#include <bitset>
#include <list>
#include <string>
+#include <queue>
-#include "arch/faults.hh"
#include "arch/utility.hh"
-#include "base/fast_alloc.hh"
#include "base/trace.hh"
-#include "config/full_system.hh"
#include "config/the_isa.hh"
+#include "cpu/checker/cpu.hh"
#include "cpu/o3/comm.hh"
#include "cpu/exetrace.hh"
#include "cpu/inst_seq.hh"
#include "cpu/translation.hh"
#include "mem/packet.hh"
#include "sim/byteswap.hh"
+#include "sim/fault_fwd.hh"
#include "sim/system.hh"
#include "sim/tlb.hh"
* Defines a dynamic instruction context.
*/
-// Forward declaration.
-class StaticInstPtr;
-
template <class Impl>
-class BaseDynInst : public FastAlloc, public RefCounted
+class BaseDynInst : public RefCounted
{
public:
// Typedef for the CPU.
// The DynInstPtr type.
typedef typename Impl::DynInstPtr DynInstPtr;
+ typedef RefCountingPtr<BaseDynInst<Impl> > BaseDynInstPtr;
// The list of instructions iterator type.
typedef typename std::list<DynInstPtr>::iterator ListIt;
enum {
MaxInstSrcRegs = TheISA::MaxInstSrcRegs, /// Max source regs
- MaxInstDestRegs = TheISA::MaxInstDestRegs, /// Max dest regs
+ MaxInstDestRegs = TheISA::MaxInstDestRegs /// Max dest regs
};
- /** The StaticInst used by this BaseDynInst. */
- StaticInstPtr staticInst;
-
- ////////////////////////////////////////////
- //
- // INSTRUCTION EXECUTION
- //
- ////////////////////////////////////////////
- /** InstRecord that tracks this instructions. */
- Trace::InstRecord *traceData;
-
- void demapPage(Addr vaddr, uint64_t asn)
- {
- cpu->demapPage(vaddr, asn);
- }
- void demapInstPage(Addr vaddr, uint64_t asn)
- {
- cpu->demapPage(vaddr, asn);
- }
- void demapDataPage(Addr vaddr, uint64_t asn)
- {
- cpu->demapPage(vaddr, asn);
- }
-
- Fault readMem(Addr addr, uint8_t *data, unsigned size, unsigned flags);
-
- Fault writeMem(uint8_t *data, unsigned size,
- Addr addr, unsigned flags, uint64_t *res);
-
- /** Splits a request in two if it crosses a dcache block. */
- void splitRequest(RequestPtr req, RequestPtr &sreqLow,
- RequestPtr &sreqHigh);
-
- /** Initiate a DTB address translation. */
- void initiateTranslation(RequestPtr req, RequestPtr sreqLow,
- RequestPtr sreqHigh, uint64_t *res,
- BaseTLB::Mode mode);
-
- /** Finish a DTB address translation. */
- void finishTranslation(WholeTranslationState *state);
-
- /** True if the DTB address translation has started. */
- bool translationStarted;
-
- /** True if the DTB address translation has completed. */
- bool translationCompleted;
-
- /**
- * Returns true if the DTB address translation is being delayed due to a hw
- * page table walk.
- */
- bool isTranslationDelayed() const
- {
- return (translationStarted && !translationCompleted);
- }
-
- /**
- * Saved memory requests (needed when the DTB address translation is
- * delayed due to a hw page table walk).
- */
- RequestPtr savedReq;
- RequestPtr savedSreqLow;
- RequestPtr savedSreqHigh;
-
- /** @todo: Consider making this private. */
- public:
- /** The sequence number of the instruction. */
- InstSeqNum seqNum;
+ union Result {
+ uint64_t integer;
+ double dbl;
+ void set(uint64_t i) { integer = i; }
+ void set(double d) { dbl = d; }
+ void get(uint64_t& i) { i = integer; }
+ void get(double& d) { d = dbl; }
+ };
+ protected:
enum Status {
IqEntry, /// Instruction is in the IQ
RobEntry, /// Instruction is in the ROB
NumStatus
};
- /** The status of this BaseDynInst. Several bits can be set. */
- std::bitset<NumStatus> status;
-
- /** The thread this instruction is from. */
- ThreadID threadNumber;
+ enum Flags {
+ TranslationStarted,
+ TranslationCompleted,
+ PossibleLoadViolation,
+ HitExternalSnoop,
+ EffAddrValid,
+ RecordResult,
+ Predicate,
+ PredTaken,
+ /** Whether or not the effective address calculation is completed.
