Fixes to get compiling to work. This is mainly fixing up some includes; changing functions within the XCs; changing MemReqPtrs to Requests or Packets where appropriate.

Currently the O3 and Ozone CPUs do not work in the new memory system; I still need to fix up the ports to work and handle responses properly.  This check-in is so that the merge between m5 and newmem is no longer outstanding.

src/SConscript:
    Need to include FU Pool for new CPU model.  I'll try to figure out a cleaner way to handle this in the future.
src/base/traceflags.py:
    Include new traces flags, fix up merge mess up.
src/cpu/SConscript:
    Include the base_dyn_inst.cc as one of othe sources.
    Don't compile the Ozone CPU for now.
src/cpu/base.cc:
    Remove an extra } from the merge.
src/cpu/base_dyn_inst.cc:
    Fixes to make compiling work.  Don't instantiate the OzoneCPU for now.
src/cpu/base_dyn_inst.hh:
src/cpu/o3/2bit_local_pred.cc:
src/cpu/o3/alpha_cpu_builder.cc:
src/cpu/o3/alpha_cpu_impl.hh:
src/cpu/o3/alpha_dyn_inst.hh:
src/cpu/o3/alpha_params.hh:
src/cpu/o3/bpred_unit.cc:
src/cpu/o3/btb.hh:
src/cpu/o3/commit.hh:
src/cpu/o3/commit_impl.hh:
src/cpu/o3/cpu.cc:
src/cpu/o3/cpu.hh:
src/cpu/o3/fetch.hh:
src/cpu/o3/fetch_impl.hh:
src/cpu/o3/free_list.hh:
src/cpu/o3/iew.hh:
src/cpu/o3/iew_impl.hh:
src/cpu/o3/inst_queue.hh:
src/cpu/o3/inst_queue_impl.hh:
src/cpu/o3/regfile.hh:
src/cpu/o3/sat_counter.hh:
src/cpu/op_class.hh:
src/cpu/ozone/cpu.hh:
src/cpu/checker/cpu.cc:
src/cpu/checker/cpu.hh:
src/cpu/checker/exec_context.hh:
src/cpu/checker/o3_cpu_builder.cc:
src/cpu/ozone/cpu_impl.hh:
src/mem/request.hh:
src/cpu/o3/fu_pool.hh:
src/cpu/o3/lsq.hh:
src/cpu/o3/lsq_unit.hh:
src/cpu/o3/lsq_unit_impl.hh:
src/cpu/o3/thread_state.hh:
src/cpu/ozone/back_end.hh:
src/cpu/ozone/dyn_inst.cc:
src/cpu/ozone/dyn_inst.hh:
src/cpu/ozone/front_end.hh:
src/cpu/ozone/inorder_back_end.hh:
src/cpu/ozone/lw_back_end.hh:
src/cpu/ozone/lw_lsq.hh:
src/cpu/ozone/ozone_impl.hh:
src/cpu/ozone/thread_state.hh:
    Fixes to get compiling to work.
src/cpu/o3/alpha_cpu.hh:
    Fixes to get compiling to work.
    Float reg accessors have changed, as well as MemReqPtrs to RequestPtrs.
src/cpu/o3/alpha_dyn_inst_impl.hh:
    Fixes to get compiling to work.
    Pass in the packet to the completeAcc function.
    Fix up syscall function.

