shuffle files around for new directory structure

--HG--
rename : cpu/base_cpu.cc => cpu/base.cc
rename : cpu/base_cpu.hh => cpu/base.hh
rename : cpu/beta_cpu/2bit_local_pred.cc => cpu/o3/2bit_local_pred.cc
rename : cpu/beta_cpu/2bit_local_pred.hh => cpu/o3/2bit_local_pred.hh
rename : cpu/beta_cpu/alpha_full_cpu.cc => cpu/o3/alpha_cpu.cc
rename : cpu/beta_cpu/alpha_full_cpu.hh => cpu/o3/alpha_cpu.hh
rename : cpu/beta_cpu/alpha_full_cpu_builder.cc => cpu/o3/alpha_cpu_builder.cc
rename : cpu/beta_cpu/alpha_full_cpu_impl.hh => cpu/o3/alpha_cpu_impl.hh
rename : cpu/beta_cpu/alpha_dyn_inst.cc => cpu/o3/alpha_dyn_inst.cc
rename : cpu/beta_cpu/alpha_dyn_inst.hh => cpu/o3/alpha_dyn_inst.hh
rename : cpu/beta_cpu/alpha_dyn_inst_impl.hh => cpu/o3/alpha_dyn_inst_impl.hh
rename : cpu/beta_cpu/alpha_impl.hh => cpu/o3/alpha_impl.hh
rename : cpu/beta_cpu/alpha_params.hh => cpu/o3/alpha_params.hh
rename : cpu/beta_cpu/bpred_unit.cc => cpu/o3/bpred_unit.cc
rename : cpu/beta_cpu/bpred_unit.hh => cpu/o3/bpred_unit.hh
rename : cpu/beta_cpu/bpred_unit_impl.hh => cpu/o3/bpred_unit_impl.hh
rename : cpu/beta_cpu/btb.cc => cpu/o3/btb.cc
rename : cpu/beta_cpu/btb.hh => cpu/o3/btb.hh
rename : cpu/beta_cpu/comm.hh => cpu/o3/comm.hh
rename : cpu/beta_cpu/commit.cc => cpu/o3/commit.cc
rename : cpu/beta_cpu/commit.hh => cpu/o3/commit.hh
rename : cpu/beta_cpu/commit_impl.hh => cpu/o3/commit_impl.hh
rename : cpu/beta_cpu/full_cpu.cc => cpu/o3/cpu.cc
rename : cpu/beta_cpu/full_cpu.hh => cpu/o3/cpu.hh
rename : cpu/beta_cpu/cpu_policy.hh => cpu/o3/cpu_policy.hh
rename : cpu/beta_cpu/decode.cc => cpu/o3/decode.cc
rename : cpu/beta_cpu/decode.hh => cpu/o3/decode.hh
rename : cpu/beta_cpu/decode_impl.hh => cpu/o3/decode_impl.hh
rename : cpu/beta_cpu/fetch.cc => cpu/o3/fetch.cc
rename : cpu/beta_cpu/fetch.hh => cpu/o3/fetch.hh
rename : cpu/beta_cpu/fetch_impl.hh => cpu/o3/fetch_impl.hh
rename : cpu/beta_cpu/free_list.cc => cpu/o3/free_list.cc
rename : cpu/beta_cpu/free_list.hh => cpu/o3/free_list.hh
rename : cpu/beta_cpu/iew.cc => cpu/o3/iew.cc
rename : cpu/beta_cpu/iew.hh => cpu/o3/iew.hh
rename : cpu/beta_cpu/iew_impl.hh => cpu/o3/iew_impl.hh
rename : cpu/beta_cpu/inst_queue.cc => cpu/o3/inst_queue.cc
rename : cpu/beta_cpu/inst_queue.hh => cpu/o3/inst_queue.hh
rename : cpu/beta_cpu/inst_queue_impl.hh => cpu/o3/inst_queue_impl.hh
rename : cpu/beta_cpu/mem_dep_unit.cc => cpu/o3/mem_dep_unit.cc
rename : cpu/beta_cpu/mem_dep_unit.hh => cpu/o3/mem_dep_unit.hh
rename : cpu/beta_cpu/mem_dep_unit_impl.hh => cpu/o3/mem_dep_unit_impl.hh
rename : cpu/beta_cpu/ras.cc => cpu/o3/ras.cc
rename : cpu/beta_cpu/ras.hh => cpu/o3/ras.hh
rename : cpu/beta_cpu/regfile.hh => cpu/o3/regfile.hh
rename : cpu/beta_cpu/rename.cc => cpu/o3/rename.cc
rename : cpu/beta_cpu/rename.hh => cpu/o3/rename.hh
rename : cpu/beta_cpu/rename_impl.hh => cpu/o3/rename_impl.hh
rename : cpu/beta_cpu/rename_map.cc => cpu/o3/rename_map.cc
rename : cpu/beta_cpu/rename_map.hh => cpu/o3/rename_map.hh
rename : cpu/beta_cpu/rob.cc => cpu/o3/rob.cc
rename : cpu/beta_cpu/rob.hh => cpu/o3/rob.hh
rename : cpu/beta_cpu/rob_impl.hh => cpu/o3/rob_impl.hh
rename : cpu/beta_cpu/sat_counter.cc => cpu/o3/sat_counter.cc
rename : cpu/beta_cpu/sat_counter.hh => cpu/o3/sat_counter.hh
rename : cpu/beta_cpu/store_set.cc => cpu/o3/store_set.cc
rename : cpu/beta_cpu/store_set.hh => cpu/o3/store_set.hh
rename : cpu/beta_cpu/tournament_pred.cc => cpu/o3/tournament_pred.cc
rename : cpu/beta_cpu/tournament_pred.hh => cpu/o3/tournament_pred.hh
rename : cpu/ooo_cpu/ooo_cpu.cc => cpu/ozone/cpu.cc
rename : cpu/ooo_cpu/ooo_cpu.hh => cpu/ozone/cpu.hh
rename : cpu/ooo_cpu/ooo_impl.hh => cpu/ozone/cpu_impl.hh
rename : cpu/ooo_cpu/ea_list.cc => cpu/ozone/ea_list.cc
rename : cpu/ooo_cpu/ea_list.hh => cpu/ozone/ea_list.hh
rename : cpu/simple_cpu/simple_cpu.cc => cpu/simple/cpu.cc
rename : cpu/simple_cpu/simple_cpu.hh => cpu/simple/cpu.hh
rename : cpu/full_cpu/smt.hh => cpu/smt.hh
rename : cpu/full_cpu/op_class.hh => encumbered/cpu/full/op_class.hh
extra : convert_revision : c4a891d8d6d3e0e9e5ea56be47d851da44d8c032

2 files changed, 1254 insertions, 0 deletions

diff --git a/cpu/simple/cpu.cc b/cpu/simple/cpu.cc
new file mode 100644
index 000000000..306398ac2
--- /dev/null
+++ b/cpu/simple/cpu.cc
@@ -0,0 +1,911 @@
+/*
+ * Copyright (c) 2002-2004 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.