+ * @todo: Consider if this is necessary or not.
+ */
+ EACalcDone,
+ IsUncacheable,
+ ReqMade,
+ MemOpDone,
+ MaxFlags
+ };
- /** data address space ID, for loads & stores. */
- short asid;
+ public:
+ /** The sequence number of the instruction. */
+ InstSeqNum seqNum;
- /** How many source registers are ready. */
- unsigned readyRegs;
+ /** The StaticInst used by this BaseDynInst. */
+ StaticInstPtr staticInst;
/** Pointer to the Impl's CPU object. */
ImplCPU *cpu;
/** The kind of fault this instruction has generated. */
Fault fault;
- /** Pointer to the data for the memory access. */
- uint8_t *memData;
+ /** InstRecord that tracks this instructions. */
+ Trace::InstRecord *traceData;
- /** The effective virtual address (lds & stores only). */
- Addr effAddr;
+ protected:
+ /** The result of the instruction; assumes an instruction can have many
+ * destination registers.
+ */
+ std::queue<Result> instResult;
- /** The size of the request */
- Addr effSize;
+ /** PC state for this instruction. */
+ TheISA::PCState pc;
- /** Is the effective virtual address valid. */
- bool effAddrValid;
+ /* An amalgamation of a lot of boolean values into one */
+ std::bitset<MaxFlags> instFlags;
+
+ /** The status of this BaseDynInst. Several bits can be set. */
+ std::bitset<NumStatus> status;
+
+ /** Whether or not the source register is ready.
+ * @todo: Not sure this should be here vs the derived class.
+ */
+ std::bitset<MaxInstSrcRegs> _readySrcRegIdx;
+
+ public:
+ /** The thread this instruction is from. */
+ ThreadID threadNumber;
+
+ /** Iterator pointing to this BaseDynInst in the list of all insts. */
+ ListIt instListIt;
+
+ ////////////////////// Branch Data ///////////////
+ /** Predicted PC state after this instruction. */
+ TheISA::PCState predPC;
+
+ /** The Macroop if one exists */
+ StaticInstPtr macroop;
+
+ /** How many source registers are ready. */
+ uint8_t readyRegs;
+
+ public:
+ /////////////////////// Load Store Data //////////////////////
+ /** The effective virtual address (lds & stores only). */
+ Addr effAddr;
/** The effective physical address. */
Addr physEffAddr;
/** The memory request flags (from translation). */
unsigned memReqFlags;
- union Result {
- uint64_t integer;
-// float fp;
- double dbl;
- };
-
- /** The result of the instruction; assumes for now that there's only one
- * destination register.
- */
- Result instResult;
+ /** data address space ID, for loads & stores. */
+ short asid;
- /** Records changes to result? */
- bool recordResult;
+ /** The size of the request */
+ uint8_t effSize;
- /** Did this instruction execute, or is it predicated false */
- bool predicate;
+ /** Pointer to the data for the memory access. */
+ uint8_t *memData;
- protected:
- /** PC state for this instruction. */
- TheISA::PCState pc;
+ /** Load queue index. */
+ int16_t lqIdx;
- /** Predicted PC state after this instruction. */
- TheISA::PCState predPC;
+ /** Store queue index. */
+ int16_t sqIdx;
- /** If this is a branch that was predicted taken */
- bool predTaken;
- public:
+ /////////////////////// TLB Miss //////////////////////
+ /**
+ * Saved memory requests (needed when the DTB address translation is
+ * delayed due to a hw page table walk).
+ */
+ RequestPtr savedReq;
+ RequestPtr savedSreqLow;
+ RequestPtr savedSreqHigh;
-#ifdef DEBUG
- void dumpSNList();
-#endif
+ /////////////////////// Checker //////////////////////
+ // Need a copy of main request pointer to verify on writes.