--HG--
rename : cpu/activity.cc => src/cpu/activity.cc
rename : cpu/activity.hh => src/cpu/activity.hh
rename : cpu/checker/cpu.cc => src/cpu/checker/cpu.cc
rename : cpu/checker/cpu.hh => src/cpu/checker/cpu.hh
rename : cpu/checker/cpu_builder.cc => src/cpu/checker/cpu_builder.cc
rename : cpu/checker/exec_context.hh => src/cpu/checker/exec_context.hh
rename : cpu/checker/o3_cpu_builder.cc => src/cpu/checker/o3_cpu_builder.cc
rename : cpu/o3/dep_graph.hh => src/cpu/o3/dep_graph.hh
rename : cpu/o3/fu_pool.cc => src/cpu/o3/fu_pool.cc
rename : cpu/o3/fu_pool.hh => src/cpu/o3/fu_pool.hh
rename : cpu/o3/lsq.cc => src/cpu/o3/lsq.cc
rename : cpu/o3/lsq.hh => src/cpu/o3/lsq.hh
rename : cpu/o3/lsq_impl.hh => src/cpu/o3/lsq_impl.hh
rename : cpu/o3/lsq_unit.cc => src/cpu/o3/lsq_unit.cc
rename : cpu/o3/lsq_unit.hh => src/cpu/o3/lsq_unit.hh
rename : cpu/o3/lsq_unit_impl.hh => src/cpu/o3/lsq_unit_impl.hh
rename : cpu/o3/scoreboard.cc => src/cpu/o3/scoreboard.cc
rename : cpu/o3/scoreboard.hh => src/cpu/o3/scoreboard.hh
rename : cpu/o3/thread_state.hh => src/cpu/o3/thread_state.hh
rename : cpu/ozone/back_end.cc => src/cpu/ozone/back_end.cc
rename : cpu/ozone/back_end.hh => src/cpu/ozone/back_end.hh
rename : cpu/ozone/back_end_impl.hh => src/cpu/ozone/back_end_impl.hh
rename : cpu/ozone/cpu_builder.cc => src/cpu/ozone/cpu_builder.cc
rename : cpu/ozone/dyn_inst.cc => src/cpu/ozone/dyn_inst.cc
rename : cpu/ozone/dyn_inst.hh => src/cpu/ozone/dyn_inst.hh
rename : cpu/ozone/dyn_inst_impl.hh => src/cpu/ozone/dyn_inst_impl.hh
rename : cpu/ozone/front_end.cc => src/cpu/ozone/front_end.cc
rename : cpu/ozone/front_end.hh => src/cpu/ozone/front_end.hh
rename : cpu/ozone/front_end_impl.hh => src/cpu/ozone/front_end_impl.hh
rename : cpu/ozone/inorder_back_end.cc => src/cpu/ozone/inorder_back_end.cc
rename : cpu/ozone/inorder_back_end.hh => src/cpu/ozone/inorder_back_end.hh
rename : cpu/ozone/inorder_back_end_impl.hh => src/cpu/ozone/inorder_back_end_impl.hh
rename : cpu/ozone/inst_queue.cc => src/cpu/ozone/inst_queue.cc
rename : cpu/ozone/inst_queue.hh => src/cpu/ozone/inst_queue.hh
rename : cpu/ozone/inst_queue_impl.hh => src/cpu/ozone/inst_queue_impl.hh
rename : cpu/ozone/lsq_unit.cc => src/cpu/ozone/lsq_unit.cc
rename : cpu/ozone/lsq_unit.hh => src/cpu/ozone/lsq_unit.hh
rename : cpu/ozone/lsq_unit_impl.hh => src/cpu/ozone/lsq_unit_impl.hh
rename : cpu/ozone/lw_back_end.cc => src/cpu/ozone/lw_back_end.cc
rename : cpu/ozone/lw_back_end.hh => src/cpu/ozone/lw_back_end.hh
rename : cpu/ozone/lw_back_end_impl.hh => src/cpu/ozone/lw_back_end_impl.hh
rename : cpu/ozone/lw_lsq.cc => src/cpu/ozone/lw_lsq.cc
rename : cpu/ozone/lw_lsq.hh => src/cpu/ozone/lw_lsq.hh
rename : cpu/ozone/lw_lsq_impl.hh => src/cpu/ozone/lw_lsq_impl.hh
rename : cpu/ozone/null_predictor.hh => src/cpu/ozone/null_predictor.hh
rename : cpu/ozone/ozone_impl.hh => src/cpu/ozone/ozone_impl.hh
rename : cpu/ozone/rename_table.cc => src/cpu/ozone/rename_table.cc
rename : cpu/ozone/rename_table.hh => src/cpu/ozone/rename_table.hh
rename : cpu/ozone/rename_table_impl.hh => src/cpu/ozone/rename_table_impl.hh
rename : cpu/ozone/simple_impl.hh => src/cpu/ozone/simple_impl.hh
rename : cpu/ozone/simple_params.hh => src/cpu/ozone/simple_params.hh