+ */
+
+#include <cmath>
+#include <cstdio>
+#include <cstdlib>
+#include <iostream>
+#include <iomanip>
+#include <list>
+#include <sstream>
+#include <string>
+
+#include "base/cprintf.hh"
+#include "base/inifile.hh"
+#include "base/loader/symtab.hh"
+#include "base/misc.hh"
+#include "base/pollevent.hh"
+#include "base/range.hh"
+#include "base/stats/events.hh"
+#include "base/trace.hh"
+#include "cpu/base.hh"
+#include "cpu/exec_context.hh"
+#include "cpu/exetrace.hh"
+#include "cpu/sampler/sampler.hh"
+#include "cpu/simple/cpu.hh"
+#include "cpu/smt.hh"
+#include "cpu/static_inst.hh"
+#include "mem/base_mem.hh"
+#include "mem/mem_interface.hh"
+#include "sim/builder.hh"
+#include "sim/debug.hh"
+#include "sim/host.hh"
+#include "sim/sim_events.hh"
+#include "sim/sim_object.hh"
+#include "sim/stats.hh"
+
+#ifdef FULL_SYSTEM
+#include "base/remote_gdb.hh"
+#include "mem/functional/memory_control.hh"
+#include "mem/functional/physical.hh"
+#include "sim/system.hh"
+#include "targetarch/alpha_memory.hh"
+#include "targetarch/vtophys.hh"
+#else // !FULL_SYSTEM
+#include "eio/eio.hh"
+#include "mem/functional/functional.hh"
+#endif // FULL_SYSTEM
+
+using namespace std;
+
+
+SimpleCPU::TickEvent::TickEvent(SimpleCPU *c, int w)
+    : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c), width(w)
+{
+}
+
+void
+SimpleCPU::TickEvent::process()
+{
+    int count = width;
+    do {
+        cpu->tick();
+    } while (--count > 0 && cpu->status() == Running);
+}
+
+const char *
+SimpleCPU::TickEvent::description()
+{
+    return "SimpleCPU tick event";
+}
+
+
+SimpleCPU::CacheCompletionEvent::CacheCompletionEvent(SimpleCPU *_cpu)
+    : Event(&mainEventQueue), cpu(_cpu)
+{
+}
+
+void SimpleCPU::CacheCompletionEvent::process()
+{
+    cpu->processCacheCompletion();
+}
+
+const char *
+SimpleCPU::CacheCompletionEvent::description()
+{
+    return "SimpleCPU cache completion event";
+}
+
+SimpleCPU::SimpleCPU(Params *p)
+    : BaseCPU(p), tickEvent(this, p->width), xc(NULL),
+      cacheCompletionEvent(this)
+{
+    _status = Idle;
+#ifdef FULL_SYSTEM
+    xc = new ExecContext(this, 0, p->system, p->itb, p->dtb, p->mem);
+
+    // initialize CPU, including PC
+    TheISA::initCPU(&xc->regs);
+#else
+    xc = new ExecContext(this, /* thread_num */ 0, p->process, /* asid */ 0);
+#endif // !FULL_SYSTEM
+
+    icacheInterface = p->icache_interface;
+    dcacheInterface = p->dcache_interface;
+
+    memReq = new MemReq();
+    memReq->xc = xc;
+    memReq->asid = 0;
+    memReq->data = new uint8_t[64];
+
+    numInst = 0;
+    startNumInst = 0;
+    numLoad = 0;
+    startNumLoad = 0;
+    lastIcacheStall = 0;
+    lastDcacheStall = 0;
+
+    execContexts.push_back(xc);
+}
+
+SimpleCPU::~SimpleCPU()
+{
+}
+
+void
+SimpleCPU::switchOut(SamplingCPU *s)
+{
+    sampler = s;
+    if (status() == DcacheMissStall) {
+        DPRINTF(Sampler,"Outstanding dcache access, waiting for completion\n");
+        _status = DcacheMissSwitch;
+    }
+    else {
+        _status = SwitchedOut;
+
+        if (tickEvent.scheduled())
+            tickEvent.squash();
+
+        sampler->signalSwitched();
+    }
+}
+
+
+void
+SimpleCPU::takeOverFrom(BaseCPU *oldCPU)
+{
+    BaseCPU::takeOverFrom(oldCPU);
+
+    assert(!tickEvent.scheduled());
+
+    // if any of this CPU's ExecContexts are active, mark the CPU as
+    // running and schedule its tick event.
