",
+                            curTick);
+                    fetchStatus[tid] = QuiescePending;
+                    ++numInst;
+                    status_change = true;
+                    break;
+                }
 
+                ++numInst;
+            } while (staticInst->isMicroOp() &&
+                     !staticInst->isLastMicroOp() &&
+                     numInst < fetchWidth);
             offset += instSize;
         }
 
-        if (offset >= cacheBlkSize) {
-            DPRINTF(Fetch, "[tid:%i]: Done fetching, reached the end of cache "
-                    "block.\n", tid);
+        if (predicted_branch) {
+            DPRINTF(Fetch, "[tid:%i]: Done fetching, predicted branch "
+                    "instruction encountered.\n", tid);
         } else if (numInst >= fetchWidth) {
             DPRINTF(Fetch, "[tid:%i]: Done fetching, reached fetch bandwidth "
                     "for this cycle.\n", tid);
-        } else if (predicted_branch) {
-            DPRINTF(Fetch, "[tid:%i]: Done fetching, predicted branch "
-                    "instruction encountered.\n", tid);
+        } else if (offset >= cacheBlkSize) {
+            DPRINTF(Fetch, "[tid:%i]: Done fetching, reached the end of cache "
+                    "block.\n", tid);
         }
     }
 
@@ -1212,12 +1216,8 @@ DefaultFetch<Impl>::fetch(bool &status_change)
     if (fault == NoFault) {
         PC[tid] = next_PC;
         nextPC[tid] = next_NPC;
-        nextNPC[tid] = next_NPC + instSize;
-#if ISA_HAS_DELAY_SLOT
-        DPRINTF(Fetch, "[tid:%i]: Setting PC to %08p.\n", tid, PC[tid]);
-#else
+        microPC[tid] = next_MicroPC;
         DPRINTF(Fetch, "[tid:%i]: Setting PC to %08p.\n", tid, next_PC);
-#endif
     } else {
         // We shouldn't be in an icache miss and also have a fault (an ITB
         // miss)
@@ -1235,8 +1235,9 @@ DefaultFetch<Impl>::fetch(bool &status_change)
         // We will use a nop in order to carry the fault.
         ext_inst = TheISA::NoopMachInst;
 
+        StaticInstPtr staticInst = new StaticInst(ext_inst);
         // Create a new DynInst from the dummy nop.
-        DynInstPtr instruction = new DynInst(ext_inst,
+        DynInstPtr instruction = new DynInst(staticInst,
                                              fetch_PC, fetch_NPC,
                                              next_PC, next_NPC,
                                              inst_seq, cpu);
diff --git a/src/cpu/o3/free_list.hh b/src/cpu/o3/free_list.hh
index c669b0b34..42fc0c533 100644
--- a/src/cpu/o3/free_list.hh
+++ b/src/cpu/o3/free_list.hh
@@ -168,7 +168,9 @@ SimpleFreeList::addReg(PhysRegIndex freed_reg)
         if (freed_reg != TheISA::ZeroReg)
             freeIntRegs.push(freed_reg);
     } else if (freed_reg < numPhysicalRegs) {
+#if THE_ISA == ALPHA_ISA
         if (freed_reg != (TheISA::ZeroReg + numPhysicalIntRegs))
+#endif
             freeFloatRegs.push(freed_reg);
     }
 }
diff --git a/src/cpu/o3/iew.hh b/src/cpu/o3/iew.hh
index ce2991cfb..eef5a15d2 100644
--- a/src/cpu/o3/iew.hh
+++ b/src/cpu/o3/iew.hh
@@ -402,9 +402,6 @@ class DefaultIEW
     /** Records if there is a fetch redirect on this cycle for each thread. */
     bool fetchRedirect[Impl::MaxThreads];
 