rename : cpu/ozone/thread_state.hh => src/cpu/ozone/thread_state.hh
rename : cpu/quiesce_event.cc => src/cpu/quiesce_event.cc
rename : cpu/quiesce_event.hh => src/cpu/quiesce_event.hh
rename : cpu/thread_state.hh => src/cpu/thread_state.hh
rename : python/m5/objects/FUPool.py => src/python/m5/objects/FUPool.py
rename : python/m5/objects/OzoneCPU.py => src/python/m5/objects/OzoneCPU.py
rename : python/m5/objects/SimpleOzoneCPU.py => src/python/m5/objects/SimpleOzoneCPU.py
extra : convert_revision : ca7f0fbf65ee1a70d482fb4eda9a1840c7f9b8f8

1 files changed, 656 insertions, 0 deletions

diff --git a/src/cpu/ozone/lw_lsq.hh b/src/cpu/ozone/lw_lsq.hh
new file mode 100644
index 000000000..e1488dd6f
--- /dev/null
+++ b/src/cpu/ozone/lw_lsq.hh
@@ -0,0 +1,656 @@
+/*
+ * Copyright (c) 2004-2006 The Regents of The University of Michigan
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met: redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer;
+ * redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution;
+ * neither the name of the copyright holders nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __CPU_OZONE_LW_LSQ_HH__
+#define __CPU_OZONE_LW_LSQ_HH__
+
+#include <list>
+#include <map>
+#include <queue>
+#include <algorithm>
+
+#include "arch/faults.hh"
+#include "arch/isa_traits.hh"
+#include "config/full_system.hh"
+#include "base/hashmap.hh"
+#include "cpu/inst_seq.hh"
+#include "mem/packet.hh"
+#include "mem/port.hh"
+//#include "mem/page_table.hh"
+#include "sim/debug.hh"
+#include "sim/sim_object.hh"
+
+//class PageTable;
+
+/**
+ * Class that implements the actual LQ and SQ for each specific thread.
+ * Both are circular queues; load entries are freed upon committing, while
+ * store entries are freed once they writeback. The LSQUnit tracks if there
+ * are memory ordering violations, and also detects partial load to store
+ * forwarding cases (a store only has part of a load's data) that requires
+ * the load to wait until the store writes back. In the former case it
+ * holds onto the instruction until the dependence unit looks at it, and
+ * in the latter it stalls the LSQ until the store writes back. At that
+ * point the load is replayed.
+ */
+template <class Impl>
+class OzoneLWLSQ {
+  public:
+    typedef typename Impl::Params Params;
+    typedef typename Impl::FullCPU FullCPU;
+    typedef typename Impl::BackEnd BackEnd;
+    typedef typename Impl::DynInstPtr DynInstPtr;
+    typedef typename Impl::IssueStruct IssueStruct;
+
+    typedef TheISA::IntReg IntReg;
+
+    typedef typename std::map<InstSeqNum, DynInstPtr>::iterator LdMapIt;
+
+  private:
+    class StoreCompletionEvent : public Event {
+      public:
+        /** Constructs a store completion event. */
+        StoreCompletionEvent(DynInstPtr &inst, BackEnd *be,
+                             Event *wb_event, OzoneLWLSQ *lsq_ptr);
+
+        /** Processes the store completion event. */
+        void process();
+
+        /** Returns the description of this event. */
+        const char *description();
+
+      private:
+        /** The store index of the store being written back. */
+        DynInstPtr inst;
+
+        BackEnd *be;
+        /** The writeback event for the store.  Needed for store
+         * conditionals.