+    for (int i = 0; i < execContexts.size(); ++i) {
+        ExecContext *xc = execContexts[i];
+        if (xc->status() == ExecContext::Active && _status != Running) {
+            _status = Running;
+            tickEvent.schedule(curTick);
+        }
+    }
+}
+
+
+void
+SimpleCPU::activateContext(int thread_num, int delay)
+{
+    assert(thread_num == 0);
+    assert(xc);
+
+    assert(_status == Idle);
+    notIdleFraction++;
+    scheduleTickEvent(delay);
+    _status = Running;
+}
+
+
+void
+SimpleCPU::suspendContext(int thread_num)
+{
+    assert(thread_num == 0);
+    assert(xc);
+
+    assert(_status == Running);
+    notIdleFraction--;
+    unscheduleTickEvent();
+    _status = Idle;
+}
+
+
+void
+SimpleCPU::deallocateContext(int thread_num)
+{
+    // for now, these are equivalent
+    suspendContext(thread_num);
+}
+
+
+void
+SimpleCPU::haltContext(int thread_num)
+{
+    // for now, these are equivalent
+    suspendContext(thread_num);
+}
+
+
+void
+SimpleCPU::regStats()
+{
+    using namespace Stats;
+
+    BaseCPU::regStats();
+
+    numInsts
+        .name(name() + ".num_insts")
+        .desc("Number of instructions executed")
+        ;
+
+    numMemRefs
+        .name(name() + ".num_refs")
+        .desc("Number of memory references")
+        ;
+
+    notIdleFraction
+        .name(name() + ".not_idle_fraction")
+        .desc("Percentage of non-idle cycles")
+        ;
+
+    idleFraction
+        .name(name() + ".idle_fraction")
+        .desc("Percentage of idle cycles")
+        ;
+
+    icacheStallCycles
+        .name(name() + ".icache_stall_cycles")
+        .desc("ICache total stall cycles")
+        .prereq(icacheStallCycles)
+        ;
+
+    dcacheStallCycles
+        .name(name() + ".dcache_stall_cycles")
+        .desc("DCache total stall cycles")
+        .prereq(dcacheStallCycles)
+        ;
+
+    idleFraction = constant(1.0) - notIdleFraction;
+}
+
+void
+SimpleCPU::resetStats()
+{
+    startNumInst = numInst;
+    notIdleFraction = (_status != Idle);
+}
+
+void
+SimpleCPU::serialize(ostream &os)
+{
+    BaseCPU::serialize(os);
+    SERIALIZE_ENUM(_status);
+    SERIALIZE_SCALAR(inst);
+    nameOut(os, csprintf("%s.xc", name()));
+    xc->serialize(os);
+    nameOut(os, csprintf("%s.tickEvent", name()));
+    tickEvent.serialize(os);
+    nameOut(os, csprintf("%s.cacheCompletionEvent", name()));
+    cacheCompletionEvent.serialize(os);
+}
+
+void
+SimpleCPU::unserialize(Checkpoint *cp, const string &section)
+{
+    BaseCPU::unserialize(cp, section);
+    UNSERIALIZE_ENUM(_status);
+    UNSERIALIZE_SCALAR(inst);
+    xc->unserialize(cp, csprintf("%s.xc", section));
+    tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
+    cacheCompletionEvent
+        .unserialize(cp, csprintf("%s.cacheCompletionEvent", section));
+}
+
+void
+change_thread_state(int thread_number, int activate, int priority)
+{
+}
+
+Fault
+SimpleCPU::copySrcTranslate(Addr src)
+{
+    static bool no_warn = true;
+    int blk_size = (dcacheInterface) ? dcacheInterface->getBlockSize() : 64;
+    // Only support block sizes of 64 atm.
+    assert(blk_size == 64);
+    int offset = src & (blk_size - 1);
+
+    // Make sure block doesn't span page
+    if (no_warn &&
+        (src & TheISA::PageMask) != ((src + blk_size) & TheISA::PageMask) &&
+        (src >> 40) != 0xfffffc) {
+        warn("Copied block source spans pages %x.", src);
+        no_warn = false;
+    }
+
+    memReq->reset(src & ~(blk_size - 1), blk_size);
+
+    // translate to physical address
+    Fault fault = xc->translateDataReadReq(memReq);
+
+    assert(fault != Alignment_Fault);
+
+    if (fault == No_Fault) {
+        xc->copySrcAddr = src;
+        xc->copySrcPhysAddr = memReq->paddr + offset;
+    } else {
+        xc->copySrcAddr = 0;
+        xc->copySrcPhysAddr = 0;
+    }
+    return fault;
+}
+
+Fault
+SimpleCPU::copy(Addr dest)
+{
+    static bool no_warn = true;
+    int blk_size = (dcacheInterface) ? dcacheInterface->getBlockSize() : 64;
+    // Only support block sizes of 64 atm.
+    assert(blk_size == 64);
+    uint8_t data[blk_size];
+    //assert(xc->copySrcAddr);
+    int offset = dest & (blk_size - 1);
+
+    // Make sure block doesn't span page
+    if (no_warn &&
+        (dest & TheISA::PageMask) != ((dest + blk_size) & TheISA::PageMask) &&
+        (dest >> 40) != 0xfffffc) {
+        no_warn = false;
+        warn("Copied block destination spans pages %x. ", dest);
+    }
+
+    memReq->reset(dest & ~(blk_size -1), blk_size);
+    // translate to physical address
+    Fault fault = xc->translateDataWriteReq(memReq);
+
+    assert(fault != Alignment_Fault);
+
+    if (fault == No_Fault) {
+        Addr dest_addr = memReq->paddr + offset;
+        // Need to read straight from memory since we have more than 8 bytes.