-    /** Keeps track of the last valid branch delay slot instss for threads */
-    InstSeqNum bdelayDoneSeqNum[Impl::MaxThreads];
-
     /** Used to track if all instructions have been dispatched this cycle.
      * If they have not, then blocking must have occurred, and the instructions
      * would already be added to the skid buffer.
diff --git a/src/cpu/o3/iew_impl.hh b/src/cpu/o3/iew_impl.hh
index 62e656e93..399c44909 100644
--- a/src/cpu/o3/iew_impl.hh
+++ b/src/cpu/o3/iew_impl.hh
@@ -69,7 +69,6 @@ DefaultIEW<Impl>::DefaultIEW(O3CPU *_cpu, Params *params)
         dispatchStatus[i] = Running;
         stalls[i].commit = false;
         fetchRedirect[i] = false;
-        bdelayDoneSeqNum[i] = 0;
     }
 
     wbMax = wbWidth * params->wbDepth;
@@ -410,31 +409,14 @@ DefaultIEW<Impl>::squash(unsigned tid)
     instQueue.squash(tid);
 
     // Tell the LDSTQ to start squashing.
-#if ISA_HAS_DELAY_SLOT
-    ldstQueue.squash(fromCommit->commitInfo[tid].bdelayDoneSeqNum, tid);
-#else
     ldstQueue.squash(fromCommit->commitInfo[tid].doneSeqNum, tid);
-#endif
     updatedQueues = true;
 
     // Clear the skid buffer in case it has any data in it.
     DPRINTF(IEW, "[tid:%i]: Removing skidbuffer instructions until [sn:%i].\n",
-            tid, fromCommit->commitInfo[tid].bdelayDoneSeqNum);
+            tid, fromCommit->commitInfo[tid].doneSeqNum);
 
     while (!skidBuffer[tid].empty()) {
-#if ISA_HAS_DELAY_SLOT
-        if (skidBuffer[tid].front()->seqNum <=
-            fromCommit->commitInfo[tid].bdelayDoneSeqNum) {
-            DPRINTF(IEW, "[tid:%i]: Cannot remove skidbuffer instructions "
-                    "that occur before delay slot [sn:%i].\n",
-                    fromCommit->commitInfo[tid].bdelayDoneSeqNum,
-                    tid);
-            break;
-        } else {
-            DPRINTF(IEW, "[tid:%i]: Removing instruction [sn:%i] from "
-                    "skidBuffer.\n", tid, skidBuffer[tid].front()->seqNum);
-        }
-#endif
         if (skidBuffer[tid].front()->isLoad() ||
             skidBuffer[tid].front()->isStore() ) {
             toRename->iewInfo[tid].dispatchedToLSQ++;
@@ -445,8 +427,6 @@ DefaultIEW<Impl>::squash(unsigned tid)
         skidBuffer[tid].pop();
     }
 
-    bdelayDoneSeqNum[tid] = fromCommit->commitInfo[tid].bdelayDoneSeqNum;
-
     emptyRenameInsts(tid);
 }
 
@@ -462,38 +442,19 @@ DefaultIEW<Impl>::squashDueToBranch(DynInstPtr &inst, unsigned tid)
     toCommit->mispredPC[tid] = inst->readPC();
     toCommit->branchMispredict[tid] = true;
 