+         */
+      public:
+        Event *wbEvent;
+        bool miss;
+      private:
+        /** The pointer to the LSQ unit that issued the store. */
+        OzoneLWLSQ<Impl> *lsqPtr;
+    };
+
+  public:
+    /** Constructs an LSQ unit. init() must be called prior to use. */
+    OzoneLWLSQ();
+
+    /** Initializes the LSQ unit with the specified number of entries. */
+    void init(Params *params, unsigned maxLQEntries,
+              unsigned maxSQEntries, unsigned id);
+
+    /** Returns the name of the LSQ unit. */
+    std::string name() const;
+
+    /** Sets the CPU pointer. */
+    void setCPU(FullCPU *cpu_ptr)
+    { cpu = cpu_ptr; }
+
+    /** Sets the back-end stage pointer. */
+    void setBE(BackEnd *be_ptr)
+    { be = be_ptr; }
+
+    /** Sets the page table pointer. */
+//    void setPageTable(PageTable *pt_ptr);
+
+    /** Ticks the LSQ unit, which in this case only resets the number of
+     * used cache ports.
+     * @todo: Move the number of used ports up to the LSQ level so it can
+     * be shared by all LSQ units.
+     */
+    void tick() { usedPorts = 0; }
+
+    /** Inserts an instruction. */
+    void insert(DynInstPtr &inst);
+    /** Inserts a load instruction. */
+    void insertLoad(DynInstPtr &load_inst);
+    /** Inserts a store instruction. */
+    void insertStore(DynInstPtr &store_inst);
+
+    /** Executes a load instruction. */
+    Fault executeLoad(DynInstPtr &inst);
+
+    /** Executes a store instruction. */
+    Fault executeStore(DynInstPtr &inst);
+
+    /** Commits the head load. */
+    void commitLoad();
+    /** Commits loads older than a specific sequence number. */
+    void commitLoads(InstSeqNum &youngest_inst);
+
+    /** Commits stores older than a specific sequence number. */
+    void commitStores(InstSeqNum &youngest_inst);
+
+    /** Writes back stores. */
+    void writebackStores();
+
+    // @todo: Include stats in the LSQ unit.
+    //void regStats();
+
+    /** Clears all the entries in the LQ. */
+    void clearLQ();
+
+    /** Clears all the entries in the SQ. */
+    void clearSQ();
+
+    /** Resizes the LQ to a given size. */
+    void resizeLQ(unsigned size);
+
+    /** Resizes the SQ to a given size. */
+    void resizeSQ(unsigned size);
+
+    /** Squashes all instructions younger than a specific sequence number. */
+    void squash(const InstSeqNum &squashed_num);
+
+    /** Returns if there is a memory ordering violation. Value is reset upon
+     * call to getMemDepViolator().
+     */
+    bool violation() { return memDepViolator; }
+
+    /** Returns the memory ordering violator. */
+    DynInstPtr getMemDepViolator();
+
+    /** Returns if a load became blocked due to the memory system.  It clears
+     *  the bool's value upon this being called.