+ RequestPtr reqToVerify;
- /** Whether or not the source register is ready.
- * @todo: Not sure this should be here vs the derived class.
+ private:
+ /** Instruction effective address.
+ * @todo: Consider if this is necessary or not.
*/
- bool _readySrcRegIdx[MaxInstSrcRegs];
+ Addr instEffAddr;
protected:
/** Flattened register index of the destination registers of this
*/
TheISA::RegIndex _flatDestRegIdx[TheISA::MaxInstDestRegs];
- /** Flattened register index of the source registers of this
- * instruction.
- */
- TheISA::RegIndex _flatSrcRegIdx[TheISA::MaxInstSrcRegs];
-
/** Physical register index of the destination registers of this
* instruction.
*/
*/
PhysRegIndex _prevDestRegIdx[TheISA::MaxInstDestRegs];
+
public:
+ /** Records changes to result? */
+ void recordResult(bool f) { instFlags[RecordResult] = f; }
+
+ /** Is the effective virtual address valid. */
+ bool effAddrValid() const { return instFlags[EffAddrValid]; }
+
+ /** Whether or not the memory operation is done. */
+ bool memOpDone() const { return instFlags[MemOpDone]; }
+ void memOpDone(bool f) { instFlags[MemOpDone] = f; }
+
+
+ ////////////////////////////////////////////
+ //
+ // INSTRUCTION EXECUTION
+ //
+ ////////////////////////////////////////////
+
+ void demapPage(Addr vaddr, uint64_t asn)
+ {
+ cpu->demapPage(vaddr, asn);
+ }
+ void demapInstPage(Addr vaddr, uint64_t asn)
+ {
+ cpu->demapPage(vaddr, asn);
+ }
+ void demapDataPage(Addr vaddr, uint64_t asn)
+ {
+ cpu->demapPage(vaddr, asn);
+ }
+
+ Fault readMem(Addr addr, uint8_t *data, unsigned size, unsigned flags);
+
+ Fault writeMem(uint8_t *data, unsigned size,
+ Addr addr, unsigned flags, uint64_t *res);
+
+ /** Splits a request in two if it crosses a dcache block. */
+ void splitRequest(RequestPtr req, RequestPtr &sreqLow,
+ RequestPtr &sreqHigh);
+
+ /** Initiate a DTB address translation. */
+ void initiateTranslation(RequestPtr req, RequestPtr sreqLow,
+ RequestPtr sreqHigh, uint64_t *res,
+ BaseTLB::Mode mode);
+
+ /** Finish a DTB address translation. */
+ void finishTranslation(WholeTranslationState *state);
+
+ /** True if the DTB address translation has started. */
+ bool translationStarted() const { return instFlags[TranslationStarted]; }
+ void translationStarted(bool f) { instFlags[TranslationStarted] = f; }
+
+ /** True if the DTB address translation has completed. */
+ bool translationCompleted() const { return instFlags[TranslationCompleted]; }
+ void translationCompleted(bool f) { instFlags[TranslationCompleted] = f; }
+
+ /** True if this address was found to match a previous load and they issued
+ * out of order. If that happend, then it's only a problem if an incoming
+ * snoop invalidate modifies the line, in which case we need to squash.
+ * If nothing modified the line the order doesn't matter.
+ */
+ bool possibleLoadViolation() const { return instFlags[PossibleLoadViolation]; }
+ void possibleLoadViolation(bool f) { instFlags[PossibleLoadViolation] = f; }
+
+ /** True if the address hit a external snoop while sitting in the LSQ.
+ * If this is true and a older instruction sees it, this instruction must
+ * reexecute
+ */
+ bool hitExternalSnoop() const { return instFlags[HitExternalSnoop]; }
+ void hitExternalSnoop(bool f) { instFlags[HitExternalSnoop] = f; }
+
+ /**
+ * Returns true if the DTB address translation is being delayed due to a hw
+ * page table walk.
+ */
+ bool isTranslationDelayed() const
+ {
+ return (translationStarted() && !translationCompleted());
+ }
+
+ public:
+#ifdef DEBUG
+ void dumpSNList();
+#endif
/** Returns the physical register index of the i'th destination
* register.