-    int instSize = sizeof(TheISA::MachInst);
 #if ISA_HAS_DELAY_SLOT
-    bool branch_taken =
+    int instSize = sizeof(TheISA::MachInst);
+    toCommit->branchTaken[tid] =
         !(inst->readNextPC() + instSize == inst->readNextNPC() &&
           (inst->readNextPC() == inst->readPC() + instSize ||
            inst->readNextPC() == inst->readPC() + 2 * instSize));
-    DPRINTF(Sparc, "Branch taken = %s [sn:%i]\n",
-            branch_taken ? "true": "false", inst->seqNum);
-
-    toCommit->branchTaken[tid] = branch_taken;
-
-    bool squashDelaySlot = true;
-//	(inst->readNextPC() != inst->readPC() + sizeof(TheISA::MachInst));
-    DPRINTF(Sparc, "Squash delay slot = %s [sn:%i]\n",
-            squashDelaySlot ? "true": "false", inst->seqNum);
-    toCommit->squashDelaySlot[tid] = squashDelaySlot;
-    //If we're squashing the delay slot, we need to pick back up at NextPC.
-    //Otherwise, NextPC isn't being squashed, so we should pick back up at
-    //NextNPC.
-    if (squashDelaySlot) {
-        toCommit->nextPC[tid] = inst->readNextPC();
-        toCommit->nextNPC[tid] = inst->readNextNPC();
-    } else {
-        toCommit->nextPC[tid] = inst->readNextNPC();
-        toCommit->nextNPC[tid] = inst->readNextNPC() + instSize;
-    }
 #else
     toCommit->branchTaken[tid] = inst->readNextPC() !=
         (inst->readPC() + sizeof(TheISA::MachInst));
-    toCommit->nextPC[tid] = inst->readNextPC();
-    toCommit->nextNPC[tid] = inst->readNextPC() + instSize;
 #endif
+    toCommit->nextPC[tid] = inst->readNextPC();
+    toCommit->nextNPC[tid] = inst->readNextNPC();
+    toCommit->nextMicroPC[tid] = inst->readNextMicroPC();
 
     toCommit->includeSquashInst[tid] = false;
 
@@ -510,11 +471,7 @@ DefaultIEW<Impl>::squashDueToMemOrder(DynInstPtr &inst, unsigned tid)
     toCommit->squash[tid] = true;
     toCommit->squashedSeqNum[tid] = inst->seqNum;
     toCommit->nextPC[tid] = inst->readNextPC();
-#if ISA_HAS_DELAY_SLOT
     toCommit->nextNPC[tid] = inst->readNextNPC();
-#else
-    toCommit->nextNPC[tid] = inst->readNextPC() + sizeof(TheISA::MachInst);
-#endif
     toCommit->branchMispredict[tid] = false;
 
     toCommit->includeSquashInst[tid] = false;
@@ -532,11 +489,7 @@ DefaultIEW<Impl>::squashDueToMemBlocked(DynInstPtr &inst, unsigned tid)
     toCommit->squash[tid] = true;
     toCommit->squashedSeqNum[tid] = inst->seqNum;
     toCommit->nextPC[tid] = inst->readPC();
-#if ISA_HAS_DELAY_SLOT
     toCommit->nextNPC[tid] = inst->readNextPC();
-#else
-    toCommit->nextNPC[tid] = inst->readPC() + sizeof(TheISA::MachInst);
-#endif
     toCommit->branchMispredict[tid] = false;
 
     // Must include the broadcasted SN in the squash.
@@ -880,11 +833,9 @@ DefaultIEW<Impl>::sortInsts()
 {
     int insts_from_rename = fromRename->size;
 #ifdef DEBUG
-#if !ISA_HAS_DELAY_SLOT
     for (int i = 0; i < numThreads; i++)
         assert(insts[i].empty());
 #endif
-#endif
     for (int i = 0; i < insts_from_rename; ++i) {
         insts[fromRename->insts[i]->threadNumber].push(fromRename->insts[i]);
     }
@@ -894,21 +845,9 @@ template <class Impl>
 void
 DefaultIEW<Impl>::emptyRenameInsts(unsigned tid)
 {
-    DPRINTF(IEW, "[tid:%i]: Removing incoming rename instructions until "
-            "[sn:%i].\n", tid, bdelayDoneSeqNum[tid]);
+    DPRINTF(IEW, "[tid:%i]: Removing incoming rename instructions\n", tid);
 
     while (!insts[tid].empty()) {
-#if ISA_HAS_DELAY_SLOT
-        if (insts[tid].front()->seqNum <= bdelayDoneSeqNum[tid]) {
-            DPRINTF(IEW, "[tid:%i]: Done removing, cannot remove instruction"
-                    " that occurs at or before delay slot [sn:%i].\n",
-                    tid, bdelayDoneSeqNum[tid]);
-            break;
-        } else {
-            DPRINTF(IEW, "[tid:%i]: Removing incoming rename instruction "
-                    "[sn:%i].\n", tid, insts[tid].front()->seqNum);
-        }
-#endif
 
