O3 CPU: Provide the squashing instruction

This patch adds a function to the ROB that will get the squashing instruction
from the ROB's list of instructions. This squashing instruction is used for
figuring out the macroop from which the fetch stage should fetch the microops.
Further, a check has been added that if the instructions are to be fetched
from the cache maintained by the fetch stage, then the data in the cache should
be valid and the PC of the thread being fetched from is same as the address of
the cache block.

4 files changed, 25 insertions, 1 deletions

diff --git a/src/cpu/o3/commit_impl.hh b/src/cpu/o3/commit_impl.hh
index edf7e861b..a9b795933 100644
--- a/src/cpu/o3/commit_impl.hh
+++ b/src/cpu/o3/commit_impl.hh
@@ -856,7 +856,8 @@ DefaultCommit<Impl>::commit()
                 fromIEW->mispredictInst[tid];
             toIEW->commitInfo[tid].branchTaken =
                 fromIEW->branchTaken[tid];
-            toIEW->commitInfo[tid].squashInst = NULL;
+            toIEW->commitInfo[tid].squashInst =
+                                    rob->findInst(tid, squashed_inst);
 
             toIEW->commitInfo[tid].pc = fromIEW->pc[tid];
 
diff --git a/src/cpu/o3/fetch_impl.hh b/src/cpu/o3/fetch_impl.hh
index c8f7b0181..48c21917a 100644
--- a/src/cpu/o3/fetch_impl.hh
+++ b/src/cpu/o3/fetch_impl.hh
@@ -1189,8 +1189,15 @@ DefaultFetch<Impl>::fetch(bool &status_change)
         // StaticInst from the rom, the current macroop, or what's already
         // in the predecoder.
         bool needMem = !inRom && !curMacroop && !predecoder.extMachInstReady();
+        fetchAddr = (thisPC.instAddr() + pcOffset) & BaseCPU::PCMask;
+        Addr block_PC = icacheBlockAlignPC(fetchAddr);
 
         if (needMem) {
+            // If buffer is no longer valid or fetchAddr has moved to point
+            // to the next cache block then start fetch from icache.
+            if (!cacheDataValid[tid] || block_PC != cacheDataPC[tid])
+                break;
+
             if (blkOffset >= numInsts) {
                 // We need to process more memory, but we've run out of the
                 // current block.
diff --git a/src/cpu/o3/rob.hh b/src/cpu/o3/rob.hh
index fadd99849..d0b156954 100644
--- a/src/cpu/o3/rob.hh
+++ b/src/cpu/o3/rob.hh
@@ -122,6 +122,11 @@ class ROB
      */
     DynInstPtr readHeadInst(ThreadID tid);
 
+    /** Returns a pointer to the instruction with the given sequence if it is
+     *  in the ROB.
+     */
+    DynInstPtr findInst(ThreadID tid, InstSeqNum squash_inst);
+
     /** Returns pointer to the tail instruction within the ROB.  There is
      *  no guarantee as to the return value if the ROB is empty.
      *  @retval Pointer to the DynInst that is at the tail of the ROB.
diff --git a/src/cpu/o3/rob_impl.hh b/src/cpu/o3/rob_impl.hh
index 0484f519c..ee4a9e576 100644
--- a/src/cpu/o3/rob_impl.hh
+++ b/src/cpu/o3/rob_impl.hh
@@ -544,3 +544,14 @@ ROB<Impl>::regStats()
         .desc("The number of ROB writes");
 }
 
+template <class Impl>
+typename Impl::DynInstPtr
+ROB<Impl>::findInst(ThreadID tid, InstSeqNum squash_inst)
+{
+    for (InstIt it = instList[tid].begin(); it != instList[tid].end(); it++) {
+        if ((*it)->seqNum == squash_inst) {
+            return *it;
+        }
+    }
+    return NULL;
+}