+     */
+    bool loadBlocked()
+    { return isLoadBlocked; }
+
+    void clearLoadBlocked()
+    { isLoadBlocked = false; }
+
+    bool isLoadBlockedHandled()
+    { return loadBlockedHandled; }
+
+    void setLoadBlockedHandled()
+    { loadBlockedHandled = true; }
+
+    /** Returns the number of free entries (min of free LQ and SQ entries). */
+    unsigned numFreeEntries();
+
+    /** Returns the number of loads ready to execute. */
+    int numLoadsReady();
+
+    /** Returns the number of loads in the LQ. */
+    int numLoads() { return loads; }
+
+    /** Returns the number of stores in the SQ. */
+    int numStores() { return stores; }
+
+    /** Returns if either the LQ or SQ is full. */
+    bool isFull() { return lqFull() || sqFull(); }
+
+    /** Returns if the LQ is full. */
+    bool lqFull() { return loads >= (LQEntries - 1); }
+
+    /** Returns if the SQ is full. */
+    bool sqFull() { return stores >= (SQEntries - 1); }
+
+    /** Debugging function to dump instructions in the LSQ. */
+    void dumpInsts();
+
+    /** Returns the number of instructions in the LSQ. */
+    unsigned getCount() { return loads + stores; }
+
+    /** Returns if there are any stores to writeback. */
+    bool hasStoresToWB() { return storesToWB; }
+
+    /** Returns the number of stores to writeback. */
+    int numStoresToWB() { return storesToWB; }
+
+    /** Returns if the LSQ unit will writeback on this cycle. */
+    bool willWB() { return storeQueue.back().canWB &&
+                        !storeQueue.back().completed/* &&
+                                                       !dcacheInterface->isBlocked()*/; }
+
+    void switchOut();
+
+    void takeOverFrom(ExecContext *old_xc = NULL);
+
+    bool isSwitchedOut() { return switchedOut; }
+
+    bool switchedOut;
+
+  private:
+    /** Completes the store at the specified index. */
+    void completeStore(int store_idx);
+
+  private:
+    /** Pointer to the CPU. */
+    FullCPU *cpu;
+
+    /** Pointer to the back-end stage. */
+    BackEnd *be;
+
+    MemObject *mem;
+
+    class DcachePort : public Port
+    {
+      protected:
+        FullCPU *cpu;
+
+      public:
+        DcachePort(const std::string &_name, FullCPU *_cpu)
+            : Port(_name), cpu(_cpu)
+        { }
+
+      protected:
+        virtual Tick recvAtomic(PacketPtr pkt);
+
+        virtual void recvFunctional(PacketPtr pkt);
+
+        virtual void recvStatusChange(Status status);
+
+        virtual void getDeviceAddressRanges(AddrRangeList &resp,
+                                            AddrRangeList &snoop)
+        { resp.clear(); snoop.clear(); }
+
+        virtual bool recvTiming(PacketPtr pkt);
+
+        virtual void recvRetry();
+    };
+
+    /** Pointer to the D-cache. */
+    DcachePort dcachePort;
+
+    /** Pointer to the page table. */
+//    PageTable *pTable;
+
+  public:
+    struct SQEntry {
+        /** Constructs an empty store queue entry. */
+        SQEntry()
+            : inst(NULL), req(NULL), size(0), data(0),
+              canWB(0), committed(0), completed(0), lqIt(NULL)
+        { }
+
+        /** Constructs a store queue entry for a given instruction. */
+        SQEntry(DynInstPtr &_inst)
+            : inst(_inst), req(NULL), size(0), data(0),
+              canWB(0), committed(0), completed(0), lqIt(NULL)
+        { }
+
+        /** The store instruction. */
+        DynInstPtr inst;
+        /** The memory request for the store. */
+        RequestPtr req;
+        /** The size of the store. */
+        int size;
+        /** The store data. */
+        IntReg data;
+        /** Whether or not the store can writeback. */
+        bool canWB;
+        /** Whether or not the store is committed. */
+        bool committed;
+        /** Whether or not the store is completed. */
+        bool completed;
+
+        typename std::list<DynInstPtr>::iterator lqIt;
+    };
+
+    enum Status {
+        Running,
+        Idle,
+        DcacheMissStall,
+        DcacheMissSwitch
+    };
+
+  private:
+    /** The OzoneLWLSQ thread id. */
+    unsigned lsqID;
+
+    /** The status of the LSQ unit. */