         if (insts[tid].front()->isLoad() ||
             insts[tid].front()->isStore() ) {
diff --git a/src/cpu/o3/inst_queue_impl.hh b/src/cpu/o3/inst_queue_impl.hh
index 10c3287f2..bdf5f07aa 100644
--- a/src/cpu/o3/inst_queue_impl.hh
+++ b/src/cpu/o3/inst_queue_impl.hh
@@ -1005,11 +1005,7 @@ InstructionQueue<Impl>::squash(unsigned tid)
 
     // Read instruction sequence number of last instruction out of the
     // time buffer.
-#if ISA_HAS_DELAY_SLOT
-    squashedSeqNum[tid] = fromCommit->commitInfo[tid].bdelayDoneSeqNum;
-#else
     squashedSeqNum[tid] = fromCommit->commitInfo[tid].doneSeqNum;
-#endif
 
     // Call doSquash if there are insts in the IQ
     if (count[tid] > 0) {
diff --git a/src/cpu/o3/mips/dyn_inst.hh b/src/cpu/o3/mips/dyn_inst.hh
index 366b4bb23..cf78c0941 100755
--- a/src/cpu/o3/mips/dyn_inst.hh
+++ b/src/cpu/o3/mips/dyn_inst.hh
@@ -70,9 +70,15 @@ class MipsDynInst : public BaseDynInst<Impl>
 
   public:
     /** BaseDynInst constructor given a binary instruction. */
+    MipsDynInst(StaticInstPtr staticInst,
+                Addr PC, Addr NPC, Addr microPC,
+                Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
+                InstSeqNum seq_num, O3CPU *cpu);
+
+    /** BaseDynInst constructor given a binary instruction. */
     MipsDynInst(ExtMachInst inst,
-                Addr PC, Addr NPC,
-                Addr Pred_PC, Addr Pred_NPC,
+                Addr PC, Addr NPC, Addr microPC,
+                Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
                 InstSeqNum seq_num, O3CPU *cpu);
 
     /** BaseDynInst constructor given a static inst pointer. */
diff --git a/src/cpu/o3/mips/dyn_inst_impl.hh b/src/cpu/o3/mips/dyn_inst_impl.hh
index c0f9ae771..7e8697b32 100755
--- a/src/cpu/o3/mips/dyn_inst_impl.hh
+++ b/src/cpu/o3/mips/dyn_inst_impl.hh
@@ -31,11 +31,23 @@
 #include "cpu/o3/mips/dyn_inst.hh"
 
 template <class Impl>
+MipsDynInst<Impl>::MipsDynInst(StaticInstPtr staticInst,
+                               Addr PC, Addr NPC, Addr microPC,
+                               Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
+                               InstSeqNum seq_num, O3CPU *cpu)
+    : BaseDynInst<Impl>(staticInst, PC, NPC, microPC,
+            Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu)
+{
+    initVars();
+}
+
+template <class Impl>
 MipsDynInst<Impl>::MipsDynInst(ExtMachInst inst,
-                               Addr PC, Addr NPC,
-                               Addr Pred_PC, Addr Pred_NPC,
+                               Addr PC, Addr NPC, Addr microPC,
+                               Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
                                InstSeqNum seq_num, O3CPU *cpu)
-    : BaseDynInst<Impl>(inst, PC, NPC, Pred_PC, Pred_NPC, seq_num, cpu)
+    : BaseDynInst<Impl>(inst, PC, NPC, microPC,
+            Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu)
 {
     initVars();
 }
diff --git a/src/cpu/o3/regfile.hh b/src/cpu/o3/regfile.hh
index b5b1cd021..75d3fa6eb 100644
--- a/src/cpu/o3/regfile.hh
+++ b/src/cpu/o3/regfile.hh
@@ -179,7 +179,9 @@ class PhysRegFile
         DPRINTF(IEW, "RegFile: Setting float register %i to %#x\n",
                 int(reg_idx), (uint64_t)val);
 