/** Returns the physical register index of the i'th source register. */
PhysRegIndex renamedSrcRegIdx(int idx) const
{
+ assert(TheISA::MaxInstSrcRegs > idx);
return _srcRegIdx[idx];
}
return _flatDestRegIdx[idx];
}
- /** Returns the flattened register index of the i'th source register */
- TheISA::RegIndex flattenedSrcRegIdx(int idx) const
- {
- return _flatSrcRegIdx[idx];
- }
-
/** Returns the physical register index of the previous physical register
* that remapped to the same logical register index.
*/
_srcRegIdx[idx] = renamed_src;
}
- /** Flattens a source architectural register index into a logical index.
- */
- void flattenSrcReg(int idx, TheISA::RegIndex flattened_src)
- {
- _flatSrcRegIdx[idx] = flattened_src;
- }
-
/** Flattens a destination architectural register index into a logical
* index.
*/
* @param seq_num The sequence number of the instruction.
* @param cpu Pointer to the instruction's CPU.
*/
- BaseDynInst(StaticInstPtr staticInst, TheISA::PCState pc,
- TheISA::PCState predPC, InstSeqNum seq_num, ImplCPU *cpu);
+ BaseDynInst(StaticInstPtr staticInst, StaticInstPtr macroop,
+ TheISA::PCState pc, TheISA::PCState predPC,
+ InstSeqNum seq_num, ImplCPU *cpu);
/** BaseDynInst constructor given a StaticInst pointer.
* @param _staticInst The StaticInst for this BaseDynInst.
*/
- BaseDynInst(StaticInstPtr &_staticInst);
+ BaseDynInst(StaticInstPtr staticInst, StaticInstPtr macroop);
/** BaseDynInst destructor. */
~BaseDynInst();
/** Read this CPU's ID. */
int cpuId() { return cpu->cpuId(); }
+ /** Read this CPU's data requestor ID */
+ MasterID masterId() { return cpu->dataMasterId(); }
+
/** Read this context's system-wide ID **/
int contextId() { return thread->contextId(); }
/** Returns whether the instruction was predicted taken or not. */
bool readPredTaken()
{
- return predTaken;
+ return instFlags[PredTaken];
}
void setPredTaken(bool predicted_taken)
{
- predTaken = predicted_taken;
+ instFlags[PredTaken] = predicted_taken;
}
/** Returns whether the instruction mispredicted. */
/** Returns the logical register index of the i'th source register. */
RegIndex srcRegIdx(int i) const { return staticInst->srcRegIdx(i); }
- /** Returns the result of an integer instruction. */
- uint64_t readIntResult() { return instResult.integer; }
+ /** Pops a result off the instResult queue */
+ template <class T>
+ void popResult(T& t)
+ {
+ if (!instResult.empty()) {
+ instResult.front().get(t);
+ instResult.pop();
+ }
+ }
- /** Returns the result of a floating point instruction. */
- float readFloatResult() { return (float)instResult.dbl; }
+ /** Read the most recent result stored by this instruction */
+ template <class T>
+ void readResult(T& t)
+ {
+ instResult.back().get(t);
+ }
- /** Returns the result of a floating point (double) instruction. */
- double readDoubleResult() { return instResult.dbl; }
+ /** Pushes a result onto the instResult queue */
+ template <class T>
+ void setResult(T t)
+ {
+ if (instFlags[RecordResult]) {
+ Result instRes;
+ instRes.set(t);
+ instResult.push(instRes);
+ }
+ }
/** Records an integer register being set to a value. */
void setIntRegOperand(const StaticInst *si, int idx, uint64_t val)
{
- if (recordResult)
- instResult.integer = val;
+ setResult<uint64_t>(val);
}
/** Records an fp register being set to a value. */
void setFloatRegOperand(const StaticInst *si, int idx, FloatReg val,
int width)
{
- if (recordResult) {
- if (width == 32)
- instResult.dbl = (double)val;
- else if (width == 64)
- instResult.dbl = val;
- else
- panic("Unsupported width!");
+ if (width == 32 || width == 64) {
+ setResult<double>(val);
+ } else {
+ panic("Unsupported width!");
}
}
/** Records an fp register being set to a value. */
void setFloatRegOperand(const StaticInst *si, int idx, FloatReg val)
{
- if (recordResult)
- instResult.dbl = (double)val;
+ setResult<double>(val);
}
/** Records an fp register being set to an integer value. */
void setFloatRegOperandBits(const StaticInst *si, int idx, uint64_t val,
int width)
{
- if (recordResult)
- instResult.integer = val;
+ setResult<uint64_t>(val);
}
/** Records an fp register being set to an integer value. */
void setFloatRegOperandBits(const StaticInst *si, int idx, uint64_t val)
{
- if (recordResult)
- instResult.integer = val;
+ setResult<uint64_t>(val);
}
/** Records that one of the source registers is ready. */
bool readPredicate()
{
- return predicate;
+ return instFlags[Predicate];
}
void setPredicate(bool val)
{
- predicate = val;
+ instFlags[Predicate] = val;
if (traceData) {
traceData->setPredicate(val);
/** Returns the thread context. */
ThreadContext *tcBase() { return thread->getTC(); }
- private:
- /** Instruction effective address.