+    Status _status;
+
+    /** The store queue. */
+    std::list<SQEntry> storeQueue;
+    /** The load queue. */
+    std::list<DynInstPtr> loadQueue;
+
+    typedef typename std::list<SQEntry>::iterator SQIt;
+    typedef typename std::list<DynInstPtr>::iterator LQIt;
+
+
+    struct HashFn {
+    size_t operator() (const int a) const
+    {
+        unsigned hash = (((a >> 14) ^ ((a >> 2) & 0xffff))) & 0x7FFFFFFF;
+
+        return hash;
+    }
+    };
+
+    m5::hash_map<int, SQIt, HashFn> SQItHash;
+    std::queue<int> SQIndices;
+    m5::hash_map<int, LQIt, HashFn> LQItHash;
+    std::queue<int> LQIndices;
+
+    typedef typename m5::hash_map<int, LQIt, HashFn>::iterator LQHashIt;
+    typedef typename m5::hash_map<int, SQIt, HashFn>::iterator SQHashIt;
+    // Consider making these 16 bits
+    /** The number of LQ entries. */
+    unsigned LQEntries;
+    /** The number of SQ entries. */
+    unsigned SQEntries;
+
+    /** The number of load instructions in the LQ. */
+    int loads;
+    /** The number of store instructions in the SQ (excludes those waiting to
+     * writeback).
+     */
+    int stores;
+
+    int storesToWB;
+
+    /// @todo Consider moving to a more advanced model with write vs read ports
+    /** The number of cache ports available each cycle. */
+    int cachePorts;
+
+    /** The number of used cache ports in this cycle. */
+    int usedPorts;
+
+    //list<InstSeqNum> mshrSeqNums;
+
+     //Stats::Scalar<> dcacheStallCycles;
+    Counter lastDcacheStall;
+
+    // Make these per thread?
+    /** Whether or not the LSQ is stalled. */
+    bool stalled;
+    /** The store that causes the stall due to partial store to load
+     * forwarding.
+     */
+    InstSeqNum stallingStoreIsn;
+    /** The index of the above store. */
+    LQIt stallingLoad;
+
+    /** Whether or not a load is blocked due to the memory system.  It is
+     *  cleared when this value is checked via loadBlocked().
+     */
+    bool isLoadBlocked;
+
+    bool loadBlockedHandled;
+
+    InstSeqNum blockedLoadSeqNum;
+
+    /** The oldest faulting load instruction. */
+    DynInstPtr loadFaultInst;
+    /** The oldest faulting store instruction. */
+    DynInstPtr storeFaultInst;
+
+    /** The oldest load that caused a memory ordering violation. */
+    DynInstPtr memDepViolator;
+
+    // Will also need how many read/write ports the Dcache has.  Or keep track
+    // of that in stage that is one level up, and only call executeLoad/Store
+    // the appropriate number of times.
+
+  public:
+    /** Executes the load at the given index. */
+    template <class T>
+    Fault read(RequestPtr req, T &data, int load_idx);
+
+    /** Executes the store at the given index. */
+    template <class T>
+    Fault write(RequestPtr req, T &data, int store_idx);
+
+    /** Returns the sequence number of the head load instruction. */
+    InstSeqNum getLoadHeadSeqNum()
+    {
+        if (!loadQueue.empty()) {
+            return loadQueue.back()->seqNum;
+        } else {
+            return 0;
+        }
+
+    }
+
+    /** Returns the sequence number of the head store instruction. */
+    InstSeqNum getStoreHeadSeqNum()
+    {
+        if (!storeQueue.empty()) {
+            return storeQueue.back().inst->seqNum;
+        } else {
+            return 0;
+        }
+
+    }
+
+    /** Returns whether or not the LSQ unit is stalled. */
+    bool isStalled()  { return stalled; }
+};
+
+template <class Impl>
+template <class T>
+Fault
+OzoneLWLSQ<Impl>::read(RequestPtr req, T &data, int load_idx)
+{
+    //Depending on issue2execute delay a squashed load could
+    //execute if it is found to be squashed in the same
+    //cycle it is scheduled to execute
+    typename m5::hash_map<int, LQIt, HashFn>::iterator
+        lq_hash_it = LQItHash.find(load_idx);
+    assert(lq_hash_it != LQItHash.end());
+    DynInstPtr inst = (*(*lq_hash_it).second);
+
+    // Make sure this isn't an uncacheable access
+    // A bit of a hackish way to get uncached accesses to work only if they're
+    // at the head of the LSQ and are ready to commit (at the head of the ROB
+    // too).