+#if THE_ISA == ALPHA_ISA
         if (reg_idx != TheISA::ZeroReg)
+#endif
             floatRegFile[reg_idx].d = val;
     }
 
@@ -194,7 +196,9 @@ class PhysRegFile
         DPRINTF(IEW, "RegFile: Setting float register %i to %#x\n",
                 int(reg_idx), (uint64_t)val);
 
+#if THE_ISA == ALPHA_ISA
         if (reg_idx != TheISA::ZeroReg)
+#endif
             floatRegFile[reg_idx].d = val;
     }
 
diff --git a/src/cpu/o3/rename_impl.hh b/src/cpu/o3/rename_impl.hh
index 431705e19..d78de2c87 100644
--- a/src/cpu/o3/rename_impl.hh
+++ b/src/cpu/o3/rename_impl.hh
@@ -356,47 +356,12 @@ DefaultRename<Impl>::squash(const InstSeqNum &squash_seq_num, unsigned tid)
     }
 
     // Clear the instruction list and skid buffer in case they have any
-    // insts in them. Since we support multiple ISAs, we cant just:
-    // "insts[tid].clear();" or "skidBuffer[tid].clear()" since there is
-    // a possible delay slot inst for different architectures
-    // insts[tid].clear();
-#if ISA_HAS_DELAY_SLOT
-    DPRINTF(Rename, "[tid:%i] Squashing incoming decode instructions until "
-            "[sn:%i].\n",tid, squash_seq_num);
-    ListIt ilist_it = insts[tid].begin();
-    while (ilist_it != insts[tid].end()) {
-        if ((*ilist_it)->seqNum > squash_seq_num) {
-            (*ilist_it)->setSquashed();
-            DPRINTF(Rename, "Squashing incoming decode instruction, "
-                    "[tid:%i] [sn:%i] PC %08p.\n", tid, (*ilist_it)->seqNum, (*ilist_it)->PC);
-        }
-        ilist_it++;
-    }
-#else
+    // insts in them.
     insts[tid].clear();
-#endif
 
     // Clear the skid buffer in case it has any data in it.
-    // See comments above.
-    //     skidBuffer[tid].clear();
-#if ISA_HAS_DELAY_SLOT
-    DPRINTF(Rename, "[tid:%i] Squashing incoming skidbuffer instructions "
-            "until [sn:%i].\n", tid, squash_seq_num);
-    ListIt slist_it = skidBuffer[tid].begin();
-    while (slist_it != skidBuffer[tid].end()) {
-        if ((*slist_it)->seqNum > squash_seq_num) {
-            (*slist_it)->setSquashed();
-            DPRINTF(Rename, "Squashing skidbuffer instruction, [tid:%i] [sn:%i]"
-                    "PC %08p.\n", tid, (*slist_it)->seqNum, (*slist_it)->PC);
-        }
-        slist_it++;
-    }
-    resumeUnblocking = (skidBuffer[tid].size() != 0);
-    DPRINTF(Rename, "Resume unblocking set to %s\n",
-            resumeUnblocking ? "true" : "false");
-#else
     skidBuffer[tid].clear();
-#endif
+
     doSquash(squash_seq_num, tid);
 }
 
@@ -776,11 +741,9 @@ DefaultRename<Impl>::sortInsts()
 {
     int insts_from_decode = fromDecode->size;
 #ifdef DEBUG
-#if !ISA_HAS_DELAY_SLOT
     for (int i=0; i < numThreads; i++)
         assert(insts[i].empty());
 #endif
-#endif
     for (int i = 0; i < insts_from_decode; ++i) {
         DynInstPtr inst = fromDecode->insts[i];
         insts[inst->threadNumber].push_back(inst);
@@ -1000,6 +963,7 @@ DefaultRename<Impl>::renameSrcRegs(DynInstPtr &inst,unsigned tid)
             // Floating point and Miscellaneous registers need their indexes
             // adjusted to account for the expanded number of flattened int regs.
             flat_src_reg = src_reg - TheISA::FP_Base_DepTag + TheISA::NumIntRegs;
+            DPRINTF(Rename, "Adjusting reg index from %d to %d.\n", src_reg, flat_src_reg);
         }
 
         inst->flattenSrcReg(src_idx, flat_src_reg);
@@ -1016,9 +980,11 @@ DefaultRename<Impl>::renameSrcRegs(DynInstPtr &inst,unsigned tid)
 