- * @todo: Consider if this is necessary or not.
- */
- Addr instEffAddr;
-
- /** Whether or not the effective address calculation is completed.
- * @todo: Consider if this is necessary or not.
- */
- bool eaCalcDone;
-
- /** Is this instruction's memory access uncacheable. */
- bool isUncacheable;
-
- /** Has this instruction generated a memory request. */
- bool reqMade;
-
public:
/** Sets the effective address. */
- void setEA(Addr &ea) { instEffAddr = ea; eaCalcDone = true; }
+ void setEA(Addr &ea) { instEffAddr = ea; instFlags[EACalcDone] = true; }
/** Returns the effective address. */
const Addr &getEA() const { return instEffAddr; }
/** Returns whether or not the eff. addr. calculation has been completed. */
- bool doneEACalc() { return eaCalcDone; }
+ bool doneEACalc() { return instFlags[EACalcDone]; }
/** Returns whether or not the eff. addr. source registers are ready. */
bool eaSrcsReady();
- /** Whether or not the memory operation is done. */
- bool memOpDone;
-
/** Is this instruction's memory access uncacheable. */
- bool uncacheable() { return isUncacheable; }
+ bool uncacheable() { return instFlags[IsUncacheable]; }
/** Has this instruction generated a memory request. */
- bool hasRequest() { return reqMade; }
-
- public:
- /** Load queue index. */
- int16_t lqIdx;
-
- /** Store queue index. */
- int16_t sqIdx;
-
- /** Iterator pointing to this BaseDynInst in the list of all insts. */
- ListIt instListIt;
+ bool hasRequest() { return instFlags[ReqMade]; }
/** Returns iterator to this instruction in the list of all insts. */
ListIt &getInstListIt() { return instListIt; }
BaseDynInst<Impl>::readMem(Addr addr, uint8_t *data,
unsigned size, unsigned flags)
{
- reqMade = true;
+ instFlags[ReqMade] = true;
Request *req = NULL;
Request *sreqLow = NULL;
Request *sreqHigh = NULL;
- if (reqMade && translationStarted) {
+ if (instFlags[ReqMade] && translationStarted()) {
req = savedReq;
sreqLow = savedSreqLow;
sreqHigh = savedSreqHigh;
} else {
- req = new Request(asid, addr, size, flags, this->pc.instAddr(),
+ req = new Request(asid, addr, size, flags, masterId(), this->pc.instAddr(),
thread->contextId(), threadNumber);
// Only split the request if the ISA supports unaligned accesses.