+        memReq->paddr = xc->copySrcPhysAddr;
+        xc->mem->read(memReq, data);
+        memReq->paddr = dest_addr;
+        xc->mem->write(memReq, data);
+        if (dcacheInterface) {
+            memReq->cmd = Copy;
+            memReq->completionEvent = NULL;
+            memReq->paddr = xc->copySrcPhysAddr;
+            memReq->dest = dest_addr;
+            memReq->size = 64;
+            memReq->time = curTick;
+            dcacheInterface->access(memReq);
+        }
+    }
+    return fault;
+}
+
+// precise architected memory state accessor macros
+template <class T>
+Fault
+SimpleCPU::read(Addr addr, T &data, unsigned flags)
+{
+    if (status() == DcacheMissStall || status() == DcacheMissSwitch) {
+        Fault fault = xc->read(memReq,data);
+
+        if (traceData) {
+            traceData->setAddr(addr);
+        }
+        return fault;
+    }
+
+    memReq->reset(addr, sizeof(T), flags);
+
+    // translate to physical address
+    Fault fault = xc->translateDataReadReq(memReq);
+
+    // if we have a cache, do cache access too
+    if (fault == No_Fault && dcacheInterface) {
+        memReq->cmd = Read;
+        memReq->completionEvent = NULL;
+        memReq->time = curTick;
+        MemAccessResult result = dcacheInterface->access(memReq);
+
+        // Ugly hack to get an event scheduled *only* if the access is
+        // a miss.  We really should add first-class support for this
+        // at some point.
+        if (result != MA_HIT && dcacheInterface->doEvents()) {
+            memReq->completionEvent = &cacheCompletionEvent;
+            lastDcacheStall = curTick;
+            unscheduleTickEvent();
+            _status = DcacheMissStall;
+        } else {
+            // do functional access
+            fault = xc->read(memReq, data);
+
+        }
+    } else if(fault == No_Fault) {
+        // do functional access
+        fault = xc->read(memReq, data);
+
+    }
+
+    if (!dcacheInterface && (memReq->flags & UNCACHEABLE))
+        recordEvent("Uncached Read");
+
+    return fault;
+}
+
+#ifndef DOXYGEN_SHOULD_SKIP_THIS
+
+template
+Fault
+SimpleCPU::read(Addr addr, uint64_t &data, unsigned flags);
+
+template
+Fault
+SimpleCPU::read(Addr addr, uint32_t &data, unsigned flags);
+
+template
+Fault
+SimpleCPU::read(Addr addr, uint16_t &data, unsigned flags);
+
+template
+Fault
+SimpleCPU::read(Addr addr, uint8_t &data, unsigned flags);
+
+#endif //DOXYGEN_SHOULD_SKIP_THIS
+
+template<>
+Fault
+SimpleCPU::read(Addr addr, double &data, unsigned flags)
+{
+    return read(addr, *(uint64_t*)&data, flags);
+}
+
+template<>
+Fault
+SimpleCPU::read(Addr addr, float &data, unsigned flags)
+{
+    return read(addr, *(uint32_t*)&data, flags);
+}
+
+
+template<>
+Fault
+SimpleCPU::read(Addr addr, int32_t &data, unsigned flags)
+{
+    return read(addr, (uint32_t&)data, flags);
+}
+
+
+template <class T>
+Fault
+SimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
+{
+    memReq->reset(addr, sizeof(T), flags);
+
+    // translate to physical address
+    Fault fault = xc->translateDataWriteReq(memReq);
+
+    // do functional access
+    if (fault == No_Fault)
+        fault = xc->write(memReq, data);
+
+    if (fault == No_Fault && dcacheInterface) {
+        memReq->cmd = Write;
+        memcpy(memReq->data,(uint8_t *)&data,memReq->size);
+        memReq->completionEvent = NULL;
+        memReq->time = curTick;
+        MemAccessResult result = dcacheInterface->access(memReq);
+
+        // Ugly hack to get an event scheduled *only* if the access is
+        // a miss.  We really should add first-class support for this
+        // at some point.