         // See if the register is ready or not.
         if (scoreboard->getReg(renamed_reg) == true) {
-            DPRINTF(Rename, "[tid:%u]: Register is ready.\n", tid);
+            DPRINTF(Rename, "[tid:%u]: Register %d is ready.\n", tid, renamed_reg);
 
             inst->markSrcRegReady(src_idx);
+        } else {
+            DPRINTF(Rename, "[tid:%u]: Register %d is not ready.\n", tid, renamed_reg);
         }
 
         ++renameRenameLookups;
@@ -1045,6 +1011,7 @@ DefaultRename<Impl>::renameDestRegs(DynInstPtr &inst,unsigned tid)
             // Floating point and Miscellaneous registers need their indexes
             // adjusted to account for the expanded number of flattened int regs.
             flat_dest_reg = dest_reg - TheISA::FP_Base_DepTag + TheISA::NumIntRegs;
+            DPRINTF(Rename, "Adjusting reg index from %d to %d.\n", dest_reg, flat_dest_reg);
         }
 
         inst->flattenDestReg(dest_idx, flat_dest_reg);
@@ -1248,13 +1215,7 @@ DefaultRename<Impl>::checkSignalsAndUpdate(unsigned tid)
         DPRINTF(Rename, "[tid:%u]: Squashing instructions due to squash from "
                 "commit.\n", tid);
 
-#if ISA_HAS_DELAY_SLOT
-        InstSeqNum squashed_seq_num = fromCommit->commitInfo[tid].bdelayDoneSeqNum;
-#else
-        InstSeqNum squashed_seq_num = fromCommit->commitInfo[tid].doneSeqNum;
-#endif
-
-        squash(squashed_seq_num, tid);
+        squash(fromCommit->commitInfo[tid].doneSeqNum, tid);
 
         return true;
     }
diff --git a/src/cpu/o3/rename_map.cc b/src/cpu/o3/rename_map.cc
index b436ec1c3..e6649ce3e 100644
--- a/src/cpu/o3/rename_map.cc
+++ b/src/cpu/o3/rename_map.cc
@@ -165,17 +165,21 @@ SimpleRenameMap::rename(RegIndex arch_reg)
 
         // If it's not referencing the zero register, then rename the
         // register.
+#if THE_ISA == ALPHA_ISA
         if (arch_reg != floatZeroReg) {
+#endif
             renamed_reg = freeList->getFloatReg();
 
             floatRenameMap[arch_reg].physical_reg = renamed_reg;
 
             assert(renamed_reg < numPhysicalRegs &&
                    renamed_reg >= numPhysicalIntRegs);
+#if THE_ISA == ALPHA_ISA
         } else {
             // Otherwise return the zero register so nothing bad happens.
             renamed_reg = floatZeroReg;
         }
+#endif
     } else {
         // Subtract off the base offset for miscellaneous registers.
         arch_reg = arch_reg - numLogicalRegs;
diff --git a/src/cpu/o3/scoreboard.cc b/src/cpu/o3/scoreboard.cc
index 1859b35a4..e7f8b7949 100644
--- a/src/cpu/o3/scoreboard.cc
+++ b/src/cpu/o3/scoreboard.cc
@@ -29,6 +29,7 @@
  *          Kevin Lim
  */
 
+#include "arch/isa_specific.hh"
 #include "cpu/o3/scoreboard.hh"
 