initiateTranslation(req, sreqLow, sreqHigh, NULL, BaseTLB::Read);
}
- if (translationCompleted) {
+ if (translationCompleted()) {
if (fault == NoFault) {
effAddr = req->getVaddr();
effSize = size;
- effAddrValid = true;
+ instFlags[EffAddrValid] = true;
+
+ if (cpu->checker) {
+ if (reqToVerify != NULL) {
+ delete reqToVerify;
+ }
+ reqToVerify = new Request(*req);
+ }
fault = cpu->read(req, sreqLow, sreqHigh, data, lqIdx);
} else {
// Commit will have to clean up whatever happened. Set this
traceData->setAddr(addr);
}
- reqMade = true;
+ instFlags[ReqMade] = true;
Request *req = NULL;
Request *sreqLow = NULL;
Request *sreqHigh = NULL;
- if (reqMade && translationStarted) {
+ if (instFlags[ReqMade] && translationStarted()) {
req = savedReq;
sreqLow = savedSreqLow;
sreqHigh = savedSreqHigh;
} else {
- req = new Request(asid, addr, size, flags, this->pc.instAddr(),
+ req = new Request(asid, addr, size, flags, masterId(), this->pc.instAddr(),
thread->contextId(), threadNumber);
// Only split the request if the ISA supports unaligned accesses.
initiateTranslation(req, sreqLow, sreqHigh, res, BaseTLB::Write);
}
- if (fault == NoFault && translationCompleted) {
+ if (fault == NoFault && translationCompleted()) {
effAddr = req->getVaddr();
effSize = size;
- effAddrValid = true;
+ instFlags[EffAddrValid] = true;
+
+ if (cpu->checker) {
+ if (reqToVerify != NULL) {
+ delete reqToVerify;
+ }
+ reqToVerify = new Request(*req);
+ }
fault = cpu->write(req, sreqLow, sreqHigh, data, sqIdx);
}
RequestPtr &sreqHigh)
{
// Check to see if the request crosses the next level block boundary.
- unsigned block_size = cpu->getDcachePort()->peerBlockSize();
+ unsigned block_size = cpu->getDataPort().peerBlockSize();
Addr addr = req->getVaddr();
Addr split_addr = roundDown(addr + req->getSize() - 1, block_size);
assert(split_addr <= addr || split_addr - addr < block_size);
RequestPtr sreqHigh, uint64_t *res,
BaseTLB::Mode mode)
{
- translationStarted = true;
+ translationStarted(true);
if (!TheISA::HasUnalignedMemAcc || sreqLow == NULL) {
WholeTranslationState *state =
new WholeTranslationState(req, NULL, res, mode);
// One translation if the request isn't split.
- DataTranslation<BaseDynInst<Impl> > *trans =
- new DataTranslation<BaseDynInst<Impl> >(this, state);
+ DataTranslation<BaseDynInstPtr> *trans =
+ new DataTranslation<BaseDynInstPtr>(this, state);
cpu->dtb->translateTiming(req, thread->getTC(), trans, mode);
- if (!translationCompleted) {
+ if (!translationCompleted()) {
// Save memory requests.
savedReq = state->mainReq;
savedSreqLow = state->sreqLow;
new WholeTranslationState(req, sreqLow, sreqHigh, NULL, res, mode);
// Two translations when the request is split.
- DataTranslation<BaseDynInst<Impl> > *stransLow =
- new DataTranslation<BaseDynInst<Impl> >(this, state, 0);
- DataTranslation<BaseDynInst<Impl> > *stransHigh =
- new DataTranslation<BaseDynInst<Impl> >(this, state, 1);
+ DataTranslation<BaseDynInstPtr> *stransLow =
+ new DataTranslation<BaseDynInstPtr>(this, state, 0);
+ DataTranslation<BaseDynInstPtr> *stransHigh =
+ new DataTranslation<BaseDynInstPtr>(this, state, 1);
cpu->dtb->translateTiming(sreqLow, thread->getTC(), stransLow, mode);
cpu->dtb->translateTiming(sreqHigh, thread->getTC(), stransHigh, mode);
- if (!translationCompleted) {
+ if (!translationCompleted()) {
// Save memory requests.
savedReq = state->mainReq;
savedSreqLow = state->sreqLow;
{
fault = state->getFault();
- if (state->isUncacheable())
- isUncacheable = true;
+ instFlags[IsUncacheable] = state->isUncacheable();
if (fault == NoFault) {
physEffAddr = state->getPaddr();
}
delete state;
- translationCompleted = true;
+ translationCompleted(true);
}
#endif // __CPU_BASE_DYN_INST_HH__
projects / gem5.git / blobdiff