+        if (result != MA_HIT && dcacheInterface->doEvents()) {
+            memReq->completionEvent = &cacheCompletionEvent;
+            lastDcacheStall = curTick;
+            unscheduleTickEvent();
+            _status = DcacheMissStall;
+        }
+    }
+
+    if (res && (fault == No_Fault))
+        *res = memReq->result;
+
+    if (!dcacheInterface && (memReq->flags & UNCACHEABLE))
+        recordEvent("Uncached Write");
+
+    return fault;
+}
+
+
+#ifndef DOXYGEN_SHOULD_SKIP_THIS
+template
+Fault
+SimpleCPU::write(uint64_t data, Addr addr, unsigned flags, uint64_t *res);
+
+template
+Fault
+SimpleCPU::write(uint32_t data, Addr addr, unsigned flags, uint64_t *res);
+
+template
+Fault
+SimpleCPU::write(uint16_t data, Addr addr, unsigned flags, uint64_t *res);
+
+template
+Fault
+SimpleCPU::write(uint8_t data, Addr addr, unsigned flags, uint64_t *res);
+
+#endif //DOXYGEN_SHOULD_SKIP_THIS
+
+template<>
+Fault
+SimpleCPU::write(double data, Addr addr, unsigned flags, uint64_t *res)
+{
+    return write(*(uint64_t*)&data, addr, flags, res);
+}
+
+template<>
+Fault
+SimpleCPU::write(float data, Addr addr, unsigned flags, uint64_t *res)
+{
+    return write(*(uint32_t*)&data, addr, flags, res);
+}
+
+
+template<>
+Fault
+SimpleCPU::write(int32_t data, Addr addr, unsigned flags, uint64_t *res)
+{
+    return write((uint32_t)data, addr, flags, res);
+}
+
+
+#ifdef FULL_SYSTEM
+Addr
+SimpleCPU::dbg_vtophys(Addr addr)
+{
+    return vtophys(xc, addr);
+}
+#endif // FULL_SYSTEM
+
+void
+SimpleCPU::processCacheCompletion()
+{
+    switch (status()) {
+      case IcacheMissStall:
+        icacheStallCycles += curTick - lastIcacheStall;
+        _status = IcacheMissComplete;
+        scheduleTickEvent(1);
+        break;
+      case DcacheMissStall:
+        if (memReq->cmd.isRead()) {
+            curStaticInst->execute(this,traceData);
+            if (traceData)
+                traceData->finalize();
+        }
+        dcacheStallCycles += curTick - lastDcacheStall;
+        _status = Running;
+        scheduleTickEvent(1);
+        break;
+      case DcacheMissSwitch:
+        if (memReq->cmd.isRead()) {
+            curStaticInst->execute(this,traceData);
+            if (traceData)
+                traceData->finalize();
+        }
+        _status = SwitchedOut;
+        sampler->signalSwitched();
+      case SwitchedOut:
+        // If this CPU has been switched out due to sampling/warm-up,
+        // ignore any further status changes (e.g., due to cache
+        // misses outstanding at the time of the switch).
+        return;
+      default:
+        panic("SimpleCPU::processCacheCompletion: bad state");
+        break;
+    }
+}
+
+#ifdef FULL_SYSTEM
+void
+SimpleCPU::post_interrupt(int int_num, int index)
+{
+    BaseCPU::post_interrupt(int_num, index);
+
+    if (xc->status() == ExecContext::Suspended) {
+                DPRINTF(IPI,"Suspended Processor awoke\n");
+        xc->activate();
+    }
+}
+#endif // FULL_SYSTEM
+
+/* start simulation, program loaded, processor precise state initialized */
+void
+SimpleCPU::tick()
+{
+    numCycles++;
+
+    traceData = NULL;
+
+    Fault fault = No_Fault;
+
+#ifdef FULL_SYSTEM
+    if (checkInterrupts && check_interrupts() && !xc->inPalMode() &&
+        status() != IcacheMissComplete) {
+        int ipl = 0;
+        int summary = 0;
+        checkInterrupts = false;
+        IntReg *ipr = xc->regs.ipr;
+
+        if (xc->regs.ipr[TheISA::IPR_SIRR]) {
+            for (int i = TheISA::INTLEVEL_SOFTWARE_MIN;
+                 i < TheISA::INTLEVEL_SOFTWARE_MAX; i++) {
+                if (ipr[TheISA::IPR_SIRR] & (ULL(1) << i)) {
+                    // See table 4-19 of 21164 hardware reference
+                    ipl = (i - TheISA::INTLEVEL_SOFTWARE_MIN) + 1;
+                    summary |= (ULL(1) << i);
+                }
+            }
+        }
+
+        uint64_t interrupts = xc->cpu->intr_status();
+        for (int i = TheISA::INTLEVEL_EXTERNAL_MIN;
+            i < TheISA::INTLEVEL_EXTERNAL_MAX; i++) {
+            if (interrupts & (ULL(1) << i)) {
+                // See table 4-19 of 21164 hardware reference
+                ipl = i;
+                summary |= (ULL(1) << i);
+            }
+        }
+
+        if (ipr[TheISA::IPR_ASTRR])
+            panic("asynchronous traps not implemented\n");
+
+        if (ipl && ipl > xc->regs.ipr[TheISA::IPR_IPLR]) {
+            ipr[TheISA::IPR_ISR] = summary;
+            ipr[TheISA::IPR_INTID] = ipl;
+            xc->ev5_trap(Interrupt_Fault);
+
+            DPRINTF(Flow, "Interrupt! IPLR=%d ipl=%d summary=%x\n",
+                    ipr[TheISA::IPR_IPLR], ipl, summary);
+        }
+    }
+#endif
+
+    // maintain $r0 semantics
+    xc->regs.intRegFile[ZeroReg] = 0;
+#ifdef TARGET_ALPHA
+    xc->regs.floatRegFile.d[ZeroReg] = 0.0;
+#endif // TARGET_ALPHA
+
+    if (status() == IcacheMissComplete) {
+        // We've already fetched an instruction and were stalled on an
+        // I-cache miss.  No need to fetch it again.
+
+        // Set status to running; tick event will get rescheduled if
+        // necessary at end of tick() function.
+        _status = Running;
+    }
+    else {
+        // Try to fetch an instruction
+
+        // set up memory request for instruction fetch
+#ifdef FULL_SYSTEM
+#define IFETCH_FLAGS(pc)	((pc) & 1) ? PHYSICAL : 0
+#else
+#define IFETCH_FLAGS(pc)	0
+#endif
+
+        memReq->cmd = Read;
+        memReq->reset(xc->regs.pc & ~3, sizeof(uint32_t),
+                     IFETCH_FLAGS(xc->regs.pc));
+
+        fault = xc->translateInstReq(memReq);
+
+        if (fault == No_Fault)
+            fault = xc->mem->read(memReq, inst);
+
+        if (icacheInterface && fault == No_Fault) {
+            memReq->completionEvent = NULL;
+
+            memReq->time = curTick;
+            MemAccessResult result = icacheInterface->access(memReq);
+
+            // Ugly hack to get an event scheduled *only* if the access is
+            // a miss.  We really should add first-class support for this
+            // at some point.