 Scoreboard::Scoreboard(unsigned activeThreads,
@@ -79,11 +80,18 @@ Scoreboard::name() const
 bool
 Scoreboard::getReg(PhysRegIndex phys_reg)
 {
+#if THE_ISA == ALPHA_ISA
     // Always ready if int or fp zero reg.
     if (phys_reg == zeroRegIdx ||
         phys_reg == (zeroRegIdx + numPhysicalIntRegs)) {
         return 1;
     }
+#else
+    // Always ready if int zero reg.
+    if (phys_reg == zeroRegIdx) {
+        return 1;
+    }
+#endif
 
     return regScoreBoard[phys_reg];
 }
@@ -99,11 +107,18 @@ Scoreboard::setReg(PhysRegIndex phys_reg)
 void
 Scoreboard::unsetReg(PhysRegIndex ready_reg)
 {
+#if THE_ISA == ALPHA_ISA
     if (ready_reg == zeroRegIdx ||
         ready_reg == (zeroRegIdx + numPhysicalIntRegs)) {
         // Don't do anything if int or fp zero reg.
         return;
     }
+#else
+    if (ready_reg == zeroRegIdx) {
+        // Don't do anything if int zero reg.
+        return;
+    }
+#endif
 
     regScoreBoard[ready_reg] = 0;
 }
diff --git a/src/cpu/o3/sparc/dyn_inst.hh b/src/cpu/o3/sparc/dyn_inst.hh
index 72242b161..a7ab6cd79 100644
--- a/src/cpu/o3/sparc/dyn_inst.hh
+++ b/src/cpu/o3/sparc/dyn_inst.hh
@@ -56,8 +56,14 @@ class SparcDynInst : public BaseDynInst<Impl>
 
   public:
     /** BaseDynInst constructor given a binary instruction. */
-    SparcDynInst(TheISA::ExtMachInst inst, Addr PC, Addr NPC,
-            Addr Pred_PC, Addr Pred_NPC, InstSeqNum seq_num, O3CPU *cpu);
+    SparcDynInst(StaticInstPtr staticInst, Addr PC, Addr NPC, Addr microPC,
+            Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
+            InstSeqNum seq_num, O3CPU *cpu);
+
+    /** BaseDynInst constructor given a binary instruction. */
+    SparcDynInst(TheISA::ExtMachInst inst, Addr PC, Addr NPC, Addr microPC,
+            Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
+            InstSeqNum seq_num, O3CPU *cpu);
 
     /** BaseDynInst constructor given a static inst pointer. */
     SparcDynInst(StaticInstPtr &_staticInst);
diff --git a/src/cpu/o3/sparc/dyn_inst_impl.hh b/src/cpu/o3/sparc/dyn_inst_impl.hh
index c4d30b6f4..6bfe97717 100644
--- a/src/cpu/o3/sparc/dyn_inst_impl.hh
+++ b/src/cpu/o3/sparc/dyn_inst_impl.hh
@@ -31,10 +31,23 @@
 #include "cpu/o3/sparc/dyn_inst.hh"
 
 template <class Impl>
+SparcDynInst<Impl>::SparcDynInst(StaticInstPtr staticInst,
+        Addr PC, Addr NPC, Addr microPC,
+        Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
+        InstSeqNum seq_num, O3CPU *cpu)
+    : BaseDynInst<Impl>(staticInst, PC, NPC, microPC,
+            Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu)
+{
+    initVars();
+}
+
+template <class Impl>
 SparcDynInst<Impl>::SparcDynInst(TheISA::ExtMachInst inst,
-        Addr PC, Addr NPC, Addr Pred_PC, Addr Pred_NPC,
+        Addr PC, Addr NPC, Addr microPC,
+        Addr Pred_PC, Addr Pred_NPC, Addr Pred_MicroPC,
         InstSeqNum seq_num, O3CPU *cpu)
-    : BaseDynInst<Impl>(inst, PC, NPC, Pred_PC, Pred_NPC, seq_num, cpu)
+    : BaseDynInst<Impl>(inst, PC, NPC, microPC,
+            Pred_PC, Pred_NPC, Pred_MicroPC, seq_num, cpu)
 {
     initVars();
 }