+    // @todo: Fix uncached accesses.
+    if (req->getFlags() & UNCACHEABLE &&
+        (inst != loadQueue.back() || !inst->reachedCommit)) {
+        DPRINTF(OzoneLSQ, "[sn:%lli] Uncached load and not head of "
+                "commit/LSQ!\n",
+                inst->seqNum);
+        be->rescheduleMemInst(inst);
+        return TheISA::genMachineCheckFault();
+    }
+
+    // Check the SQ for any previous stores that might lead to forwarding
+    SQIt sq_it = storeQueue.begin();
+    int store_size = 0;
+
+    DPRINTF(OzoneLSQ, "Read called, load idx: %i addr: %#x\n",
+            load_idx, req->getPaddr());
+
+    while (sq_it != storeQueue.end() && (*sq_it).inst->seqNum > inst->seqNum)
+        ++sq_it;
+
+    while (1) {
+        // End once we've reached the top of the LSQ
+        if (sq_it == storeQueue.end()) {
+            break;
+        }
+
+        assert((*sq_it).inst);
+
+        store_size = (*sq_it).size;
+
+        if (store_size == 0) {
+            sq_it++;
+            continue;
+        }
+
+        // Check if the store data is within the lower and upper bounds of
+        // addresses that the request needs.
+        bool store_has_lower_limit =
+            req->getVaddr() >= (*sq_it).inst->effAddr;
+        bool store_has_upper_limit =
+            (req->getVaddr() + req->getSize()) <= ((*sq_it).inst->effAddr +
+                                                   store_size);
+        bool lower_load_has_store_part =
+            req->getVaddr() < ((*sq_it).inst->effAddr +
+                               store_size);
+        bool upper_load_has_store_part =
+            (req->getVaddr() + req->getSize()) > (*sq_it).inst->effAddr;
+
+        // If the store's data has all of the data needed, we can forward.
+        if (store_has_lower_limit && store_has_upper_limit) {
+            int shift_amt = req->getVaddr() & (store_size - 1);
+            // Assumes byte addressing
+            shift_amt = shift_amt << 3;
+
+            // Cast this to type T?
+            data = (*sq_it).data >> shift_amt;
+
+            assert(!inst->memData);
+            inst->memData = new uint8_t[64];
+
+            memcpy(inst->memData, &data, req->getSize());
+
+            DPRINTF(OzoneLSQ, "Forwarding from store [sn:%lli] to load to "
+                    "[sn:%lli] addr %#x, data %#x\n",
+                    (*sq_it).inst->seqNum, inst->seqNum, req->vaddr, *(inst->memData));
+/*
+            typename BackEnd::LdWritebackEvent *wb =
+                new typename BackEnd::LdWritebackEvent(inst,
+                                                       be);
+
+            // We'll say this has a 1 cycle load-store forwarding latency
+            // for now.
+            // FIXME - Need to make this a parameter.
+            wb->schedule(curTick);
+*/
+            // Should keep track of stat for forwarded data
+            return NoFault;
+        } else if ((store_has_lower_limit && lower_load_has_store_part) ||
+                   (store_has_upper_limit && upper_load_has_store_part) ||
+                   (lower_load_has_store_part && upper_load_has_store_part)) {
+            // This is the partial store-load forwarding case where a store
+            // has only part of the load's data.