+            if (result != MA_HIT && icacheInterface->doEvents()) {
+                memReq->completionEvent = &cacheCompletionEvent;
+                lastIcacheStall = curTick;
+                unscheduleTickEvent();
+                _status = IcacheMissStall;
+                return;
+            }
+        }
+    }
+
+    // If we've got a valid instruction (i.e., no fault on instruction
+    // fetch), then execute it.
+    if (fault == No_Fault) {
+
+        // keep an instruction count
+        numInst++;
+        numInsts++;
+
+        // check for instruction-count-based events
+        comInstEventQueue[0]->serviceEvents(numInst);
+
+        // decode the instruction
+        inst = htoa(inst);
+        curStaticInst = StaticInst<TheISA>::decode(inst);
+
+        traceData = Trace::getInstRecord(curTick, xc, this, curStaticInst,
+                                         xc->regs.pc);
+
+#ifdef FULL_SYSTEM
+        xc->setInst(inst);
+#endif // FULL_SYSTEM
+
+        xc->func_exe_inst++;
+
+        fault = curStaticInst->execute(this, traceData);
+
+#ifdef FULL_SYSTEM
+        if (xc->fnbin)
+            xc->execute(curStaticInst.get());
+#endif
+
+        if (curStaticInst->isMemRef()) {
+            numMemRefs++;
+        }
+
+        if (curStaticInst->isLoad()) {
+            ++numLoad;
+            comLoadEventQueue[0]->serviceEvents(numLoad);
+        }
+
+        // If we have a dcache miss, then we can't finialize the instruction
+        // trace yet because we want to populate it with the data later
+        if (traceData &&
+                !(status() == DcacheMissStall && memReq->cmd.isRead())) {
+            traceData->finalize();
+        }
+
+        traceFunctions(xc->regs.pc);
+
+    }	// if (fault == No_Fault)
+
+    if (fault != No_Fault) {
+#ifdef FULL_SYSTEM
+        xc->ev5_trap(fault);
+#else // !FULL_SYSTEM
+        fatal("fault (%d) detected @ PC 0x%08p", fault, xc->regs.pc);
+#endif // FULL_SYSTEM
+    }
+    else {
+        // go to the next instruction
+        xc->regs.pc = xc->regs.npc;
+        xc->regs.npc += sizeof(MachInst);
+    }
+
+#ifdef FULL_SYSTEM
+    Addr oldpc;
+    do {
+        oldpc = xc->regs.pc;
+        system->pcEventQueue.service(xc);
+    } while (oldpc != xc->regs.pc);
+#endif
+
+    assert(status() == Running ||
+           status() == Idle ||
+           status() == DcacheMissStall);
+
+    if (status() == Running && !tickEvent.scheduled())
+        tickEvent.schedule(curTick + cycles(1));
+}
+
+
+////////////////////////////////////////////////////////////////////////
+//
+//  SimpleCPU Simulation Object
+//
+BEGIN_DECLARE_SIM_OBJECT_PARAMS(SimpleCPU)
+
+    Param<Counter> max_insts_any_thread;
+    Param<Counter> max_insts_all_threads;
+    Param<Counter> max_loads_any_thread;
+    Param<Counter> max_loads_all_threads;
+
+#ifdef FULL_SYSTEM
+    SimObjectParam<AlphaITB *> itb;
+    SimObjectParam<AlphaDTB *> dtb;
+    SimObjectParam<FunctionalMemory *> mem;
+    SimObjectParam<System *> system;
+    Param<int> mult;
+#else
+    SimObjectParam<Process *> workload;
+#endif // FULL_SYSTEM
+
+    Param<int> clock;
+    SimObjectParam<BaseMem *> icache;
+    SimObjectParam<BaseMem *> dcache;
+
+    Param<bool> defer_registration;
+    Param<int> width;
+    Param<bool> function_trace;
+    Param<Tick> function_trace_start;
+
+END_DECLARE_SIM_OBJECT_PARAMS(SimpleCPU)
+
+BEGIN_INIT_SIM_OBJECT_PARAMS(SimpleCPU)
+
+    INIT_PARAM(max_insts_any_thread,
+               "terminate when any thread reaches this inst count"),
+    INIT_PARAM(max_insts_all_threads,
+               "terminate when all threads have reached this inst count"),
+    INIT_PARAM(max_loads_any_thread,
+               "terminate when any thread reaches this load count"),
+    INIT_PARAM(max_loads_all_threads,
+               "terminate when all threads have reached this load count"),
+
+#ifdef FULL_SYSTEM
+    INIT_PARAM(itb, "Instruction TLB"),
+    INIT_PARAM(dtb, "Data TLB"),
+    INIT_PARAM(mem, "memory"),
+    INIT_PARAM(system, "system object"),
+    INIT_PARAM(mult, "system clock multiplier"),
+#else
+    INIT_PARAM(workload, "processes to run"),
+#endif // FULL_SYSTEM
+
+    INIT_PARAM(clock, "clock speed"),
+    INIT_PARAM(icache, "L1 instruction cache object"),
+    INIT_PARAM(dcache, "L1 data cache object"),
+    INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
+    INIT_PARAM(width, "cpu width"),