+
+            // If it's already been written back, then don't worry about
+            // stalling on it.
+            if ((*sq_it).completed) {
+                sq_it++;
+                break;
+            }
+
+            // Must stall load and force it to retry, so long as it's the oldest
+            // load that needs to do so.
+            if (!stalled ||
+                (stalled &&
+                 inst->seqNum <
+                 (*stallingLoad)->seqNum)) {
+                stalled = true;
+                stallingStoreIsn = (*sq_it).inst->seqNum;
+                stallingLoad = (*lq_hash_it).second;
+            }
+
+            // Tell IQ/mem dep unit that this instruction will need to be
+            // rescheduled eventually
+            be->rescheduleMemInst(inst);
+
+            DPRINTF(OzoneLSQ, "Load-store forwarding mis-match. "
+                    "Store [sn:%lli] to load addr %#x\n",
+                    (*sq_it).inst->seqNum, req->vaddr);
+
+            return NoFault;
+        }
+        sq_it++;
+    }
+
+    // If there's no forwarding case, then go access memory
+    DPRINTF(OzoneLSQ, "Doing functional access for inst PC %#x\n",
+            inst->readPC());
+
+    assert(!inst->memData);
+    inst->memData = new uint8_t[64];
+
+    ++usedPorts;
+
+    DPRINTF(OzoneLSQ, "Doing timing access for inst PC %#x\n",
+            inst->readPC());
+
+    PacketPtr data_pkt = new Packet(req, Packet::ReadReq, Packet::Broadcast);
+    data_pkt->dataStatic(inst->memData);
+
+    // if we have a cache, do cache access too
+    if (!dcachePort.sendTiming(data_pkt)) {
+        // There's an older load that's already going to squash.
+        if (isLoadBlocked && blockedLoadSeqNum < inst->seqNum)
+            return NoFault;
+
+        // Record that the load was blocked due to memory.  This
+        // load will squash all instructions after it, be
+        // refetched, and re-executed.
+        isLoadBlocked = true;
+        loadBlockedHandled = false;
+        blockedLoadSeqNum = inst->seqNum;
+        // No fault occurred, even though the interface is blocked.
+        return NoFault;
+    }
+
+    if (data_pkt->result != Packet::Success) {
+        DPRINTF(OzoneLSQ, "OzoneLSQ: D-cache miss!\n");
+        DPRINTF(Activity, "Activity: ld accessing mem miss [sn:%lli]\n",
+                inst->seqNum);
+    } else {
+        DPRINTF(OzoneLSQ, "OzoneLSQ: D-cache hit!\n");
+        DPRINTF(Activity, "Activity: ld accessing mem hit [sn:%lli]\n",
+                inst->seqNum);
+    }
+
+    return NoFault;
+}
+
+template <class Impl>
+template <class T>
+Fault
+OzoneLWLSQ<Impl>::write(RequestPtr req, T &data, int store_idx)
+{
+    SQHashIt sq_hash_it = SQItHash.find(store_idx);
+    assert(sq_hash_it != SQItHash.end());
+
+    SQIt sq_it = (*sq_hash_it).second;
+    assert((*sq_it).inst);
+
+    DPRINTF(OzoneLSQ, "Doing write to store idx %i, addr %#x data %#x"
+            " | [sn:%lli]\n",
+            store_idx, req->getPaddr(), data, (*sq_it).inst->seqNum);
+
+    (*sq_it).req = req;
+    (*sq_it).size = sizeof(T);
+    (*sq_it).data = data;
+/*
+    assert(!req->data);
+    req->data = new uint8_t[64];
+    memcpy(req->data, (uint8_t *)&(*sq_it).data, req->size);
+*/
+
+    // This function only writes the data to the store queue, so no fault
+    // can happen here.
+    return NoFault;
+}
+
+#endif // __CPU_OZONE_LW_LSQ_HH__