+    INIT_PARAM(function_trace, "Enable function trace"),
+    INIT_PARAM(function_trace_start, "Cycle to start function trace")
+
+END_INIT_SIM_OBJECT_PARAMS(SimpleCPU)
+
+
+CREATE_SIM_OBJECT(SimpleCPU)
+{
+#ifdef FULL_SYSTEM
+    if (mult != 1)
+        panic("processor clock multiplier must be 1\n");
+#endif
+
+    SimpleCPU::Params *params = new SimpleCPU::Params();
+    params->name = getInstanceName();
+    params->numberOfThreads = 1;
+    params->max_insts_any_thread = max_insts_any_thread;
+    params->max_insts_all_threads = max_insts_all_threads;
+    params->max_loads_any_thread = max_loads_any_thread;
+    params->max_loads_all_threads = max_loads_all_threads;
+    params->deferRegistration = defer_registration;
+    params->clock = clock;
+    params->functionTrace = function_trace;
+    params->functionTraceStart = function_trace_start;
+    params->icache_interface = (icache) ? icache->getInterface() : NULL;
+    params->dcache_interface = (dcache) ? dcache->getInterface() : NULL;
+    params->width = width;
+
+#ifdef FULL_SYSTEM
+    params->itb = itb;
+    params->dtb = dtb;
+    params->mem = mem;
+    params->system = system;
+#else
+    params->process = workload;
+#endif
+
+    SimpleCPU *cpu = new SimpleCPU(params);
+    return cpu;
+}
+
+REGISTER_SIM_OBJECT("SimpleCPU", SimpleCPU)
+
diff --git a/cpu/simple/cpu.hh b/cpu/simple/cpu.hh
new file mode 100644
index 000000000..9a0c2952a
--- /dev/null
+++ b/cpu/simple/cpu.hh
@@ -0,0 +1,343 @@
+/*
+ * Copyright (c) 2002-2004 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_SIMPLE_CPU_SIMPLE_CPU_HH__
+#define __CPU_SIMPLE_CPU_SIMPLE_CPU_HH__
+
+#include "base/statistics.hh"
+#include "cpu/base.hh"
+#include "cpu/exec_context.hh"
+#include "cpu/pc_event.hh"
+#include "cpu/sampler/sampler.hh"
+#include "cpu/static_inst.hh"
+#include "sim/eventq.hh"
+
+// forward declarations
+#ifdef FULL_SYSTEM
+class Processor;
+class AlphaITB;
+class AlphaDTB;
+class PhysicalMemory;
+
+class RemoteGDB;
+class GDBListener;
+
+#else
+
+class Process;
+
+#endif // FULL_SYSTEM
+
+class MemInterface;
+class Checkpoint;
+
+namespace Trace {
+    class InstRecord;
+}
+
+class SimpleCPU : public BaseCPU
+{
+  public:
+    // main simulation loop (one cycle)
+    void tick();
+
+  private:
+    struct TickEvent : public Event
+    {
+        SimpleCPU *cpu;
+        int width;
+
+        TickEvent(SimpleCPU *c, int w);
+        void process();
+        const char *description();
+    };
+
+    TickEvent tickEvent;
+
+    /// Schedule tick event, regardless of its current state.
+    void scheduleTickEvent(int numCycles)
+    {
+        if (tickEvent.squashed())
+            tickEvent.reschedule(curTick + cycles(numCycles));
+        else if (!tickEvent.scheduled())
+            tickEvent.schedule(curTick + cycles(numCycles));
+    }
+
+    /// Unschedule tick event, regardless of its current state.
+    void unscheduleTickEvent()
+    {
+        if (tickEvent.scheduled())
+            tickEvent.squash();
+    }
+
+  private:
+    Trace::InstRecord *traceData;
+
+  public:
+    //
+    enum Status {
+        Running,
+        Idle,
+        IcacheMissStall,
+        IcacheMissComplete,
+        DcacheMissStall,
+        DcacheMissSwitch,
+        SwitchedOut
+    };
+
+  private:
+    Status _status;
+
+  public:
+    void post_interrupt(int int_num, int index);
+
+    void zero_fill_64(Addr addr) {
+      static int warned = 0;
+      if (!warned) {
+        warn ("WH64 is not implemented");
+        warned = 1;
+      }
+    };
+
+  public:
+    struct Params : public BaseCPU::Params
+    {
+        MemInterface *icache_interface;
+        MemInterface *dcache_interface;
+        int width;
+#ifdef FULL_SYSTEM
+        AlphaITB *itb;
+        AlphaDTB *dtb;
+        FunctionalMemory *mem;
+#else
+        Process *process;
+#endif
+    };
+    SimpleCPU(Params *params);
+    virtual ~SimpleCPU();
+
+  public:
+    // execution context
+    ExecContext *xc;
+
+    void switchOut(SamplingCPU *s);
+    void takeOverFrom(BaseCPU *oldCPU);
+
+#ifdef FULL_SYSTEM
+    Addr dbg_vtophys(Addr addr);
+
+    bool interval_stats;
+#endif
+
+    // L1 instruction cache
+    MemInterface *icacheInterface;
+
+    // L1 data cache
+    MemInterface *dcacheInterface;
+
+    // current instruction
+    MachInst inst;
+
+    // Refcounted pointer to the one memory request.
+    MemReqPtr memReq;
+
+    // Pointer to the sampler that is telling us to switchover.
+    // Used to signal the completion of the pipe drain and schedule
+    // the next switchover
+    SamplingCPU *sampler;
+
+    StaticInstPtr<TheISA> curStaticInst;
+
+    class CacheCompletionEvent : public Event
+    {
+      private:
+        SimpleCPU *cpu;
+
+      public:
+        CacheCompletionEvent(SimpleCPU *_cpu);
+
+        virtual void process();
+        virtual const char *description();
+    };
+
+    CacheCompletionEvent cacheCompletionEvent;
+
+    Status status() const { return _status; }
+
+    virtual void activateContext(int thread_num, int delay);
+    virtual void suspendContext(int thread_num);
+    virtual void deallocateContext(int thread_num);
+    virtual void haltContext(int thread_num);
+
+    // statistics
+    virtual void regStats();
+    virtual void resetStats();
+
+    // number of simulated instructions
+    Counter numInst;
+    Counter startNumInst;
+    Stats::Scalar<> numInsts;
+
+    virtual Counter totalInstructions() const
+    {
+        return numInst - startNumInst;
+    }
+
+    // number of simulated memory references
+    Stats::Scalar<> numMemRefs;
+
+    // number of simulated loads
+    Counter numLoad;
+    Counter startNumLoad;
+
+    // number of idle cycles
+    Stats::Average<> notIdleFraction;
+    Stats::Formula idleFraction;
+
+    // number of cycles stalled for I-cache misses
+    Stats::Scalar<> icacheStallCycles;
+    Counter lastIcacheStall;
+
+    // number of cycles stalled for D-cache misses
+    Stats::Scalar<> dcacheStallCycles;
+    Counter lastDcacheStall;
+
+    void processCacheCompletion();
+
+    virtual void serialize(std::ostream &os);
+    virtual void unserialize(Checkpoint *cp, const std::string &section);
+
+    template <class T>
+    Fault read(Addr addr, T &data, unsigned flags);
+
+    template <class T>
+    Fault write(T data, Addr addr, unsigned flags, uint64_t *res);
+
+    // These functions are only used in CPU models that split
+    // effective address computation from the actual memory access.
+    void setEA(Addr EA) { panic("SimpleCPU::setEA() not implemented\n"); }
+    Addr getEA() 	{ panic("SimpleCPU::getEA() not implemented\n"); }
+
+    void prefetch(Addr addr, unsigned flags)
+    {
+        // need to do this...
+    }
+
+    void writeHint(Addr addr, int size, unsigned flags)
+    {
+        // need to do this...
+    }
+
+    Fault copySrcTranslate(Addr src);
+
+    Fault copy(Addr dest);
+
+    // The register accessor methods provide the index of the
+    // instruction's operand (e.g., 0 or 1), not the architectural
+    // register index, to simplify the implementation of register
+    // renaming.  We find the architectural register index by indexing
+    // into the instruction's own operand index table.  Note that a
+    // raw pointer to the StaticInst is provided instead of a
+    // ref-counted StaticInstPtr to redice overhead.  This is fine as
+    // long as these methods don't copy the pointer into any long-term
+    // storage (which is pretty hard to imagine they would have reason
+    // to do).
+
+    uint64_t readIntReg(const StaticInst<TheISA> *si, int idx)
+    {
+        return xc->readIntReg(si->srcRegIdx(idx));
+    }
+
+    float readFloatRegSingle(const StaticInst<TheISA> *si, int idx)
+    {
+        int reg_idx = si->srcRegIdx(idx) - TheISA::FP_Base_DepTag;
+        return xc->readFloatRegSingle(reg_idx);
+    }
+
+    double readFloatRegDouble(const StaticInst<TheISA> *si, int idx)
+    {
+        int reg_idx = si->srcRegIdx(idx) - TheISA::FP_Base_DepTag;
+        return xc->readFloatRegDouble(reg_idx);
+    }
+
+    uint64_t readFloatRegInt(const StaticInst<TheISA> *si, int idx)
+    {
+        int reg_idx = si->srcRegIdx(idx) - TheISA::FP_Base_DepTag;
+        return xc->readFloatRegInt(reg_idx);
+    }
+
+    void setIntReg(const StaticInst<TheISA> *si, int idx, uint64_t val)
+    {
+        xc->setIntReg(si->destRegIdx(idx), val);
+    }
+
+    void setFloatRegSingle(const StaticInst<TheISA> *si, int idx, float val)
+    {
+        int reg_idx = si->destRegIdx(idx) - TheISA::FP_Base_DepTag;
+        xc->setFloatRegSingle(reg_idx, val);
+    }
+
+    void setFloatRegDouble(const StaticInst<TheISA> *si, int idx, double val)
+    {
+        int reg_idx = si->destRegIdx(idx) - TheISA::FP_Base_DepTag;
+        xc->setFloatRegDouble(reg_idx, val);
+    }
+
+    void setFloatRegInt(const StaticInst<TheISA> *si, int idx, uint64_t val)
+    {
+        int reg_idx = si->destRegIdx(idx) - TheISA::FP_Base_DepTag;
+        xc->setFloatRegInt(reg_idx, val);
+    }
+
+    uint64_t readPC() { return xc->readPC(); }
+    void setNextPC(uint64_t val) { xc->setNextPC(val); }
+
+    uint64_t readUniq() { return xc->readUniq(); }
+    void setUniq(uint64_t val) { xc->setUniq(val); }
+
+    uint64_t readFpcr() { return xc->readFpcr(); }
+    void setFpcr(uint64_t val) { xc->setFpcr(val); }
+
+#ifdef FULL_SYSTEM
+    uint64_t readIpr(int idx, Fault &fault) { return xc->readIpr(idx, fault); }
+    Fault setIpr(int idx, uint64_t val) { return xc->setIpr(idx, val); }
+    Fault hwrei() { return xc->hwrei(); }
+    int readIntrFlag() { return xc->readIntrFlag(); }
+    void setIntrFlag(int val) { xc->setIntrFlag(val); }
+    bool inPalMode() { return xc->inPalMode(); }
+    void ev5_trap(Fault fault) { xc->ev5_trap(fault); }
+    bool simPalCheck(int palFunc) { return xc->simPalCheck(palFunc); }
+#else
+    void syscall() { xc->syscall(); }
+#endif
+
+    bool misspeculating() { return xc->misspeculating(); }
+    ExecContext *xcBase() { return xc; }
+};
+
+#endif // __CPU_SIMPLE_CPU_SIMPLE_CPU_HH__